5 * 'A fair jaw-cracker dwarf-language must be.' --Samwise Gamgee
7 * [p.285 of _The Lord of the Rings_, II/iii: "The Ring Goes South"]
10 /* This file contains functions for compiling a regular expression. See
11 * also regexec.c which funnily enough, contains functions for executing
12 * a regular expression.
14 * This file is also copied at build time to ext/re/re_comp.c, where
15 * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT.
16 * This causes the main functions to be compiled under new names and with
17 * debugging support added, which makes "use re 'debug'" work.
20 /* NOTE: this is derived from Henry Spencer's regexp code, and should not
21 * confused with the original package (see point 3 below). Thanks, Henry!
24 /* Additional note: this code is very heavily munged from Henry's version
25 * in places. In some spots I've traded clarity for efficiency, so don't
26 * blame Henry for some of the lack of readability.
29 /* The names of the functions have been changed from regcomp and
30 * regexec to pregcomp and pregexec in order to avoid conflicts
31 * with the POSIX routines of the same names.
34 #ifdef PERL_EXT_RE_BUILD
39 * pregcomp and pregexec -- regsub and regerror are not used in perl
41 * Copyright (c) 1986 by University of Toronto.
42 * Written by Henry Spencer. Not derived from licensed software.
44 * Permission is granted to anyone to use this software for any
45 * purpose on any computer system, and to redistribute it freely,
46 * subject to the following restrictions:
48 * 1. The author is not responsible for the consequences of use of
49 * this software, no matter how awful, even if they arise
52 * 2. The origin of this software must not be misrepresented, either
53 * by explicit claim or by omission.
55 * 3. Altered versions must be plainly marked as such, and must not
56 * be misrepresented as being the original software.
59 **** Alterations to Henry's code are...
61 **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
62 **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
63 **** by Larry Wall and others
65 **** You may distribute under the terms of either the GNU General Public
66 **** License or the Artistic License, as specified in the README file.
69 * Beware that some of this code is subtly aware of the way operator
70 * precedence is structured in regular expressions. Serious changes in
71 * regular-expression syntax might require a total rethink.
74 #define PERL_IN_REGCOMP_C
77 #ifndef PERL_IN_XSUB_RE
82 #ifdef PERL_IN_XSUB_RE
93 # if defined(BUGGY_MSC6)
94 /* MSC 6.00A breaks on op/regexp.t test 85 unless we turn this off */
95 # pragma optimize("a",off)
96 /* But MSC 6.00A is happy with 'w', for aliases only across function calls*/
97 # pragma optimize("w",on )
98 # endif /* BUGGY_MSC6 */
102 #define STATIC static
105 typedef struct RExC_state_t {
106 U32 flags; /* are we folding, multilining? */
107 char *precomp; /* uncompiled string. */
108 REGEXP *rx_sv; /* The SV that is the regexp. */
109 regexp *rx; /* perl core regexp structure */
110 regexp_internal *rxi; /* internal data for regexp object pprivate field */
111 char *start; /* Start of input for compile */
112 char *end; /* End of input for compile */
113 char *parse; /* Input-scan pointer. */
114 I32 whilem_seen; /* number of WHILEM in this expr */
115 regnode *emit_start; /* Start of emitted-code area */
116 regnode *emit_bound; /* First regnode outside of the allocated space */
117 regnode *emit; /* Code-emit pointer; ®dummy = don't = compiling */
118 I32 naughty; /* How bad is this pattern? */
119 I32 sawback; /* Did we see \1, ...? */
121 I32 size; /* Code size. */
122 I32 npar; /* Capture buffer count, (OPEN). */
123 I32 cpar; /* Capture buffer count, (CLOSE). */
124 I32 nestroot; /* root parens we are in - used by accept */
128 regnode **open_parens; /* pointers to open parens */
129 regnode **close_parens; /* pointers to close parens */
130 regnode *opend; /* END node in program */
131 I32 utf8; /* whether the pattern is utf8 or not */
132 I32 orig_utf8; /* whether the pattern was originally in utf8 */
133 /* XXX use this for future optimisation of case
134 * where pattern must be upgraded to utf8. */
135 HV *paren_names; /* Paren names */
137 regnode **recurse; /* Recurse regops */
138 I32 recurse_count; /* Number of recurse regops */
140 char *starttry; /* -Dr: where regtry was called. */
141 #define RExC_starttry (pRExC_state->starttry)
144 const char *lastparse;
146 AV *paren_name_list; /* idx -> name */
147 #define RExC_lastparse (pRExC_state->lastparse)
148 #define RExC_lastnum (pRExC_state->lastnum)
149 #define RExC_paren_name_list (pRExC_state->paren_name_list)
153 #define RExC_flags (pRExC_state->flags)
154 #define RExC_precomp (pRExC_state->precomp)
155 #define RExC_rx_sv (pRExC_state->rx_sv)
156 #define RExC_rx (pRExC_state->rx)
157 #define RExC_rxi (pRExC_state->rxi)
158 #define RExC_start (pRExC_state->start)
159 #define RExC_end (pRExC_state->end)
160 #define RExC_parse (pRExC_state->parse)
161 #define RExC_whilem_seen (pRExC_state->whilem_seen)
162 #ifdef RE_TRACK_PATTERN_OFFSETS
163 #define RExC_offsets (pRExC_state->rxi->u.offsets) /* I am not like the others */
165 #define RExC_emit (pRExC_state->emit)
166 #define RExC_emit_start (pRExC_state->emit_start)
167 #define RExC_emit_bound (pRExC_state->emit_bound)
168 #define RExC_naughty (pRExC_state->naughty)
169 #define RExC_sawback (pRExC_state->sawback)
170 #define RExC_seen (pRExC_state->seen)
171 #define RExC_size (pRExC_state->size)
172 #define RExC_npar (pRExC_state->npar)
173 #define RExC_nestroot (pRExC_state->nestroot)
174 #define RExC_extralen (pRExC_state->extralen)
175 #define RExC_seen_zerolen (pRExC_state->seen_zerolen)
176 #define RExC_seen_evals (pRExC_state->seen_evals)
177 #define RExC_utf8 (pRExC_state->utf8)
178 #define RExC_orig_utf8 (pRExC_state->orig_utf8)
179 #define RExC_open_parens (pRExC_state->open_parens)
180 #define RExC_close_parens (pRExC_state->close_parens)
181 #define RExC_opend (pRExC_state->opend)
182 #define RExC_paren_names (pRExC_state->paren_names)
183 #define RExC_recurse (pRExC_state->recurse)
184 #define RExC_recurse_count (pRExC_state->recurse_count)
187 #define ISMULT1(c) ((c) == '*' || (c) == '+' || (c) == '?')
188 #define ISMULT2(s) ((*s) == '*' || (*s) == '+' || (*s) == '?' || \
189 ((*s) == '{' && regcurly(s)))
192 #undef SPSTART /* dratted cpp namespace... */
195 * Flags to be passed up and down.
197 #define WORST 0 /* Worst case. */
198 #define HASWIDTH 0x01 /* Known to match non-null strings. */
199 #define SIMPLE 0x02 /* Simple enough to be STAR/PLUS operand. */
200 #define SPSTART 0x04 /* Starts with * or +. */
201 #define TRYAGAIN 0x08 /* Weeded out a declaration. */
202 #define POSTPONED 0x10 /* (?1),(?&name), (??{...}) or similar */
204 #define REG_NODE_NUM(x) ((x) ? (int)((x)-RExC_emit_start) : -1)
206 /* whether trie related optimizations are enabled */
207 #if PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION
208 #define TRIE_STUDY_OPT
209 #define FULL_TRIE_STUDY
215 #define PBYTE(u8str,paren) ((U8*)(u8str))[(paren) >> 3]
216 #define PBITVAL(paren) (1 << ((paren) & 7))
217 #define PAREN_TEST(u8str,paren) ( PBYTE(u8str,paren) & PBITVAL(paren))
218 #define PAREN_SET(u8str,paren) PBYTE(u8str,paren) |= PBITVAL(paren)
219 #define PAREN_UNSET(u8str,paren) PBYTE(u8str,paren) &= (~PBITVAL(paren))
222 /* About scan_data_t.
224 During optimisation we recurse through the regexp program performing
225 various inplace (keyhole style) optimisations. In addition study_chunk
226 and scan_commit populate this data structure with information about
227 what strings MUST appear in the pattern. We look for the longest
228 string that must appear for at a fixed location, and we look for the
229 longest string that may appear at a floating location. So for instance
234 Both 'FOO' and 'A' are fixed strings. Both 'B' and 'BAR' are floating
235 strings (because they follow a .* construct). study_chunk will identify
236 both FOO and BAR as being the longest fixed and floating strings respectively.
238 The strings can be composites, for instance
242 will result in a composite fixed substring 'foo'.
244 For each string some basic information is maintained:
246 - offset or min_offset
247 This is the position the string must appear at, or not before.
248 It also implicitly (when combined with minlenp) tells us how many
249 character must match before the string we are searching.
250 Likewise when combined with minlenp and the length of the string
251 tells us how many characters must appear after the string we have
255 Only used for floating strings. This is the rightmost point that
256 the string can appear at. Ifset to I32 max it indicates that the
257 string can occur infinitely far to the right.
260 A pointer to the minimum length of the pattern that the string
261 was found inside. This is important as in the case of positive
262 lookahead or positive lookbehind we can have multiple patterns
267 The minimum length of the pattern overall is 3, the minimum length
268 of the lookahead part is 3, but the minimum length of the part that
269 will actually match is 1. So 'FOO's minimum length is 3, but the
270 minimum length for the F is 1. This is important as the minimum length
271 is used to determine offsets in front of and behind the string being
272 looked for. Since strings can be composites this is the length of the
273 pattern at the time it was commited with a scan_commit. Note that
274 the length is calculated by study_chunk, so that the minimum lengths
275 are not known until the full pattern has been compiled, thus the
276 pointer to the value.
280 In the case of lookbehind the string being searched for can be
281 offset past the start point of the final matching string.
282 If this value was just blithely removed from the min_offset it would
283 invalidate some of the calculations for how many chars must match
284 before or after (as they are derived from min_offset and minlen and
285 the length of the string being searched for).
286 When the final pattern is compiled and the data is moved from the
287 scan_data_t structure into the regexp structure the information
288 about lookbehind is factored in, with the information that would
289 have been lost precalculated in the end_shift field for the
292 The fields pos_min and pos_delta are used to store the minimum offset
293 and the delta to the maximum offset at the current point in the pattern.
297 typedef struct scan_data_t {
298 /*I32 len_min; unused */
299 /*I32 len_delta; unused */
303 I32 last_end; /* min value, <0 unless valid. */
306 SV **longest; /* Either &l_fixed, or &l_float. */
307 SV *longest_fixed; /* longest fixed string found in pattern */
308 I32 offset_fixed; /* offset where it starts */
309 I32 *minlen_fixed; /* pointer to the minlen relevent to the string */
310 I32 lookbehind_fixed; /* is the position of the string modfied by LB */
311 SV *longest_float; /* longest floating string found in pattern */
312 I32 offset_float_min; /* earliest point in string it can appear */
313 I32 offset_float_max; /* latest point in string it can appear */
314 I32 *minlen_float; /* pointer to the minlen relevent to the string */
315 I32 lookbehind_float; /* is the position of the string modified by LB */
319 struct regnode_charclass_class *start_class;
323 * Forward declarations for pregcomp()'s friends.
326 static const scan_data_t zero_scan_data =
327 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ,0};
329 #define SF_BEFORE_EOL (SF_BEFORE_SEOL|SF_BEFORE_MEOL)
330 #define SF_BEFORE_SEOL 0x0001
331 #define SF_BEFORE_MEOL 0x0002
332 #define SF_FIX_BEFORE_EOL (SF_FIX_BEFORE_SEOL|SF_FIX_BEFORE_MEOL)
333 #define SF_FL_BEFORE_EOL (SF_FL_BEFORE_SEOL|SF_FL_BEFORE_MEOL)
336 # define SF_FIX_SHIFT_EOL (0+2)
337 # define SF_FL_SHIFT_EOL (0+4)
339 # define SF_FIX_SHIFT_EOL (+2)
340 # define SF_FL_SHIFT_EOL (+4)
343 #define SF_FIX_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FIX_SHIFT_EOL)
344 #define SF_FIX_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FIX_SHIFT_EOL)
346 #define SF_FL_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FL_SHIFT_EOL)
347 #define SF_FL_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FL_SHIFT_EOL) /* 0x20 */
348 #define SF_IS_INF 0x0040
349 #define SF_HAS_PAR 0x0080
350 #define SF_IN_PAR 0x0100
351 #define SF_HAS_EVAL 0x0200
352 #define SCF_DO_SUBSTR 0x0400
353 #define SCF_DO_STCLASS_AND 0x0800
354 #define SCF_DO_STCLASS_OR 0x1000
355 #define SCF_DO_STCLASS (SCF_DO_STCLASS_AND|SCF_DO_STCLASS_OR)
356 #define SCF_WHILEM_VISITED_POS 0x2000
358 #define SCF_TRIE_RESTUDY 0x4000 /* Do restudy? */
359 #define SCF_SEEN_ACCEPT 0x8000
361 #define UTF (RExC_utf8 != 0)
362 #define LOC ((RExC_flags & RXf_PMf_LOCALE) != 0)
363 #define FOLD ((RExC_flags & RXf_PMf_FOLD) != 0)
365 #define OOB_UNICODE 12345678
366 #define OOB_NAMEDCLASS -1
368 #define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv))
369 #define CHR_DIST(a,b) (UTF ? utf8_distance(a,b) : a - b)
372 /* length of regex to show in messages that don't mark a position within */
373 #define RegexLengthToShowInErrorMessages 127
376 * If MARKER[12] are adjusted, be sure to adjust the constants at the top
377 * of t/op/regmesg.t, the tests in t/op/re_tests, and those in
378 * op/pragma/warn/regcomp.
380 #define MARKER1 "<-- HERE" /* marker as it appears in the description */
381 #define MARKER2 " <-- HERE " /* marker as it appears within the regex */
383 #define REPORT_LOCATION " in regex; marked by " MARKER1 " in m/%.*s" MARKER2 "%s/"
386 * Calls SAVEDESTRUCTOR_X if needed, then calls Perl_croak with the given
387 * arg. Show regex, up to a maximum length. If it's too long, chop and add
390 #define _FAIL(code) STMT_START { \
391 const char *ellipses = ""; \
392 IV len = RExC_end - RExC_precomp; \
395 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx_sv); \
396 if (len > RegexLengthToShowInErrorMessages) { \
397 /* chop 10 shorter than the max, to ensure meaning of "..." */ \
398 len = RegexLengthToShowInErrorMessages - 10; \
404 #define FAIL(msg) _FAIL( \
405 Perl_croak(aTHX_ "%s in regex m/%.*s%s/", \
406 msg, (int)len, RExC_precomp, ellipses))
408 #define FAIL2(msg,arg) _FAIL( \
409 Perl_croak(aTHX_ msg " in regex m/%.*s%s/", \
410 arg, (int)len, RExC_precomp, ellipses))
413 * Simple_vFAIL -- like FAIL, but marks the current location in the scan
415 #define Simple_vFAIL(m) STMT_START { \
416 const IV offset = RExC_parse - RExC_precomp; \
417 Perl_croak(aTHX_ "%s" REPORT_LOCATION, \
418 m, (int)offset, RExC_precomp, RExC_precomp + offset); \
422 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL()
424 #define vFAIL(m) STMT_START { \
426 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx_sv); \
431 * Like Simple_vFAIL(), but accepts two arguments.
433 #define Simple_vFAIL2(m,a1) STMT_START { \
434 const IV offset = RExC_parse - RExC_precomp; \
435 S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, \
436 (int)offset, RExC_precomp, RExC_precomp + offset); \
440 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL2().
442 #define vFAIL2(m,a1) STMT_START { \
444 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx_sv); \
445 Simple_vFAIL2(m, a1); \
450 * Like Simple_vFAIL(), but accepts three arguments.
452 #define Simple_vFAIL3(m, a1, a2) STMT_START { \
453 const IV offset = RExC_parse - RExC_precomp; \
454 S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, a2, \
455 (int)offset, RExC_precomp, RExC_precomp + offset); \
459 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL3().
461 #define vFAIL3(m,a1,a2) STMT_START { \
463 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx_sv); \
464 Simple_vFAIL3(m, a1, a2); \
468 * Like Simple_vFAIL(), but accepts four arguments.
470 #define Simple_vFAIL4(m, a1, a2, a3) STMT_START { \
471 const IV offset = RExC_parse - RExC_precomp; \
472 S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, a2, a3, \
473 (int)offset, RExC_precomp, RExC_precomp + offset); \
476 #define ckWARNreg(loc,m) STMT_START { \
477 const IV offset = loc - RExC_precomp; \
478 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
479 (int)offset, RExC_precomp, RExC_precomp + offset); \
482 #define ckWARNregdep(loc,m) STMT_START { \
483 const IV offset = loc - RExC_precomp; \
484 Perl_ck_warner_d(aTHX_ packWARN2(WARN_DEPRECATED, WARN_REGEXP), \
486 (int)offset, RExC_precomp, RExC_precomp + offset); \
489 #define ckWARN2reg(loc, m, a1) STMT_START { \
490 const IV offset = loc - RExC_precomp; \
491 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
492 a1, (int)offset, RExC_precomp, RExC_precomp + offset); \
495 #define vWARN3(loc, m, a1, a2) STMT_START { \
496 const IV offset = loc - RExC_precomp; \
497 Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
498 a1, a2, (int)offset, RExC_precomp, RExC_precomp + offset); \
501 #define ckWARN3reg(loc, m, a1, a2) STMT_START { \
502 const IV offset = loc - RExC_precomp; \
503 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
504 a1, a2, (int)offset, RExC_precomp, RExC_precomp + offset); \
507 #define vWARN4(loc, m, a1, a2, a3) STMT_START { \
508 const IV offset = loc - RExC_precomp; \
509 Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
510 a1, a2, a3, (int)offset, RExC_precomp, RExC_precomp + offset); \
513 #define ckWARN4reg(loc, m, a1, a2, a3) STMT_START { \
514 const IV offset = loc - RExC_precomp; \
515 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
516 a1, a2, a3, (int)offset, RExC_precomp, RExC_precomp + offset); \
519 #define vWARN5(loc, m, a1, a2, a3, a4) STMT_START { \
520 const IV offset = loc - RExC_precomp; \
521 Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
522 a1, a2, a3, a4, (int)offset, RExC_precomp, RExC_precomp + offset); \
526 /* Allow for side effects in s */
527 #define REGC(c,s) STMT_START { \
528 if (!SIZE_ONLY) *(s) = (c); else (void)(s); \
531 /* Macros for recording node offsets. 20001227 mjd@plover.com
532 * Nodes are numbered 1, 2, 3, 4. Node #n's position is recorded in
533 * element 2*n-1 of the array. Element #2n holds the byte length node #n.
534 * Element 0 holds the number n.
535 * Position is 1 indexed.
537 #ifndef RE_TRACK_PATTERN_OFFSETS
538 #define Set_Node_Offset_To_R(node,byte)
539 #define Set_Node_Offset(node,byte)
540 #define Set_Cur_Node_Offset
541 #define Set_Node_Length_To_R(node,len)
542 #define Set_Node_Length(node,len)
543 #define Set_Node_Cur_Length(node)
544 #define Node_Offset(n)
545 #define Node_Length(n)
546 #define Set_Node_Offset_Length(node,offset,len)
547 #define ProgLen(ri) ri->u.proglen
548 #define SetProgLen(ri,x) ri->u.proglen = x
550 #define ProgLen(ri) ri->u.offsets[0]
551 #define SetProgLen(ri,x) ri->u.offsets[0] = x
552 #define Set_Node_Offset_To_R(node,byte) STMT_START { \
554 MJD_OFFSET_DEBUG(("** (%d) offset of node %d is %d.\n", \
555 __LINE__, (int)(node), (int)(byte))); \
557 Perl_croak(aTHX_ "value of node is %d in Offset macro", (int)(node)); \
559 RExC_offsets[2*(node)-1] = (byte); \
564 #define Set_Node_Offset(node,byte) \
565 Set_Node_Offset_To_R((node)-RExC_emit_start, (byte)-RExC_start)
566 #define Set_Cur_Node_Offset Set_Node_Offset(RExC_emit, RExC_parse)
568 #define Set_Node_Length_To_R(node,len) STMT_START { \
570 MJD_OFFSET_DEBUG(("** (%d) size of node %d is %d.\n", \
571 __LINE__, (int)(node), (int)(len))); \
573 Perl_croak(aTHX_ "value of node is %d in Length macro", (int)(node)); \
575 RExC_offsets[2*(node)] = (len); \
580 #define Set_Node_Length(node,len) \
581 Set_Node_Length_To_R((node)-RExC_emit_start, len)
582 #define Set_Cur_Node_Length(len) Set_Node_Length(RExC_emit, len)
583 #define Set_Node_Cur_Length(node) \
584 Set_Node_Length(node, RExC_parse - parse_start)
586 /* Get offsets and lengths */
587 #define Node_Offset(n) (RExC_offsets[2*((n)-RExC_emit_start)-1])
588 #define Node_Length(n) (RExC_offsets[2*((n)-RExC_emit_start)])
590 #define Set_Node_Offset_Length(node,offset,len) STMT_START { \
591 Set_Node_Offset_To_R((node)-RExC_emit_start, (offset)); \
592 Set_Node_Length_To_R((node)-RExC_emit_start, (len)); \
596 #if PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS
597 #define EXPERIMENTAL_INPLACESCAN
598 #endif /*RE_TRACK_PATTERN_OFFSETS*/
600 #define DEBUG_STUDYDATA(str,data,depth) \
601 DEBUG_OPTIMISE_MORE_r(if(data){ \
602 PerlIO_printf(Perl_debug_log, \
603 "%*s" str "Pos:%"IVdf"/%"IVdf \
604 " Flags: 0x%"UVXf" Whilem_c: %"IVdf" Lcp: %"IVdf" %s", \
605 (int)(depth)*2, "", \
606 (IV)((data)->pos_min), \
607 (IV)((data)->pos_delta), \
608 (UV)((data)->flags), \
609 (IV)((data)->whilem_c), \
610 (IV)((data)->last_closep ? *((data)->last_closep) : -1), \
611 is_inf ? "INF " : "" \
613 if ((data)->last_found) \
614 PerlIO_printf(Perl_debug_log, \
615 "Last:'%s' %"IVdf":%"IVdf"/%"IVdf" %sFixed:'%s' @ %"IVdf \
616 " %sFloat: '%s' @ %"IVdf"/%"IVdf"", \
617 SvPVX_const((data)->last_found), \
618 (IV)((data)->last_end), \
619 (IV)((data)->last_start_min), \
620 (IV)((data)->last_start_max), \
621 ((data)->longest && \
622 (data)->longest==&((data)->longest_fixed)) ? "*" : "", \
623 SvPVX_const((data)->longest_fixed), \
624 (IV)((data)->offset_fixed), \
625 ((data)->longest && \
626 (data)->longest==&((data)->longest_float)) ? "*" : "", \
627 SvPVX_const((data)->longest_float), \
628 (IV)((data)->offset_float_min), \
629 (IV)((data)->offset_float_max) \
631 PerlIO_printf(Perl_debug_log,"\n"); \
634 static void clear_re(pTHX_ void *r);
636 /* Mark that we cannot extend a found fixed substring at this point.
637 Update the longest found anchored substring and the longest found
638 floating substrings if needed. */
641 S_scan_commit(pTHX_ const RExC_state_t *pRExC_state, scan_data_t *data, I32 *minlenp, int is_inf)
643 const STRLEN l = CHR_SVLEN(data->last_found);
644 const STRLEN old_l = CHR_SVLEN(*data->longest);
645 GET_RE_DEBUG_FLAGS_DECL;
647 PERL_ARGS_ASSERT_SCAN_COMMIT;
649 if ((l >= old_l) && ((l > old_l) || (data->flags & SF_BEFORE_EOL))) {
650 SvSetMagicSV(*data->longest, data->last_found);
651 if (*data->longest == data->longest_fixed) {
652 data->offset_fixed = l ? data->last_start_min : data->pos_min;
653 if (data->flags & SF_BEFORE_EOL)
655 |= ((data->flags & SF_BEFORE_EOL) << SF_FIX_SHIFT_EOL);
657 data->flags &= ~SF_FIX_BEFORE_EOL;
658 data->minlen_fixed=minlenp;
659 data->lookbehind_fixed=0;
661 else { /* *data->longest == data->longest_float */
662 data->offset_float_min = l ? data->last_start_min : data->pos_min;
663 data->offset_float_max = (l
664 ? data->last_start_max
665 : data->pos_min + data->pos_delta);
666 if (is_inf || (U32)data->offset_float_max > (U32)I32_MAX)
667 data->offset_float_max = I32_MAX;
668 if (data->flags & SF_BEFORE_EOL)
670 |= ((data->flags & SF_BEFORE_EOL) << SF_FL_SHIFT_EOL);
672 data->flags &= ~SF_FL_BEFORE_EOL;
673 data->minlen_float=minlenp;
674 data->lookbehind_float=0;
677 SvCUR_set(data->last_found, 0);
679 SV * const sv = data->last_found;
680 if (SvUTF8(sv) && SvMAGICAL(sv)) {
681 MAGIC * const mg = mg_find(sv, PERL_MAGIC_utf8);
687 data->flags &= ~SF_BEFORE_EOL;
688 DEBUG_STUDYDATA("commit: ",data,0);
691 /* Can match anything (initialization) */
693 S_cl_anything(const RExC_state_t *pRExC_state, struct regnode_charclass_class *cl)
695 PERL_ARGS_ASSERT_CL_ANYTHING;
697 ANYOF_CLASS_ZERO(cl);
698 ANYOF_BITMAP_SETALL(cl);
699 cl->flags = ANYOF_EOS|ANYOF_UNICODE_ALL;
701 cl->flags |= ANYOF_LOCALE;
704 /* Can match anything (initialization) */
706 S_cl_is_anything(const struct regnode_charclass_class *cl)
710 PERL_ARGS_ASSERT_CL_IS_ANYTHING;
712 for (value = 0; value <= ANYOF_MAX; value += 2)
713 if (ANYOF_CLASS_TEST(cl, value) && ANYOF_CLASS_TEST(cl, value + 1))
715 if (!(cl->flags & ANYOF_UNICODE_ALL))
717 if (!ANYOF_BITMAP_TESTALLSET((const void*)cl))
722 /* Can match anything (initialization) */
724 S_cl_init(const RExC_state_t *pRExC_state, struct regnode_charclass_class *cl)
726 PERL_ARGS_ASSERT_CL_INIT;
728 Zero(cl, 1, struct regnode_charclass_class);
730 cl_anything(pRExC_state, cl);
734 S_cl_init_zero(const RExC_state_t *pRExC_state, struct regnode_charclass_class *cl)
736 PERL_ARGS_ASSERT_CL_INIT_ZERO;
738 Zero(cl, 1, struct regnode_charclass_class);
740 cl_anything(pRExC_state, cl);
742 cl->flags |= ANYOF_LOCALE;
745 /* 'And' a given class with another one. Can create false positives */
746 /* We assume that cl is not inverted */
748 S_cl_and(struct regnode_charclass_class *cl,
749 const struct regnode_charclass_class *and_with)
751 PERL_ARGS_ASSERT_CL_AND;
753 assert(and_with->type == ANYOF);
754 if (!(and_with->flags & ANYOF_CLASS)
755 && !(cl->flags & ANYOF_CLASS)
756 && (and_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
757 && !(and_with->flags & ANYOF_FOLD)
758 && !(cl->flags & ANYOF_FOLD)) {
761 if (and_with->flags & ANYOF_INVERT)
762 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
763 cl->bitmap[i] &= ~and_with->bitmap[i];
765 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
766 cl->bitmap[i] &= and_with->bitmap[i];
767 } /* XXXX: logic is complicated otherwise, leave it along for a moment. */
768 if (!(and_with->flags & ANYOF_EOS))
769 cl->flags &= ~ANYOF_EOS;
771 if (cl->flags & ANYOF_UNICODE_ALL && and_with->flags & ANYOF_UNICODE &&
772 !(and_with->flags & ANYOF_INVERT)) {
773 cl->flags &= ~ANYOF_UNICODE_ALL;
774 cl->flags |= ANYOF_UNICODE;
775 ARG_SET(cl, ARG(and_with));
777 if (!(and_with->flags & ANYOF_UNICODE_ALL) &&
778 !(and_with->flags & ANYOF_INVERT))
779 cl->flags &= ~ANYOF_UNICODE_ALL;
780 if (!(and_with->flags & (ANYOF_UNICODE|ANYOF_UNICODE_ALL)) &&
781 !(and_with->flags & ANYOF_INVERT))
782 cl->flags &= ~ANYOF_UNICODE;
785 /* 'OR' a given class with another one. Can create false positives */
786 /* We assume that cl is not inverted */
788 S_cl_or(const RExC_state_t *pRExC_state, struct regnode_charclass_class *cl, const struct regnode_charclass_class *or_with)
790 PERL_ARGS_ASSERT_CL_OR;
792 if (or_with->flags & ANYOF_INVERT) {
794 * (B1 | CL1) | (!B2 & !CL2) = (B1 | !B2 & !CL2) | (CL1 | (!B2 & !CL2))
795 * <= (B1 | !B2) | (CL1 | !CL2)
796 * which is wasteful if CL2 is small, but we ignore CL2:
797 * (B1 | CL1) | (!B2 & !CL2) <= (B1 | CL1) | !B2 = (B1 | !B2) | CL1
798 * XXXX Can we handle case-fold? Unclear:
799 * (OK1(i) | OK1(i')) | !(OK1(i) | OK1(i')) =
800 * (OK1(i) | OK1(i')) | (!OK1(i) & !OK1(i'))
802 if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
803 && !(or_with->flags & ANYOF_FOLD)
804 && !(cl->flags & ANYOF_FOLD) ) {
807 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
808 cl->bitmap[i] |= ~or_with->bitmap[i];
809 } /* XXXX: logic is complicated otherwise */
811 cl_anything(pRExC_state, cl);
814 /* (B1 | CL1) | (B2 | CL2) = (B1 | B2) | (CL1 | CL2)) */
815 if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
816 && (!(or_with->flags & ANYOF_FOLD)
817 || (cl->flags & ANYOF_FOLD)) ) {
820 /* OR char bitmap and class bitmap separately */
821 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
822 cl->bitmap[i] |= or_with->bitmap[i];
823 if (or_with->flags & ANYOF_CLASS) {
824 for (i = 0; i < ANYOF_CLASSBITMAP_SIZE; i++)
825 cl->classflags[i] |= or_with->classflags[i];
826 cl->flags |= ANYOF_CLASS;
829 else { /* XXXX: logic is complicated, leave it along for a moment. */
830 cl_anything(pRExC_state, cl);
833 if (or_with->flags & ANYOF_EOS)
834 cl->flags |= ANYOF_EOS;
836 if (cl->flags & ANYOF_UNICODE && or_with->flags & ANYOF_UNICODE &&
837 ARG(cl) != ARG(or_with)) {
838 cl->flags |= ANYOF_UNICODE_ALL;
839 cl->flags &= ~ANYOF_UNICODE;
841 if (or_with->flags & ANYOF_UNICODE_ALL) {
842 cl->flags |= ANYOF_UNICODE_ALL;
843 cl->flags &= ~ANYOF_UNICODE;
847 #define TRIE_LIST_ITEM(state,idx) (trie->states[state].trans.list)[ idx ]
848 #define TRIE_LIST_CUR(state) ( TRIE_LIST_ITEM( state, 0 ).forid )
849 #define TRIE_LIST_LEN(state) ( TRIE_LIST_ITEM( state, 0 ).newstate )
850 #define TRIE_LIST_USED(idx) ( trie->states[state].trans.list ? (TRIE_LIST_CUR( idx ) - 1) : 0 )
855 dump_trie(trie,widecharmap,revcharmap)
856 dump_trie_interim_list(trie,widecharmap,revcharmap,next_alloc)
857 dump_trie_interim_table(trie,widecharmap,revcharmap,next_alloc)
859 These routines dump out a trie in a somewhat readable format.
860 The _interim_ variants are used for debugging the interim
861 tables that are used to generate the final compressed
862 representation which is what dump_trie expects.
864 Part of the reason for their existance is to provide a form
865 of documentation as to how the different representations function.
870 Dumps the final compressed table form of the trie to Perl_debug_log.
871 Used for debugging make_trie().
875 S_dump_trie(pTHX_ const struct _reg_trie_data *trie, HV *widecharmap,
876 AV *revcharmap, U32 depth)
879 SV *sv=sv_newmortal();
880 int colwidth= widecharmap ? 6 : 4;
882 GET_RE_DEBUG_FLAGS_DECL;
884 PERL_ARGS_ASSERT_DUMP_TRIE;
886 PerlIO_printf( Perl_debug_log, "%*sChar : %-6s%-6s%-4s ",
887 (int)depth * 2 + 2,"",
888 "Match","Base","Ofs" );
890 for( state = 0 ; state < trie->uniquecharcount ; state++ ) {
891 SV ** const tmp = av_fetch( revcharmap, state, 0);
893 PerlIO_printf( Perl_debug_log, "%*s",
895 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
896 PL_colors[0], PL_colors[1],
897 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
898 PERL_PV_ESCAPE_FIRSTCHAR
903 PerlIO_printf( Perl_debug_log, "\n%*sState|-----------------------",
904 (int)depth * 2 + 2,"");
906 for( state = 0 ; state < trie->uniquecharcount ; state++ )
907 PerlIO_printf( Perl_debug_log, "%.*s", colwidth, "--------");
908 PerlIO_printf( Perl_debug_log, "\n");
910 for( state = 1 ; state < trie->statecount ; state++ ) {
911 const U32 base = trie->states[ state ].trans.base;
913 PerlIO_printf( Perl_debug_log, "%*s#%4"UVXf"|", (int)depth * 2 + 2,"", (UV)state);
915 if ( trie->states[ state ].wordnum ) {
916 PerlIO_printf( Perl_debug_log, " W%4X", trie->states[ state ].wordnum );
918 PerlIO_printf( Perl_debug_log, "%6s", "" );
921 PerlIO_printf( Perl_debug_log, " @%4"UVXf" ", (UV)base );
926 while( ( base + ofs < trie->uniquecharcount ) ||
927 ( base + ofs - trie->uniquecharcount < trie->lasttrans
928 && trie->trans[ base + ofs - trie->uniquecharcount ].check != state))
931 PerlIO_printf( Perl_debug_log, "+%2"UVXf"[ ", (UV)ofs);
933 for ( ofs = 0 ; ofs < trie->uniquecharcount ; ofs++ ) {
934 if ( ( base + ofs >= trie->uniquecharcount ) &&
935 ( base + ofs - trie->uniquecharcount < trie->lasttrans ) &&
936 trie->trans[ base + ofs - trie->uniquecharcount ].check == state )
938 PerlIO_printf( Perl_debug_log, "%*"UVXf,
940 (UV)trie->trans[ base + ofs - trie->uniquecharcount ].next );
942 PerlIO_printf( Perl_debug_log, "%*s",colwidth," ." );
946 PerlIO_printf( Perl_debug_log, "]");
949 PerlIO_printf( Perl_debug_log, "\n" );
951 PerlIO_printf(Perl_debug_log, "%*sword_info N:(prev,len)=", (int)depth*2, "");
952 for (word=1; word <= trie->wordcount; word++) {
953 PerlIO_printf(Perl_debug_log, " %d:(%d,%d)",
954 (int)word, (int)(trie->wordinfo[word].prev),
955 (int)(trie->wordinfo[word].len));
957 PerlIO_printf(Perl_debug_log, "\n" );
960 Dumps a fully constructed but uncompressed trie in list form.
961 List tries normally only are used for construction when the number of
962 possible chars (trie->uniquecharcount) is very high.
963 Used for debugging make_trie().
966 S_dump_trie_interim_list(pTHX_ const struct _reg_trie_data *trie,
967 HV *widecharmap, AV *revcharmap, U32 next_alloc,
971 SV *sv=sv_newmortal();
972 int colwidth= widecharmap ? 6 : 4;
973 GET_RE_DEBUG_FLAGS_DECL;
975 PERL_ARGS_ASSERT_DUMP_TRIE_INTERIM_LIST;
977 /* print out the table precompression. */
978 PerlIO_printf( Perl_debug_log, "%*sState :Word | Transition Data\n%*s%s",
979 (int)depth * 2 + 2,"", (int)depth * 2 + 2,"",
980 "------:-----+-----------------\n" );
982 for( state=1 ; state < next_alloc ; state ++ ) {
985 PerlIO_printf( Perl_debug_log, "%*s %4"UVXf" :",
986 (int)depth * 2 + 2,"", (UV)state );
987 if ( ! trie->states[ state ].wordnum ) {
988 PerlIO_printf( Perl_debug_log, "%5s| ","");
990 PerlIO_printf( Perl_debug_log, "W%4x| ",
991 trie->states[ state ].wordnum
994 for( charid = 1 ; charid <= TRIE_LIST_USED( state ) ; charid++ ) {
995 SV ** const tmp = av_fetch( revcharmap, TRIE_LIST_ITEM(state,charid).forid, 0);
997 PerlIO_printf( Perl_debug_log, "%*s:%3X=%4"UVXf" | ",
999 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
1000 PL_colors[0], PL_colors[1],
1001 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
1002 PERL_PV_ESCAPE_FIRSTCHAR
1004 TRIE_LIST_ITEM(state,charid).forid,
1005 (UV)TRIE_LIST_ITEM(state,charid).newstate
1008 PerlIO_printf(Perl_debug_log, "\n%*s| ",
1009 (int)((depth * 2) + 14), "");
1012 PerlIO_printf( Perl_debug_log, "\n");
1017 Dumps a fully constructed but uncompressed trie in table form.
1018 This is the normal DFA style state transition table, with a few
1019 twists to facilitate compression later.
1020 Used for debugging make_trie().
1023 S_dump_trie_interim_table(pTHX_ const struct _reg_trie_data *trie,
1024 HV *widecharmap, AV *revcharmap, U32 next_alloc,
1029 SV *sv=sv_newmortal();
1030 int colwidth= widecharmap ? 6 : 4;
1031 GET_RE_DEBUG_FLAGS_DECL;
1033 PERL_ARGS_ASSERT_DUMP_TRIE_INTERIM_TABLE;
1036 print out the table precompression so that we can do a visual check
1037 that they are identical.
1040 PerlIO_printf( Perl_debug_log, "%*sChar : ",(int)depth * 2 + 2,"" );
1042 for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) {
1043 SV ** const tmp = av_fetch( revcharmap, charid, 0);
1045 PerlIO_printf( Perl_debug_log, "%*s",
1047 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
1048 PL_colors[0], PL_colors[1],
1049 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
1050 PERL_PV_ESCAPE_FIRSTCHAR
1056 PerlIO_printf( Perl_debug_log, "\n%*sState+-",(int)depth * 2 + 2,"" );
1058 for( charid=0 ; charid < trie->uniquecharcount ; charid++ ) {
1059 PerlIO_printf( Perl_debug_log, "%.*s", colwidth,"--------");
1062 PerlIO_printf( Perl_debug_log, "\n" );
1064 for( state=1 ; state < next_alloc ; state += trie->uniquecharcount ) {
1066 PerlIO_printf( Perl_debug_log, "%*s%4"UVXf" : ",
1067 (int)depth * 2 + 2,"",
1068 (UV)TRIE_NODENUM( state ) );
1070 for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) {
1071 UV v=(UV)SAFE_TRIE_NODENUM( trie->trans[ state + charid ].next );
1073 PerlIO_printf( Perl_debug_log, "%*"UVXf, colwidth, v );
1075 PerlIO_printf( Perl_debug_log, "%*s", colwidth, "." );
1077 if ( ! trie->states[ TRIE_NODENUM( state ) ].wordnum ) {
1078 PerlIO_printf( Perl_debug_log, " (%4"UVXf")\n", (UV)trie->trans[ state ].check );
1080 PerlIO_printf( Perl_debug_log, " (%4"UVXf") W%4X\n", (UV)trie->trans[ state ].check,
1081 trie->states[ TRIE_NODENUM( state ) ].wordnum );
1089 /* make_trie(startbranch,first,last,tail,word_count,flags,depth)
1090 startbranch: the first branch in the whole branch sequence
1091 first : start branch of sequence of branch-exact nodes.
1092 May be the same as startbranch
1093 last : Thing following the last branch.
1094 May be the same as tail.
1095 tail : item following the branch sequence
1096 count : words in the sequence
1097 flags : currently the OP() type we will be building one of /EXACT(|F|Fl)/
1098 depth : indent depth
1100 Inplace optimizes a sequence of 2 or more Branch-Exact nodes into a TRIE node.
1102 A trie is an N'ary tree where the branches are determined by digital
1103 decomposition of the key. IE, at the root node you look up the 1st character and
1104 follow that branch repeat until you find the end of the branches. Nodes can be
1105 marked as "accepting" meaning they represent a complete word. Eg:
1109 would convert into the following structure. Numbers represent states, letters
1110 following numbers represent valid transitions on the letter from that state, if
1111 the number is in square brackets it represents an accepting state, otherwise it
1112 will be in parenthesis.
1114 +-h->+-e->[3]-+-r->(8)-+-s->[9]
1118 (1) +-i->(6)-+-s->[7]
1120 +-s->(3)-+-h->(4)-+-e->[5]
1122 Accept Word Mapping: 3=>1 (he),5=>2 (she), 7=>3 (his), 9=>4 (hers)
1124 This shows that when matching against the string 'hers' we will begin at state 1
1125 read 'h' and move to state 2, read 'e' and move to state 3 which is accepting,
1126 then read 'r' and go to state 8 followed by 's' which takes us to state 9 which
1127 is also accepting. Thus we know that we can match both 'he' and 'hers' with a
1128 single traverse. We store a mapping from accepting to state to which word was
1129 matched, and then when we have multiple possibilities we try to complete the
1130 rest of the regex in the order in which they occured in the alternation.
1132 The only prior NFA like behaviour that would be changed by the TRIE support is
1133 the silent ignoring of duplicate alternations which are of the form:
1135 / (DUPE|DUPE) X? (?{ ... }) Y /x
1137 Thus EVAL blocks follwing a trie may be called a different number of times with
1138 and without the optimisation. With the optimisations dupes will be silently
1139 ignored. This inconsistant behaviour of EVAL type nodes is well established as
1140 the following demonstrates:
1142 'words'=~/(word|word|word)(?{ print $1 })[xyz]/
1144 which prints out 'word' three times, but
1146 'words'=~/(word|word|word)(?{ print $1 })S/
1148 which doesnt print it out at all. This is due to other optimisations kicking in.
1150 Example of what happens on a structural level:
1152 The regexp /(ac|ad|ab)+/ will produce the folowing debug output:
1154 1: CURLYM[1] {1,32767}(18)
1165 This would be optimizable with startbranch=5, first=5, last=16, tail=16
1166 and should turn into:
1168 1: CURLYM[1] {1,32767}(18)
1170 [Words:3 Chars Stored:6 Unique Chars:4 States:5 NCP:1]
1178 Cases where tail != last would be like /(?foo|bar)baz/:
1188 which would be optimizable with startbranch=1, first=1, last=7, tail=8
1189 and would end up looking like:
1192 [Words:2 Chars Stored:6 Unique Chars:5 States:7 NCP:1]
1199 d = uvuni_to_utf8_flags(d, uv, 0);
1201 is the recommended Unicode-aware way of saying
1206 #define TRIE_STORE_REVCHAR \
1209 SV *zlopp = newSV(2); \
1210 unsigned char *flrbbbbb = (unsigned char *) SvPVX(zlopp); \
1211 unsigned const char *const kapow = uvuni_to_utf8(flrbbbbb, uvc & 0xFF); \
1212 SvCUR_set(zlopp, kapow - flrbbbbb); \
1215 av_push(revcharmap, zlopp); \
1217 char ooooff = (char)uvc; \
1218 av_push(revcharmap, newSVpvn(&ooooff, 1)); \
1222 #define TRIE_READ_CHAR STMT_START { \
1226 if ( foldlen > 0 ) { \
1227 uvc = utf8n_to_uvuni( scan, UTF8_MAXLEN, &len, uniflags ); \
1232 uvc = utf8n_to_uvuni( (const U8*)uc, UTF8_MAXLEN, &len, uniflags);\
1233 uvc = to_uni_fold( uvc, foldbuf, &foldlen ); \
1234 foldlen -= UNISKIP( uvc ); \
1235 scan = foldbuf + UNISKIP( uvc ); \
1238 uvc = utf8n_to_uvuni( (const U8*)uc, UTF8_MAXLEN, &len, uniflags);\
1248 #define TRIE_LIST_PUSH(state,fid,ns) STMT_START { \
1249 if ( TRIE_LIST_CUR( state ) >=TRIE_LIST_LEN( state ) ) { \
1250 U32 ging = TRIE_LIST_LEN( state ) *= 2; \
1251 Renew( trie->states[ state ].trans.list, ging, reg_trie_trans_le ); \
1253 TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).forid = fid; \
1254 TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).newstate = ns; \
1255 TRIE_LIST_CUR( state )++; \
1258 #define TRIE_LIST_NEW(state) STMT_START { \
1259 Newxz( trie->states[ state ].trans.list, \
1260 4, reg_trie_trans_le ); \
1261 TRIE_LIST_CUR( state ) = 1; \
1262 TRIE_LIST_LEN( state ) = 4; \
1265 #define TRIE_HANDLE_WORD(state) STMT_START { \
1266 U16 dupe= trie->states[ state ].wordnum; \
1267 regnode * const noper_next = regnext( noper ); \
1270 /* store the word for dumping */ \
1272 if (OP(noper) != NOTHING) \
1273 tmp = newSVpvn_utf8(STRING(noper), STR_LEN(noper), UTF); \
1275 tmp = newSVpvn_utf8( "", 0, UTF ); \
1276 av_push( trie_words, tmp ); \
1280 trie->wordinfo[curword].prev = 0; \
1281 trie->wordinfo[curword].len = wordlen; \
1282 trie->wordinfo[curword].accept = state; \
1284 if ( noper_next < tail ) { \
1286 trie->jump = (U16 *) PerlMemShared_calloc( word_count + 1, sizeof(U16) ); \
1287 trie->jump[curword] = (U16)(noper_next - convert); \
1289 jumper = noper_next; \
1291 nextbranch= regnext(cur); \
1295 /* It's a dupe. Pre-insert into the wordinfo[].prev */\
1296 /* chain, so that when the bits of chain are later */\
1297 /* linked together, the dups appear in the chain */\
1298 trie->wordinfo[curword].prev = trie->wordinfo[dupe].prev; \
1299 trie->wordinfo[dupe].prev = curword; \
1301 /* we haven't inserted this word yet. */ \
1302 trie->states[ state ].wordnum = curword; \
1307 #define TRIE_TRANS_STATE(state,base,ucharcount,charid,special) \
1308 ( ( base + charid >= ucharcount \
1309 && base + charid < ubound \
1310 && state == trie->trans[ base - ucharcount + charid ].check \
1311 && trie->trans[ base - ucharcount + charid ].next ) \
1312 ? trie->trans[ base - ucharcount + charid ].next \
1313 : ( state==1 ? special : 0 ) \
1317 #define MADE_JUMP_TRIE 2
1318 #define MADE_EXACT_TRIE 4
1321 S_make_trie(pTHX_ RExC_state_t *pRExC_state, regnode *startbranch, regnode *first, regnode *last, regnode *tail, U32 word_count, U32 flags, U32 depth)
1324 /* first pass, loop through and scan words */
1325 reg_trie_data *trie;
1326 HV *widecharmap = NULL;
1327 AV *revcharmap = newAV();
1329 const U32 uniflags = UTF8_ALLOW_DEFAULT;
1334 regnode *jumper = NULL;
1335 regnode *nextbranch = NULL;
1336 regnode *convert = NULL;
1337 U32 *prev_states; /* temp array mapping each state to previous one */
1338 /* we just use folder as a flag in utf8 */
1339 const U8 * const folder = ( flags == EXACTF
1341 : ( flags == EXACTFL
1348 const U32 data_slot = add_data( pRExC_state, 4, "tuuu" );
1349 AV *trie_words = NULL;
1350 /* along with revcharmap, this only used during construction but both are
1351 * useful during debugging so we store them in the struct when debugging.
1354 const U32 data_slot = add_data( pRExC_state, 2, "tu" );
1355 STRLEN trie_charcount=0;
1357 SV *re_trie_maxbuff;
1358 GET_RE_DEBUG_FLAGS_DECL;
1360 PERL_ARGS_ASSERT_MAKE_TRIE;
1362 PERL_UNUSED_ARG(depth);
1365 trie = (reg_trie_data *) PerlMemShared_calloc( 1, sizeof(reg_trie_data) );
1367 trie->startstate = 1;
1368 trie->wordcount = word_count;
1369 RExC_rxi->data->data[ data_slot ] = (void*)trie;
1370 trie->charmap = (U16 *) PerlMemShared_calloc( 256, sizeof(U16) );
1371 if (!(UTF && folder))
1372 trie->bitmap = (char *) PerlMemShared_calloc( ANYOF_BITMAP_SIZE, 1 );
1373 trie->wordinfo = (reg_trie_wordinfo *) PerlMemShared_calloc(
1374 trie->wordcount+1, sizeof(reg_trie_wordinfo));
1377 trie_words = newAV();
1380 re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, 1);
1381 if (!SvIOK(re_trie_maxbuff)) {
1382 sv_setiv(re_trie_maxbuff, RE_TRIE_MAXBUF_INIT);
1385 PerlIO_printf( Perl_debug_log,
1386 "%*smake_trie start==%d, first==%d, last==%d, tail==%d depth=%d\n",
1387 (int)depth * 2 + 2, "",
1388 REG_NODE_NUM(startbranch),REG_NODE_NUM(first),
1389 REG_NODE_NUM(last), REG_NODE_NUM(tail),
1393 /* Find the node we are going to overwrite */
1394 if ( first == startbranch && OP( last ) != BRANCH ) {
1395 /* whole branch chain */
1398 /* branch sub-chain */
1399 convert = NEXTOPER( first );
1402 /* -- First loop and Setup --
1404 We first traverse the branches and scan each word to determine if it
1405 contains widechars, and how many unique chars there are, this is
1406 important as we have to build a table with at least as many columns as we
1409 We use an array of integers to represent the character codes 0..255
1410 (trie->charmap) and we use a an HV* to store Unicode characters. We use the
1411 native representation of the character value as the key and IV's for the
1414 *TODO* If we keep track of how many times each character is used we can
1415 remap the columns so that the table compression later on is more
1416 efficient in terms of memory by ensuring most common value is in the
1417 middle and the least common are on the outside. IMO this would be better
1418 than a most to least common mapping as theres a decent chance the most
1419 common letter will share a node with the least common, meaning the node
1420 will not be compressable. With a middle is most common approach the worst
1421 case is when we have the least common nodes twice.
1425 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
1426 regnode * const noper = NEXTOPER( cur );
1427 const U8 *uc = (U8*)STRING( noper );
1428 const U8 * const e = uc + STR_LEN( noper );
1430 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
1431 const U8 *scan = (U8*)NULL;
1432 U32 wordlen = 0; /* required init */
1434 bool set_bit = trie->bitmap ? 1 : 0; /*store the first char in the bitmap?*/
1436 if (OP(noper) == NOTHING) {
1440 if ( set_bit ) /* bitmap only alloced when !(UTF&&Folding) */
1441 TRIE_BITMAP_SET(trie,*uc); /* store the raw first byte
1442 regardless of encoding */
1444 for ( ; uc < e ; uc += len ) {
1445 TRIE_CHARCOUNT(trie)++;
1449 if ( !trie->charmap[ uvc ] ) {
1450 trie->charmap[ uvc ]=( ++trie->uniquecharcount );
1452 trie->charmap[ folder[ uvc ] ] = trie->charmap[ uvc ];
1456 /* store the codepoint in the bitmap, and if its ascii
1457 also store its folded equivelent. */
1458 TRIE_BITMAP_SET(trie,uvc);
1460 /* store the folded codepoint */
1461 if ( folder ) TRIE_BITMAP_SET(trie,folder[ uvc ]);
1464 /* store first byte of utf8 representation of
1465 codepoints in the 127 < uvc < 256 range */
1466 if (127 < uvc && uvc < 192) {
1467 TRIE_BITMAP_SET(trie,194);
1468 } else if (191 < uvc ) {
1469 TRIE_BITMAP_SET(trie,195);
1470 /* && uvc < 256 -- we know uvc is < 256 already */
1473 set_bit = 0; /* We've done our bit :-) */
1478 widecharmap = newHV();
1480 svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 1 );
1483 Perl_croak( aTHX_ "error creating/fetching widecharmap entry for 0x%"UVXf, uvc );
1485 if ( !SvTRUE( *svpp ) ) {
1486 sv_setiv( *svpp, ++trie->uniquecharcount );
1491 if( cur == first ) {
1494 } else if (chars < trie->minlen) {
1496 } else if (chars > trie->maxlen) {
1500 } /* end first pass */
1501 DEBUG_TRIE_COMPILE_r(
1502 PerlIO_printf( Perl_debug_log, "%*sTRIE(%s): W:%d C:%d Uq:%d Min:%d Max:%d\n",
1503 (int)depth * 2 + 2,"",
1504 ( widecharmap ? "UTF8" : "NATIVE" ), (int)word_count,
1505 (int)TRIE_CHARCOUNT(trie), trie->uniquecharcount,
1506 (int)trie->minlen, (int)trie->maxlen )
1510 We now know what we are dealing with in terms of unique chars and
1511 string sizes so we can calculate how much memory a naive
1512 representation using a flat table will take. If it's over a reasonable
1513 limit (as specified by ${^RE_TRIE_MAXBUF}) we use a more memory
1514 conservative but potentially much slower representation using an array
1517 At the end we convert both representations into the same compressed
1518 form that will be used in regexec.c for matching with. The latter
1519 is a form that cannot be used to construct with but has memory
1520 properties similar to the list form and access properties similar
1521 to the table form making it both suitable for fast searches and
1522 small enough that its feasable to store for the duration of a program.
1524 See the comment in the code where the compressed table is produced
1525 inplace from the flat tabe representation for an explanation of how
1526 the compression works.
1531 Newx(prev_states, TRIE_CHARCOUNT(trie) + 2, U32);
1534 if ( (IV)( ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount + 1) > SvIV(re_trie_maxbuff) ) {
1536 Second Pass -- Array Of Lists Representation
1538 Each state will be represented by a list of charid:state records
1539 (reg_trie_trans_le) the first such element holds the CUR and LEN
1540 points of the allocated array. (See defines above).
1542 We build the initial structure using the lists, and then convert
1543 it into the compressed table form which allows faster lookups
1544 (but cant be modified once converted).
1547 STRLEN transcount = 1;
1549 DEBUG_TRIE_COMPILE_MORE_r( PerlIO_printf( Perl_debug_log,
1550 "%*sCompiling trie using list compiler\n",
1551 (int)depth * 2 + 2, ""));
1553 trie->states = (reg_trie_state *)
1554 PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2,
1555 sizeof(reg_trie_state) );
1559 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
1561 regnode * const noper = NEXTOPER( cur );
1562 U8 *uc = (U8*)STRING( noper );
1563 const U8 * const e = uc + STR_LEN( noper );
1564 U32 state = 1; /* required init */
1565 U16 charid = 0; /* sanity init */
1566 U8 *scan = (U8*)NULL; /* sanity init */
1567 STRLEN foldlen = 0; /* required init */
1568 U32 wordlen = 0; /* required init */
1569 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
1571 if (OP(noper) != NOTHING) {
1572 for ( ; uc < e ; uc += len ) {
1577 charid = trie->charmap[ uvc ];
1579 SV** const svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 0);
1583 charid=(U16)SvIV( *svpp );
1586 /* charid is now 0 if we dont know the char read, or nonzero if we do */
1593 if ( !trie->states[ state ].trans.list ) {
1594 TRIE_LIST_NEW( state );
1596 for ( check = 1; check <= TRIE_LIST_USED( state ); check++ ) {
1597 if ( TRIE_LIST_ITEM( state, check ).forid == charid ) {
1598 newstate = TRIE_LIST_ITEM( state, check ).newstate;
1603 newstate = next_alloc++;
1604 prev_states[newstate] = state;
1605 TRIE_LIST_PUSH( state, charid, newstate );
1610 Perl_croak( aTHX_ "panic! In trie construction, no char mapping for %"IVdf, uvc );
1614 TRIE_HANDLE_WORD(state);
1616 } /* end second pass */
1618 /* next alloc is the NEXT state to be allocated */
1619 trie->statecount = next_alloc;
1620 trie->states = (reg_trie_state *)
1621 PerlMemShared_realloc( trie->states,
1623 * sizeof(reg_trie_state) );
1625 /* and now dump it out before we compress it */
1626 DEBUG_TRIE_COMPILE_MORE_r(dump_trie_interim_list(trie, widecharmap,
1627 revcharmap, next_alloc,
1631 trie->trans = (reg_trie_trans *)
1632 PerlMemShared_calloc( transcount, sizeof(reg_trie_trans) );
1639 for( state=1 ; state < next_alloc ; state ++ ) {
1643 DEBUG_TRIE_COMPILE_MORE_r(
1644 PerlIO_printf( Perl_debug_log, "tp: %d zp: %d ",tp,zp)
1648 if (trie->states[state].trans.list) {
1649 U16 minid=TRIE_LIST_ITEM( state, 1).forid;
1653 for( idx = 2 ; idx <= TRIE_LIST_USED( state ) ; idx++ ) {
1654 const U16 forid = TRIE_LIST_ITEM( state, idx).forid;
1655 if ( forid < minid ) {
1657 } else if ( forid > maxid ) {
1661 if ( transcount < tp + maxid - minid + 1) {
1663 trie->trans = (reg_trie_trans *)
1664 PerlMemShared_realloc( trie->trans,
1666 * sizeof(reg_trie_trans) );
1667 Zero( trie->trans + (transcount / 2), transcount / 2 , reg_trie_trans );
1669 base = trie->uniquecharcount + tp - minid;
1670 if ( maxid == minid ) {
1672 for ( ; zp < tp ; zp++ ) {
1673 if ( ! trie->trans[ zp ].next ) {
1674 base = trie->uniquecharcount + zp - minid;
1675 trie->trans[ zp ].next = TRIE_LIST_ITEM( state, 1).newstate;
1676 trie->trans[ zp ].check = state;
1682 trie->trans[ tp ].next = TRIE_LIST_ITEM( state, 1).newstate;
1683 trie->trans[ tp ].check = state;
1688 for ( idx=1; idx <= TRIE_LIST_USED( state ) ; idx++ ) {
1689 const U32 tid = base - trie->uniquecharcount + TRIE_LIST_ITEM( state, idx ).forid;
1690 trie->trans[ tid ].next = TRIE_LIST_ITEM( state, idx ).newstate;
1691 trie->trans[ tid ].check = state;
1693 tp += ( maxid - minid + 1 );
1695 Safefree(trie->states[ state ].trans.list);
1698 DEBUG_TRIE_COMPILE_MORE_r(
1699 PerlIO_printf( Perl_debug_log, " base: %d\n",base);
1702 trie->states[ state ].trans.base=base;
1704 trie->lasttrans = tp + 1;
1708 Second Pass -- Flat Table Representation.
1710 we dont use the 0 slot of either trans[] or states[] so we add 1 to each.
1711 We know that we will need Charcount+1 trans at most to store the data
1712 (one row per char at worst case) So we preallocate both structures
1713 assuming worst case.
1715 We then construct the trie using only the .next slots of the entry
1718 We use the .check field of the first entry of the node temporarily to
1719 make compression both faster and easier by keeping track of how many non
1720 zero fields are in the node.
1722 Since trans are numbered from 1 any 0 pointer in the table is a FAIL
1725 There are two terms at use here: state as a TRIE_NODEIDX() which is a
1726 number representing the first entry of the node, and state as a
1727 TRIE_NODENUM() which is the trans number. state 1 is TRIE_NODEIDX(1) and
1728 TRIE_NODENUM(1), state 2 is TRIE_NODEIDX(2) and TRIE_NODENUM(3) if there
1729 are 2 entrys per node. eg:
1737 The table is internally in the right hand, idx form. However as we also
1738 have to deal with the states array which is indexed by nodenum we have to
1739 use TRIE_NODENUM() to convert.
1742 DEBUG_TRIE_COMPILE_MORE_r( PerlIO_printf( Perl_debug_log,
1743 "%*sCompiling trie using table compiler\n",
1744 (int)depth * 2 + 2, ""));
1746 trie->trans = (reg_trie_trans *)
1747 PerlMemShared_calloc( ( TRIE_CHARCOUNT(trie) + 1 )
1748 * trie->uniquecharcount + 1,
1749 sizeof(reg_trie_trans) );
1750 trie->states = (reg_trie_state *)
1751 PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2,
1752 sizeof(reg_trie_state) );
1753 next_alloc = trie->uniquecharcount + 1;
1756 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
1758 regnode * const noper = NEXTOPER( cur );
1759 const U8 *uc = (U8*)STRING( noper );
1760 const U8 * const e = uc + STR_LEN( noper );
1762 U32 state = 1; /* required init */
1764 U16 charid = 0; /* sanity init */
1765 U32 accept_state = 0; /* sanity init */
1766 U8 *scan = (U8*)NULL; /* sanity init */
1768 STRLEN foldlen = 0; /* required init */
1769 U32 wordlen = 0; /* required init */
1770 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
1772 if ( OP(noper) != NOTHING ) {
1773 for ( ; uc < e ; uc += len ) {
1778 charid = trie->charmap[ uvc ];
1780 SV* const * const svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 0);
1781 charid = svpp ? (U16)SvIV(*svpp) : 0;
1785 if ( !trie->trans[ state + charid ].next ) {
1786 trie->trans[ state + charid ].next = next_alloc;
1787 trie->trans[ state ].check++;
1788 prev_states[TRIE_NODENUM(next_alloc)]
1789 = TRIE_NODENUM(state);
1790 next_alloc += trie->uniquecharcount;
1792 state = trie->trans[ state + charid ].next;
1794 Perl_croak( aTHX_ "panic! In trie construction, no char mapping for %"IVdf, uvc );
1796 /* charid is now 0 if we dont know the char read, or nonzero if we do */
1799 accept_state = TRIE_NODENUM( state );
1800 TRIE_HANDLE_WORD(accept_state);
1802 } /* end second pass */
1804 /* and now dump it out before we compress it */
1805 DEBUG_TRIE_COMPILE_MORE_r(dump_trie_interim_table(trie, widecharmap,
1807 next_alloc, depth+1));
1811 * Inplace compress the table.*
1813 For sparse data sets the table constructed by the trie algorithm will
1814 be mostly 0/FAIL transitions or to put it another way mostly empty.
1815 (Note that leaf nodes will not contain any transitions.)
1817 This algorithm compresses the tables by eliminating most such
1818 transitions, at the cost of a modest bit of extra work during lookup:
1820 - Each states[] entry contains a .base field which indicates the
1821 index in the state[] array wheres its transition data is stored.
1823 - If .base is 0 there are no valid transitions from that node.
1825 - If .base is nonzero then charid is added to it to find an entry in
1828 -If trans[states[state].base+charid].check!=state then the
1829 transition is taken to be a 0/Fail transition. Thus if there are fail
1830 transitions at the front of the node then the .base offset will point
1831 somewhere inside the previous nodes data (or maybe even into a node
1832 even earlier), but the .check field determines if the transition is
1836 The following process inplace converts the table to the compressed
1837 table: We first do not compress the root node 1,and mark its all its
1838 .check pointers as 1 and set its .base pointer as 1 as well. This
1839 allows to do a DFA construction from the compressed table later, and
1840 ensures that any .base pointers we calculate later are greater than
1843 - We set 'pos' to indicate the first entry of the second node.
1845 - We then iterate over the columns of the node, finding the first and
1846 last used entry at l and m. We then copy l..m into pos..(pos+m-l),
1847 and set the .check pointers accordingly, and advance pos
1848 appropriately and repreat for the next node. Note that when we copy
1849 the next pointers we have to convert them from the original
1850 NODEIDX form to NODENUM form as the former is not valid post
1853 - If a node has no transitions used we mark its base as 0 and do not
1854 advance the pos pointer.
1856 - If a node only has one transition we use a second pointer into the
1857 structure to fill in allocated fail transitions from other states.
1858 This pointer is independent of the main pointer and scans forward
1859 looking for null transitions that are allocated to a state. When it
1860 finds one it writes the single transition into the "hole". If the
1861 pointer doesnt find one the single transition is appended as normal.
1863 - Once compressed we can Renew/realloc the structures to release the
1866 See "Table-Compression Methods" in sec 3.9 of the Red Dragon,
1867 specifically Fig 3.47 and the associated pseudocode.
1871 const U32 laststate = TRIE_NODENUM( next_alloc );
1874 trie->statecount = laststate;
1876 for ( state = 1 ; state < laststate ; state++ ) {
1878 const U32 stateidx = TRIE_NODEIDX( state );
1879 const U32 o_used = trie->trans[ stateidx ].check;
1880 U32 used = trie->trans[ stateidx ].check;
1881 trie->trans[ stateidx ].check = 0;
1883 for ( charid = 0 ; used && charid < trie->uniquecharcount ; charid++ ) {
1884 if ( flag || trie->trans[ stateidx + charid ].next ) {
1885 if ( trie->trans[ stateidx + charid ].next ) {
1887 for ( ; zp < pos ; zp++ ) {
1888 if ( ! trie->trans[ zp ].next ) {
1892 trie->states[ state ].trans.base = zp + trie->uniquecharcount - charid ;
1893 trie->trans[ zp ].next = SAFE_TRIE_NODENUM( trie->trans[ stateidx + charid ].next );
1894 trie->trans[ zp ].check = state;
1895 if ( ++zp > pos ) pos = zp;
1902 trie->states[ state ].trans.base = pos + trie->uniquecharcount - charid ;
1904 trie->trans[ pos ].next = SAFE_TRIE_NODENUM( trie->trans[ stateidx + charid ].next );
1905 trie->trans[ pos ].check = state;
1910 trie->lasttrans = pos + 1;
1911 trie->states = (reg_trie_state *)
1912 PerlMemShared_realloc( trie->states, laststate
1913 * sizeof(reg_trie_state) );
1914 DEBUG_TRIE_COMPILE_MORE_r(
1915 PerlIO_printf( Perl_debug_log,
1916 "%*sAlloc: %d Orig: %"IVdf" elements, Final:%"IVdf". Savings of %%%5.2f\n",
1917 (int)depth * 2 + 2,"",
1918 (int)( ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount + 1 ),
1921 ( ( next_alloc - pos ) * 100 ) / (double)next_alloc );
1924 } /* end table compress */
1926 DEBUG_TRIE_COMPILE_MORE_r(
1927 PerlIO_printf(Perl_debug_log, "%*sStatecount:%"UVxf" Lasttrans:%"UVxf"\n",
1928 (int)depth * 2 + 2, "",
1929 (UV)trie->statecount,
1930 (UV)trie->lasttrans)
1932 /* resize the trans array to remove unused space */
1933 trie->trans = (reg_trie_trans *)
1934 PerlMemShared_realloc( trie->trans, trie->lasttrans
1935 * sizeof(reg_trie_trans) );
1937 { /* Modify the program and insert the new TRIE node*/
1938 U8 nodetype =(U8)(flags & 0xFF);
1942 regnode *optimize = NULL;
1943 #ifdef RE_TRACK_PATTERN_OFFSETS
1946 U32 mjd_nodelen = 0;
1947 #endif /* RE_TRACK_PATTERN_OFFSETS */
1948 #endif /* DEBUGGING */
1950 This means we convert either the first branch or the first Exact,
1951 depending on whether the thing following (in 'last') is a branch
1952 or not and whther first is the startbranch (ie is it a sub part of
1953 the alternation or is it the whole thing.)
1954 Assuming its a sub part we conver the EXACT otherwise we convert
1955 the whole branch sequence, including the first.
1957 /* Find the node we are going to overwrite */
1958 if ( first != startbranch || OP( last ) == BRANCH ) {
1959 /* branch sub-chain */
1960 NEXT_OFF( first ) = (U16)(last - first);
1961 #ifdef RE_TRACK_PATTERN_OFFSETS
1963 mjd_offset= Node_Offset((convert));
1964 mjd_nodelen= Node_Length((convert));
1967 /* whole branch chain */
1969 #ifdef RE_TRACK_PATTERN_OFFSETS
1972 const regnode *nop = NEXTOPER( convert );
1973 mjd_offset= Node_Offset((nop));
1974 mjd_nodelen= Node_Length((nop));
1978 PerlIO_printf(Perl_debug_log, "%*sMJD offset:%"UVuf" MJD length:%"UVuf"\n",
1979 (int)depth * 2 + 2, "",
1980 (UV)mjd_offset, (UV)mjd_nodelen)
1983 /* But first we check to see if there is a common prefix we can
1984 split out as an EXACT and put in front of the TRIE node. */
1985 trie->startstate= 1;
1986 if ( trie->bitmap && !widecharmap && !trie->jump ) {
1988 for ( state = 1 ; state < trie->statecount-1 ; state++ ) {
1992 const U32 base = trie->states[ state ].trans.base;
1994 if ( trie->states[state].wordnum )
1997 for ( ofs = 0 ; ofs < trie->uniquecharcount ; ofs++ ) {
1998 if ( ( base + ofs >= trie->uniquecharcount ) &&
1999 ( base + ofs - trie->uniquecharcount < trie->lasttrans ) &&
2000 trie->trans[ base + ofs - trie->uniquecharcount ].check == state )
2002 if ( ++count > 1 ) {
2003 SV **tmp = av_fetch( revcharmap, ofs, 0);
2004 const U8 *ch = (U8*)SvPV_nolen_const( *tmp );
2005 if ( state == 1 ) break;
2007 Zero(trie->bitmap, ANYOF_BITMAP_SIZE, char);
2009 PerlIO_printf(Perl_debug_log,
2010 "%*sNew Start State=%"UVuf" Class: [",
2011 (int)depth * 2 + 2, "",
2014 SV ** const tmp = av_fetch( revcharmap, idx, 0);
2015 const U8 * const ch = (U8*)SvPV_nolen_const( *tmp );
2017 TRIE_BITMAP_SET(trie,*ch);
2019 TRIE_BITMAP_SET(trie, folder[ *ch ]);
2021 PerlIO_printf(Perl_debug_log, "%s", (char*)ch)
2025 TRIE_BITMAP_SET(trie,*ch);
2027 TRIE_BITMAP_SET(trie,folder[ *ch ]);
2028 DEBUG_OPTIMISE_r(PerlIO_printf( Perl_debug_log,"%s", ch));
2034 SV **tmp = av_fetch( revcharmap, idx, 0);
2036 char *ch = SvPV( *tmp, len );
2038 SV *sv=sv_newmortal();
2039 PerlIO_printf( Perl_debug_log,
2040 "%*sPrefix State: %"UVuf" Idx:%"UVuf" Char='%s'\n",
2041 (int)depth * 2 + 2, "",
2043 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 6,
2044 PL_colors[0], PL_colors[1],
2045 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
2046 PERL_PV_ESCAPE_FIRSTCHAR
2051 OP( convert ) = nodetype;
2052 str=STRING(convert);
2055 STR_LEN(convert) += len;
2061 DEBUG_OPTIMISE_r(PerlIO_printf( Perl_debug_log,"]\n"));
2066 trie->prefixlen = (state-1);
2068 regnode *n = convert+NODE_SZ_STR(convert);
2069 NEXT_OFF(convert) = NODE_SZ_STR(convert);
2070 trie->startstate = state;
2071 trie->minlen -= (state - 1);
2072 trie->maxlen -= (state - 1);
2074 /* At least the UNICOS C compiler choked on this
2075 * being argument to DEBUG_r(), so let's just have
2078 #ifdef PERL_EXT_RE_BUILD
2084 regnode *fix = convert;
2085 U32 word = trie->wordcount;
2087 Set_Node_Offset_Length(convert, mjd_offset, state - 1);
2088 while( ++fix < n ) {
2089 Set_Node_Offset_Length(fix, 0, 0);
2092 SV ** const tmp = av_fetch( trie_words, word, 0 );
2094 if ( STR_LEN(convert) <= SvCUR(*tmp) )
2095 sv_chop(*tmp, SvPV_nolen(*tmp) + STR_LEN(convert));
2097 sv_chop(*tmp, SvPV_nolen(*tmp) + SvCUR(*tmp));
2105 NEXT_OFF(convert) = (U16)(tail - convert);
2106 DEBUG_r(optimize= n);
2112 if ( trie->maxlen ) {
2113 NEXT_OFF( convert ) = (U16)(tail - convert);
2114 ARG_SET( convert, data_slot );
2115 /* Store the offset to the first unabsorbed branch in
2116 jump[0], which is otherwise unused by the jump logic.
2117 We use this when dumping a trie and during optimisation. */
2119 trie->jump[0] = (U16)(nextbranch - convert);
2122 if ( !trie->states[trie->startstate].wordnum && trie->bitmap &&
2123 ( (char *)jumper - (char *)convert) >= (int)sizeof(struct regnode_charclass) )
2125 OP( convert ) = TRIEC;
2126 Copy(trie->bitmap, ((struct regnode_charclass *)convert)->bitmap, ANYOF_BITMAP_SIZE, char);
2127 PerlMemShared_free(trie->bitmap);
2130 OP( convert ) = TRIE;
2132 /* store the type in the flags */
2133 convert->flags = nodetype;
2137 + regarglen[ OP( convert ) ];
2139 /* XXX We really should free up the resource in trie now,
2140 as we won't use them - (which resources?) dmq */
2142 /* needed for dumping*/
2143 DEBUG_r(if (optimize) {
2144 regnode *opt = convert;
2146 while ( ++opt < optimize) {
2147 Set_Node_Offset_Length(opt,0,0);
2150 Try to clean up some of the debris left after the
2153 while( optimize < jumper ) {
2154 mjd_nodelen += Node_Length((optimize));
2155 OP( optimize ) = OPTIMIZED;
2156 Set_Node_Offset_Length(optimize,0,0);
2159 Set_Node_Offset_Length(convert,mjd_offset,mjd_nodelen);
2161 } /* end node insert */
2162 REH_CALL_COMP_NODE_HOOK(pRExC_state->rx, convert);
2164 /* Finish populating the prev field of the wordinfo array. Walk back
2165 * from each accept state until we find another accept state, and if
2166 * so, point the first word's .prev field at the second word. If the
2167 * second already has a .prev field set, stop now. This will be the
2168 * case either if we've already processed that word's accept state,
2169 * or that that state had multiple words, and the overspill words
2170 * were already linked up earlier.
2177 for (word=1; word <= trie->wordcount; word++) {
2179 if (trie->wordinfo[word].prev)
2181 state = trie->wordinfo[word].accept;
2183 state = prev_states[state];
2186 prev = trie->states[state].wordnum;
2190 trie->wordinfo[word].prev = prev;
2192 Safefree(prev_states);
2196 /* and now dump out the compressed format */
2197 DEBUG_TRIE_COMPILE_r(dump_trie(trie, widecharmap, revcharmap, depth+1));
2199 RExC_rxi->data->data[ data_slot + 1 ] = (void*)widecharmap;
2201 RExC_rxi->data->data[ data_slot + TRIE_WORDS_OFFSET ] = (void*)trie_words;
2202 RExC_rxi->data->data[ data_slot + 3 ] = (void*)revcharmap;
2204 SvREFCNT_dec(revcharmap);
2208 : trie->startstate>1
2214 S_make_trie_failtable(pTHX_ RExC_state_t *pRExC_state, regnode *source, regnode *stclass, U32 depth)
2216 /* The Trie is constructed and compressed now so we can build a fail array now if its needed
2218 This is basically the Aho-Corasick algorithm. Its from exercise 3.31 and 3.32 in the
2219 "Red Dragon" -- Compilers, principles, techniques, and tools. Aho, Sethi, Ullman 1985/88
2222 We find the fail state for each state in the trie, this state is the longest proper
2223 suffix of the current states 'word' that is also a proper prefix of another word in our
2224 trie. State 1 represents the word '' and is the thus the default fail state. This allows
2225 the DFA not to have to restart after its tried and failed a word at a given point, it
2226 simply continues as though it had been matching the other word in the first place.
2228 'abcdgu'=~/abcdefg|cdgu/
2229 When we get to 'd' we are still matching the first word, we would encounter 'g' which would
2230 fail, which would bring use to the state representing 'd' in the second word where we would
2231 try 'g' and succeed, prodceding to match 'cdgu'.
2233 /* add a fail transition */
2234 const U32 trie_offset = ARG(source);
2235 reg_trie_data *trie=(reg_trie_data *)RExC_rxi->data->data[trie_offset];
2237 const U32 ucharcount = trie->uniquecharcount;
2238 const U32 numstates = trie->statecount;
2239 const U32 ubound = trie->lasttrans + ucharcount;
2243 U32 base = trie->states[ 1 ].trans.base;
2246 const U32 data_slot = add_data( pRExC_state, 1, "T" );
2247 GET_RE_DEBUG_FLAGS_DECL;
2249 PERL_ARGS_ASSERT_MAKE_TRIE_FAILTABLE;
2251 PERL_UNUSED_ARG(depth);
2255 ARG_SET( stclass, data_slot );
2256 aho = (reg_ac_data *) PerlMemShared_calloc( 1, sizeof(reg_ac_data) );
2257 RExC_rxi->data->data[ data_slot ] = (void*)aho;
2258 aho->trie=trie_offset;
2259 aho->states=(reg_trie_state *)PerlMemShared_malloc( numstates * sizeof(reg_trie_state) );
2260 Copy( trie->states, aho->states, numstates, reg_trie_state );
2261 Newxz( q, numstates, U32);
2262 aho->fail = (U32 *) PerlMemShared_calloc( numstates, sizeof(U32) );
2265 /* initialize fail[0..1] to be 1 so that we always have
2266 a valid final fail state */
2267 fail[ 0 ] = fail[ 1 ] = 1;
2269 for ( charid = 0; charid < ucharcount ; charid++ ) {
2270 const U32 newstate = TRIE_TRANS_STATE( 1, base, ucharcount, charid, 0 );
2272 q[ q_write ] = newstate;
2273 /* set to point at the root */
2274 fail[ q[ q_write++ ] ]=1;
2277 while ( q_read < q_write) {
2278 const U32 cur = q[ q_read++ % numstates ];
2279 base = trie->states[ cur ].trans.base;
2281 for ( charid = 0 ; charid < ucharcount ; charid++ ) {
2282 const U32 ch_state = TRIE_TRANS_STATE( cur, base, ucharcount, charid, 1 );
2284 U32 fail_state = cur;
2287 fail_state = fail[ fail_state ];
2288 fail_base = aho->states[ fail_state ].trans.base;
2289 } while ( !TRIE_TRANS_STATE( fail_state, fail_base, ucharcount, charid, 1 ) );
2291 fail_state = TRIE_TRANS_STATE( fail_state, fail_base, ucharcount, charid, 1 );
2292 fail[ ch_state ] = fail_state;
2293 if ( !aho->states[ ch_state ].wordnum && aho->states[ fail_state ].wordnum )
2295 aho->states[ ch_state ].wordnum = aho->states[ fail_state ].wordnum;
2297 q[ q_write++ % numstates] = ch_state;
2301 /* restore fail[0..1] to 0 so that we "fall out" of the AC loop
2302 when we fail in state 1, this allows us to use the
2303 charclass scan to find a valid start char. This is based on the principle
2304 that theres a good chance the string being searched contains lots of stuff
2305 that cant be a start char.
2307 fail[ 0 ] = fail[ 1 ] = 0;
2308 DEBUG_TRIE_COMPILE_r({
2309 PerlIO_printf(Perl_debug_log,
2310 "%*sStclass Failtable (%"UVuf" states): 0",
2311 (int)(depth * 2), "", (UV)numstates
2313 for( q_read=1; q_read<numstates; q_read++ ) {
2314 PerlIO_printf(Perl_debug_log, ", %"UVuf, (UV)fail[q_read]);
2316 PerlIO_printf(Perl_debug_log, "\n");
2319 /*RExC_seen |= REG_SEEN_TRIEDFA;*/
2324 * There are strange code-generation bugs caused on sparc64 by gcc-2.95.2.
2325 * These need to be revisited when a newer toolchain becomes available.
2327 #if defined(__sparc64__) && defined(__GNUC__)
2328 # if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 96)
2329 # undef SPARC64_GCC_WORKAROUND
2330 # define SPARC64_GCC_WORKAROUND 1
2334 #define DEBUG_PEEP(str,scan,depth) \
2335 DEBUG_OPTIMISE_r({if (scan){ \
2336 SV * const mysv=sv_newmortal(); \
2337 regnode *Next = regnext(scan); \
2338 regprop(RExC_rx, mysv, scan); \
2339 PerlIO_printf(Perl_debug_log, "%*s" str ">%3d: %s (%d)\n", \
2340 (int)depth*2, "", REG_NODE_NUM(scan), SvPV_nolen_const(mysv),\
2341 Next ? (REG_NODE_NUM(Next)) : 0 ); \
2348 #define JOIN_EXACT(scan,min,flags) \
2349 if (PL_regkind[OP(scan)] == EXACT) \
2350 join_exact(pRExC_state,(scan),(min),(flags),NULL,depth+1)
2353 S_join_exact(pTHX_ RExC_state_t *pRExC_state, regnode *scan, I32 *min, U32 flags,regnode *val, U32 depth) {
2354 /* Merge several consecutive EXACTish nodes into one. */
2355 regnode *n = regnext(scan);
2357 regnode *next = scan + NODE_SZ_STR(scan);
2361 regnode *stop = scan;
2362 GET_RE_DEBUG_FLAGS_DECL;
2364 PERL_UNUSED_ARG(depth);
2367 PERL_ARGS_ASSERT_JOIN_EXACT;
2368 #ifndef EXPERIMENTAL_INPLACESCAN
2369 PERL_UNUSED_ARG(flags);
2370 PERL_UNUSED_ARG(val);
2372 DEBUG_PEEP("join",scan,depth);
2374 /* Skip NOTHING, merge EXACT*. */
2376 ( PL_regkind[OP(n)] == NOTHING ||
2377 (stringok && (OP(n) == OP(scan))))
2379 && NEXT_OFF(scan) + NEXT_OFF(n) < I16_MAX) {
2381 if (OP(n) == TAIL || n > next)
2383 if (PL_regkind[OP(n)] == NOTHING) {
2384 DEBUG_PEEP("skip:",n,depth);
2385 NEXT_OFF(scan) += NEXT_OFF(n);
2386 next = n + NODE_STEP_REGNODE;
2393 else if (stringok) {
2394 const unsigned int oldl = STR_LEN(scan);
2395 regnode * const nnext = regnext(n);
2397 DEBUG_PEEP("merg",n,depth);
2400 if (oldl + STR_LEN(n) > U8_MAX)
2402 NEXT_OFF(scan) += NEXT_OFF(n);
2403 STR_LEN(scan) += STR_LEN(n);
2404 next = n + NODE_SZ_STR(n);
2405 /* Now we can overwrite *n : */
2406 Move(STRING(n), STRING(scan) + oldl, STR_LEN(n), char);
2414 #ifdef EXPERIMENTAL_INPLACESCAN
2415 if (flags && !NEXT_OFF(n)) {
2416 DEBUG_PEEP("atch", val, depth);
2417 if (reg_off_by_arg[OP(n)]) {
2418 ARG_SET(n, val - n);
2421 NEXT_OFF(n) = val - n;
2428 if (UTF && ( OP(scan) == EXACTF ) && ( STR_LEN(scan) >= 6 ) ) {
2430 Two problematic code points in Unicode casefolding of EXACT nodes:
2432 U+0390 - GREEK SMALL LETTER IOTA WITH DIALYTIKA AND TONOS
2433 U+03B0 - GREEK SMALL LETTER UPSILON WITH DIALYTIKA AND TONOS
2439 U+03B9 U+0308 U+0301 0xCE 0xB9 0xCC 0x88 0xCC 0x81
2440 U+03C5 U+0308 U+0301 0xCF 0x85 0xCC 0x88 0xCC 0x81
2442 This means that in case-insensitive matching (or "loose matching",
2443 as Unicode calls it), an EXACTF of length six (the UTF-8 encoded byte
2444 length of the above casefolded versions) can match a target string
2445 of length two (the byte length of UTF-8 encoded U+0390 or U+03B0).
2446 This would rather mess up the minimum length computation.
2448 What we'll do is to look for the tail four bytes, and then peek
2449 at the preceding two bytes to see whether we need to decrease
2450 the minimum length by four (six minus two).
2452 Thanks to the design of UTF-8, there cannot be false matches:
2453 A sequence of valid UTF-8 bytes cannot be a subsequence of
2454 another valid sequence of UTF-8 bytes.
2457 char * const s0 = STRING(scan), *s, *t;
2458 char * const s1 = s0 + STR_LEN(scan) - 1;
2459 char * const s2 = s1 - 4;
2460 #ifdef EBCDIC /* RD tunifold greek 0390 and 03B0 */
2461 const char t0[] = "\xaf\x49\xaf\x42";
2463 const char t0[] = "\xcc\x88\xcc\x81";
2465 const char * const t1 = t0 + 3;
2468 s < s2 && (t = ninstr(s, s1, t0, t1));
2471 if (((U8)t[-1] == 0x68 && (U8)t[-2] == 0xB4) ||
2472 ((U8)t[-1] == 0x46 && (U8)t[-2] == 0xB5))
2474 if (((U8)t[-1] == 0xB9 && (U8)t[-2] == 0xCE) ||
2475 ((U8)t[-1] == 0x85 && (U8)t[-2] == 0xCF))
2483 n = scan + NODE_SZ_STR(scan);
2485 if (PL_regkind[OP(n)] != NOTHING || OP(n) == NOTHING) {
2492 DEBUG_OPTIMISE_r(if (merged){DEBUG_PEEP("finl",scan,depth)});
2496 /* REx optimizer. Converts nodes into quickier variants "in place".
2497 Finds fixed substrings. */
2499 /* Stops at toplevel WHILEM as well as at "last". At end *scanp is set
2500 to the position after last scanned or to NULL. */
2502 #define INIT_AND_WITHP \
2503 assert(!and_withp); \
2504 Newx(and_withp,1,struct regnode_charclass_class); \
2505 SAVEFREEPV(and_withp)
2507 /* this is a chain of data about sub patterns we are processing that
2508 need to be handled seperately/specially in study_chunk. Its so
2509 we can simulate recursion without losing state. */
2511 typedef struct scan_frame {
2512 regnode *last; /* last node to process in this frame */
2513 regnode *next; /* next node to process when last is reached */
2514 struct scan_frame *prev; /*previous frame*/
2515 I32 stop; /* what stopparen do we use */
2519 #define SCAN_COMMIT(s, data, m) scan_commit(s, data, m, is_inf)
2521 #define CASE_SYNST_FNC(nAmE) \
2523 if (flags & SCF_DO_STCLASS_AND) { \
2524 for (value = 0; value < 256; value++) \
2525 if (!is_ ## nAmE ## _cp(value)) \
2526 ANYOF_BITMAP_CLEAR(data->start_class, value); \
2529 for (value = 0; value < 256; value++) \
2530 if (is_ ## nAmE ## _cp(value)) \
2531 ANYOF_BITMAP_SET(data->start_class, value); \
2535 if (flags & SCF_DO_STCLASS_AND) { \
2536 for (value = 0; value < 256; value++) \
2537 if (is_ ## nAmE ## _cp(value)) \
2538 ANYOF_BITMAP_CLEAR(data->start_class, value); \
2541 for (value = 0; value < 256; value++) \
2542 if (!is_ ## nAmE ## _cp(value)) \
2543 ANYOF_BITMAP_SET(data->start_class, value); \
2550 S_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp,
2551 I32 *minlenp, I32 *deltap,
2556 struct regnode_charclass_class *and_withp,
2557 U32 flags, U32 depth)
2558 /* scanp: Start here (read-write). */
2559 /* deltap: Write maxlen-minlen here. */
2560 /* last: Stop before this one. */
2561 /* data: string data about the pattern */
2562 /* stopparen: treat close N as END */
2563 /* recursed: which subroutines have we recursed into */
2564 /* and_withp: Valid if flags & SCF_DO_STCLASS_OR */
2567 I32 min = 0, pars = 0, code;
2568 regnode *scan = *scanp, *next;
2570 int is_inf = (flags & SCF_DO_SUBSTR) && (data->flags & SF_IS_INF);
2571 int is_inf_internal = 0; /* The studied chunk is infinite */
2572 I32 is_par = OP(scan) == OPEN ? ARG(scan) : 0;
2573 scan_data_t data_fake;
2574 SV *re_trie_maxbuff = NULL;
2575 regnode *first_non_open = scan;
2576 I32 stopmin = I32_MAX;
2577 scan_frame *frame = NULL;
2578 GET_RE_DEBUG_FLAGS_DECL;
2580 PERL_ARGS_ASSERT_STUDY_CHUNK;
2583 StructCopy(&zero_scan_data, &data_fake, scan_data_t);
2587 while (first_non_open && OP(first_non_open) == OPEN)
2588 first_non_open=regnext(first_non_open);
2593 while ( scan && OP(scan) != END && scan < last ){
2594 /* Peephole optimizer: */
2595 DEBUG_STUDYDATA("Peep:", data,depth);
2596 DEBUG_PEEP("Peep",scan,depth);
2597 JOIN_EXACT(scan,&min,0);
2599 /* Follow the next-chain of the current node and optimize
2600 away all the NOTHINGs from it. */
2601 if (OP(scan) != CURLYX) {
2602 const int max = (reg_off_by_arg[OP(scan)]
2604 /* I32 may be smaller than U16 on CRAYs! */
2605 : (I32_MAX < U16_MAX ? I32_MAX : U16_MAX));
2606 int off = (reg_off_by_arg[OP(scan)] ? ARG(scan) : NEXT_OFF(scan));
2610 /* Skip NOTHING and LONGJMP. */
2611 while ((n = regnext(n))
2612 && ((PL_regkind[OP(n)] == NOTHING && (noff = NEXT_OFF(n)))
2613 || ((OP(n) == LONGJMP) && (noff = ARG(n))))
2614 && off + noff < max)
2616 if (reg_off_by_arg[OP(scan)])
2619 NEXT_OFF(scan) = off;
2624 /* The principal pseudo-switch. Cannot be a switch, since we
2625 look into several different things. */
2626 if (OP(scan) == BRANCH || OP(scan) == BRANCHJ
2627 || OP(scan) == IFTHEN) {
2628 next = regnext(scan);
2630 /* demq: the op(next)==code check is to see if we have "branch-branch" AFAICT */
2632 if (OP(next) == code || code == IFTHEN) {
2633 /* NOTE - There is similar code to this block below for handling
2634 TRIE nodes on a re-study. If you change stuff here check there
2636 I32 max1 = 0, min1 = I32_MAX, num = 0;
2637 struct regnode_charclass_class accum;
2638 regnode * const startbranch=scan;
2640 if (flags & SCF_DO_SUBSTR)
2641 SCAN_COMMIT(pRExC_state, data, minlenp); /* Cannot merge strings after this. */
2642 if (flags & SCF_DO_STCLASS)
2643 cl_init_zero(pRExC_state, &accum);
2645 while (OP(scan) == code) {
2646 I32 deltanext, minnext, f = 0, fake;
2647 struct regnode_charclass_class this_class;
2650 data_fake.flags = 0;
2652 data_fake.whilem_c = data->whilem_c;
2653 data_fake.last_closep = data->last_closep;
2656 data_fake.last_closep = &fake;
2658 data_fake.pos_delta = delta;
2659 next = regnext(scan);
2660 scan = NEXTOPER(scan);
2662 scan = NEXTOPER(scan);
2663 if (flags & SCF_DO_STCLASS) {
2664 cl_init(pRExC_state, &this_class);
2665 data_fake.start_class = &this_class;
2666 f = SCF_DO_STCLASS_AND;
2668 if (flags & SCF_WHILEM_VISITED_POS)
2669 f |= SCF_WHILEM_VISITED_POS;
2671 /* we suppose the run is continuous, last=next...*/
2672 minnext = study_chunk(pRExC_state, &scan, minlenp, &deltanext,
2674 stopparen, recursed, NULL, f,depth+1);
2677 if (max1 < minnext + deltanext)
2678 max1 = minnext + deltanext;
2679 if (deltanext == I32_MAX)
2680 is_inf = is_inf_internal = 1;
2682 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
2684 if (data_fake.flags & SCF_SEEN_ACCEPT) {
2685 if ( stopmin > minnext)
2686 stopmin = min + min1;
2687 flags &= ~SCF_DO_SUBSTR;
2689 data->flags |= SCF_SEEN_ACCEPT;
2692 if (data_fake.flags & SF_HAS_EVAL)
2693 data->flags |= SF_HAS_EVAL;
2694 data->whilem_c = data_fake.whilem_c;
2696 if (flags & SCF_DO_STCLASS)
2697 cl_or(pRExC_state, &accum, &this_class);
2699 if (code == IFTHEN && num < 2) /* Empty ELSE branch */
2701 if (flags & SCF_DO_SUBSTR) {
2702 data->pos_min += min1;
2703 data->pos_delta += max1 - min1;
2704 if (max1 != min1 || is_inf)
2705 data->longest = &(data->longest_float);
2708 delta += max1 - min1;
2709 if (flags & SCF_DO_STCLASS_OR) {
2710 cl_or(pRExC_state, data->start_class, &accum);
2712 cl_and(data->start_class, and_withp);
2713 flags &= ~SCF_DO_STCLASS;
2716 else if (flags & SCF_DO_STCLASS_AND) {
2718 cl_and(data->start_class, &accum);
2719 flags &= ~SCF_DO_STCLASS;
2722 /* Switch to OR mode: cache the old value of
2723 * data->start_class */
2725 StructCopy(data->start_class, and_withp,
2726 struct regnode_charclass_class);
2727 flags &= ~SCF_DO_STCLASS_AND;
2728 StructCopy(&accum, data->start_class,
2729 struct regnode_charclass_class);
2730 flags |= SCF_DO_STCLASS_OR;
2731 data->start_class->flags |= ANYOF_EOS;
2735 if (PERL_ENABLE_TRIE_OPTIMISATION && OP( startbranch ) == BRANCH ) {
2738 Assuming this was/is a branch we are dealing with: 'scan' now
2739 points at the item that follows the branch sequence, whatever
2740 it is. We now start at the beginning of the sequence and look
2747 which would be constructed from a pattern like /A|LIST|OF|WORDS/
2749 If we can find such a subseqence we need to turn the first
2750 element into a trie and then add the subsequent branch exact
2751 strings to the trie.
2755 1. patterns where the whole set of branch can be converted.
2757 2. patterns where only a subset can be converted.
2759 In case 1 we can replace the whole set with a single regop
2760 for the trie. In case 2 we need to keep the start and end
2763 'BRANCH EXACT; BRANCH EXACT; BRANCH X'
2764 becomes BRANCH TRIE; BRANCH X;
2766 There is an additional case, that being where there is a
2767 common prefix, which gets split out into an EXACT like node
2768 preceding the TRIE node.
2770 If x(1..n)==tail then we can do a simple trie, if not we make
2771 a "jump" trie, such that when we match the appropriate word
2772 we "jump" to the appopriate tail node. Essentailly we turn
2773 a nested if into a case structure of sorts.
2778 if (!re_trie_maxbuff) {
2779 re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, 1);
2780 if (!SvIOK(re_trie_maxbuff))
2781 sv_setiv(re_trie_maxbuff, RE_TRIE_MAXBUF_INIT);
2783 if ( SvIV(re_trie_maxbuff)>=0 ) {
2785 regnode *first = (regnode *)NULL;
2786 regnode *last = (regnode *)NULL;
2787 regnode *tail = scan;
2792 SV * const mysv = sv_newmortal(); /* for dumping */
2794 /* var tail is used because there may be a TAIL
2795 regop in the way. Ie, the exacts will point to the
2796 thing following the TAIL, but the last branch will
2797 point at the TAIL. So we advance tail. If we
2798 have nested (?:) we may have to move through several
2802 while ( OP( tail ) == TAIL ) {
2803 /* this is the TAIL generated by (?:) */
2804 tail = regnext( tail );
2809 regprop(RExC_rx, mysv, tail );
2810 PerlIO_printf( Perl_debug_log, "%*s%s%s\n",
2811 (int)depth * 2 + 2, "",
2812 "Looking for TRIE'able sequences. Tail node is: ",
2813 SvPV_nolen_const( mysv )
2819 step through the branches, cur represents each
2820 branch, noper is the first thing to be matched
2821 as part of that branch and noper_next is the
2822 regnext() of that node. if noper is an EXACT
2823 and noper_next is the same as scan (our current
2824 position in the regex) then the EXACT branch is
2825 a possible optimization target. Once we have
2826 two or more consequetive such branches we can
2827 create a trie of the EXACT's contents and stich
2828 it in place. If the sequence represents all of
2829 the branches we eliminate the whole thing and
2830 replace it with a single TRIE. If it is a
2831 subsequence then we need to stitch it in. This
2832 means the first branch has to remain, and needs
2833 to be repointed at the item on the branch chain
2834 following the last branch optimized. This could
2835 be either a BRANCH, in which case the
2836 subsequence is internal, or it could be the
2837 item following the branch sequence in which
2838 case the subsequence is at the end.
2842 /* dont use tail as the end marker for this traverse */
2843 for ( cur = startbranch ; cur != scan ; cur = regnext( cur ) ) {
2844 regnode * const noper = NEXTOPER( cur );
2845 #if defined(DEBUGGING) || defined(NOJUMPTRIE)
2846 regnode * const noper_next = regnext( noper );
2850 regprop(RExC_rx, mysv, cur);
2851 PerlIO_printf( Perl_debug_log, "%*s- %s (%d)",
2852 (int)depth * 2 + 2,"", SvPV_nolen_const( mysv ), REG_NODE_NUM(cur) );
2854 regprop(RExC_rx, mysv, noper);
2855 PerlIO_printf( Perl_debug_log, " -> %s",
2856 SvPV_nolen_const(mysv));
2859 regprop(RExC_rx, mysv, noper_next );
2860 PerlIO_printf( Perl_debug_log,"\t=> %s\t",
2861 SvPV_nolen_const(mysv));
2863 PerlIO_printf( Perl_debug_log, "(First==%d,Last==%d,Cur==%d)\n",
2864 REG_NODE_NUM(first), REG_NODE_NUM(last), REG_NODE_NUM(cur) );
2866 if ( (((first && optype!=NOTHING) ? OP( noper ) == optype
2867 : PL_regkind[ OP( noper ) ] == EXACT )
2868 || OP(noper) == NOTHING )
2870 && noper_next == tail
2875 if ( !first || optype == NOTHING ) {
2876 if (!first) first = cur;
2877 optype = OP( noper );
2883 Currently we do not believe that the trie logic can
2884 handle case insensitive matching properly when the
2885 pattern is not unicode (thus forcing unicode semantics).
2887 If/when this is fixed the following define can be swapped
2888 in below to fully enable trie logic.
2890 #define TRIE_TYPE_IS_SAFE 1
2893 #define TRIE_TYPE_IS_SAFE (UTF || optype==EXACT)
2895 if ( last && TRIE_TYPE_IS_SAFE ) {
2896 make_trie( pRExC_state,
2897 startbranch, first, cur, tail, count,
2900 if ( PL_regkind[ OP( noper ) ] == EXACT
2902 && noper_next == tail
2907 optype = OP( noper );
2917 regprop(RExC_rx, mysv, cur);
2918 PerlIO_printf( Perl_debug_log,
2919 "%*s- %s (%d) <SCAN FINISHED>\n", (int)depth * 2 + 2,
2920 "", SvPV_nolen_const( mysv ),REG_NODE_NUM(cur));
2924 if ( last && TRIE_TYPE_IS_SAFE ) {
2925 made= make_trie( pRExC_state, startbranch, first, scan, tail, count, optype, depth+1 );
2926 #ifdef TRIE_STUDY_OPT
2927 if ( ((made == MADE_EXACT_TRIE &&
2928 startbranch == first)
2929 || ( first_non_open == first )) &&
2931 flags |= SCF_TRIE_RESTUDY;
2932 if ( startbranch == first
2935 RExC_seen &=~REG_TOP_LEVEL_BRANCHES;
2945 else if ( code == BRANCHJ ) { /* single branch is optimized. */
2946 scan = NEXTOPER(NEXTOPER(scan));
2947 } else /* single branch is optimized. */
2948 scan = NEXTOPER(scan);
2950 } else if (OP(scan) == SUSPEND || OP(scan) == GOSUB || OP(scan) == GOSTART) {
2951 scan_frame *newframe = NULL;
2956 if (OP(scan) != SUSPEND) {
2957 /* set the pointer */
2958 if (OP(scan) == GOSUB) {
2960 RExC_recurse[ARG2L(scan)] = scan;
2961 start = RExC_open_parens[paren-1];
2962 end = RExC_close_parens[paren-1];
2965 start = RExC_rxi->program + 1;
2969 Newxz(recursed, (((RExC_npar)>>3) +1), U8);
2970 SAVEFREEPV(recursed);
2972 if (!PAREN_TEST(recursed,paren+1)) {
2973 PAREN_SET(recursed,paren+1);
2974 Newx(newframe,1,scan_frame);
2976 if (flags & SCF_DO_SUBSTR) {
2977 SCAN_COMMIT(pRExC_state,data,minlenp);
2978 data->longest = &(data->longest_float);
2980 is_inf = is_inf_internal = 1;
2981 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
2982 cl_anything(pRExC_state, data->start_class);
2983 flags &= ~SCF_DO_STCLASS;
2986 Newx(newframe,1,scan_frame);
2989 end = regnext(scan);
2994 SAVEFREEPV(newframe);
2995 newframe->next = regnext(scan);
2996 newframe->last = last;
2997 newframe->stop = stopparen;
2998 newframe->prev = frame;
3008 else if (OP(scan) == EXACT) {
3009 I32 l = STR_LEN(scan);
3012 const U8 * const s = (U8*)STRING(scan);
3013 l = utf8_length(s, s + l);
3014 uc = utf8_to_uvchr(s, NULL);
3016 uc = *((U8*)STRING(scan));
3019 if (flags & SCF_DO_SUBSTR) { /* Update longest substr. */
3020 /* The code below prefers earlier match for fixed
3021 offset, later match for variable offset. */
3022 if (data->last_end == -1) { /* Update the start info. */
3023 data->last_start_min = data->pos_min;
3024 data->last_start_max = is_inf
3025 ? I32_MAX : data->pos_min + data->pos_delta;
3027 sv_catpvn(data->last_found, STRING(scan), STR_LEN(scan));
3029 SvUTF8_on(data->last_found);
3031 SV * const sv = data->last_found;
3032 MAGIC * const mg = SvUTF8(sv) && SvMAGICAL(sv) ?
3033 mg_find(sv, PERL_MAGIC_utf8) : NULL;
3034 if (mg && mg->mg_len >= 0)
3035 mg->mg_len += utf8_length((U8*)STRING(scan),
3036 (U8*)STRING(scan)+STR_LEN(scan));
3038 data->last_end = data->pos_min + l;
3039 data->pos_min += l; /* As in the first entry. */
3040 data->flags &= ~SF_BEFORE_EOL;
3042 if (flags & SCF_DO_STCLASS_AND) {
3043 /* Check whether it is compatible with what we know already! */
3047 (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
3048 && !ANYOF_BITMAP_TEST(data->start_class, uc)
3049 && (!(data->start_class->flags & ANYOF_FOLD)
3050 || !ANYOF_BITMAP_TEST(data->start_class, PL_fold[uc])))
3053 ANYOF_CLASS_ZERO(data->start_class);
3054 ANYOF_BITMAP_ZERO(data->start_class);
3056 ANYOF_BITMAP_SET(data->start_class, uc);
3057 data->start_class->flags &= ~ANYOF_EOS;
3059 data->start_class->flags &= ~ANYOF_UNICODE_ALL;
3061 else if (flags & SCF_DO_STCLASS_OR) {
3062 /* false positive possible if the class is case-folded */
3064 ANYOF_BITMAP_SET(data->start_class, uc);
3066 data->start_class->flags |= ANYOF_UNICODE_ALL;
3067 data->start_class->flags &= ~ANYOF_EOS;
3068 cl_and(data->start_class, and_withp);
3070 flags &= ~SCF_DO_STCLASS;
3072 else if (PL_regkind[OP(scan)] == EXACT) { /* But OP != EXACT! */
3073 I32 l = STR_LEN(scan);
3074 UV uc = *((U8*)STRING(scan));
3076 /* Search for fixed substrings supports EXACT only. */
3077 if (flags & SCF_DO_SUBSTR) {
3079 SCAN_COMMIT(pRExC_state, data, minlenp);
3082 const U8 * const s = (U8 *)STRING(scan);
3083 l = utf8_length(s, s + l);
3084 uc = utf8_to_uvchr(s, NULL);
3087 if (flags & SCF_DO_SUBSTR)
3089 if (flags & SCF_DO_STCLASS_AND) {
3090 /* Check whether it is compatible with what we know already! */
3094 (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
3095 && !ANYOF_BITMAP_TEST(data->start_class, uc)
3096 && !ANYOF_BITMAP_TEST(data->start_class, PL_fold[uc])))
3098 ANYOF_CLASS_ZERO(data->start_class);
3099 ANYOF_BITMAP_ZERO(data->start_class);
3101 ANYOF_BITMAP_SET(data->start_class, uc);
3102 data->start_class->flags &= ~ANYOF_EOS;
3103 data->start_class->flags |= ANYOF_FOLD;
3104 if (OP(scan) == EXACTFL)
3105 data->start_class->flags |= ANYOF_LOCALE;
3108 else if (flags & SCF_DO_STCLASS_OR) {
3109 if (data->start_class->flags & ANYOF_FOLD) {
3110 /* false positive possible if the class is case-folded.
3111 Assume that the locale settings are the same... */
3113 ANYOF_BITMAP_SET(data->start_class, uc);
3114 data->start_class->flags &= ~ANYOF_EOS;
3116 cl_and(data->start_class, and_withp);
3118 flags &= ~SCF_DO_STCLASS;
3120 else if (REGNODE_VARIES(OP(scan))) {
3121 I32 mincount, maxcount, minnext, deltanext, fl = 0;
3122 I32 f = flags, pos_before = 0;
3123 regnode * const oscan = scan;
3124 struct regnode_charclass_class this_class;
3125 struct regnode_charclass_class *oclass = NULL;
3126 I32 next_is_eval = 0;
3128 switch (PL_regkind[OP(scan)]) {
3129 case WHILEM: /* End of (?:...)* . */
3130 scan = NEXTOPER(scan);
3133 if (flags & (SCF_DO_SUBSTR | SCF_DO_STCLASS)) {
3134 next = NEXTOPER(scan);
3135 if (OP(next) == EXACT || (flags & SCF_DO_STCLASS)) {
3137 maxcount = REG_INFTY;
3138 next = regnext(scan);
3139 scan = NEXTOPER(scan);
3143 if (flags & SCF_DO_SUBSTR)
3148 if (flags & SCF_DO_STCLASS) {
3150 maxcount = REG_INFTY;
3151 next = regnext(scan);
3152 scan = NEXTOPER(scan);
3155 is_inf = is_inf_internal = 1;
3156 scan = regnext(scan);
3157 if (flags & SCF_DO_SUBSTR) {
3158 SCAN_COMMIT(pRExC_state, data, minlenp); /* Cannot extend fixed substrings */
3159 data->longest = &(data->longest_float);
3161 goto optimize_curly_tail;
3163 if (stopparen>0 && (OP(scan)==CURLYN || OP(scan)==CURLYM)
3164 && (scan->flags == stopparen))
3169 mincount = ARG1(scan);
3170 maxcount = ARG2(scan);
3172 next = regnext(scan);
3173 if (OP(scan) == CURLYX) {
3174 I32 lp = (data ? *(data->last_closep) : 0);
3175 scan->flags = ((lp <= (I32)U8_MAX) ? (U8)lp : U8_MAX);
3177 scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3178 next_is_eval = (OP(scan) == EVAL);
3180 if (flags & SCF_DO_SUBSTR) {
3181 if (mincount == 0) SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot extend fixed substrings */
3182 pos_before = data->pos_min;
3186 data->flags &= ~(SF_HAS_PAR|SF_IN_PAR|SF_HAS_EVAL);
3188 data->flags |= SF_IS_INF;
3190 if (flags & SCF_DO_STCLASS) {
3191 cl_init(pRExC_state, &this_class);
3192 oclass = data->start_class;
3193 data->start_class = &this_class;
3194 f |= SCF_DO_STCLASS_AND;
3195 f &= ~SCF_DO_STCLASS_OR;
3197 /* These are the cases when once a subexpression
3198 fails at a particular position, it cannot succeed
3199 even after backtracking at the enclosing scope.
3201 XXXX what if minimal match and we are at the
3202 initial run of {n,m}? */
3203 if ((mincount != maxcount - 1) && (maxcount != REG_INFTY))
3204 f &= ~SCF_WHILEM_VISITED_POS;
3206 /* This will finish on WHILEM, setting scan, or on NULL: */
3207 minnext = study_chunk(pRExC_state, &scan, minlenp, &deltanext,
3208 last, data, stopparen, recursed, NULL,
3210 ? (f & ~SCF_DO_SUBSTR) : f),depth+1);
3212 if (flags & SCF_DO_STCLASS)
3213 data->start_class = oclass;
3214 if (mincount == 0 || minnext == 0) {
3215 if (flags & SCF_DO_STCLASS_OR) {
3216 cl_or(pRExC_state, data->start_class, &this_class);
3218 else if (flags & SCF_DO_STCLASS_AND) {
3219 /* Switch to OR mode: cache the old value of
3220 * data->start_class */
3222 StructCopy(data->start_class, and_withp,
3223 struct regnode_charclass_class);
3224 flags &= ~SCF_DO_STCLASS_AND;
3225 StructCopy(&this_class, data->start_class,
3226 struct regnode_charclass_class);
3227 flags |= SCF_DO_STCLASS_OR;
3228 data->start_class->flags |= ANYOF_EOS;
3230 } else { /* Non-zero len */
3231 if (flags & SCF_DO_STCLASS_OR) {
3232 cl_or(pRExC_state, data->start_class, &this_class);
3233 cl_and(data->start_class, and_withp);
3235 else if (flags & SCF_DO_STCLASS_AND)
3236 cl_and(data->start_class, &this_class);
3237 flags &= ~SCF_DO_STCLASS;
3239 if (!scan) /* It was not CURLYX, but CURLY. */
3241 if ( /* ? quantifier ok, except for (?{ ... }) */
3242 (next_is_eval || !(mincount == 0 && maxcount == 1))
3243 && (minnext == 0) && (deltanext == 0)
3244 && data && !(data->flags & (SF_HAS_PAR|SF_IN_PAR))
3245 && maxcount <= REG_INFTY/3) /* Complement check for big count */
3247 ckWARNreg(RExC_parse,
3248 "Quantifier unexpected on zero-length expression");
3251 min += minnext * mincount;
3252 is_inf_internal |= ((maxcount == REG_INFTY
3253 && (minnext + deltanext) > 0)
3254 || deltanext == I32_MAX);
3255 is_inf |= is_inf_internal;
3256 delta += (minnext + deltanext) * maxcount - minnext * mincount;
3258 /* Try powerful optimization CURLYX => CURLYN. */
3259 if ( OP(oscan) == CURLYX && data
3260 && data->flags & SF_IN_PAR
3261 && !(data->flags & SF_HAS_EVAL)
3262 && !deltanext && minnext == 1 ) {
3263 /* Try to optimize to CURLYN. */
3264 regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS;
3265 regnode * const nxt1 = nxt;
3272 if (!REGNODE_SIMPLE(OP(nxt))
3273 && !(PL_regkind[OP(nxt)] == EXACT
3274 && STR_LEN(nxt) == 1))
3280 if (OP(nxt) != CLOSE)
3282 if (RExC_open_parens) {
3283 RExC_open_parens[ARG(nxt1)-1]=oscan; /*open->CURLYM*/
3284 RExC_close_parens[ARG(nxt1)-1]=nxt+2; /*close->while*/
3286 /* Now we know that nxt2 is the only contents: */
3287 oscan->flags = (U8)ARG(nxt);
3289 OP(nxt1) = NOTHING; /* was OPEN. */
3292 OP(nxt1 + 1) = OPTIMIZED; /* was count. */
3293 NEXT_OFF(nxt1+ 1) = 0; /* just for consistancy. */
3294 NEXT_OFF(nxt2) = 0; /* just for consistancy with CURLY. */
3295 OP(nxt) = OPTIMIZED; /* was CLOSE. */
3296 OP(nxt + 1) = OPTIMIZED; /* was count. */
3297 NEXT_OFF(nxt+ 1) = 0; /* just for consistancy. */
3302 /* Try optimization CURLYX => CURLYM. */
3303 if ( OP(oscan) == CURLYX && data
3304 && !(data->flags & SF_HAS_PAR)
3305 && !(data->flags & SF_HAS_EVAL)
3306 && !deltanext /* atom is fixed width */
3307 && minnext != 0 /* CURLYM can't handle zero width */
3309 /* XXXX How to optimize if data == 0? */
3310 /* Optimize to a simpler form. */
3311 regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN */
3315 while ( (nxt2 = regnext(nxt)) /* skip over embedded stuff*/
3316 && (OP(nxt2) != WHILEM))
3318 OP(nxt2) = SUCCEED; /* Whas WHILEM */
3319 /* Need to optimize away parenths. */
3320 if (data->flags & SF_IN_PAR) {
3321 /* Set the parenth number. */
3322 regnode *nxt1 = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN*/
3324 if (OP(nxt) != CLOSE)
3325 FAIL("Panic opt close");
3326 oscan->flags = (U8)ARG(nxt);
3327 if (RExC_open_parens) {
3328 RExC_open_parens[ARG(nxt1)-1]=oscan; /*open->CURLYM*/
3329 RExC_close_parens[ARG(nxt1)-1]=nxt2+1; /*close->NOTHING*/
3331 OP(nxt1) = OPTIMIZED; /* was OPEN. */
3332 OP(nxt) = OPTIMIZED; /* was CLOSE. */
3335 OP(nxt1 + 1) = OPTIMIZED; /* was count. */
3336 OP(nxt + 1) = OPTIMIZED; /* was count. */
3337 NEXT_OFF(nxt1 + 1) = 0; /* just for consistancy. */
3338 NEXT_OFF(nxt + 1) = 0; /* just for consistancy. */
3341 while ( nxt1 && (OP(nxt1) != WHILEM)) {
3342 regnode *nnxt = regnext(nxt1);
3345 if (reg_off_by_arg[OP(nxt1)])
3346 ARG_SET(nxt1, nxt2 - nxt1);
3347 else if (nxt2 - nxt1 < U16_MAX)
3348 NEXT_OFF(nxt1) = nxt2 - nxt1;
3350 OP(nxt) = NOTHING; /* Cannot beautify */
3355 /* Optimize again: */
3356 study_chunk(pRExC_state, &nxt1, minlenp, &deltanext, nxt,
3357 NULL, stopparen, recursed, NULL, 0,depth+1);
3362 else if ((OP(oscan) == CURLYX)
3363 && (flags & SCF_WHILEM_VISITED_POS)
3364 /* See the comment on a similar expression above.
3365 However, this time it not a subexpression
3366 we care about, but the expression itself. */
3367 && (maxcount == REG_INFTY)
3368 && data && ++data->whilem_c < 16) {
3369 /* This stays as CURLYX, we can put the count/of pair. */
3370 /* Find WHILEM (as in regexec.c) */
3371 regnode *nxt = oscan + NEXT_OFF(oscan);
3373 if (OP(PREVOPER(nxt)) == NOTHING) /* LONGJMP */
3375 PREVOPER(nxt)->flags = (U8)(data->whilem_c
3376 | (RExC_whilem_seen << 4)); /* On WHILEM */
3378 if (data && fl & (SF_HAS_PAR|SF_IN_PAR))
3380 if (flags & SCF_DO_SUBSTR) {
3381 SV *last_str = NULL;
3382 int counted = mincount != 0;
3384 if (data->last_end > 0 && mincount != 0) { /* Ends with a string. */
3385 #if defined(SPARC64_GCC_WORKAROUND)
3388 const char *s = NULL;
3391 if (pos_before >= data->last_start_min)
3394 b = data->last_start_min;
3397 s = SvPV_const(data->last_found, l);
3398 old = b - data->last_start_min;
3401 I32 b = pos_before >= data->last_start_min
3402 ? pos_before : data->last_start_min;
3404 const char * const s = SvPV_const(data->last_found, l);
3405 I32 old = b - data->last_start_min;
3409 old = utf8_hop((U8*)s, old) - (U8*)s;
3412 /* Get the added string: */
3413 last_str = newSVpvn_utf8(s + old, l, UTF);
3414 if (deltanext == 0 && pos_before == b) {
3415 /* What was added is a constant string */
3417 SvGROW(last_str, (mincount * l) + 1);
3418 repeatcpy(SvPVX(last_str) + l,
3419 SvPVX_const(last_str), l, mincount - 1);
3420 SvCUR_set(last_str, SvCUR(last_str) * mincount);
3421 /* Add additional parts. */
3422 SvCUR_set(data->last_found,
3423 SvCUR(data->last_found) - l);
3424 sv_catsv(data->last_found, last_str);
3426 SV * sv = data->last_found;
3428 SvUTF8(sv) && SvMAGICAL(sv) ?
3429 mg_find(sv, PERL_MAGIC_utf8) : NULL;
3430 if (mg && mg->mg_len >= 0)
3431 mg->mg_len += CHR_SVLEN(last_str) - l;
3433 data->last_end += l * (mincount - 1);
3436 /* start offset must point into the last copy */
3437 data->last_start_min += minnext * (mincount - 1);
3438 data->last_start_max += is_inf ? I32_MAX
3439 : (maxcount - 1) * (minnext + data->pos_delta);
3442 /* It is counted once already... */
3443 data->pos_min += minnext * (mincount - counted);
3444 data->pos_delta += - counted * deltanext +
3445 (minnext + deltanext) * maxcount - minnext * mincount;
3446 if (mincount != maxcount) {
3447 /* Cannot extend fixed substrings found inside
3449 SCAN_COMMIT(pRExC_state,data,minlenp);
3450 if (mincount && last_str) {
3451 SV * const sv = data->last_found;
3452 MAGIC * const mg = SvUTF8(sv) && SvMAGICAL(sv) ?
3453 mg_find(sv, PERL_MAGIC_utf8) : NULL;
3457 sv_setsv(sv, last_str);
3458 data->last_end = data->pos_min;
3459 data->last_start_min =
3460 data->pos_min - CHR_SVLEN(last_str);
3461 data->last_start_max = is_inf
3463 : data->pos_min + data->pos_delta
3464 - CHR_SVLEN(last_str);
3466 data->longest = &(data->longest_float);
3468 SvREFCNT_dec(last_str);
3470 if (data && (fl & SF_HAS_EVAL))
3471 data->flags |= SF_HAS_EVAL;
3472 optimize_curly_tail:
3473 if (OP(oscan) != CURLYX) {
3474 while (PL_regkind[OP(next = regnext(oscan))] == NOTHING
3476 NEXT_OFF(oscan) += NEXT_OFF(next);
3479 default: /* REF and CLUMP only? */
3480 if (flags & SCF_DO_SUBSTR) {
3481 SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot expect anything... */
3482 data->longest = &(data->longest_float);
3484 is_inf = is_inf_internal = 1;
3485 if (flags & SCF_DO_STCLASS_OR)
3486 cl_anything(pRExC_state, data->start_class);
3487 flags &= ~SCF_DO_STCLASS;
3491 else if (OP(scan) == LNBREAK) {
3492 if (flags & SCF_DO_STCLASS) {
3494 data->start_class->flags &= ~ANYOF_EOS; /* No match on empty */
3495 if (flags & SCF_DO_STCLASS_AND) {
3496 for (value = 0; value < 256; value++)
3497 if (!is_VERTWS_cp(value))
3498 ANYOF_BITMAP_CLEAR(data->start_class, value);
3501 for (value = 0; value < 256; value++)
3502 if (is_VERTWS_cp(value))
3503 ANYOF_BITMAP_SET(data->start_class, value);
3505 if (flags & SCF_DO_STCLASS_OR)
3506 cl_and(data->start_class, and_withp);
3507 flags &= ~SCF_DO_STCLASS;
3511 if (flags & SCF_DO_SUBSTR) {
3512 SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot expect anything... */
3514 data->pos_delta += 1;
3515 data->longest = &(data->longest_float);
3519 else if (OP(scan) == FOLDCHAR) {
3520 int d = ARG(scan)==0xDF ? 1 : 2;
3521 flags &= ~SCF_DO_STCLASS;
3524 if (flags & SCF_DO_SUBSTR) {
3525 SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot expect anything... */
3527 data->pos_delta += d;
3528 data->longest = &(data->longest_float);
3531 else if (REGNODE_SIMPLE(OP(scan))) {
3534 if (flags & SCF_DO_SUBSTR) {
3535 SCAN_COMMIT(pRExC_state,data,minlenp);
3539 if (flags & SCF_DO_STCLASS) {
3540 data->start_class->flags &= ~ANYOF_EOS; /* No match on empty */
3542 /* Some of the logic below assumes that switching
3543 locale on will only add false positives. */
3544 switch (PL_regkind[OP(scan)]) {
3548 /* Perl_croak(aTHX_ "panic: unexpected simple REx opcode %d", OP(scan)); */
3549 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
3550 cl_anything(pRExC_state, data->start_class);
3553 if (OP(scan) == SANY)
3555 if (flags & SCF_DO_STCLASS_OR) { /* Everything but \n */
3556 value = (ANYOF_BITMAP_TEST(data->start_class,'\n')
3557 || (data->start_class->flags & ANYOF_CLASS));
3558 cl_anything(pRExC_state, data->start_class);
3560 if (flags & SCF_DO_STCLASS_AND || !value)
3561 ANYOF_BITMAP_CLEAR(data->start_class,'\n');
3564 if (flags & SCF_DO_STCLASS_AND)
3565 cl_and(data->start_class,
3566 (struct regnode_charclass_class*)scan);
3568 cl_or(pRExC_state, data->start_class,
3569 (struct regnode_charclass_class*)scan);
3572 if (flags & SCF_DO_STCLASS_AND) {
3573 if (!(data->start_class->flags & ANYOF_LOCALE)) {
3574 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM);
3575 for (value = 0; value < 256; value++)
3576 if (!isALNUM(value))
3577 ANYOF_BITMAP_CLEAR(data->start_class, value);
3581 if (data->start_class->flags & ANYOF_LOCALE)
3582 ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM);
3584 for (value = 0; value < 256; value++)
3586 ANYOF_BITMAP_SET(data->start_class, value);
3591 if (flags & SCF_DO_STCLASS_AND) {
3592 if (data->start_class->flags & ANYOF_LOCALE)
3593 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM);
3596 ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM);
3597 data->start_class->flags |= ANYOF_LOCALE;
3601 if (flags & SCF_DO_STCLASS_AND) {
3602 if (!(data->start_class->flags & ANYOF_LOCALE)) {
3603 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM);
3604 for (value = 0; value < 256; value++)
3606 ANYOF_BITMAP_CLEAR(data->start_class, value);
3610 if (data->start_class->flags & ANYOF_LOCALE)
3611 ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM);
3613 for (value = 0; value < 256; value++)
3614 if (!isALNUM(value))
3615 ANYOF_BITMAP_SET(data->start_class, value);
3620 if (flags & SCF_DO_STCLASS_AND) {
3621 if (data->start_class->flags & ANYOF_LOCALE)
3622 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM);
3625 data->start_class->flags |= ANYOF_LOCALE;
3626 ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM);
3630 if (flags & SCF_DO_STCLASS_AND) {
3631 if (!(data->start_class->flags & ANYOF_LOCALE)) {
3632 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE);
3633 for (value = 0; value < 256; value++)
3634 if (!isSPACE(value))
3635 ANYOF_BITMAP_CLEAR(data->start_class, value);
3639 if (data->start_class->flags & ANYOF_LOCALE)
3640 ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE);
3642 for (value = 0; value < 256; value++)
3644 ANYOF_BITMAP_SET(data->start_class, value);
3649 if (flags & SCF_DO_STCLASS_AND) {
3650 if (data->start_class->flags & ANYOF_LOCALE)
3651 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE);
3654 data->start_class->flags |= ANYOF_LOCALE;
3655 ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE);
3659 if (flags & SCF_DO_STCLASS_AND) {
3660 if (!(data->start_class->flags & ANYOF_LOCALE)) {
3661 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE);
3662 for (value = 0; value < 256; value++)
3664 ANYOF_BITMAP_CLEAR(data->start_class, value);
3668 if (data->start_class->flags & ANYOF_LOCALE)
3669 ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE);
3671 for (value = 0; value < 256; value++)
3672 if (!isSPACE(value))
3673 ANYOF_BITMAP_SET(data->start_class, value);
3678 if (flags & SCF_DO_STCLASS_AND) {
3679 if (data->start_class->flags & ANYOF_LOCALE) {
3680 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE);
3681 for (value = 0; value < 256; value++)
3682 if (!isSPACE(value))
3683 ANYOF_BITMAP_CLEAR(data->start_class, value);
3687 data->start_class->flags |= ANYOF_LOCALE;
3688 ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE);
3692 if (flags & SCF_DO_STCLASS_AND) {
3693 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NDIGIT);
3694 for (value = 0; value < 256; value++)
3695 if (!isDIGIT(value))
3696 ANYOF_BITMAP_CLEAR(data->start_class, value);
3699 if (data->start_class->flags & ANYOF_LOCALE)
3700 ANYOF_CLASS_SET(data->start_class,ANYOF_DIGIT);
3702 for (value = 0; value < 256; value++)
3704 ANYOF_BITMAP_SET(data->start_class, value);
3709 if (flags & SCF_DO_STCLASS_AND) {
3710 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_DIGIT);
3711 for (value = 0; value < 256; value++)
3713 ANYOF_BITMAP_CLEAR(data->start_class, value);
3716 if (data->start_class->flags & ANYOF_LOCALE)
3717 ANYOF_CLASS_SET(data->start_class,ANYOF_NDIGIT);
3719 for (value = 0; value < 256; value++)
3720 if (!isDIGIT(value))
3721 ANYOF_BITMAP_SET(data->start_class, value);
3725 CASE_SYNST_FNC(VERTWS);
3726 CASE_SYNST_FNC(HORIZWS);
3729 if (flags & SCF_DO_STCLASS_OR)
3730 cl_and(data->start_class, and_withp);
3731 flags &= ~SCF_DO_STCLASS;
3734 else if (PL_regkind[OP(scan)] == EOL && flags & SCF_DO_SUBSTR) {
3735 data->flags |= (OP(scan) == MEOL
3739 else if ( PL_regkind[OP(scan)] == BRANCHJ
3740 /* Lookbehind, or need to calculate parens/evals/stclass: */
3741 && (scan->flags || data || (flags & SCF_DO_STCLASS))
3742 && (OP(scan) == IFMATCH || OP(scan) == UNLESSM)) {
3743 if ( !PERL_ENABLE_POSITIVE_ASSERTION_STUDY
3744 || OP(scan) == UNLESSM )
3746 /* Negative Lookahead/lookbehind
3747 In this case we can't do fixed string optimisation.
3750 I32 deltanext, minnext, fake = 0;
3752 struct regnode_charclass_class intrnl;
3755 data_fake.flags = 0;
3757 data_fake.whilem_c = data->whilem_c;
3758 data_fake.last_closep = data->last_closep;
3761 data_fake.last_closep = &fake;
3762 data_fake.pos_delta = delta;
3763 if ( flags & SCF_DO_STCLASS && !scan->flags
3764 && OP(scan) == IFMATCH ) { /* Lookahead */
3765 cl_init(pRExC_state, &intrnl);
3766 data_fake.start_class = &intrnl;
3767 f |= SCF_DO_STCLASS_AND;
3769 if (flags & SCF_WHILEM_VISITED_POS)
3770 f |= SCF_WHILEM_VISITED_POS;
3771 next = regnext(scan);
3772 nscan = NEXTOPER(NEXTOPER(scan));
3773 minnext = study_chunk(pRExC_state, &nscan, minlenp, &deltanext,
3774 last, &data_fake, stopparen, recursed, NULL, f, depth+1);
3777 FAIL("Variable length lookbehind not implemented");
3779 else if (minnext > (I32)U8_MAX) {
3780 FAIL2("Lookbehind longer than %"UVuf" not implemented", (UV)U8_MAX);
3782 scan->flags = (U8)minnext;
3785 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
3787 if (data_fake.flags & SF_HAS_EVAL)
3788 data->flags |= SF_HAS_EVAL;
3789 data->whilem_c = data_fake.whilem_c;
3791 if (f & SCF_DO_STCLASS_AND) {
3792 if (flags & SCF_DO_STCLASS_OR) {
3793 /* OR before, AND after: ideally we would recurse with
3794 * data_fake to get the AND applied by study of the
3795 * remainder of the pattern, and then derecurse;
3796 * *** HACK *** for now just treat as "no information".
3797 * See [perl #56690].
3799 cl_init(pRExC_state, data->start_class);
3801 /* AND before and after: combine and continue */
3802 const int was = (data->start_class->flags & ANYOF_EOS);
3804 cl_and(data->start_class, &intrnl);
3806 data->start_class->flags |= ANYOF_EOS;
3810 #if PERL_ENABLE_POSITIVE_ASSERTION_STUDY
3812 /* Positive Lookahead/lookbehind
3813 In this case we can do fixed string optimisation,
3814 but we must be careful about it. Note in the case of
3815 lookbehind the positions will be offset by the minimum
3816 length of the pattern, something we won't know about
3817 until after the recurse.
3819 I32 deltanext, fake = 0;
3821 struct regnode_charclass_class intrnl;
3823 /* We use SAVEFREEPV so that when the full compile
3824 is finished perl will clean up the allocated
3825 minlens when its all done. This was we don't
3826 have to worry about freeing them when we know
3827 they wont be used, which would be a pain.
3830 Newx( minnextp, 1, I32 );
3831 SAVEFREEPV(minnextp);
3834 StructCopy(data, &data_fake, scan_data_t);
3835 if ((flags & SCF_DO_SUBSTR) && data->last_found) {
3838 SCAN_COMMIT(pRExC_state, &data_fake,minlenp);
3839 data_fake.last_found=newSVsv(data->last_found);
3843 data_fake.last_closep = &fake;
3844 data_fake.flags = 0;
3845 data_fake.pos_delta = delta;
3847 data_fake.flags |= SF_IS_INF;
3848 if ( flags & SCF_DO_STCLASS && !scan->flags
3849 && OP(scan) == IFMATCH ) { /* Lookahead */
3850 cl_init(pRExC_state, &intrnl);
3851 data_fake.start_class = &intrnl;
3852 f |= SCF_DO_STCLASS_AND;
3854 if (flags & SCF_WHILEM_VISITED_POS)
3855 f |= SCF_WHILEM_VISITED_POS;
3856 next = regnext(scan);
3857 nscan = NEXTOPER(NEXTOPER(scan));
3859 *minnextp = study_chunk(pRExC_state, &nscan, minnextp, &deltanext,
3860 last, &data_fake, stopparen, recursed, NULL, f,depth+1);
3863 FAIL("Variable length lookbehind not implemented");
3865 else if (*minnextp > (I32)U8_MAX) {
3866 FAIL2("Lookbehind longer than %"UVuf" not implemented", (UV)U8_MAX);
3868 scan->flags = (U8)*minnextp;
3873 if (f & SCF_DO_STCLASS_AND) {
3874 const int was = (data->start_class->flags & ANYOF_EOS);
3876 cl_and(data->start_class, &intrnl);
3878 data->start_class->flags |= ANYOF_EOS;
3881 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
3883 if (data_fake.flags & SF_HAS_EVAL)
3884 data->flags |= SF_HAS_EVAL;
3885 data->whilem_c = data_fake.whilem_c;
3886 if ((flags & SCF_DO_SUBSTR) && data_fake.last_found) {
3887 if (RExC_rx->minlen<*minnextp)
3888 RExC_rx->minlen=*minnextp;
3889 SCAN_COMMIT(pRExC_state, &data_fake, minnextp);
3890 SvREFCNT_dec(data_fake.last_found);
3892 if ( data_fake.minlen_fixed != minlenp )
3894 data->offset_fixed= data_fake.offset_fixed;
3895 data->minlen_fixed= data_fake.minlen_fixed;
3896 data->lookbehind_fixed+= scan->flags;
3898 if ( data_fake.minlen_float != minlenp )
3900 data->minlen_float= data_fake.minlen_float;
3901 data->offset_float_min=data_fake.offset_float_min;
3902 data->offset_float_max=data_fake.offset_float_max;
3903 data->lookbehind_float+= scan->flags;
3912 else if (OP(scan) == OPEN) {
3913 if (stopparen != (I32)ARG(scan))
3916 else if (OP(scan) == CLOSE) {
3917 if (stopparen == (I32)ARG(scan)) {
3920 if ((I32)ARG(scan) == is_par) {
3921 next = regnext(scan);
3923 if ( next && (OP(next) != WHILEM) && next < last)
3924 is_par = 0; /* Disable optimization */
3927 *(data->last_closep) = ARG(scan);
3929 else if (OP(scan) == EVAL) {
3931 data->flags |= SF_HAS_EVAL;
3933 else if ( PL_regkind[OP(scan)] == ENDLIKE ) {
3934 if (flags & SCF_DO_SUBSTR) {
3935 SCAN_COMMIT(pRExC_state,data,minlenp);
3936 flags &= ~SCF_DO_SUBSTR;
3938 if (data && OP(scan)==ACCEPT) {
3939 data->flags |= SCF_SEEN_ACCEPT;
3944 else if (OP(scan) == LOGICAL && scan->flags == 2) /* Embedded follows */
3946 if (flags & SCF_DO_SUBSTR) {
3947 SCAN_COMMIT(pRExC_state,data,minlenp);
3948 data->longest = &(data->longest_float);
3950 is_inf = is_inf_internal = 1;
3951 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
3952 cl_anything(pRExC_state, data->start_class);
3953 flags &= ~SCF_DO_STCLASS;
3955 else if (OP(scan) == GPOS) {
3956 if (!(RExC_rx->extflags & RXf_GPOS_FLOAT) &&
3957 !(delta || is_inf || (data && data->pos_delta)))
3959 if (!(RExC_rx->extflags & RXf_ANCH) && (flags & SCF_DO_SUBSTR))
3960 RExC_rx->extflags |= RXf_ANCH_GPOS;
3961 if (RExC_rx->gofs < (U32)min)
3962 RExC_rx->gofs = min;
3964 RExC_rx->extflags |= RXf_GPOS_FLOAT;
3968 #ifdef TRIE_STUDY_OPT
3969 #ifdef FULL_TRIE_STUDY
3970 else if (PL_regkind[OP(scan)] == TRIE) {
3971 /* NOTE - There is similar code to this block above for handling
3972 BRANCH nodes on the initial study. If you change stuff here
3974 regnode *trie_node= scan;
3975 regnode *tail= regnext(scan);
3976 reg_trie_data *trie = (reg_trie_data*)RExC_rxi->data->data[ ARG(scan) ];
3977 I32 max1 = 0, min1 = I32_MAX;
3978 struct regnode_charclass_class accum;
3980 if (flags & SCF_DO_SUBSTR) /* XXXX Add !SUSPEND? */
3981 SCAN_COMMIT(pRExC_state, data,minlenp); /* Cannot merge strings after this. */
3982 if (flags & SCF_DO_STCLASS)
3983 cl_init_zero(pRExC_state, &accum);
3989 const regnode *nextbranch= NULL;
3992 for ( word=1 ; word <= trie->wordcount ; word++)
3994 I32 deltanext=0, minnext=0, f = 0, fake;
3995 struct regnode_charclass_class this_class;
3997 data_fake.flags = 0;
3999 data_fake.whilem_c = data->whilem_c;
4000 data_fake.last_closep = data->last_closep;
4003 data_fake.last_closep = &fake;
4004 data_fake.pos_delta = delta;
4005 if (flags & SCF_DO_STCLASS) {
4006 cl_init(pRExC_state, &this_class);
4007 data_fake.start_class = &this_class;
4008 f = SCF_DO_STCLASS_AND;
4010 if (flags & SCF_WHILEM_VISITED_POS)
4011 f |= SCF_WHILEM_VISITED_POS;
4013 if (trie->jump[word]) {
4015 nextbranch = trie_node + trie->jump[0];
4016 scan= trie_node + trie->jump[word];
4017 /* We go from the jump point to the branch that follows
4018 it. Note this means we need the vestigal unused branches
4019 even though they arent otherwise used.
4021 minnext = study_chunk(pRExC_state, &scan, minlenp,
4022 &deltanext, (regnode *)nextbranch, &data_fake,
4023 stopparen, recursed, NULL, f,depth+1);
4025 if (nextbranch && PL_regkind[OP(nextbranch)]==BRANCH)
4026 nextbranch= regnext((regnode*)nextbranch);
4028 if (min1 > (I32)(minnext + trie->minlen))
4029 min1 = minnext + trie->minlen;
4030 if (max1 < (I32)(minnext + deltanext + trie->maxlen))
4031 max1 = minnext + deltanext + trie->maxlen;
4032 if (deltanext == I32_MAX)
4033 is_inf = is_inf_internal = 1;
4035 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
4037 if (data_fake.flags & SCF_SEEN_ACCEPT) {
4038 if ( stopmin > min + min1)
4039 stopmin = min + min1;
4040 flags &= ~SCF_DO_SUBSTR;
4042 data->flags |= SCF_SEEN_ACCEPT;
4045 if (data_fake.flags & SF_HAS_EVAL)
4046 data->flags |= SF_HAS_EVAL;
4047 data->whilem_c = data_fake.whilem_c;
4049 if (flags & SCF_DO_STCLASS)
4050 cl_or(pRExC_state, &accum, &this_class);
4053 if (flags & SCF_DO_SUBSTR) {
4054 data->pos_min += min1;
4055 data->pos_delta += max1 - min1;
4056 if (max1 != min1 || is_inf)
4057 data->longest = &(data->longest_float);
4060 delta += max1 - min1;
4061 if (flags & SCF_DO_STCLASS_OR) {
4062 cl_or(pRExC_state, data->start_class, &accum);
4064 cl_and(data->start_class, and_withp);
4065 flags &= ~SCF_DO_STCLASS;
4068 else if (flags & SCF_DO_STCLASS_AND) {
4070 cl_and(data->start_class, &accum);
4071 flags &= ~SCF_DO_STCLASS;
4074 /* Switch to OR mode: cache the old value of
4075 * data->start_class */
4077 StructCopy(data->start_class, and_withp,
4078 struct regnode_charclass_class);
4079 flags &= ~SCF_DO_STCLASS_AND;
4080 StructCopy(&accum, data->start_class,
4081 struct regnode_charclass_class);
4082 flags |= SCF_DO_STCLASS_OR;
4083 data->start_class->flags |= ANYOF_EOS;
4090 else if (PL_regkind[OP(scan)] == TRIE) {
4091 reg_trie_data *trie = (reg_trie_data*)RExC_rxi->data->data[ ARG(scan) ];
4094 min += trie->minlen;
4095 delta += (trie->maxlen - trie->minlen);
4096 flags &= ~SCF_DO_STCLASS; /* xxx */
4097 if (flags & SCF_DO_SUBSTR) {
4098 SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot expect anything... */
4099 data->pos_min += trie->minlen;
4100 data->pos_delta += (trie->maxlen - trie->minlen);
4101 if (trie->maxlen != trie->minlen)
4102 data->longest = &(data->longest_float);
4104 if (trie->jump) /* no more substrings -- for now /grr*/
4105 flags &= ~SCF_DO_SUBSTR;
4107 #endif /* old or new */
4108 #endif /* TRIE_STUDY_OPT */
4110 /* Else: zero-length, ignore. */
4111 scan = regnext(scan);
4116 stopparen = frame->stop;
4117 frame = frame->prev;
4118 goto fake_study_recurse;
4123 DEBUG_STUDYDATA("pre-fin:",data,depth);
4126 *deltap = is_inf_internal ? I32_MAX : delta;
4127 if (flags & SCF_DO_SUBSTR && is_inf)
4128 data->pos_delta = I32_MAX - data->pos_min;
4129 if (is_par > (I32)U8_MAX)
4131 if (is_par && pars==1 && data) {
4132 data->flags |= SF_IN_PAR;
4133 data->flags &= ~SF_HAS_PAR;
4135 else if (pars && data) {
4136 data->flags |= SF_HAS_PAR;
4137 data->flags &= ~SF_IN_PAR;
4139 if (flags & SCF_DO_STCLASS_OR)
4140 cl_and(data->start_class, and_withp);
4141 if (flags & SCF_TRIE_RESTUDY)
4142 data->flags |= SCF_TRIE_RESTUDY;
4144 DEBUG_STUDYDATA("post-fin:",data,depth);
4146 return min < stopmin ? min : stopmin;
4150 S_add_data(RExC_state_t *pRExC_state, U32 n, const char *s)
4152 U32 count = RExC_rxi->data ? RExC_rxi->data->count : 0;
4154 PERL_ARGS_ASSERT_ADD_DATA;
4156 Renewc(RExC_rxi->data,
4157 sizeof(*RExC_rxi->data) + sizeof(void*) * (count + n - 1),
4158 char, struct reg_data);
4160 Renew(RExC_rxi->data->what, count + n, U8);
4162 Newx(RExC_rxi->data->what, n, U8);
4163 RExC_rxi->data->count = count + n;
4164 Copy(s, RExC_rxi->data->what + count, n, U8);
4168 /*XXX: todo make this not included in a non debugging perl */
4169 #ifndef PERL_IN_XSUB_RE
4171 Perl_reginitcolors(pTHX)
4174 const char * const s = PerlEnv_getenv("PERL_RE_COLORS");
4176 char *t = savepv(s);
4180 t = strchr(t, '\t');
4186 PL_colors[i] = t = (char *)"";
4191 PL_colors[i++] = (char *)"";
4198 #ifdef TRIE_STUDY_OPT
4199 #define CHECK_RESTUDY_GOTO \
4201 (data.flags & SCF_TRIE_RESTUDY) \
4205 #define CHECK_RESTUDY_GOTO
4209 - pregcomp - compile a regular expression into internal code
4211 * We can't allocate space until we know how big the compiled form will be,
4212 * but we can't compile it (and thus know how big it is) until we've got a
4213 * place to put the code. So we cheat: we compile it twice, once with code
4214 * generation turned off and size counting turned on, and once "for real".
4215 * This also means that we don't allocate space until we are sure that the
4216 * thing really will compile successfully, and we never have to move the
4217 * code and thus invalidate pointers into it. (Note that it has to be in
4218 * one piece because free() must be able to free it all.) [NB: not true in perl]
4220 * Beware that the optimization-preparation code in here knows about some
4221 * of the structure of the compiled regexp. [I'll say.]
4226 #ifndef PERL_IN_XSUB_RE
4227 #define RE_ENGINE_PTR &reh_regexp_engine
4229 extern const struct regexp_engine my_reg_engine;
4230 #define RE_ENGINE_PTR &my_reg_engine
4233 #ifndef PERL_IN_XSUB_RE
4235 Perl_pregcomp(pTHX_ SV * const pattern, const U32 flags)
4238 HV * const table = GvHV(PL_hintgv);
4240 PERL_ARGS_ASSERT_PREGCOMP;
4242 /* Dispatch a request to compile a regexp to correct
4245 SV **ptr= hv_fetchs(table, "regcomp", FALSE);
4246 GET_RE_DEBUG_FLAGS_DECL;
4247 if (ptr && SvIOK(*ptr) && SvIV(*ptr)) {
4248 const regexp_engine *eng=INT2PTR(regexp_engine*,SvIV(*ptr));
4250 PerlIO_printf(Perl_debug_log, "Using engine %"UVxf"\n",
4253 return CALLREGCOMP_ENG(eng, pattern, flags);
4256 return Perl_re_compile(aTHX_ pattern, flags);
4261 Perl_re_compile(pTHX_ SV * const pattern, U32 pm_flags)
4266 register regexp_internal *ri;
4268 char *exp = SvPV(pattern, plen);
4269 char* xend = exp + plen;
4276 RExC_state_t RExC_state;
4277 RExC_state_t * const pRExC_state = &RExC_state;
4278 #ifdef TRIE_STUDY_OPT
4280 RExC_state_t copyRExC_state;
4282 GET_RE_DEBUG_FLAGS_DECL;
4284 PERL_ARGS_ASSERT_RE_COMPILE;
4286 DEBUG_r(if (!PL_colorset) reginitcolors());
4288 RExC_utf8 = RExC_orig_utf8 = SvUTF8(pattern);
4291 SV *dsv= sv_newmortal();
4292 RE_PV_QUOTED_DECL(s, RExC_utf8,
4293 dsv, exp, plen, 60);
4294 PerlIO_printf(Perl_debug_log, "%sCompiling REx%s %s\n",
4295 PL_colors[4],PL_colors[5],s);
4300 RExC_flags = pm_flags;
4304 RExC_seen_zerolen = *exp == '^' ? -1 : 0;
4305 RExC_seen_evals = 0;
4308 /* First pass: determine size, legality. */
4316 RExC_emit = &PL_regdummy;
4317 RExC_whilem_seen = 0;
4318 RExC_open_parens = NULL;
4319 RExC_close_parens = NULL;
4321 RExC_paren_names = NULL;
4323 RExC_paren_name_list = NULL;
4325 RExC_recurse = NULL;
4326 RExC_recurse_count = 0;
4328 #if 0 /* REGC() is (currently) a NOP at the first pass.
4329 * Clever compilers notice this and complain. --jhi */
4330 REGC((U8)REG_MAGIC, (char*)RExC_emit);
4332 DEBUG_PARSE_r(PerlIO_printf(Perl_debug_log, "Starting first pass (sizing)\n"));
4333 if (reg(pRExC_state, 0, &flags,1) == NULL) {
4334 RExC_precomp = NULL;
4337 if (RExC_utf8 && !RExC_orig_utf8) {
4338 /* It's possible to write a regexp in ascii that represents Unicode
4339 codepoints outside of the byte range, such as via \x{100}. If we
4340 detect such a sequence we have to convert the entire pattern to utf8
4341 and then recompile, as our sizing calculation will have been based
4342 on 1 byte == 1 character, but we will need to use utf8 to encode
4343 at least some part of the pattern, and therefore must convert the whole
4345 XXX: somehow figure out how to make this less expensive...
4348 DEBUG_PARSE_r(PerlIO_printf(Perl_debug_log,
4349 "UTF8 mismatch! Converting to utf8 for resizing and compile\n"));
4350 exp = (char*)Perl_bytes_to_utf8(aTHX_ (U8*)exp, &len);
4352 RExC_orig_utf8 = RExC_utf8;
4354 goto redo_first_pass;
4357 PerlIO_printf(Perl_debug_log,
4358 "Required size %"IVdf" nodes\n"
4359 "Starting second pass (creation)\n",
4362 RExC_lastparse=NULL;
4364 /* Small enough for pointer-storage convention?
4365 If extralen==0, this means that we will not need long jumps. */
4366 if (RExC_size >= 0x10000L && RExC_extralen)
4367 RExC_size += RExC_extralen;
4370 if (RExC_whilem_seen > 15)
4371 RExC_whilem_seen = 15;
4373 /* Allocate space and zero-initialize. Note, the two step process
4374 of zeroing when in debug mode, thus anything assigned has to
4375 happen after that */
4376 rx = (REGEXP*) newSV_type(SVt_REGEXP);
4377 r = (struct regexp*)SvANY(rx);
4378 Newxc(ri, sizeof(regexp_internal) + (unsigned)RExC_size * sizeof(regnode),
4379 char, regexp_internal);
4380 if ( r == NULL || ri == NULL )
4381 FAIL("Regexp out of space");
4383 /* avoid reading uninitialized memory in DEBUGGING code in study_chunk() */
4384 Zero(ri, sizeof(regexp_internal) + (unsigned)RExC_size * sizeof(regnode), char);
4386 /* bulk initialize base fields with 0. */
4387 Zero(ri, sizeof(regexp_internal), char);
4390 /* non-zero initialization begins here */
4392 r->engine= RE_ENGINE_PTR;
4393 r->extflags = pm_flags;
4395 bool has_p = ((r->extflags & RXf_PMf_KEEPCOPY) == RXf_PMf_KEEPCOPY);
4396 bool has_minus = ((r->extflags & RXf_PMf_STD_PMMOD) != RXf_PMf_STD_PMMOD);
4397 bool has_runon = ((RExC_seen & REG_SEEN_RUN_ON_COMMENT)==REG_SEEN_RUN_ON_COMMENT);
4398 U16 reganch = (U16)((r->extflags & RXf_PMf_STD_PMMOD)
4399 >> RXf_PMf_STD_PMMOD_SHIFT);
4400 const char *fptr = STD_PAT_MODS; /*"msix"*/
4402 const STRLEN wraplen = plen + has_minus + has_p + has_runon
4403 + (sizeof(STD_PAT_MODS) - 1)
4404 + (sizeof("(?:)") - 1);
4406 p = sv_grow(MUTABLE_SV(rx), wraplen + 1);
4407 SvCUR_set(rx, wraplen);
4409 SvFLAGS(rx) |= SvUTF8(pattern);
4412 *p++ = KEEPCOPY_PAT_MOD; /*'p'*/
4414 char *r = p + (sizeof(STD_PAT_MODS) - 1) + has_minus - 1;
4415 char *colon = r + 1;
4418 while((ch = *fptr++)) {
4432 Copy(RExC_precomp, p, plen, char);
4433 assert ((RX_WRAPPED(rx) - p) < 16);
4434 r->pre_prefix = p - RX_WRAPPED(rx);
4443 r->nparens = RExC_npar - 1; /* set early to validate backrefs */
4445 if (RExC_seen & REG_SEEN_RECURSE) {
4446 Newxz(RExC_open_parens, RExC_npar,regnode *);
4447 SAVEFREEPV(RExC_open_parens);
4448 Newxz(RExC_close_parens,RExC_npar,regnode *);
4449 SAVEFREEPV(RExC_close_parens);
4452 /* Useful during FAIL. */
4453 #ifdef RE_TRACK_PATTERN_OFFSETS
4454 Newxz(ri->u.offsets, 2*RExC_size+1, U32); /* MJD 20001228 */
4455 DEBUG_OFFSETS_r(PerlIO_printf(Perl_debug_log,
4456 "%s %"UVuf" bytes for offset annotations.\n",
4457 ri->u.offsets ? "Got" : "Couldn't get",
4458 (UV)((2*RExC_size+1) * sizeof(U32))));
4460 SetProgLen(ri,RExC_size);
4464 REH_CALL_COMP_BEGIN_HOOK(pRExC_state->rx);
4466 /* Second pass: emit code. */
4467 RExC_flags = pm_flags; /* don't let top level (?i) bleed */
4472 RExC_emit_start = ri->program;
4473 RExC_emit = ri->program;
4474 RExC_emit_bound = ri->program + RExC_size + 1;
4476 /* Store the count of eval-groups for security checks: */
4477 RExC_rx->seen_evals = RExC_seen_evals;
4478 REGC((U8)REG_MAGIC, (char*) RExC_emit++);
4479 if (reg(pRExC_state, 0, &flags,1) == NULL) {
4483 /* XXXX To minimize changes to RE engine we always allocate
4484 3-units-long substrs field. */
4485 Newx(r->substrs, 1, struct reg_substr_data);
4486 if (RExC_recurse_count) {
4487 Newxz(RExC_recurse,RExC_recurse_count,regnode *);
4488 SAVEFREEPV(RExC_recurse);
4492 r->minlen = minlen = sawplus = sawopen = 0;
4493 Zero(r->substrs, 1, struct reg_substr_data);
4495 #ifdef TRIE_STUDY_OPT
4497 StructCopy(&zero_scan_data, &data, scan_data_t);
4498 copyRExC_state = RExC_state;
4501 DEBUG_OPTIMISE_r(PerlIO_printf(Perl_debug_log,"Restudying\n"));
4503 RExC_state = copyRExC_state;
4504 if (seen & REG_TOP_LEVEL_BRANCHES)
4505 RExC_seen |= REG_TOP_LEVEL_BRANCHES;
4507 RExC_seen &= ~REG_TOP_LEVEL_BRANCHES;
4508 if (data.last_found) {
4509 SvREFCNT_dec(data.longest_fixed);
4510 SvREFCNT_dec(data.longest_float);
4511 SvREFCNT_dec(data.last_found);
4513 StructCopy(&zero_scan_data, &data, scan_data_t);
4516 StructCopy(&zero_scan_data, &data, scan_data_t);
4519 /* Dig out information for optimizations. */
4520 r->extflags = RExC_flags; /* was pm_op */
4521 /*dmq: removed as part of de-PMOP: pm->op_pmflags = RExC_flags; */
4524 SvUTF8_on(rx); /* Unicode in it? */
4525 ri->regstclass = NULL;
4526 if (RExC_naughty >= 10) /* Probably an expensive pattern. */
4527 r->intflags |= PREGf_NAUGHTY;
4528 scan = ri->program + 1; /* First BRANCH. */
4530 /* testing for BRANCH here tells us whether there is "must appear"
4531 data in the pattern. If there is then we can use it for optimisations */
4532 if (!(RExC_seen & REG_TOP_LEVEL_BRANCHES)) { /* Only one top-level choice. */
4534 STRLEN longest_float_length, longest_fixed_length;
4535 struct regnode_charclass_class ch_class; /* pointed to by data */
4537 I32 last_close = 0; /* pointed to by data */
4538 regnode *first= scan;
4539 regnode *first_next= regnext(first);
4542 * Skip introductions and multiplicators >= 1
4543 * so that we can extract the 'meat' of the pattern that must
4544 * match in the large if() sequence following.
4545 * NOTE that EXACT is NOT covered here, as it is normally
4546 * picked up by the optimiser separately.
4548 * This is unfortunate as the optimiser isnt handling lookahead
4549 * properly currently.
4552 while ((OP(first) == OPEN && (sawopen = 1)) ||
4553 /* An OR of *one* alternative - should not happen now. */
4554 (OP(first) == BRANCH && OP(first_next) != BRANCH) ||
4555 /* for now we can't handle lookbehind IFMATCH*/
4556 (OP(first) == IFMATCH && !first->flags) ||
4557 (OP(first) == PLUS) ||
4558 (OP(first) == MINMOD) ||
4559 /* An {n,m} with n>0 */
4560 (PL_regkind[OP(first)] == CURLY && ARG1(first) > 0) ||
4561 (OP(first) == NOTHING && PL_regkind[OP(first_next)] != END ))
4564 * the only op that could be a regnode is PLUS, all the rest
4565 * will be regnode_1 or regnode_2.
4568 if (OP(first) == PLUS)
4571 first += regarglen[OP(first)];
4573 first = NEXTOPER(first);
4574 first_next= regnext(first);
4577 /* Starting-point info. */
4579 DEBUG_PEEP("first:",first,0);
4580 /* Ignore EXACT as we deal with it later. */
4581 if (PL_regkind[OP(first)] == EXACT) {
4582 if (OP(first) == EXACT)
4583 NOOP; /* Empty, get anchored substr later. */
4584 else if ((OP(first) == EXACTF || OP(first) == EXACTFL))
4585 ri->regstclass = first;
4588 else if (PL_regkind[OP(first)] == TRIE &&
4589 ((reg_trie_data *)ri->data->data[ ARG(first) ])->minlen>0)
4592 /* this can happen only on restudy */
4593 if ( OP(first) == TRIE ) {
4594 struct regnode_1 *trieop = (struct regnode_1 *)
4595 PerlMemShared_calloc(1, sizeof(struct regnode_1));
4596 StructCopy(first,trieop,struct regnode_1);
4597 trie_op=(regnode *)trieop;
4599 struct regnode_charclass *trieop = (struct regnode_charclass *)
4600 PerlMemShared_calloc(1, sizeof(struct regnode_charclass));
4601 StructCopy(first,trieop,struct regnode_charclass);
4602 trie_op=(regnode *)trieop;
4605 make_trie_failtable(pRExC_state, (regnode *)first, trie_op, 0);
4606 ri->regstclass = trie_op;
4609 else if (REGNODE_SIMPLE(OP(first)))
4610 ri->regstclass = first;
4611 else if (PL_regkind[OP(first)] == BOUND ||
4612 PL_regkind[OP(first)] == NBOUND)
4613 ri->regstclass = first;
4614 else if (PL_regkind[OP(first)] == BOL) {
4615 r->extflags |= (OP(first) == MBOL
4617 : (OP(first) == SBOL
4620 first = NEXTOPER(first);
4623 else if (OP(first) == GPOS) {
4624 r->extflags |= RXf_ANCH_GPOS;
4625 first = NEXTOPER(first);
4628 else if ((!sawopen || !RExC_sawback) &&
4629 (OP(first) == STAR &&
4630 PL_regkind[OP(NEXTOPER(first))] == REG_ANY) &&
4631 !(r->extflags & RXf_ANCH) && !(RExC_seen & REG_SEEN_EVAL))
4633 /* turn .* into ^.* with an implied $*=1 */
4635 (OP(NEXTOPER(first)) == REG_ANY)
4638 r->extflags |= type;
4639 r->intflags |= PREGf_IMPLICIT;
4640 first = NEXTOPER(first);
4643 if (sawplus && (!sawopen || !RExC_sawback)
4644 && !(RExC_seen & REG_SEEN_EVAL)) /* May examine pos and $& */
4645 /* x+ must match at the 1st pos of run of x's */
4646 r->intflags |= PREGf_SKIP;
4648 /* Scan is after the zeroth branch, first is atomic matcher. */
4649 #ifdef TRIE_STUDY_OPT
4652 PerlIO_printf(Perl_debug_log, "first at %"IVdf"\n",
4653 (IV)(first - scan + 1))
4657 PerlIO_printf(Perl_debug_log, "first at %"IVdf"\n",
4658 (IV)(first - scan + 1))
4664 * If there's something expensive in the r.e., find the
4665 * longest literal string that must appear and make it the
4666 * regmust. Resolve ties in favor of later strings, since
4667 * the regstart check works with the beginning of the r.e.
4668 * and avoiding duplication strengthens checking. Not a
4669 * strong reason, but sufficient in the absence of others.
4670 * [Now we resolve ties in favor of the earlier string if
4671 * it happens that c_offset_min has been invalidated, since the
4672 * earlier string may buy us something the later one won't.]
4675 data.longest_fixed = newSVpvs("");
4676 data.longest_float = newSVpvs("");
4677 data.last_found = newSVpvs("");
4678 data.longest = &(data.longest_fixed);
4680 if (!ri->regstclass) {
4681 cl_init(pRExC_state, &ch_class);
4682 data.start_class = &ch_class;
4683 stclass_flag = SCF_DO_STCLASS_AND;
4684 } else /* XXXX Check for BOUND? */
4686 data.last_closep = &last_close;
4688 minlen = study_chunk(pRExC_state, &first, &minlen, &fake, scan + RExC_size, /* Up to end */
4689 &data, -1, NULL, NULL,
4690 SCF_DO_SUBSTR | SCF_WHILEM_VISITED_POS | stclass_flag,0);
4696 if ( RExC_npar == 1 && data.longest == &(data.longest_fixed)
4697 && data.last_start_min == 0 && data.last_end > 0
4698 && !RExC_seen_zerolen
4699 && !(RExC_seen & REG_SEEN_VERBARG)
4700 && (!(RExC_seen & REG_SEEN_GPOS) || (r->extflags & RXf_ANCH_GPOS)))
4701 r->extflags |= RXf_CHECK_ALL;
4702 scan_commit(pRExC_state, &data,&minlen,0);
4703 SvREFCNT_dec(data.last_found);
4705 /* Note that code very similar to this but for anchored string
4706 follows immediately below, changes may need to be made to both.
4709 longest_float_length = CHR_SVLEN(data.longest_float);
4710 if (longest_float_length
4711 || (data.flags & SF_FL_BEFORE_EOL
4712 && (!(data.flags & SF_FL_BEFORE_MEOL)
4713 || (RExC_flags & RXf_PMf_MULTILINE))))
4717 if (SvCUR(data.longest_fixed) /* ok to leave SvCUR */
4718 && data.offset_fixed == data.offset_float_min
4719 && SvCUR(data.longest_fixed) == SvCUR(data.longest_float))
4720 goto remove_float; /* As in (a)+. */
4722 /* copy the information about the longest float from the reg_scan_data
4723 over to the program. */
4724 if (SvUTF8(data.longest_float)) {
4725 r->float_utf8 = data.longest_float;
4726 r->float_substr = NULL;
4728 r->float_substr = data.longest_float;
4729 r->float_utf8 = NULL;
4731 /* float_end_shift is how many chars that must be matched that
4732 follow this item. We calculate it ahead of time as once the
4733 lookbehind offset is added in we lose the ability to correctly
4735 ml = data.minlen_float ? *(data.minlen_float)
4736 : (I32)longest_float_length;
4737 r->float_end_shift = ml - data.offset_float_min
4738 - longest_float_length + (SvTAIL(data.longest_float) != 0)
4739 + data.lookbehind_float;
4740 r->float_min_offset = data.offset_float_min - data.lookbehind_float;
4741 r->float_max_offset = data.offset_float_max;
4742 if (data.offset_float_max < I32_MAX) /* Don't offset infinity */
4743 r->float_max_offset -= data.lookbehind_float;
4745 t = (data.flags & SF_FL_BEFORE_EOL /* Can't have SEOL and MULTI */
4746 && (!(data.flags & SF_FL_BEFORE_MEOL)
4747 || (RExC_flags & RXf_PMf_MULTILINE)));
4748 fbm_compile(data.longest_float, t ? FBMcf_TAIL : 0);
4752 r->float_substr = r->float_utf8 = NULL;
4753 SvREFCNT_dec(data.longest_float);
4754 longest_float_length = 0;
4757 /* Note that code very similar to this but for floating string
4758 is immediately above, changes may need to be made to both.
4761 longest_fixed_length = CHR_SVLEN(data.longest_fixed);
4762 if (longest_fixed_length
4763 || (data.flags & SF_FIX_BEFORE_EOL /* Cannot have SEOL and MULTI */
4764 && (!(data.flags & SF_FIX_BEFORE_MEOL)
4765 || (RExC_flags & RXf_PMf_MULTILINE))))
4769 /* copy the information about the longest fixed
4770 from the reg_scan_data over to the program. */
4771 if (SvUTF8(data.longest_fixed)) {
4772 r->anchored_utf8 = data.longest_fixed;
4773 r->anchored_substr = NULL;
4775 r->anchored_substr = data.longest_fixed;
4776 r->anchored_utf8 = NULL;
4778 /* fixed_end_shift is how many chars that must be matched that
4779 follow this item. We calculate it ahead of time as once the
4780 lookbehind offset is added in we lose the ability to correctly
4782 ml = data.minlen_fixed ? *(data.minlen_fixed)
4783 : (I32)longest_fixed_length;
4784 r->anchored_end_shift = ml - data.offset_fixed
4785 - longest_fixed_length + (SvTAIL(data.longest_fixed) != 0)
4786 + data.lookbehind_fixed;
4787 r->anchored_offset = data.offset_fixed - data.lookbehind_fixed;
4789 t = (data.flags & SF_FIX_BEFORE_EOL /* Can't have SEOL and MULTI */
4790 && (!(data.flags & SF_FIX_BEFORE_MEOL)
4791 || (RExC_flags & RXf_PMf_MULTILINE)));
4792 fbm_compile(data.longest_fixed, t ? FBMcf_TAIL : 0);
4795 r->anchored_substr = r->anchored_utf8 = NULL;
4796 SvREFCNT_dec(data.longest_fixed);
4797 longest_fixed_length = 0;
4800 && (OP(ri->regstclass) == REG_ANY || OP(ri->regstclass) == SANY))
4801 ri->regstclass = NULL;
4802 if ((!(r->anchored_substr || r->anchored_utf8) || r->anchored_offset)
4804 && !(data.start_class->flags & ANYOF_EOS)
4805 && !cl_is_anything(data.start_class))
4807 const U32 n = add_data(pRExC_state, 1, "f");
4809 Newx(RExC_rxi->data->data[n], 1,
4810 struct regnode_charclass_class);
4811 StructCopy(data.start_class,
4812 (struct regnode_charclass_class*)RExC_rxi->data->data[n],
4813 struct regnode_charclass_class);
4814 ri->regstclass = (regnode*)RExC_rxi->data->data[n];
4815 r->intflags &= ~PREGf_SKIP; /* Used in find_byclass(). */
4816 DEBUG_COMPILE_r({ SV *sv = sv_newmortal();
4817 regprop(r, sv, (regnode*)data.start_class);
4818 PerlIO_printf(Perl_debug_log,
4819 "synthetic stclass \"%s\".\n",
4820 SvPVX_const(sv));});
4823 /* A temporary algorithm prefers floated substr to fixed one to dig more info. */
4824 if (longest_fixed_length > longest_float_length) {
4825 r->check_end_shift = r->anchored_end_shift;
4826 r->check_substr = r->anchored_substr;
4827 r->check_utf8 = r->anchored_utf8;
4828 r->check_offset_min = r->check_offset_max = r->anchored_offset;
4829 if (r->extflags & RXf_ANCH_SINGLE)
4830 r->extflags |= RXf_NOSCAN;
4833 r->check_end_shift = r->float_end_shift;
4834 r->check_substr = r->float_substr;
4835 r->check_utf8 = r->float_utf8;
4836 r->check_offset_min = r->float_min_offset;
4837 r->check_offset_max = r->float_max_offset;
4839 /* XXXX Currently intuiting is not compatible with ANCH_GPOS.
4840 This should be changed ASAP! */
4841 if ((r->check_substr || r->check_utf8) && !(r->extflags & RXf_ANCH_GPOS)) {
4842 r->extflags |= RXf_USE_INTUIT;
4843 if (SvTAIL(r->check_substr ? r->check_substr : r->check_utf8))
4844 r->extflags |= RXf_INTUIT_TAIL;
4846 /* XXX Unneeded? dmq (shouldn't as this is handled elsewhere)
4847 if ( (STRLEN)minlen < longest_float_length )
4848 minlen= longest_float_length;
4849 if ( (STRLEN)minlen < longest_fixed_length )
4850 minlen= longest_fixed_length;
4854 /* Several toplevels. Best we can is to set minlen. */
4856 struct regnode_charclass_class ch_class;
4859 DEBUG_PARSE_r(PerlIO_printf(Perl_debug_log, "\nMulti Top Level\n"));
4861 scan = ri->program + 1;
4862 cl_init(pRExC_state, &ch_class);
4863 data.start_class = &ch_class;
4864 data.last_closep = &last_close;
4867 minlen = study_chunk(pRExC_state, &scan, &minlen, &fake, scan + RExC_size,
4868 &data, -1, NULL, NULL, SCF_DO_STCLASS_AND|SCF_WHILEM_VISITED_POS,0);
4872 r->check_substr = r->check_utf8 = r->anchored_substr = r->anchored_utf8
4873 = r->float_substr = r->float_utf8 = NULL;
4874 if (!(data.start_class->flags & ANYOF_EOS)
4875 && !cl_is_anything(data.start_class))
4877 const U32 n = add_data(pRExC_state, 1, "f");
4879 Newx(RExC_rxi->data->data[n], 1,
4880 struct regnode_charclass_class);
4881 StructCopy(data.start_class,
4882 (struct regnode_charclass_class*)RExC_rxi->data->data[n],
4883 struct regnode_charclass_class);
4884 ri->regstclass = (regnode*)RExC_rxi->data->data[n];
4885 r->intflags &= ~PREGf_SKIP; /* Used in find_byclass(). */
4886 DEBUG_COMPILE_r({ SV* sv = sv_newmortal();
4887 regprop(r, sv, (regnode*)data.start_class);
4888 PerlIO_printf(Perl_debug_log,
4889 "synthetic stclass \"%s\".\n",
4890 SvPVX_const(sv));});
4894 /* Guard against an embedded (?=) or (?<=) with a longer minlen than
4895 the "real" pattern. */
4897 PerlIO_printf(Perl_debug_log,"minlen: %"IVdf" r->minlen:%"IVdf"\n",
4898 (IV)minlen, (IV)r->minlen);
4900 r->minlenret = minlen;
4901 if (r->minlen < minlen)
4904 if (RExC_seen & REG_SEEN_GPOS)
4905 r->extflags |= RXf_GPOS_SEEN;
4906 if (RExC_seen & REG_SEEN_LOOKBEHIND)
4907 r->extflags |= RXf_LOOKBEHIND_SEEN;
4908 if (RExC_seen & REG_SEEN_EVAL)
4909 r->extflags |= RXf_EVAL_SEEN;
4910 if (RExC_seen & REG_SEEN_CANY)
4911 r->extflags |= RXf_CANY_SEEN;
4912 if (RExC_seen & REG_SEEN_VERBARG)
4913 r->intflags |= PREGf_VERBARG_SEEN;
4914 if (RExC_seen & REG_SEEN_CUTGROUP)
4915 r->intflags |= PREGf_CUTGROUP_SEEN;
4916 if (RExC_paren_names)
4917 RXp_PAREN_NAMES(r) = MUTABLE_HV(SvREFCNT_inc(RExC_paren_names));
4919 RXp_PAREN_NAMES(r) = NULL;
4921 #ifdef STUPID_PATTERN_CHECKS
4922 if (RX_PRELEN(rx) == 0)
4923 r->extflags |= RXf_NULL;
4924 if (r->extflags & RXf_SPLIT && RX_PRELEN(rx) == 1 && RX_PRECOMP(rx)[0] == ' ')
4925 /* XXX: this should happen BEFORE we compile */
4926 r->extflags |= (RXf_SKIPWHITE|RXf_WHITE);
4927 else if (RX_PRELEN(rx) == 3 && memEQ("\\s+", RX_PRECOMP(rx), 3))
4928 r->extflags |= RXf_WHITE;
4929 else if (RX_PRELEN(rx) == 1 && RXp_PRECOMP(rx)[0] == '^')
4930 r->extflags |= RXf_START_ONLY;
4932 if (r->extflags & RXf_SPLIT && RX_PRELEN(rx) == 1 && RX_PRECOMP(rx)[0] == ' ')
4933 /* XXX: this should happen BEFORE we compile */
4934 r->extflags |= (RXf_SKIPWHITE|RXf_WHITE);
4936 regnode *first = ri->program + 1;
4938 U8 nop = OP(NEXTOPER(first));
4940 if (PL_regkind[fop] == NOTHING && nop == END)
4941 r->extflags |= RXf_NULL;
4942 else if (PL_regkind[fop] == BOL && nop == END)
4943 r->extflags |= RXf_START_ONLY;
4944 else if (fop == PLUS && nop ==SPACE && OP(regnext(first))==END)
4945 r->extflags |= RXf_WHITE;
4949 if (RExC_paren_names) {
4950 ri->name_list_idx = add_data( pRExC_state, 1, "a" );
4951 ri->data->data[ri->name_list_idx] = (void*)SvREFCNT_inc(RExC_paren_name_list);
4954 ri->name_list_idx = 0;
4956 if (RExC_recurse_count) {
4957 for ( ; RExC_recurse_count ; RExC_recurse_count-- ) {
4958 const regnode *scan = RExC_recurse[RExC_recurse_count-1];
4959 ARG2L_SET( scan, RExC_open_parens[ARG(scan)-1] - scan );
4962 Newxz(r->offs, RExC_npar, regexp_paren_pair);
4963 /* assume we don't need to swap parens around before we match */
4966 PerlIO_printf(Perl_debug_log,"Final program:\n");
4969 #ifdef RE_TRACK_PATTERN_OFFSETS
4970 DEBUG_OFFSETS_r(if (ri->u.offsets) {
4971 const U32 len = ri->u.offsets[0];
4973 GET_RE_DEBUG_FLAGS_DECL;
4974 PerlIO_printf(Perl_debug_log, "Offsets: [%"UVuf"]\n\t", (UV)ri->u.offsets[0]);
4975 for (i = 1; i <= len; i++) {
4976 if (ri->u.offsets[i*2-1] || ri->u.offsets[i*2])
4977 PerlIO_printf(Perl_debug_log, "%"UVuf":%"UVuf"[%"UVuf"] ",
4978 (UV)i, (UV)ri->u.offsets[i*2-1], (UV)ri->u.offsets[i*2]);
4980 PerlIO_printf(Perl_debug_log, "\n");
4986 #undef RE_ENGINE_PTR
4990 Perl_reg_named_buff(pTHX_ REGEXP * const rx, SV * const key, SV * const value,
4993 PERL_ARGS_ASSERT_REG_NAMED_BUFF;
4995 PERL_UNUSED_ARG(value);
4997 if (flags & RXapif_FETCH) {
4998 return reg_named_buff_fetch(rx, key, flags);
4999 } else if (flags & (RXapif_STORE | RXapif_DELETE | RXapif_CLEAR)) {
5000 Perl_croak_no_modify(aTHX);
5002 } else if (flags & RXapif_EXISTS) {
5003 return reg_named_buff_exists(rx, key, flags)
5006 } else if (flags & RXapif_REGNAMES) {
5007 return reg_named_buff_all(rx, flags);
5008 } else if (flags & (RXapif_SCALAR | RXapif_REGNAMES_COUNT)) {
5009 return reg_named_buff_scalar(rx, flags);
5011 Perl_croak(aTHX_ "panic: Unknown flags %d in named_buff", (int)flags);
5017 Perl_reg_named_buff_iter(pTHX_ REGEXP * const rx, const SV * const lastkey,
5020 PERL_ARGS_ASSERT_REG_NAMED_BUFF_ITER;
5021 PERL_UNUSED_ARG(lastkey);
5023 if (flags & RXapif_FIRSTKEY)
5024 return reg_named_buff_firstkey(rx, flags);
5025 else if (flags & RXapif_NEXTKEY)
5026 return reg_named_buff_nextkey(rx, flags);
5028 Perl_croak(aTHX_ "panic: Unknown flags %d in named_buff_iter", (int)flags);
5034 Perl_reg_named_buff_fetch(pTHX_ REGEXP * const r, SV * const namesv,
5037 AV *retarray = NULL;
5039 struct regexp *const rx = (struct regexp *)SvANY(r);
5041 PERL_ARGS_ASSERT_REG_NAMED_BUFF_FETCH;
5043 if (flags & RXapif_ALL)
5046 if (rx && RXp_PAREN_NAMES(rx)) {
5047 HE *he_str = hv_fetch_ent( RXp_PAREN_NAMES(rx), namesv, 0, 0 );
5050 SV* sv_dat=HeVAL(he_str);
5051 I32 *nums=(I32*)SvPVX(sv_dat);
5052 for ( i=0; i<SvIVX(sv_dat); i++ ) {
5053 if ((I32)(rx->nparens) >= nums[i]
5054 && rx->offs[nums[i]].start != -1
5055 && rx->offs[nums[i]].end != -1)
5058 CALLREG_NUMBUF_FETCH(r,nums[i],ret);
5062 ret = newSVsv(&PL_sv_undef);
5065 av_push(retarray, ret);
5068 return newRV_noinc(MUTABLE_SV(retarray));
5075 Perl_reg_named_buff_exists(pTHX_ REGEXP * const r, SV * const key,
5078 struct regexp *const rx = (struct regexp *)SvANY(r);
5080 PERL_ARGS_ASSERT_REG_NAMED_BUFF_EXISTS;
5082 if (rx && RXp_PAREN_NAMES(rx)) {
5083 if (flags & RXapif_ALL) {
5084 return hv_exists_ent(RXp_PAREN_NAMES(rx), key, 0);
5086 SV *sv = CALLREG_NAMED_BUFF_FETCH(r, key, flags);
5100 Perl_reg_named_buff_firstkey(pTHX_ REGEXP * const r, const U32 flags)
5102 struct regexp *const rx = (struct regexp *)SvANY(r);
5104 PERL_ARGS_ASSERT_REG_NAMED_BUFF_FIRSTKEY;
5106 if ( rx && RXp_PAREN_NAMES(rx) ) {
5107 (void)hv_iterinit(RXp_PAREN_NAMES(rx));
5109 return CALLREG_NAMED_BUFF_NEXTKEY(r, NULL, flags & ~RXapif_FIRSTKEY);
5116 Perl_reg_named_buff_nextkey(pTHX_ REGEXP * const r, const U32 flags)
5118 struct regexp *const rx = (struct regexp *)SvANY(r);
5119 GET_RE_DEBUG_FLAGS_DECL;
5121 PERL_ARGS_ASSERT_REG_NAMED_BUFF_NEXTKEY;
5123 if (rx && RXp_PAREN_NAMES(rx)) {
5124 HV *hv = RXp_PAREN_NAMES(rx);
5126 while ( (temphe = hv_iternext_flags(hv,0)) ) {
5129 SV* sv_dat = HeVAL(temphe);
5130 I32 *nums = (I32*)SvPVX(sv_dat);
5131 for ( i = 0; i < SvIVX(sv_dat); i++ ) {
5132 if ((I32)(rx->lastparen) >= nums[i] &&
5133 rx->offs[nums[i]].start != -1 &&
5134 rx->offs[nums[i]].end != -1)
5140 if (parno || flags & RXapif_ALL) {
5141 return newSVhek(HeKEY_hek(temphe));
5149 Perl_reg_named_buff_scalar(pTHX_ REGEXP * const r, const U32 flags)
5154 struct regexp *const rx = (struct regexp *)SvANY(r);
5156 PERL_ARGS_ASSERT_REG_NAMED_BUFF_SCALAR;
5158 if (rx && RXp_PAREN_NAMES(rx)) {
5159 if (flags & (RXapif_ALL | RXapif_REGNAMES_COUNT)) {
5160 return newSViv(HvTOTALKEYS(RXp_PAREN_NAMES(rx)));
5161 } else if (flags & RXapif_ONE) {
5162 ret = CALLREG_NAMED_BUFF_ALL(r, (flags | RXapif_REGNAMES));
5163 av = MUTABLE_AV(SvRV(ret));
5164 length = av_len(av);
5166 return newSViv(length + 1);
5168 Perl_croak(aTHX_ "panic: Unknown flags %d in named_buff_scalar", (int)flags);
5172 return &PL_sv_undef;
5176 Perl_reg_named_buff_all(pTHX_ REGEXP * const r, const U32 flags)
5178 struct regexp *const rx = (struct regexp *)SvANY(r);
5181 PERL_ARGS_ASSERT_REG_NAMED_BUFF_ALL;
5183 if (rx && RXp_PAREN_NAMES(rx)) {
5184 HV *hv= RXp_PAREN_NAMES(rx);
5186 (void)hv_iterinit(hv);
5187 while ( (temphe = hv_iternext_flags(hv,0)) ) {
5190 SV* sv_dat = HeVAL(temphe);
5191 I32 *nums = (I32*)SvPVX(sv_dat);
5192 for ( i = 0; i < SvIVX(sv_dat); i++ ) {
5193 if ((I32)(rx->lastparen) >= nums[i] &&
5194 rx->offs[nums[i]].start != -1 &&
5195 rx->offs[nums[i]].end != -1)
5201 if (parno || flags & RXapif_ALL) {
5202 av_push(av, newSVhek(HeKEY_hek(temphe)));
5207 return newRV_noinc(MUTABLE_SV(av));
5211 Perl_reg_numbered_buff_fetch(pTHX_ REGEXP * const r, const I32 paren,
5214 struct regexp *const rx = (struct regexp *)SvANY(r);
5219 PERL_ARGS_ASSERT_REG_NUMBERED_BUFF_FETCH;
5222 sv_setsv(sv,&PL_sv_undef);
5226 if (paren == RX_BUFF_IDX_PREMATCH && rx->offs[0].start != -1) {
5228 i = rx->offs[0].start;
5232 if (paren == RX_BUFF_IDX_POSTMATCH && rx->offs[0].end != -1) {
5234 s = rx->subbeg + rx->offs[0].end;
5235 i = rx->sublen - rx->offs[0].end;
5238 if ( 0 <= paren && paren <= (I32)rx->nparens &&
5239 (s1 = rx->offs[paren].start) != -1 &&
5240 (t1 = rx->offs[paren].end) != -1)
5244 s = rx->subbeg + s1;
5246 sv_setsv(sv,&PL_sv_undef);
5249 assert(rx->sublen >= (s - rx->subbeg) + i );
5251 const int oldtainted = PL_tainted;
5253 sv_setpvn(sv, s, i);
5254 PL_tainted = oldtainted;
5255 if ( (rx->extflags & RXf_CANY_SEEN)
5256 ? (RXp_MATCH_UTF8(rx)
5257 && (!i || is_utf8_string((U8*)s, i)))
5258 : (RXp_MATCH_UTF8(rx)) )
5265 if (RXp_MATCH_TAINTED(rx)) {
5266 if (SvTYPE(sv) >= SVt_PVMG) {
5267 MAGIC* const mg = SvMAGIC(sv);
5270 SvMAGIC_set(sv, mg->mg_moremagic);
5272 if ((mgt = SvMAGIC(sv))) {
5273 mg->mg_moremagic = mgt;
5274 SvMAGIC_set(sv, mg);
5284 sv_setsv(sv,&PL_sv_undef);
5290 Perl_reg_numbered_buff_store(pTHX_ REGEXP * const rx, const I32 paren,
5291 SV const * const value)
5293 PERL_ARGS_ASSERT_REG_NUMBERED_BUFF_STORE;
5295 PERL_UNUSED_ARG(rx);
5296 PERL_UNUSED_ARG(paren);
5297 PERL_UNUSED_ARG(value);
5300 Perl_croak_no_modify(aTHX);
5304 Perl_reg_numbered_buff_length(pTHX_ REGEXP * const r, const SV * const sv,
5307 struct regexp *const rx = (struct regexp *)SvANY(r);
5311 PERL_ARGS_ASSERT_REG_NUMBERED_BUFF_LENGTH;
5313 /* Some of this code was originally in C<Perl_magic_len> in F<mg.c> */
5315 /* $` / ${^PREMATCH} */
5316 case RX_BUFF_IDX_PREMATCH:
5317 if (rx->offs[0].start != -1) {
5318 i = rx->offs[0].start;
5326 /* $' / ${^POSTMATCH} */
5327 case RX_BUFF_IDX_POSTMATCH:
5328 if (rx->offs[0].end != -1) {
5329 i = rx->sublen - rx->offs[0].end;
5331 s1 = rx->offs[0].end;
5337 /* $& / ${^MATCH}, $1, $2, ... */
5339 if (paren <= (I32)rx->nparens &&
5340 (s1 = rx->offs[paren].start) != -1 &&
5341 (t1 = rx->offs[paren].end) != -1)
5346 if (ckWARN(WARN_UNINITIALIZED))
5347 report_uninit((const SV *)sv);
5352 if (i > 0 && RXp_MATCH_UTF8(rx)) {
5353 const char * const s = rx->subbeg + s1;
5358 if (is_utf8_string_loclen((U8*)s, i, &ep, &el))
5365 Perl_reg_qr_package(pTHX_ REGEXP * const rx)
5367 PERL_ARGS_ASSERT_REG_QR_PACKAGE;
5368 PERL_UNUSED_ARG(rx);
5372 return newSVpvs("Regexp");
5375 /* Scans the name of a named buffer from the pattern.
5376 * If flags is REG_RSN_RETURN_NULL returns null.
5377 * If flags is REG_RSN_RETURN_NAME returns an SV* containing the name
5378 * If flags is REG_RSN_RETURN_DATA returns the data SV* corresponding
5379 * to the parsed name as looked up in the RExC_paren_names hash.
5380 * If there is an error throws a vFAIL().. type exception.
5383 #define REG_RSN_RETURN_NULL 0
5384 #define REG_RSN_RETURN_NAME 1
5385 #define REG_RSN_RETURN_DATA 2
5388 S_reg_scan_name(pTHX_ RExC_state_t *pRExC_state, U32 flags)
5390 char *name_start = RExC_parse;
5392 PERL_ARGS_ASSERT_REG_SCAN_NAME;
5394 if (isIDFIRST_lazy_if(RExC_parse, UTF)) {
5395 /* skip IDFIRST by using do...while */
5398 RExC_parse += UTF8SKIP(RExC_parse);
5399 } while (isALNUM_utf8((U8*)RExC_parse));
5403 } while (isALNUM(*RExC_parse));
5408 = newSVpvn_flags(name_start, (int)(RExC_parse - name_start),
5409 SVs_TEMP | (UTF ? SVf_UTF8 : 0));
5410 if ( flags == REG_RSN_RETURN_NAME)
5412 else if (flags==REG_RSN_RETURN_DATA) {
5415 if ( ! sv_name ) /* should not happen*/
5416 Perl_croak(aTHX_ "panic: no svname in reg_scan_name");
5417 if (RExC_paren_names)
5418 he_str = hv_fetch_ent( RExC_paren_names, sv_name, 0, 0 );
5420 sv_dat = HeVAL(he_str);
5422 vFAIL("Reference to nonexistent named group");
5426 Perl_croak(aTHX_ "panic: bad flag in reg_scan_name");
5433 #define DEBUG_PARSE_MSG(funcname) DEBUG_PARSE_r({ \
5434 int rem=(int)(RExC_end - RExC_parse); \
5443 if (RExC_lastparse!=RExC_parse) \
5444 PerlIO_printf(Perl_debug_log," >%.*s%-*s", \
5447 iscut ? "..." : "<" \
5450 PerlIO_printf(Perl_debug_log,"%16s",""); \
5453 num = RExC_size + 1; \
5455 num=REG_NODE_NUM(RExC_emit); \
5456 if (RExC_lastnum!=num) \
5457 PerlIO_printf(Perl_debug_log,"|%4d",num); \
5459 PerlIO_printf(Perl_debug_log,"|%4s",""); \
5460 PerlIO_printf(Perl_debug_log,"|%*s%-4s", \
5461 (int)((depth*2)), "", \
5465 RExC_lastparse=RExC_parse; \
5470 #define DEBUG_PARSE(funcname) DEBUG_PARSE_r({ \
5471 DEBUG_PARSE_MSG((funcname)); \
5472 PerlIO_printf(Perl_debug_log,"%4s","\n"); \
5474 #define DEBUG_PARSE_FMT(funcname,fmt,args) DEBUG_PARSE_r({ \
5475 DEBUG_PARSE_MSG((funcname)); \
5476 PerlIO_printf(Perl_debug_log,fmt "\n",args); \
5479 - reg - regular expression, i.e. main body or parenthesized thing
5481 * Caller must absorb opening parenthesis.
5483 * Combining parenthesis handling with the base level of regular expression
5484 * is a trifle forced, but the need to tie the tails of the branches to what
5485 * follows makes it hard to avoid.
5487 #define REGTAIL(x,y,z) regtail((x),(y),(z),depth+1)
5489 #define REGTAIL_STUDY(x,y,z) regtail_study((x),(y),(z),depth+1)
5491 #define REGTAIL_STUDY(x,y,z) regtail((x),(y),(z),depth+1)
5495 S_reg(pTHX_ RExC_state_t *pRExC_state, I32 paren, I32 *flagp,U32 depth)
5496 /* paren: Parenthesized? 0=top, 1=(, inside: changed to letter. */
5499 register regnode *ret; /* Will be the head of the group. */
5500 register regnode *br;
5501 register regnode *lastbr;
5502 register regnode *ender = NULL;
5503 register I32 parno = 0;
5505 U32 oregflags = RExC_flags;
5506 bool have_branch = 0;
5508 I32 freeze_paren = 0;
5509 I32 after_freeze = 0;
5511 /* for (?g), (?gc), and (?o) warnings; warning
5512 about (?c) will warn about (?g) -- japhy */
5514 #define WASTED_O 0x01
5515 #define WASTED_G 0x02
5516 #define WASTED_C 0x04
5517 #define WASTED_GC (0x02|0x04)
5518 I32 wastedflags = 0x00;
5520 char * parse_start = RExC_parse; /* MJD */
5521 char * const oregcomp_parse = RExC_parse;
5523 GET_RE_DEBUG_FLAGS_DECL;
5525 PERL_ARGS_ASSERT_REG;
5526 DEBUG_PARSE("reg ");
5528 *flagp = 0; /* Tentatively. */
5531 /* Make an OPEN node, if parenthesized. */
5533 if ( *RExC_parse == '*') { /* (*VERB:ARG) */
5534 char *start_verb = RExC_parse;
5535 STRLEN verb_len = 0;
5536 char *start_arg = NULL;
5537 unsigned char op = 0;
5539 int internal_argval = 0; /* internal_argval is only useful if !argok */
5540 while ( *RExC_parse && *RExC_parse != ')' ) {
5541 if ( *RExC_parse == ':' ) {
5542 start_arg = RExC_parse + 1;
5548 verb_len = RExC_parse - start_verb;
5551 while ( *RExC_parse && *RExC_parse != ')' )
5553 if ( *RExC_parse != ')' )
5554 vFAIL("Unterminated verb pattern argument");
5555 if ( RExC_parse == start_arg )
5558 if ( *RExC_parse != ')' )
5559 vFAIL("Unterminated verb pattern");
5562 switch ( *start_verb ) {
5563 case 'A': /* (*ACCEPT) */
5564 if ( memEQs(start_verb,verb_len,"ACCEPT") ) {
5566 internal_argval = RExC_nestroot;
5569 case 'C': /* (*COMMIT) */
5570 if ( memEQs(start_verb,verb_len,"COMMIT") )
5573 case 'F': /* (*FAIL) */
5574 if ( verb_len==1 || memEQs(start_verb,verb_len,"FAIL") ) {
5579 case ':': /* (*:NAME) */
5580 case 'M': /* (*MARK:NAME) */
5581 if ( verb_len==0 || memEQs(start_verb,verb_len,"MARK") ) {
5586 case 'P': /* (*PRUNE) */
5587 if ( memEQs(start_verb,verb_len,"PRUNE") )
5590 case 'S': /* (*SKIP) */
5591 if ( memEQs(start_verb,verb_len,"SKIP") )
5594 case 'T': /* (*THEN) */
5595 /* [19:06] <TimToady> :: is then */
5596 if ( memEQs(start_verb,verb_len,"THEN") ) {
5598 RExC_seen |= REG_SEEN_CUTGROUP;
5604 vFAIL3("Unknown verb pattern '%.*s'",
5605 verb_len, start_verb);
5608 if ( start_arg && internal_argval ) {
5609 vFAIL3("Verb pattern '%.*s' may not have an argument",
5610 verb_len, start_verb);
5611 } else if ( argok < 0 && !start_arg ) {
5612 vFAIL3("Verb pattern '%.*s' has a mandatory argument",
5613 verb_len, start_verb);
5615 ret = reganode(pRExC_state, op, internal_argval);
5616 if ( ! internal_argval && ! SIZE_ONLY ) {
5618 SV *sv = newSVpvn( start_arg, RExC_parse - start_arg);
5619 ARG(ret) = add_data( pRExC_state, 1, "S" );
5620 RExC_rxi->data->data[ARG(ret)]=(void*)sv;
5627 if (!internal_argval)
5628 RExC_seen |= REG_SEEN_VERBARG;
5629 } else if ( start_arg ) {
5630 vFAIL3("Verb pattern '%.*s' may not have an argument",
5631 verb_len, start_verb);
5633 ret = reg_node(pRExC_state, op);
5635 nextchar(pRExC_state);
5638 if (*RExC_parse == '?') { /* (?...) */
5639 bool is_logical = 0;
5640 const char * const seqstart = RExC_parse;
5643 paren = *RExC_parse++;
5644 ret = NULL; /* For look-ahead/behind. */
5647 case 'P': /* (?P...) variants for those used to PCRE/Python */
5648 paren = *RExC_parse++;
5649 if ( paren == '<') /* (?P<...>) named capture */
5651 else if (paren == '>') { /* (?P>name) named recursion */
5652 goto named_recursion;
5654 else if (paren == '=') { /* (?P=...) named backref */
5655 /* this pretty much dupes the code for \k<NAME> in regatom(), if
5656 you change this make sure you change that */
5657 char* name_start = RExC_parse;
5659 SV *sv_dat = reg_scan_name(pRExC_state,
5660 SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
5661 if (RExC_parse == name_start || *RExC_parse != ')')
5662 vFAIL2("Sequence %.3s... not terminated",parse_start);
5665 num = add_data( pRExC_state, 1, "S" );
5666 RExC_rxi->data->data[num]=(void*)sv_dat;
5667 SvREFCNT_inc_simple_void(sv_dat);
5670 ret = reganode(pRExC_state,
5671 (U8)(FOLD ? (LOC ? NREFFL : NREFF) : NREF),
5675 Set_Node_Offset(ret, parse_start+1);
5676 Set_Node_Cur_Length(ret); /* MJD */
5678 nextchar(pRExC_state);
5682 vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
5684 case '<': /* (?<...) */
5685 if (*RExC_parse == '!')
5687 else if (*RExC_parse != '=')
5693 case '\'': /* (?'...') */
5694 name_start= RExC_parse;
5695 svname = reg_scan_name(pRExC_state,
5696 SIZE_ONLY ? /* reverse test from the others */
5697 REG_RSN_RETURN_NAME :
5698 REG_RSN_RETURN_NULL);
5699 if (RExC_parse == name_start) {
5701 vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
5704 if (*RExC_parse != paren)
5705 vFAIL2("Sequence (?%c... not terminated",
5706 paren=='>' ? '<' : paren);
5710 if (!svname) /* shouldnt happen */
5712 "panic: reg_scan_name returned NULL");
5713 if (!RExC_paren_names) {
5714 RExC_paren_names= newHV();
5715 sv_2mortal(MUTABLE_SV(RExC_paren_names));
5717 RExC_paren_name_list= newAV();
5718 sv_2mortal(MUTABLE_SV(RExC_paren_name_list));
5721 he_str = hv_fetch_ent( RExC_paren_names, svname, 1, 0 );
5723 sv_dat = HeVAL(he_str);
5725 /* croak baby croak */
5727 "panic: paren_name hash element allocation failed");
5728 } else if ( SvPOK(sv_dat) ) {
5729 /* (?|...) can mean we have dupes so scan to check
5730 its already been stored. Maybe a flag indicating
5731 we are inside such a construct would be useful,
5732 but the arrays are likely to be quite small, so
5733 for now we punt -- dmq */
5734 IV count = SvIV(sv_dat);
5735 I32 *pv = (I32*)SvPVX(sv_dat);
5737 for ( i = 0 ; i < count ; i++ ) {
5738 if ( pv[i] == RExC_npar ) {
5744 pv = (I32*)SvGROW(sv_dat, SvCUR(sv_dat) + sizeof(I32)+1);
5745 SvCUR_set(sv_dat, SvCUR(sv_dat) + sizeof(I32));
5746 pv[count] = RExC_npar;
5747 SvIV_set(sv_dat, SvIVX(sv_dat) + 1);
5750 (void)SvUPGRADE(sv_dat,SVt_PVNV);
5751 sv_setpvn(sv_dat, (char *)&(RExC_npar), sizeof(I32));
5753 SvIV_set(sv_dat, 1);
5756 if (!av_store(RExC_paren_name_list, RExC_npar, SvREFCNT_inc(svname)))
5757 SvREFCNT_dec(svname);
5760 /*sv_dump(sv_dat);*/
5762 nextchar(pRExC_state);
5764 goto capturing_parens;
5766 RExC_seen |= REG_SEEN_LOOKBEHIND;
5768 case '=': /* (?=...) */
5769 RExC_seen_zerolen++;
5771 case '!': /* (?!...) */
5772 RExC_seen_zerolen++;
5773 if (*RExC_parse == ')') {
5774 ret=reg_node(pRExC_state, OPFAIL);
5775 nextchar(pRExC_state);
5779 case '|': /* (?|...) */
5780 /* branch reset, behave like a (?:...) except that
5781 buffers in alternations share the same numbers */
5783 after_freeze = freeze_paren = RExC_npar;
5785 case ':': /* (?:...) */
5786 case '>': /* (?>...) */
5788 case '$': /* (?$...) */
5789 case '@': /* (?@...) */
5790 vFAIL2("Sequence (?%c...) not implemented", (int)paren);
5792 case '#': /* (?#...) */
5793 while (*RExC_parse && *RExC_parse != ')')
5795 if (*RExC_parse != ')')
5796 FAIL("Sequence (?#... not terminated");
5797 nextchar(pRExC_state);
5800 case '0' : /* (?0) */
5801 case 'R' : /* (?R) */
5802 if (*RExC_parse != ')')
5803 FAIL("Sequence (?R) not terminated");
5804 ret = reg_node(pRExC_state, GOSTART);
5805 *flagp |= POSTPONED;
5806 nextchar(pRExC_state);
5809 { /* named and numeric backreferences */
5811 case '&': /* (?&NAME) */
5812 parse_start = RExC_parse - 1;
5815 SV *sv_dat = reg_scan_name(pRExC_state,
5816 SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
5817 num = sv_dat ? *((I32 *)SvPVX(sv_dat)) : 0;
5819 goto gen_recurse_regop;
5822 if (!(RExC_parse[0] >= '1' && RExC_parse[0] <= '9')) {
5824 vFAIL("Illegal pattern");
5826 goto parse_recursion;
5828 case '-': /* (?-1) */
5829 if (!(RExC_parse[0] >= '1' && RExC_parse[0] <= '9')) {
5830 RExC_parse--; /* rewind to let it be handled later */
5834 case '1': case '2': case '3': case '4': /* (?1) */
5835 case '5': case '6': case '7': case '8': case '9':
5838 num = atoi(RExC_parse);
5839 parse_start = RExC_parse - 1; /* MJD */
5840 if (*RExC_parse == '-')
5842 while (isDIGIT(*RExC_parse))
5844 if (*RExC_parse!=')')
5845 vFAIL("Expecting close bracket");
5848 if ( paren == '-' ) {
5850 Diagram of capture buffer numbering.
5851 Top line is the normal capture buffer numbers
5852 Botton line is the negative indexing as from
5856 /(a(x)y)(a(b(c(?-2)d)e)f)(g(h))/
5860 num = RExC_npar + num;
5863 vFAIL("Reference to nonexistent group");
5865 } else if ( paren == '+' ) {
5866 num = RExC_npar + num - 1;
5869 ret = reganode(pRExC_state, GOSUB, num);
5871 if (num > (I32)RExC_rx->nparens) {
5873 vFAIL("Reference to nonexistent group");
5875 ARG2L_SET( ret, RExC_recurse_count++);
5877 DEBUG_OPTIMISE_MORE_r(PerlIO_printf(Perl_debug_log,
5878 "Recurse #%"UVuf" to %"IVdf"\n", (UV)ARG(ret), (IV)ARG2L(ret)));
5882 RExC_seen |= REG_SEEN_RECURSE;
5883 Set_Node_Length(ret, 1 + regarglen[OP(ret)]); /* MJD */
5884 Set_Node_Offset(ret, parse_start); /* MJD */
5886 *flagp |= POSTPONED;
5887 nextchar(pRExC_state);
5889 } /* named and numeric backreferences */
5892 case '?': /* (??...) */
5894 if (*RExC_parse != '{') {
5896 vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
5899 *flagp |= POSTPONED;
5900 paren = *RExC_parse++;
5902 case '{': /* (?{...}) */
5907 char *s = RExC_parse;
5909 RExC_seen_zerolen++;
5910 RExC_seen |= REG_SEEN_EVAL;
5911 while (count && (c = *RExC_parse)) {
5922 if (*RExC_parse != ')') {
5924 vFAIL("Sequence (?{...}) not terminated or not {}-balanced");
5928 OP_4tree *sop, *rop;
5929 SV * const sv = newSVpvn(s, RExC_parse - 1 - s);
5932 Perl_save_re_context(aTHX);
5933 rop = sv_compile_2op(sv, &sop, "re", &pad);
5934 sop->op_private |= OPpREFCOUNTED;
5935 /* re_dup will OpREFCNT_inc */
5936 OpREFCNT_set(sop, 1);
5939 n = add_data(pRExC_state, 3, "nop");
5940 RExC_rxi->data->data[n] = (void*)rop;
5941 RExC_rxi->data->data[n+1] = (void*)sop;
5942 RExC_rxi->data->data[n+2] = (void*)pad;
5945 else { /* First pass */
5946 if (PL_reginterp_cnt < ++RExC_seen_evals
5948 /* No compiled RE interpolated, has runtime
5949 components ===> unsafe. */
5950 FAIL("Eval-group not allowed at runtime, use re 'eval'");
5951 if (PL_tainting && PL_tainted)
5952 FAIL("Eval-group in insecure regular expression");
5953 #if PERL_VERSION > 8
5954 if (IN_PERL_COMPILETIME)
5959 nextchar(pRExC_state);
5961 ret = reg_node(pRExC_state, LOGICAL);
5964 REGTAIL(pRExC_state, ret, reganode(pRExC_state, EVAL, n));
5965 /* deal with the length of this later - MJD */
5968 ret = reganode(pRExC_state, EVAL, n);
5969 Set_Node_Length(ret, RExC_parse - parse_start + 1);
5970 Set_Node_Offset(ret, parse_start);
5973 case '(': /* (?(?{...})...) and (?(?=...)...) */
5976 if (RExC_parse[0] == '?') { /* (?(?...)) */
5977 if (RExC_parse[1] == '=' || RExC_parse[1] == '!'
5978 || RExC_parse[1] == '<'
5979 || RExC_parse[1] == '{') { /* Lookahead or eval. */
5982 ret = reg_node(pRExC_state, LOGICAL);
5985 REGTAIL(pRExC_state, ret, reg(pRExC_state, 1, &flag,depth+1));
5989 else if ( RExC_parse[0] == '<' /* (?(<NAME>)...) */
5990 || RExC_parse[0] == '\'' ) /* (?('NAME')...) */
5992 char ch = RExC_parse[0] == '<' ? '>' : '\'';
5993 char *name_start= RExC_parse++;
5995 SV *sv_dat=reg_scan_name(pRExC_state,
5996 SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
5997 if (RExC_parse == name_start || *RExC_parse != ch)
5998 vFAIL2("Sequence (?(%c... not terminated",
5999 (ch == '>' ? '<' : ch));
6002 num = add_data( pRExC_state, 1, "S" );
6003 RExC_rxi->data->data[num]=(void*)sv_dat;
6004 SvREFCNT_inc_simple_void(sv_dat);
6006 ret = reganode(pRExC_state,NGROUPP,num);
6007 goto insert_if_check_paren;
6009 else if (RExC_parse[0] == 'D' &&
6010 RExC_parse[1] == 'E' &&
6011 RExC_parse[2] == 'F' &&
6012 RExC_parse[3] == 'I' &&
6013 RExC_parse[4] == 'N' &&
6014 RExC_parse[5] == 'E')
6016 ret = reganode(pRExC_state,DEFINEP,0);
6019 goto insert_if_check_paren;
6021 else if (RExC_parse[0] == 'R') {
6024 if (RExC_parse[0] >= '1' && RExC_parse[0] <= '9' ) {
6025 parno = atoi(RExC_parse++);
6026 while (isDIGIT(*RExC_parse))
6028 } else if (RExC_parse[0] == '&') {
6031 sv_dat = reg_scan_name(pRExC_state,
6032 SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
6033 parno = sv_dat ? *((I32 *)SvPVX(sv_dat)) : 0;
6035 ret = reganode(pRExC_state,INSUBP,parno);
6036 goto insert_if_check_paren;
6038 else if (RExC_parse[0] >= '1' && RExC_parse[0] <= '9' ) {
6041 parno = atoi(RExC_parse++);
6043 while (isDIGIT(*RExC_parse))
6045 ret = reganode(pRExC_state, GROUPP, parno);
6047 insert_if_check_paren:
6048 if ((c = *nextchar(pRExC_state)) != ')')
6049 vFAIL("Switch condition not recognized");
6051 REGTAIL(pRExC_state, ret, reganode(pRExC_state, IFTHEN, 0));
6052 br = regbranch(pRExC_state, &flags, 1,depth+1);
6054 br = reganode(pRExC_state, LONGJMP, 0);
6056 REGTAIL(pRExC_state, br, reganode(pRExC_state, LONGJMP, 0));
6057 c = *nextchar(pRExC_state);
6062 vFAIL("(?(DEFINE)....) does not allow branches");
6063 lastbr = reganode(pRExC_state, IFTHEN, 0); /* Fake one for optimizer. */
6064 regbranch(pRExC_state, &flags, 1,depth+1);
6065 REGTAIL(pRExC_state, ret, lastbr);
6068 c = *nextchar(pRExC_state);
6073 vFAIL("Switch (?(condition)... contains too many branches");
6074 ender = reg_node(pRExC_state, TAIL);
6075 REGTAIL(pRExC_state, br, ender);
6077 REGTAIL(pRExC_state, lastbr, ender);
6078 REGTAIL(pRExC_state, NEXTOPER(NEXTOPER(lastbr)), ender);
6081 REGTAIL(pRExC_state, ret, ender);
6082 RExC_size++; /* XXX WHY do we need this?!!
6083 For large programs it seems to be required
6084 but I can't figure out why. -- dmq*/
6088 vFAIL2("Unknown switch condition (?(%.2s", RExC_parse);
6092 RExC_parse--; /* for vFAIL to print correctly */
6093 vFAIL("Sequence (? incomplete");
6097 parse_flags: /* (?i) */
6099 U32 posflags = 0, negflags = 0;
6100 U32 *flagsp = &posflags;
6102 while (*RExC_parse) {
6103 /* && strchr("iogcmsx", *RExC_parse) */
6104 /* (?g), (?gc) and (?o) are useless here
6105 and must be globally applied -- japhy */
6106 switch (*RExC_parse) {
6107 CASE_STD_PMMOD_FLAGS_PARSE_SET(flagsp);
6108 case ONCE_PAT_MOD: /* 'o' */
6109 case GLOBAL_PAT_MOD: /* 'g' */
6110 if (SIZE_ONLY && ckWARN(WARN_REGEXP)) {
6111 const I32 wflagbit = *RExC_parse == 'o' ? WASTED_O : WASTED_G;
6112 if (! (wastedflags & wflagbit) ) {
6113 wastedflags |= wflagbit;
6116 "Useless (%s%c) - %suse /%c modifier",
6117 flagsp == &negflags ? "?-" : "?",
6119 flagsp == &negflags ? "don't " : "",
6126 case CONTINUE_PAT_MOD: /* 'c' */
6127 if (SIZE_ONLY && ckWARN(WARN_REGEXP)) {
6128 if (! (wastedflags & WASTED_C) ) {
6129 wastedflags |= WASTED_GC;
6132 "Useless (%sc) - %suse /gc modifier",
6133 flagsp == &negflags ? "?-" : "?",
6134 flagsp == &negflags ? "don't " : ""
6139 case KEEPCOPY_PAT_MOD: /* 'p' */
6140 if (flagsp == &negflags) {
6142 ckWARNreg(RExC_parse + 1,"Useless use of (?-p)");
6144 *flagsp |= RXf_PMf_KEEPCOPY;
6148 if (flagsp == &negflags) {
6150 vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
6154 wastedflags = 0; /* reset so (?g-c) warns twice */
6160 RExC_flags |= posflags;
6161 RExC_flags &= ~negflags;
6163 oregflags |= posflags;
6164 oregflags &= ~negflags;
6166 nextchar(pRExC_state);
6177 vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
6182 }} /* one for the default block, one for the switch */
6189 ret = reganode(pRExC_state, OPEN, parno);
6192 RExC_nestroot = parno;
6193 if (RExC_seen & REG_SEEN_RECURSE
6194 && !RExC_open_parens[parno-1])
6196 DEBUG_OPTIMISE_MORE_r(PerlIO_printf(Perl_debug_log,
6197 "Setting open paren #%"IVdf" to %d\n",
6198 (IV)parno, REG_NODE_NUM(ret)));
6199 RExC_open_parens[parno-1]= ret;
6202 Set_Node_Length(ret, 1); /* MJD */
6203 Set_Node_Offset(ret, RExC_parse); /* MJD */
6211 /* Pick up the branches, linking them together. */
6212 parse_start = RExC_parse; /* MJD */
6213 br = regbranch(pRExC_state, &flags, 1,depth+1);
6216 if (RExC_npar > after_freeze)
6217 after_freeze = RExC_npar;
6218 RExC_npar = freeze_paren;
6221 /* branch_len = (paren != 0); */
6225 if (*RExC_parse == '|') {
6226 if (!SIZE_ONLY && RExC_extralen) {
6227 reginsert(pRExC_state, BRANCHJ, br, depth+1);
6230 reginsert(pRExC_state, BRANCH, br, depth+1);
6231 Set_Node_Length(br, paren != 0);
6232 Set_Node_Offset_To_R(br-RExC_emit_start, parse_start-RExC_start);
6236 RExC_extralen += 1; /* For BRANCHJ-BRANCH. */
6238 else if (paren == ':') {
6239 *flagp |= flags&SIMPLE;
6241 if (is_open) { /* Starts with OPEN. */
6242 REGTAIL(pRExC_state, ret, br); /* OPEN -> first. */
6244 else if (paren != '?') /* Not Conditional */
6246 *flagp |= flags & (SPSTART | HASWIDTH | POSTPONED);
6248 while (*RExC_parse == '|') {
6249 if (!SIZE_ONLY && RExC_extralen) {
6250 ender = reganode(pRExC_state, LONGJMP,0);
6251 REGTAIL(pRExC_state, NEXTOPER(NEXTOPER(lastbr)), ender); /* Append to the previous. */
6254 RExC_extralen += 2; /* Account for LONGJMP. */
6255 nextchar(pRExC_state);
6257 if (RExC_npar > after_freeze)
6258 after_freeze = RExC_npar;
6259 RExC_npar = freeze_paren;
6261 br = regbranch(pRExC_state, &flags, 0, depth+1);
6265 REGTAIL(pRExC_state, lastbr, br); /* BRANCH -> BRANCH. */
6267 *flagp |= flags & (SPSTART | HASWIDTH | POSTPONED);
6270 if (have_branch || paren != ':') {
6271 /* Make a closing node, and hook it on the end. */
6274 ender = reg_node(pRExC_state, TAIL);
6277 ender = reganode(pRExC_state, CLOSE, parno);
6278 if (!SIZE_ONLY && RExC_seen & REG_SEEN_RECURSE) {
6279 DEBUG_OPTIMISE_MORE_r(PerlIO_printf(Perl_debug_log,
6280 "Setting close paren #%"IVdf" to %d\n",
6281 (IV)parno, REG_NODE_NUM(ender)));
6282 RExC_close_parens[parno-1]= ender;
6283 if (RExC_nestroot == parno)
6286 Set_Node_Offset(ender,RExC_parse+1); /* MJD */
6287 Set_Node_Length(ender,1); /* MJD */
6293 *flagp &= ~HASWIDTH;
6296 ender = reg_node(pRExC_state, SUCCEED);
6299 ender = reg_node(pRExC_state, END);
6301 assert(!RExC_opend); /* there can only be one! */
6306 REGTAIL(pRExC_state, lastbr, ender);
6308 if (have_branch && !SIZE_ONLY) {
6310 RExC_seen |= REG_TOP_LEVEL_BRANCHES;
6312 /* Hook the tails of the branches to the closing node. */
6313 for (br = ret; br; br = regnext(br)) {
6314 const U8 op = PL_regkind[OP(br)];
6316 REGTAIL_STUDY(pRExC_state, NEXTOPER(br), ender);
6318 else if (op == BRANCHJ) {
6319 REGTAIL_STUDY(pRExC_state, NEXTOPER(NEXTOPER(br)), ender);
6327 static const char parens[] = "=!<,>";
6329 if (paren && (p = strchr(parens, paren))) {
6330 U8 node = ((p - parens) % 2) ? UNLESSM : IFMATCH;
6331 int flag = (p - parens) > 1;
6334 node = SUSPEND, flag = 0;
6335 reginsert(pRExC_state, node,ret, depth+1);
6336 Set_Node_Cur_Length(ret);
6337 Set_Node_Offset(ret, parse_start + 1);
6339 REGTAIL_STUDY(pRExC_state, ret, reg_node(pRExC_state, TAIL));
6343 /* Check for proper termination. */
6345 RExC_flags = oregflags;
6346 if (RExC_parse >= RExC_end || *nextchar(pRExC_state) != ')') {
6347 RExC_parse = oregcomp_parse;
6348 vFAIL("Unmatched (");
6351 else if (!paren && RExC_parse < RExC_end) {
6352 if (*RExC_parse == ')') {
6354 vFAIL("Unmatched )");
6357 FAIL("Junk on end of regexp"); /* "Can't happen". */
6361 RExC_npar = after_freeze;
6366 - regbranch - one alternative of an | operator
6368 * Implements the concatenation operator.
6371 S_regbranch(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, I32 first, U32 depth)
6374 register regnode *ret;
6375 register regnode *chain = NULL;
6376 register regnode *latest;
6377 I32 flags = 0, c = 0;
6378 GET_RE_DEBUG_FLAGS_DECL;
6380 PERL_ARGS_ASSERT_REGBRANCH;
6382 DEBUG_PARSE("brnc");
6387 if (!SIZE_ONLY && RExC_extralen)
6388 ret = reganode(pRExC_state, BRANCHJ,0);
6390 ret = reg_node(pRExC_state, BRANCH);
6391 Set_Node_Length(ret, 1);
6395 if (!first && SIZE_ONLY)
6396 RExC_extralen += 1; /* BRANCHJ */
6398 *flagp = WORST; /* Tentatively. */
6401 nextchar(pRExC_state);
6402 while (RExC_parse < RExC_end && *RExC_parse != '|' && *RExC_parse != ')') {
6404 latest = regpiece(pRExC_state, &flags,depth+1);
6405 if (latest == NULL) {
6406 if (flags & TRYAGAIN)
6410 else if (ret == NULL)
6412 *flagp |= flags&(HASWIDTH|POSTPONED);
6413 if (chain == NULL) /* First piece. */
6414 *flagp |= flags&SPSTART;
6417 REGTAIL(pRExC_state, chain, latest);
6422 if (chain == NULL) { /* Loop ran zero times. */
6423 chain = reg_node(pRExC_state, NOTHING);
6428 *flagp |= flags&SIMPLE;
6435 - regpiece - something followed by possible [*+?]
6437 * Note that the branching code sequences used for ? and the general cases
6438 * of * and + are somewhat optimized: they use the same NOTHING node as
6439 * both the endmarker for their branch list and the body of the last branch.
6440 * It might seem that this node could be dispensed with entirely, but the
6441 * endmarker role is not redundant.
6444 S_regpiece(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, U32 depth)
6447 register regnode *ret;
6449 register char *next;
6451 const char * const origparse = RExC_parse;
6453 I32 max = REG_INFTY;
6455 const char *maxpos = NULL;
6456 GET_RE_DEBUG_FLAGS_DECL;
6458 PERL_ARGS_ASSERT_REGPIECE;
6460 DEBUG_PARSE("piec");
6462 ret = regatom(pRExC_state, &flags,depth+1);
6464 if (flags & TRYAGAIN)
6471 if (op == '{' && regcurly(RExC_parse)) {
6473 parse_start = RExC_parse; /* MJD */
6474 next = RExC_parse + 1;
6475 while (isDIGIT(*next) || *next == ',') {
6484 if (*next == '}') { /* got one */
6488 min = atoi(RExC_parse);
6492 maxpos = RExC_parse;
6494 if (!max && *maxpos != '0')
6495 max = REG_INFTY; /* meaning "infinity" */
6496 else if (max >= REG_INFTY)
6497 vFAIL2("Quantifier in {,} bigger than %d", REG_INFTY - 1);
6499 nextchar(pRExC_state);
6502 if ((flags&SIMPLE)) {
6503 RExC_naughty += 2 + RExC_naughty / 2;
6504 reginsert(pRExC_state, CURLY, ret, depth+1);
6505 Set_Node_Offset(ret, parse_start+1); /* MJD */
6506 Set_Node_Cur_Length(ret);
6509 regnode * const w = reg_node(pRExC_state, WHILEM);
6512 REGTAIL(pRExC_state, ret, w);
6513 if (!SIZE_ONLY && RExC_extralen) {
6514 reginsert(pRExC_state, LONGJMP,ret, depth+1);
6515 reginsert(pRExC_state, NOTHING,ret, depth+1);
6516 NEXT_OFF(ret) = 3; /* Go over LONGJMP. */
6518 reginsert(pRExC_state, CURLYX,ret, depth+1);
6520 Set_Node_Offset(ret, parse_start+1);
6521 Set_Node_Length(ret,
6522 op == '{' ? (RExC_parse - parse_start) : 1);
6524 if (!SIZE_ONLY && RExC_extralen)
6525 NEXT_OFF(ret) = 3; /* Go over NOTHING to LONGJMP. */
6526 REGTAIL(pRExC_state, ret, reg_node(pRExC_state, NOTHING));
6528 RExC_whilem_seen++, RExC_extralen += 3;
6529 RExC_naughty += 4 + RExC_naughty; /* compound interest */
6538 vFAIL("Can't do {n,m} with n > m");
6540 ARG1_SET(ret, (U16)min);
6541 ARG2_SET(ret, (U16)max);
6553 #if 0 /* Now runtime fix should be reliable. */
6555 /* if this is reinstated, don't forget to put this back into perldiag:
6557 =item Regexp *+ operand could be empty at {#} in regex m/%s/
6559 (F) The part of the regexp subject to either the * or + quantifier
6560 could match an empty string. The {#} shows in the regular
6561 expression about where the problem was discovered.
6565 if (!(flags&HASWIDTH) && op != '?')
6566 vFAIL("Regexp *+ operand could be empty");
6569 parse_start = RExC_parse;
6570 nextchar(pRExC_state);
6572 *flagp = (op != '+') ? (WORST|SPSTART|HASWIDTH) : (WORST|HASWIDTH);
6574 if (op == '*' && (flags&SIMPLE)) {
6575 reginsert(pRExC_state, STAR, ret, depth+1);
6579 else if (op == '*') {
6583 else if (op == '+' && (flags&SIMPLE)) {
6584 reginsert(pRExC_state, PLUS, ret, depth+1);
6588 else if (op == '+') {
6592 else if (op == '?') {
6597 if (!SIZE_ONLY && !(flags&(HASWIDTH|POSTPONED)) && max > REG_INFTY/3) {
6598 ckWARN3reg(RExC_parse,
6599 "%.*s matches null string many times",
6600 (int)(RExC_parse >= origparse ? RExC_parse - origparse : 0),
6604 if (RExC_parse < RExC_end && *RExC_parse == '?') {
6605 nextchar(pRExC_state);
6606 reginsert(pRExC_state, MINMOD, ret, depth+1);
6607 REGTAIL(pRExC_state, ret, ret + NODE_STEP_REGNODE);
6609 #ifndef REG_ALLOW_MINMOD_SUSPEND
6612 if (RExC_parse < RExC_end && *RExC_parse == '+') {
6614 nextchar(pRExC_state);
6615 ender = reg_node(pRExC_state, SUCCEED);
6616 REGTAIL(pRExC_state, ret, ender);
6617 reginsert(pRExC_state, SUSPEND, ret, depth+1);
6619 ender = reg_node(pRExC_state, TAIL);
6620 REGTAIL(pRExC_state, ret, ender);
6624 if (RExC_parse < RExC_end && ISMULT2(RExC_parse)) {
6626 vFAIL("Nested quantifiers");
6633 /* reg_namedseq(pRExC_state,UVp)
6635 This is expected to be called by a parser routine that has
6636 recognized '\N' and needs to handle the rest. RExC_parse is
6637 expected to point at the first char following the N at the time
6640 The \N may be inside (indicated by valuep not being NULL) or outside a
6643 \N may begin either a named sequence, or if outside a character class, mean
6644 to match a non-newline. For non single-quoted regexes, the tokenizer has
6645 attempted to decide which, and in the case of a named sequence converted it
6646 into one of the forms: \N{} (if the sequence is null), or \N{U+c1.c2...},
6647 where c1... are the characters in the sequence. For single-quoted regexes,
6648 the tokenizer passes the \N sequence through unchanged; this code will not
6649 attempt to determine this nor expand those. The net effect is that if the
6650 beginning of the passed-in pattern isn't '{U+' or there is no '}', it
6651 signals that this \N occurrence means to match a non-newline.
6653 Only the \N{U+...} form should occur in a character class, for the same
6654 reason that '.' inside a character class means to just match a period: it
6655 just doesn't make sense.
6657 If valuep is non-null then it is assumed that we are parsing inside
6658 of a charclass definition and the first codepoint in the resolved
6659 string is returned via *valuep and the routine will return NULL.
6660 In this mode if a multichar string is returned from the charnames
6661 handler, a warning will be issued, and only the first char in the
6662 sequence will be examined. If the string returned is zero length
6663 then the value of *valuep is undefined and NON-NULL will
6664 be returned to indicate failure. (This will NOT be a valid pointer
6667 If valuep is null then it is assumed that we are parsing normal text and a
6668 new EXACT node is inserted into the program containing the resolved string,
6669 and a pointer to the new node is returned. But if the string is zero length
6670 a NOTHING node is emitted instead.
6672 On success RExC_parse is set to the char following the endbrace.
6673 Parsing failures will generate a fatal error via vFAIL(...)
6676 S_reg_namedseq(pTHX_ RExC_state_t *pRExC_state, UV *valuep, I32 *flagp)
6678 char * endbrace; /* '}' following the name */
6679 regnode *ret = NULL;
6681 char* parse_start = RExC_parse - 2; /* points to the '\N' */
6685 GET_RE_DEBUG_FLAGS_DECL;
6687 PERL_ARGS_ASSERT_REG_NAMEDSEQ;
6691 /* The [^\n] meaning of \N ignores spaces and comments under the /x
6692 * modifier. The other meaning does not */
6693 p = (RExC_flags & RXf_PMf_EXTENDED)
6694 ? regwhite( pRExC_state, RExC_parse )
6697 /* Disambiguate between \N meaning a named character versus \N meaning
6698 * [^\n]. The former is assumed when it can't be the latter. */
6699 if (*p != '{' || regcurly(p)) {
6702 /* no bare \N in a charclass */
6703 vFAIL("\\N in a character class must be a named character: \\N{...}");
6705 nextchar(pRExC_state);
6706 ret = reg_node(pRExC_state, REG_ANY);
6707 *flagp |= HASWIDTH|SIMPLE;
6710 Set_Node_Length(ret, 1); /* MJD */
6714 /* Here, we have decided it should be a named sequence */
6716 /* The test above made sure that the next real character is a '{', but
6717 * under the /x modifier, it could be separated by space (or a comment and
6718 * \n) and this is not allowed (for consistency with \x{...} and the
6719 * tokenizer handling of \N{NAME}). */
6720 if (*RExC_parse != '{') {
6721 vFAIL("Missing braces on \\N{}");
6724 RExC_parse++; /* Skip past the '{' */
6726 if (! (endbrace = strchr(RExC_parse, '}')) /* no trailing brace */
6727 || ! (endbrace == RExC_parse /* nothing between the {} */
6728 || (endbrace - RExC_parse >= 2 /* U+ (bad hex is checked below */
6729 && strnEQ(RExC_parse, "U+", 2)))) /* for a better error msg) */
6731 if (endbrace) RExC_parse = endbrace; /* position msg's '<--HERE' */
6732 vFAIL("\\N{NAME} must be resolved by the lexer");
6735 if (endbrace == RExC_parse) { /* empty: \N{} */
6737 RExC_parse = endbrace + 1;
6738 return reg_node(pRExC_state,NOTHING);
6742 ckWARNreg(RExC_parse,
6743 "Ignoring zero length \\N{} in character class"
6745 RExC_parse = endbrace + 1;
6748 return (regnode *) &RExC_parse; /* Invalid regnode pointer */
6751 RExC_utf8 = 1; /* named sequences imply Unicode semantics */
6752 RExC_parse += 2; /* Skip past the 'U+' */
6754 if (valuep) { /* In a bracketed char class */
6755 /* We only pay attention to the first char of
6756 multichar strings being returned. I kinda wonder
6757 if this makes sense as it does change the behaviour
6758 from earlier versions, OTOH that behaviour was broken
6759 as well. XXX Solution is to recharacterize as
6760 [rest-of-class]|multi1|multi2... */
6762 STRLEN length_of_hex;
6763 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
6764 | PERL_SCAN_DISALLOW_PREFIX
6765 | (SIZE_ONLY ? PERL_SCAN_SILENT_ILLDIGIT : 0);
6767 char * endchar = RExC_parse + strcspn(RExC_parse, ".}");
6768 if (endchar < endbrace) {
6769 ckWARNreg(endchar, "Using just the first character returned by \\N{} in character class");
6772 length_of_hex = (STRLEN)(endchar - RExC_parse);
6773 *valuep = grok_hex(RExC_parse, &length_of_hex, &flags, NULL);
6775 /* The tokenizer should have guaranteed validity, but it's possible to
6776 * bypass it by using single quoting, so check */
6777 if (length_of_hex == 0
6778 || length_of_hex != (STRLEN)(endchar - RExC_parse) )
6780 RExC_parse += length_of_hex; /* Includes all the valid */
6781 RExC_parse += (RExC_orig_utf8) /* point to after 1st invalid */
6782 ? UTF8SKIP(RExC_parse)
6784 /* Guard against malformed utf8 */
6785 if (RExC_parse >= endchar) RExC_parse = endchar;
6786 vFAIL("Invalid hexadecimal number in \\N{U+...}");
6789 RExC_parse = endbrace + 1;
6790 if (endchar == endbrace) return NULL;
6792 ret = (regnode *) &RExC_parse; /* Invalid regnode pointer */
6794 else { /* Not a char class */
6795 char *s; /* String to put in generated EXACT node */
6796 STRLEN len = 0; /* Its current length */
6797 char *endchar; /* Points to '.' or '}' ending cur char in the input
6800 ret = reg_node(pRExC_state,
6801 (U8)(FOLD ? (LOC ? EXACTFL : EXACTF) : EXACT));
6804 /* Exact nodes can hold only a U8 length's of text = 255. Loop through
6805 * the input which is of the form now 'c1.c2.c3...}' until find the
6806 * ending brace or exeed length 255. The characters that exceed this
6807 * limit are dropped. The limit could be relaxed should it become
6808 * desirable by reparsing this as (?:\N{NAME}), so could generate
6809 * multiple EXACT nodes, as is done for just regular input. But this
6810 * is primarily a named character, and not intended to be a huge long
6811 * string, so 255 bytes should be good enough */
6813 STRLEN length_of_hex;
6814 I32 grok_flags = PERL_SCAN_ALLOW_UNDERSCORES
6815 | PERL_SCAN_DISALLOW_PREFIX
6816 | (SIZE_ONLY ? PERL_SCAN_SILENT_ILLDIGIT : 0);
6817 UV cp; /* Ord of current character */
6819 /* Code points are separated by dots. If none, there is only one
6820 * code point, and is terminated by the brace */
6821 endchar = RExC_parse + strcspn(RExC_parse, ".}");
6823 /* The values are Unicode even on EBCDIC machines */
6824 length_of_hex = (STRLEN)(endchar - RExC_parse);
6825 cp = grok_hex(RExC_parse, &length_of_hex, &grok_flags, NULL);
6826 if ( length_of_hex == 0
6827 || length_of_hex != (STRLEN)(endchar - RExC_parse) )
6829 RExC_parse += length_of_hex; /* Includes all the valid */
6830 RExC_parse += (RExC_orig_utf8) /* point to after 1st invalid */
6831 ? UTF8SKIP(RExC_parse)
6833 /* Guard against malformed utf8 */
6834 if (RExC_parse >= endchar) RExC_parse = endchar;
6835 vFAIL("Invalid hexadecimal number in \\N{U+...}");
6838 if (! FOLD) { /* Not folding, just append to the string */
6841 /* Quit before adding this character if would exceed limit */
6842 if (len + UNISKIP(cp) > U8_MAX) break;
6844 unilen = reguni(pRExC_state, cp, s);
6849 } else { /* Folding, output the folded equivalent */
6850 STRLEN foldlen,numlen;
6851 U8 tmpbuf[UTF8_MAXBYTES_CASE+1], *foldbuf;
6852 cp = toFOLD_uni(cp, tmpbuf, &foldlen);
6854 /* Quit before exceeding size limit */
6855 if (len + foldlen > U8_MAX) break;
6857 for (foldbuf = tmpbuf;
6861 cp = utf8_to_uvchr(foldbuf, &numlen);
6863 const STRLEN unilen = reguni(pRExC_state, cp, s);
6866 /* In EBCDIC the numlen and unilen can differ. */
6868 if (numlen >= foldlen)
6872 break; /* "Can't happen." */
6876 /* Point to the beginning of the next character in the sequence. */
6877 RExC_parse = endchar + 1;
6879 /* Quit if no more characters */
6880 if (RExC_parse >= endbrace) break;
6885 if (RExC_parse < endbrace) {
6886 ckWARNreg(RExC_parse - 1,
6887 "Using just the first characters returned by \\N{}");
6890 RExC_size += STR_SZ(len);
6893 RExC_emit += STR_SZ(len);
6896 RExC_parse = endbrace + 1;
6898 *flagp |= HASWIDTH; /* Not SIMPLE, as that causes the engine to fail
6899 with malformed in t/re/pat_advanced.t */
6901 Set_Node_Cur_Length(ret); /* MJD */
6902 nextchar(pRExC_state);
6912 * It returns the code point in utf8 for the value in *encp.
6913 * value: a code value in the source encoding
6914 * encp: a pointer to an Encode object
6916 * If the result from Encode is not a single character,
6917 * it returns U+FFFD (Replacement character) and sets *encp to NULL.
6920 S_reg_recode(pTHX_ const char value, SV **encp)
6923 SV * const sv = newSVpvn_flags(&value, numlen, SVs_TEMP);
6924 const char * const s = *encp ? sv_recode_to_utf8(sv, *encp) : SvPVX(sv);
6925 const STRLEN newlen = SvCUR(sv);
6926 UV uv = UNICODE_REPLACEMENT;
6928 PERL_ARGS_ASSERT_REG_RECODE;
6932 ? utf8n_to_uvchr((U8*)s, newlen, &numlen, UTF8_ALLOW_DEFAULT)
6935 if (!newlen || numlen != newlen) {
6936 uv = UNICODE_REPLACEMENT;
6944 - regatom - the lowest level
6946 Try to identify anything special at the start of the pattern. If there
6947 is, then handle it as required. This may involve generating a single regop,
6948 such as for an assertion; or it may involve recursing, such as to
6949 handle a () structure.
6951 If the string doesn't start with something special then we gobble up
6952 as much literal text as we can.
6954 Once we have been able to handle whatever type of thing started the
6955 sequence, we return.
6957 Note: we have to be careful with escapes, as they can be both literal
6958 and special, and in the case of \10 and friends can either, depending
6959 on context. Specifically there are two seperate switches for handling
6960 escape sequences, with the one for handling literal escapes requiring
6961 a dummy entry for all of the special escapes that are actually handled
6966 S_regatom(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, U32 depth)
6969 register regnode *ret = NULL;
6971 char *parse_start = RExC_parse;
6972 GET_RE_DEBUG_FLAGS_DECL;
6973 DEBUG_PARSE("atom");
6974 *flagp = WORST; /* Tentatively. */
6976 PERL_ARGS_ASSERT_REGATOM;
6979 switch ((U8)*RExC_parse) {
6981 RExC_seen_zerolen++;
6982 nextchar(pRExC_state);
6983 if (RExC_flags & RXf_PMf_MULTILINE)
6984 ret = reg_node(pRExC_state, MBOL);
6985 else if (RExC_flags & RXf_PMf_SINGLELINE)
6986 ret = reg_node(pRExC_state, SBOL);
6988 ret = reg_node(pRExC_state, BOL);
6989 Set_Node_Length(ret, 1); /* MJD */
6992 nextchar(pRExC_state);
6994 RExC_seen_zerolen++;
6995 if (RExC_flags & RXf_PMf_MULTILINE)
6996 ret = reg_node(pRExC_state, MEOL);
6997 else if (RExC_flags & RXf_PMf_SINGLELINE)
6998 ret = reg_node(pRExC_state, SEOL);
7000 ret = reg_node(pRExC_state, EOL);
7001 Set_Node_Length(ret, 1); /* MJD */
7004 nextchar(pRExC_state);
7005 if (RExC_flags & RXf_PMf_SINGLELINE)
7006 ret = reg_node(pRExC_state, SANY);
7008 ret = reg_node(pRExC_state, REG_ANY);
7009 *flagp |= HASWIDTH|SIMPLE;
7011 Set_Node_Length(ret, 1); /* MJD */
7015 char * const oregcomp_parse = ++RExC_parse;
7016 ret = regclass(pRExC_state,depth+1);
7017 if (*RExC_parse != ']') {
7018 RExC_parse = oregcomp_parse;
7019 vFAIL("Unmatched [");
7021 nextchar(pRExC_state);
7022 *flagp |= HASWIDTH|SIMPLE;
7023 Set_Node_Length(ret, RExC_parse - oregcomp_parse + 1); /* MJD */
7027 nextchar(pRExC_state);
7028 ret = reg(pRExC_state, 1, &flags,depth+1);
7030 if (flags & TRYAGAIN) {
7031 if (RExC_parse == RExC_end) {
7032 /* Make parent create an empty node if needed. */
7040 *flagp |= flags&(HASWIDTH|SPSTART|SIMPLE|POSTPONED);
7044 if (flags & TRYAGAIN) {
7048 vFAIL("Internal urp");
7049 /* Supposed to be caught earlier. */
7052 if (!regcurly(RExC_parse)) {
7061 vFAIL("Quantifier follows nothing");
7069 len=0; /* silence a spurious compiler warning */
7070 if ((cp = what_len_TRICKYFOLD_safe(RExC_parse,RExC_end,UTF,len))) {
7071 *flagp |= HASWIDTH; /* could be SIMPLE too, but needs a handler in regexec.regrepeat */
7072 RExC_parse+=len-1; /* we get one from nextchar() as well. :-( */
7073 ret = reganode(pRExC_state, FOLDCHAR, cp);
7074 Set_Node_Length(ret, 1); /* MJD */
7075 nextchar(pRExC_state); /* kill whitespace under /x */
7083 This switch handles escape sequences that resolve to some kind
7084 of special regop and not to literal text. Escape sequnces that
7085 resolve to literal text are handled below in the switch marked
7088 Every entry in this switch *must* have a corresponding entry
7089 in the literal escape switch. However, the opposite is not
7090 required, as the default for this switch is to jump to the
7091 literal text handling code.
7093 switch ((U8)*++RExC_parse) {
7098 /* Special Escapes */
7100 RExC_seen_zerolen++;
7101 ret = reg_node(pRExC_state, SBOL);
7103 goto finish_meta_pat;
7105 ret = reg_node(pRExC_state, GPOS);
7106 RExC_seen |= REG_SEEN_GPOS;
7108 goto finish_meta_pat;
7110 RExC_seen_zerolen++;
7111 ret = reg_node(pRExC_state, KEEPS);
7113 /* XXX:dmq : disabling in-place substitution seems to
7114 * be necessary here to avoid cases of memory corruption, as
7115 * with: C<$_="x" x 80; s/x\K/y/> -- rgs
7117 RExC_seen |= REG_SEEN_LOOKBEHIND;
7118 goto finish_meta_pat;
7120 ret = reg_node(pRExC_state, SEOL);
7122 RExC_seen_zerolen++; /* Do not optimize RE away */
7123 goto finish_meta_pat;
7125 ret = reg_node(pRExC_state, EOS);
7127 RExC_seen_zerolen++; /* Do not optimize RE away */
7128 goto finish_meta_pat;
7130 ret = reg_node(pRExC_state, CANY);
7131 RExC_seen |= REG_SEEN_CANY;
7132 *flagp |= HASWIDTH|SIMPLE;
7133 goto finish_meta_pat;
7135 ret = reg_node(pRExC_state, CLUMP);
7137 goto finish_meta_pat;
7139 ret = reg_node(pRExC_state, (U8)(LOC ? ALNUML : ALNUM));
7140 *flagp |= HASWIDTH|SIMPLE;
7141 goto finish_meta_pat;
7143 ret = reg_node(pRExC_state, (U8)(LOC ? NALNUML : NALNUM));
7144 *flagp |= HASWIDTH|SIMPLE;
7145 goto finish_meta_pat;
7147 RExC_seen_zerolen++;
7148 RExC_seen |= REG_SEEN_LOOKBEHIND;
7149 ret = reg_node(pRExC_state, (U8)(LOC ? BOUNDL : BOUND));
7151 goto finish_meta_pat;
7153 RExC_seen_zerolen++;
7154 RExC_seen |= REG_SEEN_LOOKBEHIND;
7155 ret = reg_node(pRExC_state, (U8)(LOC ? NBOUNDL : NBOUND));
7157 goto finish_meta_pat;
7159 ret = reg_node(pRExC_state, (U8)(LOC ? SPACEL : SPACE));
7160 *flagp |= HASWIDTH|SIMPLE;
7161 goto finish_meta_pat;
7163 ret = reg_node(pRExC_state, (U8)(LOC ? NSPACEL : NSPACE));
7164 *flagp |= HASWIDTH|SIMPLE;
7165 goto finish_meta_pat;
7167 ret = reg_node(pRExC_state, DIGIT);
7168 *flagp |= HASWIDTH|SIMPLE;
7169 goto finish_meta_pat;
7171 ret = reg_node(pRExC_state, NDIGIT);
7172 *flagp |= HASWIDTH|SIMPLE;
7173 goto finish_meta_pat;
7175 ret = reg_node(pRExC_state, LNBREAK);
7176 *flagp |= HASWIDTH|SIMPLE;
7177 goto finish_meta_pat;
7179 ret = reg_node(pRExC_state, HORIZWS);
7180 *flagp |= HASWIDTH|SIMPLE;
7181 goto finish_meta_pat;
7183 ret = reg_node(pRExC_state, NHORIZWS);
7184 *flagp |= HASWIDTH|SIMPLE;
7185 goto finish_meta_pat;
7187 ret = reg_node(pRExC_state, VERTWS);
7188 *flagp |= HASWIDTH|SIMPLE;
7189 goto finish_meta_pat;
7191 ret = reg_node(pRExC_state, NVERTWS);
7192 *flagp |= HASWIDTH|SIMPLE;
7194 nextchar(pRExC_state);
7195 Set_Node_Length(ret, 2); /* MJD */
7200 char* const oldregxend = RExC_end;
7202 char* parse_start = RExC_parse - 2;
7205 if (RExC_parse[1] == '{') {
7206 /* a lovely hack--pretend we saw [\pX] instead */
7207 RExC_end = strchr(RExC_parse, '}');
7209 const U8 c = (U8)*RExC_parse;
7211 RExC_end = oldregxend;
7212 vFAIL2("Missing right brace on \\%c{}", c);
7217 RExC_end = RExC_parse + 2;
7218 if (RExC_end > oldregxend)
7219 RExC_end = oldregxend;
7223 ret = regclass(pRExC_state,depth+1);
7225 RExC_end = oldregxend;
7228 Set_Node_Offset(ret, parse_start + 2);
7229 Set_Node_Cur_Length(ret);
7230 nextchar(pRExC_state);
7231 *flagp |= HASWIDTH|SIMPLE;
7235 /* Handle \N and \N{NAME} here and not below because it can be
7236 multicharacter. join_exact() will join them up later on.
7237 Also this makes sure that things like /\N{BLAH}+/ and
7238 \N{BLAH} being multi char Just Happen. dmq*/
7240 ret= reg_namedseq(pRExC_state, NULL, flagp);
7242 case 'k': /* Handle \k<NAME> and \k'NAME' */
7245 char ch= RExC_parse[1];
7246 if (ch != '<' && ch != '\'' && ch != '{') {
7248 vFAIL2("Sequence %.2s... not terminated",parse_start);
7250 /* this pretty much dupes the code for (?P=...) in reg(), if
7251 you change this make sure you change that */
7252 char* name_start = (RExC_parse += 2);
7254 SV *sv_dat = reg_scan_name(pRExC_state,
7255 SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
7256 ch= (ch == '<') ? '>' : (ch == '{') ? '}' : '\'';
7257 if (RExC_parse == name_start || *RExC_parse != ch)
7258 vFAIL2("Sequence %.3s... not terminated",parse_start);
7261 num = add_data( pRExC_state, 1, "S" );
7262 RExC_rxi->data->data[num]=(void*)sv_dat;
7263 SvREFCNT_inc_simple_void(sv_dat);
7267 ret = reganode(pRExC_state,
7268 (U8)(FOLD ? (LOC ? NREFFL : NREFF) : NREF),
7272 /* override incorrect value set in reganode MJD */
7273 Set_Node_Offset(ret, parse_start+1);
7274 Set_Node_Cur_Length(ret); /* MJD */
7275 nextchar(pRExC_state);
7281 case '1': case '2': case '3': case '4':
7282 case '5': case '6': case '7': case '8': case '9':
7285 bool isg = *RExC_parse == 'g';
7290 if (*RExC_parse == '{') {
7294 if (*RExC_parse == '-') {
7298 if (hasbrace && !isDIGIT(*RExC_parse)) {
7299 if (isrel) RExC_parse--;
7301 goto parse_named_seq;
7303 num = atoi(RExC_parse);
7304 if (isg && num == 0)
7305 vFAIL("Reference to invalid group 0");
7307 num = RExC_npar - num;
7309 vFAIL("Reference to nonexistent or unclosed group");
7311 if (!isg && num > 9 && num >= RExC_npar)
7314 char * const parse_start = RExC_parse - 1; /* MJD */
7315 while (isDIGIT(*RExC_parse))
7317 if (parse_start == RExC_parse - 1)
7318 vFAIL("Unterminated \\g... pattern");
7320 if (*RExC_parse != '}')
7321 vFAIL("Unterminated \\g{...} pattern");
7325 if (num > (I32)RExC_rx->nparens)
7326 vFAIL("Reference to nonexistent group");
7329 ret = reganode(pRExC_state,
7330 (U8)(FOLD ? (LOC ? REFFL : REFF) : REF),
7334 /* override incorrect value set in reganode MJD */
7335 Set_Node_Offset(ret, parse_start+1);
7336 Set_Node_Cur_Length(ret); /* MJD */
7338 nextchar(pRExC_state);
7343 if (RExC_parse >= RExC_end)
7344 FAIL("Trailing \\");
7347 /* Do not generate "unrecognized" warnings here, we fall
7348 back into the quick-grab loop below */
7355 if (RExC_flags & RXf_PMf_EXTENDED) {
7356 if ( reg_skipcomment( pRExC_state ) )
7363 register STRLEN len;
7369 U8 tmpbuf[UTF8_MAXBYTES_CASE+1], *foldbuf;
7371 parse_start = RExC_parse - 1;
7377 ret = reg_node(pRExC_state,
7378 (U8)(FOLD ? (LOC ? EXACTFL : EXACTF) : EXACT));
7380 for (len = 0, p = RExC_parse - 1;
7381 len < 127 && p < RExC_end;
7384 char * const oldp = p;
7386 if (RExC_flags & RXf_PMf_EXTENDED)
7387 p = regwhite( pRExC_state, p );
7392 if (LOC || !FOLD || !is_TRICKYFOLD_safe(p,RExC_end,UTF))
7393 goto normal_default;
7403 /* Literal Escapes Switch
7405 This switch is meant to handle escape sequences that
7406 resolve to a literal character.
7408 Every escape sequence that represents something
7409 else, like an assertion or a char class, is handled
7410 in the switch marked 'Special Escapes' above in this
7411 routine, but also has an entry here as anything that
7412 isn't explicitly mentioned here will be treated as
7413 an unescaped equivalent literal.
7417 /* These are all the special escapes. */
7421 if (LOC || !FOLD || !is_TRICKYFOLD_safe(p,RExC_end,UTF))
7422 goto normal_default;
7423 case 'A': /* Start assertion */
7424 case 'b': case 'B': /* Word-boundary assertion*/
7425 case 'C': /* Single char !DANGEROUS! */
7426 case 'd': case 'D': /* digit class */
7427 case 'g': case 'G': /* generic-backref, pos assertion */
7428 case 'h': case 'H': /* HORIZWS */
7429 case 'k': case 'K': /* named backref, keep marker */
7430 case 'N': /* named char sequence */
7431 case 'p': case 'P': /* Unicode property */
7432 case 'R': /* LNBREAK */
7433 case 's': case 'S': /* space class */
7434 case 'v': case 'V': /* VERTWS */
7435 case 'w': case 'W': /* word class */
7436 case 'X': /* eXtended Unicode "combining character sequence" */
7437 case 'z': case 'Z': /* End of line/string assertion */
7441 /* Anything after here is an escape that resolves to a
7442 literal. (Except digits, which may or may not)
7461 ender = ASCII_TO_NATIVE('\033');
7465 ender = ASCII_TO_NATIVE('\007');
7470 STRLEN brace_len = len;
7472 if ((error_msg = grok_bslash_o(p,
7485 if (PL_encoding && ender < 0x100) {
7486 goto recode_encoding;
7495 char* const e = strchr(p, '}');
7499 vFAIL("Missing right brace on \\x{}");
7502 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
7503 | PERL_SCAN_DISALLOW_PREFIX;
7504 STRLEN numlen = e - p - 1;
7505 ender = grok_hex(p + 1, &numlen, &flags, NULL);
7512 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
7514 ender = grok_hex(p, &numlen, &flags, NULL);
7517 if (PL_encoding && ender < 0x100)
7518 goto recode_encoding;
7522 ender = grok_bslash_c(*p++, SIZE_ONLY);
7524 case '0': case '1': case '2': case '3':case '4':
7525 case '5': case '6': case '7': case '8':case '9':
7527 (isDIGIT(p[1]) && atoi(p) >= RExC_npar) ) {
7530 ender = grok_oct(p, &numlen, &flags, NULL);
7540 if (PL_encoding && ender < 0x100)
7541 goto recode_encoding;
7545 SV* enc = PL_encoding;
7546 ender = reg_recode((const char)(U8)ender, &enc);
7547 if (!enc && SIZE_ONLY)
7548 ckWARNreg(p, "Invalid escape in the specified encoding");
7554 FAIL("Trailing \\");
7557 if (!SIZE_ONLY&& isALPHA(*p))
7558 ckWARN2reg(p + 1, "Unrecognized escape \\%c passed through", UCHARAT(p));
7559 goto normal_default;
7564 if (UTF8_IS_START(*p) && UTF) {
7566 ender = utf8n_to_uvchr((U8*)p, RExC_end - p,
7567 &numlen, UTF8_ALLOW_DEFAULT);
7574 if ( RExC_flags & RXf_PMf_EXTENDED)
7575 p = regwhite( pRExC_state, p );
7577 /* Prime the casefolded buffer. */
7578 ender = toFOLD_uni(ender, tmpbuf, &foldlen);
7580 if (p < RExC_end && ISMULT2(p)) { /* Back off on ?+*. */
7585 /* Emit all the Unicode characters. */
7587 for (foldbuf = tmpbuf;
7589 foldlen -= numlen) {
7590 ender = utf8_to_uvchr(foldbuf, &numlen);
7592 const STRLEN unilen = reguni(pRExC_state, ender, s);
7595 /* In EBCDIC the numlen
7596 * and unilen can differ. */
7598 if (numlen >= foldlen)
7602 break; /* "Can't happen." */
7606 const STRLEN unilen = reguni(pRExC_state, ender, s);
7615 REGC((char)ender, s++);
7621 /* Emit all the Unicode characters. */
7623 for (foldbuf = tmpbuf;
7625 foldlen -= numlen) {
7626 ender = utf8_to_uvchr(foldbuf, &numlen);
7628 const STRLEN unilen = reguni(pRExC_state, ender, s);
7631 /* In EBCDIC the numlen
7632 * and unilen can differ. */
7634 if (numlen >= foldlen)
7642 const STRLEN unilen = reguni(pRExC_state, ender, s);
7651 REGC((char)ender, s++);
7655 Set_Node_Cur_Length(ret); /* MJD */
7656 nextchar(pRExC_state);
7658 /* len is STRLEN which is unsigned, need to copy to signed */
7661 vFAIL("Internal disaster");
7665 if (len == 1 && UNI_IS_INVARIANT(ender))
7669 RExC_size += STR_SZ(len);
7672 RExC_emit += STR_SZ(len);
7682 S_regwhite( RExC_state_t *pRExC_state, char *p )
7684 const char *e = RExC_end;
7686 PERL_ARGS_ASSERT_REGWHITE;
7691 else if (*p == '#') {
7700 RExC_seen |= REG_SEEN_RUN_ON_COMMENT;
7708 /* Parse POSIX character classes: [[:foo:]], [[=foo=]], [[.foo.]].
7709 Character classes ([:foo:]) can also be negated ([:^foo:]).
7710 Returns a named class id (ANYOF_XXX) if successful, -1 otherwise.
7711 Equivalence classes ([=foo=]) and composites ([.foo.]) are parsed,
7712 but trigger failures because they are currently unimplemented. */
7714 #define POSIXCC_DONE(c) ((c) == ':')
7715 #define POSIXCC_NOTYET(c) ((c) == '=' || (c) == '.')
7716 #define POSIXCC(c) (POSIXCC_DONE(c) || POSIXCC_NOTYET(c))
7719 S_regpposixcc(pTHX_ RExC_state_t *pRExC_state, I32 value)
7722 I32 namedclass = OOB_NAMEDCLASS;
7724 PERL_ARGS_ASSERT_REGPPOSIXCC;
7726 if (value == '[' && RExC_parse + 1 < RExC_end &&
7727 /* I smell either [: or [= or [. -- POSIX has been here, right? */
7728 POSIXCC(UCHARAT(RExC_parse))) {
7729 const char c = UCHARAT(RExC_parse);
7730 char* const s = RExC_parse++;
7732 while (RExC_parse < RExC_end && UCHARAT(RExC_parse) != c)
7734 if (RExC_parse == RExC_end)
7735 /* Grandfather lone [:, [=, [. */
7738 const char* const t = RExC_parse++; /* skip over the c */
7741 if (UCHARAT(RExC_parse) == ']') {
7742 const char *posixcc = s + 1;
7743 RExC_parse++; /* skip over the ending ] */
7746 const I32 complement = *posixcc == '^' ? *posixcc++ : 0;
7747 const I32 skip = t - posixcc;
7749 /* Initially switch on the length of the name. */
7752 if (memEQ(posixcc, "word", 4)) /* this is not POSIX, this is the Perl \w */
7753 namedclass = complement ? ANYOF_NALNUM : ANYOF_ALNUM;
7756 /* Names all of length 5. */
7757 /* alnum alpha ascii blank cntrl digit graph lower
7758 print punct space upper */
7759 /* Offset 4 gives the best switch position. */
7760 switch (posixcc[4]) {
7762 if (memEQ(posixcc, "alph", 4)) /* alpha */
7763 namedclass = complement ? ANYOF_NALPHA : ANYOF_ALPHA;
7766 if (memEQ(posixcc, "spac", 4)) /* space */
7767 namedclass = complement ? ANYOF_NPSXSPC : ANYOF_PSXSPC;
7770 if (memEQ(posixcc, "grap", 4)) /* graph */
7771 namedclass = complement ? ANYOF_NGRAPH : ANYOF_GRAPH;
7774 if (memEQ(posixcc, "asci", 4)) /* ascii */
7775 namedclass = complement ? ANYOF_NASCII : ANYOF_ASCII;
7778 if (memEQ(posixcc, "blan", 4)) /* blank */
7779 namedclass = complement ? ANYOF_NBLANK : ANYOF_BLANK;
7782 if (memEQ(posixcc, "cntr", 4)) /* cntrl */
7783 namedclass = complement ? ANYOF_NCNTRL : ANYOF_CNTRL;
7786 if (memEQ(posixcc, "alnu", 4)) /* alnum */
7787 namedclass = complement ? ANYOF_NALNUMC : ANYOF_ALNUMC;
7790 if (memEQ(posixcc, "lowe", 4)) /* lower */
7791 namedclass = complement ? ANYOF_NLOWER : ANYOF_LOWER;
7792 else if (memEQ(posixcc, "uppe", 4)) /* upper */
7793 namedclass = complement ? ANYOF_NUPPER : ANYOF_UPPER;
7796 if (memEQ(posixcc, "digi", 4)) /* digit */
7797 namedclass = complement ? ANYOF_NDIGIT : ANYOF_DIGIT;
7798 else if (memEQ(posixcc, "prin", 4)) /* print */
7799 namedclass = complement ? ANYOF_NPRINT : ANYOF_PRINT;
7800 else if (memEQ(posixcc, "punc", 4)) /* punct */
7801 namedclass = complement ? ANYOF_NPUNCT : ANYOF_PUNCT;
7806 if (memEQ(posixcc, "xdigit", 6))
7807 namedclass = complement ? ANYOF_NXDIGIT : ANYOF_XDIGIT;
7811 if (namedclass == OOB_NAMEDCLASS)
7812 Simple_vFAIL3("POSIX class [:%.*s:] unknown",
7814 assert (posixcc[skip] == ':');
7815 assert (posixcc[skip+1] == ']');
7816 } else if (!SIZE_ONLY) {
7817 /* [[=foo=]] and [[.foo.]] are still future. */
7819 /* adjust RExC_parse so the warning shows after
7821 while (UCHARAT(RExC_parse) && UCHARAT(RExC_parse) != ']')
7823 Simple_vFAIL3("POSIX syntax [%c %c] is reserved for future extensions", c, c);
7826 /* Maternal grandfather:
7827 * "[:" ending in ":" but not in ":]" */
7837 S_checkposixcc(pTHX_ RExC_state_t *pRExC_state)
7841 PERL_ARGS_ASSERT_CHECKPOSIXCC;
7843 if (POSIXCC(UCHARAT(RExC_parse))) {
7844 const char *s = RExC_parse;
7845 const char c = *s++;
7849 if (*s && c == *s && s[1] == ']') {
7851 "POSIX syntax [%c %c] belongs inside character classes",
7854 /* [[=foo=]] and [[.foo.]] are still future. */
7855 if (POSIXCC_NOTYET(c)) {
7856 /* adjust RExC_parse so the error shows after
7858 while (UCHARAT(RExC_parse) && UCHARAT(RExC_parse++) != ']')
7860 Simple_vFAIL3("POSIX syntax [%c %c] is reserved for future extensions", c, c);
7867 #define _C_C_T_(NAME,TEST,WORD) \
7870 ANYOF_CLASS_SET(ret, ANYOF_##NAME); \
7872 for (value = 0; value < 256; value++) \
7874 ANYOF_BITMAP_SET(ret, value); \
7879 case ANYOF_N##NAME: \
7881 ANYOF_CLASS_SET(ret, ANYOF_N##NAME); \
7883 for (value = 0; value < 256; value++) \
7885 ANYOF_BITMAP_SET(ret, value); \
7891 #define _C_C_T_NOLOC_(NAME,TEST,WORD) \
7893 for (value = 0; value < 256; value++) \
7895 ANYOF_BITMAP_SET(ret, value); \
7899 case ANYOF_N##NAME: \
7900 for (value = 0; value < 256; value++) \
7902 ANYOF_BITMAP_SET(ret, value); \
7908 We dont use PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS as the direct test
7909 so that it is possible to override the option here without having to
7910 rebuild the entire core. as we are required to do if we change regcomp.h
7911 which is where PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS is defined.
7913 #if PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS
7914 #define BROKEN_UNICODE_CHARCLASS_MAPPINGS
7917 #ifdef BROKEN_UNICODE_CHARCLASS_MAPPINGS
7918 #define POSIX_CC_UNI_NAME(CCNAME) CCNAME
7920 #define POSIX_CC_UNI_NAME(CCNAME) "Posix" CCNAME
7924 parse a class specification and produce either an ANYOF node that
7925 matches the pattern or if the pattern matches a single char only and
7926 that char is < 256 and we are case insensitive then we produce an
7931 S_regclass(pTHX_ RExC_state_t *pRExC_state, U32 depth)
7934 register UV nextvalue;
7935 register IV prevvalue = OOB_UNICODE;
7936 register IV range = 0;
7937 UV value = 0; /* XXX:dmq: needs to be referenceable (unfortunately) */
7938 register regnode *ret;
7941 char *rangebegin = NULL;
7942 bool need_class = 0;
7945 bool optimize_invert = TRUE;
7946 AV* unicode_alternate = NULL;
7948 UV literal_endpoint = 0;
7950 UV stored = 0; /* number of chars stored in the class */
7952 regnode * const orig_emit = RExC_emit; /* Save the original RExC_emit in
7953 case we need to change the emitted regop to an EXACT. */
7954 const char * orig_parse = RExC_parse;
7955 GET_RE_DEBUG_FLAGS_DECL;
7957 PERL_ARGS_ASSERT_REGCLASS;
7959 PERL_UNUSED_ARG(depth);
7962 DEBUG_PARSE("clas");
7964 /* Assume we are going to generate an ANYOF node. */
7965 ret = reganode(pRExC_state, ANYOF, 0);
7968 ANYOF_FLAGS(ret) = 0;
7970 if (UCHARAT(RExC_parse) == '^') { /* Complement of range. */
7974 ANYOF_FLAGS(ret) |= ANYOF_INVERT;
7978 RExC_size += ANYOF_SKIP;
7979 listsv = &PL_sv_undef; /* For code scanners: listsv always non-NULL. */
7982 RExC_emit += ANYOF_SKIP;
7984 ANYOF_FLAGS(ret) |= ANYOF_FOLD;
7986 ANYOF_FLAGS(ret) |= ANYOF_LOCALE;
7987 ANYOF_BITMAP_ZERO(ret);
7988 listsv = newSVpvs("# comment\n");
7991 nextvalue = RExC_parse < RExC_end ? UCHARAT(RExC_parse) : 0;
7993 if (!SIZE_ONLY && POSIXCC(nextvalue))
7994 checkposixcc(pRExC_state);
7996 /* allow 1st char to be ] (allowing it to be - is dealt with later) */
7997 if (UCHARAT(RExC_parse) == ']')
8001 while (RExC_parse < RExC_end && UCHARAT(RExC_parse) != ']') {
8006 namedclass = OOB_NAMEDCLASS; /* initialize as illegal */
8009 rangebegin = RExC_parse;
8011 value = utf8n_to_uvchr((U8*)RExC_parse,
8012 RExC_end - RExC_parse,
8013 &numlen, UTF8_ALLOW_DEFAULT);
8014 RExC_parse += numlen;
8017 value = UCHARAT(RExC_parse++);
8019 nextvalue = RExC_parse < RExC_end ? UCHARAT(RExC_parse) : 0;
8020 if (value == '[' && POSIXCC(nextvalue))
8021 namedclass = regpposixcc(pRExC_state, value);
8022 else if (value == '\\') {
8024 value = utf8n_to_uvchr((U8*)RExC_parse,
8025 RExC_end - RExC_parse,
8026 &numlen, UTF8_ALLOW_DEFAULT);
8027 RExC_parse += numlen;
8030 value = UCHARAT(RExC_parse++);
8031 /* Some compilers cannot handle switching on 64-bit integer
8032 * values, therefore value cannot be an UV. Yes, this will
8033 * be a problem later if we want switch on Unicode.
8034 * A similar issue a little bit later when switching on
8035 * namedclass. --jhi */
8036 switch ((I32)value) {
8037 case 'w': namedclass = ANYOF_ALNUM; break;
8038 case 'W': namedclass = ANYOF_NALNUM; break;
8039 case 's': namedclass = ANYOF_SPACE; break;
8040 case 'S': namedclass = ANYOF_NSPACE; break;
8041 case 'd': namedclass = ANYOF_DIGIT; break;
8042 case 'D': namedclass = ANYOF_NDIGIT; break;
8043 case 'v': namedclass = ANYOF_VERTWS; break;
8044 case 'V': namedclass = ANYOF_NVERTWS; break;
8045 case 'h': namedclass = ANYOF_HORIZWS; break;
8046 case 'H': namedclass = ANYOF_NHORIZWS; break;
8047 case 'N': /* Handle \N{NAME} in class */
8049 /* We only pay attention to the first char of
8050 multichar strings being returned. I kinda wonder
8051 if this makes sense as it does change the behaviour
8052 from earlier versions, OTOH that behaviour was broken
8054 UV v; /* value is register so we cant & it /grrr */
8055 if (reg_namedseq(pRExC_state, &v, NULL)) {
8065 if (RExC_parse >= RExC_end)
8066 vFAIL2("Empty \\%c{}", (U8)value);
8067 if (*RExC_parse == '{') {
8068 const U8 c = (U8)value;
8069 e = strchr(RExC_parse++, '}');
8071 vFAIL2("Missing right brace on \\%c{}", c);
8072 while (isSPACE(UCHARAT(RExC_parse)))
8074 if (e == RExC_parse)
8075 vFAIL2("Empty \\%c{}", c);
8077 while (isSPACE(UCHARAT(RExC_parse + n - 1)))
8085 if (UCHARAT(RExC_parse) == '^') {
8088 value = value == 'p' ? 'P' : 'p'; /* toggle */
8089 while (isSPACE(UCHARAT(RExC_parse))) {
8094 Perl_sv_catpvf(aTHX_ listsv, "%cutf8::%.*s\n",
8095 (value=='p' ? '+' : '!'), (int)n, RExC_parse);
8098 ANYOF_FLAGS(ret) |= ANYOF_UNICODE;
8099 namedclass = ANYOF_MAX; /* no official name, but it's named */
8102 case 'n': value = '\n'; break;
8103 case 'r': value = '\r'; break;
8104 case 't': value = '\t'; break;
8105 case 'f': value = '\f'; break;
8106 case 'b': value = '\b'; break;
8107 case 'e': value = ASCII_TO_NATIVE('\033');break;
8108 case 'a': value = ASCII_TO_NATIVE('\007');break;
8110 RExC_parse--; /* function expects to be pointed at the 'o' */
8111 if ((error_msg = grok_bslash_o(RExC_parse,
8119 RExC_parse += numlen;
8120 if (PL_encoding && value < 0x100) {
8121 goto recode_encoding;
8125 if (*RExC_parse == '{') {
8126 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
8127 | PERL_SCAN_DISALLOW_PREFIX;
8128 char * const e = strchr(RExC_parse++, '}');
8130 vFAIL("Missing right brace on \\x{}");
8132 numlen = e - RExC_parse;
8133 value = grok_hex(RExC_parse, &numlen, &flags, NULL);
8137 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
8139 value = grok_hex(RExC_parse, &numlen, &flags, NULL);
8140 RExC_parse += numlen;
8142 if (PL_encoding && value < 0x100)
8143 goto recode_encoding;
8146 value = grok_bslash_c(*RExC_parse++, SIZE_ONLY);
8148 case '0': case '1': case '2': case '3': case '4':
8149 case '5': case '6': case '7': case '8': case '9':
8153 value = grok_oct(--RExC_parse, &numlen, &flags, NULL);
8154 RExC_parse += numlen;
8155 if (PL_encoding && value < 0x100)
8156 goto recode_encoding;
8161 SV* enc = PL_encoding;
8162 value = reg_recode((const char)(U8)value, &enc);
8163 if (!enc && SIZE_ONLY)
8164 ckWARNreg(RExC_parse,
8165 "Invalid escape in the specified encoding");
8169 if (!SIZE_ONLY && isALPHA(value))
8170 ckWARN2reg(RExC_parse,
8171 "Unrecognized escape \\%c in character class passed through",
8175 } /* end of \blah */
8181 if (namedclass > OOB_NAMEDCLASS) { /* this is a named class \blah */
8183 if (!SIZE_ONLY && !need_class)
8184 ANYOF_CLASS_ZERO(ret);
8188 /* a bad range like a-\d, a-[:digit:] ? */
8192 RExC_parse >= rangebegin ?
8193 RExC_parse - rangebegin : 0;
8194 ckWARN4reg(RExC_parse,
8195 "False [] range \"%*.*s\"",
8198 if (prevvalue < 256) {
8199 ANYOF_BITMAP_SET(ret, prevvalue);
8200 ANYOF_BITMAP_SET(ret, '-');
8203 ANYOF_FLAGS(ret) |= ANYOF_UNICODE;
8204 Perl_sv_catpvf(aTHX_ listsv,
8205 "%04"UVxf"\n%04"UVxf"\n", (UV)prevvalue, (UV) '-');
8209 range = 0; /* this was not a true range */
8215 const char *what = NULL;
8218 if (namedclass > OOB_NAMEDCLASS)
8219 optimize_invert = FALSE;
8220 /* Possible truncation here but in some 64-bit environments
8221 * the compiler gets heartburn about switch on 64-bit values.
8222 * A similar issue a little earlier when switching on value.
8224 switch ((I32)namedclass) {
8226 case _C_C_T_(ALNUMC, isALNUMC(value), POSIX_CC_UNI_NAME("Alnum"));
8227 case _C_C_T_(ALPHA, isALPHA(value), POSIX_CC_UNI_NAME("Alpha"));
8228 case _C_C_T_(BLANK, isBLANK(value), POSIX_CC_UNI_NAME("Blank"));
8229 case _C_C_T_(CNTRL, isCNTRL(value), POSIX_CC_UNI_NAME("Cntrl"));
8230 case _C_C_T_(GRAPH, isGRAPH(value), POSIX_CC_UNI_NAME("Graph"));
8231 case _C_C_T_(LOWER, isLOWER(value), POSIX_CC_UNI_NAME("Lower"));
8232 case _C_C_T_(PRINT, isPRINT(value), POSIX_CC_UNI_NAME("Print"));
8233 case _C_C_T_(PSXSPC, isPSXSPC(value), POSIX_CC_UNI_NAME("Space"));
8234 case _C_C_T_(PUNCT, isPUNCT(value), POSIX_CC_UNI_NAME("Punct"));
8235 case _C_C_T_(UPPER, isUPPER(value), POSIX_CC_UNI_NAME("Upper"));
8236 #ifdef BROKEN_UNICODE_CHARCLASS_MAPPINGS
8237 case _C_C_T_(ALNUM, isALNUM(value), "Word");
8238 case _C_C_T_(SPACE, isSPACE(value), "SpacePerl");
8240 case _C_C_T_(SPACE, isSPACE(value), "PerlSpace");
8241 case _C_C_T_(ALNUM, isALNUM(value), "PerlWord");
8243 case _C_C_T_(XDIGIT, isXDIGIT(value), "XDigit");
8244 case _C_C_T_NOLOC_(VERTWS, is_VERTWS_latin1(&value), "VertSpace");
8245 case _C_C_T_NOLOC_(HORIZWS, is_HORIZWS_latin1(&value), "HorizSpace");
8248 ANYOF_CLASS_SET(ret, ANYOF_ASCII);
8251 for (value = 0; value < 128; value++)
8252 ANYOF_BITMAP_SET(ret, value);
8254 for (value = 0; value < 256; value++) {
8256 ANYOF_BITMAP_SET(ret, value);
8265 ANYOF_CLASS_SET(ret, ANYOF_NASCII);
8268 for (value = 128; value < 256; value++)
8269 ANYOF_BITMAP_SET(ret, value);
8271 for (value = 0; value < 256; value++) {
8272 if (!isASCII(value))
8273 ANYOF_BITMAP_SET(ret, value);
8282 ANYOF_CLASS_SET(ret, ANYOF_DIGIT);
8284 /* consecutive digits assumed */
8285 for (value = '0'; value <= '9'; value++)
8286 ANYOF_BITMAP_SET(ret, value);
8289 what = POSIX_CC_UNI_NAME("Digit");
8293 ANYOF_CLASS_SET(ret, ANYOF_NDIGIT);
8295 /* consecutive digits assumed */
8296 for (value = 0; value < '0'; value++)
8297 ANYOF_BITMAP_SET(ret, value);
8298 for (value = '9' + 1; value < 256; value++)
8299 ANYOF_BITMAP_SET(ret, value);
8302 what = POSIX_CC_UNI_NAME("Digit");
8305 /* this is to handle \p and \P */
8308 vFAIL("Invalid [::] class");
8312 /* Strings such as "+utf8::isWord\n" */
8313 Perl_sv_catpvf(aTHX_ listsv, "%cutf8::Is%s\n", yesno, what);
8316 ANYOF_FLAGS(ret) |= ANYOF_CLASS;
8319 } /* end of namedclass \blah */
8322 if (prevvalue > (IV)value) /* b-a */ {
8323 const int w = RExC_parse - rangebegin;
8324 Simple_vFAIL4("Invalid [] range \"%*.*s\"", w, w, rangebegin);
8325 range = 0; /* not a valid range */
8329 prevvalue = value; /* save the beginning of the range */
8330 if (*RExC_parse == '-' && RExC_parse+1 < RExC_end &&
8331 RExC_parse[1] != ']') {
8334 /* a bad range like \w-, [:word:]- ? */
8335 if (namedclass > OOB_NAMEDCLASS) {
8336 if (ckWARN(WARN_REGEXP)) {
8338 RExC_parse >= rangebegin ?
8339 RExC_parse - rangebegin : 0;
8341 "False [] range \"%*.*s\"",
8345 ANYOF_BITMAP_SET(ret, '-');
8347 range = 1; /* yeah, it's a range! */
8348 continue; /* but do it the next time */
8352 /* now is the next time */
8353 /*stored += (value - prevvalue + 1);*/
8355 if (prevvalue < 256) {
8356 const IV ceilvalue = value < 256 ? value : 255;
8359 /* In EBCDIC [\x89-\x91] should include
8360 * the \x8e but [i-j] should not. */
8361 if (literal_endpoint == 2 &&
8362 ((isLOWER(prevvalue) && isLOWER(ceilvalue)) ||
8363 (isUPPER(prevvalue) && isUPPER(ceilvalue))))
8365 if (isLOWER(prevvalue)) {
8366 for (i = prevvalue; i <= ceilvalue; i++)
8367 if (isLOWER(i) && !ANYOF_BITMAP_TEST(ret,i)) {
8369 ANYOF_BITMAP_SET(ret, i);
8372 for (i = prevvalue; i <= ceilvalue; i++)
8373 if (isUPPER(i) && !ANYOF_BITMAP_TEST(ret,i)) {
8375 ANYOF_BITMAP_SET(ret, i);
8381 for (i = prevvalue; i <= ceilvalue; i++) {
8382 if (!ANYOF_BITMAP_TEST(ret,i)) {
8384 ANYOF_BITMAP_SET(ret, i);
8388 if (value > 255 || UTF) {
8389 const UV prevnatvalue = NATIVE_TO_UNI(prevvalue);
8390 const UV natvalue = NATIVE_TO_UNI(value);
8391 stored+=2; /* can't optimize this class */
8392 ANYOF_FLAGS(ret) |= ANYOF_UNICODE;
8393 if (prevnatvalue < natvalue) { /* what about > ? */
8394 Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\t%04"UVxf"\n",
8395 prevnatvalue, natvalue);
8397 else if (prevnatvalue == natvalue) {
8398 Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\n", natvalue);
8400 U8 foldbuf[UTF8_MAXBYTES_CASE+1];
8402 const UV f = to_uni_fold(natvalue, foldbuf, &foldlen);
8404 #ifdef EBCDIC /* RD t/uni/fold ff and 6b */
8405 if (RExC_precomp[0] == ':' &&
8406 RExC_precomp[1] == '[' &&
8407 (f == 0xDF || f == 0x92)) {
8408 f = NATIVE_TO_UNI(f);
8411 /* If folding and foldable and a single
8412 * character, insert also the folded version
8413 * to the charclass. */
8415 #ifdef EBCDIC /* RD tunifold ligatures s,t fb05, fb06 */
8416 if ((RExC_precomp[0] == ':' &&
8417 RExC_precomp[1] == '[' &&
8419 (value == 0xFB05 || value == 0xFB06))) ?
8420 foldlen == ((STRLEN)UNISKIP(f) - 1) :
8421 foldlen == (STRLEN)UNISKIP(f) )
8423 if (foldlen == (STRLEN)UNISKIP(f))
8425 Perl_sv_catpvf(aTHX_ listsv,
8428 /* Any multicharacter foldings
8429 * require the following transform:
8430 * [ABCDEF] -> (?:[ABCabcDEFd]|pq|rst)
8431 * where E folds into "pq" and F folds
8432 * into "rst", all other characters
8433 * fold to single characters. We save
8434 * away these multicharacter foldings,
8435 * to be later saved as part of the
8436 * additional "s" data. */
8439 if (!unicode_alternate)
8440 unicode_alternate = newAV();
8441 sv = newSVpvn_utf8((char*)foldbuf, foldlen,
8443 av_push(unicode_alternate, sv);
8447 /* If folding and the value is one of the Greek
8448 * sigmas insert a few more sigmas to make the
8449 * folding rules of the sigmas to work right.
8450 * Note that not all the possible combinations
8451 * are handled here: some of them are handled
8452 * by the standard folding rules, and some of
8453 * them (literal or EXACTF cases) are handled
8454 * during runtime in regexec.c:S_find_byclass(). */
8455 if (value == UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA) {
8456 Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\n",
8457 (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA);
8458 Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\n",
8459 (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA);
8461 else if (value == UNICODE_GREEK_CAPITAL_LETTER_SIGMA)
8462 Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\n",
8463 (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA);
8468 literal_endpoint = 0;
8472 range = 0; /* this range (if it was one) is done now */
8476 ANYOF_FLAGS(ret) |= ANYOF_LARGE;
8478 RExC_size += ANYOF_CLASS_ADD_SKIP;
8480 RExC_emit += ANYOF_CLASS_ADD_SKIP;
8486 /****** !SIZE_ONLY AFTER HERE *********/
8488 if( stored == 1 && (value < 128 || (value < 256 && !UTF))
8489 && !( ANYOF_FLAGS(ret) & ( ANYOF_FLAGS_ALL ^ ANYOF_FOLD ) )
8491 /* optimize single char class to an EXACT node
8492 but *only* when its not a UTF/high char */
8493 const char * cur_parse= RExC_parse;
8494 RExC_emit = (regnode *)orig_emit;
8495 RExC_parse = (char *)orig_parse;
8496 ret = reg_node(pRExC_state,
8497 (U8)((ANYOF_FLAGS(ret) & ANYOF_FOLD) ? EXACTF : EXACT));
8498 RExC_parse = (char *)cur_parse;
8499 *STRING(ret)= (char)value;
8501 RExC_emit += STR_SZ(1);
8502 SvREFCNT_dec(listsv);
8505 /* optimize case-insensitive simple patterns (e.g. /[a-z]/i) */
8506 if ( /* If the only flag is folding (plus possibly inversion). */
8507 ((ANYOF_FLAGS(ret) & (ANYOF_FLAGS_ALL ^ ANYOF_INVERT)) == ANYOF_FOLD)
8509 for (value = 0; value < 256; ++value) {
8510 if (ANYOF_BITMAP_TEST(ret, value)) {
8511 UV fold = PL_fold[value];
8514 ANYOF_BITMAP_SET(ret, fold);
8517 ANYOF_FLAGS(ret) &= ~ANYOF_FOLD;
8520 /* optimize inverted simple patterns (e.g. [^a-z]) */
8521 if (optimize_invert &&
8522 /* If the only flag is inversion. */
8523 (ANYOF_FLAGS(ret) & ANYOF_FLAGS_ALL) == ANYOF_INVERT) {
8524 for (value = 0; value < ANYOF_BITMAP_SIZE; ++value)
8525 ANYOF_BITMAP(ret)[value] ^= ANYOF_FLAGS_ALL;
8526 ANYOF_FLAGS(ret) = ANYOF_UNICODE_ALL;
8529 AV * const av = newAV();
8531 /* The 0th element stores the character class description
8532 * in its textual form: used later (regexec.c:Perl_regclass_swash())
8533 * to initialize the appropriate swash (which gets stored in
8534 * the 1st element), and also useful for dumping the regnode.
8535 * The 2nd element stores the multicharacter foldings,
8536 * used later (regexec.c:S_reginclass()). */
8537 av_store(av, 0, listsv);
8538 av_store(av, 1, NULL);
8539 av_store(av, 2, MUTABLE_SV(unicode_alternate));
8540 rv = newRV_noinc(MUTABLE_SV(av));
8541 n = add_data(pRExC_state, 1, "s");
8542 RExC_rxi->data->data[n] = (void*)rv;
8550 /* reg_skipcomment()
8552 Absorbs an /x style # comments from the input stream.
8553 Returns true if there is more text remaining in the stream.
8554 Will set the REG_SEEN_RUN_ON_COMMENT flag if the comment
8555 terminates the pattern without including a newline.
8557 Note its the callers responsibility to ensure that we are
8563 S_reg_skipcomment(pTHX_ RExC_state_t *pRExC_state)
8567 PERL_ARGS_ASSERT_REG_SKIPCOMMENT;
8569 while (RExC_parse < RExC_end)
8570 if (*RExC_parse++ == '\n') {
8575 /* we ran off the end of the pattern without ending
8576 the comment, so we have to add an \n when wrapping */
8577 RExC_seen |= REG_SEEN_RUN_ON_COMMENT;
8585 Advance that parse position, and optionall absorbs
8586 "whitespace" from the inputstream.
8588 Without /x "whitespace" means (?#...) style comments only,
8589 with /x this means (?#...) and # comments and whitespace proper.
8591 Returns the RExC_parse point from BEFORE the scan occurs.
8593 This is the /x friendly way of saying RExC_parse++.
8597 S_nextchar(pTHX_ RExC_state_t *pRExC_state)
8599 char* const retval = RExC_parse++;
8601 PERL_ARGS_ASSERT_NEXTCHAR;
8604 if (*RExC_parse == '(' && RExC_parse[1] == '?' &&
8605 RExC_parse[2] == '#') {
8606 while (*RExC_parse != ')') {
8607 if (RExC_parse == RExC_end)
8608 FAIL("Sequence (?#... not terminated");
8614 if (RExC_flags & RXf_PMf_EXTENDED) {
8615 if (isSPACE(*RExC_parse)) {
8619 else if (*RExC_parse == '#') {
8620 if ( reg_skipcomment( pRExC_state ) )
8629 - reg_node - emit a node
8631 STATIC regnode * /* Location. */
8632 S_reg_node(pTHX_ RExC_state_t *pRExC_state, U8 op)
8635 register regnode *ptr;
8636 regnode * const ret = RExC_emit;
8637 GET_RE_DEBUG_FLAGS_DECL;
8639 PERL_ARGS_ASSERT_REG_NODE;
8642 SIZE_ALIGN(RExC_size);
8646 if (RExC_emit >= RExC_emit_bound)
8647 Perl_croak(aTHX_ "panic: reg_node overrun trying to emit %d", op);
8649 NODE_ALIGN_FILL(ret);
8651 FILL_ADVANCE_NODE(ptr, op);
8652 REH_CALL_COMP_NODE_HOOK(pRExC_state->rx, (ptr) - 1);
8653 #ifdef RE_TRACK_PATTERN_OFFSETS
8654 if (RExC_offsets) { /* MJD */
8655 MJD_OFFSET_DEBUG(("%s:%d: (op %s) %s %"UVuf" (len %"UVuf") (max %"UVuf").\n",
8656 "reg_node", __LINE__,
8658 (UV)(RExC_emit - RExC_emit_start) > RExC_offsets[0]
8659 ? "Overwriting end of array!\n" : "OK",
8660 (UV)(RExC_emit - RExC_emit_start),
8661 (UV)(RExC_parse - RExC_start),
8662 (UV)RExC_offsets[0]));
8663 Set_Node_Offset(RExC_emit, RExC_parse + (op == END));
8671 - reganode - emit a node with an argument
8673 STATIC regnode * /* Location. */
8674 S_reganode(pTHX_ RExC_state_t *pRExC_state, U8 op, U32 arg)
8677 register regnode *ptr;
8678 regnode * const ret = RExC_emit;
8679 GET_RE_DEBUG_FLAGS_DECL;
8681 PERL_ARGS_ASSERT_REGANODE;
8684 SIZE_ALIGN(RExC_size);
8689 assert(2==regarglen[op]+1);
8691 Anything larger than this has to allocate the extra amount.
8692 If we changed this to be:
8694 RExC_size += (1 + regarglen[op]);
8696 then it wouldn't matter. Its not clear what side effect
8697 might come from that so its not done so far.
8702 if (RExC_emit >= RExC_emit_bound)
8703 Perl_croak(aTHX_ "panic: reg_node overrun trying to emit %d", op);
8705 NODE_ALIGN_FILL(ret);
8707 FILL_ADVANCE_NODE_ARG(ptr, op, arg);
8708 REH_CALL_COMP_NODE_HOOK(pRExC_state->rx, (ptr) - 2);
8709 #ifdef RE_TRACK_PATTERN_OFFSETS
8710 if (RExC_offsets) { /* MJD */
8711 MJD_OFFSET_DEBUG(("%s(%d): (op %s) %s %"UVuf" <- %"UVuf" (max %"UVuf").\n",
8715 (UV)(RExC_emit - RExC_emit_start) > RExC_offsets[0] ?
8716 "Overwriting end of array!\n" : "OK",
8717 (UV)(RExC_emit - RExC_emit_start),
8718 (UV)(RExC_parse - RExC_start),
8719 (UV)RExC_offsets[0]));
8720 Set_Cur_Node_Offset;
8728 - reguni - emit (if appropriate) a Unicode character
8731 S_reguni(pTHX_ const RExC_state_t *pRExC_state, UV uv, char* s)
8735 PERL_ARGS_ASSERT_REGUNI;
8737 return SIZE_ONLY ? UNISKIP(uv) : (uvchr_to_utf8((U8*)s, uv) - (U8*)s);
8741 - reginsert - insert an operator in front of already-emitted operand
8743 * Means relocating the operand.
8746 S_reginsert(pTHX_ RExC_state_t *pRExC_state, U8 op, regnode *opnd, U32 depth)
8749 register regnode *src;
8750 register regnode *dst;
8751 register regnode *place;
8752 const int offset = regarglen[(U8)op];
8753 const int size = NODE_STEP_REGNODE + offset;
8754 GET_RE_DEBUG_FLAGS_DECL;
8756 PERL_ARGS_ASSERT_REGINSERT;
8757 PERL_UNUSED_ARG(depth);
8758 /* (PL_regkind[(U8)op] == CURLY ? EXTRA_STEP_2ARGS : 0); */
8759 DEBUG_PARSE_FMT("inst"," - %s",PL_reg_name[op]);
8768 if (RExC_open_parens) {
8770 /*DEBUG_PARSE_FMT("inst"," - %"IVdf, (IV)RExC_npar);*/
8771 for ( paren=0 ; paren < RExC_npar ; paren++ ) {
8772 if ( RExC_open_parens[paren] >= opnd ) {
8773 /*DEBUG_PARSE_FMT("open"," - %d",size);*/
8774 RExC_open_parens[paren] += size;
8776 /*DEBUG_PARSE_FMT("open"," - %s","ok");*/
8778 if ( RExC_close_parens[paren] >= opnd ) {
8779 /*DEBUG_PARSE_FMT("close"," - %d",size);*/
8780 RExC_close_parens[paren] += size;
8782 /*DEBUG_PARSE_FMT("close"," - %s","ok");*/
8787 while (src > opnd) {
8788 StructCopy(--src, --dst, regnode);
8789 #ifdef RE_TRACK_PATTERN_OFFSETS
8790 if (RExC_offsets) { /* MJD 20010112 */
8791 MJD_OFFSET_DEBUG(("%s(%d): (op %s) %s copy %"UVuf" -> %"UVuf" (max %"UVuf").\n",
8795 (UV)(dst - RExC_emit_start) > RExC_offsets[0]
8796 ? "Overwriting end of array!\n" : "OK",
8797 (UV)(src - RExC_emit_start),
8798 (UV)(dst - RExC_emit_start),
8799 (UV)RExC_offsets[0]));
8800 Set_Node_Offset_To_R(dst-RExC_emit_start, Node_Offset(src));
8801 Set_Node_Length_To_R(dst-RExC_emit_start, Node_Length(src));
8807 place = opnd; /* Op node, where operand used to be. */
8808 #ifdef RE_TRACK_PATTERN_OFFSETS
8809 if (RExC_offsets) { /* MJD */
8810 MJD_OFFSET_DEBUG(("%s(%d): (op %s) %s %"UVuf" <- %"UVuf" (max %"UVuf").\n",
8814 (UV)(place - RExC_emit_start) > RExC_offsets[0]
8815 ? "Overwriting end of array!\n" : "OK",
8816 (UV)(place - RExC_emit_start),
8817 (UV)(RExC_parse - RExC_start),
8818 (UV)RExC_offsets[0]));
8819 Set_Node_Offset(place, RExC_parse);
8820 Set_Node_Length(place, 1);
8823 src = NEXTOPER(place);
8824 FILL_ADVANCE_NODE(place, op);
8825 REH_CALL_COMP_NODE_HOOK(pRExC_state->rx, (place) - 1);
8826 Zero(src, offset, regnode);
8830 - regtail - set the next-pointer at the end of a node chain of p to val.
8831 - SEE ALSO: regtail_study
8833 /* TODO: All three parms should be const */
8835 S_regtail(pTHX_ RExC_state_t *pRExC_state, regnode *p, const regnode *val,U32 depth)
8838 register regnode *scan;
8839 GET_RE_DEBUG_FLAGS_DECL;
8841 PERL_ARGS_ASSERT_REGTAIL;
8843 PERL_UNUSED_ARG(depth);
8849 /* Find last node. */
8852 regnode * const temp = regnext(scan);
8854 SV * const mysv=sv_newmortal();
8855 DEBUG_PARSE_MSG((scan==p ? "tail" : ""));
8856 regprop(RExC_rx, mysv, scan);
8857 PerlIO_printf(Perl_debug_log, "~ %s (%d) %s %s\n",
8858 SvPV_nolen_const(mysv), REG_NODE_NUM(scan),
8859 (temp == NULL ? "->" : ""),
8860 (temp == NULL ? PL_reg_name[OP(val)] : "")
8868 if (reg_off_by_arg[OP(scan)]) {
8869 ARG_SET(scan, val - scan);
8872 NEXT_OFF(scan) = val - scan;
8878 - regtail_study - set the next-pointer at the end of a node chain of p to val.
8879 - Look for optimizable sequences at the same time.
8880 - currently only looks for EXACT chains.
8882 This is expermental code. The idea is to use this routine to perform
8883 in place optimizations on branches and groups as they are constructed,
8884 with the long term intention of removing optimization from study_chunk so
8885 that it is purely analytical.
8887 Currently only used when in DEBUG mode. The macro REGTAIL_STUDY() is used
8888 to control which is which.
8891 /* TODO: All four parms should be const */
8894 S_regtail_study(pTHX_ RExC_state_t *pRExC_state, regnode *p, const regnode *val,U32 depth)
8897 register regnode *scan;
8899 #ifdef EXPERIMENTAL_INPLACESCAN
8902 GET_RE_DEBUG_FLAGS_DECL;
8904 PERL_ARGS_ASSERT_REGTAIL_STUDY;
8910 /* Find last node. */
8914 regnode * const temp = regnext(scan);
8915 #ifdef EXPERIMENTAL_INPLACESCAN
8916 if (PL_regkind[OP(scan)] == EXACT)
8917 if (join_exact(pRExC_state,scan,&min,1,val,depth+1))
8925 if( exact == PSEUDO )
8927 else if ( exact != OP(scan) )
8936 SV * const mysv=sv_newmortal();
8937 DEBUG_PARSE_MSG((scan==p ? "tsdy" : ""));
8938 regprop(RExC_rx, mysv, scan);
8939 PerlIO_printf(Perl_debug_log, "~ %s (%d) -> %s\n",
8940 SvPV_nolen_const(mysv),
8942 PL_reg_name[exact]);
8949 SV * const mysv_val=sv_newmortal();
8950 DEBUG_PARSE_MSG("");
8951 regprop(RExC_rx, mysv_val, val);
8952 PerlIO_printf(Perl_debug_log, "~ attach to %s (%"IVdf") offset to %"IVdf"\n",
8953 SvPV_nolen_const(mysv_val),
8954 (IV)REG_NODE_NUM(val),
8958 if (reg_off_by_arg[OP(scan)]) {
8959 ARG_SET(scan, val - scan);
8962 NEXT_OFF(scan) = val - scan;
8970 - regcurly - a little FSA that accepts {\d+,?\d*}
8972 #ifndef PERL_IN_XSUB_RE
8974 Perl_regcurly(register const char *s)
8976 PERL_ARGS_ASSERT_REGCURLY;
8995 - regdump - dump a regexp onto Perl_debug_log in vaguely comprehensible form
8999 S_regdump_extflags(pTHX_ const char *lead, const U32 flags)
9004 for (bit=0; bit<32; bit++) {
9005 if (flags & (1<<bit)) {
9007 PerlIO_printf(Perl_debug_log, "%s",lead);
9008 PerlIO_printf(Perl_debug_log, "%s ",PL_reg_extflags_name[bit]);
9013 PerlIO_printf(Perl_debug_log, "\n");
9015 PerlIO_printf(Perl_debug_log, "%s[none-set]\n",lead);
9021 Perl_regdump(pTHX_ const regexp *r)
9025 SV * const sv = sv_newmortal();
9026 SV *dsv= sv_newmortal();
9028 GET_RE_DEBUG_FLAGS_DECL;
9030 PERL_ARGS_ASSERT_REGDUMP;
9032 (void)dumpuntil(r, ri->program, ri->program + 1, NULL, NULL, sv, 0, 0);
9034 /* Header fields of interest. */
9035 if (r->anchored_substr) {
9036 RE_PV_QUOTED_DECL(s, 0, dsv, SvPVX_const(r->anchored_substr),
9037 RE_SV_DUMPLEN(r->anchored_substr), 30);
9038 PerlIO_printf(Perl_debug_log,
9039 "anchored %s%s at %"IVdf" ",
9040 s, RE_SV_TAIL(r->anchored_substr),
9041 (IV)r->anchored_offset);
9042 } else if (r->anchored_utf8) {
9043 RE_PV_QUOTED_DECL(s, 1, dsv, SvPVX_const(r->anchored_utf8),
9044 RE_SV_DUMPLEN(r->anchored_utf8), 30);
9045 PerlIO_printf(Perl_debug_log,
9046 "anchored utf8 %s%s at %"IVdf" ",
9047 s, RE_SV_TAIL(r->anchored_utf8),
9048 (IV)r->anchored_offset);
9050 if (r->float_substr) {
9051 RE_PV_QUOTED_DECL(s, 0, dsv, SvPVX_const(r->float_substr),
9052 RE_SV_DUMPLEN(r->float_substr), 30);
9053 PerlIO_printf(Perl_debug_log,
9054 "floating %s%s at %"IVdf"..%"UVuf" ",
9055 s, RE_SV_TAIL(r->float_substr),
9056 (IV)r->float_min_offset, (UV)r->float_max_offset);
9057 } else if (r->float_utf8) {
9058 RE_PV_QUOTED_DECL(s, 1, dsv, SvPVX_const(r->float_utf8),
9059 RE_SV_DUMPLEN(r->float_utf8), 30);
9060 PerlIO_printf(Perl_debug_log,
9061 "floating utf8 %s%s at %"IVdf"..%"UVuf" ",
9062 s, RE_SV_TAIL(r->float_utf8),
9063 (IV)r->float_min_offset, (UV)r->float_max_offset);
9065 if (r->check_substr || r->check_utf8)
9066 PerlIO_printf(Perl_debug_log,
9068 (r->check_substr == r->float_substr
9069 && r->check_utf8 == r->float_utf8
9070 ? "(checking floating" : "(checking anchored"));
9071 if (r->extflags & RXf_NOSCAN)
9072 PerlIO_printf(Perl_debug_log, " noscan");
9073 if (r->extflags & RXf_CHECK_ALL)
9074 PerlIO_printf(Perl_debug_log, " isall");
9075 if (r->check_substr || r->check_utf8)
9076 PerlIO_printf(Perl_debug_log, ") ");
9078 if (ri->regstclass) {
9079 regprop(r, sv, ri->regstclass);
9080 PerlIO_printf(Perl_debug_log, "stclass %s ", SvPVX_const(sv));
9082 if (r->extflags & RXf_ANCH) {
9083 PerlIO_printf(Perl_debug_log, "anchored");
9084 if (r->extflags & RXf_ANCH_BOL)
9085 PerlIO_printf(Perl_debug_log, "(BOL)");
9086 if (r->extflags & RXf_ANCH_MBOL)
9087 PerlIO_printf(Perl_debug_log, "(MBOL)");
9088 if (r->extflags & RXf_ANCH_SBOL)
9089 PerlIO_printf(Perl_debug_log, "(SBOL)");
9090 if (r->extflags & RXf_ANCH_GPOS)
9091 PerlIO_printf(Perl_debug_log, "(GPOS)");
9092 PerlIO_putc(Perl_debug_log, ' ');
9094 if (r->extflags & RXf_GPOS_SEEN)
9095 PerlIO_printf(Perl_debug_log, "GPOS:%"UVuf" ", (UV)r->gofs);
9096 if (r->intflags & PREGf_SKIP)
9097 PerlIO_printf(Perl_debug_log, "plus ");
9098 if (r->intflags & PREGf_IMPLICIT)
9099 PerlIO_printf(Perl_debug_log, "implicit ");
9100 PerlIO_printf(Perl_debug_log, "minlen %"IVdf" ", (IV)r->minlen);
9101 if (r->extflags & RXf_EVAL_SEEN)
9102 PerlIO_printf(Perl_debug_log, "with eval ");
9103 PerlIO_printf(Perl_debug_log, "\n");
9104 DEBUG_FLAGS_r(regdump_extflags("r->extflags: ",r->extflags));
9106 PERL_ARGS_ASSERT_REGDUMP;
9107 PERL_UNUSED_CONTEXT;
9109 #endif /* DEBUGGING */
9113 - regprop - printable representation of opcode
9115 #define EMIT_ANYOF_TEST_SEPARATOR(do_sep,sv,flags) \
9118 Perl_sv_catpvf(aTHX_ sv,"%s][%s",PL_colors[1],PL_colors[0]); \
9119 if (flags & ANYOF_INVERT) \
9120 /*make sure the invert info is in each */ \
9121 sv_catpvs(sv, "^"); \
9127 Perl_regprop(pTHX_ const regexp *prog, SV *sv, const regnode *o)
9132 RXi_GET_DECL(prog,progi);
9133 GET_RE_DEBUG_FLAGS_DECL;
9135 PERL_ARGS_ASSERT_REGPROP;
9139 if (OP(o) > REGNODE_MAX) /* regnode.type is unsigned */
9140 /* It would be nice to FAIL() here, but this may be called from
9141 regexec.c, and it would be hard to supply pRExC_state. */
9142 Perl_croak(aTHX_ "Corrupted regexp opcode %d > %d", (int)OP(o), (int)REGNODE_MAX);
9143 sv_catpv(sv, PL_reg_name[OP(o)]); /* Take off const! */
9145 k = PL_regkind[OP(o)];
9149 /* Using is_utf8_string() (via PERL_PV_UNI_DETECT)
9150 * is a crude hack but it may be the best for now since
9151 * we have no flag "this EXACTish node was UTF-8"
9153 pv_pretty(sv, STRING(o), STR_LEN(o), 60, PL_colors[0], PL_colors[1],
9154 PERL_PV_ESCAPE_UNI_DETECT |
9155 PERL_PV_PRETTY_ELLIPSES |
9156 PERL_PV_PRETTY_LTGT |
9157 PERL_PV_PRETTY_NOCLEAR
9159 } else if (k == TRIE) {
9160 /* print the details of the trie in dumpuntil instead, as
9161 * progi->data isn't available here */
9162 const char op = OP(o);
9163 const U32 n = ARG(o);
9164 const reg_ac_data * const ac = IS_TRIE_AC(op) ?
9165 (reg_ac_data *)progi->data->data[n] :
9167 const reg_trie_data * const trie
9168 = (reg_trie_data*)progi->data->data[!IS_TRIE_AC(op) ? n : ac->trie];
9170 Perl_sv_catpvf(aTHX_ sv, "-%s",PL_reg_name[o->flags]);
9171 DEBUG_TRIE_COMPILE_r(
9172 Perl_sv_catpvf(aTHX_ sv,
9173 "<S:%"UVuf"/%"IVdf" W:%"UVuf" L:%"UVuf"/%"UVuf" C:%"UVuf"/%"UVuf">",
9174 (UV)trie->startstate,
9175 (IV)trie->statecount-1, /* -1 because of the unused 0 element */
9176 (UV)trie->wordcount,
9179 (UV)TRIE_CHARCOUNT(trie),
9180 (UV)trie->uniquecharcount
9183 if ( IS_ANYOF_TRIE(op) || trie->bitmap ) {
9185 int rangestart = -1;
9186 U8* bitmap = IS_ANYOF_TRIE(op) ? (U8*)ANYOF_BITMAP(o) : (U8*)TRIE_BITMAP(trie);
9188 for (i = 0; i <= 256; i++) {
9189 if (i < 256 && BITMAP_TEST(bitmap,i)) {
9190 if (rangestart == -1)
9192 } else if (rangestart != -1) {
9193 if (i <= rangestart + 3)
9194 for (; rangestart < i; rangestart++)
9195 put_byte(sv, rangestart);
9197 put_byte(sv, rangestart);
9199 put_byte(sv, i - 1);
9207 } else if (k == CURLY) {
9208 if (OP(o) == CURLYM || OP(o) == CURLYN || OP(o) == CURLYX)
9209 Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* Parenth number */
9210 Perl_sv_catpvf(aTHX_ sv, " {%d,%d}", ARG1(o), ARG2(o));
9212 else if (k == WHILEM && o->flags) /* Ordinal/of */
9213 Perl_sv_catpvf(aTHX_ sv, "[%d/%d]", o->flags & 0xf, o->flags>>4);
9214 else if (k == REF || k == OPEN || k == CLOSE || k == GROUPP || OP(o)==ACCEPT) {
9215 Perl_sv_catpvf(aTHX_ sv, "%d", (int)ARG(o)); /* Parenth number */
9216 if ( RXp_PAREN_NAMES(prog) ) {
9217 if ( k != REF || OP(o) < NREF) {
9218 AV *list= MUTABLE_AV(progi->data->data[progi->name_list_idx]);
9219 SV **name= av_fetch(list, ARG(o), 0 );
9221 Perl_sv_catpvf(aTHX_ sv, " '%"SVf"'", SVfARG(*name));
9224 AV *list= MUTABLE_AV(progi->data->data[ progi->name_list_idx ]);
9225 SV *sv_dat= MUTABLE_SV(progi->data->data[ ARG( o ) ]);
9226 I32 *nums=(I32*)SvPVX(sv_dat);
9227 SV **name= av_fetch(list, nums[0], 0 );
9230 for ( n=0; n<SvIVX(sv_dat); n++ ) {
9231 Perl_sv_catpvf(aTHX_ sv, "%s%"IVdf,
9232 (n ? "," : ""), (IV)nums[n]);
9234 Perl_sv_catpvf(aTHX_ sv, " '%"SVf"'", SVfARG(*name));
9238 } else if (k == GOSUB)
9239 Perl_sv_catpvf(aTHX_ sv, "%d[%+d]", (int)ARG(o),(int)ARG2L(o)); /* Paren and offset */
9240 else if (k == VERB) {
9242 Perl_sv_catpvf(aTHX_ sv, ":%"SVf,
9243 SVfARG((MUTABLE_SV(progi->data->data[ ARG( o ) ]))));
9244 } else if (k == LOGICAL)
9245 Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* 2: embedded, otherwise 1 */
9246 else if (k == FOLDCHAR)
9247 Perl_sv_catpvf(aTHX_ sv, "[0x%"UVXf"]", PTR2UV(ARG(o)) );
9248 else if (k == ANYOF) {
9249 int i, rangestart = -1;
9250 const U8 flags = ANYOF_FLAGS(o);
9253 /* Should be synchronized with * ANYOF_ #xdefines in regcomp.h */
9254 static const char * const anyofs[] = {
9287 if (flags & ANYOF_LOCALE)
9288 sv_catpvs(sv, "{loc}");
9289 if (flags & ANYOF_FOLD)
9290 sv_catpvs(sv, "{i}");
9291 Perl_sv_catpvf(aTHX_ sv, "[%s", PL_colors[0]);
9292 if (flags & ANYOF_INVERT)
9295 /* output what the standard cp 0-255 bitmap matches */
9296 for (i = 0; i <= 256; i++) {
9297 if (i < 256 && ANYOF_BITMAP_TEST(o,i)) {
9298 if (rangestart == -1)
9300 } else if (rangestart != -1) {
9301 if (i <= rangestart + 3)
9302 for (; rangestart < i; rangestart++)
9303 put_byte(sv, rangestart);
9305 put_byte(sv, rangestart);
9307 put_byte(sv, i - 1);
9314 EMIT_ANYOF_TEST_SEPARATOR(do_sep,sv,flags);
9315 /* output any special charclass tests (used mostly under use locale) */
9316 if (o->flags & ANYOF_CLASS)
9317 for (i = 0; i < (int)(sizeof(anyofs)/sizeof(char*)); i++)
9318 if (ANYOF_CLASS_TEST(o,i)) {
9319 sv_catpv(sv, anyofs[i]);
9323 EMIT_ANYOF_TEST_SEPARATOR(do_sep,sv,flags);
9325 /* output information about the unicode matching */
9326 if (flags & ANYOF_UNICODE)
9327 sv_catpvs(sv, "{unicode}");
9328 else if (flags & ANYOF_UNICODE_ALL)
9329 sv_catpvs(sv, "{unicode_all}");
9333 SV * const sw = regclass_swash(prog, o, FALSE, &lv, 0);
9337 U8 s[UTF8_MAXBYTES_CASE+1];
9339 for (i = 0; i <= 256; i++) { /* just the first 256 */
9340 uvchr_to_utf8(s, i);
9342 if (i < 256 && swash_fetch(sw, s, TRUE)) {
9343 if (rangestart == -1)
9345 } else if (rangestart != -1) {
9346 if (i <= rangestart + 3)
9347 for (; rangestart < i; rangestart++) {
9348 const U8 * const e = uvchr_to_utf8(s,rangestart);
9350 for(p = s; p < e; p++)
9354 const U8 *e = uvchr_to_utf8(s,rangestart);
9356 for (p = s; p < e; p++)
9359 e = uvchr_to_utf8(s, i-1);
9360 for (p = s; p < e; p++)
9367 sv_catpvs(sv, "..."); /* et cetera */
9371 char *s = savesvpv(lv);
9372 char * const origs = s;
9374 while (*s && *s != '\n')
9378 const char * const t = ++s;
9396 Perl_sv_catpvf(aTHX_ sv, "%s]", PL_colors[1]);
9398 else if (k == BRANCHJ && (OP(o) == UNLESSM || OP(o) == IFMATCH))
9399 Perl_sv_catpvf(aTHX_ sv, "[%d]", -(o->flags));
9401 PERL_UNUSED_CONTEXT;
9402 PERL_UNUSED_ARG(sv);
9404 PERL_UNUSED_ARG(prog);
9405 #endif /* DEBUGGING */
9409 Perl_re_intuit_string(pTHX_ REGEXP * const r)
9410 { /* Assume that RE_INTUIT is set */
9412 struct regexp *const prog = (struct regexp *)SvANY(r);
9413 GET_RE_DEBUG_FLAGS_DECL;
9415 PERL_ARGS_ASSERT_RE_INTUIT_STRING;
9416 PERL_UNUSED_CONTEXT;
9420 const char * const s = SvPV_nolen_const(prog->check_substr
9421 ? prog->check_substr : prog->check_utf8);
9423 if (!PL_colorset) reginitcolors();
9424 PerlIO_printf(Perl_debug_log,
9425 "%sUsing REx %ssubstr:%s \"%s%.60s%s%s\"\n",
9427 prog->check_substr ? "" : "utf8 ",
9428 PL_colors[5],PL_colors[0],
9431 (strlen(s) > 60 ? "..." : ""));
9434 return prog->check_substr ? prog->check_substr : prog->check_utf8;
9440 handles refcounting and freeing the perl core regexp structure. When
9441 it is necessary to actually free the structure the first thing it
9442 does is call the 'free' method of the regexp_engine associated to to
9443 the regexp, allowing the handling of the void *pprivate; member
9444 first. (This routine is not overridable by extensions, which is why
9445 the extensions free is called first.)
9447 See regdupe and regdupe_internal if you change anything here.
9449 #ifndef PERL_IN_XSUB_RE
9451 Perl_pregfree(pTHX_ REGEXP *r)
9457 Perl_pregfree2(pTHX_ REGEXP *rx)
9460 struct regexp *const r = (struct regexp *)SvANY(rx);
9461 GET_RE_DEBUG_FLAGS_DECL;
9463 PERL_ARGS_ASSERT_PREGFREE2;
9466 ReREFCNT_dec(r->mother_re);
9468 CALLREGFREE_PVT(rx); /* free the private data */
9469 SvREFCNT_dec(RXp_PAREN_NAMES(r));
9472 SvREFCNT_dec(r->anchored_substr);
9473 SvREFCNT_dec(r->anchored_utf8);
9474 SvREFCNT_dec(r->float_substr);
9475 SvREFCNT_dec(r->float_utf8);
9476 Safefree(r->substrs);
9478 RX_MATCH_COPY_FREE(rx);
9479 #ifdef PERL_OLD_COPY_ON_WRITE
9480 SvREFCNT_dec(r->saved_copy);
9487 This is a hacky workaround to the structural issue of match results
9488 being stored in the regexp structure which is in turn stored in
9489 PL_curpm/PL_reg_curpm. The problem is that due to qr// the pattern
9490 could be PL_curpm in multiple contexts, and could require multiple
9491 result sets being associated with the pattern simultaneously, such
9492 as when doing a recursive match with (??{$qr})
9494 The solution is to make a lightweight copy of the regexp structure
9495 when a qr// is returned from the code executed by (??{$qr}) this
9496 lightweight copy doesnt actually own any of its data except for
9497 the starp/end and the actual regexp structure itself.
9503 Perl_reg_temp_copy (pTHX_ REGEXP *ret_x, REGEXP *rx)
9506 struct regexp *const r = (struct regexp *)SvANY(rx);
9507 register const I32 npar = r->nparens+1;
9509 PERL_ARGS_ASSERT_REG_TEMP_COPY;
9512 ret_x = (REGEXP*) newSV_type(SVt_REGEXP);
9513 ret = (struct regexp *)SvANY(ret_x);
9515 (void)ReREFCNT_inc(rx);
9516 /* We can take advantage of the existing "copied buffer" mechanism in SVs
9517 by pointing directly at the buffer, but flagging that the allocated
9518 space in the copy is zero. As we've just done a struct copy, it's now
9519 a case of zero-ing that, rather than copying the current length. */
9520 SvPV_set(ret_x, RX_WRAPPED(rx));
9521 SvFLAGS(ret_x) |= SvFLAGS(rx) & (SVf_POK|SVp_POK|SVf_UTF8);
9522 memcpy(&(ret->xpv_cur), &(r->xpv_cur),
9523 sizeof(regexp) - STRUCT_OFFSET(regexp, xpv_cur));
9524 SvLEN_set(ret_x, 0);
9525 SvSTASH_set(ret_x, NULL);
9526 SvMAGIC_set(ret_x, NULL);
9527 Newx(ret->offs, npar, regexp_paren_pair);
9528 Copy(r->offs, ret->offs, npar, regexp_paren_pair);
9530 Newx(ret->substrs, 1, struct reg_substr_data);
9531 StructCopy(r->substrs, ret->substrs, struct reg_substr_data);
9533 SvREFCNT_inc_void(ret->anchored_substr);
9534 SvREFCNT_inc_void(ret->anchored_utf8);
9535 SvREFCNT_inc_void(ret->float_substr);
9536 SvREFCNT_inc_void(ret->float_utf8);
9538 /* check_substr and check_utf8, if non-NULL, point to either their
9539 anchored or float namesakes, and don't hold a second reference. */
9541 RX_MATCH_COPIED_off(ret_x);
9542 #ifdef PERL_OLD_COPY_ON_WRITE
9543 ret->saved_copy = NULL;
9545 ret->mother_re = rx;
9551 /* regfree_internal()
9553 Free the private data in a regexp. This is overloadable by
9554 extensions. Perl takes care of the regexp structure in pregfree(),
9555 this covers the *pprivate pointer which technically perldoesnt
9556 know about, however of course we have to handle the
9557 regexp_internal structure when no extension is in use.
9559 Note this is called before freeing anything in the regexp
9564 Perl_regfree_internal(pTHX_ REGEXP * const rx)
9567 struct regexp *const r = (struct regexp *)SvANY(rx);
9569 GET_RE_DEBUG_FLAGS_DECL;
9571 PERL_ARGS_ASSERT_REGFREE_INTERNAL;
9577 SV *dsv= sv_newmortal();
9578 RE_PV_QUOTED_DECL(s, RX_UTF8(rx),
9579 dsv, RX_PRECOMP(rx), RX_PRELEN(rx), 60);
9580 PerlIO_printf(Perl_debug_log,"%sFreeing REx:%s %s\n",
9581 PL_colors[4],PL_colors[5],s);
9584 #ifdef RE_TRACK_PATTERN_OFFSETS
9586 Safefree(ri->u.offsets); /* 20010421 MJD */
9589 int n = ri->data->count;
9590 PAD* new_comppad = NULL;
9595 /* If you add a ->what type here, update the comment in regcomp.h */
9596 switch (ri->data->what[n]) {
9601 SvREFCNT_dec(MUTABLE_SV(ri->data->data[n]));
9604 Safefree(ri->data->data[n]);
9607 new_comppad = MUTABLE_AV(ri->data->data[n]);
9610 if (new_comppad == NULL)
9611 Perl_croak(aTHX_ "panic: pregfree comppad");
9612 PAD_SAVE_LOCAL(old_comppad,
9613 /* Watch out for global destruction's random ordering. */
9614 (SvTYPE(new_comppad) == SVt_PVAV) ? new_comppad : NULL
9617 refcnt = OpREFCNT_dec((OP_4tree*)ri->data->data[n]);
9620 op_free((OP_4tree*)ri->data->data[n]);
9622 PAD_RESTORE_LOCAL(old_comppad);
9623 SvREFCNT_dec(MUTABLE_SV(new_comppad));
9629 { /* Aho Corasick add-on structure for a trie node.
9630 Used in stclass optimization only */
9632 reg_ac_data *aho=(reg_ac_data*)ri->data->data[n];
9634 refcount = --aho->refcount;
9637 PerlMemShared_free(aho->states);
9638 PerlMemShared_free(aho->fail);
9639 /* do this last!!!! */
9640 PerlMemShared_free(ri->data->data[n]);
9641 PerlMemShared_free(ri->regstclass);
9647 /* trie structure. */
9649 reg_trie_data *trie=(reg_trie_data*)ri->data->data[n];
9651 refcount = --trie->refcount;
9654 PerlMemShared_free(trie->charmap);
9655 PerlMemShared_free(trie->states);
9656 PerlMemShared_free(trie->trans);
9658 PerlMemShared_free(trie->bitmap);
9660 PerlMemShared_free(trie->jump);
9661 PerlMemShared_free(trie->wordinfo);
9662 /* do this last!!!! */
9663 PerlMemShared_free(ri->data->data[n]);
9668 Perl_croak(aTHX_ "panic: regfree data code '%c'", ri->data->what[n]);
9671 Safefree(ri->data->what);
9678 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
9679 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
9680 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
9683 re_dup - duplicate a regexp.
9685 This routine is expected to clone a given regexp structure. It is only
9686 compiled under USE_ITHREADS.
9688 After all of the core data stored in struct regexp is duplicated
9689 the regexp_engine.dupe method is used to copy any private data
9690 stored in the *pprivate pointer. This allows extensions to handle
9691 any duplication it needs to do.
9693 See pregfree() and regfree_internal() if you change anything here.
9695 #if defined(USE_ITHREADS)
9696 #ifndef PERL_IN_XSUB_RE
9698 Perl_re_dup_guts(pTHX_ const REGEXP *sstr, REGEXP *dstr, CLONE_PARAMS *param)
9702 const struct regexp *r = (const struct regexp *)SvANY(sstr);
9703 struct regexp *ret = (struct regexp *)SvANY(dstr);
9705 PERL_ARGS_ASSERT_RE_DUP_GUTS;
9707 npar = r->nparens+1;
9708 Newx(ret->offs, npar, regexp_paren_pair);
9709 Copy(r->offs, ret->offs, npar, regexp_paren_pair);
9711 /* no need to copy these */
9712 Newx(ret->swap, npar, regexp_paren_pair);
9716 /* Do it this way to avoid reading from *r after the StructCopy().
9717 That way, if any of the sv_dup_inc()s dislodge *r from the L1
9718 cache, it doesn't matter. */
9719 const bool anchored = r->check_substr
9720 ? r->check_substr == r->anchored_substr
9721 : r->check_utf8 == r->anchored_utf8;
9722 Newx(ret->substrs, 1, struct reg_substr_data);
9723 StructCopy(r->substrs, ret->substrs, struct reg_substr_data);
9725 ret->anchored_substr = sv_dup_inc(ret->anchored_substr, param);
9726 ret->anchored_utf8 = sv_dup_inc(ret->anchored_utf8, param);
9727 ret->float_substr = sv_dup_inc(ret->float_substr, param);
9728 ret->float_utf8 = sv_dup_inc(ret->float_utf8, param);
9730 /* check_substr and check_utf8, if non-NULL, point to either their
9731 anchored or float namesakes, and don't hold a second reference. */
9733 if (ret->check_substr) {
9735 assert(r->check_utf8 == r->anchored_utf8);
9736 ret->check_substr = ret->anchored_substr;
9737 ret->check_utf8 = ret->anchored_utf8;
9739 assert(r->check_substr == r->float_substr);
9740 assert(r->check_utf8 == r->float_utf8);
9741 ret->check_substr = ret->float_substr;
9742 ret->check_utf8 = ret->float_utf8;
9744 } else if (ret->check_utf8) {
9746 ret->check_utf8 = ret->anchored_utf8;
9748 ret->check_utf8 = ret->float_utf8;
9753 RXp_PAREN_NAMES(ret) = hv_dup_inc(RXp_PAREN_NAMES(ret), param);
9756 RXi_SET(ret,CALLREGDUPE_PVT(dstr,param));
9758 if (RX_MATCH_COPIED(dstr))
9759 ret->subbeg = SAVEPVN(ret->subbeg, ret->sublen);
9762 #ifdef PERL_OLD_COPY_ON_WRITE
9763 ret->saved_copy = NULL;
9766 if (ret->mother_re) {
9767 if (SvPVX_const(dstr) == SvPVX_const(ret->mother_re)) {
9768 /* Our storage points directly to our mother regexp, but that's
9769 1: a buffer in a different thread
9770 2: something we no longer hold a reference on
9771 so we need to copy it locally. */
9772 /* Note we need to sue SvCUR() on our mother_re, because it, in
9773 turn, may well be pointing to its own mother_re. */
9774 SvPV_set(dstr, SAVEPVN(SvPVX_const(ret->mother_re),
9775 SvCUR(ret->mother_re)+1));
9776 SvLEN_set(dstr, SvCUR(ret->mother_re)+1);
9778 ret->mother_re = NULL;
9782 #endif /* PERL_IN_XSUB_RE */
9787 This is the internal complement to regdupe() which is used to copy
9788 the structure pointed to by the *pprivate pointer in the regexp.
9789 This is the core version of the extension overridable cloning hook.
9790 The regexp structure being duplicated will be copied by perl prior
9791 to this and will be provided as the regexp *r argument, however
9792 with the /old/ structures pprivate pointer value. Thus this routine
9793 may override any copying normally done by perl.
9795 It returns a pointer to the new regexp_internal structure.
9799 Perl_regdupe_internal(pTHX_ REGEXP * const rx, CLONE_PARAMS *param)
9802 struct regexp *const r = (struct regexp *)SvANY(rx);
9803 regexp_internal *reti;
9807 PERL_ARGS_ASSERT_REGDUPE_INTERNAL;
9809 npar = r->nparens+1;
9812 Newxc(reti, sizeof(regexp_internal) + len*sizeof(regnode), char, regexp_internal);
9813 Copy(ri->program, reti->program, len+1, regnode);
9816 reti->regstclass = NULL;
9820 const int count = ri->data->count;
9823 Newxc(d, sizeof(struct reg_data) + count*sizeof(void *),
9824 char, struct reg_data);
9825 Newx(d->what, count, U8);
9828 for (i = 0; i < count; i++) {
9829 d->what[i] = ri->data->what[i];
9830 switch (d->what[i]) {
9831 /* legal options are one of: sSfpontTua
9832 see also regcomp.h and pregfree() */
9833 case 'a': /* actually an AV, but the dup function is identical. */
9836 case 'p': /* actually an AV, but the dup function is identical. */
9837 case 'u': /* actually an HV, but the dup function is identical. */
9838 d->data[i] = sv_dup_inc((const SV *)ri->data->data[i], param);
9841 /* This is cheating. */
9842 Newx(d->data[i], 1, struct regnode_charclass_class);
9843 StructCopy(ri->data->data[i], d->data[i],
9844 struct regnode_charclass_class);
9845 reti->regstclass = (regnode*)d->data[i];
9848 /* Compiled op trees are readonly and in shared memory,
9849 and can thus be shared without duplication. */
9851 d->data[i] = (void*)OpREFCNT_inc((OP*)ri->data->data[i]);
9855 /* Trie stclasses are readonly and can thus be shared
9856 * without duplication. We free the stclass in pregfree
9857 * when the corresponding reg_ac_data struct is freed.
9859 reti->regstclass= ri->regstclass;
9863 ((reg_trie_data*)ri->data->data[i])->refcount++;
9867 d->data[i] = ri->data->data[i];
9870 Perl_croak(aTHX_ "panic: re_dup unknown data code '%c'", ri->data->what[i]);
9879 reti->name_list_idx = ri->name_list_idx;
9881 #ifdef RE_TRACK_PATTERN_OFFSETS
9882 if (ri->u.offsets) {
9883 Newx(reti->u.offsets, 2*len+1, U32);
9884 Copy(ri->u.offsets, reti->u.offsets, 2*len+1, U32);
9887 SetProgLen(reti,len);
9893 #endif /* USE_ITHREADS */
9895 #ifndef PERL_IN_XSUB_RE
9898 - regnext - dig the "next" pointer out of a node
9901 Perl_regnext(pTHX_ register regnode *p)
9904 register I32 offset;
9909 if (OP(p) > REGNODE_MAX) { /* regnode.type is unsigned */
9910 Perl_croak(aTHX_ "Corrupted regexp opcode %d > %d", (int)OP(p), (int)REGNODE_MAX);
9913 offset = (reg_off_by_arg[OP(p)] ? ARG(p) : NEXT_OFF(p));
9922 S_re_croak2(pTHX_ const char* pat1,const char* pat2,...)
9925 STRLEN l1 = strlen(pat1);
9926 STRLEN l2 = strlen(pat2);
9929 const char *message;
9931 PERL_ARGS_ASSERT_RE_CROAK2;
9937 Copy(pat1, buf, l1 , char);
9938 Copy(pat2, buf + l1, l2 , char);
9939 buf[l1 + l2] = '\n';
9940 buf[l1 + l2 + 1] = '\0';
9942 /* ANSI variant takes additional second argument */
9943 va_start(args, pat2);
9947 msv = vmess(buf, &args);
9949 message = SvPV_const(msv,l1);
9952 Copy(message, buf, l1 , char);
9953 buf[l1-1] = '\0'; /* Overwrite \n */
9954 Perl_croak(aTHX_ "%s", buf);
9957 /* XXX Here's a total kludge. But we need to re-enter for swash routines. */
9959 #ifndef PERL_IN_XSUB_RE
9961 Perl_save_re_context(pTHX)
9965 struct re_save_state *state;
9967 SAVEVPTR(PL_curcop);
9968 SSGROW(SAVESTACK_ALLOC_FOR_RE_SAVE_STATE + 1);
9970 state = (struct re_save_state *)(PL_savestack + PL_savestack_ix);
9971 PL_savestack_ix += SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
9972 SSPUSHUV(SAVEt_RE_STATE);
9974 Copy(&PL_reg_state, state, 1, struct re_save_state);
9976 PL_reg_start_tmp = 0;
9977 PL_reg_start_tmpl = 0;
9978 PL_reg_oldsaved = NULL;
9979 PL_reg_oldsavedlen = 0;
9981 PL_reg_leftiter = 0;
9982 PL_reg_poscache = NULL;
9983 PL_reg_poscache_size = 0;
9984 #ifdef PERL_OLD_COPY_ON_WRITE
9988 /* Save $1..$n (#18107: UTF-8 s/(\w+)/uc($1)/e); AMS 20021106. */
9990 const REGEXP * const rx = PM_GETRE(PL_curpm);
9993 for (i = 1; i <= RX_NPARENS(rx); i++) {
9994 char digits[TYPE_CHARS(long)];
9995 const STRLEN len = my_snprintf(digits, sizeof(digits), "%lu", (long)i);
9996 GV *const *const gvp
9997 = (GV**)hv_fetch(PL_defstash, digits, len, 0);
10000 GV * const gv = *gvp;
10001 if (SvTYPE(gv) == SVt_PVGV && GvSV(gv))
10011 clear_re(pTHX_ void *r)
10014 ReREFCNT_dec((REGEXP *)r);
10020 S_put_byte(pTHX_ SV *sv, int c)
10022 PERL_ARGS_ASSERT_PUT_BYTE;
10024 /* Our definition of isPRINT() ignores locales, so only bytes that are
10025 not part of UTF-8 are considered printable. I assume that the same
10026 holds for UTF-EBCDIC.
10027 Also, code point 255 is not printable in either (it's E0 in EBCDIC,
10028 which Wikipedia says:
10030 EO, or Eight Ones, is an 8-bit EBCDIC character code represented as all
10031 ones (binary 1111 1111, hexadecimal FF). It is similar, but not
10032 identical, to the ASCII delete (DEL) or rubout control character.
10033 ) So the old condition can be simplified to !isPRINT(c) */
10035 Perl_sv_catpvf(aTHX_ sv, "\\%o", c);
10037 const char string = c;
10038 if (c == '-' || c == ']' || c == '\\' || c == '^')
10039 sv_catpvs(sv, "\\");
10040 sv_catpvn(sv, &string, 1);
10045 #define CLEAR_OPTSTART \
10046 if (optstart) STMT_START { \
10047 DEBUG_OPTIMISE_r(PerlIO_printf(Perl_debug_log, " (%"IVdf" nodes)\n", (IV)(node - optstart))); \
10051 #define DUMPUNTIL(b,e) CLEAR_OPTSTART; node=dumpuntil(r,start,(b),(e),last,sv,indent+1,depth+1);
10053 STATIC const regnode *
10054 S_dumpuntil(pTHX_ const regexp *r, const regnode *start, const regnode *node,
10055 const regnode *last, const regnode *plast,
10056 SV* sv, I32 indent, U32 depth)
10059 register U8 op = PSEUDO; /* Arbitrary non-END op. */
10060 register const regnode *next;
10061 const regnode *optstart= NULL;
10063 RXi_GET_DECL(r,ri);
10064 GET_RE_DEBUG_FLAGS_DECL;
10066 PERL_ARGS_ASSERT_DUMPUNTIL;
10068 #ifdef DEBUG_DUMPUNTIL
10069 PerlIO_printf(Perl_debug_log, "--- %d : %d - %d - %d\n",indent,node-start,
10070 last ? last-start : 0,plast ? plast-start : 0);
10073 if (plast && plast < last)
10076 while (PL_regkind[op] != END && (!last || node < last)) {
10077 /* While that wasn't END last time... */
10080 if (op == CLOSE || op == WHILEM)
10082 next = regnext((regnode *)node);
10085 if (OP(node) == OPTIMIZED) {
10086 if (!optstart && RE_DEBUG_FLAG(RE_DEBUG_COMPILE_OPTIMISE))
10093 regprop(r, sv, node);
10094 PerlIO_printf(Perl_debug_log, "%4"IVdf":%*s%s", (IV)(node - start),
10095 (int)(2*indent + 1), "", SvPVX_const(sv));
10097 if (OP(node) != OPTIMIZED) {
10098 if (next == NULL) /* Next ptr. */
10099 PerlIO_printf(Perl_debug_log, " (0)");
10100 else if (PL_regkind[(U8)op] == BRANCH && PL_regkind[OP(next)] != BRANCH )
10101 PerlIO_printf(Perl_debug_log, " (FAIL)");
10103 PerlIO_printf(Perl_debug_log, " (%"IVdf")", (IV)(next - start));
10104 (void)PerlIO_putc(Perl_debug_log, '\n');
10108 if (PL_regkind[(U8)op] == BRANCHJ) {
10111 register const regnode *nnode = (OP(next) == LONGJMP
10112 ? regnext((regnode *)next)
10114 if (last && nnode > last)
10116 DUMPUNTIL(NEXTOPER(NEXTOPER(node)), nnode);
10119 else if (PL_regkind[(U8)op] == BRANCH) {
10121 DUMPUNTIL(NEXTOPER(node), next);
10123 else if ( PL_regkind[(U8)op] == TRIE ) {
10124 const regnode *this_trie = node;
10125 const char op = OP(node);
10126 const U32 n = ARG(node);
10127 const reg_ac_data * const ac = op>=AHOCORASICK ?
10128 (reg_ac_data *)ri->data->data[n] :
10130 const reg_trie_data * const trie =
10131 (reg_trie_data*)ri->data->data[op<AHOCORASICK ? n : ac->trie];
10133 AV *const trie_words = MUTABLE_AV(ri->data->data[n + TRIE_WORDS_OFFSET]);
10135 const regnode *nextbranch= NULL;
10138 for (word_idx= 0; word_idx < (I32)trie->wordcount; word_idx++) {
10139 SV ** const elem_ptr = av_fetch(trie_words,word_idx,0);
10141 PerlIO_printf(Perl_debug_log, "%*s%s ",
10142 (int)(2*(indent+3)), "",
10143 elem_ptr ? pv_pretty(sv, SvPV_nolen_const(*elem_ptr), SvCUR(*elem_ptr), 60,
10144 PL_colors[0], PL_colors[1],
10145 (SvUTF8(*elem_ptr) ? PERL_PV_ESCAPE_UNI : 0) |
10146 PERL_PV_PRETTY_ELLIPSES |
10147 PERL_PV_PRETTY_LTGT
10152 U16 dist= trie->jump[word_idx+1];
10153 PerlIO_printf(Perl_debug_log, "(%"UVuf")\n",
10154 (UV)((dist ? this_trie + dist : next) - start));
10157 nextbranch= this_trie + trie->jump[0];
10158 DUMPUNTIL(this_trie + dist, nextbranch);
10160 if (nextbranch && PL_regkind[OP(nextbranch)]==BRANCH)
10161 nextbranch= regnext((regnode *)nextbranch);
10163 PerlIO_printf(Perl_debug_log, "\n");
10166 if (last && next > last)
10171 else if ( op == CURLY ) { /* "next" might be very big: optimizer */
10172 DUMPUNTIL(NEXTOPER(node) + EXTRA_STEP_2ARGS,
10173 NEXTOPER(node) + EXTRA_STEP_2ARGS + 1);
10175 else if (PL_regkind[(U8)op] == CURLY && op != CURLYX) {
10177 DUMPUNTIL(NEXTOPER(node) + EXTRA_STEP_2ARGS, next);
10179 else if ( op == PLUS || op == STAR) {
10180 DUMPUNTIL(NEXTOPER(node), NEXTOPER(node) + 1);
10182 else if (op == ANYOF) {
10183 /* arglen 1 + class block */
10184 node += 1 + ((ANYOF_FLAGS(node) & ANYOF_LARGE)
10185 ? ANYOF_CLASS_SKIP : ANYOF_SKIP);
10186 node = NEXTOPER(node);
10188 else if (PL_regkind[(U8)op] == EXACT) {
10189 /* Literal string, where present. */
10190 node += NODE_SZ_STR(node) - 1;
10191 node = NEXTOPER(node);
10194 node = NEXTOPER(node);
10195 node += regarglen[(U8)op];
10197 if (op == CURLYX || op == OPEN)
10201 #ifdef DEBUG_DUMPUNTIL
10202 PerlIO_printf(Perl_debug_log, "--- %d\n", (int)indent);
10207 #endif /* DEBUGGING */
10211 * c-indentation-style: bsd
10212 * c-basic-offset: 4
10213 * indent-tabs-mode: t
10216 * ex: set ts=8 sts=4 sw=4 noet:
projects / perl / modules / re-engine-Hooks.git / blob