5 * "A fair jaw-cracker dwarf-language must be." --Samwise Gamgee
8 /* This file contains functions for compiling a regular expression. See
9 * also regexec.c which funnily enough, contains functions for executing
10 * a regular expression.
12 * This file is also copied at build time to ext/re/re_comp.c, where
13 * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT.
14 * This causes the main functions to be compiled under new names and with
15 * debugging support added, which makes "use re 'debug'" work.
18 /* NOTE: this is derived from Henry Spencer's regexp code, and should not
19 * confused with the original package (see point 3 below). Thanks, Henry!
22 /* Additional note: this code is very heavily munged from Henry's version
23 * in places. In some spots I've traded clarity for efficiency, so don't
24 * blame Henry for some of the lack of readability.
27 /* The names of the functions have been changed from regcomp and
28 * regexec to pregcomp and pregexec in order to avoid conflicts
29 * with the POSIX routines of the same names.
32 #ifdef PERL_EXT_RE_BUILD
37 * pregcomp and pregexec -- regsub and regerror are not used in perl
39 * Copyright (c) 1986 by University of Toronto.
40 * Written by Henry Spencer. Not derived from licensed software.
42 * Permission is granted to anyone to use this software for any
43 * purpose on any computer system, and to redistribute it freely,
44 * subject to the following restrictions:
46 * 1. The author is not responsible for the consequences of use of
47 * this software, no matter how awful, even if they arise
50 * 2. The origin of this software must not be misrepresented, either
51 * by explicit claim or by omission.
53 * 3. Altered versions must be plainly marked as such, and must not
54 * be misrepresented as being the original software.
57 **** Alterations to Henry's code are...
59 **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
60 **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007 by Larry Wall and others
62 **** You may distribute under the terms of either the GNU General Public
63 **** License or the Artistic License, as specified in the README file.
66 * Beware that some of this code is subtly aware of the way operator
67 * precedence is structured in regular expressions. Serious changes in
68 * regular-expression syntax might require a total rethink.
71 #define PERL_IN_REGCOMP_C
74 #ifndef PERL_IN_XSUB_RE
79 #ifdef PERL_IN_XSUB_RE
90 # if defined(BUGGY_MSC6)
91 /* MSC 6.00A breaks on op/regexp.t test 85 unless we turn this off */
92 # pragma optimize("a",off)
93 /* But MSC 6.00A is happy with 'w', for aliases only across function calls*/
94 # pragma optimize("w",on )
95 # endif /* BUGGY_MSC6 */
102 typedef struct RExC_state_t {
103 U32 flags; /* are we folding, multilining? */
104 char *precomp; /* uncompiled string. */
105 regexp *rx; /* perl core regexp structure */
106 regexp_internal *rxi; /* internal data for regexp object pprivate field */
107 char *start; /* Start of input for compile */
108 char *end; /* End of input for compile */
109 char *parse; /* Input-scan pointer. */
110 I32 whilem_seen; /* number of WHILEM in this expr */
111 regnode *emit_start; /* Start of emitted-code area */
112 regnode *emit; /* Code-emit pointer; ®dummy = don't = compiling */
113 I32 naughty; /* How bad is this pattern? */
114 I32 sawback; /* Did we see \1, ...? */
116 I32 size; /* Code size. */
117 I32 npar; /* Capture buffer count, (OPEN). */
118 I32 cpar; /* Capture buffer count, (CLOSE). */
119 I32 nestroot; /* root parens we are in - used by accept */
123 regnode **open_parens; /* pointers to open parens */
124 regnode **close_parens; /* pointers to close parens */
125 regnode *opend; /* END node in program */
127 HV *charnames; /* cache of named sequences */
128 HV *paren_names; /* Paren names */
130 regnode **recurse; /* Recurse regops */
131 I32 recurse_count; /* Number of recurse regops */
133 char *starttry; /* -Dr: where regtry was called. */
134 #define RExC_starttry (pRExC_state->starttry)
137 const char *lastparse;
139 AV *paren_name_list; /* idx -> name */
140 #define RExC_lastparse (pRExC_state->lastparse)
141 #define RExC_lastnum (pRExC_state->lastnum)
142 #define RExC_paren_name_list (pRExC_state->paren_name_list)
146 #define RExC_flags (pRExC_state->flags)
147 #define RExC_precomp (pRExC_state->precomp)
148 #define RExC_rx (pRExC_state->rx)
149 #define RExC_rxi (pRExC_state->rxi)
150 #define RExC_start (pRExC_state->start)
151 #define RExC_end (pRExC_state->end)
152 #define RExC_parse (pRExC_state->parse)
153 #define RExC_whilem_seen (pRExC_state->whilem_seen)
154 #ifdef RE_TRACK_PATTERN_OFFSETS
155 #define RExC_offsets (pRExC_state->rxi->u.offsets) /* I am not like the others */
157 #define RExC_emit (pRExC_state->emit)
158 #define RExC_emit_start (pRExC_state->emit_start)
159 #define RExC_naughty (pRExC_state->naughty)
160 #define RExC_sawback (pRExC_state->sawback)
161 #define RExC_seen (pRExC_state->seen)
162 #define RExC_size (pRExC_state->size)
163 #define RExC_npar (pRExC_state->npar)
164 #define RExC_nestroot (pRExC_state->nestroot)
165 #define RExC_extralen (pRExC_state->extralen)
166 #define RExC_seen_zerolen (pRExC_state->seen_zerolen)
167 #define RExC_seen_evals (pRExC_state->seen_evals)
168 #define RExC_utf8 (pRExC_state->utf8)
169 #define RExC_charnames (pRExC_state->charnames)
170 #define RExC_open_parens (pRExC_state->open_parens)
171 #define RExC_close_parens (pRExC_state->close_parens)
172 #define RExC_opend (pRExC_state->opend)
173 #define RExC_paren_names (pRExC_state->paren_names)
174 #define RExC_recurse (pRExC_state->recurse)
175 #define RExC_recurse_count (pRExC_state->recurse_count)
178 #define ISMULT1(c) ((c) == '*' || (c) == '+' || (c) == '?')
179 #define ISMULT2(s) ((*s) == '*' || (*s) == '+' || (*s) == '?' || \
180 ((*s) == '{' && regcurly(s)))
183 #undef SPSTART /* dratted cpp namespace... */
186 * Flags to be passed up and down.
188 #define WORST 0 /* Worst case. */
189 #define HASWIDTH 0x01 /* Known to match non-null strings. */
190 #define SIMPLE 0x02 /* Simple enough to be STAR/PLUS operand. */
191 #define SPSTART 0x04 /* Starts with * or +. */
192 #define TRYAGAIN 0x08 /* Weeded out a declaration. */
193 #define POSTPONED 0x10 /* (?1),(?&name), (??{...}) or similar */
195 #define REG_NODE_NUM(x) ((x) ? (int)((x)-RExC_emit_start) : -1)
197 /* whether trie related optimizations are enabled */
198 #if PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION
199 #define TRIE_STUDY_OPT
200 #define FULL_TRIE_STUDY
206 #define PBYTE(u8str,paren) ((U8*)(u8str))[(paren) >> 3]
207 #define PBITVAL(paren) (1 << ((paren) & 7))
208 #define PAREN_TEST(u8str,paren) ( PBYTE(u8str,paren) & PBITVAL(paren))
209 #define PAREN_SET(u8str,paren) PBYTE(u8str,paren) |= PBITVAL(paren)
210 #define PAREN_UNSET(u8str,paren) PBYTE(u8str,paren) &= (~PBITVAL(paren))
213 /* About scan_data_t.
215 During optimisation we recurse through the regexp program performing
216 various inplace (keyhole style) optimisations. In addition study_chunk
217 and scan_commit populate this data structure with information about
218 what strings MUST appear in the pattern. We look for the longest
219 string that must appear for at a fixed location, and we look for the
220 longest string that may appear at a floating location. So for instance
225 Both 'FOO' and 'A' are fixed strings. Both 'B' and 'BAR' are floating
226 strings (because they follow a .* construct). study_chunk will identify
227 both FOO and BAR as being the longest fixed and floating strings respectively.
229 The strings can be composites, for instance
233 will result in a composite fixed substring 'foo'.
235 For each string some basic information is maintained:
237 - offset or min_offset
238 This is the position the string must appear at, or not before.
239 It also implicitly (when combined with minlenp) tells us how many
240 character must match before the string we are searching.
241 Likewise when combined with minlenp and the length of the string
242 tells us how many characters must appear after the string we have
246 Only used for floating strings. This is the rightmost point that
247 the string can appear at. Ifset to I32 max it indicates that the
248 string can occur infinitely far to the right.
251 A pointer to the minimum length of the pattern that the string
252 was found inside. This is important as in the case of positive
253 lookahead or positive lookbehind we can have multiple patterns
258 The minimum length of the pattern overall is 3, the minimum length
259 of the lookahead part is 3, but the minimum length of the part that
260 will actually match is 1. So 'FOO's minimum length is 3, but the
261 minimum length for the F is 1. This is important as the minimum length
262 is used to determine offsets in front of and behind the string being
263 looked for. Since strings can be composites this is the length of the
264 pattern at the time it was commited with a scan_commit. Note that
265 the length is calculated by study_chunk, so that the minimum lengths
266 are not known until the full pattern has been compiled, thus the
267 pointer to the value.
271 In the case of lookbehind the string being searched for can be
272 offset past the start point of the final matching string.
273 If this value was just blithely removed from the min_offset it would
274 invalidate some of the calculations for how many chars must match
275 before or after (as they are derived from min_offset and minlen and
276 the length of the string being searched for).
277 When the final pattern is compiled and the data is moved from the
278 scan_data_t structure into the regexp structure the information
279 about lookbehind is factored in, with the information that would
280 have been lost precalculated in the end_shift field for the
283 The fields pos_min and pos_delta are used to store the minimum offset
284 and the delta to the maximum offset at the current point in the pattern.
288 typedef struct scan_data_t {
289 /*I32 len_min; unused */
290 /*I32 len_delta; unused */
294 I32 last_end; /* min value, <0 unless valid. */
297 SV **longest; /* Either &l_fixed, or &l_float. */
298 SV *longest_fixed; /* longest fixed string found in pattern */
299 I32 offset_fixed; /* offset where it starts */
300 I32 *minlen_fixed; /* pointer to the minlen relevent to the string */
301 I32 lookbehind_fixed; /* is the position of the string modfied by LB */
302 SV *longest_float; /* longest floating string found in pattern */
303 I32 offset_float_min; /* earliest point in string it can appear */
304 I32 offset_float_max; /* latest point in string it can appear */
305 I32 *minlen_float; /* pointer to the minlen relevent to the string */
306 I32 lookbehind_float; /* is the position of the string modified by LB */
310 struct regnode_charclass_class *start_class;
314 * Forward declarations for pregcomp()'s friends.
317 static const scan_data_t zero_scan_data =
318 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ,0};
320 #define SF_BEFORE_EOL (SF_BEFORE_SEOL|SF_BEFORE_MEOL)
321 #define SF_BEFORE_SEOL 0x0001
322 #define SF_BEFORE_MEOL 0x0002
323 #define SF_FIX_BEFORE_EOL (SF_FIX_BEFORE_SEOL|SF_FIX_BEFORE_MEOL)
324 #define SF_FL_BEFORE_EOL (SF_FL_BEFORE_SEOL|SF_FL_BEFORE_MEOL)
327 # define SF_FIX_SHIFT_EOL (0+2)
328 # define SF_FL_SHIFT_EOL (0+4)
330 # define SF_FIX_SHIFT_EOL (+2)
331 # define SF_FL_SHIFT_EOL (+4)
334 #define SF_FIX_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FIX_SHIFT_EOL)
335 #define SF_FIX_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FIX_SHIFT_EOL)
337 #define SF_FL_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FL_SHIFT_EOL)
338 #define SF_FL_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FL_SHIFT_EOL) /* 0x20 */
339 #define SF_IS_INF 0x0040
340 #define SF_HAS_PAR 0x0080
341 #define SF_IN_PAR 0x0100
342 #define SF_HAS_EVAL 0x0200
343 #define SCF_DO_SUBSTR 0x0400
344 #define SCF_DO_STCLASS_AND 0x0800
345 #define SCF_DO_STCLASS_OR 0x1000
346 #define SCF_DO_STCLASS (SCF_DO_STCLASS_AND|SCF_DO_STCLASS_OR)
347 #define SCF_WHILEM_VISITED_POS 0x2000
349 #define SCF_TRIE_RESTUDY 0x4000 /* Do restudy? */
350 #define SCF_SEEN_ACCEPT 0x8000
352 #define UTF (RExC_utf8 != 0)
353 #define LOC ((RExC_flags & RXf_PMf_LOCALE) != 0)
354 #define FOLD ((RExC_flags & RXf_PMf_FOLD) != 0)
356 #define OOB_UNICODE 12345678
357 #define OOB_NAMEDCLASS -1
359 #define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv))
360 #define CHR_DIST(a,b) (UTF ? utf8_distance(a,b) : a - b)
363 /* length of regex to show in messages that don't mark a position within */
364 #define RegexLengthToShowInErrorMessages 127
367 * If MARKER[12] are adjusted, be sure to adjust the constants at the top
368 * of t/op/regmesg.t, the tests in t/op/re_tests, and those in
369 * op/pragma/warn/regcomp.
371 #define MARKER1 "<-- HERE" /* marker as it appears in the description */
372 #define MARKER2 " <-- HERE " /* marker as it appears within the regex */
374 #define REPORT_LOCATION " in regex; marked by " MARKER1 " in m/%.*s" MARKER2 "%s/"
377 * Calls SAVEDESTRUCTOR_X if needed, then calls Perl_croak with the given
378 * arg. Show regex, up to a maximum length. If it's too long, chop and add
381 #define _FAIL(code) STMT_START { \
382 const char *ellipses = ""; \
383 IV len = RExC_end - RExC_precomp; \
386 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \
387 if (len > RegexLengthToShowInErrorMessages) { \
388 /* chop 10 shorter than the max, to ensure meaning of "..." */ \
389 len = RegexLengthToShowInErrorMessages - 10; \
395 #define FAIL(msg) _FAIL( \
396 Perl_croak(aTHX_ "%s in regex m/%.*s%s/", \
397 msg, (int)len, RExC_precomp, ellipses))
399 #define FAIL2(msg,arg) _FAIL( \
400 Perl_croak(aTHX_ msg " in regex m/%.*s%s/", \
401 arg, (int)len, RExC_precomp, ellipses))
404 * Simple_vFAIL -- like FAIL, but marks the current location in the scan
406 #define Simple_vFAIL(m) STMT_START { \
407 const IV offset = RExC_parse - RExC_precomp; \
408 Perl_croak(aTHX_ "%s" REPORT_LOCATION, \
409 m, (int)offset, RExC_precomp, RExC_precomp + offset); \
413 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL()
415 #define vFAIL(m) STMT_START { \
417 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \
422 * Like Simple_vFAIL(), but accepts two arguments.
424 #define Simple_vFAIL2(m,a1) STMT_START { \
425 const IV offset = RExC_parse - RExC_precomp; \
426 S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, \
427 (int)offset, RExC_precomp, RExC_precomp + offset); \
431 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL2().
433 #define vFAIL2(m,a1) STMT_START { \
435 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \
436 Simple_vFAIL2(m, a1); \
441 * Like Simple_vFAIL(), but accepts three arguments.
443 #define Simple_vFAIL3(m, a1, a2) STMT_START { \
444 const IV offset = RExC_parse - RExC_precomp; \
445 S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, a2, \
446 (int)offset, RExC_precomp, RExC_precomp + offset); \
450 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL3().
452 #define vFAIL3(m,a1,a2) STMT_START { \
454 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \
455 Simple_vFAIL3(m, a1, a2); \
459 * Like Simple_vFAIL(), but accepts four arguments.
461 #define Simple_vFAIL4(m, a1, a2, a3) STMT_START { \
462 const IV offset = RExC_parse - RExC_precomp; \
463 S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, a2, a3, \
464 (int)offset, RExC_precomp, RExC_precomp + offset); \
467 #define vWARN(loc,m) STMT_START { \
468 const IV offset = loc - RExC_precomp; \
469 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s" REPORT_LOCATION, \
470 m, (int)offset, RExC_precomp, RExC_precomp + offset); \
473 #define vWARNdep(loc,m) STMT_START { \
474 const IV offset = loc - RExC_precomp; \
475 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_REGEXP), \
476 "%s" REPORT_LOCATION, \
477 m, (int)offset, RExC_precomp, RExC_precomp + offset); \
481 #define vWARN2(loc, m, a1) STMT_START { \
482 const IV offset = loc - RExC_precomp; \
483 Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
484 a1, (int)offset, RExC_precomp, RExC_precomp + offset); \
487 #define vWARN3(loc, m, a1, a2) STMT_START { \
488 const IV offset = loc - RExC_precomp; \
489 Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
490 a1, a2, (int)offset, RExC_precomp, RExC_precomp + offset); \
493 #define vWARN4(loc, m, a1, a2, a3) STMT_START { \
494 const IV offset = loc - RExC_precomp; \
495 Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
496 a1, a2, a3, (int)offset, RExC_precomp, RExC_precomp + offset); \
499 #define vWARN5(loc, m, a1, a2, a3, a4) STMT_START { \
500 const IV offset = loc - RExC_precomp; \
501 Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
502 a1, a2, a3, a4, (int)offset, RExC_precomp, RExC_precomp + offset); \
506 /* Allow for side effects in s */
507 #define REGC(c,s) STMT_START { \
508 if (!SIZE_ONLY) *(s) = (c); else (void)(s); \
511 /* Macros for recording node offsets. 20001227 mjd@plover.com
512 * Nodes are numbered 1, 2, 3, 4. Node #n's position is recorded in
513 * element 2*n-1 of the array. Element #2n holds the byte length node #n.
514 * Element 0 holds the number n.
515 * Position is 1 indexed.
517 #ifndef RE_TRACK_PATTERN_OFFSETS
518 #define Set_Node_Offset_To_R(node,byte)
519 #define Set_Node_Offset(node,byte)
520 #define Set_Cur_Node_Offset
521 #define Set_Node_Length_To_R(node,len)
522 #define Set_Node_Length(node,len)
523 #define Set_Node_Cur_Length(node)
524 #define Node_Offset(n)
525 #define Node_Length(n)
526 #define Set_Node_Offset_Length(node,offset,len)
527 #define ProgLen(ri) ri->u.proglen
528 #define SetProgLen(ri,x) ri->u.proglen = x
530 #define ProgLen(ri) ri->u.offsets[0]
531 #define SetProgLen(ri,x) ri->u.offsets[0] = x
532 #define Set_Node_Offset_To_R(node,byte) STMT_START { \
534 MJD_OFFSET_DEBUG(("** (%d) offset of node %d is %d.\n", \
535 __LINE__, (int)(node), (int)(byte))); \
537 Perl_croak(aTHX_ "value of node is %d in Offset macro", (int)(node)); \
539 RExC_offsets[2*(node)-1] = (byte); \
544 #define Set_Node_Offset(node,byte) \
545 Set_Node_Offset_To_R((node)-RExC_emit_start, (byte)-RExC_start)
546 #define Set_Cur_Node_Offset Set_Node_Offset(RExC_emit, RExC_parse)
548 #define Set_Node_Length_To_R(node,len) STMT_START { \
550 MJD_OFFSET_DEBUG(("** (%d) size of node %d is %d.\n", \
551 __LINE__, (int)(node), (int)(len))); \
553 Perl_croak(aTHX_ "value of node is %d in Length macro", (int)(node)); \
555 RExC_offsets[2*(node)] = (len); \
560 #define Set_Node_Length(node,len) \
561 Set_Node_Length_To_R((node)-RExC_emit_start, len)
562 #define Set_Cur_Node_Length(len) Set_Node_Length(RExC_emit, len)
563 #define Set_Node_Cur_Length(node) \
564 Set_Node_Length(node, RExC_parse - parse_start)
566 /* Get offsets and lengths */
567 #define Node_Offset(n) (RExC_offsets[2*((n)-RExC_emit_start)-1])
568 #define Node_Length(n) (RExC_offsets[2*((n)-RExC_emit_start)])
570 #define Set_Node_Offset_Length(node,offset,len) STMT_START { \
571 Set_Node_Offset_To_R((node)-RExC_emit_start, (offset)); \
572 Set_Node_Length_To_R((node)-RExC_emit_start, (len)); \
576 #if PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS
577 #define EXPERIMENTAL_INPLACESCAN
578 #endif /*RE_TRACK_PATTERN_OFFSETS*/
580 #define DEBUG_STUDYDATA(str,data,depth) \
581 DEBUG_OPTIMISE_MORE_r(if(data){ \
582 PerlIO_printf(Perl_debug_log, \
583 "%*s" str "Pos:%"IVdf"/%"IVdf \
584 " Flags: 0x%"UVXf" Whilem_c: %"IVdf" Lcp: %"IVdf" %s", \
585 (int)(depth)*2, "", \
586 (IV)((data)->pos_min), \
587 (IV)((data)->pos_delta), \
588 (UV)((data)->flags), \
589 (IV)((data)->whilem_c), \
590 (IV)((data)->last_closep ? *((data)->last_closep) : -1), \
591 is_inf ? "INF " : "" \
593 if ((data)->last_found) \
594 PerlIO_printf(Perl_debug_log, \
595 "Last:'%s' %"IVdf":%"IVdf"/%"IVdf" %sFixed:'%s' @ %"IVdf \
596 " %sFloat: '%s' @ %"IVdf"/%"IVdf"", \
597 SvPVX_const((data)->last_found), \
598 (IV)((data)->last_end), \
599 (IV)((data)->last_start_min), \
600 (IV)((data)->last_start_max), \
601 ((data)->longest && \
602 (data)->longest==&((data)->longest_fixed)) ? "*" : "", \
603 SvPVX_const((data)->longest_fixed), \
604 (IV)((data)->offset_fixed), \
605 ((data)->longest && \
606 (data)->longest==&((data)->longest_float)) ? "*" : "", \
607 SvPVX_const((data)->longest_float), \
608 (IV)((data)->offset_float_min), \
609 (IV)((data)->offset_float_max) \
611 PerlIO_printf(Perl_debug_log,"\n"); \
614 static void clear_re(pTHX_ void *r);
616 /* Mark that we cannot extend a found fixed substring at this point.
617 Update the longest found anchored substring and the longest found
618 floating substrings if needed. */
621 S_scan_commit(pTHX_ const RExC_state_t *pRExC_state, scan_data_t *data, I32 *minlenp, int is_inf)
623 const STRLEN l = CHR_SVLEN(data->last_found);
624 const STRLEN old_l = CHR_SVLEN(*data->longest);
625 GET_RE_DEBUG_FLAGS_DECL;
627 if ((l >= old_l) && ((l > old_l) || (data->flags & SF_BEFORE_EOL))) {
628 SvSetMagicSV(*data->longest, data->last_found);
629 if (*data->longest == data->longest_fixed) {
630 data->offset_fixed = l ? data->last_start_min : data->pos_min;
631 if (data->flags & SF_BEFORE_EOL)
633 |= ((data->flags & SF_BEFORE_EOL) << SF_FIX_SHIFT_EOL);
635 data->flags &= ~SF_FIX_BEFORE_EOL;
636 data->minlen_fixed=minlenp;
637 data->lookbehind_fixed=0;
639 else { /* *data->longest == data->longest_float */
640 data->offset_float_min = l ? data->last_start_min : data->pos_min;
641 data->offset_float_max = (l
642 ? data->last_start_max
643 : data->pos_min + data->pos_delta);
644 if (is_inf || (U32)data->offset_float_max > (U32)I32_MAX)
645 data->offset_float_max = I32_MAX;
646 if (data->flags & SF_BEFORE_EOL)
648 |= ((data->flags & SF_BEFORE_EOL) << SF_FL_SHIFT_EOL);
650 data->flags &= ~SF_FL_BEFORE_EOL;
651 data->minlen_float=minlenp;
652 data->lookbehind_float=0;
655 SvCUR_set(data->last_found, 0);
657 SV * const sv = data->last_found;
658 if (SvUTF8(sv) && SvMAGICAL(sv)) {
659 MAGIC * const mg = mg_find(sv, PERL_MAGIC_utf8);
665 data->flags &= ~SF_BEFORE_EOL;
666 DEBUG_STUDYDATA("commit: ",data,0);
669 /* Can match anything (initialization) */
671 S_cl_anything(const RExC_state_t *pRExC_state, struct regnode_charclass_class *cl)
673 ANYOF_CLASS_ZERO(cl);
674 ANYOF_BITMAP_SETALL(cl);
675 cl->flags = ANYOF_EOS|ANYOF_UNICODE_ALL;
677 cl->flags |= ANYOF_LOCALE;
680 /* Can match anything (initialization) */
682 S_cl_is_anything(const struct regnode_charclass_class *cl)
686 for (value = 0; value <= ANYOF_MAX; value += 2)
687 if (ANYOF_CLASS_TEST(cl, value) && ANYOF_CLASS_TEST(cl, value + 1))
689 if (!(cl->flags & ANYOF_UNICODE_ALL))
691 if (!ANYOF_BITMAP_TESTALLSET((const void*)cl))
696 /* Can match anything (initialization) */
698 S_cl_init(const RExC_state_t *pRExC_state, struct regnode_charclass_class *cl)
700 Zero(cl, 1, struct regnode_charclass_class);
702 cl_anything(pRExC_state, cl);
706 S_cl_init_zero(const RExC_state_t *pRExC_state, struct regnode_charclass_class *cl)
708 Zero(cl, 1, struct regnode_charclass_class);
710 cl_anything(pRExC_state, cl);
712 cl->flags |= ANYOF_LOCALE;
715 /* 'And' a given class with another one. Can create false positives */
716 /* We assume that cl is not inverted */
718 S_cl_and(struct regnode_charclass_class *cl,
719 const struct regnode_charclass_class *and_with)
722 assert(and_with->type == ANYOF);
723 if (!(and_with->flags & ANYOF_CLASS)
724 && !(cl->flags & ANYOF_CLASS)
725 && (and_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
726 && !(and_with->flags & ANYOF_FOLD)
727 && !(cl->flags & ANYOF_FOLD)) {
730 if (and_with->flags & ANYOF_INVERT)
731 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
732 cl->bitmap[i] &= ~and_with->bitmap[i];
734 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
735 cl->bitmap[i] &= and_with->bitmap[i];
736 } /* XXXX: logic is complicated otherwise, leave it along for a moment. */
737 if (!(and_with->flags & ANYOF_EOS))
738 cl->flags &= ~ANYOF_EOS;
740 if (cl->flags & ANYOF_UNICODE_ALL && and_with->flags & ANYOF_UNICODE &&
741 !(and_with->flags & ANYOF_INVERT)) {
742 cl->flags &= ~ANYOF_UNICODE_ALL;
743 cl->flags |= ANYOF_UNICODE;
744 ARG_SET(cl, ARG(and_with));
746 if (!(and_with->flags & ANYOF_UNICODE_ALL) &&
747 !(and_with->flags & ANYOF_INVERT))
748 cl->flags &= ~ANYOF_UNICODE_ALL;
749 if (!(and_with->flags & (ANYOF_UNICODE|ANYOF_UNICODE_ALL)) &&
750 !(and_with->flags & ANYOF_INVERT))
751 cl->flags &= ~ANYOF_UNICODE;
754 /* 'OR' a given class with another one. Can create false positives */
755 /* We assume that cl is not inverted */
757 S_cl_or(const RExC_state_t *pRExC_state, struct regnode_charclass_class *cl, const struct regnode_charclass_class *or_with)
759 if (or_with->flags & ANYOF_INVERT) {
761 * (B1 | CL1) | (!B2 & !CL2) = (B1 | !B2 & !CL2) | (CL1 | (!B2 & !CL2))
762 * <= (B1 | !B2) | (CL1 | !CL2)
763 * which is wasteful if CL2 is small, but we ignore CL2:
764 * (B1 | CL1) | (!B2 & !CL2) <= (B1 | CL1) | !B2 = (B1 | !B2) | CL1
765 * XXXX Can we handle case-fold? Unclear:
766 * (OK1(i) | OK1(i')) | !(OK1(i) | OK1(i')) =
767 * (OK1(i) | OK1(i')) | (!OK1(i) & !OK1(i'))
769 if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
770 && !(or_with->flags & ANYOF_FOLD)
771 && !(cl->flags & ANYOF_FOLD) ) {
774 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
775 cl->bitmap[i] |= ~or_with->bitmap[i];
776 } /* XXXX: logic is complicated otherwise */
778 cl_anything(pRExC_state, cl);
781 /* (B1 | CL1) | (B2 | CL2) = (B1 | B2) | (CL1 | CL2)) */
782 if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
783 && (!(or_with->flags & ANYOF_FOLD)
784 || (cl->flags & ANYOF_FOLD)) ) {
787 /* OR char bitmap and class bitmap separately */
788 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
789 cl->bitmap[i] |= or_with->bitmap[i];
790 if (or_with->flags & ANYOF_CLASS) {
791 for (i = 0; i < ANYOF_CLASSBITMAP_SIZE; i++)
792 cl->classflags[i] |= or_with->classflags[i];
793 cl->flags |= ANYOF_CLASS;
796 else { /* XXXX: logic is complicated, leave it along for a moment. */
797 cl_anything(pRExC_state, cl);
800 if (or_with->flags & ANYOF_EOS)
801 cl->flags |= ANYOF_EOS;
803 if (cl->flags & ANYOF_UNICODE && or_with->flags & ANYOF_UNICODE &&
804 ARG(cl) != ARG(or_with)) {
805 cl->flags |= ANYOF_UNICODE_ALL;
806 cl->flags &= ~ANYOF_UNICODE;
808 if (or_with->flags & ANYOF_UNICODE_ALL) {
809 cl->flags |= ANYOF_UNICODE_ALL;
810 cl->flags &= ~ANYOF_UNICODE;
814 #define TRIE_LIST_ITEM(state,idx) (trie->states[state].trans.list)[ idx ]
815 #define TRIE_LIST_CUR(state) ( TRIE_LIST_ITEM( state, 0 ).forid )
816 #define TRIE_LIST_LEN(state) ( TRIE_LIST_ITEM( state, 0 ).newstate )
817 #define TRIE_LIST_USED(idx) ( trie->states[state].trans.list ? (TRIE_LIST_CUR( idx ) - 1) : 0 )
822 dump_trie(trie,widecharmap,revcharmap)
823 dump_trie_interim_list(trie,widecharmap,revcharmap,next_alloc)
824 dump_trie_interim_table(trie,widecharmap,revcharmap,next_alloc)
826 These routines dump out a trie in a somewhat readable format.
827 The _interim_ variants are used for debugging the interim
828 tables that are used to generate the final compressed
829 representation which is what dump_trie expects.
831 Part of the reason for their existance is to provide a form
832 of documentation as to how the different representations function.
837 Dumps the final compressed table form of the trie to Perl_debug_log.
838 Used for debugging make_trie().
842 S_dump_trie(pTHX_ const struct _reg_trie_data *trie, HV *widecharmap,
843 AV *revcharmap, U32 depth)
846 SV *sv=sv_newmortal();
847 int colwidth= widecharmap ? 6 : 4;
848 GET_RE_DEBUG_FLAGS_DECL;
851 PerlIO_printf( Perl_debug_log, "%*sChar : %-6s%-6s%-4s ",
852 (int)depth * 2 + 2,"",
853 "Match","Base","Ofs" );
855 for( state = 0 ; state < trie->uniquecharcount ; state++ ) {
856 SV ** const tmp = av_fetch( revcharmap, state, 0);
858 PerlIO_printf( Perl_debug_log, "%*s",
860 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
861 PL_colors[0], PL_colors[1],
862 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
863 PERL_PV_ESCAPE_FIRSTCHAR
868 PerlIO_printf( Perl_debug_log, "\n%*sState|-----------------------",
869 (int)depth * 2 + 2,"");
871 for( state = 0 ; state < trie->uniquecharcount ; state++ )
872 PerlIO_printf( Perl_debug_log, "%.*s", colwidth, "--------");
873 PerlIO_printf( Perl_debug_log, "\n");
875 for( state = 1 ; state < trie->statecount ; state++ ) {
876 const U32 base = trie->states[ state ].trans.base;
878 PerlIO_printf( Perl_debug_log, "%*s#%4"UVXf"|", (int)depth * 2 + 2,"", (UV)state);
880 if ( trie->states[ state ].wordnum ) {
881 PerlIO_printf( Perl_debug_log, " W%4X", trie->states[ state ].wordnum );
883 PerlIO_printf( Perl_debug_log, "%6s", "" );
886 PerlIO_printf( Perl_debug_log, " @%4"UVXf" ", (UV)base );
891 while( ( base + ofs < trie->uniquecharcount ) ||
892 ( base + ofs - trie->uniquecharcount < trie->lasttrans
893 && trie->trans[ base + ofs - trie->uniquecharcount ].check != state))
896 PerlIO_printf( Perl_debug_log, "+%2"UVXf"[ ", (UV)ofs);
898 for ( ofs = 0 ; ofs < trie->uniquecharcount ; ofs++ ) {
899 if ( ( base + ofs >= trie->uniquecharcount ) &&
900 ( base + ofs - trie->uniquecharcount < trie->lasttrans ) &&
901 trie->trans[ base + ofs - trie->uniquecharcount ].check == state )
903 PerlIO_printf( Perl_debug_log, "%*"UVXf,
905 (UV)trie->trans[ base + ofs - trie->uniquecharcount ].next );
907 PerlIO_printf( Perl_debug_log, "%*s",colwidth," ." );
911 PerlIO_printf( Perl_debug_log, "]");
914 PerlIO_printf( Perl_debug_log, "\n" );
918 Dumps a fully constructed but uncompressed trie in list form.
919 List tries normally only are used for construction when the number of
920 possible chars (trie->uniquecharcount) is very high.
921 Used for debugging make_trie().
924 S_dump_trie_interim_list(pTHX_ const struct _reg_trie_data *trie,
925 HV *widecharmap, AV *revcharmap, U32 next_alloc,
929 SV *sv=sv_newmortal();
930 int colwidth= widecharmap ? 6 : 4;
931 GET_RE_DEBUG_FLAGS_DECL;
932 /* print out the table precompression. */
933 PerlIO_printf( Perl_debug_log, "%*sState :Word | Transition Data\n%*s%s",
934 (int)depth * 2 + 2,"", (int)depth * 2 + 2,"",
935 "------:-----+-----------------\n" );
937 for( state=1 ; state < next_alloc ; state ++ ) {
940 PerlIO_printf( Perl_debug_log, "%*s %4"UVXf" :",
941 (int)depth * 2 + 2,"", (UV)state );
942 if ( ! trie->states[ state ].wordnum ) {
943 PerlIO_printf( Perl_debug_log, "%5s| ","");
945 PerlIO_printf( Perl_debug_log, "W%4x| ",
946 trie->states[ state ].wordnum
949 for( charid = 1 ; charid <= TRIE_LIST_USED( state ) ; charid++ ) {
950 SV ** const tmp = av_fetch( revcharmap, TRIE_LIST_ITEM(state,charid).forid, 0);
952 PerlIO_printf( Perl_debug_log, "%*s:%3X=%4"UVXf" | ",
954 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
955 PL_colors[0], PL_colors[1],
956 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
957 PERL_PV_ESCAPE_FIRSTCHAR
959 TRIE_LIST_ITEM(state,charid).forid,
960 (UV)TRIE_LIST_ITEM(state,charid).newstate
963 PerlIO_printf(Perl_debug_log, "\n%*s| ",
964 (int)((depth * 2) + 14), "");
967 PerlIO_printf( Perl_debug_log, "\n");
972 Dumps a fully constructed but uncompressed trie in table form.
973 This is the normal DFA style state transition table, with a few
974 twists to facilitate compression later.
975 Used for debugging make_trie().
978 S_dump_trie_interim_table(pTHX_ const struct _reg_trie_data *trie,
979 HV *widecharmap, AV *revcharmap, U32 next_alloc,
984 SV *sv=sv_newmortal();
985 int colwidth= widecharmap ? 6 : 4;
986 GET_RE_DEBUG_FLAGS_DECL;
989 print out the table precompression so that we can do a visual check
990 that they are identical.
993 PerlIO_printf( Perl_debug_log, "%*sChar : ",(int)depth * 2 + 2,"" );
995 for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) {
996 SV ** const tmp = av_fetch( revcharmap, charid, 0);
998 PerlIO_printf( Perl_debug_log, "%*s",
1000 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
1001 PL_colors[0], PL_colors[1],
1002 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
1003 PERL_PV_ESCAPE_FIRSTCHAR
1009 PerlIO_printf( Perl_debug_log, "\n%*sState+-",(int)depth * 2 + 2,"" );
1011 for( charid=0 ; charid < trie->uniquecharcount ; charid++ ) {
1012 PerlIO_printf( Perl_debug_log, "%.*s", colwidth,"--------");
1015 PerlIO_printf( Perl_debug_log, "\n" );
1017 for( state=1 ; state < next_alloc ; state += trie->uniquecharcount ) {
1019 PerlIO_printf( Perl_debug_log, "%*s%4"UVXf" : ",
1020 (int)depth * 2 + 2,"",
1021 (UV)TRIE_NODENUM( state ) );
1023 for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) {
1024 UV v=(UV)SAFE_TRIE_NODENUM( trie->trans[ state + charid ].next );
1026 PerlIO_printf( Perl_debug_log, "%*"UVXf, colwidth, v );
1028 PerlIO_printf( Perl_debug_log, "%*s", colwidth, "." );
1030 if ( ! trie->states[ TRIE_NODENUM( state ) ].wordnum ) {
1031 PerlIO_printf( Perl_debug_log, " (%4"UVXf")\n", (UV)trie->trans[ state ].check );
1033 PerlIO_printf( Perl_debug_log, " (%4"UVXf") W%4X\n", (UV)trie->trans[ state ].check,
1034 trie->states[ TRIE_NODENUM( state ) ].wordnum );
1041 /* make_trie(startbranch,first,last,tail,word_count,flags,depth)
1042 startbranch: the first branch in the whole branch sequence
1043 first : start branch of sequence of branch-exact nodes.
1044 May be the same as startbranch
1045 last : Thing following the last branch.
1046 May be the same as tail.
1047 tail : item following the branch sequence
1048 count : words in the sequence
1049 flags : currently the OP() type we will be building one of /EXACT(|F|Fl)/
1050 depth : indent depth
1052 Inplace optimizes a sequence of 2 or more Branch-Exact nodes into a TRIE node.
1054 A trie is an N'ary tree where the branches are determined by digital
1055 decomposition of the key. IE, at the root node you look up the 1st character and
1056 follow that branch repeat until you find the end of the branches. Nodes can be
1057 marked as "accepting" meaning they represent a complete word. Eg:
1061 would convert into the following structure. Numbers represent states, letters
1062 following numbers represent valid transitions on the letter from that state, if
1063 the number is in square brackets it represents an accepting state, otherwise it
1064 will be in parenthesis.
1066 +-h->+-e->[3]-+-r->(8)-+-s->[9]
1070 (1) +-i->(6)-+-s->[7]
1072 +-s->(3)-+-h->(4)-+-e->[5]
1074 Accept Word Mapping: 3=>1 (he),5=>2 (she), 7=>3 (his), 9=>4 (hers)
1076 This shows that when matching against the string 'hers' we will begin at state 1
1077 read 'h' and move to state 2, read 'e' and move to state 3 which is accepting,
1078 then read 'r' and go to state 8 followed by 's' which takes us to state 9 which
1079 is also accepting. Thus we know that we can match both 'he' and 'hers' with a
1080 single traverse. We store a mapping from accepting to state to which word was
1081 matched, and then when we have multiple possibilities we try to complete the
1082 rest of the regex in the order in which they occured in the alternation.
1084 The only prior NFA like behaviour that would be changed by the TRIE support is
1085 the silent ignoring of duplicate alternations which are of the form:
1087 / (DUPE|DUPE) X? (?{ ... }) Y /x
1089 Thus EVAL blocks follwing a trie may be called a different number of times with
1090 and without the optimisation. With the optimisations dupes will be silently
1091 ignored. This inconsistant behaviour of EVAL type nodes is well established as
1092 the following demonstrates:
1094 'words'=~/(word|word|word)(?{ print $1 })[xyz]/
1096 which prints out 'word' three times, but
1098 'words'=~/(word|word|word)(?{ print $1 })S/
1100 which doesnt print it out at all. This is due to other optimisations kicking in.
1102 Example of what happens on a structural level:
1104 The regexp /(ac|ad|ab)+/ will produce the folowing debug output:
1106 1: CURLYM[1] {1,32767}(18)
1117 This would be optimizable with startbranch=5, first=5, last=16, tail=16
1118 and should turn into:
1120 1: CURLYM[1] {1,32767}(18)
1122 [Words:3 Chars Stored:6 Unique Chars:4 States:5 NCP:1]
1130 Cases where tail != last would be like /(?foo|bar)baz/:
1140 which would be optimizable with startbranch=1, first=1, last=7, tail=8
1141 and would end up looking like:
1144 [Words:2 Chars Stored:6 Unique Chars:5 States:7 NCP:1]
1151 d = uvuni_to_utf8_flags(d, uv, 0);
1153 is the recommended Unicode-aware way of saying
1158 #define TRIE_STORE_REVCHAR \
1160 SV *tmp = newSVpvs(""); \
1161 if (UTF) SvUTF8_on(tmp); \
1162 Perl_sv_catpvf( aTHX_ tmp, "%c", (int)uvc ); \
1163 av_push( revcharmap, tmp ); \
1166 #define TRIE_READ_CHAR STMT_START { \
1170 if ( foldlen > 0 ) { \
1171 uvc = utf8n_to_uvuni( scan, UTF8_MAXLEN, &len, uniflags ); \
1176 uvc = utf8n_to_uvuni( (const U8*)uc, UTF8_MAXLEN, &len, uniflags);\
1177 uvc = to_uni_fold( uvc, foldbuf, &foldlen ); \
1178 foldlen -= UNISKIP( uvc ); \
1179 scan = foldbuf + UNISKIP( uvc ); \
1182 uvc = utf8n_to_uvuni( (const U8*)uc, UTF8_MAXLEN, &len, uniflags);\
1192 #define TRIE_LIST_PUSH(state,fid,ns) STMT_START { \
1193 if ( TRIE_LIST_CUR( state ) >=TRIE_LIST_LEN( state ) ) { \
1194 U32 ging = TRIE_LIST_LEN( state ) *= 2; \
1195 Renew( trie->states[ state ].trans.list, ging, reg_trie_trans_le ); \
1197 TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).forid = fid; \
1198 TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).newstate = ns; \
1199 TRIE_LIST_CUR( state )++; \
1202 #define TRIE_LIST_NEW(state) STMT_START { \
1203 Newxz( trie->states[ state ].trans.list, \
1204 4, reg_trie_trans_le ); \
1205 TRIE_LIST_CUR( state ) = 1; \
1206 TRIE_LIST_LEN( state ) = 4; \
1209 #define TRIE_HANDLE_WORD(state) STMT_START { \
1210 U16 dupe= trie->states[ state ].wordnum; \
1211 regnode * const noper_next = regnext( noper ); \
1213 if (trie->wordlen) \
1214 trie->wordlen[ curword ] = wordlen; \
1216 /* store the word for dumping */ \
1218 if (OP(noper) != NOTHING) \
1219 tmp = newSVpvn(STRING(noper), STR_LEN(noper)); \
1221 tmp = newSVpvn( "", 0 ); \
1222 if ( UTF ) SvUTF8_on( tmp ); \
1223 av_push( trie_words, tmp ); \
1228 if ( noper_next < tail ) { \
1230 trie->jump = (U16 *) PerlMemShared_calloc( word_count + 1, sizeof(U16) ); \
1231 trie->jump[curword] = (U16)(noper_next - convert); \
1233 jumper = noper_next; \
1235 nextbranch= regnext(cur); \
1239 /* So it's a dupe. This means we need to maintain a */\
1240 /* linked-list from the first to the next. */\
1241 /* we only allocate the nextword buffer when there */\
1242 /* a dupe, so first time we have to do the allocation */\
1243 if (!trie->nextword) \
1244 trie->nextword = (U16 *) \
1245 PerlMemShared_calloc( word_count + 1, sizeof(U16)); \
1246 while ( trie->nextword[dupe] ) \
1247 dupe= trie->nextword[dupe]; \
1248 trie->nextword[dupe]= curword; \
1250 /* we haven't inserted this word yet. */ \
1251 trie->states[ state ].wordnum = curword; \
1256 #define TRIE_TRANS_STATE(state,base,ucharcount,charid,special) \
1257 ( ( base + charid >= ucharcount \
1258 && base + charid < ubound \
1259 && state == trie->trans[ base - ucharcount + charid ].check \
1260 && trie->trans[ base - ucharcount + charid ].next ) \
1261 ? trie->trans[ base - ucharcount + charid ].next \
1262 : ( state==1 ? special : 0 ) \
1266 #define MADE_JUMP_TRIE 2
1267 #define MADE_EXACT_TRIE 4
1270 S_make_trie(pTHX_ RExC_state_t *pRExC_state, regnode *startbranch, regnode *first, regnode *last, regnode *tail, U32 word_count, U32 flags, U32 depth)
1273 /* first pass, loop through and scan words */
1274 reg_trie_data *trie;
1275 HV *widecharmap = NULL;
1276 AV *revcharmap = newAV();
1278 const U32 uniflags = UTF8_ALLOW_DEFAULT;
1283 regnode *jumper = NULL;
1284 regnode *nextbranch = NULL;
1285 regnode *convert = NULL;
1286 /* we just use folder as a flag in utf8 */
1287 const U8 * const folder = ( flags == EXACTF
1289 : ( flags == EXACTFL
1296 const U32 data_slot = add_data( pRExC_state, 4, "tuuu" );
1297 AV *trie_words = NULL;
1298 /* along with revcharmap, this only used during construction but both are
1299 * useful during debugging so we store them in the struct when debugging.
1302 const U32 data_slot = add_data( pRExC_state, 2, "tu" );
1303 STRLEN trie_charcount=0;
1305 SV *re_trie_maxbuff;
1306 GET_RE_DEBUG_FLAGS_DECL;
1308 PERL_UNUSED_ARG(depth);
1311 trie = (reg_trie_data *) PerlMemShared_calloc( 1, sizeof(reg_trie_data) );
1313 trie->startstate = 1;
1314 trie->wordcount = word_count;
1315 RExC_rxi->data->data[ data_slot ] = (void*)trie;
1316 trie->charmap = (U16 *) PerlMemShared_calloc( 256, sizeof(U16) );
1317 if (!(UTF && folder))
1318 trie->bitmap = (char *) PerlMemShared_calloc( ANYOF_BITMAP_SIZE, 1 );
1320 trie_words = newAV();
1323 re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, 1);
1324 if (!SvIOK(re_trie_maxbuff)) {
1325 sv_setiv(re_trie_maxbuff, RE_TRIE_MAXBUF_INIT);
1328 PerlIO_printf( Perl_debug_log,
1329 "%*smake_trie start==%d, first==%d, last==%d, tail==%d depth=%d\n",
1330 (int)depth * 2 + 2, "",
1331 REG_NODE_NUM(startbranch),REG_NODE_NUM(first),
1332 REG_NODE_NUM(last), REG_NODE_NUM(tail),
1336 /* Find the node we are going to overwrite */
1337 if ( first == startbranch && OP( last ) != BRANCH ) {
1338 /* whole branch chain */
1341 /* branch sub-chain */
1342 convert = NEXTOPER( first );
1345 /* -- First loop and Setup --
1347 We first traverse the branches and scan each word to determine if it
1348 contains widechars, and how many unique chars there are, this is
1349 important as we have to build a table with at least as many columns as we
1352 We use an array of integers to represent the character codes 0..255
1353 (trie->charmap) and we use a an HV* to store unicode characters. We use the
1354 native representation of the character value as the key and IV's for the
1357 *TODO* If we keep track of how many times each character is used we can
1358 remap the columns so that the table compression later on is more
1359 efficient in terms of memory by ensuring most common value is in the
1360 middle and the least common are on the outside. IMO this would be better
1361 than a most to least common mapping as theres a decent chance the most
1362 common letter will share a node with the least common, meaning the node
1363 will not be compressable. With a middle is most common approach the worst
1364 case is when we have the least common nodes twice.
1368 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
1369 regnode * const noper = NEXTOPER( cur );
1370 const U8 *uc = (U8*)STRING( noper );
1371 const U8 * const e = uc + STR_LEN( noper );
1373 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
1374 const U8 *scan = (U8*)NULL;
1375 U32 wordlen = 0; /* required init */
1378 if (OP(noper) == NOTHING) {
1383 TRIE_BITMAP_SET(trie,*uc);
1384 if ( folder ) TRIE_BITMAP_SET(trie,folder[ *uc ]);
1386 for ( ; uc < e ; uc += len ) {
1387 TRIE_CHARCOUNT(trie)++;
1391 if ( !trie->charmap[ uvc ] ) {
1392 trie->charmap[ uvc ]=( ++trie->uniquecharcount );
1394 trie->charmap[ folder[ uvc ] ] = trie->charmap[ uvc ];
1400 widecharmap = newHV();
1402 svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 1 );
1405 Perl_croak( aTHX_ "error creating/fetching widecharmap entry for 0x%"UVXf, uvc );
1407 if ( !SvTRUE( *svpp ) ) {
1408 sv_setiv( *svpp, ++trie->uniquecharcount );
1413 if( cur == first ) {
1416 } else if (chars < trie->minlen) {
1418 } else if (chars > trie->maxlen) {
1422 } /* end first pass */
1423 DEBUG_TRIE_COMPILE_r(
1424 PerlIO_printf( Perl_debug_log, "%*sTRIE(%s): W:%d C:%d Uq:%d Min:%d Max:%d\n",
1425 (int)depth * 2 + 2,"",
1426 ( widecharmap ? "UTF8" : "NATIVE" ), (int)word_count,
1427 (int)TRIE_CHARCOUNT(trie), trie->uniquecharcount,
1428 (int)trie->minlen, (int)trie->maxlen )
1430 trie->wordlen = (U32 *) PerlMemShared_calloc( word_count, sizeof(U32) );
1433 We now know what we are dealing with in terms of unique chars and
1434 string sizes so we can calculate how much memory a naive
1435 representation using a flat table will take. If it's over a reasonable
1436 limit (as specified by ${^RE_TRIE_MAXBUF}) we use a more memory
1437 conservative but potentially much slower representation using an array
1440 At the end we convert both representations into the same compressed
1441 form that will be used in regexec.c for matching with. The latter
1442 is a form that cannot be used to construct with but has memory
1443 properties similar to the list form and access properties similar
1444 to the table form making it both suitable for fast searches and
1445 small enough that its feasable to store for the duration of a program.
1447 See the comment in the code where the compressed table is produced
1448 inplace from the flat tabe representation for an explanation of how
1449 the compression works.
1454 if ( (IV)( ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount + 1) > SvIV(re_trie_maxbuff) ) {
1456 Second Pass -- Array Of Lists Representation
1458 Each state will be represented by a list of charid:state records
1459 (reg_trie_trans_le) the first such element holds the CUR and LEN
1460 points of the allocated array. (See defines above).
1462 We build the initial structure using the lists, and then convert
1463 it into the compressed table form which allows faster lookups
1464 (but cant be modified once converted).
1467 STRLEN transcount = 1;
1469 DEBUG_TRIE_COMPILE_MORE_r( PerlIO_printf( Perl_debug_log,
1470 "%*sCompiling trie using list compiler\n",
1471 (int)depth * 2 + 2, ""));
1473 trie->states = (reg_trie_state *)
1474 PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2,
1475 sizeof(reg_trie_state) );
1479 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
1481 regnode * const noper = NEXTOPER( cur );
1482 U8 *uc = (U8*)STRING( noper );
1483 const U8 * const e = uc + STR_LEN( noper );
1484 U32 state = 1; /* required init */
1485 U16 charid = 0; /* sanity init */
1486 U8 *scan = (U8*)NULL; /* sanity init */
1487 STRLEN foldlen = 0; /* required init */
1488 U32 wordlen = 0; /* required init */
1489 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
1491 if (OP(noper) != NOTHING) {
1492 for ( ; uc < e ; uc += len ) {
1497 charid = trie->charmap[ uvc ];
1499 SV** const svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 0);
1503 charid=(U16)SvIV( *svpp );
1506 /* charid is now 0 if we dont know the char read, or nonzero if we do */
1513 if ( !trie->states[ state ].trans.list ) {
1514 TRIE_LIST_NEW( state );
1516 for ( check = 1; check <= TRIE_LIST_USED( state ); check++ ) {
1517 if ( TRIE_LIST_ITEM( state, check ).forid == charid ) {
1518 newstate = TRIE_LIST_ITEM( state, check ).newstate;
1523 newstate = next_alloc++;
1524 TRIE_LIST_PUSH( state, charid, newstate );
1529 Perl_croak( aTHX_ "panic! In trie construction, no char mapping for %"IVdf, uvc );
1533 TRIE_HANDLE_WORD(state);
1535 } /* end second pass */
1537 /* next alloc is the NEXT state to be allocated */
1538 trie->statecount = next_alloc;
1539 trie->states = (reg_trie_state *)
1540 PerlMemShared_realloc( trie->states,
1542 * sizeof(reg_trie_state) );
1544 /* and now dump it out before we compress it */
1545 DEBUG_TRIE_COMPILE_MORE_r(dump_trie_interim_list(trie, widecharmap,
1546 revcharmap, next_alloc,
1550 trie->trans = (reg_trie_trans *)
1551 PerlMemShared_calloc( transcount, sizeof(reg_trie_trans) );
1558 for( state=1 ; state < next_alloc ; state ++ ) {
1562 DEBUG_TRIE_COMPILE_MORE_r(
1563 PerlIO_printf( Perl_debug_log, "tp: %d zp: %d ",tp,zp)
1567 if (trie->states[state].trans.list) {
1568 U16 minid=TRIE_LIST_ITEM( state, 1).forid;
1572 for( idx = 2 ; idx <= TRIE_LIST_USED( state ) ; idx++ ) {
1573 const U16 forid = TRIE_LIST_ITEM( state, idx).forid;
1574 if ( forid < minid ) {
1576 } else if ( forid > maxid ) {
1580 if ( transcount < tp + maxid - minid + 1) {
1582 trie->trans = (reg_trie_trans *)
1583 PerlMemShared_realloc( trie->trans,
1585 * sizeof(reg_trie_trans) );
1586 Zero( trie->trans + (transcount / 2), transcount / 2 , reg_trie_trans );
1588 base = trie->uniquecharcount + tp - minid;
1589 if ( maxid == minid ) {
1591 for ( ; zp < tp ; zp++ ) {
1592 if ( ! trie->trans[ zp ].next ) {
1593 base = trie->uniquecharcount + zp - minid;
1594 trie->trans[ zp ].next = TRIE_LIST_ITEM( state, 1).newstate;
1595 trie->trans[ zp ].check = state;
1601 trie->trans[ tp ].next = TRIE_LIST_ITEM( state, 1).newstate;
1602 trie->trans[ tp ].check = state;
1607 for ( idx=1; idx <= TRIE_LIST_USED( state ) ; idx++ ) {
1608 const U32 tid = base - trie->uniquecharcount + TRIE_LIST_ITEM( state, idx ).forid;
1609 trie->trans[ tid ].next = TRIE_LIST_ITEM( state, idx ).newstate;
1610 trie->trans[ tid ].check = state;
1612 tp += ( maxid - minid + 1 );
1614 Safefree(trie->states[ state ].trans.list);
1617 DEBUG_TRIE_COMPILE_MORE_r(
1618 PerlIO_printf( Perl_debug_log, " base: %d\n",base);
1621 trie->states[ state ].trans.base=base;
1623 trie->lasttrans = tp + 1;
1627 Second Pass -- Flat Table Representation.
1629 we dont use the 0 slot of either trans[] or states[] so we add 1 to each.
1630 We know that we will need Charcount+1 trans at most to store the data
1631 (one row per char at worst case) So we preallocate both structures
1632 assuming worst case.
1634 We then construct the trie using only the .next slots of the entry
1637 We use the .check field of the first entry of the node temporarily to
1638 make compression both faster and easier by keeping track of how many non
1639 zero fields are in the node.
1641 Since trans are numbered from 1 any 0 pointer in the table is a FAIL
1644 There are two terms at use here: state as a TRIE_NODEIDX() which is a
1645 number representing the first entry of the node, and state as a
1646 TRIE_NODENUM() which is the trans number. state 1 is TRIE_NODEIDX(1) and
1647 TRIE_NODENUM(1), state 2 is TRIE_NODEIDX(2) and TRIE_NODENUM(3) if there
1648 are 2 entrys per node. eg:
1656 The table is internally in the right hand, idx form. However as we also
1657 have to deal with the states array which is indexed by nodenum we have to
1658 use TRIE_NODENUM() to convert.
1661 DEBUG_TRIE_COMPILE_MORE_r( PerlIO_printf( Perl_debug_log,
1662 "%*sCompiling trie using table compiler\n",
1663 (int)depth * 2 + 2, ""));
1665 trie->trans = (reg_trie_trans *)
1666 PerlMemShared_calloc( ( TRIE_CHARCOUNT(trie) + 1 )
1667 * trie->uniquecharcount + 1,
1668 sizeof(reg_trie_trans) );
1669 trie->states = (reg_trie_state *)
1670 PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2,
1671 sizeof(reg_trie_state) );
1672 next_alloc = trie->uniquecharcount + 1;
1675 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
1677 regnode * const noper = NEXTOPER( cur );
1678 const U8 *uc = (U8*)STRING( noper );
1679 const U8 * const e = uc + STR_LEN( noper );
1681 U32 state = 1; /* required init */
1683 U16 charid = 0; /* sanity init */
1684 U32 accept_state = 0; /* sanity init */
1685 U8 *scan = (U8*)NULL; /* sanity init */
1687 STRLEN foldlen = 0; /* required init */
1688 U32 wordlen = 0; /* required init */
1689 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
1691 if ( OP(noper) != NOTHING ) {
1692 for ( ; uc < e ; uc += len ) {
1697 charid = trie->charmap[ uvc ];
1699 SV* const * const svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 0);
1700 charid = svpp ? (U16)SvIV(*svpp) : 0;
1704 if ( !trie->trans[ state + charid ].next ) {
1705 trie->trans[ state + charid ].next = next_alloc;
1706 trie->trans[ state ].check++;
1707 next_alloc += trie->uniquecharcount;
1709 state = trie->trans[ state + charid ].next;
1711 Perl_croak( aTHX_ "panic! In trie construction, no char mapping for %"IVdf, uvc );
1713 /* charid is now 0 if we dont know the char read, or nonzero if we do */
1716 accept_state = TRIE_NODENUM( state );
1717 TRIE_HANDLE_WORD(accept_state);
1719 } /* end second pass */
1721 /* and now dump it out before we compress it */
1722 DEBUG_TRIE_COMPILE_MORE_r(dump_trie_interim_table(trie, widecharmap,
1724 next_alloc, depth+1));
1728 * Inplace compress the table.*
1730 For sparse data sets the table constructed by the trie algorithm will
1731 be mostly 0/FAIL transitions or to put it another way mostly empty.
1732 (Note that leaf nodes will not contain any transitions.)
1734 This algorithm compresses the tables by eliminating most such
1735 transitions, at the cost of a modest bit of extra work during lookup:
1737 - Each states[] entry contains a .base field which indicates the
1738 index in the state[] array wheres its transition data is stored.
1740 - If .base is 0 there are no valid transitions from that node.
1742 - If .base is nonzero then charid is added to it to find an entry in
1745 -If trans[states[state].base+charid].check!=state then the
1746 transition is taken to be a 0/Fail transition. Thus if there are fail
1747 transitions at the front of the node then the .base offset will point
1748 somewhere inside the previous nodes data (or maybe even into a node
1749 even earlier), but the .check field determines if the transition is
1753 The following process inplace converts the table to the compressed
1754 table: We first do not compress the root node 1,and mark its all its
1755 .check pointers as 1 and set its .base pointer as 1 as well. This
1756 allows to do a DFA construction from the compressed table later, and
1757 ensures that any .base pointers we calculate later are greater than
1760 - We set 'pos' to indicate the first entry of the second node.
1762 - We then iterate over the columns of the node, finding the first and
1763 last used entry at l and m. We then copy l..m into pos..(pos+m-l),
1764 and set the .check pointers accordingly, and advance pos
1765 appropriately and repreat for the next node. Note that when we copy
1766 the next pointers we have to convert them from the original
1767 NODEIDX form to NODENUM form as the former is not valid post
1770 - If a node has no transitions used we mark its base as 0 and do not
1771 advance the pos pointer.
1773 - If a node only has one transition we use a second pointer into the
1774 structure to fill in allocated fail transitions from other states.
1775 This pointer is independent of the main pointer and scans forward
1776 looking for null transitions that are allocated to a state. When it
1777 finds one it writes the single transition into the "hole". If the
1778 pointer doesnt find one the single transition is appended as normal.
1780 - Once compressed we can Renew/realloc the structures to release the
1783 See "Table-Compression Methods" in sec 3.9 of the Red Dragon,
1784 specifically Fig 3.47 and the associated pseudocode.
1788 const U32 laststate = TRIE_NODENUM( next_alloc );
1791 trie->statecount = laststate;
1793 for ( state = 1 ; state < laststate ; state++ ) {
1795 const U32 stateidx = TRIE_NODEIDX( state );
1796 const U32 o_used = trie->trans[ stateidx ].check;
1797 U32 used = trie->trans[ stateidx ].check;
1798 trie->trans[ stateidx ].check = 0;
1800 for ( charid = 0 ; used && charid < trie->uniquecharcount ; charid++ ) {
1801 if ( flag || trie->trans[ stateidx + charid ].next ) {
1802 if ( trie->trans[ stateidx + charid ].next ) {
1804 for ( ; zp < pos ; zp++ ) {
1805 if ( ! trie->trans[ zp ].next ) {
1809 trie->states[ state ].trans.base = zp + trie->uniquecharcount - charid ;
1810 trie->trans[ zp ].next = SAFE_TRIE_NODENUM( trie->trans[ stateidx + charid ].next );
1811 trie->trans[ zp ].check = state;
1812 if ( ++zp > pos ) pos = zp;
1819 trie->states[ state ].trans.base = pos + trie->uniquecharcount - charid ;
1821 trie->trans[ pos ].next = SAFE_TRIE_NODENUM( trie->trans[ stateidx + charid ].next );
1822 trie->trans[ pos ].check = state;
1827 trie->lasttrans = pos + 1;
1828 trie->states = (reg_trie_state *)
1829 PerlMemShared_realloc( trie->states, laststate
1830 * sizeof(reg_trie_state) );
1831 DEBUG_TRIE_COMPILE_MORE_r(
1832 PerlIO_printf( Perl_debug_log,
1833 "%*sAlloc: %d Orig: %"IVdf" elements, Final:%"IVdf". Savings of %%%5.2f\n",
1834 (int)depth * 2 + 2,"",
1835 (int)( ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount + 1 ),
1838 ( ( next_alloc - pos ) * 100 ) / (double)next_alloc );
1841 } /* end table compress */
1843 DEBUG_TRIE_COMPILE_MORE_r(
1844 PerlIO_printf(Perl_debug_log, "%*sStatecount:%"UVxf" Lasttrans:%"UVxf"\n",
1845 (int)depth * 2 + 2, "",
1846 (UV)trie->statecount,
1847 (UV)trie->lasttrans)
1849 /* resize the trans array to remove unused space */
1850 trie->trans = (reg_trie_trans *)
1851 PerlMemShared_realloc( trie->trans, trie->lasttrans
1852 * sizeof(reg_trie_trans) );
1854 /* and now dump out the compressed format */
1855 DEBUG_TRIE_COMPILE_r(dump_trie(trie, widecharmap, revcharmap, depth+1));
1857 { /* Modify the program and insert the new TRIE node*/
1858 U8 nodetype =(U8)(flags & 0xFF);
1862 regnode *optimize = NULL;
1863 #ifdef RE_TRACK_PATTERN_OFFSETS
1866 U32 mjd_nodelen = 0;
1867 #endif /* RE_TRACK_PATTERN_OFFSETS */
1868 #endif /* DEBUGGING */
1870 This means we convert either the first branch or the first Exact,
1871 depending on whether the thing following (in 'last') is a branch
1872 or not and whther first is the startbranch (ie is it a sub part of
1873 the alternation or is it the whole thing.)
1874 Assuming its a sub part we conver the EXACT otherwise we convert
1875 the whole branch sequence, including the first.
1877 /* Find the node we are going to overwrite */
1878 if ( first != startbranch || OP( last ) == BRANCH ) {
1879 /* branch sub-chain */
1880 NEXT_OFF( first ) = (U16)(last - first);
1881 #ifdef RE_TRACK_PATTERN_OFFSETS
1883 mjd_offset= Node_Offset((convert));
1884 mjd_nodelen= Node_Length((convert));
1887 /* whole branch chain */
1889 #ifdef RE_TRACK_PATTERN_OFFSETS
1892 const regnode *nop = NEXTOPER( convert );
1893 mjd_offset= Node_Offset((nop));
1894 mjd_nodelen= Node_Length((nop));
1898 PerlIO_printf(Perl_debug_log, "%*sMJD offset:%"UVuf" MJD length:%"UVuf"\n",
1899 (int)depth * 2 + 2, "",
1900 (UV)mjd_offset, (UV)mjd_nodelen)
1903 /* But first we check to see if there is a common prefix we can
1904 split out as an EXACT and put in front of the TRIE node. */
1905 trie->startstate= 1;
1906 if ( trie->bitmap && !widecharmap && !trie->jump ) {
1908 for ( state = 1 ; state < trie->statecount-1 ; state++ ) {
1912 const U32 base = trie->states[ state ].trans.base;
1914 if ( trie->states[state].wordnum )
1917 for ( ofs = 0 ; ofs < trie->uniquecharcount ; ofs++ ) {
1918 if ( ( base + ofs >= trie->uniquecharcount ) &&
1919 ( base + ofs - trie->uniquecharcount < trie->lasttrans ) &&
1920 trie->trans[ base + ofs - trie->uniquecharcount ].check == state )
1922 if ( ++count > 1 ) {
1923 SV **tmp = av_fetch( revcharmap, ofs, 0);
1924 const U8 *ch = (U8*)SvPV_nolen_const( *tmp );
1925 if ( state == 1 ) break;
1927 Zero(trie->bitmap, ANYOF_BITMAP_SIZE, char);
1929 PerlIO_printf(Perl_debug_log,
1930 "%*sNew Start State=%"UVuf" Class: [",
1931 (int)depth * 2 + 2, "",
1934 SV ** const tmp = av_fetch( revcharmap, idx, 0);
1935 const U8 * const ch = (U8*)SvPV_nolen_const( *tmp );
1937 TRIE_BITMAP_SET(trie,*ch);
1939 TRIE_BITMAP_SET(trie, folder[ *ch ]);
1941 PerlIO_printf(Perl_debug_log, (char*)ch)
1945 TRIE_BITMAP_SET(trie,*ch);
1947 TRIE_BITMAP_SET(trie,folder[ *ch ]);
1948 DEBUG_OPTIMISE_r(PerlIO_printf( Perl_debug_log,"%s", ch));
1954 SV **tmp = av_fetch( revcharmap, idx, 0);
1955 char *ch = SvPV_nolen( *tmp );
1957 SV *sv=sv_newmortal();
1958 PerlIO_printf( Perl_debug_log,
1959 "%*sPrefix State: %"UVuf" Idx:%"UVuf" Char='%s'\n",
1960 (int)depth * 2 + 2, "",
1962 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 6,
1963 PL_colors[0], PL_colors[1],
1964 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
1965 PERL_PV_ESCAPE_FIRSTCHAR
1970 OP( convert ) = nodetype;
1971 str=STRING(convert);
1982 DEBUG_OPTIMISE_r(PerlIO_printf( Perl_debug_log,"]\n"));
1988 regnode *n = convert+NODE_SZ_STR(convert);
1989 NEXT_OFF(convert) = NODE_SZ_STR(convert);
1990 trie->startstate = state;
1991 trie->minlen -= (state - 1);
1992 trie->maxlen -= (state - 1);
1994 regnode *fix = convert;
1995 U32 word = trie->wordcount;
1997 Set_Node_Offset_Length(convert, mjd_offset, state - 1);
1998 while( ++fix < n ) {
1999 Set_Node_Offset_Length(fix, 0, 0);
2002 SV ** const tmp = av_fetch( trie_words, word, 0 );
2004 if ( STR_LEN(convert) <= SvCUR(*tmp) )
2005 sv_chop(*tmp, SvPV_nolen(*tmp) + STR_LEN(convert));
2007 sv_chop(*tmp, SvPV_nolen(*tmp) + SvCUR(*tmp));
2014 NEXT_OFF(convert) = (U16)(tail - convert);
2015 DEBUG_r(optimize= n);
2021 if ( trie->maxlen ) {
2022 NEXT_OFF( convert ) = (U16)(tail - convert);
2023 ARG_SET( convert, data_slot );
2024 /* Store the offset to the first unabsorbed branch in
2025 jump[0], which is otherwise unused by the jump logic.
2026 We use this when dumping a trie and during optimisation. */
2028 trie->jump[0] = (U16)(nextbranch - convert);
2031 if ( !trie->states[trie->startstate].wordnum && trie->bitmap &&
2032 ( (char *)jumper - (char *)convert) >= (int)sizeof(struct regnode_charclass) )
2034 OP( convert ) = TRIEC;
2035 Copy(trie->bitmap, ((struct regnode_charclass *)convert)->bitmap, ANYOF_BITMAP_SIZE, char);
2036 PerlMemShared_free(trie->bitmap);
2039 OP( convert ) = TRIE;
2041 /* store the type in the flags */
2042 convert->flags = nodetype;
2046 + regarglen[ OP( convert ) ];
2048 /* XXX We really should free up the resource in trie now,
2049 as we won't use them - (which resources?) dmq */
2051 /* needed for dumping*/
2052 DEBUG_r(if (optimize) {
2053 regnode *opt = convert;
2055 while ( ++opt < optimize) {
2056 Set_Node_Offset_Length(opt,0,0);
2059 Try to clean up some of the debris left after the
2062 while( optimize < jumper ) {
2063 mjd_nodelen += Node_Length((optimize));
2064 OP( optimize ) = OPTIMIZED;
2065 Set_Node_Offset_Length(optimize,0,0);
2068 Set_Node_Offset_Length(convert,mjd_offset,mjd_nodelen);
2070 } /* end node insert */
2071 RExC_rxi->data->data[ data_slot + 1 ] = (void*)widecharmap;
2073 RExC_rxi->data->data[ data_slot + TRIE_WORDS_OFFSET ] = (void*)trie_words;
2074 RExC_rxi->data->data[ data_slot + 3 ] = (void*)revcharmap;
2076 SvREFCNT_dec(revcharmap);
2080 : trie->startstate>1
2086 S_make_trie_failtable(pTHX_ RExC_state_t *pRExC_state, regnode *source, regnode *stclass, U32 depth)
2088 /* The Trie is constructed and compressed now so we can build a fail array now if its needed
2090 This is basically the Aho-Corasick algorithm. Its from exercise 3.31 and 3.32 in the
2091 "Red Dragon" -- Compilers, principles, techniques, and tools. Aho, Sethi, Ullman 1985/88
2094 We find the fail state for each state in the trie, this state is the longest proper
2095 suffix of the current states 'word' that is also a proper prefix of another word in our
2096 trie. State 1 represents the word '' and is the thus the default fail state. This allows
2097 the DFA not to have to restart after its tried and failed a word at a given point, it
2098 simply continues as though it had been matching the other word in the first place.
2100 'abcdgu'=~/abcdefg|cdgu/
2101 When we get to 'd' we are still matching the first word, we would encounter 'g' which would
2102 fail, which would bring use to the state representing 'd' in the second word where we would
2103 try 'g' and succeed, prodceding to match 'cdgu'.
2105 /* add a fail transition */
2106 const U32 trie_offset = ARG(source);
2107 reg_trie_data *trie=(reg_trie_data *)RExC_rxi->data->data[trie_offset];
2109 const U32 ucharcount = trie->uniquecharcount;
2110 const U32 numstates = trie->statecount;
2111 const U32 ubound = trie->lasttrans + ucharcount;
2115 U32 base = trie->states[ 1 ].trans.base;
2118 const U32 data_slot = add_data( pRExC_state, 1, "T" );
2119 GET_RE_DEBUG_FLAGS_DECL;
2121 PERL_UNUSED_ARG(depth);
2125 ARG_SET( stclass, data_slot );
2126 aho = (reg_ac_data *) PerlMemShared_calloc( 1, sizeof(reg_ac_data) );
2127 RExC_rxi->data->data[ data_slot ] = (void*)aho;
2128 aho->trie=trie_offset;
2129 aho->states=(reg_trie_state *)PerlMemShared_malloc( numstates * sizeof(reg_trie_state) );
2130 Copy( trie->states, aho->states, numstates, reg_trie_state );
2131 Newxz( q, numstates, U32);
2132 aho->fail = (U32 *) PerlMemShared_calloc( numstates, sizeof(U32) );
2135 /* initialize fail[0..1] to be 1 so that we always have
2136 a valid final fail state */
2137 fail[ 0 ] = fail[ 1 ] = 1;
2139 for ( charid = 0; charid < ucharcount ; charid++ ) {
2140 const U32 newstate = TRIE_TRANS_STATE( 1, base, ucharcount, charid, 0 );
2142 q[ q_write ] = newstate;
2143 /* set to point at the root */
2144 fail[ q[ q_write++ ] ]=1;
2147 while ( q_read < q_write) {
2148 const U32 cur = q[ q_read++ % numstates ];
2149 base = trie->states[ cur ].trans.base;
2151 for ( charid = 0 ; charid < ucharcount ; charid++ ) {
2152 const U32 ch_state = TRIE_TRANS_STATE( cur, base, ucharcount, charid, 1 );
2154 U32 fail_state = cur;
2157 fail_state = fail[ fail_state ];
2158 fail_base = aho->states[ fail_state ].trans.base;
2159 } while ( !TRIE_TRANS_STATE( fail_state, fail_base, ucharcount, charid, 1 ) );
2161 fail_state = TRIE_TRANS_STATE( fail_state, fail_base, ucharcount, charid, 1 );
2162 fail[ ch_state ] = fail_state;
2163 if ( !aho->states[ ch_state ].wordnum && aho->states[ fail_state ].wordnum )
2165 aho->states[ ch_state ].wordnum = aho->states[ fail_state ].wordnum;
2167 q[ q_write++ % numstates] = ch_state;
2171 /* restore fail[0..1] to 0 so that we "fall out" of the AC loop
2172 when we fail in state 1, this allows us to use the
2173 charclass scan to find a valid start char. This is based on the principle
2174 that theres a good chance the string being searched contains lots of stuff
2175 that cant be a start char.
2177 fail[ 0 ] = fail[ 1 ] = 0;
2178 DEBUG_TRIE_COMPILE_r({
2179 PerlIO_printf(Perl_debug_log,
2180 "%*sStclass Failtable (%"UVuf" states): 0",
2181 (int)(depth * 2), "", (UV)numstates
2183 for( q_read=1; q_read<numstates; q_read++ ) {
2184 PerlIO_printf(Perl_debug_log, ", %"UVuf, (UV)fail[q_read]);
2186 PerlIO_printf(Perl_debug_log, "\n");
2189 /*RExC_seen |= REG_SEEN_TRIEDFA;*/
2194 * There are strange code-generation bugs caused on sparc64 by gcc-2.95.2.
2195 * These need to be revisited when a newer toolchain becomes available.
2197 #if defined(__sparc64__) && defined(__GNUC__)
2198 # if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 96)
2199 # undef SPARC64_GCC_WORKAROUND
2200 # define SPARC64_GCC_WORKAROUND 1
2204 #define DEBUG_PEEP(str,scan,depth) \
2205 DEBUG_OPTIMISE_r({if (scan){ \
2206 SV * const mysv=sv_newmortal(); \
2207 regnode *Next = regnext(scan); \
2208 regprop(RExC_rx, mysv, scan); \
2209 PerlIO_printf(Perl_debug_log, "%*s" str ">%3d: %s (%d)\n", \
2210 (int)depth*2, "", REG_NODE_NUM(scan), SvPV_nolen_const(mysv),\
2211 Next ? (REG_NODE_NUM(Next)) : 0 ); \
2218 #define JOIN_EXACT(scan,min,flags) \
2219 if (PL_regkind[OP(scan)] == EXACT) \
2220 join_exact(pRExC_state,(scan),(min),(flags),NULL,depth+1)
2223 S_join_exact(pTHX_ RExC_state_t *pRExC_state, regnode *scan, I32 *min, U32 flags,regnode *val, U32 depth) {
2224 /* Merge several consecutive EXACTish nodes into one. */
2225 regnode *n = regnext(scan);
2227 regnode *next = scan + NODE_SZ_STR(scan);
2231 regnode *stop = scan;
2232 GET_RE_DEBUG_FLAGS_DECL;
2234 PERL_UNUSED_ARG(depth);
2236 #ifndef EXPERIMENTAL_INPLACESCAN
2237 PERL_UNUSED_ARG(flags);
2238 PERL_UNUSED_ARG(val);
2240 DEBUG_PEEP("join",scan,depth);
2242 /* Skip NOTHING, merge EXACT*. */
2244 ( PL_regkind[OP(n)] == NOTHING ||
2245 (stringok && (OP(n) == OP(scan))))
2247 && NEXT_OFF(scan) + NEXT_OFF(n) < I16_MAX) {
2249 if (OP(n) == TAIL || n > next)
2251 if (PL_regkind[OP(n)] == NOTHING) {
2252 DEBUG_PEEP("skip:",n,depth);
2253 NEXT_OFF(scan) += NEXT_OFF(n);
2254 next = n + NODE_STEP_REGNODE;
2261 else if (stringok) {
2262 const unsigned int oldl = STR_LEN(scan);
2263 regnode * const nnext = regnext(n);
2265 DEBUG_PEEP("merg",n,depth);
2268 if (oldl + STR_LEN(n) > U8_MAX)
2270 NEXT_OFF(scan) += NEXT_OFF(n);
2271 STR_LEN(scan) += STR_LEN(n);
2272 next = n + NODE_SZ_STR(n);
2273 /* Now we can overwrite *n : */
2274 Move(STRING(n), STRING(scan) + oldl, STR_LEN(n), char);
2282 #ifdef EXPERIMENTAL_INPLACESCAN
2283 if (flags && !NEXT_OFF(n)) {
2284 DEBUG_PEEP("atch", val, depth);
2285 if (reg_off_by_arg[OP(n)]) {
2286 ARG_SET(n, val - n);
2289 NEXT_OFF(n) = val - n;
2296 if (UTF && ( OP(scan) == EXACTF ) && ( STR_LEN(scan) >= 6 ) ) {
2298 Two problematic code points in Unicode casefolding of EXACT nodes:
2300 U+0390 - GREEK SMALL LETTER IOTA WITH DIALYTIKA AND TONOS
2301 U+03B0 - GREEK SMALL LETTER UPSILON WITH DIALYTIKA AND TONOS
2307 U+03B9 U+0308 U+0301 0xCE 0xB9 0xCC 0x88 0xCC 0x81
2308 U+03C5 U+0308 U+0301 0xCF 0x85 0xCC 0x88 0xCC 0x81
2310 This means that in case-insensitive matching (or "loose matching",
2311 as Unicode calls it), an EXACTF of length six (the UTF-8 encoded byte
2312 length of the above casefolded versions) can match a target string
2313 of length two (the byte length of UTF-8 encoded U+0390 or U+03B0).
2314 This would rather mess up the minimum length computation.
2316 What we'll do is to look for the tail four bytes, and then peek
2317 at the preceding two bytes to see whether we need to decrease
2318 the minimum length by four (six minus two).
2320 Thanks to the design of UTF-8, there cannot be false matches:
2321 A sequence of valid UTF-8 bytes cannot be a subsequence of
2322 another valid sequence of UTF-8 bytes.
2325 char * const s0 = STRING(scan), *s, *t;
2326 char * const s1 = s0 + STR_LEN(scan) - 1;
2327 char * const s2 = s1 - 4;
2328 #ifdef EBCDIC /* RD tunifold greek 0390 and 03B0 */
2329 const char t0[] = "\xaf\x49\xaf\x42";
2331 const char t0[] = "\xcc\x88\xcc\x81";
2333 const char * const t1 = t0 + 3;
2336 s < s2 && (t = ninstr(s, s1, t0, t1));
2339 if (((U8)t[-1] == 0x68 && (U8)t[-2] == 0xB4) ||
2340 ((U8)t[-1] == 0x46 && (U8)t[-2] == 0xB5))
2342 if (((U8)t[-1] == 0xB9 && (U8)t[-2] == 0xCE) ||
2343 ((U8)t[-1] == 0x85 && (U8)t[-2] == 0xCF))
2351 n = scan + NODE_SZ_STR(scan);
2353 if (PL_regkind[OP(n)] != NOTHING || OP(n) == NOTHING) {
2360 DEBUG_OPTIMISE_r(if (merged){DEBUG_PEEP("finl",scan,depth)});
2364 /* REx optimizer. Converts nodes into quickier variants "in place".
2365 Finds fixed substrings. */
2367 /* Stops at toplevel WHILEM as well as at "last". At end *scanp is set
2368 to the position after last scanned or to NULL. */
2370 #define INIT_AND_WITHP \
2371 assert(!and_withp); \
2372 Newx(and_withp,1,struct regnode_charclass_class); \
2373 SAVEFREEPV(and_withp)
2375 /* this is a chain of data about sub patterns we are processing that
2376 need to be handled seperately/specially in study_chunk. Its so
2377 we can simulate recursion without losing state. */
2379 typedef struct scan_frame {
2380 regnode *last; /* last node to process in this frame */
2381 regnode *next; /* next node to process when last is reached */
2382 struct scan_frame *prev; /*previous frame*/
2383 I32 stop; /* what stopparen do we use */
2387 #define SCAN_COMMIT(s, data, m) scan_commit(s, data, m, is_inf)
2390 S_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp,
2391 I32 *minlenp, I32 *deltap,
2396 struct regnode_charclass_class *and_withp,
2397 U32 flags, U32 depth)
2398 /* scanp: Start here (read-write). */
2399 /* deltap: Write maxlen-minlen here. */
2400 /* last: Stop before this one. */
2401 /* data: string data about the pattern */
2402 /* stopparen: treat close N as END */
2403 /* recursed: which subroutines have we recursed into */
2404 /* and_withp: Valid if flags & SCF_DO_STCLASS_OR */
2407 I32 min = 0, pars = 0, code;
2408 regnode *scan = *scanp, *next;
2410 int is_inf = (flags & SCF_DO_SUBSTR) && (data->flags & SF_IS_INF);
2411 int is_inf_internal = 0; /* The studied chunk is infinite */
2412 I32 is_par = OP(scan) == OPEN ? ARG(scan) : 0;
2413 scan_data_t data_fake;
2414 SV *re_trie_maxbuff = NULL;
2415 regnode *first_non_open = scan;
2416 I32 stopmin = I32_MAX;
2417 scan_frame *frame = NULL;
2419 GET_RE_DEBUG_FLAGS_DECL;
2422 StructCopy(&zero_scan_data, &data_fake, scan_data_t);
2426 while (first_non_open && OP(first_non_open) == OPEN)
2427 first_non_open=regnext(first_non_open);
2432 while ( scan && OP(scan) != END && scan < last ){
2433 /* Peephole optimizer: */
2434 DEBUG_STUDYDATA("Peep:", data,depth);
2435 DEBUG_PEEP("Peep",scan,depth);
2436 JOIN_EXACT(scan,&min,0);
2438 /* Follow the next-chain of the current node and optimize
2439 away all the NOTHINGs from it. */
2440 if (OP(scan) != CURLYX) {
2441 const int max = (reg_off_by_arg[OP(scan)]
2443 /* I32 may be smaller than U16 on CRAYs! */
2444 : (I32_MAX < U16_MAX ? I32_MAX : U16_MAX));
2445 int off = (reg_off_by_arg[OP(scan)] ? ARG(scan) : NEXT_OFF(scan));
2449 /* Skip NOTHING and LONGJMP. */
2450 while ((n = regnext(n))
2451 && ((PL_regkind[OP(n)] == NOTHING && (noff = NEXT_OFF(n)))
2452 || ((OP(n) == LONGJMP) && (noff = ARG(n))))
2453 && off + noff < max)
2455 if (reg_off_by_arg[OP(scan)])
2458 NEXT_OFF(scan) = off;
2463 /* The principal pseudo-switch. Cannot be a switch, since we
2464 look into several different things. */
2465 if (OP(scan) == BRANCH || OP(scan) == BRANCHJ
2466 || OP(scan) == IFTHEN) {
2467 next = regnext(scan);
2469 /* demq: the op(next)==code check is to see if we have "branch-branch" AFAICT */
2471 if (OP(next) == code || code == IFTHEN) {
2472 /* NOTE - There is similar code to this block below for handling
2473 TRIE nodes on a re-study. If you change stuff here check there
2475 I32 max1 = 0, min1 = I32_MAX, num = 0;
2476 struct regnode_charclass_class accum;
2477 regnode * const startbranch=scan;
2479 if (flags & SCF_DO_SUBSTR)
2480 SCAN_COMMIT(pRExC_state, data, minlenp); /* Cannot merge strings after this. */
2481 if (flags & SCF_DO_STCLASS)
2482 cl_init_zero(pRExC_state, &accum);
2484 while (OP(scan) == code) {
2485 I32 deltanext, minnext, f = 0, fake;
2486 struct regnode_charclass_class this_class;
2489 data_fake.flags = 0;
2491 data_fake.whilem_c = data->whilem_c;
2492 data_fake.last_closep = data->last_closep;
2495 data_fake.last_closep = &fake;
2497 data_fake.pos_delta = delta;
2498 next = regnext(scan);
2499 scan = NEXTOPER(scan);
2501 scan = NEXTOPER(scan);
2502 if (flags & SCF_DO_STCLASS) {
2503 cl_init(pRExC_state, &this_class);
2504 data_fake.start_class = &this_class;
2505 f = SCF_DO_STCLASS_AND;
2507 if (flags & SCF_WHILEM_VISITED_POS)
2508 f |= SCF_WHILEM_VISITED_POS;
2510 /* we suppose the run is continuous, last=next...*/
2511 minnext = study_chunk(pRExC_state, &scan, minlenp, &deltanext,
2513 stopparen, recursed, NULL, f,depth+1);
2516 if (max1 < minnext + deltanext)
2517 max1 = minnext + deltanext;
2518 if (deltanext == I32_MAX)
2519 is_inf = is_inf_internal = 1;
2521 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
2523 if (data_fake.flags & SCF_SEEN_ACCEPT) {
2524 if ( stopmin > minnext)
2525 stopmin = min + min1;
2526 flags &= ~SCF_DO_SUBSTR;
2528 data->flags |= SCF_SEEN_ACCEPT;
2531 if (data_fake.flags & SF_HAS_EVAL)
2532 data->flags |= SF_HAS_EVAL;
2533 data->whilem_c = data_fake.whilem_c;
2535 if (flags & SCF_DO_STCLASS)
2536 cl_or(pRExC_state, &accum, &this_class);
2538 if (code == IFTHEN && num < 2) /* Empty ELSE branch */
2540 if (flags & SCF_DO_SUBSTR) {
2541 data->pos_min += min1;
2542 data->pos_delta += max1 - min1;
2543 if (max1 != min1 || is_inf)
2544 data->longest = &(data->longest_float);
2547 delta += max1 - min1;
2548 if (flags & SCF_DO_STCLASS_OR) {
2549 cl_or(pRExC_state, data->start_class, &accum);
2551 cl_and(data->start_class, and_withp);
2552 flags &= ~SCF_DO_STCLASS;
2555 else if (flags & SCF_DO_STCLASS_AND) {
2557 cl_and(data->start_class, &accum);
2558 flags &= ~SCF_DO_STCLASS;
2561 /* Switch to OR mode: cache the old value of
2562 * data->start_class */
2564 StructCopy(data->start_class, and_withp,
2565 struct regnode_charclass_class);
2566 flags &= ~SCF_DO_STCLASS_AND;
2567 StructCopy(&accum, data->start_class,
2568 struct regnode_charclass_class);
2569 flags |= SCF_DO_STCLASS_OR;
2570 data->start_class->flags |= ANYOF_EOS;
2574 if (PERL_ENABLE_TRIE_OPTIMISATION && OP( startbranch ) == BRANCH ) {
2577 Assuming this was/is a branch we are dealing with: 'scan' now
2578 points at the item that follows the branch sequence, whatever
2579 it is. We now start at the beginning of the sequence and look
2586 which would be constructed from a pattern like /A|LIST|OF|WORDS/
2588 If we can find such a subseqence we need to turn the first
2589 element into a trie and then add the subsequent branch exact
2590 strings to the trie.
2594 1. patterns where the whole set of branch can be converted.
2596 2. patterns where only a subset can be converted.
2598 In case 1 we can replace the whole set with a single regop
2599 for the trie. In case 2 we need to keep the start and end
2602 'BRANCH EXACT; BRANCH EXACT; BRANCH X'
2603 becomes BRANCH TRIE; BRANCH X;
2605 There is an additional case, that being where there is a
2606 common prefix, which gets split out into an EXACT like node
2607 preceding the TRIE node.
2609 If x(1..n)==tail then we can do a simple trie, if not we make
2610 a "jump" trie, such that when we match the appropriate word
2611 we "jump" to the appopriate tail node. Essentailly we turn
2612 a nested if into a case structure of sorts.
2617 if (!re_trie_maxbuff) {
2618 re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, 1);
2619 if (!SvIOK(re_trie_maxbuff))
2620 sv_setiv(re_trie_maxbuff, RE_TRIE_MAXBUF_INIT);
2622 if ( SvIV(re_trie_maxbuff)>=0 ) {
2624 regnode *first = (regnode *)NULL;
2625 regnode *last = (regnode *)NULL;
2626 regnode *tail = scan;
2631 SV * const mysv = sv_newmortal(); /* for dumping */
2633 /* var tail is used because there may be a TAIL
2634 regop in the way. Ie, the exacts will point to the
2635 thing following the TAIL, but the last branch will
2636 point at the TAIL. So we advance tail. If we
2637 have nested (?:) we may have to move through several
2641 while ( OP( tail ) == TAIL ) {
2642 /* this is the TAIL generated by (?:) */
2643 tail = regnext( tail );
2648 regprop(RExC_rx, mysv, tail );
2649 PerlIO_printf( Perl_debug_log, "%*s%s%s\n",
2650 (int)depth * 2 + 2, "",
2651 "Looking for TRIE'able sequences. Tail node is: ",
2652 SvPV_nolen_const( mysv )
2658 step through the branches, cur represents each
2659 branch, noper is the first thing to be matched
2660 as part of that branch and noper_next is the
2661 regnext() of that node. if noper is an EXACT
2662 and noper_next is the same as scan (our current
2663 position in the regex) then the EXACT branch is
2664 a possible optimization target. Once we have
2665 two or more consequetive such branches we can
2666 create a trie of the EXACT's contents and stich
2667 it in place. If the sequence represents all of
2668 the branches we eliminate the whole thing and
2669 replace it with a single TRIE. If it is a
2670 subsequence then we need to stitch it in. This
2671 means the first branch has to remain, and needs
2672 to be repointed at the item on the branch chain
2673 following the last branch optimized. This could
2674 be either a BRANCH, in which case the
2675 subsequence is internal, or it could be the
2676 item following the branch sequence in which
2677 case the subsequence is at the end.
2681 /* dont use tail as the end marker for this traverse */
2682 for ( cur = startbranch ; cur != scan ; cur = regnext( cur ) ) {
2683 regnode * const noper = NEXTOPER( cur );
2684 #if defined(DEBUGGING) || defined(NOJUMPTRIE)
2685 regnode * const noper_next = regnext( noper );
2689 regprop(RExC_rx, mysv, cur);
2690 PerlIO_printf( Perl_debug_log, "%*s- %s (%d)",
2691 (int)depth * 2 + 2,"", SvPV_nolen_const( mysv ), REG_NODE_NUM(cur) );
2693 regprop(RExC_rx, mysv, noper);
2694 PerlIO_printf( Perl_debug_log, " -> %s",
2695 SvPV_nolen_const(mysv));
2698 regprop(RExC_rx, mysv, noper_next );
2699 PerlIO_printf( Perl_debug_log,"\t=> %s\t",
2700 SvPV_nolen_const(mysv));
2702 PerlIO_printf( Perl_debug_log, "(First==%d,Last==%d,Cur==%d)\n",
2703 REG_NODE_NUM(first), REG_NODE_NUM(last), REG_NODE_NUM(cur) );
2705 if ( (((first && optype!=NOTHING) ? OP( noper ) == optype
2706 : PL_regkind[ OP( noper ) ] == EXACT )
2707 || OP(noper) == NOTHING )
2709 && noper_next == tail
2714 if ( !first || optype == NOTHING ) {
2715 if (!first) first = cur;
2716 optype = OP( noper );
2722 make_trie( pRExC_state,
2723 startbranch, first, cur, tail, count,
2726 if ( PL_regkind[ OP( noper ) ] == EXACT
2728 && noper_next == tail
2733 optype = OP( noper );
2743 regprop(RExC_rx, mysv, cur);
2744 PerlIO_printf( Perl_debug_log,
2745 "%*s- %s (%d) <SCAN FINISHED>\n", (int)depth * 2 + 2,
2746 "", SvPV_nolen_const( mysv ),REG_NODE_NUM(cur));
2750 made= make_trie( pRExC_state, startbranch, first, scan, tail, count, optype, depth+1 );
2751 #ifdef TRIE_STUDY_OPT
2752 if ( ((made == MADE_EXACT_TRIE &&
2753 startbranch == first)
2754 || ( first_non_open == first )) &&
2756 flags |= SCF_TRIE_RESTUDY;
2757 if ( startbranch == first
2760 RExC_seen &=~REG_TOP_LEVEL_BRANCHES;
2770 else if ( code == BRANCHJ ) { /* single branch is optimized. */
2771 scan = NEXTOPER(NEXTOPER(scan));
2772 } else /* single branch is optimized. */
2773 scan = NEXTOPER(scan);
2775 } else if (OP(scan) == SUSPEND || OP(scan) == GOSUB || OP(scan) == GOSTART) {
2776 scan_frame *newframe = NULL;
2781 if (OP(scan) != SUSPEND) {
2782 /* set the pointer */
2783 if (OP(scan) == GOSUB) {
2785 RExC_recurse[ARG2L(scan)] = scan;
2786 start = RExC_open_parens[paren-1];
2787 end = RExC_close_parens[paren-1];
2790 start = RExC_rxi->program + 1;
2794 Newxz(recursed, (((RExC_npar)>>3) +1), U8);
2795 SAVEFREEPV(recursed);
2797 if (!PAREN_TEST(recursed,paren+1)) {
2798 PAREN_SET(recursed,paren+1);
2799 Newx(newframe,1,scan_frame);
2801 if (flags & SCF_DO_SUBSTR) {
2802 SCAN_COMMIT(pRExC_state,data,minlenp);
2803 data->longest = &(data->longest_float);
2805 is_inf = is_inf_internal = 1;
2806 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
2807 cl_anything(pRExC_state, data->start_class);
2808 flags &= ~SCF_DO_STCLASS;
2811 Newx(newframe,1,scan_frame);
2814 end = regnext(scan);
2819 SAVEFREEPV(newframe);
2820 newframe->next = regnext(scan);
2821 newframe->last = last;
2822 newframe->stop = stopparen;
2823 newframe->prev = frame;
2833 else if (OP(scan) == EXACT) {
2834 I32 l = STR_LEN(scan);
2837 const U8 * const s = (U8*)STRING(scan);
2838 l = utf8_length(s, s + l);
2839 uc = utf8_to_uvchr(s, NULL);
2841 uc = *((U8*)STRING(scan));
2844 if (flags & SCF_DO_SUBSTR) { /* Update longest substr. */
2845 /* The code below prefers earlier match for fixed
2846 offset, later match for variable offset. */
2847 if (data->last_end == -1) { /* Update the start info. */
2848 data->last_start_min = data->pos_min;
2849 data->last_start_max = is_inf
2850 ? I32_MAX : data->pos_min + data->pos_delta;
2852 sv_catpvn(data->last_found, STRING(scan), STR_LEN(scan));
2854 SvUTF8_on(data->last_found);
2856 SV * const sv = data->last_found;
2857 MAGIC * const mg = SvUTF8(sv) && SvMAGICAL(sv) ?
2858 mg_find(sv, PERL_MAGIC_utf8) : NULL;
2859 if (mg && mg->mg_len >= 0)
2860 mg->mg_len += utf8_length((U8*)STRING(scan),
2861 (U8*)STRING(scan)+STR_LEN(scan));
2863 data->last_end = data->pos_min + l;
2864 data->pos_min += l; /* As in the first entry. */
2865 data->flags &= ~SF_BEFORE_EOL;
2867 if (flags & SCF_DO_STCLASS_AND) {
2868 /* Check whether it is compatible with what we know already! */
2872 (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
2873 && !ANYOF_BITMAP_TEST(data->start_class, uc)
2874 && (!(data->start_class->flags & ANYOF_FOLD)
2875 || !ANYOF_BITMAP_TEST(data->start_class, PL_fold[uc])))
2878 ANYOF_CLASS_ZERO(data->start_class);
2879 ANYOF_BITMAP_ZERO(data->start_class);
2881 ANYOF_BITMAP_SET(data->start_class, uc);
2882 data->start_class->flags &= ~ANYOF_EOS;
2884 data->start_class->flags &= ~ANYOF_UNICODE_ALL;
2886 else if (flags & SCF_DO_STCLASS_OR) {
2887 /* false positive possible if the class is case-folded */
2889 ANYOF_BITMAP_SET(data->start_class, uc);
2891 data->start_class->flags |= ANYOF_UNICODE_ALL;
2892 data->start_class->flags &= ~ANYOF_EOS;
2893 cl_and(data->start_class, and_withp);
2895 flags &= ~SCF_DO_STCLASS;
2897 else if (PL_regkind[OP(scan)] == EXACT) { /* But OP != EXACT! */
2898 I32 l = STR_LEN(scan);
2899 UV uc = *((U8*)STRING(scan));
2901 /* Search for fixed substrings supports EXACT only. */
2902 if (flags & SCF_DO_SUBSTR) {
2904 SCAN_COMMIT(pRExC_state, data, minlenp);
2907 const U8 * const s = (U8 *)STRING(scan);
2908 l = utf8_length(s, s + l);
2909 uc = utf8_to_uvchr(s, NULL);
2912 if (flags & SCF_DO_SUBSTR)
2914 if (flags & SCF_DO_STCLASS_AND) {
2915 /* Check whether it is compatible with what we know already! */
2919 (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
2920 && !ANYOF_BITMAP_TEST(data->start_class, uc)
2921 && !ANYOF_BITMAP_TEST(data->start_class, PL_fold[uc])))
2923 ANYOF_CLASS_ZERO(data->start_class);
2924 ANYOF_BITMAP_ZERO(data->start_class);
2926 ANYOF_BITMAP_SET(data->start_class, uc);
2927 data->start_class->flags &= ~ANYOF_EOS;
2928 data->start_class->flags |= ANYOF_FOLD;
2929 if (OP(scan) == EXACTFL)
2930 data->start_class->flags |= ANYOF_LOCALE;
2933 else if (flags & SCF_DO_STCLASS_OR) {
2934 if (data->start_class->flags & ANYOF_FOLD) {
2935 /* false positive possible if the class is case-folded.
2936 Assume that the locale settings are the same... */
2938 ANYOF_BITMAP_SET(data->start_class, uc);
2939 data->start_class->flags &= ~ANYOF_EOS;
2941 cl_and(data->start_class, and_withp);
2943 flags &= ~SCF_DO_STCLASS;
2945 else if (strchr((const char*)PL_varies,OP(scan))) {
2946 I32 mincount, maxcount, minnext, deltanext, fl = 0;
2947 I32 f = flags, pos_before = 0;
2948 regnode * const oscan = scan;
2949 struct regnode_charclass_class this_class;
2950 struct regnode_charclass_class *oclass = NULL;
2951 I32 next_is_eval = 0;
2953 switch (PL_regkind[OP(scan)]) {
2954 case WHILEM: /* End of (?:...)* . */
2955 scan = NEXTOPER(scan);
2958 if (flags & (SCF_DO_SUBSTR | SCF_DO_STCLASS)) {
2959 next = NEXTOPER(scan);
2960 if (OP(next) == EXACT || (flags & SCF_DO_STCLASS)) {
2962 maxcount = REG_INFTY;
2963 next = regnext(scan);
2964 scan = NEXTOPER(scan);
2968 if (flags & SCF_DO_SUBSTR)
2973 if (flags & SCF_DO_STCLASS) {
2975 maxcount = REG_INFTY;
2976 next = regnext(scan);
2977 scan = NEXTOPER(scan);
2980 is_inf = is_inf_internal = 1;
2981 scan = regnext(scan);
2982 if (flags & SCF_DO_SUBSTR) {
2983 SCAN_COMMIT(pRExC_state, data, minlenp); /* Cannot extend fixed substrings */
2984 data->longest = &(data->longest_float);
2986 goto optimize_curly_tail;
2988 if (stopparen>0 && (OP(scan)==CURLYN || OP(scan)==CURLYM)
2989 && (scan->flags == stopparen))
2994 mincount = ARG1(scan);
2995 maxcount = ARG2(scan);
2997 next = regnext(scan);
2998 if (OP(scan) == CURLYX) {
2999 I32 lp = (data ? *(data->last_closep) : 0);
3000 scan->flags = ((lp <= (I32)U8_MAX) ? (U8)lp : U8_MAX);
3002 scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3003 next_is_eval = (OP(scan) == EVAL);
3005 if (flags & SCF_DO_SUBSTR) {
3006 if (mincount == 0) SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot extend fixed substrings */
3007 pos_before = data->pos_min;
3011 data->flags &= ~(SF_HAS_PAR|SF_IN_PAR|SF_HAS_EVAL);
3013 data->flags |= SF_IS_INF;
3015 if (flags & SCF_DO_STCLASS) {
3016 cl_init(pRExC_state, &this_class);
3017 oclass = data->start_class;
3018 data->start_class = &this_class;
3019 f |= SCF_DO_STCLASS_AND;
3020 f &= ~SCF_DO_STCLASS_OR;
3022 /* These are the cases when once a subexpression
3023 fails at a particular position, it cannot succeed
3024 even after backtracking at the enclosing scope.
3026 XXXX what if minimal match and we are at the
3027 initial run of {n,m}? */
3028 if ((mincount != maxcount - 1) && (maxcount != REG_INFTY))
3029 f &= ~SCF_WHILEM_VISITED_POS;
3031 /* This will finish on WHILEM, setting scan, or on NULL: */
3032 minnext = study_chunk(pRExC_state, &scan, minlenp, &deltanext,
3033 last, data, stopparen, recursed, NULL,
3035 ? (f & ~SCF_DO_SUBSTR) : f),depth+1);
3037 if (flags & SCF_DO_STCLASS)
3038 data->start_class = oclass;
3039 if (mincount == 0 || minnext == 0) {
3040 if (flags & SCF_DO_STCLASS_OR) {
3041 cl_or(pRExC_state, data->start_class, &this_class);
3043 else if (flags & SCF_DO_STCLASS_AND) {
3044 /* Switch to OR mode: cache the old value of
3045 * data->start_class */
3047 StructCopy(data->start_class, and_withp,
3048 struct regnode_charclass_class);
3049 flags &= ~SCF_DO_STCLASS_AND;
3050 StructCopy(&this_class, data->start_class,
3051 struct regnode_charclass_class);
3052 flags |= SCF_DO_STCLASS_OR;
3053 data->start_class->flags |= ANYOF_EOS;
3055 } else { /* Non-zero len */
3056 if (flags & SCF_DO_STCLASS_OR) {
3057 cl_or(pRExC_state, data->start_class, &this_class);
3058 cl_and(data->start_class, and_withp);
3060 else if (flags & SCF_DO_STCLASS_AND)
3061 cl_and(data->start_class, &this_class);
3062 flags &= ~SCF_DO_STCLASS;
3064 if (!scan) /* It was not CURLYX, but CURLY. */
3066 if ( /* ? quantifier ok, except for (?{ ... }) */
3067 (next_is_eval || !(mincount == 0 && maxcount == 1))
3068 && (minnext == 0) && (deltanext == 0)
3069 && data && !(data->flags & (SF_HAS_PAR|SF_IN_PAR))
3070 && maxcount <= REG_INFTY/3 /* Complement check for big count */
3071 && ckWARN(WARN_REGEXP))
3074 "Quantifier unexpected on zero-length expression");
3077 min += minnext * mincount;
3078 is_inf_internal |= ((maxcount == REG_INFTY
3079 && (minnext + deltanext) > 0)
3080 || deltanext == I32_MAX);
3081 is_inf |= is_inf_internal;
3082 delta += (minnext + deltanext) * maxcount - minnext * mincount;
3084 /* Try powerful optimization CURLYX => CURLYN. */
3085 if ( OP(oscan) == CURLYX && data
3086 && data->flags & SF_IN_PAR
3087 && !(data->flags & SF_HAS_EVAL)
3088 && !deltanext && minnext == 1 ) {
3089 /* Try to optimize to CURLYN. */
3090 regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS;
3091 regnode * const nxt1 = nxt;
3098 if (!strchr((const char*)PL_simple,OP(nxt))
3099 && !(PL_regkind[OP(nxt)] == EXACT
3100 && STR_LEN(nxt) == 1))
3106 if (OP(nxt) != CLOSE)
3108 if (RExC_open_parens) {
3109 RExC_open_parens[ARG(nxt1)-1]=oscan; /*open->CURLYM*/
3110 RExC_close_parens[ARG(nxt1)-1]=nxt+2; /*close->while*/
3112 /* Now we know that nxt2 is the only contents: */
3113 oscan->flags = (U8)ARG(nxt);
3115 OP(nxt1) = NOTHING; /* was OPEN. */
3118 OP(nxt1 + 1) = OPTIMIZED; /* was count. */
3119 NEXT_OFF(nxt1+ 1) = 0; /* just for consistancy. */
3120 NEXT_OFF(nxt2) = 0; /* just for consistancy with CURLY. */
3121 OP(nxt) = OPTIMIZED; /* was CLOSE. */
3122 OP(nxt + 1) = OPTIMIZED; /* was count. */
3123 NEXT_OFF(nxt+ 1) = 0; /* just for consistancy. */
3128 /* Try optimization CURLYX => CURLYM. */
3129 if ( OP(oscan) == CURLYX && data
3130 && !(data->flags & SF_HAS_PAR)
3131 && !(data->flags & SF_HAS_EVAL)
3132 && !deltanext /* atom is fixed width */
3133 && minnext != 0 /* CURLYM can't handle zero width */
3135 /* XXXX How to optimize if data == 0? */
3136 /* Optimize to a simpler form. */
3137 regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN */
3141 while ( (nxt2 = regnext(nxt)) /* skip over embedded stuff*/
3142 && (OP(nxt2) != WHILEM))
3144 OP(nxt2) = SUCCEED; /* Whas WHILEM */
3145 /* Need to optimize away parenths. */
3146 if (data->flags & SF_IN_PAR) {
3147 /* Set the parenth number. */
3148 regnode *nxt1 = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN*/
3150 if (OP(nxt) != CLOSE)
3151 FAIL("Panic opt close");
3152 oscan->flags = (U8)ARG(nxt);
3153 if (RExC_open_parens) {
3154 RExC_open_parens[ARG(nxt1)-1]=oscan; /*open->CURLYM*/
3155 RExC_close_parens[ARG(nxt1)-1]=nxt2+1; /*close->NOTHING*/
3157 OP(nxt1) = OPTIMIZED; /* was OPEN. */
3158 OP(nxt) = OPTIMIZED; /* was CLOSE. */
3161 OP(nxt1 + 1) = OPTIMIZED; /* was count. */
3162 OP(nxt + 1) = OPTIMIZED; /* was count. */
3163 NEXT_OFF(nxt1 + 1) = 0; /* just for consistancy. */
3164 NEXT_OFF(nxt + 1) = 0; /* just for consistancy. */
3167 while ( nxt1 && (OP(nxt1) != WHILEM)) {
3168 regnode *nnxt = regnext(nxt1);
3171 if (reg_off_by_arg[OP(nxt1)])
3172 ARG_SET(nxt1, nxt2 - nxt1);
3173 else if (nxt2 - nxt1 < U16_MAX)
3174 NEXT_OFF(nxt1) = nxt2 - nxt1;
3176 OP(nxt) = NOTHING; /* Cannot beautify */
3181 /* Optimize again: */
3182 study_chunk(pRExC_state, &nxt1, minlenp, &deltanext, nxt,
3183 NULL, stopparen, recursed, NULL, 0,depth+1);
3188 else if ((OP(oscan) == CURLYX)
3189 && (flags & SCF_WHILEM_VISITED_POS)
3190 /* See the comment on a similar expression above.
3191 However, this time it not a subexpression
3192 we care about, but the expression itself. */
3193 && (maxcount == REG_INFTY)
3194 && data && ++data->whilem_c < 16) {
3195 /* This stays as CURLYX, we can put the count/of pair. */
3196 /* Find WHILEM (as in regexec.c) */
3197 regnode *nxt = oscan + NEXT_OFF(oscan);
3199 if (OP(PREVOPER(nxt)) == NOTHING) /* LONGJMP */
3201 PREVOPER(nxt)->flags = (U8)(data->whilem_c
3202 | (RExC_whilem_seen << 4)); /* On WHILEM */
3204 if (data && fl & (SF_HAS_PAR|SF_IN_PAR))
3206 if (flags & SCF_DO_SUBSTR) {
3207 SV *last_str = NULL;
3208 int counted = mincount != 0;
3210 if (data->last_end > 0 && mincount != 0) { /* Ends with a string. */
3211 #if defined(SPARC64_GCC_WORKAROUND)
3214 const char *s = NULL;
3217 if (pos_before >= data->last_start_min)
3220 b = data->last_start_min;
3223 s = SvPV_const(data->last_found, l);
3224 old = b - data->last_start_min;
3227 I32 b = pos_before >= data->last_start_min
3228 ? pos_before : data->last_start_min;
3230 const char * const s = SvPV_const(data->last_found, l);
3231 I32 old = b - data->last_start_min;
3235 old = utf8_hop((U8*)s, old) - (U8*)s;
3238 /* Get the added string: */
3239 last_str = newSVpvn(s + old, l);
3241 SvUTF8_on(last_str);
3242 if (deltanext == 0 && pos_before == b) {
3243 /* What was added is a constant string */
3245 SvGROW(last_str, (mincount * l) + 1);
3246 repeatcpy(SvPVX(last_str) + l,
3247 SvPVX_const(last_str), l, mincount - 1);
3248 SvCUR_set(last_str, SvCUR(last_str) * mincount);
3249 /* Add additional parts. */
3250 SvCUR_set(data->last_found,
3251 SvCUR(data->last_found) - l);
3252 sv_catsv(data->last_found, last_str);
3254 SV * sv = data->last_found;
3256 SvUTF8(sv) && SvMAGICAL(sv) ?
3257 mg_find(sv, PERL_MAGIC_utf8) : NULL;
3258 if (mg && mg->mg_len >= 0)
3259 mg->mg_len += CHR_SVLEN(last_str);
3261 data->last_end += l * (mincount - 1);
3264 /* start offset must point into the last copy */
3265 data->last_start_min += minnext * (mincount - 1);
3266 data->last_start_max += is_inf ? I32_MAX
3267 : (maxcount - 1) * (minnext + data->pos_delta);
3270 /* It is counted once already... */
3271 data->pos_min += minnext * (mincount - counted);
3272 data->pos_delta += - counted * deltanext +
3273 (minnext + deltanext) * maxcount - minnext * mincount;
3274 if (mincount != maxcount) {
3275 /* Cannot extend fixed substrings found inside
3277 SCAN_COMMIT(pRExC_state,data,minlenp);
3278 if (mincount && last_str) {
3279 SV * const sv = data->last_found;
3280 MAGIC * const mg = SvUTF8(sv) && SvMAGICAL(sv) ?
3281 mg_find(sv, PERL_MAGIC_utf8) : NULL;
3285 sv_setsv(sv, last_str);
3286 data->last_end = data->pos_min;
3287 data->last_start_min =
3288 data->pos_min - CHR_SVLEN(last_str);
3289 data->last_start_max = is_inf
3291 : data->pos_min + data->pos_delta
3292 - CHR_SVLEN(last_str);
3294 data->longest = &(data->longest_float);
3296 SvREFCNT_dec(last_str);
3298 if (data && (fl & SF_HAS_EVAL))
3299 data->flags |= SF_HAS_EVAL;
3300 optimize_curly_tail:
3301 if (OP(oscan) != CURLYX) {
3302 while (PL_regkind[OP(next = regnext(oscan))] == NOTHING
3304 NEXT_OFF(oscan) += NEXT_OFF(next);
3307 default: /* REF and CLUMP only? */
3308 if (flags & SCF_DO_SUBSTR) {
3309 SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot expect anything... */
3310 data->longest = &(data->longest_float);
3312 is_inf = is_inf_internal = 1;
3313 if (flags & SCF_DO_STCLASS_OR)
3314 cl_anything(pRExC_state, data->start_class);
3315 flags &= ~SCF_DO_STCLASS;
3319 else if (strchr((const char*)PL_simple,OP(scan))) {
3322 if (flags & SCF_DO_SUBSTR) {
3323 SCAN_COMMIT(pRExC_state,data,minlenp);
3327 if (flags & SCF_DO_STCLASS) {
3328 data->start_class->flags &= ~ANYOF_EOS; /* No match on empty */
3330 /* Some of the logic below assumes that switching
3331 locale on will only add false positives. */
3332 switch (PL_regkind[OP(scan)]) {
3336 /* Perl_croak(aTHX_ "panic: unexpected simple REx opcode %d", OP(scan)); */
3337 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
3338 cl_anything(pRExC_state, data->start_class);
3341 if (OP(scan) == SANY)
3343 if (flags & SCF_DO_STCLASS_OR) { /* Everything but \n */
3344 value = (ANYOF_BITMAP_TEST(data->start_class,'\n')
3345 || (data->start_class->flags & ANYOF_CLASS));
3346 cl_anything(pRExC_state, data->start_class);
3348 if (flags & SCF_DO_STCLASS_AND || !value)
3349 ANYOF_BITMAP_CLEAR(data->start_class,'\n');
3352 if (flags & SCF_DO_STCLASS_AND)
3353 cl_and(data->start_class,
3354 (struct regnode_charclass_class*)scan);
3356 cl_or(pRExC_state, data->start_class,
3357 (struct regnode_charclass_class*)scan);
3360 if (flags & SCF_DO_STCLASS_AND) {
3361 if (!(data->start_class->flags & ANYOF_LOCALE)) {
3362 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM);
3363 for (value = 0; value < 256; value++)
3364 if (!isALNUM(value))
3365 ANYOF_BITMAP_CLEAR(data->start_class, value);
3369 if (data->start_class->flags & ANYOF_LOCALE)
3370 ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM);
3372 for (value = 0; value < 256; value++)
3374 ANYOF_BITMAP_SET(data->start_class, value);
3379 if (flags & SCF_DO_STCLASS_AND) {
3380 if (data->start_class->flags & ANYOF_LOCALE)
3381 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM);
3384 ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM);
3385 data->start_class->flags |= ANYOF_LOCALE;
3389 if (flags & SCF_DO_STCLASS_AND) {
3390 if (!(data->start_class->flags & ANYOF_LOCALE)) {
3391 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM);
3392 for (value = 0; value < 256; value++)
3394 ANYOF_BITMAP_CLEAR(data->start_class, value);
3398 if (data->start_class->flags & ANYOF_LOCALE)
3399 ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM);
3401 for (value = 0; value < 256; value++)
3402 if (!isALNUM(value))
3403 ANYOF_BITMAP_SET(data->start_class, value);
3408 if (flags & SCF_DO_STCLASS_AND) {
3409 if (data->start_class->flags & ANYOF_LOCALE)
3410 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM);
3413 data->start_class->flags |= ANYOF_LOCALE;
3414 ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM);
3418 if (flags & SCF_DO_STCLASS_AND) {
3419 if (!(data->start_class->flags & ANYOF_LOCALE)) {
3420 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE);
3421 for (value = 0; value < 256; value++)
3422 if (!isSPACE(value))
3423 ANYOF_BITMAP_CLEAR(data->start_class, value);
3427 if (data->start_class->flags & ANYOF_LOCALE)
3428 ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE);
3430 for (value = 0; value < 256; value++)
3432 ANYOF_BITMAP_SET(data->start_class, value);
3437 if (flags & SCF_DO_STCLASS_AND) {
3438 if (data->start_class->flags & ANYOF_LOCALE)
3439 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE);
3442 data->start_class->flags |= ANYOF_LOCALE;
3443 ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE);
3447 if (flags & SCF_DO_STCLASS_AND) {
3448 if (!(data->start_class->flags & ANYOF_LOCALE)) {
3449 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE);
3450 for (value = 0; value < 256; value++)
3452 ANYOF_BITMAP_CLEAR(data->start_class, value);
3456 if (data->start_class->flags & ANYOF_LOCALE)
3457 ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE);
3459 for (value = 0; value < 256; value++)
3460 if (!isSPACE(value))
3461 ANYOF_BITMAP_SET(data->start_class, value);
3466 if (flags & SCF_DO_STCLASS_AND) {
3467 if (data->start_class->flags & ANYOF_LOCALE) {
3468 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE);
3469 for (value = 0; value < 256; value++)
3470 if (!isSPACE(value))
3471 ANYOF_BITMAP_CLEAR(data->start_class, value);
3475 data->start_class->flags |= ANYOF_LOCALE;
3476 ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE);
3480 if (flags & SCF_DO_STCLASS_AND) {
3481 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NDIGIT);
3482 for (value = 0; value < 256; value++)
3483 if (!isDIGIT(value))
3484 ANYOF_BITMAP_CLEAR(data->start_class, value);
3487 if (data->start_class->flags & ANYOF_LOCALE)
3488 ANYOF_CLASS_SET(data->start_class,ANYOF_DIGIT);
3490 for (value = 0; value < 256; value++)
3492 ANYOF_BITMAP_SET(data->start_class, value);
3497 if (flags & SCF_DO_STCLASS_AND) {
3498 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_DIGIT);
3499 for (value = 0; value < 256; value++)
3501 ANYOF_BITMAP_CLEAR(data->start_class, value);
3504 if (data->start_class->flags & ANYOF_LOCALE)
3505 ANYOF_CLASS_SET(data->start_class,ANYOF_NDIGIT);
3507 for (value = 0; value < 256; value++)
3508 if (!isDIGIT(value))
3509 ANYOF_BITMAP_SET(data->start_class, value);
3514 if (flags & SCF_DO_STCLASS_OR)
3515 cl_and(data->start_class, and_withp);
3516 flags &= ~SCF_DO_STCLASS;
3519 else if (PL_regkind[OP(scan)] == EOL && flags & SCF_DO_SUBSTR) {
3520 data->flags |= (OP(scan) == MEOL
3524 else if ( PL_regkind[OP(scan)] == BRANCHJ
3525 /* Lookbehind, or need to calculate parens/evals/stclass: */
3526 && (scan->flags || data || (flags & SCF_DO_STCLASS))
3527 && (OP(scan) == IFMATCH || OP(scan) == UNLESSM)) {
3528 if ( !PERL_ENABLE_POSITIVE_ASSERTION_STUDY
3529 || OP(scan) == UNLESSM )
3531 /* Negative Lookahead/lookbehind
3532 In this case we can't do fixed string optimisation.
3535 I32 deltanext, minnext, fake = 0;
3537 struct regnode_charclass_class intrnl;
3540 data_fake.flags = 0;
3542 data_fake.whilem_c = data->whilem_c;
3543 data_fake.last_closep = data->last_closep;
3546 data_fake.last_closep = &fake;
3547 data_fake.pos_delta = delta;
3548 if ( flags & SCF_DO_STCLASS && !scan->flags
3549 && OP(scan) == IFMATCH ) { /* Lookahead */
3550 cl_init(pRExC_state, &intrnl);
3551 data_fake.start_class = &intrnl;
3552 f |= SCF_DO_STCLASS_AND;
3554 if (flags & SCF_WHILEM_VISITED_POS)
3555 f |= SCF_WHILEM_VISITED_POS;
3556 next = regnext(scan);
3557 nscan = NEXTOPER(NEXTOPER(scan));
3558 minnext = study_chunk(pRExC_state, &nscan, minlenp, &deltanext,
3559 last, &data_fake, stopparen, recursed, NULL, f, depth+1);
3562 FAIL("Variable length lookbehind not implemented");
3564 else if (minnext > (I32)U8_MAX) {
3565 FAIL2("Lookbehind longer than %"UVuf" not implemented", (UV)U8_MAX);
3567 scan->flags = (U8)minnext;
3570 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
3572 if (data_fake.flags & SF_HAS_EVAL)
3573 data->flags |= SF_HAS_EVAL;
3574 data->whilem_c = data_fake.whilem_c;
3576 if (f & SCF_DO_STCLASS_AND) {
3577 const int was = (data->start_class->flags & ANYOF_EOS);
3579 cl_and(data->start_class, &intrnl);
3581 data->start_class->flags |= ANYOF_EOS;
3584 #if PERL_ENABLE_POSITIVE_ASSERTION_STUDY
3586 /* Positive Lookahead/lookbehind
3587 In this case we can do fixed string optimisation,
3588 but we must be careful about it. Note in the case of
3589 lookbehind the positions will be offset by the minimum
3590 length of the pattern, something we won't know about
3591 until after the recurse.
3593 I32 deltanext, fake = 0;
3595 struct regnode_charclass_class intrnl;
3597 /* We use SAVEFREEPV so that when the full compile
3598 is finished perl will clean up the allocated
3599 minlens when its all done. This was we don't
3600 have to worry about freeing them when we know
3601 they wont be used, which would be a pain.
3604 Newx( minnextp, 1, I32 );
3605 SAVEFREEPV(minnextp);
3608 StructCopy(data, &data_fake, scan_data_t);
3609 if ((flags & SCF_DO_SUBSTR) && data->last_found) {
3612 SCAN_COMMIT(pRExC_state, &data_fake,minlenp);
3613 data_fake.last_found=newSVsv(data->last_found);
3617 data_fake.last_closep = &fake;
3618 data_fake.flags = 0;
3619 data_fake.pos_delta = delta;
3621 data_fake.flags |= SF_IS_INF;
3622 if ( flags & SCF_DO_STCLASS && !scan->flags
3623 && OP(scan) == IFMATCH ) { /* Lookahead */
3624 cl_init(pRExC_state, &intrnl);
3625 data_fake.start_class = &intrnl;
3626 f |= SCF_DO_STCLASS_AND;
3628 if (flags & SCF_WHILEM_VISITED_POS)
3629 f |= SCF_WHILEM_VISITED_POS;
3630 next = regnext(scan);
3631 nscan = NEXTOPER(NEXTOPER(scan));
3633 *minnextp = study_chunk(pRExC_state, &nscan, minnextp, &deltanext,
3634 last, &data_fake, stopparen, recursed, NULL, f,depth+1);
3637 FAIL("Variable length lookbehind not implemented");
3639 else if (*minnextp > (I32)U8_MAX) {
3640 FAIL2("Lookbehind longer than %"UVuf" not implemented", (UV)U8_MAX);
3642 scan->flags = (U8)*minnextp;
3647 if (f & SCF_DO_STCLASS_AND) {
3648 const int was = (data->start_class->flags & ANYOF_EOS);
3650 cl_and(data->start_class, &intrnl);
3652 data->start_class->flags |= ANYOF_EOS;
3655 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
3657 if (data_fake.flags & SF_HAS_EVAL)
3658 data->flags |= SF_HAS_EVAL;
3659 data->whilem_c = data_fake.whilem_c;
3660 if ((flags & SCF_DO_SUBSTR) && data_fake.last_found) {
3661 if (RExC_rx->minlen<*minnextp)
3662 RExC_rx->minlen=*minnextp;
3663 SCAN_COMMIT(pRExC_state, &data_fake, minnextp);
3664 SvREFCNT_dec(data_fake.last_found);
3666 if ( data_fake.minlen_fixed != minlenp )
3668 data->offset_fixed= data_fake.offset_fixed;
3669 data->minlen_fixed= data_fake.minlen_fixed;
3670 data->lookbehind_fixed+= scan->flags;
3672 if ( data_fake.minlen_float != minlenp )
3674 data->minlen_float= data_fake.minlen_float;
3675 data->offset_float_min=data_fake.offset_float_min;
3676 data->offset_float_max=data_fake.offset_float_max;
3677 data->lookbehind_float+= scan->flags;
3686 else if (OP(scan) == OPEN) {
3687 if (stopparen != (I32)ARG(scan))
3690 else if (OP(scan) == CLOSE) {
3691 if (stopparen == (I32)ARG(scan)) {
3694 if ((I32)ARG(scan) == is_par) {
3695 next = regnext(scan);
3697 if ( next && (OP(next) != WHILEM) && next < last)
3698 is_par = 0; /* Disable optimization */
3701 *(data->last_closep) = ARG(scan);
3703 else if (OP(scan) == EVAL) {
3705 data->flags |= SF_HAS_EVAL;
3707 else if ( PL_regkind[OP(scan)] == ENDLIKE ) {
3708 if (flags & SCF_DO_SUBSTR) {
3709 SCAN_COMMIT(pRExC_state,data,minlenp);
3710 flags &= ~SCF_DO_SUBSTR;
3712 if (data && OP(scan)==ACCEPT) {
3713 data->flags |= SCF_SEEN_ACCEPT;
3718 else if (OP(scan) == LOGICAL && scan->flags == 2) /* Embedded follows */
3720 if (flags & SCF_DO_SUBSTR) {
3721 SCAN_COMMIT(pRExC_state,data,minlenp);
3722 data->longest = &(data->longest_float);
3724 is_inf = is_inf_internal = 1;
3725 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
3726 cl_anything(pRExC_state, data->start_class);
3727 flags &= ~SCF_DO_STCLASS;
3729 else if (OP(scan) == GPOS) {
3730 if (!(RExC_rx->extflags & RXf_GPOS_FLOAT) &&
3731 !(delta || is_inf || (data && data->pos_delta)))
3733 if (!(RExC_rx->extflags & RXf_ANCH) && (flags & SCF_DO_SUBSTR))
3734 RExC_rx->extflags |= RXf_ANCH_GPOS;
3735 if (RExC_rx->gofs < (U32)min)
3736 RExC_rx->gofs = min;
3738 RExC_rx->extflags |= RXf_GPOS_FLOAT;
3742 #ifdef TRIE_STUDY_OPT
3743 #ifdef FULL_TRIE_STUDY
3744 else if (PL_regkind[OP(scan)] == TRIE) {
3745 /* NOTE - There is similar code to this block above for handling
3746 BRANCH nodes on the initial study. If you change stuff here
3748 regnode *trie_node= scan;
3749 regnode *tail= regnext(scan);
3750 reg_trie_data *trie = (reg_trie_data*)RExC_rxi->data->data[ ARG(scan) ];
3751 I32 max1 = 0, min1 = I32_MAX;
3752 struct regnode_charclass_class accum;
3754 if (flags & SCF_DO_SUBSTR) /* XXXX Add !SUSPEND? */
3755 SCAN_COMMIT(pRExC_state, data,minlenp); /* Cannot merge strings after this. */
3756 if (flags & SCF_DO_STCLASS)
3757 cl_init_zero(pRExC_state, &accum);
3763 const regnode *nextbranch= NULL;
3766 for ( word=1 ; word <= trie->wordcount ; word++)
3768 I32 deltanext=0, minnext=0, f = 0, fake;
3769 struct regnode_charclass_class this_class;
3771 data_fake.flags = 0;
3773 data_fake.whilem_c = data->whilem_c;
3774 data_fake.last_closep = data->last_closep;
3777 data_fake.last_closep = &fake;
3778 data_fake.pos_delta = delta;
3779 if (flags & SCF_DO_STCLASS) {
3780 cl_init(pRExC_state, &this_class);
3781 data_fake.start_class = &this_class;
3782 f = SCF_DO_STCLASS_AND;
3784 if (flags & SCF_WHILEM_VISITED_POS)
3785 f |= SCF_WHILEM_VISITED_POS;
3787 if (trie->jump[word]) {
3789 nextbranch = trie_node + trie->jump[0];
3790 scan= trie_node + trie->jump[word];
3791 /* We go from the jump point to the branch that follows
3792 it. Note this means we need the vestigal unused branches
3793 even though they arent otherwise used.
3795 minnext = study_chunk(pRExC_state, &scan, minlenp,
3796 &deltanext, (regnode *)nextbranch, &data_fake,
3797 stopparen, recursed, NULL, f,depth+1);
3799 if (nextbranch && PL_regkind[OP(nextbranch)]==BRANCH)
3800 nextbranch= regnext((regnode*)nextbranch);
3802 if (min1 > (I32)(minnext + trie->minlen))
3803 min1 = minnext + trie->minlen;
3804 if (max1 < (I32)(minnext + deltanext + trie->maxlen))
3805 max1 = minnext + deltanext + trie->maxlen;
3806 if (deltanext == I32_MAX)
3807 is_inf = is_inf_internal = 1;
3809 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
3811 if (data_fake.flags & SCF_SEEN_ACCEPT) {
3812 if ( stopmin > min + min1)
3813 stopmin = min + min1;
3814 flags &= ~SCF_DO_SUBSTR;
3816 data->flags |= SCF_SEEN_ACCEPT;
3819 if (data_fake.flags & SF_HAS_EVAL)
3820 data->flags |= SF_HAS_EVAL;
3821 data->whilem_c = data_fake.whilem_c;
3823 if (flags & SCF_DO_STCLASS)
3824 cl_or(pRExC_state, &accum, &this_class);
3827 if (flags & SCF_DO_SUBSTR) {
3828 data->pos_min += min1;
3829 data->pos_delta += max1 - min1;
3830 if (max1 != min1 || is_inf)
3831 data->longest = &(data->longest_float);
3834 delta += max1 - min1;
3835 if (flags & SCF_DO_STCLASS_OR) {
3836 cl_or(pRExC_state, data->start_class, &accum);
3838 cl_and(data->start_class, and_withp);
3839 flags &= ~SCF_DO_STCLASS;
3842 else if (flags & SCF_DO_STCLASS_AND) {
3844 cl_and(data->start_class, &accum);
3845 flags &= ~SCF_DO_STCLASS;
3848 /* Switch to OR mode: cache the old value of
3849 * data->start_class */
3851 StructCopy(data->start_class, and_withp,
3852 struct regnode_charclass_class);
3853 flags &= ~SCF_DO_STCLASS_AND;
3854 StructCopy(&accum, data->start_class,
3855 struct regnode_charclass_class);
3856 flags |= SCF_DO_STCLASS_OR;
3857 data->start_class->flags |= ANYOF_EOS;
3864 else if (PL_regkind[OP(scan)] == TRIE) {
3865 reg_trie_data *trie = (reg_trie_data*)RExC_rxi->data->data[ ARG(scan) ];
3868 min += trie->minlen;
3869 delta += (trie->maxlen - trie->minlen);
3870 flags &= ~SCF_DO_STCLASS; /* xxx */
3871 if (flags & SCF_DO_SUBSTR) {
3872 SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot expect anything... */
3873 data->pos_min += trie->minlen;
3874 data->pos_delta += (trie->maxlen - trie->minlen);
3875 if (trie->maxlen != trie->minlen)
3876 data->longest = &(data->longest_float);
3878 if (trie->jump) /* no more substrings -- for now /grr*/
3879 flags &= ~SCF_DO_SUBSTR;
3881 #endif /* old or new */
3882 #endif /* TRIE_STUDY_OPT */
3883 /* Else: zero-length, ignore. */
3884 scan = regnext(scan);
3889 stopparen = frame->stop;
3890 frame = frame->prev;
3891 goto fake_study_recurse;
3896 DEBUG_STUDYDATA("pre-fin:",data,depth);
3899 *deltap = is_inf_internal ? I32_MAX : delta;
3900 if (flags & SCF_DO_SUBSTR && is_inf)
3901 data->pos_delta = I32_MAX - data->pos_min;
3902 if (is_par > (I32)U8_MAX)
3904 if (is_par && pars==1 && data) {
3905 data->flags |= SF_IN_PAR;
3906 data->flags &= ~SF_HAS_PAR;
3908 else if (pars && data) {
3909 data->flags |= SF_HAS_PAR;
3910 data->flags &= ~SF_IN_PAR;
3912 if (flags & SCF_DO_STCLASS_OR)
3913 cl_and(data->start_class, and_withp);
3914 if (flags & SCF_TRIE_RESTUDY)
3915 data->flags |= SCF_TRIE_RESTUDY;
3917 DEBUG_STUDYDATA("post-fin:",data,depth);
3919 return min < stopmin ? min : stopmin;
3923 S_add_data(RExC_state_t *pRExC_state, U32 n, const char *s)
3925 U32 count = RExC_rxi->data ? RExC_rxi->data->count : 0;
3927 Renewc(RExC_rxi->data,
3928 sizeof(*RExC_rxi->data) + sizeof(void*) * (count + n - 1),
3929 char, struct reg_data);
3931 Renew(RExC_rxi->data->what, count + n, U8);
3933 Newx(RExC_rxi->data->what, n, U8);
3934 RExC_rxi->data->count = count + n;
3935 Copy(s, RExC_rxi->data->what + count, n, U8);
3939 /*XXX: todo make this not included in a non debugging perl */
3940 #ifndef PERL_IN_XSUB_RE
3942 Perl_reginitcolors(pTHX)
3945 const char * const s = PerlEnv_getenv("PERL_RE_COLORS");
3947 char *t = savepv(s);
3951 t = strchr(t, '\t');
3957 PL_colors[i] = t = (char *)"";
3962 PL_colors[i++] = (char *)"";
3969 #ifdef TRIE_STUDY_OPT
3970 #define CHECK_RESTUDY_GOTO \
3972 (data.flags & SCF_TRIE_RESTUDY) \
3976 #define CHECK_RESTUDY_GOTO
3980 - pregcomp - compile a regular expression into internal code
3982 * We can't allocate space until we know how big the compiled form will be,
3983 * but we can't compile it (and thus know how big it is) until we've got a
3984 * place to put the code. So we cheat: we compile it twice, once with code
3985 * generation turned off and size counting turned on, and once "for real".
3986 * This also means that we don't allocate space until we are sure that the
3987 * thing really will compile successfully, and we never have to move the
3988 * code and thus invalidate pointers into it. (Note that it has to be in
3989 * one piece because free() must be able to free it all.) [NB: not true in perl]
3991 * Beware that the optimization-preparation code in here knows about some
3992 * of the structure of the compiled regexp. [I'll say.]
3997 #ifndef PERL_IN_XSUB_RE
3998 #define RE_ENGINE_PTR &PL_core_reg_engine
4000 extern const struct regexp_engine my_reg_engine;
4001 #define RE_ENGINE_PTR &my_reg_engine
4004 #ifndef PERL_IN_XSUB_RE
4006 Perl_pregcomp(pTHX_ char *exp, char *xend, PMOP *pm)
4009 HV * const table = GvHV(PL_hintgv);
4010 /* Dispatch a request to compile a regexp to correct
4013 SV **ptr= hv_fetchs(table, "regcomp", FALSE);
4014 GET_RE_DEBUG_FLAGS_DECL;
4015 if (ptr && SvIOK(*ptr) && SvIV(*ptr)) {
4016 const regexp_engine *eng=INT2PTR(regexp_engine*,SvIV(*ptr));
4018 PerlIO_printf(Perl_debug_log, "Using engine %"UVxf"\n",
4021 return CALLREGCOMP_ENG(eng, exp, xend, pm);
4024 return Perl_re_compile(aTHX_ exp, xend, pm);
4029 Perl_re_compile(pTHX_ char *exp, char *xend, PMOP *pm)
4033 register regexp_internal *ri;
4041 RExC_state_t RExC_state;
4042 RExC_state_t * const pRExC_state = &RExC_state;
4043 #ifdef TRIE_STUDY_OPT
4045 RExC_state_t copyRExC_state;
4047 GET_RE_DEBUG_FLAGS_DECL;
4048 DEBUG_r(if (!PL_colorset) reginitcolors());
4051 FAIL("NULL regexp argument");
4053 RExC_utf8 = pm->op_pmdynflags & PMdf_CMP_UTF8;
4057 SV *dsv= sv_newmortal();
4058 RE_PV_QUOTED_DECL(s, RExC_utf8,
4059 dsv, RExC_precomp, (xend - exp), 60);
4060 PerlIO_printf(Perl_debug_log, "%sCompiling REx%s %s\n",
4061 PL_colors[4],PL_colors[5],s);
4063 RExC_flags = pm->op_pmflags;
4067 RExC_seen_zerolen = *exp == '^' ? -1 : 0;
4068 RExC_seen_evals = 0;
4071 /* First pass: determine size, legality. */
4079 RExC_emit = &PL_regdummy;
4080 RExC_whilem_seen = 0;
4081 RExC_charnames = NULL;
4082 RExC_open_parens = NULL;
4083 RExC_close_parens = NULL;
4085 RExC_paren_names = NULL;
4087 RExC_paren_name_list = NULL;
4089 RExC_recurse = NULL;
4090 RExC_recurse_count = 0;
4092 #if 0 /* REGC() is (currently) a NOP at the first pass.
4093 * Clever compilers notice this and complain. --jhi */
4094 REGC((U8)REG_MAGIC, (char*)RExC_emit);
4096 DEBUG_PARSE_r(PerlIO_printf(Perl_debug_log, "Starting first pass (sizing)\n"));
4097 if (reg(pRExC_state, 0, &flags,1) == NULL) {
4098 RExC_precomp = NULL;
4102 PerlIO_printf(Perl_debug_log,
4103 "Required size %"IVdf" nodes\n"
4104 "Starting second pass (creation)\n",
4107 RExC_lastparse=NULL;
4109 /* Small enough for pointer-storage convention?
4110 If extralen==0, this means that we will not need long jumps. */
4111 if (RExC_size >= 0x10000L && RExC_extralen)
4112 RExC_size += RExC_extralen;
4115 if (RExC_whilem_seen > 15)
4116 RExC_whilem_seen = 15;
4119 /* Make room for a sentinel value at the end of the program */
4123 /* Allocate space and zero-initialize. Note, the two step process
4124 of zeroing when in debug mode, thus anything assigned has to
4125 happen after that */
4126 Newxz(r, 1, regexp);
4127 Newxc(ri, sizeof(regexp_internal) + (unsigned)RExC_size * sizeof(regnode),
4128 char, regexp_internal);
4129 if ( r == NULL || ri == NULL )
4130 FAIL("Regexp out of space");
4132 /* avoid reading uninitialized memory in DEBUGGING code in study_chunk() */
4133 Zero(ri, sizeof(regexp_internal) + (unsigned)RExC_size * sizeof(regnode), char);
4135 /* bulk initialize base fields with 0. */
4136 Zero(ri, sizeof(regexp_internal), char);
4139 /* non-zero initialization begins here */
4141 r->engine= RE_ENGINE_PTR;
4143 r->prelen = xend - exp;
4144 r->extflags = pm->op_pmflags & RXf_PMf_COMPILETIME;
4146 bool has_k = ((r->extflags & RXf_PMf_KEEPCOPY) == RXf_PMf_KEEPCOPY);
4147 bool has_minus = ((r->extflags & RXf_PMf_STD_PMMOD) != RXf_PMf_STD_PMMOD);
4148 bool has_runon = ((RExC_seen & REG_SEEN_RUN_ON_COMMENT)==REG_SEEN_RUN_ON_COMMENT);
4149 U16 reganch = (U16)((r->extflags & RXf_PMf_STD_PMMOD) >> 12);
4150 const char *fptr = STD_PAT_MODS; /*"msix"*/
4152 r->wraplen = r->prelen + has_minus + has_k + has_runon
4153 + (sizeof(STD_PAT_MODS) - 1)
4154 + (sizeof("(?:)") - 1);
4156 Newx(r->wrapped, r->wraplen, char );
4160 *p++ = KEEPCOPY_PAT_MOD; /*'k'*/
4162 char *r = p + (sizeof(STD_PAT_MODS) - 1) + has_minus - 1;
4163 char *colon = r + 1;
4166 while((ch = *fptr++)) {
4180 Copy(RExC_precomp, p, r->prelen, char);
4189 r->nparens = RExC_npar - 1; /* set early to validate backrefs */
4191 if (RExC_seen & REG_SEEN_RECURSE) {
4192 Newxz(RExC_open_parens, RExC_npar,regnode *);
4193 SAVEFREEPV(RExC_open_parens);
4194 Newxz(RExC_close_parens,RExC_npar,regnode *);
4195 SAVEFREEPV(RExC_close_parens);
4198 /* Useful during FAIL. */
4199 #ifdef RE_TRACK_PATTERN_OFFSETS
4200 Newxz(ri->u.offsets, 2*RExC_size+1, U32); /* MJD 20001228 */
4201 DEBUG_OFFSETS_r(PerlIO_printf(Perl_debug_log,
4202 "%s %"UVuf" bytes for offset annotations.\n",
4203 ri->u.offsets ? "Got" : "Couldn't get",
4204 (UV)((2*RExC_size+1) * sizeof(U32))));
4206 SetProgLen(ri,RExC_size);
4210 /* Second pass: emit code. */
4211 RExC_flags = pm->op_pmflags; /* don't let top level (?i) bleed */
4216 RExC_emit_start = ri->program;
4217 RExC_emit = ri->program;
4219 /* put a sentinal on the end of the program so we can check for
4221 ri->program[RExC_size].type = 255;
4223 /* Store the count of eval-groups for security checks: */
4224 RExC_rx->seen_evals = RExC_seen_evals;
4225 REGC((U8)REG_MAGIC, (char*) RExC_emit++);
4226 if (reg(pRExC_state, 0, &flags,1) == NULL)
4229 /* XXXX To minimize changes to RE engine we always allocate
4230 3-units-long substrs field. */
4231 Newx(r->substrs, 1, struct reg_substr_data);
4232 if (RExC_recurse_count) {
4233 Newxz(RExC_recurse,RExC_recurse_count,regnode *);
4234 SAVEFREEPV(RExC_recurse);
4238 r->minlen = minlen = sawplus = sawopen = 0;
4239 Zero(r->substrs, 1, struct reg_substr_data);
4241 #ifdef TRIE_STUDY_OPT
4244 DEBUG_OPTIMISE_r(PerlIO_printf(Perl_debug_log,"Restudying\n"));
4246 RExC_state = copyRExC_state;
4247 if (seen & REG_TOP_LEVEL_BRANCHES)
4248 RExC_seen |= REG_TOP_LEVEL_BRANCHES;
4250 RExC_seen &= ~REG_TOP_LEVEL_BRANCHES;
4251 if (data.last_found) {
4252 SvREFCNT_dec(data.longest_fixed);
4253 SvREFCNT_dec(data.longest_float);
4254 SvREFCNT_dec(data.last_found);
4256 StructCopy(&zero_scan_data, &data, scan_data_t);
4258 StructCopy(&zero_scan_data, &data, scan_data_t);
4259 copyRExC_state = RExC_state;
4262 StructCopy(&zero_scan_data, &data, scan_data_t);
4265 /* Dig out information for optimizations. */
4266 r->extflags = pm->op_pmflags & RXf_PMf_COMPILETIME; /* Again? */
4267 pm->op_pmflags = RExC_flags;
4269 r->extflags |= RXf_UTF8; /* Unicode in it? */
4270 ri->regstclass = NULL;
4271 if (RExC_naughty >= 10) /* Probably an expensive pattern. */
4272 r->intflags |= PREGf_NAUGHTY;
4273 scan = ri->program + 1; /* First BRANCH. */
4275 /* testing for BRANCH here tells us whether there is "must appear"
4276 data in the pattern. If there is then we can use it for optimisations */
4277 if (!(RExC_seen & REG_TOP_LEVEL_BRANCHES)) { /* Only one top-level choice. */
4279 STRLEN longest_float_length, longest_fixed_length;
4280 struct regnode_charclass_class ch_class; /* pointed to by data */
4282 I32 last_close = 0; /* pointed to by data */
4285 /* Skip introductions and multiplicators >= 1. */
4286 while ((OP(first) == OPEN && (sawopen = 1)) ||
4287 /* An OR of *one* alternative - should not happen now. */
4288 (OP(first) == BRANCH && OP(regnext(first)) != BRANCH) ||
4289 /* for now we can't handle lookbehind IFMATCH*/
4290 (OP(first) == IFMATCH && !first->flags) ||
4291 (OP(first) == PLUS) ||
4292 (OP(first) == MINMOD) ||
4293 /* An {n,m} with n>0 */
4294 (PL_regkind[OP(first)] == CURLY && ARG1(first) > 0) )
4297 if (OP(first) == PLUS)
4300 first += regarglen[OP(first)];
4301 if (OP(first) == IFMATCH) {
4302 first = NEXTOPER(first);
4303 first += EXTRA_STEP_2ARGS;
4304 } else /* XXX possible optimisation for /(?=)/ */
4305 first = NEXTOPER(first);
4308 /* Starting-point info. */
4310 DEBUG_PEEP("first:",first,0);
4311 /* Ignore EXACT as we deal with it later. */
4312 if (PL_regkind[OP(first)] == EXACT) {
4313 if (OP(first) == EXACT)
4314 NOOP; /* Empty, get anchored substr later. */
4315 else if ((OP(first) == EXACTF || OP(first) == EXACTFL))
4316 ri->regstclass = first;
4319 else if (PL_regkind[OP(first)] == TRIE &&
4320 ((reg_trie_data *)ri->data->data[ ARG(first) ])->minlen>0)
4323 /* this can happen only on restudy */
4324 if ( OP(first) == TRIE ) {
4325 struct regnode_1 *trieop = (struct regnode_1 *)
4326 PerlMemShared_calloc(1, sizeof(struct regnode_1));
4327 StructCopy(first,trieop,struct regnode_1);
4328 trie_op=(regnode *)trieop;
4330 struct regnode_charclass *trieop = (struct regnode_charclass *)
4331 PerlMemShared_calloc(1, sizeof(struct regnode_charclass));
4332 StructCopy(first,trieop,struct regnode_charclass);
4333 trie_op=(regnode *)trieop;
4336 make_trie_failtable(pRExC_state, (regnode *)first, trie_op, 0);
4337 ri->regstclass = trie_op;
4340 else if (strchr((const char*)PL_simple,OP(first)))
4341 ri->regstclass = first;
4342 else if (PL_regkind[OP(first)] == BOUND ||
4343 PL_regkind[OP(first)] == NBOUND)
4344 ri->regstclass = first;
4345 else if (PL_regkind[OP(first)] == BOL) {
4346 r->extflags |= (OP(first) == MBOL
4348 : (OP(first) == SBOL
4351 first = NEXTOPER(first);
4354 else if (OP(first) == GPOS) {
4355 r->extflags |= RXf_ANCH_GPOS;
4356 first = NEXTOPER(first);
4359 else if ((!sawopen || !RExC_sawback) &&
4360 (OP(first) == STAR &&
4361 PL_regkind[OP(NEXTOPER(first))] == REG_ANY) &&
4362 !(r->extflags & RXf_ANCH) && !(RExC_seen & REG_SEEN_EVAL))
4364 /* turn .* into ^.* with an implied $*=1 */
4366 (OP(NEXTOPER(first)) == REG_ANY)
4369 r->extflags |= type;
4370 r->intflags |= PREGf_IMPLICIT;
4371 first = NEXTOPER(first);
4374 if (sawplus && (!sawopen || !RExC_sawback)
4375 && !(RExC_seen & REG_SEEN_EVAL)) /* May examine pos and $& */
4376 /* x+ must match at the 1st pos of run of x's */
4377 r->intflags |= PREGf_SKIP;
4379 /* Scan is after the zeroth branch, first is atomic matcher. */
4380 #ifdef TRIE_STUDY_OPT
4383 PerlIO_printf(Perl_debug_log, "first at %"IVdf"\n",
4384 (IV)(first - scan + 1))
4388 PerlIO_printf(Perl_debug_log, "first at %"IVdf"\n",
4389 (IV)(first - scan + 1))
4395 * If there's something expensive in the r.e., find the
4396 * longest literal string that must appear and make it the
4397 * regmust. Resolve ties in favor of later strings, since
4398 * the regstart check works with the beginning of the r.e.
4399 * and avoiding duplication strengthens checking. Not a
4400 * strong reason, but sufficient in the absence of others.
4401 * [Now we resolve ties in favor of the earlier string if
4402 * it happens that c_offset_min has been invalidated, since the
4403 * earlier string may buy us something the later one won't.]
4406 data.longest_fixed = newSVpvs("");
4407 data.longest_float = newSVpvs("");
4408 data.last_found = newSVpvs("");
4409 data.longest = &(data.longest_fixed);
4411 if (!ri->regstclass) {
4412 cl_init(pRExC_state, &ch_class);
4413 data.start_class = &ch_class;
4414 stclass_flag = SCF_DO_STCLASS_AND;
4415 } else /* XXXX Check for BOUND? */
4417 data.last_closep = &last_close;
4419 minlen = study_chunk(pRExC_state, &first, &minlen, &fake, scan + RExC_size, /* Up to end */
4420 &data, -1, NULL, NULL,
4421 SCF_DO_SUBSTR | SCF_WHILEM_VISITED_POS | stclass_flag,0);
4427 if ( RExC_npar == 1 && data.longest == &(data.longest_fixed)
4428 && data.last_start_min == 0 && data.last_end > 0
4429 && !RExC_seen_zerolen
4430 && !(RExC_seen & REG_SEEN_VERBARG)
4431 && (!(RExC_seen & REG_SEEN_GPOS) || (r->extflags & RXf_ANCH_GPOS)))
4432 r->extflags |= RXf_CHECK_ALL;
4433 scan_commit(pRExC_state, &data,&minlen,0);
4434 SvREFCNT_dec(data.last_found);
4436 /* Note that code very similar to this but for anchored string
4437 follows immediately below, changes may need to be made to both.
4440 longest_float_length = CHR_SVLEN(data.longest_float);
4441 if (longest_float_length
4442 || (data.flags & SF_FL_BEFORE_EOL
4443 && (!(data.flags & SF_FL_BEFORE_MEOL)
4444 || (RExC_flags & RXf_PMf_MULTILINE))))
4448 if (SvCUR(data.longest_fixed) /* ok to leave SvCUR */
4449 && data.offset_fixed == data.offset_float_min
4450 && SvCUR(data.longest_fixed) == SvCUR(data.longest_float))
4451 goto remove_float; /* As in (a)+. */
4453 /* copy the information about the longest float from the reg_scan_data
4454 over to the program. */
4455 if (SvUTF8(data.longest_float)) {
4456 r->float_utf8 = data.longest_float;
4457 r->float_substr = NULL;
4459 r->float_substr = data.longest_float;
4460 r->float_utf8 = NULL;
4462 /* float_end_shift is how many chars that must be matched that
4463 follow this item. We calculate it ahead of time as once the
4464 lookbehind offset is added in we lose the ability to correctly
4466 ml = data.minlen_float ? *(data.minlen_float)
4467 : (I32)longest_float_length;
4468 r->float_end_shift = ml - data.offset_float_min
4469 - longest_float_length + (SvTAIL(data.longest_float) != 0)
4470 + data.lookbehind_float;
4471 r->float_min_offset = data.offset_float_min - data.lookbehind_float;
4472 r->float_max_offset = data.offset_float_max;
4473 if (data.offset_float_max < I32_MAX) /* Don't offset infinity */
4474 r->float_max_offset -= data.lookbehind_float;
4476 t = (data.flags & SF_FL_BEFORE_EOL /* Can't have SEOL and MULTI */
4477 && (!(data.flags & SF_FL_BEFORE_MEOL)
4478 || (RExC_flags & RXf_PMf_MULTILINE)));
4479 fbm_compile(data.longest_float, t ? FBMcf_TAIL : 0);
4483 r->float_substr = r->float_utf8 = NULL;
4484 SvREFCNT_dec(data.longest_float);
4485 longest_float_length = 0;
4488 /* Note that code very similar to this but for floating string
4489 is immediately above, changes may need to be made to both.
4492 longest_fixed_length = CHR_SVLEN(data.longest_fixed);
4493 if (longest_fixed_length
4494 || (data.flags & SF_FIX_BEFORE_EOL /* Cannot have SEOL and MULTI */
4495 && (!(data.flags & SF_FIX_BEFORE_MEOL)
4496 || (RExC_flags & RXf_PMf_MULTILINE))))
4500 /* copy the information about the longest fixed
4501 from the reg_scan_data over to the program. */
4502 if (SvUTF8(data.longest_fixed)) {
4503 r->anchored_utf8 = data.longest_fixed;
4504 r->anchored_substr = NULL;
4506 r->anchored_substr = data.longest_fixed;
4507 r->anchored_utf8 = NULL;
4509 /* fixed_end_shift is how many chars that must be matched that
4510 follow this item. We calculate it ahead of time as once the
4511 lookbehind offset is added in we lose the ability to correctly
4513 ml = data.minlen_fixed ? *(data.minlen_fixed)
4514 : (I32)longest_fixed_length;
4515 r->anchored_end_shift = ml - data.offset_fixed
4516 - longest_fixed_length + (SvTAIL(data.longest_fixed) != 0)
4517 + data.lookbehind_fixed;
4518 r->anchored_offset = data.offset_fixed - data.lookbehind_fixed;
4520 t = (data.flags & SF_FIX_BEFORE_EOL /* Can't have SEOL and MULTI */
4521 && (!(data.flags & SF_FIX_BEFORE_MEOL)
4522 || (RExC_flags & RXf_PMf_MULTILINE)));
4523 fbm_compile(data.longest_fixed, t ? FBMcf_TAIL : 0);
4526 r->anchored_substr = r->anchored_utf8 = NULL;
4527 SvREFCNT_dec(data.longest_fixed);
4528 longest_fixed_length = 0;
4531 && (OP(ri->regstclass) == REG_ANY || OP(ri->regstclass) == SANY))
4532 ri->regstclass = NULL;
4533 if ((!(r->anchored_substr || r->anchored_utf8) || r->anchored_offset)
4535 && !(data.start_class->flags & ANYOF_EOS)
4536 && !cl_is_anything(data.start_class))
4538 const U32 n = add_data(pRExC_state, 1, "f");
4540 Newx(RExC_rxi->data->data[n], 1,
4541 struct regnode_charclass_class);
4542 StructCopy(data.start_class,
4543 (struct regnode_charclass_class*)RExC_rxi->data->data[n],
4544 struct regnode_charclass_class);
4545 ri->regstclass = (regnode*)RExC_rxi->data->data[n];
4546 r->intflags &= ~PREGf_SKIP; /* Used in find_byclass(). */
4547 DEBUG_COMPILE_r({ SV *sv = sv_newmortal();
4548 regprop(r, sv, (regnode*)data.start_class);
4549 PerlIO_printf(Perl_debug_log,
4550 "synthetic stclass \"%s\".\n",
4551 SvPVX_const(sv));});
4554 /* A temporary algorithm prefers floated substr to fixed one to dig more info. */
4555 if (longest_fixed_length > longest_float_length) {
4556 r->check_end_shift = r->anchored_end_shift;
4557 r->check_substr = r->anchored_substr;
4558 r->check_utf8 = r->anchored_utf8;
4559 r->check_offset_min = r->check_offset_max = r->anchored_offset;
4560 if (r->extflags & RXf_ANCH_SINGLE)
4561 r->extflags |= RXf_NOSCAN;
4564 r->check_end_shift = r->float_end_shift;
4565 r->check_substr = r->float_substr;
4566 r->check_utf8 = r->float_utf8;
4567 r->check_offset_min = r->float_min_offset;
4568 r->check_offset_max = r->float_max_offset;
4570 /* XXXX Currently intuiting is not compatible with ANCH_GPOS.
4571 This should be changed ASAP! */
4572 if ((r->check_substr || r->check_utf8) && !(r->extflags & RXf_ANCH_GPOS)) {
4573 r->extflags |= RXf_USE_INTUIT;
4574 if (SvTAIL(r->check_substr ? r->check_substr : r->check_utf8))
4575 r->extflags |= RXf_INTUIT_TAIL;
4577 /* XXX Unneeded? dmq (shouldn't as this is handled elsewhere)
4578 if ( (STRLEN)minlen < longest_float_length )
4579 minlen= longest_float_length;
4580 if ( (STRLEN)minlen < longest_fixed_length )
4581 minlen= longest_fixed_length;
4585 /* Several toplevels. Best we can is to set minlen. */
4587 struct regnode_charclass_class ch_class;
4590 DEBUG_PARSE_r(PerlIO_printf(Perl_debug_log, "\nMulti Top Level\n"));
4592 scan = ri->program + 1;
4593 cl_init(pRExC_state, &ch_class);
4594 data.start_class = &ch_class;
4595 data.last_closep = &last_close;
4598 minlen = study_chunk(pRExC_state, &scan, &minlen, &fake, scan + RExC_size,
4599 &data, -1, NULL, NULL, SCF_DO_STCLASS_AND|SCF_WHILEM_VISITED_POS,0);
4603 r->check_substr = r->check_utf8 = r->anchored_substr = r->anchored_utf8
4604 = r->float_substr = r->float_utf8 = NULL;
4605 if (!(data.start_class->flags & ANYOF_EOS)
4606 && !cl_is_anything(data.start_class))
4608 const U32 n = add_data(pRExC_state, 1, "f");
4610 Newx(RExC_rxi->data->data[n], 1,
4611 struct regnode_charclass_class);
4612 StructCopy(data.start_class,
4613 (struct regnode_charclass_class*)RExC_rxi->data->data[n],
4614 struct regnode_charclass_class);
4615 ri->regstclass = (regnode*)RExC_rxi->data->data[n];
4616 r->intflags &= ~PREGf_SKIP; /* Used in find_byclass(). */
4617 DEBUG_COMPILE_r({ SV* sv = sv_newmortal();
4618 regprop(r, sv, (regnode*)data.start_class);
4619 PerlIO_printf(Perl_debug_log,
4620 "synthetic stclass \"%s\".\n",
4621 SvPVX_const(sv));});
4625 /* Guard against an embedded (?=) or (?<=) with a longer minlen than
4626 the "real" pattern. */
4628 PerlIO_printf(Perl_debug_log,"minlen: %"IVdf" r->minlen:%"IVdf"\n",
4629 (IV)minlen, (IV)r->minlen);
4631 r->minlenret = minlen;
4632 if (r->minlen < minlen)
4635 if (RExC_seen & REG_SEEN_GPOS)
4636 r->extflags |= RXf_GPOS_SEEN;
4637 if (RExC_seen & REG_SEEN_LOOKBEHIND)
4638 r->extflags |= RXf_LOOKBEHIND_SEEN;
4639 if (RExC_seen & REG_SEEN_EVAL)
4640 r->extflags |= RXf_EVAL_SEEN;
4641 if (RExC_seen & REG_SEEN_CANY)
4642 r->extflags |= RXf_CANY_SEEN;
4643 if (RExC_seen & REG_SEEN_VERBARG)
4644 r->intflags |= PREGf_VERBARG_SEEN;
4645 if (RExC_seen & REG_SEEN_CUTGROUP)
4646 r->intflags |= PREGf_CUTGROUP_SEEN;
4647 if (RExC_paren_names)
4648 r->paren_names = (HV*)SvREFCNT_inc(RExC_paren_names);
4650 r->paren_names = NULL;
4651 if (r->prelen == 3 && strEQ("\\s+", r->precomp))
4652 r->extflags |= RXf_WHITE;
4653 else if (r->prelen == 1 && r->precomp[0] == '^')
4654 r->extflags |= RXf_START_ONLY;
4657 if (RExC_paren_names) {
4658 ri->name_list_idx = add_data( pRExC_state, 1, "p" );
4659 ri->data->data[ri->name_list_idx] = (void*)SvREFCNT_inc(RExC_paren_name_list);
4662 ri->name_list_idx = 0;
4664 if (RExC_recurse_count) {
4665 for ( ; RExC_recurse_count ; RExC_recurse_count-- ) {
4666 const regnode *scan = RExC_recurse[RExC_recurse_count-1];
4667 ARG2L_SET( scan, RExC_open_parens[ARG(scan)-1] - scan );
4670 Newxz(r->startp, RExC_npar, I32);
4671 Newxz(r->endp, RExC_npar, I32);
4672 /* assume we don't need to swap parens around before we match */
4675 PerlIO_printf(Perl_debug_log,"Final program:\n");
4678 #ifdef RE_TRACK_PATTERN_OFFSETS
4679 DEBUG_OFFSETS_r(if (ri->u.offsets) {
4680 const U32 len = ri->u.offsets[0];
4682 GET_RE_DEBUG_FLAGS_DECL;
4683 PerlIO_printf(Perl_debug_log, "Offsets: [%"UVuf"]\n\t", (UV)ri->u.offsets[0]);
4684 for (i = 1; i <= len; i++) {
4685 if (ri->u.offsets[i*2-1] || ri->u.offsets[i*2])
4686 PerlIO_printf(Perl_debug_log, "%"UVuf":%"UVuf"[%"UVuf"] ",
4687 (UV)i, (UV)ri->u.offsets[i*2-1], (UV)ri->u.offsets[i*2]);
4689 PerlIO_printf(Perl_debug_log, "\n");
4695 #undef RE_ENGINE_PTR
4699 Perl_reg_named_buff_get(pTHX_ const REGEXP * const rx, SV* namesv, U32 flags)
4701 AV *retarray = NULL;
4706 if (rx && rx->paren_names) {
4707 HE *he_str = hv_fetch_ent( rx->paren_names, namesv, 0, 0 );
4710 SV* sv_dat=HeVAL(he_str);
4711 I32 *nums=(I32*)SvPVX(sv_dat);
4712 for ( i=0; i<SvIVX(sv_dat); i++ ) {
4713 if ((I32)(rx->nparens) >= nums[i]
4714 && rx->startp[nums[i]] != -1
4715 && rx->endp[nums[i]] != -1)
4717 ret = CALLREG_NUMBUF(rx,nums[i],NULL);
4721 ret = newSVsv(&PL_sv_undef);
4725 av_push(retarray, ret);
4729 return (SV*)retarray;
4736 Perl_reg_numbered_buff_get(pTHX_ const REGEXP * const rx, I32 paren, SV* usesv)
4741 SV *sv = usesv ? usesv : newSVpvs("");
4744 sv_setsv(sv,&PL_sv_undef);
4748 if (paren == -2 && rx->startp[0] != -1) {
4754 if (paren == -1 && rx->endp[0] != -1) {
4756 s = rx->subbeg + rx->endp[0];
4757 i = rx->sublen - rx->endp[0];
4760 if ( 0 <= paren && paren <= (I32)rx->nparens &&
4761 (s1 = rx->startp[paren]) != -1 &&
4762 (t1 = rx->endp[paren]) != -1)
4766 s = rx->subbeg + s1;
4768 sv_setsv(sv,&PL_sv_undef);
4771 assert(rx->sublen >= (s - rx->subbeg) + i );
4773 const int oldtainted = PL_tainted;
4775 sv_setpvn(sv, s, i);
4776 PL_tainted = oldtainted;
4777 if ( (rx->extflags & RXf_CANY_SEEN)
4778 ? (RX_MATCH_UTF8(rx)
4779 && (!i || is_utf8_string((U8*)s, i)))
4780 : (RX_MATCH_UTF8(rx)) )
4787 if (RX_MATCH_TAINTED(rx)) {
4788 if (SvTYPE(sv) >= SVt_PVMG) {
4789 MAGIC* const mg = SvMAGIC(sv);
4792 SvMAGIC_set(sv, mg->mg_moremagic);
4794 if ((mgt = SvMAGIC(sv))) {
4795 mg->mg_moremagic = mgt;
4796 SvMAGIC_set(sv, mg);
4806 sv_setsv(sv,&PL_sv_undef);
4812 /* Scans the name of a named buffer from the pattern.
4813 * If flags is REG_RSN_RETURN_NULL returns null.
4814 * If flags is REG_RSN_RETURN_NAME returns an SV* containing the name
4815 * If flags is REG_RSN_RETURN_DATA returns the data SV* corresponding
4816 * to the parsed name as looked up in the RExC_paren_names hash.
4817 * If there is an error throws a vFAIL().. type exception.
4820 #define REG_RSN_RETURN_NULL 0
4821 #define REG_RSN_RETURN_NAME 1
4822 #define REG_RSN_RETURN_DATA 2
4825 S_reg_scan_name(pTHX_ RExC_state_t *pRExC_state, U32 flags) {
4826 char *name_start = RExC_parse;
4828 if (isIDFIRST_lazy_if(RExC_parse, UTF)) {
4829 /* skip IDFIRST by using do...while */
4832 RExC_parse += UTF8SKIP(RExC_parse);
4833 } while (isALNUM_utf8((U8*)RExC_parse));
4837 } while (isALNUM(*RExC_parse));
4841 SV* sv_name = sv_2mortal(Perl_newSVpvn(aTHX_ name_start,
4842 (int)(RExC_parse - name_start)));
4845 if ( flags == REG_RSN_RETURN_NAME)
4847 else if (flags==REG_RSN_RETURN_DATA) {
4850 if ( ! sv_name ) /* should not happen*/
4851 Perl_croak(aTHX_ "panic: no svname in reg_scan_name");
4852 if (RExC_paren_names)
4853 he_str = hv_fetch_ent( RExC_paren_names, sv_name, 0, 0 );
4855 sv_dat = HeVAL(he_str);
4857 vFAIL("Reference to nonexistent named group");
4861 Perl_croak(aTHX_ "panic: bad flag in reg_scan_name");
4868 #define DEBUG_PARSE_MSG(funcname) DEBUG_PARSE_r({ \
4869 int rem=(int)(RExC_end - RExC_parse); \
4878 if (RExC_lastparse!=RExC_parse) \
4879 PerlIO_printf(Perl_debug_log," >%.*s%-*s", \
4882 iscut ? "..." : "<" \
4885 PerlIO_printf(Perl_debug_log,"%16s",""); \
4890 num=REG_NODE_NUM(RExC_emit); \
4891 if (RExC_lastnum!=num) \
4892 PerlIO_printf(Perl_debug_log,"|%4d",num); \
4894 PerlIO_printf(Perl_debug_log,"|%4s",""); \
4895 PerlIO_printf(Perl_debug_log,"|%*s%-4s", \
4896 (int)((depth*2)), "", \
4900 RExC_lastparse=RExC_parse; \
4905 #define DEBUG_PARSE(funcname) DEBUG_PARSE_r({ \
4906 DEBUG_PARSE_MSG((funcname)); \
4907 PerlIO_printf(Perl_debug_log,"%4s","\n"); \
4909 #define DEBUG_PARSE_FMT(funcname,fmt,args) DEBUG_PARSE_r({ \
4910 DEBUG_PARSE_MSG((funcname)); \
4911 PerlIO_printf(Perl_debug_log,fmt "\n",args); \
4914 - reg - regular expression, i.e. main body or parenthesized thing
4916 * Caller must absorb opening parenthesis.
4918 * Combining parenthesis handling with the base level of regular expression
4919 * is a trifle forced, but the need to tie the tails of the branches to what
4920 * follows makes it hard to avoid.
4922 #define REGTAIL(x,y,z) regtail((x),(y),(z),depth+1)
4924 #define REGTAIL_STUDY(x,y,z) regtail_study((x),(y),(z),depth+1)
4926 #define REGTAIL_STUDY(x,y,z) regtail((x),(y),(z),depth+1)
4929 /* this idea is borrowed from STR_WITH_LEN in handy.h */
4930 #define CHECK_WORD(s,v,l) \
4931 (((sizeof(s)-1)==(l)) && (strnEQ(start_verb, (s ""), (sizeof(s)-1))))
4934 S_reg(pTHX_ RExC_state_t *pRExC_state, I32 paren, I32 *flagp,U32 depth)
4935 /* paren: Parenthesized? 0=top, 1=(, inside: changed to letter. */
4938 register regnode *ret; /* Will be the head of the group. */
4939 register regnode *br;
4940 register regnode *lastbr;
4941 register regnode *ender = NULL;
4942 register I32 parno = 0;
4944 const I32 oregflags = RExC_flags;
4945 bool have_branch = 0;
4947 I32 freeze_paren = 0;
4948 I32 after_freeze = 0;
4950 /* for (?g), (?gc), and (?o) warnings; warning
4951 about (?c) will warn about (?g) -- japhy */
4953 #define WASTED_O 0x01
4954 #define WASTED_G 0x02
4955 #define WASTED_C 0x04
4956 #define WASTED_GC (0x02|0x04)
4957 I32 wastedflags = 0x00;
4959 char * parse_start = RExC_parse; /* MJD */
4960 char * const oregcomp_parse = RExC_parse;
4962 GET_RE_DEBUG_FLAGS_DECL;
4963 DEBUG_PARSE("reg ");
4966 *flagp = 0; /* Tentatively. */
4969 /* Make an OPEN node, if parenthesized. */
4971 if ( *RExC_parse == '*') { /* (*VERB:ARG) */
4972 char *start_verb = RExC_parse;
4973 STRLEN verb_len = 0;
4974 char *start_arg = NULL;
4975 unsigned char op = 0;
4977 int internal_argval = 0; /* internal_argval is only useful if !argok */
4978 while ( *RExC_parse && *RExC_parse != ')' ) {
4979 if ( *RExC_parse == ':' ) {
4980 start_arg = RExC_parse + 1;
4986 verb_len = RExC_parse - start_verb;
4989 while ( *RExC_parse && *RExC_parse != ')' )
4991 if ( *RExC_parse != ')' )
4992 vFAIL("Unterminated verb pattern argument");
4993 if ( RExC_parse == start_arg )
4996 if ( *RExC_parse != ')' )
4997 vFAIL("Unterminated verb pattern");
5000 switch ( *start_verb ) {
5001 case 'A': /* (*ACCEPT) */
5002 if ( CHECK_WORD("ACCEPT",start_verb,verb_len) ) {
5004 internal_argval = RExC_nestroot;
5007 case 'C': /* (*COMMIT) */
5008 if ( CHECK_WORD("COMMIT",start_verb,verb_len) )
5011 case 'F': /* (*FAIL) */
5012 if ( verb_len==1 || CHECK_WORD("FAIL",start_verb,verb_len) ) {
5017 case ':': /* (*:NAME) */
5018 case 'M': /* (*MARK:NAME) */
5019 if ( verb_len==0 || CHECK_WORD("MARK",start_verb,verb_len) ) {
5024 case 'P': /* (*PRUNE) */
5025 if ( CHECK_WORD("PRUNE",start_verb,verb_len) )
5028 case 'S': /* (*SKIP) */
5029 if ( CHECK_WORD("SKIP",start_verb,verb_len) )
5032 case 'T': /* (*THEN) */
5033 /* [19:06] <TimToady> :: is then */
5034 if ( CHECK_WORD("THEN",start_verb,verb_len) ) {
5036 RExC_seen |= REG_SEEN_CUTGROUP;
5042 vFAIL3("Unknown verb pattern '%.*s'",
5043 verb_len, start_verb);
5046 if ( start_arg && internal_argval ) {
5047 vFAIL3("Verb pattern '%.*s' may not have an argument",
5048 verb_len, start_verb);
5049 } else if ( argok < 0 && !start_arg ) {
5050 vFAIL3("Verb pattern '%.*s' has a mandatory argument",
5051 verb_len, start_verb);
5053 ret = reganode(pRExC_state, op, internal_argval);
5054 if ( ! internal_argval && ! SIZE_ONLY ) {
5056 SV *sv = newSVpvn( start_arg, RExC_parse - start_arg);
5057 ARG(ret) = add_data( pRExC_state, 1, "S" );
5058 RExC_rxi->data->data[ARG(ret)]=(void*)sv;
5065 if (!internal_argval)
5066 RExC_seen |= REG_SEEN_VERBARG;
5067 } else if ( start_arg ) {
5068 vFAIL3("Verb pattern '%.*s' may not have an argument",
5069 verb_len, start_verb);
5071 ret = reg_node(pRExC_state, op);
5073 nextchar(pRExC_state);
5076 if (*RExC_parse == '?') { /* (?...) */
5077 bool is_logical = 0;
5078 const char * const seqstart = RExC_parse;
5081 paren = *RExC_parse++;
5082 ret = NULL; /* For look-ahead/behind. */
5085 case 'P': /* (?P...) variants for those used to PCRE/Python */
5086 paren = *RExC_parse++;
5087 if ( paren == '<') /* (?P<...>) named capture */
5089 else if (paren == '>') { /* (?P>name) named recursion */
5090 goto named_recursion;
5092 else if (paren == '=') { /* (?P=...) named backref */
5093 /* this pretty much dupes the code for \k<NAME> in regatom(), if
5094 you change this make sure you change that */
5095 char* name_start = RExC_parse;
5097 SV *sv_dat = reg_scan_name(pRExC_state,
5098 SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
5099 if (RExC_parse == name_start || *RExC_parse != ')')
5100 vFAIL2("Sequence %.3s... not terminated",parse_start);
5103 num = add_data( pRExC_state, 1, "S" );
5104 RExC_rxi->data->data[num]=(void*)sv_dat;
5105 SvREFCNT_inc(sv_dat);
5108 ret = reganode(pRExC_state,
5109 (U8)(FOLD ? (LOC ? NREFFL : NREFF) : NREF),
5113 Set_Node_Offset(ret, parse_start+1);
5114 Set_Node_Cur_Length(ret); /* MJD */
5116 nextchar(pRExC_state);
5120 vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
5122 case '<': /* (?<...) */
5123 if (*RExC_parse == '!')
5125 else if (*RExC_parse != '=')
5131 case '\'': /* (?'...') */
5132 name_start= RExC_parse;
5133 svname = reg_scan_name(pRExC_state,
5134 SIZE_ONLY ? /* reverse test from the others */
5135 REG_RSN_RETURN_NAME :
5136 REG_RSN_RETURN_NULL);
5137 if (RExC_parse == name_start) {
5139 vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
5142 if (*RExC_parse != paren)
5143 vFAIL2("Sequence (?%c... not terminated",
5144 paren=='>' ? '<' : paren);
5148 if (!svname) /* shouldnt happen */
5150 "panic: reg_scan_name returned NULL");
5151 if (!RExC_paren_names) {
5152 RExC_paren_names= newHV();
5153 sv_2mortal((SV*)RExC_paren_names);
5155 RExC_paren_name_list= newAV();
5156 sv_2mortal((SV*)RExC_paren_name_list);
5159 he_str = hv_fetch_ent( RExC_paren_names, svname, 1, 0 );
5161 sv_dat = HeVAL(he_str);
5163 /* croak baby croak */
5165 "panic: paren_name hash element allocation failed");
5166 } else if ( SvPOK(sv_dat) ) {
5167 /* (?|...) can mean we have dupes so scan to check
5168 its already been stored. Maybe a flag indicating
5169 we are inside such a construct would be useful,
5170 but the arrays are likely to be quite small, so
5171 for now we punt -- dmq */
5172 IV count = SvIV(sv_dat);
5173 I32 *pv = (I32*)SvPVX(sv_dat);
5175 for ( i = 0 ; i < count ; i++ ) {
5176 if ( pv[i] == RExC_npar ) {
5182 pv = (I32*)SvGROW(sv_dat, SvCUR(sv_dat) + sizeof(I32)+1);
5183 SvCUR_set(sv_dat, SvCUR(sv_dat) + sizeof(I32));
5184 pv[count] = RExC_npar;
5188 (void)SvUPGRADE(sv_dat,SVt_PVNV);
5189 sv_setpvn(sv_dat, (char *)&(RExC_npar), sizeof(I32));
5194 if (!av_store(RExC_paren_name_list, RExC_npar, SvREFCNT_inc(svname)))
5195 SvREFCNT_dec(svname);
5198 /*sv_dump(sv_dat);*/
5200 nextchar(pRExC_state);
5202 goto capturing_parens;
5204 RExC_seen |= REG_SEEN_LOOKBEHIND;
5206 case '=': /* (?=...) */
5207 case '!': /* (?!...) */
5208 RExC_seen_zerolen++;
5209 if (*RExC_parse == ')') {
5210 ret=reg_node(pRExC_state, OPFAIL);
5211 nextchar(pRExC_state);
5215 case '|': /* (?|...) */
5216 /* branch reset, behave like a (?:...) except that
5217 buffers in alternations share the same numbers */
5219 after_freeze = freeze_paren = RExC_npar;
5221 case ':': /* (?:...) */
5222 case '>': /* (?>...) */
5224 case '$': /* (?$...) */
5225 case '@': /* (?@...) */
5226 vFAIL2("Sequence (?%c...) not implemented", (int)paren);
5228 case '#': /* (?#...) */
5229 while (*RExC_parse && *RExC_parse != ')')
5231 if (*RExC_parse != ')')
5232 FAIL("Sequence (?#... not terminated");
5233 nextchar(pRExC_state);
5236 case '0' : /* (?0) */
5237 case 'R' : /* (?R) */
5238 if (*RExC_parse != ')')
5239 FAIL("Sequence (?R) not terminated");
5240 ret = reg_node(pRExC_state, GOSTART);
5241 *flagp |= POSTPONED;
5242 nextchar(pRExC_state);
5245 { /* named and numeric backreferences */
5247 case '&': /* (?&NAME) */
5248 parse_start = RExC_parse - 1;
5251 SV *sv_dat = reg_scan_name(pRExC_state,
5252 SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
5253 num = sv_dat ? *((I32 *)SvPVX(sv_dat)) : 0;
5255 goto gen_recurse_regop;
5258 if (!(RExC_parse[0] >= '1' && RExC_parse[0] <= '9')) {
5260 vFAIL("Illegal pattern");
5262 goto parse_recursion;
5264 case '-': /* (?-1) */
5265 if (!(RExC_parse[0] >= '1' && RExC_parse[0] <= '9')) {
5266 RExC_parse--; /* rewind to let it be handled later */
5270 case '1': case '2': case '3': case '4': /* (?1) */
5271 case '5': case '6': case '7': case '8': case '9':
5274 num = atoi(RExC_parse);
5275 parse_start = RExC_parse - 1; /* MJD */
5276 if (*RExC_parse == '-')
5278 while (isDIGIT(*RExC_parse))
5280 if (*RExC_parse!=')')
5281 vFAIL("Expecting close bracket");
5284 if ( paren == '-' ) {
5286 Diagram of capture buffer numbering.
5287 Top line is the normal capture buffer numbers
5288 Botton line is the negative indexing as from
5292 /(a(x)y)(a(b(c(?-2)d)e)f)(g(h))/
5296 num = RExC_npar + num;
5299 vFAIL("Reference to nonexistent group");
5301 } else if ( paren == '+' ) {
5302 num = RExC_npar + num - 1;
5305 ret = reganode(pRExC_state, GOSUB, num);
5307 if (num > (I32)RExC_rx->nparens) {
5309 vFAIL("Reference to nonexistent group");
5311 ARG2L_SET( ret, RExC_recurse_count++);
5313 DEBUG_OPTIMISE_MORE_r(PerlIO_printf(Perl_debug_log,
5314 "Recurse #%"UVuf" to %"IVdf"\n", (UV)ARG(ret), (IV)ARG2L(ret)));
5318 RExC_seen |= REG_SEEN_RECURSE;
5319 Set_Node_Length(ret, 1 + regarglen[OP(ret)]); /* MJD */
5320 Set_Node_Offset(ret, parse_start); /* MJD */
5322 *flagp |= POSTPONED;
5323 nextchar(pRExC_state);
5325 } /* named and numeric backreferences */
5328 case 'p': /* (?p...) */
5329 if (SIZE_ONLY && ckWARN2(WARN_DEPRECATED, WARN_REGEXP))
5330 vWARNdep(RExC_parse, "(?p{}) is deprecated - use (??{})");
5332 case '?': /* (??...) */
5334 if (*RExC_parse != '{') {
5336 vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
5339 *flagp |= POSTPONED;
5340 paren = *RExC_parse++;
5342 case '{': /* (?{...}) */
5347 char *s = RExC_parse;
5349 RExC_seen_zerolen++;
5350 RExC_seen |= REG_SEEN_EVAL;
5351 while (count && (c = *RExC_parse)) {
5362 if (*RExC_parse != ')') {
5364 vFAIL("Sequence (?{...}) not terminated or not {}-balanced");
5368 OP_4tree *sop, *rop;
5369 SV * const sv = newSVpvn(s, RExC_parse - 1 - s);
5372 Perl_save_re_context(aTHX);
5373 rop = sv_compile_2op(sv, &sop, "re", &pad);
5374 sop->op_private |= OPpREFCOUNTED;
5375 /* re_dup will OpREFCNT_inc */
5376 OpREFCNT_set(sop, 1);
5379 n = add_data(pRExC_state, 3, "nop");
5380 RExC_rxi->data->data[n] = (void*)rop;
5381 RExC_rxi->data->data[n+1] = (void*)sop;
5382 RExC_rxi->data->data[n+2] = (void*)pad;
5385 else { /* First pass */
5386 if (PL_reginterp_cnt < ++RExC_seen_evals
5388 /* No compiled RE interpolated, has runtime
5389 components ===> unsafe. */
5390 FAIL("Eval-group not allowed at runtime, use re 'eval'");
5391 if (PL_tainting && PL_tainted)
5392 FAIL("Eval-group in insecure regular expression");
5393 #if PERL_VERSION > 8
5394 if (IN_PERL_COMPILETIME)
5399 nextchar(pRExC_state);
5401 ret = reg_node(pRExC_state, LOGICAL);
5404 REGTAIL(pRExC_state, ret, reganode(pRExC_state, EVAL, n));
5405 /* deal with the length of this later - MJD */
5408 ret = reganode(pRExC_state, EVAL, n);
5409 Set_Node_Length(ret, RExC_parse - parse_start + 1);
5410 Set_Node_Offset(ret, parse_start);
5413 case '(': /* (?(?{...})...) and (?(?=...)...) */
5416 if (RExC_parse[0] == '?') { /* (?(?...)) */
5417 if (RExC_parse[1] == '=' || RExC_parse[1] == '!'
5418 || RExC_parse[1] == '<'
5419 || RExC_parse[1] == '{') { /* Lookahead or eval. */
5422 ret = reg_node(pRExC_state, LOGICAL);
5425 REGTAIL(pRExC_state, ret, reg(pRExC_state, 1, &flag,depth+1));
5429 else if ( RExC_parse[0] == '<' /* (?(<NAME>)...) */
5430 || RExC_parse[0] == '\'' ) /* (?('NAME')...) */
5432 char ch = RExC_parse[0] == '<' ? '>' : '\'';
5433 char *name_start= RExC_parse++;
5435 SV *sv_dat=reg_scan_name(pRExC_state,
5436 SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
5437 if (RExC_parse == name_start || *RExC_parse != ch)
5438 vFAIL2("Sequence (?(%c... not terminated",
5439 (ch == '>' ? '<' : ch));
5442 num = add_data( pRExC_state, 1, "S" );
5443 RExC_rxi->data->data[num]=(void*)sv_dat;
5444 SvREFCNT_inc(sv_dat);
5446 ret = reganode(pRExC_state,NGROUPP,num);
5447 goto insert_if_check_paren;
5449 else if (RExC_parse[0] == 'D' &&
5450 RExC_parse[1] == 'E' &&
5451 RExC_parse[2] == 'F' &&
5452 RExC_parse[3] == 'I' &&
5453 RExC_parse[4] == 'N' &&
5454 RExC_parse[5] == 'E')
5456 ret = reganode(pRExC_state,DEFINEP,0);
5459 goto insert_if_check_paren;
5461 else if (RExC_parse[0] == 'R') {
5464 if (RExC_parse[0] >= '1' && RExC_parse[0] <= '9' ) {
5465 parno = atoi(RExC_parse++);
5466 while (isDIGIT(*RExC_parse))
5468 } else if (RExC_parse[0] == '&') {
5471 sv_dat = reg_scan_name(pRExC_state,
5472 SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
5473 parno = sv_dat ? *((I32 *)SvPVX(sv_dat)) : 0;
5475 ret = reganode(pRExC_state,INSUBP,parno);
5476 goto insert_if_check_paren;
5478 else if (RExC_parse[0] >= '1' && RExC_parse[0] <= '9' ) {
5481 parno = atoi(RExC_parse++);
5483 while (isDIGIT(*RExC_parse))
5485 ret = reganode(pRExC_state, GROUPP, parno);
5487 insert_if_check_paren:
5488 if ((c = *nextchar(pRExC_state)) != ')')
5489 vFAIL("Switch condition not recognized");
5491 REGTAIL(pRExC_state, ret, reganode(pRExC_state, IFTHEN, 0));
5492 br = regbranch(pRExC_state, &flags, 1,depth+1);
5494 br = reganode(pRExC_state, LONGJMP, 0);
5496 REGTAIL(pRExC_state, br, reganode(pRExC_state, LONGJMP, 0));
5497 c = *nextchar(pRExC_state);
5502 vFAIL("(?(DEFINE)....) does not allow branches");
5503 lastbr = reganode(pRExC_state, IFTHEN, 0); /* Fake one for optimizer. */
5504 regbranch(pRExC_state, &flags, 1,depth+1);
5505 REGTAIL(pRExC_state, ret, lastbr);
5508 c = *nextchar(pRExC_state);
5513 vFAIL("Switch (?(condition)... contains too many branches");
5514 ender = reg_node(pRExC_state, TAIL);
5515 REGTAIL(pRExC_state, br, ender);
5517 REGTAIL(pRExC_state, lastbr, ender);
5518 REGTAIL(pRExC_state, NEXTOPER(NEXTOPER(lastbr)), ender);
5521 REGTAIL(pRExC_state, ret, ender);
5525 vFAIL2("Unknown switch condition (?(%.2s", RExC_parse);
5529 RExC_parse--; /* for vFAIL to print correctly */
5530 vFAIL("Sequence (? incomplete");
5534 parse_flags: /* (?i) */
5536 U32 posflags = 0, negflags = 0;
5537 U32 *flagsp = &posflags;
5539 while (*RExC_parse) {
5540 /* && strchr("iogcmsx", *RExC_parse) */
5541 /* (?g), (?gc) and (?o) are useless here
5542 and must be globally applied -- japhy */
5543 switch (*RExC_parse) {
5544 CASE_STD_PMMOD_FLAGS_PARSE_SET(flagsp);
5547 if (SIZE_ONLY && ckWARN(WARN_REGEXP)) {
5548 const I32 wflagbit = *RExC_parse == 'o' ? WASTED_O : WASTED_G;
5549 if (! (wastedflags & wflagbit) ) {
5550 wastedflags |= wflagbit;
5553 "Useless (%s%c) - %suse /%c modifier",
5554 flagsp == &negflags ? "?-" : "?",
5556 flagsp == &negflags ? "don't " : "",
5564 if (SIZE_ONLY && ckWARN(WARN_REGEXP)) {
5565 if (! (wastedflags & WASTED_C) ) {
5566 wastedflags |= WASTED_GC;
5569 "Useless (%sc) - %suse /gc modifier",
5570 flagsp == &negflags ? "?-" : "?",
5571 flagsp == &negflags ? "don't " : ""
5577 if (flagsp == &negflags) {
5578 if (SIZE_ONLY && ckWARN(WARN_REGEXP))
5579 vWARN(RExC_parse + 1,"Useless use of (?-k)");
5581 *flagsp |= RXf_PMf_KEEPCOPY;
5585 if (flagsp == &negflags) {
5587 vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
5591 wastedflags = 0; /* reset so (?g-c) warns twice */
5597 RExC_flags |= posflags;
5598 RExC_flags &= ~negflags;
5599 nextchar(pRExC_state);
5610 vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
5615 }} /* one for the default block, one for the switch */
5622 ret = reganode(pRExC_state, OPEN, parno);
5625 RExC_nestroot = parno;
5626 if (RExC_seen & REG_SEEN_RECURSE
5627 && !RExC_open_parens[parno-1])
5629 DEBUG_OPTIMISE_MORE_r(PerlIO_printf(Perl_debug_log,
5630 "Setting open paren #%"IVdf" to %d\n",
5631 (IV)parno, REG_NODE_NUM(ret)));
5632 RExC_open_parens[parno-1]= ret;
5635 Set_Node_Length(ret, 1); /* MJD */
5636 Set_Node_Offset(ret, RExC_parse); /* MJD */
5644 /* Pick up the branches, linking them together. */
5645 parse_start = RExC_parse; /* MJD */
5646 br = regbranch(pRExC_state, &flags, 1,depth+1);
5647 /* branch_len = (paren != 0); */
5651 if (*RExC_parse == '|') {
5652 if (!SIZE_ONLY && RExC_extralen) {
5653 reginsert(pRExC_state, BRANCHJ, br, depth+1);
5656 reginsert(pRExC_state, BRANCH, br, depth+1);
5657 Set_Node_Length(br, paren != 0);
5658 Set_Node_Offset_To_R(br-RExC_emit_start, parse_start-RExC_start);
5662 RExC_extralen += 1; /* For BRANCHJ-BRANCH. */
5664 else if (paren == ':') {
5665 *flagp |= flags&SIMPLE;
5667 if (is_open) { /* Starts with OPEN. */
5668 REGTAIL(pRExC_state, ret, br); /* OPEN -> first. */
5670 else if (paren != '?') /* Not Conditional */
5672 *flagp |= flags & (SPSTART | HASWIDTH | POSTPONED);
5674 while (*RExC_parse == '|') {
5675 if (!SIZE_ONLY && RExC_extralen) {
5676 ender = reganode(pRExC_state, LONGJMP,0);
5677 REGTAIL(pRExC_state, NEXTOPER(NEXTOPER(lastbr)), ender); /* Append to the previous. */
5680 RExC_extralen += 2; /* Account for LONGJMP. */
5681 nextchar(pRExC_state);
5683 if (RExC_npar > after_freeze)
5684 after_freeze = RExC_npar;
5685 RExC_npar = freeze_paren;
5687 br = regbranch(pRExC_state, &flags, 0, depth+1);
5691 REGTAIL(pRExC_state, lastbr, br); /* BRANCH -> BRANCH. */
5693 *flagp |= flags & (SPSTART | HASWIDTH | POSTPONED);
5696 if (have_branch || paren != ':') {
5697 /* Make a closing node, and hook it on the end. */
5700 ender = reg_node(pRExC_state, TAIL);
5703 ender = reganode(pRExC_state, CLOSE, parno);
5704 if (!SIZE_ONLY && RExC_seen & REG_SEEN_RECURSE) {
5705 DEBUG_OPTIMISE_MORE_r(PerlIO_printf(Perl_debug_log,
5706 "Setting close paren #%"IVdf" to %d\n",
5707 (IV)parno, REG_NODE_NUM(ender)));
5708 RExC_close_parens[parno-1]= ender;
5709 if (RExC_nestroot == parno)
5712 Set_Node_Offset(ender,RExC_parse+1); /* MJD */
5713 Set_Node_Length(ender,1); /* MJD */
5719 *flagp &= ~HASWIDTH;
5722 ender = reg_node(pRExC_state, SUCCEED);
5725 ender = reg_node(pRExC_state, END);
5727 assert(!RExC_opend); /* there can only be one! */
5732 REGTAIL(pRExC_state, lastbr, ender);
5734 if (have_branch && !SIZE_ONLY) {
5736 RExC_seen |= REG_TOP_LEVEL_BRANCHES;
5738 /* Hook the tails of the branches to the closing node. */
5739 for (br = ret; br; br = regnext(br)) {
5740 const U8 op = PL_regkind[OP(br)];
5742 REGTAIL_STUDY(pRExC_state, NEXTOPER(br), ender);
5744 else if (op == BRANCHJ) {
5745 REGTAIL_STUDY(pRExC_state, NEXTOPER(NEXTOPER(br)), ender);
5753 static const char parens[] = "=!<,>";
5755 if (paren && (p = strchr(parens, paren))) {
5756 U8 node = ((p - parens) % 2) ? UNLESSM : IFMATCH;
5757 int flag = (p - parens) > 1;
5760 node = SUSPEND, flag = 0;
5761 reginsert(pRExC_state, node,ret, depth+1);
5762 Set_Node_Cur_Length(ret);
5763 Set_Node_Offset(ret, parse_start + 1);
5765 REGTAIL_STUDY(pRExC_state, ret, reg_node(pRExC_state, TAIL));
5769 /* Check for proper termination. */
5771 RExC_flags = oregflags;
5772 if (RExC_parse >= RExC_end || *nextchar(pRExC_state) != ')') {
5773 RExC_parse = oregcomp_parse;
5774 vFAIL("Unmatched (");
5777 else if (!paren && RExC_parse < RExC_end) {
5778 if (*RExC_parse == ')') {
5780 vFAIL("Unmatched )");
5783 FAIL("Junk on end of regexp"); /* "Can't happen". */
5787 RExC_npar = after_freeze;
5792 - regbranch - one alternative of an | operator
5794 * Implements the concatenation operator.
5797 S_regbranch(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, I32 first, U32 depth)
5800 register regnode *ret;
5801 register regnode *chain = NULL;
5802 register regnode *latest;
5803 I32 flags = 0, c = 0;
5804 GET_RE_DEBUG_FLAGS_DECL;
5805 DEBUG_PARSE("brnc");
5809 if (!SIZE_ONLY && RExC_extralen)
5810 ret = reganode(pRExC_state, BRANCHJ,0);
5812 ret = reg_node(pRExC_state, BRANCH);
5813 Set_Node_Length(ret, 1);
5817 if (!first && SIZE_ONLY)
5818 RExC_extralen += 1; /* BRANCHJ */
5820 *flagp = WORST; /* Tentatively. */
5823 nextchar(pRExC_state);
5824 while (RExC_parse < RExC_end && *RExC_parse != '|' && *RExC_parse != ')') {
5826 latest = regpiece(pRExC_state, &flags,depth+1);
5827 if (latest == NULL) {
5828 if (flags & TRYAGAIN)
5832 else if (ret == NULL)
5834 *flagp |= flags&(HASWIDTH|POSTPONED);
5835 if (chain == NULL) /* First piece. */
5836 *flagp |= flags&SPSTART;
5839 REGTAIL(pRExC_state, chain, latest);
5844 if (chain == NULL) { /* Loop ran zero times. */
5845 chain = reg_node(pRExC_state, NOTHING);
5850 *flagp |= flags&SIMPLE;
5857 - regpiece - something followed by possible [*+?]
5859 * Note that the branching code sequences used for ? and the general cases
5860 * of * and + are somewhat optimized: they use the same NOTHING node as
5861 * both the endmarker for their branch list and the body of the last branch.
5862 * It might seem that this node could be dispensed with entirely, but the
5863 * endmarker role is not redundant.
5866 S_regpiece(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, U32 depth)
5869 register regnode *ret;
5871 register char *next;
5873 const char * const origparse = RExC_parse;
5875 I32 max = REG_INFTY;
5877 const char *maxpos = NULL;
5878 GET_RE_DEBUG_FLAGS_DECL;
5879 DEBUG_PARSE("piec");
5881 ret = regatom(pRExC_state, &flags,depth+1);
5883 if (flags & TRYAGAIN)
5890 if (op == '{' && regcurly(RExC_parse)) {
5892 parse_start = RExC_parse; /* MJD */
5893 next = RExC_parse + 1;
5894 while (isDIGIT(*next) || *next == ',') {
5903 if (*next == '}') { /* got one */
5907 min = atoi(RExC_parse);
5911 maxpos = RExC_parse;
5913 if (!max && *maxpos != '0')
5914 max = REG_INFTY; /* meaning "infinity" */
5915 else if (max >= REG_INFTY)
5916 vFAIL2("Quantifier in {,} bigger than %d", REG_INFTY - 1);
5918 nextchar(pRExC_state);
5921 if ((flags&SIMPLE)) {
5922 RExC_naughty += 2 + RExC_naughty / 2;
5923 reginsert(pRExC_state, CURLY, ret, depth+1);
5924 Set_Node_Offset(ret, parse_start+1); /* MJD */
5925 Set_Node_Cur_Length(ret);
5928 regnode * const w = reg_node(pRExC_state, WHILEM);
5931 REGTAIL(pRExC_state, ret, w);
5932 if (!SIZE_ONLY && RExC_extralen) {
5933 reginsert(pRExC_state, LONGJMP,ret, depth+1);
5934 reginsert(pRExC_state, NOTHING,ret, depth+1);
5935 NEXT_OFF(ret) = 3; /* Go over LONGJMP. */
5937 reginsert(pRExC_state, CURLYX,ret, depth+1);
5939 Set_Node_Offset(ret, parse_start+1);
5940 Set_Node_Length(ret,
5941 op == '{' ? (RExC_parse - parse_start) : 1);
5943 if (!SIZE_ONLY && RExC_extralen)
5944 NEXT_OFF(ret) = 3; /* Go over NOTHING to LONGJMP. */
5945 REGTAIL(pRExC_state, ret, reg_node(pRExC_state, NOTHING));
5947 RExC_whilem_seen++, RExC_extralen += 3;
5948 RExC_naughty += 4 + RExC_naughty; /* compound interest */
5956 if (max && max < min)
5957 vFAIL("Can't do {n,m} with n > m");
5959 ARG1_SET(ret, (U16)min);
5960 ARG2_SET(ret, (U16)max);
5972 #if 0 /* Now runtime fix should be reliable. */
5974 /* if this is reinstated, don't forget to put this back into perldiag:
5976 =item Regexp *+ operand could be empty at {#} in regex m/%s/
5978 (F) The part of the regexp subject to either the * or + quantifier
5979 could match an empty string. The {#} shows in the regular
5980 expression about where the problem was discovered.
5984 if (!(flags&HASWIDTH) && op != '?')
5985 vFAIL("Regexp *+ operand could be empty");
5988 parse_start = RExC_parse;
5989 nextchar(pRExC_state);
5991 *flagp = (op != '+') ? (WORST|SPSTART|HASWIDTH) : (WORST|HASWIDTH);
5993 if (op == '*' && (flags&SIMPLE)) {
5994 reginsert(pRExC_state, STAR, ret, depth+1);
5998 else if (op == '*') {
6002 else if (op == '+' && (flags&SIMPLE)) {
6003 reginsert(pRExC_state, PLUS, ret, depth+1);
6007 else if (op == '+') {
6011 else if (op == '?') {
6016 if (!SIZE_ONLY && !(flags&(HASWIDTH|POSTPONED)) && max > REG_INFTY/3 && ckWARN(WARN_REGEXP)) {
6018 "%.*s matches null string many times",
6019 (int)(RExC_parse >= origparse ? RExC_parse - origparse : 0),
6023 if (RExC_parse < RExC_end && *RExC_parse == '?') {
6024 nextchar(pRExC_state);
6025 reginsert(pRExC_state, MINMOD, ret, depth+1);
6026 REGTAIL(pRExC_state, ret, ret + NODE_STEP_REGNODE);
6028 #ifndef REG_ALLOW_MINMOD_SUSPEND
6031 if (RExC_parse < RExC_end && *RExC_parse == '+') {
6033 nextchar(pRExC_state);
6034 ender = reg_node(pRExC_state, SUCCEED);
6035 REGTAIL(pRExC_state, ret, ender);
6036 reginsert(pRExC_state, SUSPEND, ret, depth+1);
6038 ender = reg_node(pRExC_state, TAIL);
6039 REGTAIL(pRExC_state, ret, ender);
6043 if (RExC_parse < RExC_end && ISMULT2(RExC_parse)) {
6045 vFAIL("Nested quantifiers");
6052 /* reg_namedseq(pRExC_state,UVp)
6054 This is expected to be called by a parser routine that has
6055 recognized'\N' and needs to handle the rest. RExC_parse is
6056 expected to point at the first char following the N at the time
6059 If valuep is non-null then it is assumed that we are parsing inside
6060 of a charclass definition and the first codepoint in the resolved
6061 string is returned via *valuep and the routine will return NULL.
6062 In this mode if a multichar string is returned from the charnames
6063 handler a warning will be issued, and only the first char in the
6064 sequence will be examined. If the string returned is zero length
6065 then the value of *valuep is undefined and NON-NULL will
6066 be returned to indicate failure. (This will NOT be a valid pointer
6069 If value is null then it is assumed that we are parsing normal text
6070 and inserts a new EXACT node into the program containing the resolved
6071 string and returns a pointer to the new node. If the string is
6072 zerolength a NOTHING node is emitted.
6074 On success RExC_parse is set to the char following the endbrace.
6075 Parsing failures will generate a fatal errorvia vFAIL(...)
6077 NOTE: We cache all results from the charnames handler locally in
6078 the RExC_charnames hash (created on first use) to prevent a charnames
6079 handler from playing silly-buggers and returning a short string and
6080 then a long string for a given pattern. Since the regexp program
6081 size is calculated during an initial parse this would result
6082 in a buffer overrun so we cache to prevent the charname result from
6083 changing during the course of the parse.
6087 S_reg_namedseq(pTHX_ RExC_state_t *pRExC_state, UV *valuep)
6089 char * name; /* start of the content of the name */
6090 char * endbrace; /* endbrace following the name */
6093 STRLEN len; /* this has various purposes throughout the code */
6094 bool cached = 0; /* if this is true then we shouldn't refcount dev sv_str */
6095 regnode *ret = NULL;
6097 if (*RExC_parse != '{') {
6098 vFAIL("Missing braces on \\N{}");
6100 name = RExC_parse+1;
6101 endbrace = strchr(RExC_parse, '}');
6104 vFAIL("Missing right brace on \\N{}");
6106 RExC_parse = endbrace + 1;
6109 /* RExC_parse points at the beginning brace,
6110 endbrace points at the last */
6111 if ( name[0]=='U' && name[1]=='+' ) {
6112 /* its a "unicode hex" notation {U+89AB} */
6113 I32 fl = PERL_SCAN_ALLOW_UNDERSCORES
6114 | PERL_SCAN_DISALLOW_PREFIX
6115 | (SIZE_ONLY ? PERL_SCAN_SILENT_ILLDIGIT : 0);
6117 len = (STRLEN)(endbrace - name - 2);
6118 cp = grok_hex(name + 2, &len, &fl, NULL);
6119 if ( len != (STRLEN)(endbrace - name - 2) ) {
6128 sv_str= Perl_newSVpvf_nocontext("%c",(int)cp);
6130 /* fetch the charnames handler for this scope */
6131 HV * const table = GvHV(PL_hintgv);
6133 hv_fetchs(table, "charnames", FALSE) :
6135 SV *cv= cvp ? *cvp : NULL;
6138 /* create an SV with the name as argument */
6139 sv_name = newSVpvn(name, endbrace - name);
6141 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
6142 vFAIL2("Constant(\\N{%s}) unknown: "
6143 "(possibly a missing \"use charnames ...\")",
6146 if (!cvp || !SvOK(*cvp)) { /* when $^H{charnames} = undef; */
6147 vFAIL2("Constant(\\N{%s}): "
6148 "$^H{charnames} is not defined",SvPVX(sv_name));
6153 if (!RExC_charnames) {
6154 /* make sure our cache is allocated */
6155 RExC_charnames = newHV();
6156 sv_2mortal((SV*)RExC_charnames);
6158 /* see if we have looked this one up before */
6159 he_str = hv_fetch_ent( RExC_charnames, sv_name, 0, 0 );
6161 sv_str = HeVAL(he_str);
6174 count= call_sv(cv, G_SCALAR);
6176 if (count == 1) { /* XXXX is this right? dmq */
6178 SvREFCNT_inc_simple_void(sv_str);
6186 if ( !sv_str || !SvOK(sv_str) ) {
6187 vFAIL2("Constant(\\N{%s}): Call to &{$^H{charnames}} "
6188 "did not return a defined value",SvPVX(sv_name));
6190 if (hv_store_ent( RExC_charnames, sv_name, sv_str, 0))
6195 char *p = SvPV(sv_str, len);
6198 if ( SvUTF8(sv_str) ) {
6199 *valuep = utf8_to_uvchr((U8*)p, &numlen);
6203 We have to turn on utf8 for high bit chars otherwise
6204 we get failures with
6206 "ss" =~ /[\N{LATIN SMALL LETTER SHARP S}]/i
6207 "SS" =~ /[\N{LATIN SMALL LETTER SHARP S}]/i
6209 This is different from what \x{} would do with the same
6210 codepoint, where the condition is > 0xFF.
6217 /* warn if we havent used the whole string? */
6219 if (numlen<len && SIZE_ONLY && ckWARN(WARN_REGEXP)) {
6221 "Ignoring excess chars from \\N{%s} in character class",
6225 } else if (SIZE_ONLY && ckWARN(WARN_REGEXP)) {
6227 "Ignoring zero length \\N{%s} in character class",
6232 SvREFCNT_dec(sv_name);
6234 SvREFCNT_dec(sv_str);
6235 return len ? NULL : (regnode *)&len;
6236 } else if(SvCUR(sv_str)) {
6242 char * parse_start = name-3; /* needed for the offsets */
6244 GET_RE_DEBUG_FLAGS_DECL; /* needed for the offsets */
6246 ret = reg_node(pRExC_state,
6247 (U8)(FOLD ? (LOC ? EXACTFL : EXACTF) : EXACT));
6250 if ( RExC_utf8 && !SvUTF8(sv_str) ) {
6251 sv_utf8_upgrade(sv_str);
6252 } else if ( !RExC_utf8 && SvUTF8(sv_str) ) {
6256 p = SvPV(sv_str, len);
6258 /* len is the length written, charlen is the size the char read */
6259 for ( len = 0; p < pend; p += charlen ) {
6261 UV uvc = utf8_to_uvchr((U8*)p, &charlen);
6263 STRLEN foldlen,numlen;
6264 U8 tmpbuf[UTF8_MAXBYTES_CASE+1], *foldbuf;
6265 uvc = toFOLD_uni(uvc, tmpbuf, &foldlen);
6266 /* Emit all the Unicode characters. */
6268 for (foldbuf = tmpbuf;
6272 uvc = utf8_to_uvchr(foldbuf, &numlen);
6274 const STRLEN unilen = reguni(pRExC_state, uvc, s);
6277 /* In EBCDIC the numlen
6278 * and unilen can differ. */
6280 if (numlen >= foldlen)
6284 break; /* "Can't happen." */
6287 const STRLEN unilen = reguni(pRExC_state, uvc, s);
6299 RExC_size += STR_SZ(len);
6302 RExC_emit += STR_SZ(len);
6304 Set_Node_Cur_Length(ret); /* MJD */
6306 nextchar(pRExC_state);
6308 ret = reg_node(pRExC_state,NOTHING);
6311 SvREFCNT_dec(sv_str);
6314 SvREFCNT_dec(sv_name);
6324 * It returns the code point in utf8 for the value in *encp.
6325 * value: a code value in the source encoding
6326 * encp: a pointer to an Encode object
6328 * If the result from Encode is not a single character,
6329 * it returns U+FFFD (Replacement character) and sets *encp to NULL.
6332 S_reg_recode(pTHX_ const char value, SV **encp)
6335 SV * const sv = sv_2mortal(newSVpvn(&value, numlen));
6336 const char * const s = encp && *encp ? sv_recode_to_utf8(sv, *encp)
6338 const STRLEN newlen = SvCUR(sv);
6339 UV uv = UNICODE_REPLACEMENT;
6343 ? utf8n_to_uvchr((U8*)s, newlen, &numlen, UTF8_ALLOW_DEFAULT)
6346 if (!newlen || numlen != newlen) {
6347 uv = UNICODE_REPLACEMENT;
6356 - regatom - the lowest level
6358 Try to identify anything special at the start of the pattern. If there
6359 is, then handle it as required. This may involve generating a single regop,
6360 such as for an assertion; or it may involve recursing, such as to
6361 handle a () structure.
6363 If the string doesn't start with something special then we gobble up
6364 as much literal text as we can.
6366 Once we have been able to handle whatever type of thing started the
6367 sequence, we return.
6369 Note: we have to be careful with escapes, as they can be both literal
6370 and special, and in the case of \10 and friends can either, depending
6371 on context. Specifically there are two seperate switches for handling
6372 escape sequences, with the one for handling literal escapes requiring
6373 a dummy entry for all of the special escapes that are actually handled
6378 S_regatom(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, U32 depth)
6381 register regnode *ret = NULL;
6383 char *parse_start = RExC_parse;
6384 GET_RE_DEBUG_FLAGS_DECL;
6385 DEBUG_PARSE("atom");
6386 *flagp = WORST; /* Tentatively. */
6390 switch (*RExC_parse) {
6392 RExC_seen_zerolen++;
6393 nextchar(pRExC_state);
6394 if (RExC_flags & RXf_PMf_MULTILINE)
6395 ret = reg_node(pRExC_state, MBOL);
6396 else if (RExC_flags & RXf_PMf_SINGLELINE)
6397 ret = reg_node(pRExC_state, SBOL);
6399 ret = reg_node(pRExC_state, BOL);
6400 Set_Node_Length(ret, 1); /* MJD */
6403 nextchar(pRExC_state);
6405 RExC_seen_zerolen++;
6406 if (RExC_flags & RXf_PMf_MULTILINE)
6407 ret = reg_node(pRExC_state, MEOL);
6408 else if (RExC_flags & RXf_PMf_SINGLELINE)
6409 ret = reg_node(pRExC_state, SEOL);
6411 ret = reg_node(pRExC_state, EOL);
6412 Set_Node_Length(ret, 1); /* MJD */
6415 nextchar(pRExC_state);
6416 if (RExC_flags & RXf_PMf_SINGLELINE)
6417 ret = reg_node(pRExC_state, SANY);
6419 ret = reg_node(pRExC_state, REG_ANY);
6420 *flagp |= HASWIDTH|SIMPLE;
6422 Set_Node_Length(ret, 1); /* MJD */
6426 char * const oregcomp_parse = ++RExC_parse;
6427 ret = regclass(pRExC_state,depth+1);
6428 if (*RExC_parse != ']') {
6429 RExC_parse = oregcomp_parse;
6430 vFAIL("Unmatched [");
6432 nextchar(pRExC_state);
6433 *flagp |= HASWIDTH|SIMPLE;
6434 Set_Node_Length(ret, RExC_parse - oregcomp_parse + 1); /* MJD */
6438 nextchar(pRExC_state);
6439 ret = reg(pRExC_state, 1, &flags,depth+1);
6441 if (flags & TRYAGAIN) {
6442 if (RExC_parse == RExC_end) {
6443 /* Make parent create an empty node if needed. */
6451 *flagp |= flags&(HASWIDTH|SPSTART|SIMPLE|POSTPONED);
6455 if (flags & TRYAGAIN) {
6459 vFAIL("Internal urp");
6460 /* Supposed to be caught earlier. */
6463 if (!regcurly(RExC_parse)) {
6472 vFAIL("Quantifier follows nothing");
6477 This switch handles escape sequences that resolve to some kind
6478 of special regop and not to literal text. Escape sequnces that
6479 resolve to literal text are handled below in the switch marked
6482 Every entry in this switch *must* have a corresponding entry
6483 in the literal escape switch. However, the opposite is not
6484 required, as the default for this switch is to jump to the
6485 literal text handling code.
6487 switch (*++RExC_parse) {
6488 /* Special Escapes */
6490 RExC_seen_zerolen++;
6491 ret = reg_node(pRExC_state, SBOL);
6493 goto finish_meta_pat;
6495 ret = reg_node(pRExC_state, GPOS);
6496 RExC_seen |= REG_SEEN_GPOS;
6498 goto finish_meta_pat;
6500 RExC_seen_zerolen++;
6501 ret = reg_node(pRExC_state, KEEPS);
6503 goto finish_meta_pat;
6505 ret = reg_node(pRExC_state, SEOL);
6507 RExC_seen_zerolen++; /* Do not optimize RE away */
6508 goto finish_meta_pat;
6510 ret = reg_node(pRExC_state, EOS);
6512 RExC_seen_zerolen++; /* Do not optimize RE away */
6513 goto finish_meta_pat;
6515 ret = reg_node(pRExC_state, CANY);
6516 RExC_seen |= REG_SEEN_CANY;
6517 *flagp |= HASWIDTH|SIMPLE;
6518 goto finish_meta_pat;
6520 ret = reg_node(pRExC_state, CLUMP);
6522 goto finish_meta_pat;
6524 ret = reg_node(pRExC_state, (U8)(LOC ? ALNUML : ALNUM));
6525 *flagp |= HASWIDTH|SIMPLE;
6526 goto finish_meta_pat;
6528 ret = reg_node(pRExC_state, (U8)(LOC ? NALNUML : NALNUM));
6529 *flagp |= HASWIDTH|SIMPLE;
6530 goto finish_meta_pat;
6532 RExC_seen_zerolen++;
6533 RExC_seen |= REG_SEEN_LOOKBEHIND;
6534 ret = reg_node(pRExC_state, (U8)(LOC ? BOUNDL : BOUND));
6536 goto finish_meta_pat;
6538 RExC_seen_zerolen++;
6539 RExC_seen |= REG_SEEN_LOOKBEHIND;
6540 ret = reg_node(pRExC_state, (U8)(LOC ? NBOUNDL : NBOUND));
6542 goto finish_meta_pat;
6544 ret = reg_node(pRExC_state, (U8)(LOC ? SPACEL : SPACE));
6545 *flagp |= HASWIDTH|SIMPLE;
6546 goto finish_meta_pat;
6548 ret = reg_node(pRExC_state, (U8)(LOC ? NSPACEL : NSPACE));
6549 *flagp |= HASWIDTH|SIMPLE;
6550 goto finish_meta_pat;
6552 ret = reg_node(pRExC_state, DIGIT);
6553 *flagp |= HASWIDTH|SIMPLE;
6554 goto finish_meta_pat;
6556 ret = reg_node(pRExC_state, NDIGIT);
6557 *flagp |= HASWIDTH|SIMPLE;
6558 goto finish_meta_pat;
6560 ret = reganode(pRExC_state, PRUNE, 0);
6563 goto finish_meta_pat;
6565 ret = reganode(pRExC_state, SKIP, 0);
6569 nextchar(pRExC_state);
6570 Set_Node_Length(ret, 2); /* MJD */
6575 char* const oldregxend = RExC_end;
6577 char* parse_start = RExC_parse - 2;
6580 if (RExC_parse[1] == '{') {
6581 /* a lovely hack--pretend we saw [\pX] instead */
6582 RExC_end = strchr(RExC_parse, '}');
6584 const U8 c = (U8)*RExC_parse;
6586 RExC_end = oldregxend;
6587 vFAIL2("Missing right brace on \\%c{}", c);
6592 RExC_end = RExC_parse + 2;
6593 if (RExC_end > oldregxend)
6594 RExC_end = oldregxend;
6598 ret = regclass(pRExC_state,depth+1);
6600 RExC_end = oldregxend;
6603 Set_Node_Offset(ret, parse_start + 2);
6604 Set_Node_Cur_Length(ret);
6605 nextchar(pRExC_state);
6606 *flagp |= HASWIDTH|SIMPLE;
6610 /* Handle \N{NAME} here and not below because it can be
6611 multicharacter. join_exact() will join them up later on.
6612 Also this makes sure that things like /\N{BLAH}+/ and
6613 \N{BLAH} being multi char Just Happen. dmq*/
6615 ret= reg_namedseq(pRExC_state, NULL);
6617 case 'k': /* Handle \k<NAME> and \k'NAME' */
6620 char ch= RExC_parse[1];
6621 if (ch != '<' && ch != '\'' && ch != '{') {
6623 vFAIL2("Sequence %.2s... not terminated",parse_start);
6625 /* this pretty much dupes the code for (?P=...) in reg(), if
6626 you change this make sure you change that */
6627 char* name_start = (RExC_parse += 2);
6629 SV *sv_dat = reg_scan_name(pRExC_state,
6630 SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
6631 ch= (ch == '<') ? '>' : (ch == '{') ? '}' : '\'';
6632 if (RExC_parse == name_start || *RExC_parse != ch)
6633 vFAIL2("Sequence %.3s... not terminated",parse_start);
6636 num = add_data( pRExC_state, 1, "S" );
6637 RExC_rxi->data->data[num]=(void*)sv_dat;
6638 SvREFCNT_inc(sv_dat);
6642 ret = reganode(pRExC_state,
6643 (U8)(FOLD ? (LOC ? NREFFL : NREFF) : NREF),
6647 /* override incorrect value set in reganode MJD */
6648 Set_Node_Offset(ret, parse_start+1);
6649 Set_Node_Cur_Length(ret); /* MJD */
6650 nextchar(pRExC_state);
6656 case '1': case '2': case '3': case '4':
6657 case '5': case '6': case '7': case '8': case '9':
6660 bool isg = *RExC_parse == 'g';
6665 if (*RExC_parse == '{') {
6669 if (*RExC_parse == '-') {
6673 if (hasbrace && !isDIGIT(*RExC_parse)) {
6674 if (isrel) RExC_parse--;
6676 goto parse_named_seq;
6678 num = atoi(RExC_parse);
6680 num = RExC_npar - num;
6682 vFAIL("Reference to nonexistent or unclosed group");
6684 if (!isg && num > 9 && num >= RExC_npar)
6687 char * const parse_start = RExC_parse - 1; /* MJD */
6688 while (isDIGIT(*RExC_parse))
6690 if (parse_start == RExC_parse - 1)
6691 vFAIL("Unterminated \\g... pattern");
6693 if (*RExC_parse != '}')
6694 vFAIL("Unterminated \\g{...} pattern");
6698 if (num > (I32)RExC_rx->nparens)
6699 vFAIL("Reference to nonexistent group");
6702 ret = reganode(pRExC_state,
6703 (U8)(FOLD ? (LOC ? REFFL : REFF) : REF),
6707 /* override incorrect value set in reganode MJD */
6708 Set_Node_Offset(ret, parse_start+1);
6709 Set_Node_Cur_Length(ret); /* MJD */
6711 nextchar(pRExC_state);
6716 if (RExC_parse >= RExC_end)
6717 FAIL("Trailing \\");
6720 /* Do not generate "unrecognized" warnings here, we fall
6721 back into the quick-grab loop below */
6728 if (RExC_flags & RXf_PMf_EXTENDED) {
6729 if ( reg_skipcomment( pRExC_state ) )
6735 register STRLEN len;
6740 U8 tmpbuf[UTF8_MAXBYTES_CASE+1], *foldbuf;
6742 parse_start = RExC_parse - 1;
6748 ret = reg_node(pRExC_state,
6749 (U8)(FOLD ? (LOC ? EXACTFL : EXACTF) : EXACT));
6751 for (len = 0, p = RExC_parse - 1;
6752 len < 127 && p < RExC_end;
6755 char * const oldp = p;
6757 if (RExC_flags & RXf_PMf_EXTENDED)
6758 p = regwhite( pRExC_state, p );
6769 /* Literal Escapes Switch
6771 This switch is meant to handle escape sequences that
6772 resolve to a literal character.
6774 Every escape sequence that represents something
6775 else, like an assertion or a char class, is handled
6776 in the switch marked 'Special Escapes' above in this
6777 routine, but also has an entry here as anything that
6778 isn't explicitly mentioned here will be treated as
6779 an unescaped equivalent literal.
6783 /* These are all the special escapes. */
6784 case 'A': /* Start assertion */
6785 case 'b': case 'B': /* Word-boundary assertion*/
6786 case 'C': /* Single char !DANGEROUS! */
6787 case 'd': case 'D': /* digit class */
6788 case 'g': case 'G': /* generic-backref, pos assertion */
6789 case 'k': case 'K': /* named backref, keep marker */
6790 case 'N': /* named char sequence */
6791 case 'p': case 'P': /* unicode property */
6792 case 's': case 'S': /* space class */
6793 case 'v': case 'V': /* (*PRUNE) and (*SKIP) */
6794 case 'w': case 'W': /* word class */
6795 case 'X': /* eXtended Unicode "combining character sequence" */
6796 case 'z': case 'Z': /* End of line/string assertion */
6800 /* Anything after here is an escape that resolves to a
6801 literal. (Except digits, which may or may not)
6820 ender = ASCII_TO_NATIVE('\033');
6824 ender = ASCII_TO_NATIVE('\007');
6829 char* const e = strchr(p, '}');
6833 vFAIL("Missing right brace on \\x{}");
6836 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
6837 | PERL_SCAN_DISALLOW_PREFIX;
6838 STRLEN numlen = e - p - 1;
6839 ender = grok_hex(p + 1, &numlen, &flags, NULL);
6846 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
6848 ender = grok_hex(p, &numlen, &flags, NULL);
6851 if (PL_encoding && ender < 0x100)
6852 goto recode_encoding;
6856 ender = UCHARAT(p++);
6857 ender = toCTRL(ender);
6859 case '0': case '1': case '2': case '3':case '4':
6860 case '5': case '6': case '7': case '8':case '9':
6862 (isDIGIT(p[1]) && atoi(p) >= RExC_npar) ) {
6865 ender = grok_oct(p, &numlen, &flags, NULL);
6872 if (PL_encoding && ender < 0x100)
6873 goto recode_encoding;
6877 SV* enc = PL_encoding;
6878 ender = reg_recode((const char)(U8)ender, &enc);
6879 if (!enc && SIZE_ONLY && ckWARN(WARN_REGEXP))
6880 vWARN(p, "Invalid escape in the specified encoding");
6886 FAIL("Trailing \\");
6889 if (!SIZE_ONLY&& isALPHA(*p) && ckWARN(WARN_REGEXP))
6890 vWARN2(p + 1, "Unrecognized escape \\%c passed through", UCHARAT(p));
6891 goto normal_default;
6896 if (UTF8_IS_START(*p) && UTF) {
6898 ender = utf8n_to_uvchr((U8*)p, RExC_end - p,
6899 &numlen, UTF8_ALLOW_DEFAULT);
6906 if ( RExC_flags & RXf_PMf_EXTENDED)
6907 p = regwhite( pRExC_state, p );
6909 /* Prime the casefolded buffer. */
6910 ender = toFOLD_uni(ender, tmpbuf, &foldlen);
6912 if (p < RExC_end && ISMULT2(p)) { /* Back off on ?+*. */
6917 /* Emit all the Unicode characters. */
6919 for (foldbuf = tmpbuf;
6921 foldlen -= numlen) {
6922 ender = utf8_to_uvchr(foldbuf, &numlen);
6924 const STRLEN unilen = reguni(pRExC_state, ender, s);
6927 /* In EBCDIC the numlen
6928 * and unilen can differ. */
6930 if (numlen >= foldlen)
6934 break; /* "Can't happen." */
6938 const STRLEN unilen = reguni(pRExC_state, ender, s);
6947 REGC((char)ender, s++);
6953 /* Emit all the Unicode characters. */
6955 for (foldbuf = tmpbuf;
6957 foldlen -= numlen) {
6958 ender = utf8_to_uvchr(foldbuf, &numlen);
6960 const STRLEN unilen = reguni(pRExC_state, ender, s);
6963 /* In EBCDIC the numlen
6964 * and unilen can differ. */
6966 if (numlen >= foldlen)
6974 const STRLEN unilen = reguni(pRExC_state, ender, s);
6983 REGC((char)ender, s++);
6987 Set_Node_Cur_Length(ret); /* MJD */
6988 nextchar(pRExC_state);
6990 /* len is STRLEN which is unsigned, need to copy to signed */
6993 vFAIL("Internal disaster");
6997 if (len == 1 && UNI_IS_INVARIANT(ender))
7001 RExC_size += STR_SZ(len);
7004 RExC_emit += STR_SZ(len);
7014 S_regwhite( RExC_state_t *pRExC_state, char *p )
7016 const char *e = RExC_end;
7020 else if (*p == '#') {
7029 RExC_seen |= REG_SEEN_RUN_ON_COMMENT;
7037 /* Parse POSIX character classes: [[:foo:]], [[=foo=]], [[.foo.]].
7038 Character classes ([:foo:]) can also be negated ([:^foo:]).
7039 Returns a named class id (ANYOF_XXX) if successful, -1 otherwise.
7040 Equivalence classes ([=foo=]) and composites ([.foo.]) are parsed,
7041 but trigger failures because they are currently unimplemented. */
7043 #define POSIXCC_DONE(c) ((c) == ':')
7044 #define POSIXCC_NOTYET(c) ((c) == '=' || (c) == '.')
7045 #define POSIXCC(c) (POSIXCC_DONE(c) || POSIXCC_NOTYET(c))
7048 S_regpposixcc(pTHX_ RExC_state_t *pRExC_state, I32 value)
7051 I32 namedclass = OOB_NAMEDCLASS;
7053 if (value == '[' && RExC_parse + 1 < RExC_end &&
7054 /* I smell either [: or [= or [. -- POSIX has been here, right? */
7055 POSIXCC(UCHARAT(RExC_parse))) {
7056 const char c = UCHARAT(RExC_parse);
7057 char* const s = RExC_parse++;
7059 while (RExC_parse < RExC_end && UCHARAT(RExC_parse) != c)
7061 if (RExC_parse == RExC_end)
7062 /* Grandfather lone [:, [=, [. */
7065 const char* const t = RExC_parse++; /* skip over the c */
7068 if (UCHARAT(RExC_parse) == ']') {
7069 const char *posixcc = s + 1;
7070 RExC_parse++; /* skip over the ending ] */
7073 const I32 complement = *posixcc == '^' ? *posixcc++ : 0;
7074 const I32 skip = t - posixcc;
7076 /* Initially switch on the length of the name. */
7079 if (memEQ(posixcc, "word", 4)) /* this is not POSIX, this is the Perl \w */
7080 namedclass = complement ? ANYOF_NALNUM : ANYOF_ALNUM;
7083 /* Names all of length 5. */
7084 /* alnum alpha ascii blank cntrl digit graph lower
7085 print punct space upper */
7086 /* Offset 4 gives the best switch position. */
7087 switch (posixcc[4]) {
7089 if (memEQ(posixcc, "alph", 4)) /* alpha */
7090 namedclass = complement ? ANYOF_NALPHA : ANYOF_ALPHA;
7093 if (memEQ(posixcc, "spac", 4)) /* space */
7094 namedclass = complement ? ANYOF_NPSXSPC : ANYOF_PSXSPC;
7097 if (memEQ(posixcc, "grap", 4)) /* graph */
7098 namedclass = complement ? ANYOF_NGRAPH : ANYOF_GRAPH;
7101 if (memEQ(posixcc, "asci", 4)) /* ascii */
7102 namedclass = complement ? ANYOF_NASCII : ANYOF_ASCII;
7105 if (memEQ(posixcc, "blan", 4)) /* blank */
7106 namedclass = complement ? ANYOF_NBLANK : ANYOF_BLANK;
7109 if (memEQ(posixcc, "cntr", 4)) /* cntrl */
7110 namedclass = complement ? ANYOF_NCNTRL : ANYOF_CNTRL;
7113 if (memEQ(posixcc, "alnu", 4)) /* alnum */
7114 namedclass = complement ? ANYOF_NALNUMC : ANYOF_ALNUMC;
7117 if (memEQ(posixcc, "lowe", 4)) /* lower */
7118 namedclass = complement ? ANYOF_NLOWER : ANYOF_LOWER;
7119 else if (memEQ(posixcc, "uppe", 4)) /* upper */
7120 namedclass = complement ? ANYOF_NUPPER : ANYOF_UPPER;
7123 if (memEQ(posixcc, "digi", 4)) /* digit */
7124 namedclass = complement ? ANYOF_NDIGIT : ANYOF_DIGIT;
7125 else if (memEQ(posixcc, "prin", 4)) /* print */
7126 namedclass = complement ? ANYOF_NPRINT : ANYOF_PRINT;
7127 else if (memEQ(posixcc, "punc", 4)) /* punct */
7128 namedclass = complement ? ANYOF_NPUNCT : ANYOF_PUNCT;
7133 if (memEQ(posixcc, "xdigit", 6))
7134 namedclass = complement ? ANYOF_NXDIGIT : ANYOF_XDIGIT;
7138 if (namedclass == OOB_NAMEDCLASS)
7139 Simple_vFAIL3("POSIX class [:%.*s:] unknown",
7141 assert (posixcc[skip] == ':');
7142 assert (posixcc[skip+1] == ']');
7143 } else if (!SIZE_ONLY) {
7144 /* [[=foo=]] and [[.foo.]] are still future. */
7146 /* adjust RExC_parse so the warning shows after
7148 while (UCHARAT(RExC_parse) && UCHARAT(RExC_parse) != ']')
7150 Simple_vFAIL3("POSIX syntax [%c %c] is reserved for future extensions", c, c);
7153 /* Maternal grandfather:
7154 * "[:" ending in ":" but not in ":]" */
7164 S_checkposixcc(pTHX_ RExC_state_t *pRExC_state)
7167 if (POSIXCC(UCHARAT(RExC_parse))) {
7168 const char *s = RExC_parse;
7169 const char c = *s++;
7173 if (*s && c == *s && s[1] == ']') {
7174 if (ckWARN(WARN_REGEXP))
7176 "POSIX syntax [%c %c] belongs inside character classes",
7179 /* [[=foo=]] and [[.foo.]] are still future. */
7180 if (POSIXCC_NOTYET(c)) {
7181 /* adjust RExC_parse so the error shows after
7183 while (UCHARAT(RExC_parse) && UCHARAT(RExC_parse++) != ']')
7185 Simple_vFAIL3("POSIX syntax [%c %c] is reserved for future extensions", c, c);
7192 #define _C_C_T_(NAME,TEST,WORD) \
7195 ANYOF_CLASS_SET(ret, ANYOF_##NAME); \
7197 for (value = 0; value < 256; value++) \
7199 ANYOF_BITMAP_SET(ret, value); \
7204 case ANYOF_N##NAME: \
7206 ANYOF_CLASS_SET(ret, ANYOF_N##NAME); \
7208 for (value = 0; value < 256; value++) \
7210 ANYOF_BITMAP_SET(ret, value); \
7218 parse a class specification and produce either an ANYOF node that
7219 matches the pattern or if the pattern matches a single char only and
7220 that char is < 256 and we are case insensitive then we produce an
7225 S_regclass(pTHX_ RExC_state_t *pRExC_state, U32 depth)
7228 register UV value = 0;
7229 register UV nextvalue;
7230 register IV prevvalue = OOB_UNICODE;
7231 register IV range = 0;
7232 register regnode *ret;
7235 char *rangebegin = NULL;
7236 bool need_class = 0;
7239 bool optimize_invert = TRUE;
7240 AV* unicode_alternate = NULL;
7242 UV literal_endpoint = 0;
7244 UV stored = 0; /* number of chars stored in the class */
7246 regnode * const orig_emit = RExC_emit; /* Save the original RExC_emit in
7247 case we need to change the emitted regop to an EXACT. */
7248 const char * orig_parse = RExC_parse;
7249 GET_RE_DEBUG_FLAGS_DECL;
7251 PERL_UNUSED_ARG(depth);
7254 DEBUG_PARSE("clas");
7256 /* Assume we are going to generate an ANYOF node. */
7257 ret = reganode(pRExC_state, ANYOF, 0);
7260 ANYOF_FLAGS(ret) = 0;
7262 if (UCHARAT(RExC_parse) == '^') { /* Complement of range. */
7266 ANYOF_FLAGS(ret) |= ANYOF_INVERT;
7270 RExC_size += ANYOF_SKIP;
7271 listsv = &PL_sv_undef; /* For code scanners: listsv always non-NULL. */
7274 RExC_emit += ANYOF_SKIP;
7276 ANYOF_FLAGS(ret) |= ANYOF_FOLD;
7278 ANYOF_FLAGS(ret) |= ANYOF_LOCALE;
7279 ANYOF_BITMAP_ZERO(ret);
7280 listsv = newSVpvs("# comment\n");
7283 nextvalue = RExC_parse < RExC_end ? UCHARAT(RExC_parse) : 0;
7285 if (!SIZE_ONLY && POSIXCC(nextvalue))
7286 checkposixcc(pRExC_state);
7288 /* allow 1st char to be ] (allowing it to be - is dealt with later) */
7289 if (UCHARAT(RExC_parse) == ']')
7293 while (RExC_parse < RExC_end && UCHARAT(RExC_parse) != ']') {
7297 namedclass = OOB_NAMEDCLASS; /* initialize as illegal */
7300 rangebegin = RExC_parse;
7302 value = utf8n_to_uvchr((U8*)RExC_parse,
7303 RExC_end - RExC_parse,
7304 &numlen, UTF8_ALLOW_DEFAULT);
7305 RExC_parse += numlen;
7308 value = UCHARAT(RExC_parse++);
7310 nextvalue = RExC_parse < RExC_end ? UCHARAT(RExC_parse) : 0;
7311 if (value == '[' && POSIXCC(nextvalue))
7312 namedclass = regpposixcc(pRExC_state, value);
7313 else if (value == '\\') {
7315 value = utf8n_to_uvchr((U8*)RExC_parse,
7316 RExC_end - RExC_parse,
7317 &numlen, UTF8_ALLOW_DEFAULT);
7318 RExC_parse += numlen;
7321 value = UCHARAT(RExC_parse++);
7322 /* Some compilers cannot handle switching on 64-bit integer
7323 * values, therefore value cannot be an UV. Yes, this will
7324 * be a problem later if we want switch on Unicode.
7325 * A similar issue a little bit later when switching on
7326 * namedclass. --jhi */
7327 switch ((I32)value) {
7328 case 'w': namedclass = ANYOF_ALNUM; break;
7329 case 'W': namedclass = ANYOF_NALNUM; break;
7330 case 's': namedclass = ANYOF_SPACE; break;
7331 case 'S': namedclass = ANYOF_NSPACE; break;
7332 case 'd': namedclass = ANYOF_DIGIT; break;
7333 case 'D': namedclass = ANYOF_NDIGIT; break;
7334 case 'N': /* Handle \N{NAME} in class */
7336 /* We only pay attention to the first char of
7337 multichar strings being returned. I kinda wonder
7338 if this makes sense as it does change the behaviour
7339 from earlier versions, OTOH that behaviour was broken
7341 UV v; /* value is register so we cant & it /grrr */
7342 if (reg_namedseq(pRExC_state, &v)) {
7352 if (RExC_parse >= RExC_end)
7353 vFAIL2("Empty \\%c{}", (U8)value);
7354 if (*RExC_parse == '{') {
7355 const U8 c = (U8)value;
7356 e = strchr(RExC_parse++, '}');
7358 vFAIL2("Missing right brace on \\%c{}", c);
7359 while (isSPACE(UCHARAT(RExC_parse)))
7361 if (e == RExC_parse)
7362 vFAIL2("Empty \\%c{}", c);
7364 while (isSPACE(UCHARAT(RExC_parse + n - 1)))
7372 if (UCHARAT(RExC_parse) == '^') {
7375 value = value == 'p' ? 'P' : 'p'; /* toggle */
7376 while (isSPACE(UCHARAT(RExC_parse))) {
7381 Perl_sv_catpvf(aTHX_ listsv, "%cutf8::%.*s\n",
7382 (value=='p' ? '+' : '!'), (int)n, RExC_parse);
7385 ANYOF_FLAGS(ret) |= ANYOF_UNICODE;
7386 namedclass = ANYOF_MAX; /* no official name, but it's named */
7389 case 'n': value = '\n'; break;
7390 case 'r': value = '\r'; break;
7391 case 't': value = '\t'; break;
7392 case 'f': value = '\f'; break;
7393 case 'b': value = '\b'; break;
7394 case 'e': value = ASCII_TO_NATIVE('\033');break;
7395 case 'a': value = ASCII_TO_NATIVE('\007');break;
7397 if (*RExC_parse == '{') {
7398 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
7399 | PERL_SCAN_DISALLOW_PREFIX;
7400 char * const e = strchr(RExC_parse++, '}');
7402 vFAIL("Missing right brace on \\x{}");
7404 numlen = e - RExC_parse;
7405 value = grok_hex(RExC_parse, &numlen, &flags, NULL);
7409 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
7411 value = grok_hex(RExC_parse, &numlen, &flags, NULL);
7412 RExC_parse += numlen;
7414 if (PL_encoding && value < 0x100)
7415 goto recode_encoding;
7418 value = UCHARAT(RExC_parse++);
7419 value = toCTRL(value);
7421 case '0': case '1': case '2': case '3': case '4':
7422 case '5': case '6': case '7': case '8': case '9':
7426 value = grok_oct(--RExC_parse, &numlen, &flags, NULL);
7427 RExC_parse += numlen;
7428 if (PL_encoding && value < 0x100)
7429 goto recode_encoding;
7434 SV* enc = PL_encoding;
7435 value = reg_recode((const char)(U8)value, &enc);
7436 if (!enc && SIZE_ONLY && ckWARN(WARN_REGEXP))
7438 "Invalid escape in the specified encoding");
7442 if (!SIZE_ONLY && isALPHA(value) && ckWARN(WARN_REGEXP))
7444 "Unrecognized escape \\%c in character class passed through",
7448 } /* end of \blah */
7454 if (namedclass > OOB_NAMEDCLASS) { /* this is a named class \blah */
7456 if (!SIZE_ONLY && !need_class)
7457 ANYOF_CLASS_ZERO(ret);
7461 /* a bad range like a-\d, a-[:digit:] ? */
7464 if (ckWARN(WARN_REGEXP)) {
7466 RExC_parse >= rangebegin ?
7467 RExC_parse - rangebegin : 0;
7469 "False [] range \"%*.*s\"",
7472 if (prevvalue < 256) {
7473 ANYOF_BITMAP_SET(ret, prevvalue);
7474 ANYOF_BITMAP_SET(ret, '-');
7477 ANYOF_FLAGS(ret) |= ANYOF_UNICODE;
7478 Perl_sv_catpvf(aTHX_ listsv,
7479 "%04"UVxf"\n%04"UVxf"\n", (UV)prevvalue, (UV) '-');
7483 range = 0; /* this was not a true range */
7489 const char *what = NULL;
7492 if (namedclass > OOB_NAMEDCLASS)
7493 optimize_invert = FALSE;
7494 /* Possible truncation here but in some 64-bit environments
7495 * the compiler gets heartburn about switch on 64-bit values.
7496 * A similar issue a little earlier when switching on value.
7498 switch ((I32)namedclass) {
7499 case _C_C_T_(ALNUM, isALNUM(value), "Word");
7500 case _C_C_T_(ALNUMC, isALNUMC(value), "Alnum");
7501 case _C_C_T_(ALPHA, isALPHA(value), "Alpha");
7502 case _C_C_T_(BLANK, isBLANK(value), "Blank");
7503 case _C_C_T_(CNTRL, isCNTRL(value), "Cntrl");
7504 case _C_C_T_(GRAPH, isGRAPH(value), "Graph");
7505 case _C_C_T_(LOWER, isLOWER(value), "Lower");
7506 case _C_C_T_(PRINT, isPRINT(value), "Print");
7507 case _C_C_T_(PSXSPC, isPSXSPC(value), "Space");
7508 case _C_C_T_(PUNCT, isPUNCT(value), "Punct");
7509 case _C_C_T_(SPACE, isSPACE(value), "SpacePerl");
7510 case _C_C_T_(UPPER, isUPPER(value), "Upper");
7511 case _C_C_T_(XDIGIT, isXDIGIT(value), "XDigit");
7514 ANYOF_CLASS_SET(ret, ANYOF_ASCII);
7517 for (value = 0; value < 128; value++)
7518 ANYOF_BITMAP_SET(ret, value);
7520 for (value = 0; value < 256; value++) {
7522 ANYOF_BITMAP_SET(ret, value);
7531 ANYOF_CLASS_SET(ret, ANYOF_NASCII);
7534 for (value = 128; value < 256; value++)
7535 ANYOF_BITMAP_SET(ret, value);
7537 for (value = 0; value < 256; value++) {
7538 if (!isASCII(value))
7539 ANYOF_BITMAP_SET(ret, value);
7548 ANYOF_CLASS_SET(ret, ANYOF_DIGIT);
7550 /* consecutive digits assumed */
7551 for (value = '0'; value <= '9'; value++)
7552 ANYOF_BITMAP_SET(ret, value);
7559 ANYOF_CLASS_SET(ret, ANYOF_NDIGIT);
7561 /* consecutive digits assumed */
7562 for (value = 0; value < '0'; value++)
7563 ANYOF_BITMAP_SET(ret, value);
7564 for (value = '9' + 1; value < 256; value++)
7565 ANYOF_BITMAP_SET(ret, value);
7571 /* this is to handle \p and \P */
7574 vFAIL("Invalid [::] class");
7578 /* Strings such as "+utf8::isWord\n" */
7579 Perl_sv_catpvf(aTHX_ listsv, "%cutf8::Is%s\n", yesno, what);
7582 ANYOF_FLAGS(ret) |= ANYOF_CLASS;
7585 } /* end of namedclass \blah */
7588 if (prevvalue > (IV)value) /* b-a */ {
7589 const int w = RExC_parse - rangebegin;
7590 Simple_vFAIL4("Invalid [] range \"%*.*s\"", w, w, rangebegin);
7591 range = 0; /* not a valid range */
7595 prevvalue = value; /* save the beginning of the range */
7596 if (*RExC_parse == '-' && RExC_parse+1 < RExC_end &&
7597 RExC_parse[1] != ']') {
7600 /* a bad range like \w-, [:word:]- ? */
7601 if (namedclass > OOB_NAMEDCLASS) {
7602 if (ckWARN(WARN_REGEXP)) {
7604 RExC_parse >= rangebegin ?
7605 RExC_parse - rangebegin : 0;
7607 "False [] range \"%*.*s\"",
7611 ANYOF_BITMAP_SET(ret, '-');
7613 range = 1; /* yeah, it's a range! */
7614 continue; /* but do it the next time */
7618 /* now is the next time */
7619 /*stored += (value - prevvalue + 1);*/
7621 if (prevvalue < 256) {
7622 const IV ceilvalue = value < 256 ? value : 255;
7625 /* In EBCDIC [\x89-\x91] should include
7626 * the \x8e but [i-j] should not. */
7627 if (literal_endpoint == 2 &&
7628 ((isLOWER(prevvalue) && isLOWER(ceilvalue)) ||
7629 (isUPPER(prevvalue) && isUPPER(ceilvalue))))
7631 if (isLOWER(prevvalue)) {
7632 for (i = prevvalue; i <= ceilvalue; i++)
7634 ANYOF_BITMAP_SET(ret, i);
7636 for (i = prevvalue; i <= ceilvalue; i++)
7638 ANYOF_BITMAP_SET(ret, i);
7643 for (i = prevvalue; i <= ceilvalue; i++) {
7644 if (!ANYOF_BITMAP_TEST(ret,i)) {
7646 ANYOF_BITMAP_SET(ret, i);
7650 if (value > 255 || UTF) {
7651 const UV prevnatvalue = NATIVE_TO_UNI(prevvalue);
7652 const UV natvalue = NATIVE_TO_UNI(value);
7653 stored+=2; /* can't optimize this class */
7654 ANYOF_FLAGS(ret) |= ANYOF_UNICODE;
7655 if (prevnatvalue < natvalue) { /* what about > ? */
7656 Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\t%04"UVxf"\n",
7657 prevnatvalue, natvalue);
7659 else if (prevnatvalue == natvalue) {
7660 Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\n", natvalue);
7662 U8 foldbuf[UTF8_MAXBYTES_CASE+1];
7664 const UV f = to_uni_fold(natvalue, foldbuf, &foldlen);
7666 #ifdef EBCDIC /* RD t/uni/fold ff and 6b */
7667 if (RExC_precomp[0] == ':' &&
7668 RExC_precomp[1] == '[' &&
7669 (f == 0xDF || f == 0x92)) {
7670 f = NATIVE_TO_UNI(f);
7673 /* If folding and foldable and a single
7674 * character, insert also the folded version
7675 * to the charclass. */
7677 #ifdef EBCDIC /* RD tunifold ligatures s,t fb05, fb06 */
7678 if ((RExC_precomp[0] == ':' &&
7679 RExC_precomp[1] == '[' &&
7681 (value == 0xFB05 || value == 0xFB06))) ?
7682 foldlen == ((STRLEN)UNISKIP(f) - 1) :
7683 foldlen == (STRLEN)UNISKIP(f) )
7685 if (foldlen == (STRLEN)UNISKIP(f))
7687 Perl_sv_catpvf(aTHX_ listsv,
7690 /* Any multicharacter foldings
7691 * require the following transform:
7692 * [ABCDEF] -> (?:[ABCabcDEFd]|pq|rst)
7693 * where E folds into "pq" and F folds
7694 * into "rst", all other characters
7695 * fold to single characters. We save
7696 * away these multicharacter foldings,
7697 * to be later saved as part of the
7698 * additional "s" data. */
7701 if (!unicode_alternate)
7702 unicode_alternate = newAV();
7703 sv = newSVpvn((char*)foldbuf, foldlen);
7705 av_push(unicode_alternate, sv);
7709 /* If folding and the value is one of the Greek
7710 * sigmas insert a few more sigmas to make the
7711 * folding rules of the sigmas to work right.
7712 * Note that not all the possible combinations
7713 * are handled here: some of them are handled
7714 * by the standard folding rules, and some of
7715 * them (literal or EXACTF cases) are handled
7716 * during runtime in regexec.c:S_find_byclass(). */
7717 if (value == UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA) {
7718 Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\n",
7719 (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA);
7720 Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\n",
7721 (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA);
7723 else if (value == UNICODE_GREEK_CAPITAL_LETTER_SIGMA)
7724 Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\n",
7725 (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA);
7730 literal_endpoint = 0;
7734 range = 0; /* this range (if it was one) is done now */
7738 ANYOF_FLAGS(ret) |= ANYOF_LARGE;
7740 RExC_size += ANYOF_CLASS_ADD_SKIP;
7742 RExC_emit += ANYOF_CLASS_ADD_SKIP;
7748 /****** !SIZE_ONLY AFTER HERE *********/
7750 if( stored == 1 && value < 256
7751 && !( ANYOF_FLAGS(ret) & ( ANYOF_FLAGS_ALL ^ ANYOF_FOLD ) )
7753 /* optimize single char class to an EXACT node
7754 but *only* when its not a UTF/high char */
7755 const char * cur_parse= RExC_parse;
7756 RExC_emit = (regnode *)orig_emit;
7757 RExC_parse = (char *)orig_parse;
7758 ret = reg_node(pRExC_state,
7759 (U8)((ANYOF_FLAGS(ret) & ANYOF_FOLD) ? EXACTF : EXACT));
7760 RExC_parse = (char *)cur_parse;
7761 *STRING(ret)= (char)value;
7763 RExC_emit += STR_SZ(1);
7766 /* optimize case-insensitive simple patterns (e.g. /[a-z]/i) */
7767 if ( /* If the only flag is folding (plus possibly inversion). */
7768 ((ANYOF_FLAGS(ret) & (ANYOF_FLAGS_ALL ^ ANYOF_INVERT)) == ANYOF_FOLD)
7770 for (value = 0; value < 256; ++value) {
7771 if (ANYOF_BITMAP_TEST(ret, value)) {
7772 UV fold = PL_fold[value];
7775 ANYOF_BITMAP_SET(ret, fold);
7778 ANYOF_FLAGS(ret) &= ~ANYOF_FOLD;
7781 /* optimize inverted simple patterns (e.g. [^a-z]) */
7782 if (optimize_invert &&
7783 /* If the only flag is inversion. */
7784 (ANYOF_FLAGS(ret) & ANYOF_FLAGS_ALL) == ANYOF_INVERT) {
7785 for (value = 0; value < ANYOF_BITMAP_SIZE; ++value)
7786 ANYOF_BITMAP(ret)[value] ^= ANYOF_FLAGS_ALL;
7787 ANYOF_FLAGS(ret) = ANYOF_UNICODE_ALL;
7790 AV * const av = newAV();
7792 /* The 0th element stores the character class description
7793 * in its textual form: used later (regexec.c:Perl_regclass_swash())
7794 * to initialize the appropriate swash (which gets stored in
7795 * the 1st element), and also useful for dumping the regnode.
7796 * The 2nd element stores the multicharacter foldings,
7797 * used later (regexec.c:S_reginclass()). */
7798 av_store(av, 0, listsv);
7799 av_store(av, 1, NULL);
7800 av_store(av, 2, (SV*)unicode_alternate);
7801 rv = newRV_noinc((SV*)av);
7802 n = add_data(pRExC_state, 1, "s");
7803 RExC_rxi->data->data[n] = (void*)rv;
7811 /* reg_skipcomment()
7813 Absorbs an /x style # comments from the input stream.
7814 Returns true if there is more text remaining in the stream.
7815 Will set the REG_SEEN_RUN_ON_COMMENT flag if the comment
7816 terminates the pattern without including a newline.
7818 Note its the callers responsibility to ensure that we are
7824 S_reg_skipcomment(pTHX_ RExC_state_t *pRExC_state)
7827 while (RExC_parse < RExC_end)
7828 if (*RExC_parse++ == '\n') {
7833 /* we ran off the end of the pattern without ending
7834 the comment, so we have to add an \n when wrapping */
7835 RExC_seen |= REG_SEEN_RUN_ON_COMMENT;
7843 Advance that parse position, and optionall absorbs
7844 "whitespace" from the inputstream.
7846 Without /x "whitespace" means (?#...) style comments only,
7847 with /x this means (?#...) and # comments and whitespace proper.
7849 Returns the RExC_parse point from BEFORE the scan occurs.
7851 This is the /x friendly way of saying RExC_parse++.
7855 S_nextchar(pTHX_ RExC_state_t *pRExC_state)
7857 char* const retval = RExC_parse++;
7860 if (*RExC_parse == '(' && RExC_parse[1] == '?' &&
7861 RExC_parse[2] == '#') {
7862 while (*RExC_parse != ')') {
7863 if (RExC_parse == RExC_end)
7864 FAIL("Sequence (?#... not terminated");
7870 if (RExC_flags & RXf_PMf_EXTENDED) {
7871 if (isSPACE(*RExC_parse)) {
7875 else if (*RExC_parse == '#') {
7876 if ( reg_skipcomment( pRExC_state ) )
7885 - reg_node - emit a node
7887 STATIC regnode * /* Location. */
7888 S_reg_node(pTHX_ RExC_state_t *pRExC_state, U8 op)
7891 register regnode *ptr;
7892 regnode * const ret = RExC_emit;
7893 GET_RE_DEBUG_FLAGS_DECL;
7896 SIZE_ALIGN(RExC_size);
7901 if (OP(RExC_emit) == 255)
7902 Perl_croak(aTHX_ "panic: reg_node overrun trying to emit %s: %d ",
7903 reg_name[op], OP(RExC_emit));
7905 NODE_ALIGN_FILL(ret);
7907 FILL_ADVANCE_NODE(ptr, op);
7908 #ifdef RE_TRACK_PATTERN_OFFSETS
7909 if (RExC_offsets) { /* MJD */
7910 MJD_OFFSET_DEBUG(("%s:%d: (op %s) %s %"UVuf" (len %"UVuf") (max %"UVuf").\n",
7911 "reg_node", __LINE__,
7913 (UV)(RExC_emit - RExC_emit_start) > RExC_offsets[0]
7914 ? "Overwriting end of array!\n" : "OK",
7915 (UV)(RExC_emit - RExC_emit_start),
7916 (UV)(RExC_parse - RExC_start),
7917 (UV)RExC_offsets[0]));
7918 Set_Node_Offset(RExC_emit, RExC_parse + (op == END));
7926 - reganode - emit a node with an argument
7928 STATIC regnode * /* Location. */
7929 S_reganode(pTHX_ RExC_state_t *pRExC_state, U8 op, U32 arg)
7932 register regnode *ptr;
7933 regnode * const ret = RExC_emit;
7934 GET_RE_DEBUG_FLAGS_DECL;
7937 SIZE_ALIGN(RExC_size);
7942 assert(2==regarglen[op]+1);
7944 Anything larger than this has to allocate the extra amount.
7945 If we changed this to be:
7947 RExC_size += (1 + regarglen[op]);
7949 then it wouldn't matter. Its not clear what side effect
7950 might come from that so its not done so far.
7956 if (OP(RExC_emit) == 255)
7957 Perl_croak(aTHX_ "panic: reganode overwriting end of allocated program space");
7959 NODE_ALIGN_FILL(ret);
7961 FILL_ADVANCE_NODE_ARG(ptr, op, arg);
7962 #ifdef RE_TRACK_PATTERN_OFFSETS
7963 if (RExC_offsets) { /* MJD */
7964 MJD_OFFSET_DEBUG(("%s(%d): (op %s) %s %"UVuf" <- %"UVuf" (max %"UVuf").\n",
7968 (UV)(RExC_emit - RExC_emit_start) > RExC_offsets[0] ?
7969 "Overwriting end of array!\n" : "OK",
7970 (UV)(RExC_emit - RExC_emit_start),
7971 (UV)(RExC_parse - RExC_start),
7972 (UV)RExC_offsets[0]));
7973 Set_Cur_Node_Offset;
7981 - reguni - emit (if appropriate) a Unicode character
7984 S_reguni(pTHX_ const RExC_state_t *pRExC_state, UV uv, char* s)
7987 return SIZE_ONLY ? UNISKIP(uv) : (uvchr_to_utf8((U8*)s, uv) - (U8*)s);
7991 - reginsert - insert an operator in front of already-emitted operand
7993 * Means relocating the operand.
7996 S_reginsert(pTHX_ RExC_state_t *pRExC_state, U8 op, regnode *opnd, U32 depth)
7999 register regnode *src;
8000 register regnode *dst;
8001 register regnode *place;
8002 const int offset = regarglen[(U8)op];
8003 const int size = NODE_STEP_REGNODE + offset;
8004 GET_RE_DEBUG_FLAGS_DECL;
8005 PERL_UNUSED_ARG(depth);
8006 /* (PL_regkind[(U8)op] == CURLY ? EXTRA_STEP_2ARGS : 0); */
8007 DEBUG_PARSE_FMT("inst"," - %s",reg_name[op]);
8016 if (RExC_open_parens) {
8018 DEBUG_PARSE_FMT("inst"," - %"IVdf, (IV)RExC_npar);
8019 for ( paren=0 ; paren < RExC_npar ; paren++ ) {
8020 if ( RExC_open_parens[paren] >= opnd ) {
8021 DEBUG_PARSE_FMT("open"," - %d",size);
8022 RExC_open_parens[paren] += size;
8024 DEBUG_PARSE_FMT("open"," - %s","ok");
8026 if ( RExC_close_parens[paren] >= opnd ) {
8027 DEBUG_PARSE_FMT("close"," - %d",size);
8028 RExC_close_parens[paren] += size;
8030 DEBUG_PARSE_FMT("close"," - %s","ok");
8035 while (src > opnd) {
8036 StructCopy(--src, --dst, regnode);
8037 #ifdef RE_TRACK_PATTERN_OFFSETS
8038 if (RExC_offsets) { /* MJD 20010112 */
8039 MJD_OFFSET_DEBUG(("%s(%d): (op %s) %s copy %"UVuf" -> %"UVuf" (max %"UVuf").\n",
8043 (UV)(dst - RExC_emit_start) > RExC_offsets[0]
8044 ? "Overwriting end of array!\n" : "OK",
8045 (UV)(src - RExC_emit_start),
8046 (UV)(dst - RExC_emit_start),
8047 (UV)RExC_offsets[0]));
8048 Set_Node_Offset_To_R(dst-RExC_emit_start, Node_Offset(src));
8049 Set_Node_Length_To_R(dst-RExC_emit_start, Node_Length(src));
8055 place = opnd; /* Op node, where operand used to be. */
8056 #ifdef RE_TRACK_PATTERN_OFFSETS
8057 if (RExC_offsets) { /* MJD */
8058 MJD_OFFSET_DEBUG(("%s(%d): (op %s) %s %"UVuf" <- %"UVuf" (max %"UVuf").\n",
8062 (UV)(place - RExC_emit_start) > RExC_offsets[0]
8063 ? "Overwriting end of array!\n" : "OK",
8064 (UV)(place - RExC_emit_start),
8065 (UV)(RExC_parse - RExC_start),
8066 (UV)RExC_offsets[0]));
8067 Set_Node_Offset(place, RExC_parse);
8068 Set_Node_Length(place, 1);
8071 src = NEXTOPER(place);
8072 FILL_ADVANCE_NODE(place, op);
8073 Zero(src, offset, regnode);
8077 - regtail - set the next-pointer at the end of a node chain of p to val.
8078 - SEE ALSO: regtail_study
8080 /* TODO: All three parms should be const */
8082 S_regtail(pTHX_ RExC_state_t *pRExC_state, regnode *p, const regnode *val,U32 depth)
8085 register regnode *scan;
8086 GET_RE_DEBUG_FLAGS_DECL;
8088 PERL_UNUSED_ARG(depth);
8094 /* Find last node. */
8097 regnode * const temp = regnext(scan);
8099 SV * const mysv=sv_newmortal();
8100 DEBUG_PARSE_MSG((scan==p ? "tail" : ""));
8101 regprop(RExC_rx, mysv, scan);
8102 PerlIO_printf(Perl_debug_log, "~ %s (%d) %s %s\n",
8103 SvPV_nolen_const(mysv), REG_NODE_NUM(scan),
8104 (temp == NULL ? "->" : ""),
8105 (temp == NULL ? reg_name[OP(val)] : "")
8113 if (reg_off_by_arg[OP(scan)]) {
8114 ARG_SET(scan, val - scan);
8117 NEXT_OFF(scan) = val - scan;
8123 - regtail_study - set the next-pointer at the end of a node chain of p to val.
8124 - Look for optimizable sequences at the same time.
8125 - currently only looks for EXACT chains.
8127 This is expermental code. The idea is to use this routine to perform
8128 in place optimizations on branches and groups as they are constructed,
8129 with the long term intention of removing optimization from study_chunk so
8130 that it is purely analytical.
8132 Currently only used when in DEBUG mode. The macro REGTAIL_STUDY() is used
8133 to control which is which.
8136 /* TODO: All four parms should be const */
8139 S_regtail_study(pTHX_ RExC_state_t *pRExC_state, regnode *p, const regnode *val,U32 depth)
8142 register regnode *scan;
8144 #ifdef EXPERIMENTAL_INPLACESCAN
8148 GET_RE_DEBUG_FLAGS_DECL;
8154 /* Find last node. */
8158 regnode * const temp = regnext(scan);
8159 #ifdef EXPERIMENTAL_INPLACESCAN
8160 if (PL_regkind[OP(scan)] == EXACT)
8161 if (join_exact(pRExC_state,scan,&min,1,val,depth+1))
8169 if( exact == PSEUDO )
8171 else if ( exact != OP(scan) )
8180 SV * const mysv=sv_newmortal();
8181 DEBUG_PARSE_MSG((scan==p ? "tsdy" : ""));
8182 regprop(RExC_rx, mysv, scan);
8183 PerlIO_printf(Perl_debug_log, "~ %s (%d) -> %s\n",
8184 SvPV_nolen_const(mysv),
8193 SV * const mysv_val=sv_newmortal();
8194 DEBUG_PARSE_MSG("");
8195 regprop(RExC_rx, mysv_val, val);
8196 PerlIO_printf(Perl_debug_log, "~ attach to %s (%"IVdf") offset to %"IVdf"\n",
8197 SvPV_nolen_const(mysv_val),
8198 (IV)REG_NODE_NUM(val),
8202 if (reg_off_by_arg[OP(scan)]) {
8203 ARG_SET(scan, val - scan);
8206 NEXT_OFF(scan) = val - scan;
8214 - regcurly - a little FSA that accepts {\d+,?\d*}
8217 S_regcurly(register const char *s)
8236 - regdump - dump a regexp onto Perl_debug_log in vaguely comprehensible form
8239 Perl_regdump(pTHX_ const regexp *r)
8243 SV * const sv = sv_newmortal();
8244 SV *dsv= sv_newmortal();
8247 (void)dumpuntil(r, ri->program, ri->program + 1, NULL, NULL, sv, 0, 0);
8249 /* Header fields of interest. */
8250 if (r->anchored_substr) {
8251 RE_PV_QUOTED_DECL(s, 0, dsv, SvPVX_const(r->anchored_substr),
8252 RE_SV_DUMPLEN(r->anchored_substr), 30);
8253 PerlIO_printf(Perl_debug_log,
8254 "anchored %s%s at %"IVdf" ",
8255 s, RE_SV_TAIL(r->anchored_substr),
8256 (IV)r->anchored_offset);
8257 } else if (r->anchored_utf8) {
8258 RE_PV_QUOTED_DECL(s, 1, dsv, SvPVX_const(r->anchored_utf8),
8259 RE_SV_DUMPLEN(r->anchored_utf8), 30);
8260 PerlIO_printf(Perl_debug_log,
8261 "anchored utf8 %s%s at %"IVdf" ",
8262 s, RE_SV_TAIL(r->anchored_utf8),
8263 (IV)r->anchored_offset);
8265 if (r->float_substr) {
8266 RE_PV_QUOTED_DECL(s, 0, dsv, SvPVX_const(r->float_substr),
8267 RE_SV_DUMPLEN(r->float_substr), 30);
8268 PerlIO_printf(Perl_debug_log,
8269 "floating %s%s at %"IVdf"..%"UVuf" ",
8270 s, RE_SV_TAIL(r->float_substr),
8271 (IV)r->float_min_offset, (UV)r->float_max_offset);
8272 } else if (r->float_utf8) {
8273 RE_PV_QUOTED_DECL(s, 1, dsv, SvPVX_const(r->float_utf8),
8274 RE_SV_DUMPLEN(r->float_utf8), 30);
8275 PerlIO_printf(Perl_debug_log,
8276 "floating utf8 %s%s at %"IVdf"..%"UVuf" ",
8277 s, RE_SV_TAIL(r->float_utf8),
8278 (IV)r->float_min_offset, (UV)r->float_max_offset);
8280 if (r->check_substr || r->check_utf8)
8281 PerlIO_printf(Perl_debug_log,
8283 (r->check_substr == r->float_substr
8284 && r->check_utf8 == r->float_utf8
8285 ? "(checking floating" : "(checking anchored"));
8286 if (r->extflags & RXf_NOSCAN)
8287 PerlIO_printf(Perl_debug_log, " noscan");
8288 if (r->extflags & RXf_CHECK_ALL)
8289 PerlIO_printf(Perl_debug_log, " isall");
8290 if (r->check_substr || r->check_utf8)
8291 PerlIO_printf(Perl_debug_log, ") ");
8293 if (ri->regstclass) {
8294 regprop(r, sv, ri->regstclass);
8295 PerlIO_printf(Perl_debug_log, "stclass %s ", SvPVX_const(sv));
8297 if (r->extflags & RXf_ANCH) {
8298 PerlIO_printf(Perl_debug_log, "anchored");
8299 if (r->extflags & RXf_ANCH_BOL)
8300 PerlIO_printf(Perl_debug_log, "(BOL)");
8301 if (r->extflags & RXf_ANCH_MBOL)
8302 PerlIO_printf(Perl_debug_log, "(MBOL)");
8303 if (r->extflags & RXf_ANCH_SBOL)
8304 PerlIO_printf(Perl_debug_log, "(SBOL)");
8305 if (r->extflags & RXf_ANCH_GPOS)
8306 PerlIO_printf(Perl_debug_log, "(GPOS)");
8307 PerlIO_putc(Perl_debug_log, ' ');
8309 if (r->extflags & RXf_GPOS_SEEN)
8310 PerlIO_printf(Perl_debug_log, "GPOS:%"UVuf" ", (UV)r->gofs);
8311 if (r->intflags & PREGf_SKIP)
8312 PerlIO_printf(Perl_debug_log, "plus ");
8313 if (r->intflags & PREGf_IMPLICIT)
8314 PerlIO_printf(Perl_debug_log, "implicit ");
8315 PerlIO_printf(Perl_debug_log, "minlen %"IVdf" ", (IV)r->minlen);
8316 if (r->extflags & RXf_EVAL_SEEN)
8317 PerlIO_printf(Perl_debug_log, "with eval ");
8318 PerlIO_printf(Perl_debug_log, "\n");
8320 PERL_UNUSED_CONTEXT;
8322 #endif /* DEBUGGING */
8326 - regprop - printable representation of opcode
8329 Perl_regprop(pTHX_ const regexp *prog, SV *sv, const regnode *o)
8334 RXi_GET_DECL(prog,progi);
8335 GET_RE_DEBUG_FLAGS_DECL;
8338 sv_setpvn(sv, "", 0);
8340 if (OP(o) > REGNODE_MAX) /* regnode.type is unsigned */
8341 /* It would be nice to FAIL() here, but this may be called from
8342 regexec.c, and it would be hard to supply pRExC_state. */
8343 Perl_croak(aTHX_ "Corrupted regexp opcode %d > %d", (int)OP(o), (int)REGNODE_MAX);
8344 sv_catpv(sv, reg_name[OP(o)]); /* Take off const! */
8346 k = PL_regkind[OP(o)];
8349 SV * const dsv = sv_2mortal(newSVpvs(""));
8350 /* Using is_utf8_string() (via PERL_PV_UNI_DETECT)
8351 * is a crude hack but it may be the best for now since
8352 * we have no flag "this EXACTish node was UTF-8"
8354 const char * const s =
8355 pv_pretty(dsv, STRING(o), STR_LEN(o), 60,
8356 PL_colors[0], PL_colors[1],
8357 PERL_PV_ESCAPE_UNI_DETECT |
8358 PERL_PV_PRETTY_ELIPSES |
8361 Perl_sv_catpvf(aTHX_ sv, " %s", s );
8362 } else if (k == TRIE) {
8363 /* print the details of the trie in dumpuntil instead, as
8364 * progi->data isn't available here */
8365 const char op = OP(o);
8366 const U32 n = ARG(o);
8367 const reg_ac_data * const ac = IS_TRIE_AC(op) ?
8368 (reg_ac_data *)progi->data->data[n] :
8370 const reg_trie_data * const trie
8371 = (reg_trie_data*)progi->data->data[!IS_TRIE_AC(op) ? n : ac->trie];
8373 Perl_sv_catpvf(aTHX_ sv, "-%s",reg_name[o->flags]);
8374 DEBUG_TRIE_COMPILE_r(
8375 Perl_sv_catpvf(aTHX_ sv,
8376 "<S:%"UVuf"/%"IVdf" W:%"UVuf" L:%"UVuf"/%"UVuf" C:%"UVuf"/%"UVuf">",
8377 (UV)trie->startstate,
8378 (IV)trie->statecount-1, /* -1 because of the unused 0 element */
8379 (UV)trie->wordcount,
8382 (UV)TRIE_CHARCOUNT(trie),
8383 (UV)trie->uniquecharcount
8386 if ( IS_ANYOF_TRIE(op) || trie->bitmap ) {
8388 int rangestart = -1;
8389 U8* bitmap = IS_ANYOF_TRIE(op) ? (U8*)ANYOF_BITMAP(o) : (U8*)TRIE_BITMAP(trie);
8390 Perl_sv_catpvf(aTHX_ sv, "[");
8391 for (i = 0; i <= 256; i++) {
8392 if (i < 256 && BITMAP_TEST(bitmap,i)) {
8393 if (rangestart == -1)
8395 } else if (rangestart != -1) {
8396 if (i <= rangestart + 3)
8397 for (; rangestart < i; rangestart++)
8398 put_byte(sv, rangestart);
8400 put_byte(sv, rangestart);
8402 put_byte(sv, i - 1);
8407 Perl_sv_catpvf(aTHX_ sv, "]");
8410 } else if (k == CURLY) {
8411 if (OP(o) == CURLYM || OP(o) == CURLYN || OP(o) == CURLYX)
8412 Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* Parenth number */
8413 Perl_sv_catpvf(aTHX_ sv, " {%d,%d}", ARG1(o), ARG2(o));
8415 else if (k == WHILEM && o->flags) /* Ordinal/of */
8416 Perl_sv_catpvf(aTHX_ sv, "[%d/%d]", o->flags & 0xf, o->flags>>4);
8417 else if (k == REF || k == OPEN || k == CLOSE || k == GROUPP || OP(o)==ACCEPT) {
8418 Perl_sv_catpvf(aTHX_ sv, "%d", (int)ARG(o)); /* Parenth number */
8419 if ( prog->paren_names ) {
8420 if ( k != REF || OP(o) < NREF) {
8421 AV *list= (AV *)progi->data->data[progi->name_list_idx];
8422 SV **name= av_fetch(list, ARG(o), 0 );
8424 Perl_sv_catpvf(aTHX_ sv, " '%"SVf"'", SVfARG(*name));
8427 AV *list= (AV *)progi->data->data[ progi->name_list_idx ];
8428 SV *sv_dat=(SV*)progi->data->data[ ARG( o ) ];
8429 I32 *nums=(I32*)SvPVX(sv_dat);
8430 SV **name= av_fetch(list, nums[0], 0 );
8433 for ( n=0; n<SvIVX(sv_dat); n++ ) {
8434 Perl_sv_catpvf(aTHX_ sv, "%s%"IVdf,
8435 (n ? "," : ""), (IV)nums[n]);
8437 Perl_sv_catpvf(aTHX_ sv, " '%"SVf"'", SVfARG(*name));
8441 } else if (k == GOSUB)
8442 Perl_sv_catpvf(aTHX_ sv, "%d[%+d]", (int)ARG(o),(int)ARG2L(o)); /* Paren and offset */
8443 else if (k == VERB) {
8445 Perl_sv_catpvf(aTHX_ sv, ":%"SVf,
8446 SVfARG((SV*)progi->data->data[ ARG( o ) ]));
8447 } else if (k == LOGICAL)
8448 Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* 2: embedded, otherwise 1 */
8449 else if (k == ANYOF) {
8450 int i, rangestart = -1;
8451 const U8 flags = ANYOF_FLAGS(o);
8453 /* Should be synchronized with * ANYOF_ #xdefines in regcomp.h */
8454 static const char * const anyofs[] = {
8487 if (flags & ANYOF_LOCALE)
8488 sv_catpvs(sv, "{loc}");
8489 if (flags & ANYOF_FOLD)
8490 sv_catpvs(sv, "{i}");
8491 Perl_sv_catpvf(aTHX_ sv, "[%s", PL_colors[0]);
8492 if (flags & ANYOF_INVERT)
8494 for (i = 0; i <= 256; i++) {
8495 if (i < 256 && ANYOF_BITMAP_TEST(o,i)) {
8496 if (rangestart == -1)
8498 } else if (rangestart != -1) {
8499 if (i <= rangestart + 3)
8500 for (; rangestart < i; rangestart++)
8501 put_byte(sv, rangestart);
8503 put_byte(sv, rangestart);
8505 put_byte(sv, i - 1);
8511 if (o->flags & ANYOF_CLASS)
8512 for (i = 0; i < (int)(sizeof(anyofs)/sizeof(char*)); i++)
8513 if (ANYOF_CLASS_TEST(o,i))
8514 sv_catpv(sv, anyofs[i]);
8516 if (flags & ANYOF_UNICODE)
8517 sv_catpvs(sv, "{unicode}");
8518 else if (flags & ANYOF_UNICODE_ALL)
8519 sv_catpvs(sv, "{unicode_all}");
8523 SV * const sw = regclass_swash(prog, o, FALSE, &lv, 0);
8527 U8 s[UTF8_MAXBYTES_CASE+1];
8529 for (i = 0; i <= 256; i++) { /* just the first 256 */
8530 uvchr_to_utf8(s, i);
8532 if (i < 256 && swash_fetch(sw, s, TRUE)) {
8533 if (rangestart == -1)
8535 } else if (rangestart != -1) {
8536 if (i <= rangestart + 3)
8537 for (; rangestart < i; rangestart++) {
8538 const U8 * const e = uvchr_to_utf8(s,rangestart);
8540 for(p = s; p < e; p++)
8544 const U8 *e = uvchr_to_utf8(s,rangestart);
8546 for (p = s; p < e; p++)
8549 e = uvchr_to_utf8(s, i-1);
8550 for (p = s; p < e; p++)
8557 sv_catpvs(sv, "..."); /* et cetera */
8561 char *s = savesvpv(lv);
8562 char * const origs = s;
8564 while (*s && *s != '\n')
8568 const char * const t = ++s;
8586 Perl_sv_catpvf(aTHX_ sv, "%s]", PL_colors[1]);
8588 else if (k == BRANCHJ && (OP(o) == UNLESSM || OP(o) == IFMATCH))
8589 Perl_sv_catpvf(aTHX_ sv, "[%d]", -(o->flags));
8591 PERL_UNUSED_CONTEXT;
8592 PERL_UNUSED_ARG(sv);
8594 PERL_UNUSED_ARG(prog);
8595 #endif /* DEBUGGING */
8599 Perl_re_intuit_string(pTHX_ regexp *prog)
8600 { /* Assume that RE_INTUIT is set */
8602 GET_RE_DEBUG_FLAGS_DECL;
8603 PERL_UNUSED_CONTEXT;
8607 const char * const s = SvPV_nolen_const(prog->check_substr
8608 ? prog->check_substr : prog->check_utf8);
8610 if (!PL_colorset) reginitcolors();
8611 PerlIO_printf(Perl_debug_log,
8612 "%sUsing REx %ssubstr:%s \"%s%.60s%s%s\"\n",
8614 prog->check_substr ? "" : "utf8 ",
8615 PL_colors[5],PL_colors[0],
8618 (strlen(s) > 60 ? "..." : ""));
8621 return prog->check_substr ? prog->check_substr : prog->check_utf8;
8627 handles refcounting and freeing the perl core regexp structure. When
8628 it is necessary to actually free the structure the first thing it
8629 does is call the 'free' method of the regexp_engine associated to to
8630 the regexp, allowing the handling of the void *pprivate; member
8631 first. (This routine is not overridable by extensions, which is why
8632 the extensions free is called first.)
8634 See regdupe and regdupe_internal if you change anything here.
8636 #ifndef PERL_IN_XSUB_RE
8638 Perl_pregfree(pTHX_ struct regexp *r)
8641 GET_RE_DEBUG_FLAGS_DECL;
8643 if (!r || (--r->refcnt > 0))
8646 CALLREGFREE_PVT(r); /* free the private data */
8647 RX_MATCH_COPY_FREE(r);
8648 #ifdef PERL_OLD_COPY_ON_WRITE
8650 SvREFCNT_dec(r->saved_copy);
8653 if (r->anchored_substr)
8654 SvREFCNT_dec(r->anchored_substr);
8655 if (r->anchored_utf8)
8656 SvREFCNT_dec(r->anchored_utf8);
8657 if (r->float_substr)
8658 SvREFCNT_dec(r->float_substr);
8660 SvREFCNT_dec(r->float_utf8);
8661 Safefree(r->substrs);
8664 SvREFCNT_dec(r->paren_names);
8665 Safefree(r->wrapped);
8666 Safefree(r->startp);
8672 /* regfree_internal()
8674 Free the private data in a regexp. This is overloadable by
8675 extensions. Perl takes care of the regexp structure in pregfree(),
8676 this covers the *pprivate pointer which technically perldoesnt
8677 know about, however of course we have to handle the
8678 regexp_internal structure when no extension is in use.
8680 Note this is called before freeing anything in the regexp
8685 Perl_regfree_internal(pTHX_ struct regexp *r)
8689 GET_RE_DEBUG_FLAGS_DECL;
8695 SV *dsv= sv_newmortal();
8696 RE_PV_QUOTED_DECL(s, (r->extflags & RXf_UTF8),
8697 dsv, r->precomp, r->prelen, 60);
8698 PerlIO_printf(Perl_debug_log,"%sFreeing REx:%s %s\n",
8699 PL_colors[4],PL_colors[5],s);
8702 #ifdef RE_TRACK_PATTERN_OFFSETS
8704 Safefree(ri->u.offsets); /* 20010421 MJD */
8707 int n = ri->data->count;
8708 PAD* new_comppad = NULL;
8713 /* If you add a ->what type here, update the comment in regcomp.h */
8714 switch (ri->data->what[n]) {
8718 SvREFCNT_dec((SV*)ri->data->data[n]);
8721 Safefree(ri->data->data[n]);
8724 new_comppad = (AV*)ri->data->data[n];
8727 if (new_comppad == NULL)
8728 Perl_croak(aTHX_ "panic: pregfree comppad");
8729 PAD_SAVE_LOCAL(old_comppad,
8730 /* Watch out for global destruction's random ordering. */
8731 (SvTYPE(new_comppad) == SVt_PVAV) ? new_comppad : NULL
8734 refcnt = OpREFCNT_dec((OP_4tree*)ri->data->data[n]);
8737 op_free((OP_4tree*)ri->data->data[n]);
8739 PAD_RESTORE_LOCAL(old_comppad);
8740 SvREFCNT_dec((SV*)new_comppad);
8746 { /* Aho Corasick add-on structure for a trie node.
8747 Used in stclass optimization only */
8749 reg_ac_data *aho=(reg_ac_data*)ri->data->data[n];
8751 refcount = --aho->refcount;
8754 PerlMemShared_free(aho->states);
8755 PerlMemShared_free(aho->fail);
8756 /* do this last!!!! */
8757 PerlMemShared_free(ri->data->data[n]);
8758 PerlMemShared_free(ri->regstclass);
8764 /* trie structure. */
8766 reg_trie_data *trie=(reg_trie_data*)ri->data->data[n];
8768 refcount = --trie->refcount;
8771 PerlMemShared_free(trie->charmap);
8772 PerlMemShared_free(trie->states);
8773 PerlMemShared_free(trie->trans);
8775 PerlMemShared_free(trie->bitmap);
8777 PerlMemShared_free(trie->wordlen);
8779 PerlMemShared_free(trie->jump);
8781 PerlMemShared_free(trie->nextword);
8782 /* do this last!!!! */
8783 PerlMemShared_free(ri->data->data[n]);
8788 Perl_croak(aTHX_ "panic: regfree data code '%c'", ri->data->what[n]);
8791 Safefree(ri->data->what);
8795 Safefree(ri->swap->startp);
8796 Safefree(ri->swap->endp);
8802 #define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
8803 #define av_dup_inc(s,t) (AV*)SvREFCNT_inc(sv_dup((SV*)s,t))
8804 #define hv_dup_inc(s,t) (HV*)SvREFCNT_inc(sv_dup((SV*)s,t))
8805 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
8808 regdupe - duplicate a regexp.
8810 This routine is called by sv.c's re_dup and is expected to clone a
8811 given regexp structure. It is a no-op when not under USE_ITHREADS.
8812 (Originally this *was* re_dup() for change history see sv.c)
8814 After all of the core data stored in struct regexp is duplicated
8815 the regexp_engine.dupe method is used to copy any private data
8816 stored in the *pprivate pointer. This allows extensions to handle
8817 any duplication it needs to do.
8819 See pregfree() and regfree_internal() if you change anything here.
8821 #if defined(USE_ITHREADS)
8822 #ifndef PERL_IN_XSUB_RE
8824 Perl_re_dup(pTHX_ const regexp *r, CLONE_PARAMS *param)
8829 struct reg_substr_datum *s;
8832 return (REGEXP *)NULL;
8834 if ((ret = (REGEXP *)ptr_table_fetch(PL_ptr_table, r)))
8838 npar = r->nparens+1;
8839 Newxz(ret, 1, regexp);
8840 Newx(ret->startp, npar, I32);
8841 Copy(r->startp, ret->startp, npar, I32);
8842 Newx(ret->endp, npar, I32);
8843 Copy(r->endp, ret->endp, npar, I32);
8846 Newx(ret->substrs, 1, struct reg_substr_data);
8847 for (s = ret->substrs->data, i = 0; i < 3; i++, s++) {
8848 s->min_offset = r->substrs->data[i].min_offset;
8849 s->max_offset = r->substrs->data[i].max_offset;
8850 s->end_shift = r->substrs->data[i].end_shift;
8851 s->substr = sv_dup_inc(r->substrs->data[i].substr, param);
8852 s->utf8_substr = sv_dup_inc(r->substrs->data[i].utf8_substr, param);
8855 ret->substrs = NULL;
8857 ret->wrapped = SAVEPVN(r->wrapped, r->wraplen);
8858 ret->precomp = ret->wrapped + (r->precomp - r->wrapped);
8859 ret->prelen = r->prelen;
8860 ret->wraplen = r->wraplen;
8862 ret->refcnt = r->refcnt;
8863 ret->minlen = r->minlen;
8864 ret->minlenret = r->minlenret;
8865 ret->nparens = r->nparens;
8866 ret->lastparen = r->lastparen;
8867 ret->lastcloseparen = r->lastcloseparen;
8868 ret->intflags = r->intflags;
8869 ret->extflags = r->extflags;
8871 ret->sublen = r->sublen;
8873 ret->engine = r->engine;
8875 ret->paren_names = hv_dup_inc(r->paren_names, param);
8877 if (RX_MATCH_COPIED(ret))
8878 ret->subbeg = SAVEPVN(r->subbeg, r->sublen);
8881 #ifdef PERL_OLD_COPY_ON_WRITE
8882 ret->saved_copy = NULL;
8885 ret->pprivate = r->pprivate;
8887 RXi_SET(ret,CALLREGDUPE_PVT(ret,param));
8889 ptr_table_store(PL_ptr_table, r, ret);
8892 #endif /* PERL_IN_XSUB_RE */
8897 This is the internal complement to regdupe() which is used to copy
8898 the structure pointed to by the *pprivate pointer in the regexp.
8899 This is the core version of the extension overridable cloning hook.
8900 The regexp structure being duplicated will be copied by perl prior
8901 to this and will be provided as the regexp *r argument, however
8902 with the /old/ structures pprivate pointer value. Thus this routine
8903 may override any copying normally done by perl.
8905 It returns a pointer to the new regexp_internal structure.
8909 Perl_regdupe_internal(pTHX_ const regexp *r, CLONE_PARAMS *param)
8912 regexp_internal *reti;
8916 npar = r->nparens+1;
8919 Newxc(reti, sizeof(regexp_internal) + (len+1)*sizeof(regnode), char, regexp_internal);
8920 Copy(ri->program, reti->program, len+1, regnode);
8923 Newx(reti->swap, 1, regexp_paren_ofs);
8924 /* no need to copy these */
8925 Newx(reti->swap->startp, npar, I32);
8926 Newx(reti->swap->endp, npar, I32);
8931 reti->regstclass = NULL;
8935 const int count = ri->data->count;
8938 Newxc(d, sizeof(struct reg_data) + count*sizeof(void *),
8939 char, struct reg_data);
8940 Newx(d->what, count, U8);
8943 for (i = 0; i < count; i++) {
8944 d->what[i] = ri->data->what[i];
8945 switch (d->what[i]) {
8946 /* legal options are one of: sSfpontTu
8947 see also regcomp.h and pregfree() */
8950 case 'p': /* actually an AV, but the dup function is identical. */
8951 case 'u': /* actually an HV, but the dup function is identical. */
8952 d->data[i] = sv_dup_inc((SV *)ri->data->data[i], param);
8955 /* This is cheating. */
8956 Newx(d->data[i], 1, struct regnode_charclass_class);
8957 StructCopy(ri->data->data[i], d->data[i],
8958 struct regnode_charclass_class);
8959 reti->regstclass = (regnode*)d->data[i];
8962 /* Compiled op trees are readonly and in shared memory,
8963 and can thus be shared without duplication. */
8965 d->data[i] = (void*)OpREFCNT_inc((OP*)ri->data->data[i]);
8969 /* Trie stclasses are readonly and can thus be shared
8970 * without duplication. We free the stclass in pregfree
8971 * when the corresponding reg_ac_data struct is freed.
8973 reti->regstclass= ri->regstclass;
8977 ((reg_trie_data*)ri->data->data[i])->refcount++;
8981 d->data[i] = ri->data->data[i];
8984 Perl_croak(aTHX_ "panic: re_dup unknown data code '%c'", ri->data->what[i]);
8993 reti->name_list_idx = ri->name_list_idx;
8995 #ifdef RE_TRACK_PATTERN_OFFSETS
8996 if (ri->u.offsets) {
8997 Newx(reti->u.offsets, 2*len+1, U32);
8998 Copy(ri->u.offsets, reti->u.offsets, 2*len+1, U32);
9001 SetProgLen(reti,len);
9007 #endif /* USE_ITHREADS */
9012 converts a regexp embedded in a MAGIC struct to its stringified form,
9013 caching the converted form in the struct and returns the cached
9016 If lp is nonnull then it is used to return the length of the
9019 If flags is nonnull and the returned string contains UTF8 then
9020 (*flags & 1) will be true.
9022 If haseval is nonnull then it is used to return whether the pattern
9025 Normally called via macro:
9027 CALLREG_STRINGIFY(mg,&len,&utf8);
9031 CALLREG_AS_STR(mg,&lp,&flags,&haseval)
9033 See sv_2pv_flags() in sv.c for an example of internal usage.
9036 #ifndef PERL_IN_XSUB_RE
9039 Perl_reg_stringify(pTHX_ MAGIC *mg, STRLEN *lp, U32 *flags, I32 *haseval ) {
9041 const regexp * const re = (regexp *)mg->mg_obj;
9043 *haseval = re->seen_evals;
9045 *flags = ((re->extflags & RXf_UTF8) ? 1 : 0);
9052 - regnext - dig the "next" pointer out of a node
9055 Perl_regnext(pTHX_ register regnode *p)
9058 register I32 offset;
9063 offset = (reg_off_by_arg[OP(p)] ? ARG(p) : NEXT_OFF(p));
9072 S_re_croak2(pTHX_ const char* pat1,const char* pat2,...)
9075 STRLEN l1 = strlen(pat1);
9076 STRLEN l2 = strlen(pat2);
9079 const char *message;
9085 Copy(pat1, buf, l1 , char);
9086 Copy(pat2, buf + l1, l2 , char);
9087 buf[l1 + l2] = '\n';
9088 buf[l1 + l2 + 1] = '\0';
9090 /* ANSI variant takes additional second argument */
9091 va_start(args, pat2);
9095 msv = vmess(buf, &args);
9097 message = SvPV_const(msv,l1);
9100 Copy(message, buf, l1 , char);
9101 buf[l1-1] = '\0'; /* Overwrite \n */
9102 Perl_croak(aTHX_ "%s", buf);
9105 /* XXX Here's a total kludge. But we need to re-enter for swash routines. */
9107 #ifndef PERL_IN_XSUB_RE
9109 Perl_save_re_context(pTHX)
9113 struct re_save_state *state;
9115 SAVEVPTR(PL_curcop);
9116 SSGROW(SAVESTACK_ALLOC_FOR_RE_SAVE_STATE + 1);
9118 state = (struct re_save_state *)(PL_savestack + PL_savestack_ix);
9119 PL_savestack_ix += SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
9120 SSPUSHINT(SAVEt_RE_STATE);
9122 Copy(&PL_reg_state, state, 1, struct re_save_state);
9124 PL_reg_start_tmp = 0;
9125 PL_reg_start_tmpl = 0;
9126 PL_reg_oldsaved = NULL;
9127 PL_reg_oldsavedlen = 0;
9129 PL_reg_leftiter = 0;
9130 PL_reg_poscache = NULL;
9131 PL_reg_poscache_size = 0;
9132 #ifdef PERL_OLD_COPY_ON_WRITE
9136 /* Save $1..$n (#18107: UTF-8 s/(\w+)/uc($1)/e); AMS 20021106. */
9138 const REGEXP * const rx = PM_GETRE(PL_curpm);
9141 for (i = 1; i <= rx->nparens; i++) {
9142 char digits[TYPE_CHARS(long)];
9143 const STRLEN len = my_snprintf(digits, sizeof(digits), "%lu", (long)i);
9144 GV *const *const gvp
9145 = (GV**)hv_fetch(PL_defstash, digits, len, 0);
9148 GV * const gv = *gvp;
9149 if (SvTYPE(gv) == SVt_PVGV && GvSV(gv))
9159 clear_re(pTHX_ void *r)
9162 ReREFCNT_dec((regexp *)r);
9168 S_put_byte(pTHX_ SV *sv, int c)
9170 if (isCNTRL(c) || c == 255 || !isPRINT(c))
9171 Perl_sv_catpvf(aTHX_ sv, "\\%o", c);
9172 else if (c == '-' || c == ']' || c == '\\' || c == '^')
9173 Perl_sv_catpvf(aTHX_ sv, "\\%c", c);
9175 Perl_sv_catpvf(aTHX_ sv, "%c", c);
9179 #define CLEAR_OPTSTART \
9180 if (optstart) STMT_START { \
9181 DEBUG_OPTIMISE_r(PerlIO_printf(Perl_debug_log, " (%"IVdf" nodes)\n", (IV)(node - optstart))); \
9185 #define DUMPUNTIL(b,e) CLEAR_OPTSTART; node=dumpuntil(r,start,(b),(e),last,sv,indent+1,depth+1);
9187 STATIC const regnode *
9188 S_dumpuntil(pTHX_ const regexp *r, const regnode *start, const regnode *node,
9189 const regnode *last, const regnode *plast,
9190 SV* sv, I32 indent, U32 depth)
9193 register U8 op = PSEUDO; /* Arbitrary non-END op. */
9194 register const regnode *next;
9195 const regnode *optstart= NULL;
9198 GET_RE_DEBUG_FLAGS_DECL;
9200 #ifdef DEBUG_DUMPUNTIL
9201 PerlIO_printf(Perl_debug_log, "--- %d : %d - %d - %d\n",indent,node-start,
9202 last ? last-start : 0,plast ? plast-start : 0);
9205 if (plast && plast < last)
9208 while (PL_regkind[op] != END && (!last || node < last)) {
9209 /* While that wasn't END last time... */
9212 if (op == CLOSE || op == WHILEM)
9214 next = regnext((regnode *)node);
9217 if (OP(node) == OPTIMIZED) {
9218 if (!optstart && RE_DEBUG_FLAG(RE_DEBUG_COMPILE_OPTIMISE))
9225 regprop(r, sv, node);
9226 PerlIO_printf(Perl_debug_log, "%4"IVdf":%*s%s", (IV)(node - start),
9227 (int)(2*indent + 1), "", SvPVX_const(sv));
9229 if (OP(node) != OPTIMIZED) {
9230 if (next == NULL) /* Next ptr. */
9231 PerlIO_printf(Perl_debug_log, " (0)");
9232 else if (PL_regkind[(U8)op] == BRANCH && PL_regkind[OP(next)] != BRANCH )
9233 PerlIO_printf(Perl_debug_log, " (FAIL)");
9235 PerlIO_printf(Perl_debug_log, " (%"IVdf")", (IV)(next - start));
9236 (void)PerlIO_putc(Perl_debug_log, '\n');
9240 if (PL_regkind[(U8)op] == BRANCHJ) {
9243 register const regnode *nnode = (OP(next) == LONGJMP
9244 ? regnext((regnode *)next)
9246 if (last && nnode > last)
9248 DUMPUNTIL(NEXTOPER(NEXTOPER(node)), nnode);
9251 else if (PL_regkind[(U8)op] == BRANCH) {
9253 DUMPUNTIL(NEXTOPER(node), next);
9255 else if ( PL_regkind[(U8)op] == TRIE ) {
9256 const regnode *this_trie = node;
9257 const char op = OP(node);
9258 const U32 n = ARG(node);
9259 const reg_ac_data * const ac = op>=AHOCORASICK ?
9260 (reg_ac_data *)ri->data->data[n] :
9262 const reg_trie_data * const trie =
9263 (reg_trie_data*)ri->data->data[op<AHOCORASICK ? n : ac->trie];
9265 AV *const trie_words = (AV *) ri->data->data[n + TRIE_WORDS_OFFSET];
9267 const regnode *nextbranch= NULL;
9269 sv_setpvn(sv, "", 0);
9270 for (word_idx= 0; word_idx < (I32)trie->wordcount; word_idx++) {
9271 SV ** const elem_ptr = av_fetch(trie_words,word_idx,0);
9273 PerlIO_printf(Perl_debug_log, "%*s%s ",
9274 (int)(2*(indent+3)), "",
9275 elem_ptr ? pv_pretty(sv, SvPV_nolen_const(*elem_ptr), SvCUR(*elem_ptr), 60,
9276 PL_colors[0], PL_colors[1],
9277 (SvUTF8(*elem_ptr) ? PERL_PV_ESCAPE_UNI : 0) |
9278 PERL_PV_PRETTY_ELIPSES |
9284 U16 dist= trie->jump[word_idx+1];
9285 PerlIO_printf(Perl_debug_log, "(%"UVuf")\n",
9286 (UV)((dist ? this_trie + dist : next) - start));
9289 nextbranch= this_trie + trie->jump[0];
9290 DUMPUNTIL(this_trie + dist, nextbranch);
9292 if (nextbranch && PL_regkind[OP(nextbranch)]==BRANCH)
9293 nextbranch= regnext((regnode *)nextbranch);
9295 PerlIO_printf(Perl_debug_log, "\n");
9298 if (last && next > last)
9303 else if ( op == CURLY ) { /* "next" might be very big: optimizer */
9304 DUMPUNTIL(NEXTOPER(node) + EXTRA_STEP_2ARGS,
9305 NEXTOPER(node) + EXTRA_STEP_2ARGS + 1);
9307 else if (PL_regkind[(U8)op] == CURLY && op != CURLYX) {
9309 DUMPUNTIL(NEXTOPER(node) + EXTRA_STEP_2ARGS, next);
9311 else if ( op == PLUS || op == STAR) {
9312 DUMPUNTIL(NEXTOPER(node), NEXTOPER(node) + 1);
9314 else if (op == ANYOF) {
9315 /* arglen 1 + class block */
9316 node += 1 + ((ANYOF_FLAGS(node) & ANYOF_LARGE)
9317 ? ANYOF_CLASS_SKIP : ANYOF_SKIP);
9318 node = NEXTOPER(node);
9320 else if (PL_regkind[(U8)op] == EXACT) {
9321 /* Literal string, where present. */
9322 node += NODE_SZ_STR(node) - 1;
9323 node = NEXTOPER(node);
9326 node = NEXTOPER(node);
9327 node += regarglen[(U8)op];
9329 if (op == CURLYX || op == OPEN)
9333 #ifdef DEBUG_DUMPUNTIL
9334 PerlIO_printf(Perl_debug_log, "--- %d\n", (int)indent);
9339 #endif /* DEBUGGING */
9343 * c-indentation-style: bsd
9345 * indent-tabs-mode: t
9348 * ex: set ts=8 sts=4 sw=4 noet:
projects / p5sagit/p5-mst-13.2.git / blob