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
88 #include "dquote_static.c"
95 # if defined(BUGGY_MSC6)
96 /* MSC 6.00A breaks on op/regexp.t test 85 unless we turn this off */
97 # pragma optimize("a",off)
98 /* But MSC 6.00A is happy with 'w', for aliases only across function calls*/
99 # pragma optimize("w",on )
100 # endif /* BUGGY_MSC6 */
104 #define STATIC static
107 typedef struct RExC_state_t {
108 U32 flags; /* are we folding, multilining? */
109 char *precomp; /* uncompiled string. */
110 REGEXP *rx_sv; /* The SV that is the regexp. */
111 regexp *rx; /* perl core regexp structure */
112 regexp_internal *rxi; /* internal data for regexp object pprivate field */
113 char *start; /* Start of input for compile */
114 char *end; /* End of input for compile */
115 char *parse; /* Input-scan pointer. */
116 I32 whilem_seen; /* number of WHILEM in this expr */
117 regnode *emit_start; /* Start of emitted-code area */
118 regnode *emit_bound; /* First regnode outside of the allocated space */
119 regnode *emit; /* Code-emit pointer; ®dummy = don't = compiling */
120 I32 naughty; /* How bad is this pattern? */
121 I32 sawback; /* Did we see \1, ...? */
123 I32 size; /* Code size. */
124 I32 npar; /* Capture buffer count, (OPEN). */
125 I32 cpar; /* Capture buffer count, (CLOSE). */
126 I32 nestroot; /* root parens we are in - used by accept */
130 regnode **open_parens; /* pointers to open parens */
131 regnode **close_parens; /* pointers to close parens */
132 regnode *opend; /* END node in program */
133 I32 utf8; /* whether the pattern is utf8 or not */
134 I32 orig_utf8; /* whether the pattern was originally in utf8 */
135 /* XXX use this for future optimisation of case
136 * where pattern must be upgraded to utf8. */
137 I32 uni_semantics; /* If a d charset modifier should use unicode
138 rules, even if the pattern is not in
140 HV *paren_names; /* Paren names */
142 regnode **recurse; /* Recurse regops */
143 I32 recurse_count; /* Number of recurse regops */
146 I32 override_recoding;
148 char *starttry; /* -Dr: where regtry was called. */
149 #define RExC_starttry (pRExC_state->starttry)
152 const char *lastparse;
154 AV *paren_name_list; /* idx -> name */
155 #define RExC_lastparse (pRExC_state->lastparse)
156 #define RExC_lastnum (pRExC_state->lastnum)
157 #define RExC_paren_name_list (pRExC_state->paren_name_list)
161 #define RExC_flags (pRExC_state->flags)
162 #define RExC_precomp (pRExC_state->precomp)
163 #define RExC_rx_sv (pRExC_state->rx_sv)
164 #define RExC_rx (pRExC_state->rx)
165 #define RExC_rxi (pRExC_state->rxi)
166 #define RExC_start (pRExC_state->start)
167 #define RExC_end (pRExC_state->end)
168 #define RExC_parse (pRExC_state->parse)
169 #define RExC_whilem_seen (pRExC_state->whilem_seen)
170 #ifdef RE_TRACK_PATTERN_OFFSETS
171 #define RExC_offsets (pRExC_state->rxi->u.offsets) /* I am not like the others */
173 #define RExC_emit (pRExC_state->emit)
174 #define RExC_emit_start (pRExC_state->emit_start)
175 #define RExC_emit_bound (pRExC_state->emit_bound)
176 #define RExC_naughty (pRExC_state->naughty)
177 #define RExC_sawback (pRExC_state->sawback)
178 #define RExC_seen (pRExC_state->seen)
179 #define RExC_size (pRExC_state->size)
180 #define RExC_npar (pRExC_state->npar)
181 #define RExC_nestroot (pRExC_state->nestroot)
182 #define RExC_extralen (pRExC_state->extralen)
183 #define RExC_seen_zerolen (pRExC_state->seen_zerolen)
184 #define RExC_seen_evals (pRExC_state->seen_evals)
185 #define RExC_utf8 (pRExC_state->utf8)
186 #define RExC_uni_semantics (pRExC_state->uni_semantics)
187 #define RExC_orig_utf8 (pRExC_state->orig_utf8)
188 #define RExC_open_parens (pRExC_state->open_parens)
189 #define RExC_close_parens (pRExC_state->close_parens)
190 #define RExC_opend (pRExC_state->opend)
191 #define RExC_paren_names (pRExC_state->paren_names)
192 #define RExC_recurse (pRExC_state->recurse)
193 #define RExC_recurse_count (pRExC_state->recurse_count)
194 #define RExC_in_lookbehind (pRExC_state->in_lookbehind)
195 #define RExC_contains_locale (pRExC_state->contains_locale)
196 #define RExC_override_recoding (pRExC_state->override_recoding)
199 #define ISMULT1(c) ((c) == '*' || (c) == '+' || (c) == '?')
200 #define ISMULT2(s) ((*s) == '*' || (*s) == '+' || (*s) == '?' || \
201 ((*s) == '{' && regcurly(s)))
204 #undef SPSTART /* dratted cpp namespace... */
207 * Flags to be passed up and down.
209 #define WORST 0 /* Worst case. */
210 #define HASWIDTH 0x01 /* Known to match non-null strings. */
212 /* Simple enough to be STAR/PLUS operand, in an EXACT node must be a single
213 * character, and if utf8, must be invariant. Note that this is not the same thing as REGNODE_SIMPLE */
215 #define SPSTART 0x04 /* Starts with * or +. */
216 #define TRYAGAIN 0x08 /* Weeded out a declaration. */
217 #define POSTPONED 0x10 /* (?1),(?&name), (??{...}) or similar */
219 #define REG_NODE_NUM(x) ((x) ? (int)((x)-RExC_emit_start) : -1)
221 /* whether trie related optimizations are enabled */
222 #if PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION
223 #define TRIE_STUDY_OPT
224 #define FULL_TRIE_STUDY
230 #define PBYTE(u8str,paren) ((U8*)(u8str))[(paren) >> 3]
231 #define PBITVAL(paren) (1 << ((paren) & 7))
232 #define PAREN_TEST(u8str,paren) ( PBYTE(u8str,paren) & PBITVAL(paren))
233 #define PAREN_SET(u8str,paren) PBYTE(u8str,paren) |= PBITVAL(paren)
234 #define PAREN_UNSET(u8str,paren) PBYTE(u8str,paren) &= (~PBITVAL(paren))
236 /* If not already in utf8, do a longjmp back to the beginning */
237 #define UTF8_LONGJMP 42 /* Choose a value not likely to ever conflict */
238 #define REQUIRE_UTF8 STMT_START { \
239 if (! UTF) JMPENV_JUMP(UTF8_LONGJMP); \
242 /* About scan_data_t.
244 During optimisation we recurse through the regexp program performing
245 various inplace (keyhole style) optimisations. In addition study_chunk
246 and scan_commit populate this data structure with information about
247 what strings MUST appear in the pattern. We look for the longest
248 string that must appear at a fixed location, and we look for the
249 longest string that may appear at a floating location. So for instance
254 Both 'FOO' and 'A' are fixed strings. Both 'B' and 'BAR' are floating
255 strings (because they follow a .* construct). study_chunk will identify
256 both FOO and BAR as being the longest fixed and floating strings respectively.
258 The strings can be composites, for instance
262 will result in a composite fixed substring 'foo'.
264 For each string some basic information is maintained:
266 - offset or min_offset
267 This is the position the string must appear at, or not before.
268 It also implicitly (when combined with minlenp) tells us how many
269 characters must match before the string we are searching for.
270 Likewise when combined with minlenp and the length of the string it
271 tells us how many characters must appear after the string we have
275 Only used for floating strings. This is the rightmost point that
276 the string can appear at. If set to I32 max it indicates that the
277 string can occur infinitely far to the right.
280 A pointer to the minimum length of the pattern that the string
281 was found inside. This is important as in the case of positive
282 lookahead or positive lookbehind we can have multiple patterns
287 The minimum length of the pattern overall is 3, the minimum length
288 of the lookahead part is 3, but the minimum length of the part that
289 will actually match is 1. So 'FOO's minimum length is 3, but the
290 minimum length for the F is 1. This is important as the minimum length
291 is used to determine offsets in front of and behind the string being
292 looked for. Since strings can be composites this is the length of the
293 pattern at the time it was committed with a scan_commit. Note that
294 the length is calculated by study_chunk, so that the minimum lengths
295 are not known until the full pattern has been compiled, thus the
296 pointer to the value.
300 In the case of lookbehind the string being searched for can be
301 offset past the start point of the final matching string.
302 If this value was just blithely removed from the min_offset it would
303 invalidate some of the calculations for how many chars must match
304 before or after (as they are derived from min_offset and minlen and
305 the length of the string being searched for).
306 When the final pattern is compiled and the data is moved from the
307 scan_data_t structure into the regexp structure the information
308 about lookbehind is factored in, with the information that would
309 have been lost precalculated in the end_shift field for the
312 The fields pos_min and pos_delta are used to store the minimum offset
313 and the delta to the maximum offset at the current point in the pattern.
317 typedef struct scan_data_t {
318 /*I32 len_min; unused */
319 /*I32 len_delta; unused */
323 I32 last_end; /* min value, <0 unless valid. */
326 SV **longest; /* Either &l_fixed, or &l_float. */
327 SV *longest_fixed; /* longest fixed string found in pattern */
328 I32 offset_fixed; /* offset where it starts */
329 I32 *minlen_fixed; /* pointer to the minlen relevant to the string */
330 I32 lookbehind_fixed; /* is the position of the string modfied by LB */
331 SV *longest_float; /* longest floating string found in pattern */
332 I32 offset_float_min; /* earliest point in string it can appear */
333 I32 offset_float_max; /* latest point in string it can appear */
334 I32 *minlen_float; /* pointer to the minlen relevant to the string */
335 I32 lookbehind_float; /* is the position of the string modified by LB */
339 struct regnode_charclass_class *start_class;
343 * Forward declarations for pregcomp()'s friends.
346 static const scan_data_t zero_scan_data =
347 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ,0};
349 #define SF_BEFORE_EOL (SF_BEFORE_SEOL|SF_BEFORE_MEOL)
350 #define SF_BEFORE_SEOL 0x0001
351 #define SF_BEFORE_MEOL 0x0002
352 #define SF_FIX_BEFORE_EOL (SF_FIX_BEFORE_SEOL|SF_FIX_BEFORE_MEOL)
353 #define SF_FL_BEFORE_EOL (SF_FL_BEFORE_SEOL|SF_FL_BEFORE_MEOL)
356 # define SF_FIX_SHIFT_EOL (0+2)
357 # define SF_FL_SHIFT_EOL (0+4)
359 # define SF_FIX_SHIFT_EOL (+2)
360 # define SF_FL_SHIFT_EOL (+4)
363 #define SF_FIX_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FIX_SHIFT_EOL)
364 #define SF_FIX_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FIX_SHIFT_EOL)
366 #define SF_FL_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FL_SHIFT_EOL)
367 #define SF_FL_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FL_SHIFT_EOL) /* 0x20 */
368 #define SF_IS_INF 0x0040
369 #define SF_HAS_PAR 0x0080
370 #define SF_IN_PAR 0x0100
371 #define SF_HAS_EVAL 0x0200
372 #define SCF_DO_SUBSTR 0x0400
373 #define SCF_DO_STCLASS_AND 0x0800
374 #define SCF_DO_STCLASS_OR 0x1000
375 #define SCF_DO_STCLASS (SCF_DO_STCLASS_AND|SCF_DO_STCLASS_OR)
376 #define SCF_WHILEM_VISITED_POS 0x2000
378 #define SCF_TRIE_RESTUDY 0x4000 /* Do restudy? */
379 #define SCF_SEEN_ACCEPT 0x8000
381 #define UTF cBOOL(RExC_utf8)
382 #define LOC (get_regex_charset(RExC_flags) == REGEX_LOCALE_CHARSET)
383 #define UNI_SEMANTICS (get_regex_charset(RExC_flags) == REGEX_UNICODE_CHARSET)
384 #define DEPENDS_SEMANTICS (get_regex_charset(RExC_flags) == REGEX_DEPENDS_CHARSET)
385 #define AT_LEAST_UNI_SEMANTICS (get_regex_charset(RExC_flags) >= REGEX_UNICODE_CHARSET)
386 #define ASCII_RESTRICTED (get_regex_charset(RExC_flags) == REGEX_ASCII_RESTRICTED_CHARSET)
387 #define MORE_ASCII_RESTRICTED (get_regex_charset(RExC_flags) == REGEX_ASCII_MORE_RESTRICTED_CHARSET)
388 #define AT_LEAST_ASCII_RESTRICTED (get_regex_charset(RExC_flags) >= REGEX_ASCII_RESTRICTED_CHARSET)
390 #define FOLD cBOOL(RExC_flags & RXf_PMf_FOLD)
392 #define OOB_UNICODE 12345678
393 #define OOB_NAMEDCLASS -1
395 #define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv))
396 #define CHR_DIST(a,b) (UTF ? utf8_distance(a,b) : a - b)
399 /* length of regex to show in messages that don't mark a position within */
400 #define RegexLengthToShowInErrorMessages 127
403 * If MARKER[12] are adjusted, be sure to adjust the constants at the top
404 * of t/op/regmesg.t, the tests in t/op/re_tests, and those in
405 * op/pragma/warn/regcomp.
407 #define MARKER1 "<-- HERE" /* marker as it appears in the description */
408 #define MARKER2 " <-- HERE " /* marker as it appears within the regex */
410 #define REPORT_LOCATION " in regex; marked by " MARKER1 " in m/%.*s" MARKER2 "%s/"
413 * Calls SAVEDESTRUCTOR_X if needed, then calls Perl_croak with the given
414 * arg. Show regex, up to a maximum length. If it's too long, chop and add
417 #define _FAIL(code) STMT_START { \
418 const char *ellipses = ""; \
419 IV len = RExC_end - RExC_precomp; \
422 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx_sv); \
423 if (len > RegexLengthToShowInErrorMessages) { \
424 /* chop 10 shorter than the max, to ensure meaning of "..." */ \
425 len = RegexLengthToShowInErrorMessages - 10; \
431 #define FAIL(msg) _FAIL( \
432 Perl_croak(aTHX_ "%s in regex m/%.*s%s/", \
433 msg, (int)len, RExC_precomp, ellipses))
435 #define FAIL2(msg,arg) _FAIL( \
436 Perl_croak(aTHX_ msg " in regex m/%.*s%s/", \
437 arg, (int)len, RExC_precomp, ellipses))
440 * Simple_vFAIL -- like FAIL, but marks the current location in the scan
442 #define Simple_vFAIL(m) STMT_START { \
443 const IV offset = RExC_parse - RExC_precomp; \
444 Perl_croak(aTHX_ "%s" REPORT_LOCATION, \
445 m, (int)offset, RExC_precomp, RExC_precomp + offset); \
449 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL()
451 #define vFAIL(m) STMT_START { \
453 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx_sv); \
458 * Like Simple_vFAIL(), but accepts two arguments.
460 #define Simple_vFAIL2(m,a1) STMT_START { \
461 const IV offset = RExC_parse - RExC_precomp; \
462 S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, \
463 (int)offset, RExC_precomp, RExC_precomp + offset); \
467 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL2().
469 #define vFAIL2(m,a1) STMT_START { \
471 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx_sv); \
472 Simple_vFAIL2(m, a1); \
477 * Like Simple_vFAIL(), but accepts three arguments.
479 #define Simple_vFAIL3(m, a1, a2) STMT_START { \
480 const IV offset = RExC_parse - RExC_precomp; \
481 S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, a2, \
482 (int)offset, RExC_precomp, RExC_precomp + offset); \
486 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL3().
488 #define vFAIL3(m,a1,a2) STMT_START { \
490 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx_sv); \
491 Simple_vFAIL3(m, a1, a2); \
495 * Like Simple_vFAIL(), but accepts four arguments.
497 #define Simple_vFAIL4(m, a1, a2, a3) STMT_START { \
498 const IV offset = RExC_parse - RExC_precomp; \
499 S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, a2, a3, \
500 (int)offset, RExC_precomp, RExC_precomp + offset); \
503 #define ckWARNreg(loc,m) STMT_START { \
504 const IV offset = loc - RExC_precomp; \
505 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
506 (int)offset, RExC_precomp, RExC_precomp + offset); \
509 #define ckWARNregdep(loc,m) STMT_START { \
510 const IV offset = loc - RExC_precomp; \
511 Perl_ck_warner_d(aTHX_ packWARN2(WARN_DEPRECATED, WARN_REGEXP), \
513 (int)offset, RExC_precomp, RExC_precomp + offset); \
516 #define ckWARN2regdep(loc,m, a1) STMT_START { \
517 const IV offset = loc - RExC_precomp; \
518 Perl_ck_warner_d(aTHX_ packWARN2(WARN_DEPRECATED, WARN_REGEXP), \
520 a1, (int)offset, RExC_precomp, RExC_precomp + offset); \
523 #define ckWARN2reg(loc, m, a1) STMT_START { \
524 const IV offset = loc - RExC_precomp; \
525 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
526 a1, (int)offset, RExC_precomp, RExC_precomp + offset); \
529 #define vWARN3(loc, m, a1, a2) STMT_START { \
530 const IV offset = loc - RExC_precomp; \
531 Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
532 a1, a2, (int)offset, RExC_precomp, RExC_precomp + offset); \
535 #define ckWARN3reg(loc, m, a1, a2) STMT_START { \
536 const IV offset = loc - RExC_precomp; \
537 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
538 a1, a2, (int)offset, RExC_precomp, RExC_precomp + offset); \
541 #define vWARN4(loc, m, a1, a2, a3) STMT_START { \
542 const IV offset = loc - RExC_precomp; \
543 Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
544 a1, a2, a3, (int)offset, RExC_precomp, RExC_precomp + offset); \
547 #define ckWARN4reg(loc, m, a1, a2, a3) STMT_START { \
548 const IV offset = loc - RExC_precomp; \
549 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
550 a1, a2, a3, (int)offset, RExC_precomp, RExC_precomp + offset); \
553 #define vWARN5(loc, m, a1, a2, a3, a4) STMT_START { \
554 const IV offset = loc - RExC_precomp; \
555 Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
556 a1, a2, a3, a4, (int)offset, RExC_precomp, RExC_precomp + offset); \
560 /* Allow for side effects in s */
561 #define REGC(c,s) STMT_START { \
562 if (!SIZE_ONLY) *(s) = (c); else (void)(s); \
565 /* Macros for recording node offsets. 20001227 mjd@plover.com
566 * Nodes are numbered 1, 2, 3, 4. Node #n's position is recorded in
567 * element 2*n-1 of the array. Element #2n holds the byte length node #n.
568 * Element 0 holds the number n.
569 * Position is 1 indexed.
571 #ifndef RE_TRACK_PATTERN_OFFSETS
572 #define Set_Node_Offset_To_R(node,byte)
573 #define Set_Node_Offset(node,byte)
574 #define Set_Cur_Node_Offset
575 #define Set_Node_Length_To_R(node,len)
576 #define Set_Node_Length(node,len)
577 #define Set_Node_Cur_Length(node)
578 #define Node_Offset(n)
579 #define Node_Length(n)
580 #define Set_Node_Offset_Length(node,offset,len)
581 #define ProgLen(ri) ri->u.proglen
582 #define SetProgLen(ri,x) ri->u.proglen = x
584 #define ProgLen(ri) ri->u.offsets[0]
585 #define SetProgLen(ri,x) ri->u.offsets[0] = x
586 #define Set_Node_Offset_To_R(node,byte) STMT_START { \
588 MJD_OFFSET_DEBUG(("** (%d) offset of node %d is %d.\n", \
589 __LINE__, (int)(node), (int)(byte))); \
591 Perl_croak(aTHX_ "value of node is %d in Offset macro", (int)(node)); \
593 RExC_offsets[2*(node)-1] = (byte); \
598 #define Set_Node_Offset(node,byte) \
599 Set_Node_Offset_To_R((node)-RExC_emit_start, (byte)-RExC_start)
600 #define Set_Cur_Node_Offset Set_Node_Offset(RExC_emit, RExC_parse)
602 #define Set_Node_Length_To_R(node,len) STMT_START { \
604 MJD_OFFSET_DEBUG(("** (%d) size of node %d is %d.\n", \
605 __LINE__, (int)(node), (int)(len))); \
607 Perl_croak(aTHX_ "value of node is %d in Length macro", (int)(node)); \
609 RExC_offsets[2*(node)] = (len); \
614 #define Set_Node_Length(node,len) \
615 Set_Node_Length_To_R((node)-RExC_emit_start, len)
616 #define Set_Cur_Node_Length(len) Set_Node_Length(RExC_emit, len)
617 #define Set_Node_Cur_Length(node) \
618 Set_Node_Length(node, RExC_parse - parse_start)
620 /* Get offsets and lengths */
621 #define Node_Offset(n) (RExC_offsets[2*((n)-RExC_emit_start)-1])
622 #define Node_Length(n) (RExC_offsets[2*((n)-RExC_emit_start)])
624 #define Set_Node_Offset_Length(node,offset,len) STMT_START { \
625 Set_Node_Offset_To_R((node)-RExC_emit_start, (offset)); \
626 Set_Node_Length_To_R((node)-RExC_emit_start, (len)); \
630 #if PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS
631 #define EXPERIMENTAL_INPLACESCAN
632 #endif /*PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS*/
634 #define DEBUG_STUDYDATA(str,data,depth) \
635 DEBUG_OPTIMISE_MORE_r(if(data){ \
636 PerlIO_printf(Perl_debug_log, \
637 "%*s" str "Pos:%"IVdf"/%"IVdf \
638 " Flags: 0x%"UVXf" Whilem_c: %"IVdf" Lcp: %"IVdf" %s", \
639 (int)(depth)*2, "", \
640 (IV)((data)->pos_min), \
641 (IV)((data)->pos_delta), \
642 (UV)((data)->flags), \
643 (IV)((data)->whilem_c), \
644 (IV)((data)->last_closep ? *((data)->last_closep) : -1), \
645 is_inf ? "INF " : "" \
647 if ((data)->last_found) \
648 PerlIO_printf(Perl_debug_log, \
649 "Last:'%s' %"IVdf":%"IVdf"/%"IVdf" %sFixed:'%s' @ %"IVdf \
650 " %sFloat: '%s' @ %"IVdf"/%"IVdf"", \
651 SvPVX_const((data)->last_found), \
652 (IV)((data)->last_end), \
653 (IV)((data)->last_start_min), \
654 (IV)((data)->last_start_max), \
655 ((data)->longest && \
656 (data)->longest==&((data)->longest_fixed)) ? "*" : "", \
657 SvPVX_const((data)->longest_fixed), \
658 (IV)((data)->offset_fixed), \
659 ((data)->longest && \
660 (data)->longest==&((data)->longest_float)) ? "*" : "", \
661 SvPVX_const((data)->longest_float), \
662 (IV)((data)->offset_float_min), \
663 (IV)((data)->offset_float_max) \
665 PerlIO_printf(Perl_debug_log,"\n"); \
668 static void clear_re(pTHX_ void *r);
670 /* Mark that we cannot extend a found fixed substring at this point.
671 Update the longest found anchored substring and the longest found
672 floating substrings if needed. */
675 S_scan_commit(pTHX_ const RExC_state_t *pRExC_state, scan_data_t *data, I32 *minlenp, int is_inf)
677 const STRLEN l = CHR_SVLEN(data->last_found);
678 const STRLEN old_l = CHR_SVLEN(*data->longest);
679 GET_RE_DEBUG_FLAGS_DECL;
681 PERL_ARGS_ASSERT_SCAN_COMMIT;
683 if ((l >= old_l) && ((l > old_l) || (data->flags & SF_BEFORE_EOL))) {
684 SvSetMagicSV(*data->longest, data->last_found);
685 if (*data->longest == data->longest_fixed) {
686 data->offset_fixed = l ? data->last_start_min : data->pos_min;
687 if (data->flags & SF_BEFORE_EOL)
689 |= ((data->flags & SF_BEFORE_EOL) << SF_FIX_SHIFT_EOL);
691 data->flags &= ~SF_FIX_BEFORE_EOL;
692 data->minlen_fixed=minlenp;
693 data->lookbehind_fixed=0;
695 else { /* *data->longest == data->longest_float */
696 data->offset_float_min = l ? data->last_start_min : data->pos_min;
697 data->offset_float_max = (l
698 ? data->last_start_max
699 : data->pos_min + data->pos_delta);
700 if (is_inf || (U32)data->offset_float_max > (U32)I32_MAX)
701 data->offset_float_max = I32_MAX;
702 if (data->flags & SF_BEFORE_EOL)
704 |= ((data->flags & SF_BEFORE_EOL) << SF_FL_SHIFT_EOL);
706 data->flags &= ~SF_FL_BEFORE_EOL;
707 data->minlen_float=minlenp;
708 data->lookbehind_float=0;
711 SvCUR_set(data->last_found, 0);
713 SV * const sv = data->last_found;
714 if (SvUTF8(sv) && SvMAGICAL(sv)) {
715 MAGIC * const mg = mg_find(sv, PERL_MAGIC_utf8);
721 data->flags &= ~SF_BEFORE_EOL;
722 DEBUG_STUDYDATA("commit: ",data,0);
725 /* Can match anything (initialization) */
727 S_cl_anything(const RExC_state_t *pRExC_state, struct regnode_charclass_class *cl)
729 PERL_ARGS_ASSERT_CL_ANYTHING;
731 ANYOF_BITMAP_SETALL(cl);
732 cl->flags = ANYOF_CLASS|ANYOF_EOS|ANYOF_UNICODE_ALL
733 |ANYOF_LOC_NONBITMAP_FOLD|ANYOF_NON_UTF8_LATIN1_ALL
734 /* Even though no bitmap is in use here, we need to set
735 * the flag below so an AND with a node that does have one
736 * doesn't lose that one. The flag should get cleared if
737 * the other one doesn't; and the code in regexec.c is
738 * structured so this being set when not needed does no
739 * harm. It seemed a little cleaner to set it here than do
740 * a special case in cl_and() */
741 |ANYOF_NONBITMAP_NON_UTF8;
743 /* If any portion of the regex is to operate under locale rules,
744 * initialization includes it. The reason this isn't done for all regexes
745 * is that the optimizer was written under the assumption that locale was
746 * all-or-nothing. Given the complexity and lack of documentation in the
747 * optimizer, and that there are inadequate test cases for locale, so many
748 * parts of it may not work properly, it is safest to avoid locale unless
750 if (RExC_contains_locale) {
751 ANYOF_CLASS_SETALL(cl); /* /l uses class */
752 cl->flags |= ANYOF_LOCALE;
755 ANYOF_CLASS_ZERO(cl); /* Only /l uses class now */
759 /* Can match anything (initialization) */
761 S_cl_is_anything(const struct regnode_charclass_class *cl)
765 PERL_ARGS_ASSERT_CL_IS_ANYTHING;
767 for (value = 0; value <= ANYOF_MAX; value += 2)
768 if (ANYOF_CLASS_TEST(cl, value) && ANYOF_CLASS_TEST(cl, value + 1))
770 if (!(cl->flags & ANYOF_UNICODE_ALL))
772 if (!ANYOF_BITMAP_TESTALLSET((const void*)cl))
777 /* Can match anything (initialization) */
779 S_cl_init(const RExC_state_t *pRExC_state, struct regnode_charclass_class *cl)
781 PERL_ARGS_ASSERT_CL_INIT;
783 Zero(cl, 1, struct regnode_charclass_class);
785 cl_anything(pRExC_state, cl);
786 ARG_SET(cl, ANYOF_NONBITMAP_EMPTY);
789 /* These two functions currently do the exact same thing */
790 #define cl_init_zero S_cl_init
792 /* 'AND' a given class with another one. Can create false positives. 'cl'
793 * should not be inverted. 'and_with->flags & ANYOF_CLASS' should be 0 if
794 * 'and_with' is a regnode_charclass instead of a regnode_charclass_class. */
796 S_cl_and(struct regnode_charclass_class *cl,
797 const struct regnode_charclass_class *and_with)
799 PERL_ARGS_ASSERT_CL_AND;
801 assert(and_with->type == ANYOF);
803 /* I (khw) am not sure all these restrictions are necessary XXX */
804 if (!(ANYOF_CLASS_TEST_ANY_SET(and_with))
805 && !(ANYOF_CLASS_TEST_ANY_SET(cl))
806 && (and_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
807 && !(and_with->flags & ANYOF_LOC_NONBITMAP_FOLD)
808 && !(cl->flags & ANYOF_LOC_NONBITMAP_FOLD)) {
811 if (and_with->flags & ANYOF_INVERT)
812 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
813 cl->bitmap[i] &= ~and_with->bitmap[i];
815 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
816 cl->bitmap[i] &= and_with->bitmap[i];
817 } /* XXXX: logic is complicated otherwise, leave it along for a moment. */
819 if (and_with->flags & ANYOF_INVERT) {
821 /* Here, the and'ed node is inverted. Get the AND of the flags that
822 * aren't affected by the inversion. Those that are affected are
823 * handled individually below */
824 U8 affected_flags = cl->flags & ~INVERSION_UNAFFECTED_FLAGS;
825 cl->flags &= (and_with->flags & INVERSION_UNAFFECTED_FLAGS);
826 cl->flags |= affected_flags;
828 /* We currently don't know how to deal with things that aren't in the
829 * bitmap, but we know that the intersection is no greater than what
830 * is already in cl, so let there be false positives that get sorted
831 * out after the synthetic start class succeeds, and the node is
832 * matched for real. */
834 /* The inversion of these two flags indicate that the resulting
835 * intersection doesn't have them */
836 if (and_with->flags & ANYOF_UNICODE_ALL) {
837 cl->flags &= ~ANYOF_UNICODE_ALL;
839 if (and_with->flags & ANYOF_NON_UTF8_LATIN1_ALL) {
840 cl->flags &= ~ANYOF_NON_UTF8_LATIN1_ALL;
843 else { /* and'd node is not inverted */
844 if (! ANYOF_NONBITMAP(and_with)) {
846 /* Here 'and_with' doesn't match anything outside the bitmap
847 * (except possibly ANYOF_UNICODE_ALL), which means the
848 * intersection can't either, except for ANYOF_UNICODE_ALL, in
849 * which case we don't know what the intersection is, but it's no
850 * greater than what cl already has, so can just leave it alone,
851 * with possible false positives */
852 if (! (and_with->flags & ANYOF_UNICODE_ALL)) {
853 ARG_SET(cl, ANYOF_NONBITMAP_EMPTY);
854 cl->flags &= ~ANYOF_NONBITMAP_NON_UTF8;
857 else if (! ANYOF_NONBITMAP(cl)) {
859 /* Here, 'and_with' does match something outside the bitmap, and cl
860 * doesn't have a list of things to match outside the bitmap. If
861 * cl can match all code points above 255, the intersection will
862 * be those above-255 code points that 'and_with' matches. There
863 * may be false positives from code points in 'and_with' that are
864 * outside the bitmap but below 256, but those get sorted out
865 * after the synthetic start class succeeds). If cl can't match
866 * all Unicode code points, it means here that it can't match *
867 * anything outside the bitmap, so we leave the bitmap empty */
868 if (cl->flags & ANYOF_UNICODE_ALL) {
869 ARG_SET(cl, ARG(and_with));
873 /* Here, both 'and_with' and cl match something outside the
874 * bitmap. Currently we do not do the intersection, so just match
875 * whatever cl had at the beginning. */
879 /* Take the intersection of the two sets of flags */
880 cl->flags &= and_with->flags;
884 /* 'OR' a given class with another one. Can create false positives. 'cl'
885 * should not be inverted. 'or_with->flags & ANYOF_CLASS' should be 0 if
886 * 'or_with' is a regnode_charclass instead of a regnode_charclass_class. */
888 S_cl_or(const RExC_state_t *pRExC_state, struct regnode_charclass_class *cl, const struct regnode_charclass_class *or_with)
890 PERL_ARGS_ASSERT_CL_OR;
892 if (or_with->flags & ANYOF_INVERT) {
894 /* Here, the or'd node is to be inverted. This means we take the
895 * complement of everything not in the bitmap, but currently we don't
896 * know what that is, so give up and match anything */
897 if (ANYOF_NONBITMAP(or_with)) {
898 cl_anything(pRExC_state, cl);
901 * (B1 | CL1) | (!B2 & !CL2) = (B1 | !B2 & !CL2) | (CL1 | (!B2 & !CL2))
902 * <= (B1 | !B2) | (CL1 | !CL2)
903 * which is wasteful if CL2 is small, but we ignore CL2:
904 * (B1 | CL1) | (!B2 & !CL2) <= (B1 | CL1) | !B2 = (B1 | !B2) | CL1
905 * XXXX Can we handle case-fold? Unclear:
906 * (OK1(i) | OK1(i')) | !(OK1(i) | OK1(i')) =
907 * (OK1(i) | OK1(i')) | (!OK1(i) & !OK1(i'))
909 else if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
910 && !(or_with->flags & ANYOF_LOC_NONBITMAP_FOLD)
911 && !(cl->flags & ANYOF_LOC_NONBITMAP_FOLD) ) {
914 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
915 cl->bitmap[i] |= ~or_with->bitmap[i];
916 } /* XXXX: logic is complicated otherwise */
918 cl_anything(pRExC_state, cl);
921 /* And, we can just take the union of the flags that aren't affected
922 * by the inversion */
923 cl->flags |= or_with->flags & INVERSION_UNAFFECTED_FLAGS;
925 /* For the remaining flags:
926 ANYOF_UNICODE_ALL and inverted means to not match anything above
927 255, which means that the union with cl should just be
928 what cl has in it, so can ignore this flag
929 ANYOF_NON_UTF8_LATIN1_ALL and inverted means if not utf8 and ord
930 is 127-255 to match them, but then invert that, so the
931 union with cl should just be what cl has in it, so can
934 } else { /* 'or_with' is not inverted */
935 /* (B1 | CL1) | (B2 | CL2) = (B1 | B2) | (CL1 | CL2)) */
936 if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
937 && (!(or_with->flags & ANYOF_LOC_NONBITMAP_FOLD)
938 || (cl->flags & ANYOF_LOC_NONBITMAP_FOLD)) ) {
941 /* OR char bitmap and class bitmap separately */
942 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
943 cl->bitmap[i] |= or_with->bitmap[i];
944 if (ANYOF_CLASS_TEST_ANY_SET(or_with)) {
945 for (i = 0; i < ANYOF_CLASSBITMAP_SIZE; i++)
946 cl->classflags[i] |= or_with->classflags[i];
947 cl->flags |= ANYOF_CLASS;
950 else { /* XXXX: logic is complicated, leave it along for a moment. */
951 cl_anything(pRExC_state, cl);
954 if (ANYOF_NONBITMAP(or_with)) {
956 /* Use the added node's outside-the-bit-map match if there isn't a
957 * conflict. If there is a conflict (both nodes match something
958 * outside the bitmap, but what they match outside is not the same
959 * pointer, and hence not easily compared until XXX we extend
960 * inversion lists this far), give up and allow the start class to
961 * match everything outside the bitmap. If that stuff is all above
962 * 255, can just set UNICODE_ALL, otherwise caould be anything. */
963 if (! ANYOF_NONBITMAP(cl)) {
964 ARG_SET(cl, ARG(or_with));
966 else if (ARG(cl) != ARG(or_with)) {
968 if ((or_with->flags & ANYOF_NONBITMAP_NON_UTF8)) {
969 cl_anything(pRExC_state, cl);
972 cl->flags |= ANYOF_UNICODE_ALL;
977 cl->flags |= or_with->flags;
982 #define TRIE_LIST_ITEM(state,idx) (trie->states[state].trans.list)[ idx ]
983 #define TRIE_LIST_CUR(state) ( TRIE_LIST_ITEM( state, 0 ).forid )
984 #define TRIE_LIST_LEN(state) ( TRIE_LIST_ITEM( state, 0 ).newstate )
985 #define TRIE_LIST_USED(idx) ( trie->states[state].trans.list ? (TRIE_LIST_CUR( idx ) - 1) : 0 )
990 dump_trie(trie,widecharmap,revcharmap)
991 dump_trie_interim_list(trie,widecharmap,revcharmap,next_alloc)
992 dump_trie_interim_table(trie,widecharmap,revcharmap,next_alloc)
994 These routines dump out a trie in a somewhat readable format.
995 The _interim_ variants are used for debugging the interim
996 tables that are used to generate the final compressed
997 representation which is what dump_trie expects.
999 Part of the reason for their existence is to provide a form
1000 of documentation as to how the different representations function.
1005 Dumps the final compressed table form of the trie to Perl_debug_log.
1006 Used for debugging make_trie().
1010 S_dump_trie(pTHX_ const struct _reg_trie_data *trie, HV *widecharmap,
1011 AV *revcharmap, U32 depth)
1014 SV *sv=sv_newmortal();
1015 int colwidth= widecharmap ? 6 : 4;
1017 GET_RE_DEBUG_FLAGS_DECL;
1019 PERL_ARGS_ASSERT_DUMP_TRIE;
1021 PerlIO_printf( Perl_debug_log, "%*sChar : %-6s%-6s%-4s ",
1022 (int)depth * 2 + 2,"",
1023 "Match","Base","Ofs" );
1025 for( state = 0 ; state < trie->uniquecharcount ; state++ ) {
1026 SV ** const tmp = av_fetch( revcharmap, state, 0);
1028 PerlIO_printf( Perl_debug_log, "%*s",
1030 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
1031 PL_colors[0], PL_colors[1],
1032 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
1033 PERL_PV_ESCAPE_FIRSTCHAR
1038 PerlIO_printf( Perl_debug_log, "\n%*sState|-----------------------",
1039 (int)depth * 2 + 2,"");
1041 for( state = 0 ; state < trie->uniquecharcount ; state++ )
1042 PerlIO_printf( Perl_debug_log, "%.*s", colwidth, "--------");
1043 PerlIO_printf( Perl_debug_log, "\n");
1045 for( state = 1 ; state < trie->statecount ; state++ ) {
1046 const U32 base = trie->states[ state ].trans.base;
1048 PerlIO_printf( Perl_debug_log, "%*s#%4"UVXf"|", (int)depth * 2 + 2,"", (UV)state);
1050 if ( trie->states[ state ].wordnum ) {
1051 PerlIO_printf( Perl_debug_log, " W%4X", trie->states[ state ].wordnum );
1053 PerlIO_printf( Perl_debug_log, "%6s", "" );
1056 PerlIO_printf( Perl_debug_log, " @%4"UVXf" ", (UV)base );
1061 while( ( base + ofs < trie->uniquecharcount ) ||
1062 ( base + ofs - trie->uniquecharcount < trie->lasttrans
1063 && trie->trans[ base + ofs - trie->uniquecharcount ].check != state))
1066 PerlIO_printf( Perl_debug_log, "+%2"UVXf"[ ", (UV)ofs);
1068 for ( ofs = 0 ; ofs < trie->uniquecharcount ; ofs++ ) {
1069 if ( ( base + ofs >= trie->uniquecharcount ) &&
1070 ( base + ofs - trie->uniquecharcount < trie->lasttrans ) &&
1071 trie->trans[ base + ofs - trie->uniquecharcount ].check == state )
1073 PerlIO_printf( Perl_debug_log, "%*"UVXf,
1075 (UV)trie->trans[ base + ofs - trie->uniquecharcount ].next );
1077 PerlIO_printf( Perl_debug_log, "%*s",colwidth," ." );
1081 PerlIO_printf( Perl_debug_log, "]");
1084 PerlIO_printf( Perl_debug_log, "\n" );
1086 PerlIO_printf(Perl_debug_log, "%*sword_info N:(prev,len)=", (int)depth*2, "");
1087 for (word=1; word <= trie->wordcount; word++) {
1088 PerlIO_printf(Perl_debug_log, " %d:(%d,%d)",
1089 (int)word, (int)(trie->wordinfo[word].prev),
1090 (int)(trie->wordinfo[word].len));
1092 PerlIO_printf(Perl_debug_log, "\n" );
1095 Dumps a fully constructed but uncompressed trie in list form.
1096 List tries normally only are used for construction when the number of
1097 possible chars (trie->uniquecharcount) is very high.
1098 Used for debugging make_trie().
1101 S_dump_trie_interim_list(pTHX_ const struct _reg_trie_data *trie,
1102 HV *widecharmap, AV *revcharmap, U32 next_alloc,
1106 SV *sv=sv_newmortal();
1107 int colwidth= widecharmap ? 6 : 4;
1108 GET_RE_DEBUG_FLAGS_DECL;
1110 PERL_ARGS_ASSERT_DUMP_TRIE_INTERIM_LIST;
1112 /* print out the table precompression. */
1113 PerlIO_printf( Perl_debug_log, "%*sState :Word | Transition Data\n%*s%s",
1114 (int)depth * 2 + 2,"", (int)depth * 2 + 2,"",
1115 "------:-----+-----------------\n" );
1117 for( state=1 ; state < next_alloc ; state ++ ) {
1120 PerlIO_printf( Perl_debug_log, "%*s %4"UVXf" :",
1121 (int)depth * 2 + 2,"", (UV)state );
1122 if ( ! trie->states[ state ].wordnum ) {
1123 PerlIO_printf( Perl_debug_log, "%5s| ","");
1125 PerlIO_printf( Perl_debug_log, "W%4x| ",
1126 trie->states[ state ].wordnum
1129 for( charid = 1 ; charid <= TRIE_LIST_USED( state ) ; charid++ ) {
1130 SV ** const tmp = av_fetch( revcharmap, TRIE_LIST_ITEM(state,charid).forid, 0);
1132 PerlIO_printf( Perl_debug_log, "%*s:%3X=%4"UVXf" | ",
1134 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
1135 PL_colors[0], PL_colors[1],
1136 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
1137 PERL_PV_ESCAPE_FIRSTCHAR
1139 TRIE_LIST_ITEM(state,charid).forid,
1140 (UV)TRIE_LIST_ITEM(state,charid).newstate
1143 PerlIO_printf(Perl_debug_log, "\n%*s| ",
1144 (int)((depth * 2) + 14), "");
1147 PerlIO_printf( Perl_debug_log, "\n");
1152 Dumps a fully constructed but uncompressed trie in table form.
1153 This is the normal DFA style state transition table, with a few
1154 twists to facilitate compression later.
1155 Used for debugging make_trie().
1158 S_dump_trie_interim_table(pTHX_ const struct _reg_trie_data *trie,
1159 HV *widecharmap, AV *revcharmap, U32 next_alloc,
1164 SV *sv=sv_newmortal();
1165 int colwidth= widecharmap ? 6 : 4;
1166 GET_RE_DEBUG_FLAGS_DECL;
1168 PERL_ARGS_ASSERT_DUMP_TRIE_INTERIM_TABLE;
1171 print out the table precompression so that we can do a visual check
1172 that they are identical.
1175 PerlIO_printf( Perl_debug_log, "%*sChar : ",(int)depth * 2 + 2,"" );
1177 for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) {
1178 SV ** const tmp = av_fetch( revcharmap, charid, 0);
1180 PerlIO_printf( Perl_debug_log, "%*s",
1182 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
1183 PL_colors[0], PL_colors[1],
1184 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
1185 PERL_PV_ESCAPE_FIRSTCHAR
1191 PerlIO_printf( Perl_debug_log, "\n%*sState+-",(int)depth * 2 + 2,"" );
1193 for( charid=0 ; charid < trie->uniquecharcount ; charid++ ) {
1194 PerlIO_printf( Perl_debug_log, "%.*s", colwidth,"--------");
1197 PerlIO_printf( Perl_debug_log, "\n" );
1199 for( state=1 ; state < next_alloc ; state += trie->uniquecharcount ) {
1201 PerlIO_printf( Perl_debug_log, "%*s%4"UVXf" : ",
1202 (int)depth * 2 + 2,"",
1203 (UV)TRIE_NODENUM( state ) );
1205 for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) {
1206 UV v=(UV)SAFE_TRIE_NODENUM( trie->trans[ state + charid ].next );
1208 PerlIO_printf( Perl_debug_log, "%*"UVXf, colwidth, v );
1210 PerlIO_printf( Perl_debug_log, "%*s", colwidth, "." );
1212 if ( ! trie->states[ TRIE_NODENUM( state ) ].wordnum ) {
1213 PerlIO_printf( Perl_debug_log, " (%4"UVXf")\n", (UV)trie->trans[ state ].check );
1215 PerlIO_printf( Perl_debug_log, " (%4"UVXf") W%4X\n", (UV)trie->trans[ state ].check,
1216 trie->states[ TRIE_NODENUM( state ) ].wordnum );
1224 /* make_trie(startbranch,first,last,tail,word_count,flags,depth)
1225 startbranch: the first branch in the whole branch sequence
1226 first : start branch of sequence of branch-exact nodes.
1227 May be the same as startbranch
1228 last : Thing following the last branch.
1229 May be the same as tail.
1230 tail : item following the branch sequence
1231 count : words in the sequence
1232 flags : currently the OP() type we will be building one of /EXACT(|F|Fl)/
1233 depth : indent depth
1235 Inplace optimizes a sequence of 2 or more Branch-Exact nodes into a TRIE node.
1237 A trie is an N'ary tree where the branches are determined by digital
1238 decomposition of the key. IE, at the root node you look up the 1st character and
1239 follow that branch repeat until you find the end of the branches. Nodes can be
1240 marked as "accepting" meaning they represent a complete word. Eg:
1244 would convert into the following structure. Numbers represent states, letters
1245 following numbers represent valid transitions on the letter from that state, if
1246 the number is in square brackets it represents an accepting state, otherwise it
1247 will be in parenthesis.
1249 +-h->+-e->[3]-+-r->(8)-+-s->[9]
1253 (1) +-i->(6)-+-s->[7]
1255 +-s->(3)-+-h->(4)-+-e->[5]
1257 Accept Word Mapping: 3=>1 (he),5=>2 (she), 7=>3 (his), 9=>4 (hers)
1259 This shows that when matching against the string 'hers' we will begin at state 1
1260 read 'h' and move to state 2, read 'e' and move to state 3 which is accepting,
1261 then read 'r' and go to state 8 followed by 's' which takes us to state 9 which
1262 is also accepting. Thus we know that we can match both 'he' and 'hers' with a
1263 single traverse. We store a mapping from accepting to state to which word was
1264 matched, and then when we have multiple possibilities we try to complete the
1265 rest of the regex in the order in which they occured in the alternation.
1267 The only prior NFA like behaviour that would be changed by the TRIE support is
1268 the silent ignoring of duplicate alternations which are of the form:
1270 / (DUPE|DUPE) X? (?{ ... }) Y /x
1272 Thus EVAL blocks following a trie may be called a different number of times with
1273 and without the optimisation. With the optimisations dupes will be silently
1274 ignored. This inconsistent behaviour of EVAL type nodes is well established as
1275 the following demonstrates:
1277 'words'=~/(word|word|word)(?{ print $1 })[xyz]/
1279 which prints out 'word' three times, but
1281 'words'=~/(word|word|word)(?{ print $1 })S/
1283 which doesnt print it out at all. This is due to other optimisations kicking in.
1285 Example of what happens on a structural level:
1287 The regexp /(ac|ad|ab)+/ will produce the following debug output:
1289 1: CURLYM[1] {1,32767}(18)
1300 This would be optimizable with startbranch=5, first=5, last=16, tail=16
1301 and should turn into:
1303 1: CURLYM[1] {1,32767}(18)
1305 [Words:3 Chars Stored:6 Unique Chars:4 States:5 NCP:1]
1313 Cases where tail != last would be like /(?foo|bar)baz/:
1323 which would be optimizable with startbranch=1, first=1, last=7, tail=8
1324 and would end up looking like:
1327 [Words:2 Chars Stored:6 Unique Chars:5 States:7 NCP:1]
1334 d = uvuni_to_utf8_flags(d, uv, 0);
1336 is the recommended Unicode-aware way of saying
1341 #define TRIE_STORE_REVCHAR \
1344 SV *zlopp = newSV(2); \
1345 unsigned char *flrbbbbb = (unsigned char *) SvPVX(zlopp); \
1346 unsigned const char *const kapow = uvuni_to_utf8(flrbbbbb, uvc & 0xFF); \
1347 SvCUR_set(zlopp, kapow - flrbbbbb); \
1350 av_push(revcharmap, zlopp); \
1352 char ooooff = (char)uvc; \
1353 av_push(revcharmap, newSVpvn(&ooooff, 1)); \
1357 #define TRIE_READ_CHAR STMT_START { \
1361 if ( foldlen > 0 ) { \
1362 uvc = utf8n_to_uvuni( scan, UTF8_MAXLEN, &len, uniflags ); \
1367 uvc = utf8n_to_uvuni( (const U8*)uc, UTF8_MAXLEN, &len, uniflags);\
1368 uvc = to_uni_fold( uvc, foldbuf, &foldlen ); \
1369 foldlen -= UNISKIP( uvc ); \
1370 scan = foldbuf + UNISKIP( uvc ); \
1373 uvc = utf8n_to_uvuni( (const U8*)uc, UTF8_MAXLEN, &len, uniflags);\
1383 #define TRIE_LIST_PUSH(state,fid,ns) STMT_START { \
1384 if ( TRIE_LIST_CUR( state ) >=TRIE_LIST_LEN( state ) ) { \
1385 U32 ging = TRIE_LIST_LEN( state ) *= 2; \
1386 Renew( trie->states[ state ].trans.list, ging, reg_trie_trans_le ); \
1388 TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).forid = fid; \
1389 TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).newstate = ns; \
1390 TRIE_LIST_CUR( state )++; \
1393 #define TRIE_LIST_NEW(state) STMT_START { \
1394 Newxz( trie->states[ state ].trans.list, \
1395 4, reg_trie_trans_le ); \
1396 TRIE_LIST_CUR( state ) = 1; \
1397 TRIE_LIST_LEN( state ) = 4; \
1400 #define TRIE_HANDLE_WORD(state) STMT_START { \
1401 U16 dupe= trie->states[ state ].wordnum; \
1402 regnode * const noper_next = regnext( noper ); \
1405 /* store the word for dumping */ \
1407 if (OP(noper) != NOTHING) \
1408 tmp = newSVpvn_utf8(STRING(noper), STR_LEN(noper), UTF); \
1410 tmp = newSVpvn_utf8( "", 0, UTF ); \
1411 av_push( trie_words, tmp ); \
1415 trie->wordinfo[curword].prev = 0; \
1416 trie->wordinfo[curword].len = wordlen; \
1417 trie->wordinfo[curword].accept = state; \
1419 if ( noper_next < tail ) { \
1421 trie->jump = (U16 *) PerlMemShared_calloc( word_count + 1, sizeof(U16) ); \
1422 trie->jump[curword] = (U16)(noper_next - convert); \
1424 jumper = noper_next; \
1426 nextbranch= regnext(cur); \
1430 /* It's a dupe. Pre-insert into the wordinfo[].prev */\
1431 /* chain, so that when the bits of chain are later */\
1432 /* linked together, the dups appear in the chain */\
1433 trie->wordinfo[curword].prev = trie->wordinfo[dupe].prev; \
1434 trie->wordinfo[dupe].prev = curword; \
1436 /* we haven't inserted this word yet. */ \
1437 trie->states[ state ].wordnum = curword; \
1442 #define TRIE_TRANS_STATE(state,base,ucharcount,charid,special) \
1443 ( ( base + charid >= ucharcount \
1444 && base + charid < ubound \
1445 && state == trie->trans[ base - ucharcount + charid ].check \
1446 && trie->trans[ base - ucharcount + charid ].next ) \
1447 ? trie->trans[ base - ucharcount + charid ].next \
1448 : ( state==1 ? special : 0 ) \
1452 #define MADE_JUMP_TRIE 2
1453 #define MADE_EXACT_TRIE 4
1456 S_make_trie(pTHX_ RExC_state_t *pRExC_state, regnode *startbranch, regnode *first, regnode *last, regnode *tail, U32 word_count, U32 flags, U32 depth)
1459 /* first pass, loop through and scan words */
1460 reg_trie_data *trie;
1461 HV *widecharmap = NULL;
1462 AV *revcharmap = newAV();
1464 const U32 uniflags = UTF8_ALLOW_DEFAULT;
1469 regnode *jumper = NULL;
1470 regnode *nextbranch = NULL;
1471 regnode *convert = NULL;
1472 U32 *prev_states; /* temp array mapping each state to previous one */
1473 /* we just use folder as a flag in utf8 */
1474 const U8 * folder = NULL;
1477 const U32 data_slot = add_data( pRExC_state, 4, "tuuu" );
1478 AV *trie_words = NULL;
1479 /* along with revcharmap, this only used during construction but both are
1480 * useful during debugging so we store them in the struct when debugging.
1483 const U32 data_slot = add_data( pRExC_state, 2, "tu" );
1484 STRLEN trie_charcount=0;
1486 SV *re_trie_maxbuff;
1487 GET_RE_DEBUG_FLAGS_DECL;
1489 PERL_ARGS_ASSERT_MAKE_TRIE;
1491 PERL_UNUSED_ARG(depth);
1496 case EXACTFU: folder = PL_fold_latin1; break;
1497 case EXACTF: folder = PL_fold; break;
1498 case EXACTFL: folder = PL_fold_locale; break;
1501 trie = (reg_trie_data *) PerlMemShared_calloc( 1, sizeof(reg_trie_data) );
1503 trie->startstate = 1;
1504 trie->wordcount = word_count;
1505 RExC_rxi->data->data[ data_slot ] = (void*)trie;
1506 trie->charmap = (U16 *) PerlMemShared_calloc( 256, sizeof(U16) );
1507 if (!(UTF && folder))
1508 trie->bitmap = (char *) PerlMemShared_calloc( ANYOF_BITMAP_SIZE, 1 );
1509 trie->wordinfo = (reg_trie_wordinfo *) PerlMemShared_calloc(
1510 trie->wordcount+1, sizeof(reg_trie_wordinfo));
1513 trie_words = newAV();
1516 re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, 1);
1517 if (!SvIOK(re_trie_maxbuff)) {
1518 sv_setiv(re_trie_maxbuff, RE_TRIE_MAXBUF_INIT);
1521 PerlIO_printf( Perl_debug_log,
1522 "%*smake_trie start==%d, first==%d, last==%d, tail==%d depth=%d\n",
1523 (int)depth * 2 + 2, "",
1524 REG_NODE_NUM(startbranch),REG_NODE_NUM(first),
1525 REG_NODE_NUM(last), REG_NODE_NUM(tail),
1529 /* Find the node we are going to overwrite */
1530 if ( first == startbranch && OP( last ) != BRANCH ) {
1531 /* whole branch chain */
1534 /* branch sub-chain */
1535 convert = NEXTOPER( first );
1538 /* -- First loop and Setup --
1540 We first traverse the branches and scan each word to determine if it
1541 contains widechars, and how many unique chars there are, this is
1542 important as we have to build a table with at least as many columns as we
1545 We use an array of integers to represent the character codes 0..255
1546 (trie->charmap) and we use a an HV* to store Unicode characters. We use the
1547 native representation of the character value as the key and IV's for the
1550 *TODO* If we keep track of how many times each character is used we can
1551 remap the columns so that the table compression later on is more
1552 efficient in terms of memory by ensuring the most common value is in the
1553 middle and the least common are on the outside. IMO this would be better
1554 than a most to least common mapping as theres a decent chance the most
1555 common letter will share a node with the least common, meaning the node
1556 will not be compressible. With a middle is most common approach the worst
1557 case is when we have the least common nodes twice.
1561 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
1562 regnode * const noper = NEXTOPER( cur );
1563 const U8 *uc = (U8*)STRING( noper );
1564 const U8 * const e = uc + STR_LEN( noper );
1566 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
1567 const U8 *scan = (U8*)NULL;
1568 U32 wordlen = 0; /* required init */
1570 bool set_bit = trie->bitmap ? 1 : 0; /*store the first char in the bitmap?*/
1572 if (OP(noper) == NOTHING) {
1576 if ( set_bit ) /* bitmap only alloced when !(UTF&&Folding) */
1577 TRIE_BITMAP_SET(trie,*uc); /* store the raw first byte
1578 regardless of encoding */
1580 for ( ; uc < e ; uc += len ) {
1581 TRIE_CHARCOUNT(trie)++;
1585 if ( !trie->charmap[ uvc ] ) {
1586 trie->charmap[ uvc ]=( ++trie->uniquecharcount );
1588 trie->charmap[ folder[ uvc ] ] = trie->charmap[ uvc ];
1592 /* store the codepoint in the bitmap, and its folded
1594 TRIE_BITMAP_SET(trie,uvc);
1596 /* store the folded codepoint */
1597 if ( folder ) TRIE_BITMAP_SET(trie,folder[ uvc ]);
1600 /* store first byte of utf8 representation of
1601 variant codepoints */
1602 if (! UNI_IS_INVARIANT(uvc)) {
1603 TRIE_BITMAP_SET(trie, UTF8_TWO_BYTE_HI(uvc));
1606 set_bit = 0; /* We've done our bit :-) */
1611 widecharmap = newHV();
1613 svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 1 );
1616 Perl_croak( aTHX_ "error creating/fetching widecharmap entry for 0x%"UVXf, uvc );
1618 if ( !SvTRUE( *svpp ) ) {
1619 sv_setiv( *svpp, ++trie->uniquecharcount );
1624 if( cur == first ) {
1627 } else if (chars < trie->minlen) {
1629 } else if (chars > trie->maxlen) {
1633 } /* end first pass */
1634 DEBUG_TRIE_COMPILE_r(
1635 PerlIO_printf( Perl_debug_log, "%*sTRIE(%s): W:%d C:%d Uq:%d Min:%d Max:%d\n",
1636 (int)depth * 2 + 2,"",
1637 ( widecharmap ? "UTF8" : "NATIVE" ), (int)word_count,
1638 (int)TRIE_CHARCOUNT(trie), trie->uniquecharcount,
1639 (int)trie->minlen, (int)trie->maxlen )
1643 We now know what we are dealing with in terms of unique chars and
1644 string sizes so we can calculate how much memory a naive
1645 representation using a flat table will take. If it's over a reasonable
1646 limit (as specified by ${^RE_TRIE_MAXBUF}) we use a more memory
1647 conservative but potentially much slower representation using an array
1650 At the end we convert both representations into the same compressed
1651 form that will be used in regexec.c for matching with. The latter
1652 is a form that cannot be used to construct with but has memory
1653 properties similar to the list form and access properties similar
1654 to the table form making it both suitable for fast searches and
1655 small enough that its feasable to store for the duration of a program.
1657 See the comment in the code where the compressed table is produced
1658 inplace from the flat tabe representation for an explanation of how
1659 the compression works.
1664 Newx(prev_states, TRIE_CHARCOUNT(trie) + 2, U32);
1667 if ( (IV)( ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount + 1) > SvIV(re_trie_maxbuff) ) {
1669 Second Pass -- Array Of Lists Representation
1671 Each state will be represented by a list of charid:state records
1672 (reg_trie_trans_le) the first such element holds the CUR and LEN
1673 points of the allocated array. (See defines above).
1675 We build the initial structure using the lists, and then convert
1676 it into the compressed table form which allows faster lookups
1677 (but cant be modified once converted).
1680 STRLEN transcount = 1;
1682 DEBUG_TRIE_COMPILE_MORE_r( PerlIO_printf( Perl_debug_log,
1683 "%*sCompiling trie using list compiler\n",
1684 (int)depth * 2 + 2, ""));
1686 trie->states = (reg_trie_state *)
1687 PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2,
1688 sizeof(reg_trie_state) );
1692 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
1694 regnode * const noper = NEXTOPER( cur );
1695 U8 *uc = (U8*)STRING( noper );
1696 const U8 * const e = uc + STR_LEN( noper );
1697 U32 state = 1; /* required init */
1698 U16 charid = 0; /* sanity init */
1699 U8 *scan = (U8*)NULL; /* sanity init */
1700 STRLEN foldlen = 0; /* required init */
1701 U32 wordlen = 0; /* required init */
1702 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
1704 if (OP(noper) != NOTHING) {
1705 for ( ; uc < e ; uc += len ) {
1710 charid = trie->charmap[ uvc ];
1712 SV** const svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 0);
1716 charid=(U16)SvIV( *svpp );
1719 /* charid is now 0 if we dont know the char read, or nonzero if we do */
1726 if ( !trie->states[ state ].trans.list ) {
1727 TRIE_LIST_NEW( state );
1729 for ( check = 1; check <= TRIE_LIST_USED( state ); check++ ) {
1730 if ( TRIE_LIST_ITEM( state, check ).forid == charid ) {
1731 newstate = TRIE_LIST_ITEM( state, check ).newstate;
1736 newstate = next_alloc++;
1737 prev_states[newstate] = state;
1738 TRIE_LIST_PUSH( state, charid, newstate );
1743 Perl_croak( aTHX_ "panic! In trie construction, no char mapping for %"IVdf, uvc );
1747 TRIE_HANDLE_WORD(state);
1749 } /* end second pass */
1751 /* next alloc is the NEXT state to be allocated */
1752 trie->statecount = next_alloc;
1753 trie->states = (reg_trie_state *)
1754 PerlMemShared_realloc( trie->states,
1756 * sizeof(reg_trie_state) );
1758 /* and now dump it out before we compress it */
1759 DEBUG_TRIE_COMPILE_MORE_r(dump_trie_interim_list(trie, widecharmap,
1760 revcharmap, next_alloc,
1764 trie->trans = (reg_trie_trans *)
1765 PerlMemShared_calloc( transcount, sizeof(reg_trie_trans) );
1772 for( state=1 ; state < next_alloc ; state ++ ) {
1776 DEBUG_TRIE_COMPILE_MORE_r(
1777 PerlIO_printf( Perl_debug_log, "tp: %d zp: %d ",tp,zp)
1781 if (trie->states[state].trans.list) {
1782 U16 minid=TRIE_LIST_ITEM( state, 1).forid;
1786 for( idx = 2 ; idx <= TRIE_LIST_USED( state ) ; idx++ ) {
1787 const U16 forid = TRIE_LIST_ITEM( state, idx).forid;
1788 if ( forid < minid ) {
1790 } else if ( forid > maxid ) {
1794 if ( transcount < tp + maxid - minid + 1) {
1796 trie->trans = (reg_trie_trans *)
1797 PerlMemShared_realloc( trie->trans,
1799 * sizeof(reg_trie_trans) );
1800 Zero( trie->trans + (transcount / 2), transcount / 2 , reg_trie_trans );
1802 base = trie->uniquecharcount + tp - minid;
1803 if ( maxid == minid ) {
1805 for ( ; zp < tp ; zp++ ) {
1806 if ( ! trie->trans[ zp ].next ) {
1807 base = trie->uniquecharcount + zp - minid;
1808 trie->trans[ zp ].next = TRIE_LIST_ITEM( state, 1).newstate;
1809 trie->trans[ zp ].check = state;
1815 trie->trans[ tp ].next = TRIE_LIST_ITEM( state, 1).newstate;
1816 trie->trans[ tp ].check = state;
1821 for ( idx=1; idx <= TRIE_LIST_USED( state ) ; idx++ ) {
1822 const U32 tid = base - trie->uniquecharcount + TRIE_LIST_ITEM( state, idx ).forid;
1823 trie->trans[ tid ].next = TRIE_LIST_ITEM( state, idx ).newstate;
1824 trie->trans[ tid ].check = state;
1826 tp += ( maxid - minid + 1 );
1828 Safefree(trie->states[ state ].trans.list);
1831 DEBUG_TRIE_COMPILE_MORE_r(
1832 PerlIO_printf( Perl_debug_log, " base: %d\n",base);
1835 trie->states[ state ].trans.base=base;
1837 trie->lasttrans = tp + 1;
1841 Second Pass -- Flat Table Representation.
1843 we dont use the 0 slot of either trans[] or states[] so we add 1 to each.
1844 We know that we will need Charcount+1 trans at most to store the data
1845 (one row per char at worst case) So we preallocate both structures
1846 assuming worst case.
1848 We then construct the trie using only the .next slots of the entry
1851 We use the .check field of the first entry of the node temporarily to
1852 make compression both faster and easier by keeping track of how many non
1853 zero fields are in the node.
1855 Since trans are numbered from 1 any 0 pointer in the table is a FAIL
1858 There are two terms at use here: state as a TRIE_NODEIDX() which is a
1859 number representing the first entry of the node, and state as a
1860 TRIE_NODENUM() which is the trans number. state 1 is TRIE_NODEIDX(1) and
1861 TRIE_NODENUM(1), state 2 is TRIE_NODEIDX(2) and TRIE_NODENUM(3) if there
1862 are 2 entrys per node. eg:
1870 The table is internally in the right hand, idx form. However as we also
1871 have to deal with the states array which is indexed by nodenum we have to
1872 use TRIE_NODENUM() to convert.
1875 DEBUG_TRIE_COMPILE_MORE_r( PerlIO_printf( Perl_debug_log,
1876 "%*sCompiling trie using table compiler\n",
1877 (int)depth * 2 + 2, ""));
1879 trie->trans = (reg_trie_trans *)
1880 PerlMemShared_calloc( ( TRIE_CHARCOUNT(trie) + 1 )
1881 * trie->uniquecharcount + 1,
1882 sizeof(reg_trie_trans) );
1883 trie->states = (reg_trie_state *)
1884 PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2,
1885 sizeof(reg_trie_state) );
1886 next_alloc = trie->uniquecharcount + 1;
1889 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
1891 regnode * const noper = NEXTOPER( cur );
1892 const U8 *uc = (U8*)STRING( noper );
1893 const U8 * const e = uc + STR_LEN( noper );
1895 U32 state = 1; /* required init */
1897 U16 charid = 0; /* sanity init */
1898 U32 accept_state = 0; /* sanity init */
1899 U8 *scan = (U8*)NULL; /* sanity init */
1901 STRLEN foldlen = 0; /* required init */
1902 U32 wordlen = 0; /* required init */
1903 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
1905 if ( OP(noper) != NOTHING ) {
1906 for ( ; uc < e ; uc += len ) {
1911 charid = trie->charmap[ uvc ];
1913 SV* const * const svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 0);
1914 charid = svpp ? (U16)SvIV(*svpp) : 0;
1918 if ( !trie->trans[ state + charid ].next ) {
1919 trie->trans[ state + charid ].next = next_alloc;
1920 trie->trans[ state ].check++;
1921 prev_states[TRIE_NODENUM(next_alloc)]
1922 = TRIE_NODENUM(state);
1923 next_alloc += trie->uniquecharcount;
1925 state = trie->trans[ state + charid ].next;
1927 Perl_croak( aTHX_ "panic! In trie construction, no char mapping for %"IVdf, uvc );
1929 /* charid is now 0 if we dont know the char read, or nonzero if we do */
1932 accept_state = TRIE_NODENUM( state );
1933 TRIE_HANDLE_WORD(accept_state);
1935 } /* end second pass */
1937 /* and now dump it out before we compress it */
1938 DEBUG_TRIE_COMPILE_MORE_r(dump_trie_interim_table(trie, widecharmap,
1940 next_alloc, depth+1));
1944 * Inplace compress the table.*
1946 For sparse data sets the table constructed by the trie algorithm will
1947 be mostly 0/FAIL transitions or to put it another way mostly empty.
1948 (Note that leaf nodes will not contain any transitions.)
1950 This algorithm compresses the tables by eliminating most such
1951 transitions, at the cost of a modest bit of extra work during lookup:
1953 - Each states[] entry contains a .base field which indicates the
1954 index in the state[] array wheres its transition data is stored.
1956 - If .base is 0 there are no valid transitions from that node.
1958 - If .base is nonzero then charid is added to it to find an entry in
1961 -If trans[states[state].base+charid].check!=state then the
1962 transition is taken to be a 0/Fail transition. Thus if there are fail
1963 transitions at the front of the node then the .base offset will point
1964 somewhere inside the previous nodes data (or maybe even into a node
1965 even earlier), but the .check field determines if the transition is
1969 The following process inplace converts the table to the compressed
1970 table: We first do not compress the root node 1,and mark all its
1971 .check pointers as 1 and set its .base pointer as 1 as well. This
1972 allows us to do a DFA construction from the compressed table later,
1973 and ensures that any .base pointers we calculate later are greater
1976 - We set 'pos' to indicate the first entry of the second node.
1978 - We then iterate over the columns of the node, finding the first and
1979 last used entry at l and m. We then copy l..m into pos..(pos+m-l),
1980 and set the .check pointers accordingly, and advance pos
1981 appropriately and repreat for the next node. Note that when we copy
1982 the next pointers we have to convert them from the original
1983 NODEIDX form to NODENUM form as the former is not valid post
1986 - If a node has no transitions used we mark its base as 0 and do not
1987 advance the pos pointer.
1989 - If a node only has one transition we use a second pointer into the
1990 structure to fill in allocated fail transitions from other states.
1991 This pointer is independent of the main pointer and scans forward
1992 looking for null transitions that are allocated to a state. When it
1993 finds one it writes the single transition into the "hole". If the
1994 pointer doesnt find one the single transition is appended as normal.
1996 - Once compressed we can Renew/realloc the structures to release the
1999 See "Table-Compression Methods" in sec 3.9 of the Red Dragon,
2000 specifically Fig 3.47 and the associated pseudocode.
2004 const U32 laststate = TRIE_NODENUM( next_alloc );
2007 trie->statecount = laststate;
2009 for ( state = 1 ; state < laststate ; state++ ) {
2011 const U32 stateidx = TRIE_NODEIDX( state );
2012 const U32 o_used = trie->trans[ stateidx ].check;
2013 U32 used = trie->trans[ stateidx ].check;
2014 trie->trans[ stateidx ].check = 0;
2016 for ( charid = 0 ; used && charid < trie->uniquecharcount ; charid++ ) {
2017 if ( flag || trie->trans[ stateidx + charid ].next ) {
2018 if ( trie->trans[ stateidx + charid ].next ) {
2020 for ( ; zp < pos ; zp++ ) {
2021 if ( ! trie->trans[ zp ].next ) {
2025 trie->states[ state ].trans.base = zp + trie->uniquecharcount - charid ;
2026 trie->trans[ zp ].next = SAFE_TRIE_NODENUM( trie->trans[ stateidx + charid ].next );
2027 trie->trans[ zp ].check = state;
2028 if ( ++zp > pos ) pos = zp;
2035 trie->states[ state ].trans.base = pos + trie->uniquecharcount - charid ;
2037 trie->trans[ pos ].next = SAFE_TRIE_NODENUM( trie->trans[ stateidx + charid ].next );
2038 trie->trans[ pos ].check = state;
2043 trie->lasttrans = pos + 1;
2044 trie->states = (reg_trie_state *)
2045 PerlMemShared_realloc( trie->states, laststate
2046 * sizeof(reg_trie_state) );
2047 DEBUG_TRIE_COMPILE_MORE_r(
2048 PerlIO_printf( Perl_debug_log,
2049 "%*sAlloc: %d Orig: %"IVdf" elements, Final:%"IVdf". Savings of %%%5.2f\n",
2050 (int)depth * 2 + 2,"",
2051 (int)( ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount + 1 ),
2054 ( ( next_alloc - pos ) * 100 ) / (double)next_alloc );
2057 } /* end table compress */
2059 DEBUG_TRIE_COMPILE_MORE_r(
2060 PerlIO_printf(Perl_debug_log, "%*sStatecount:%"UVxf" Lasttrans:%"UVxf"\n",
2061 (int)depth * 2 + 2, "",
2062 (UV)trie->statecount,
2063 (UV)trie->lasttrans)
2065 /* resize the trans array to remove unused space */
2066 trie->trans = (reg_trie_trans *)
2067 PerlMemShared_realloc( trie->trans, trie->lasttrans
2068 * sizeof(reg_trie_trans) );
2070 { /* Modify the program and insert the new TRIE node */
2071 U8 nodetype =(U8)(flags & 0xFF);
2075 regnode *optimize = NULL;
2076 #ifdef RE_TRACK_PATTERN_OFFSETS
2079 U32 mjd_nodelen = 0;
2080 #endif /* RE_TRACK_PATTERN_OFFSETS */
2081 #endif /* DEBUGGING */
2083 This means we convert either the first branch or the first Exact,
2084 depending on whether the thing following (in 'last') is a branch
2085 or not and whther first is the startbranch (ie is it a sub part of
2086 the alternation or is it the whole thing.)
2087 Assuming its a sub part we convert the EXACT otherwise we convert
2088 the whole branch sequence, including the first.
2090 /* Find the node we are going to overwrite */
2091 if ( first != startbranch || OP( last ) == BRANCH ) {
2092 /* branch sub-chain */
2093 NEXT_OFF( first ) = (U16)(last - first);
2094 #ifdef RE_TRACK_PATTERN_OFFSETS
2096 mjd_offset= Node_Offset((convert));
2097 mjd_nodelen= Node_Length((convert));
2100 /* whole branch chain */
2102 #ifdef RE_TRACK_PATTERN_OFFSETS
2105 const regnode *nop = NEXTOPER( convert );
2106 mjd_offset= Node_Offset((nop));
2107 mjd_nodelen= Node_Length((nop));
2111 PerlIO_printf(Perl_debug_log, "%*sMJD offset:%"UVuf" MJD length:%"UVuf"\n",
2112 (int)depth * 2 + 2, "",
2113 (UV)mjd_offset, (UV)mjd_nodelen)
2116 /* But first we check to see if there is a common prefix we can
2117 split out as an EXACT and put in front of the TRIE node. */
2118 trie->startstate= 1;
2119 if ( trie->bitmap && !widecharmap && !trie->jump ) {
2121 for ( state = 1 ; state < trie->statecount-1 ; state++ ) {
2125 const U32 base = trie->states[ state ].trans.base;
2127 if ( trie->states[state].wordnum )
2130 for ( ofs = 0 ; ofs < trie->uniquecharcount ; ofs++ ) {
2131 if ( ( base + ofs >= trie->uniquecharcount ) &&
2132 ( base + ofs - trie->uniquecharcount < trie->lasttrans ) &&
2133 trie->trans[ base + ofs - trie->uniquecharcount ].check == state )
2135 if ( ++count > 1 ) {
2136 SV **tmp = av_fetch( revcharmap, ofs, 0);
2137 const U8 *ch = (U8*)SvPV_nolen_const( *tmp );
2138 if ( state == 1 ) break;
2140 Zero(trie->bitmap, ANYOF_BITMAP_SIZE, char);
2142 PerlIO_printf(Perl_debug_log,
2143 "%*sNew Start State=%"UVuf" Class: [",
2144 (int)depth * 2 + 2, "",
2147 SV ** const tmp = av_fetch( revcharmap, idx, 0);
2148 const U8 * const ch = (U8*)SvPV_nolen_const( *tmp );
2150 TRIE_BITMAP_SET(trie,*ch);
2152 TRIE_BITMAP_SET(trie, folder[ *ch ]);
2154 PerlIO_printf(Perl_debug_log, "%s", (char*)ch)
2158 TRIE_BITMAP_SET(trie,*ch);
2160 TRIE_BITMAP_SET(trie,folder[ *ch ]);
2161 DEBUG_OPTIMISE_r(PerlIO_printf( Perl_debug_log,"%s", ch));
2167 SV **tmp = av_fetch( revcharmap, idx, 0);
2169 char *ch = SvPV( *tmp, len );
2171 SV *sv=sv_newmortal();
2172 PerlIO_printf( Perl_debug_log,
2173 "%*sPrefix State: %"UVuf" Idx:%"UVuf" Char='%s'\n",
2174 (int)depth * 2 + 2, "",
2176 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 6,
2177 PL_colors[0], PL_colors[1],
2178 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
2179 PERL_PV_ESCAPE_FIRSTCHAR
2184 OP( convert ) = nodetype;
2185 str=STRING(convert);
2188 STR_LEN(convert) += len;
2194 DEBUG_OPTIMISE_r(PerlIO_printf( Perl_debug_log,"]\n"));
2199 trie->prefixlen = (state-1);
2201 regnode *n = convert+NODE_SZ_STR(convert);
2202 NEXT_OFF(convert) = NODE_SZ_STR(convert);
2203 trie->startstate = state;
2204 trie->minlen -= (state - 1);
2205 trie->maxlen -= (state - 1);
2207 /* At least the UNICOS C compiler choked on this
2208 * being argument to DEBUG_r(), so let's just have
2211 #ifdef PERL_EXT_RE_BUILD
2217 regnode *fix = convert;
2218 U32 word = trie->wordcount;
2220 Set_Node_Offset_Length(convert, mjd_offset, state - 1);
2221 while( ++fix < n ) {
2222 Set_Node_Offset_Length(fix, 0, 0);
2225 SV ** const tmp = av_fetch( trie_words, word, 0 );
2227 if ( STR_LEN(convert) <= SvCUR(*tmp) )
2228 sv_chop(*tmp, SvPV_nolen(*tmp) + STR_LEN(convert));
2230 sv_chop(*tmp, SvPV_nolen(*tmp) + SvCUR(*tmp));
2238 NEXT_OFF(convert) = (U16)(tail - convert);
2239 DEBUG_r(optimize= n);
2245 if ( trie->maxlen ) {
2246 NEXT_OFF( convert ) = (U16)(tail - convert);
2247 ARG_SET( convert, data_slot );
2248 /* Store the offset to the first unabsorbed branch in
2249 jump[0], which is otherwise unused by the jump logic.
2250 We use this when dumping a trie and during optimisation. */
2252 trie->jump[0] = (U16)(nextbranch - convert);
2254 /* If the start state is not accepting (meaning there is no empty string/NOTHING)
2255 * and there is a bitmap
2256 * and the first "jump target" node we found leaves enough room
2257 * then convert the TRIE node into a TRIEC node, with the bitmap
2258 * embedded inline in the opcode - this is hypothetically faster.
2260 if ( !trie->states[trie->startstate].wordnum
2262 && ( (char *)jumper - (char *)convert) >= (int)sizeof(struct regnode_charclass) )
2264 OP( convert ) = TRIEC;
2265 Copy(trie->bitmap, ((struct regnode_charclass *)convert)->bitmap, ANYOF_BITMAP_SIZE, char);
2266 PerlMemShared_free(trie->bitmap);
2269 OP( convert ) = TRIE;
2271 /* store the type in the flags */
2272 convert->flags = nodetype;
2276 + regarglen[ OP( convert ) ];
2278 /* XXX We really should free up the resource in trie now,
2279 as we won't use them - (which resources?) dmq */
2281 /* needed for dumping*/
2282 DEBUG_r(if (optimize) {
2283 regnode *opt = convert;
2285 while ( ++opt < optimize) {
2286 Set_Node_Offset_Length(opt,0,0);
2289 Try to clean up some of the debris left after the
2292 while( optimize < jumper ) {
2293 mjd_nodelen += Node_Length((optimize));
2294 OP( optimize ) = OPTIMIZED;
2295 Set_Node_Offset_Length(optimize,0,0);
2298 Set_Node_Offset_Length(convert,mjd_offset,mjd_nodelen);
2300 } /* end node insert */
2302 /* Finish populating the prev field of the wordinfo array. Walk back
2303 * from each accept state until we find another accept state, and if
2304 * so, point the first word's .prev field at the second word. If the
2305 * second already has a .prev field set, stop now. This will be the
2306 * case either if we've already processed that word's accept state,
2307 * or that state had multiple words, and the overspill words were
2308 * already linked up earlier.
2315 for (word=1; word <= trie->wordcount; word++) {
2317 if (trie->wordinfo[word].prev)
2319 state = trie->wordinfo[word].accept;
2321 state = prev_states[state];
2324 prev = trie->states[state].wordnum;
2328 trie->wordinfo[word].prev = prev;
2330 Safefree(prev_states);
2334 /* and now dump out the compressed format */
2335 DEBUG_TRIE_COMPILE_r(dump_trie(trie, widecharmap, revcharmap, depth+1));
2337 RExC_rxi->data->data[ data_slot + 1 ] = (void*)widecharmap;
2339 RExC_rxi->data->data[ data_slot + TRIE_WORDS_OFFSET ] = (void*)trie_words;
2340 RExC_rxi->data->data[ data_slot + 3 ] = (void*)revcharmap;
2342 SvREFCNT_dec(revcharmap);
2346 : trie->startstate>1
2352 S_make_trie_failtable(pTHX_ RExC_state_t *pRExC_state, regnode *source, regnode *stclass, U32 depth)
2354 /* The Trie is constructed and compressed now so we can build a fail array if it's needed
2356 This is basically the Aho-Corasick algorithm. Its from exercise 3.31 and 3.32 in the
2357 "Red Dragon" -- Compilers, principles, techniques, and tools. Aho, Sethi, Ullman 1985/88
2360 We find the fail state for each state in the trie, this state is the longest proper
2361 suffix of the current state's 'word' that is also a proper prefix of another word in our
2362 trie. State 1 represents the word '' and is thus the default fail state. This allows
2363 the DFA not to have to restart after its tried and failed a word at a given point, it
2364 simply continues as though it had been matching the other word in the first place.
2366 'abcdgu'=~/abcdefg|cdgu/
2367 When we get to 'd' we are still matching the first word, we would encounter 'g' which would
2368 fail, which would bring us to the state representing 'd' in the second word where we would
2369 try 'g' and succeed, proceeding to match 'cdgu'.
2371 /* add a fail transition */
2372 const U32 trie_offset = ARG(source);
2373 reg_trie_data *trie=(reg_trie_data *)RExC_rxi->data->data[trie_offset];
2375 const U32 ucharcount = trie->uniquecharcount;
2376 const U32 numstates = trie->statecount;
2377 const U32 ubound = trie->lasttrans + ucharcount;
2381 U32 base = trie->states[ 1 ].trans.base;
2384 const U32 data_slot = add_data( pRExC_state, 1, "T" );
2385 GET_RE_DEBUG_FLAGS_DECL;
2387 PERL_ARGS_ASSERT_MAKE_TRIE_FAILTABLE;
2389 PERL_UNUSED_ARG(depth);
2393 ARG_SET( stclass, data_slot );
2394 aho = (reg_ac_data *) PerlMemShared_calloc( 1, sizeof(reg_ac_data) );
2395 RExC_rxi->data->data[ data_slot ] = (void*)aho;
2396 aho->trie=trie_offset;
2397 aho->states=(reg_trie_state *)PerlMemShared_malloc( numstates * sizeof(reg_trie_state) );
2398 Copy( trie->states, aho->states, numstates, reg_trie_state );
2399 Newxz( q, numstates, U32);
2400 aho->fail = (U32 *) PerlMemShared_calloc( numstates, sizeof(U32) );
2403 /* initialize fail[0..1] to be 1 so that we always have
2404 a valid final fail state */
2405 fail[ 0 ] = fail[ 1 ] = 1;
2407 for ( charid = 0; charid < ucharcount ; charid++ ) {
2408 const U32 newstate = TRIE_TRANS_STATE( 1, base, ucharcount, charid, 0 );
2410 q[ q_write ] = newstate;
2411 /* set to point at the root */
2412 fail[ q[ q_write++ ] ]=1;
2415 while ( q_read < q_write) {
2416 const U32 cur = q[ q_read++ % numstates ];
2417 base = trie->states[ cur ].trans.base;
2419 for ( charid = 0 ; charid < ucharcount ; charid++ ) {
2420 const U32 ch_state = TRIE_TRANS_STATE( cur, base, ucharcount, charid, 1 );
2422 U32 fail_state = cur;
2425 fail_state = fail[ fail_state ];
2426 fail_base = aho->states[ fail_state ].trans.base;
2427 } while ( !TRIE_TRANS_STATE( fail_state, fail_base, ucharcount, charid, 1 ) );
2429 fail_state = TRIE_TRANS_STATE( fail_state, fail_base, ucharcount, charid, 1 );
2430 fail[ ch_state ] = fail_state;
2431 if ( !aho->states[ ch_state ].wordnum && aho->states[ fail_state ].wordnum )
2433 aho->states[ ch_state ].wordnum = aho->states[ fail_state ].wordnum;
2435 q[ q_write++ % numstates] = ch_state;
2439 /* restore fail[0..1] to 0 so that we "fall out" of the AC loop
2440 when we fail in state 1, this allows us to use the
2441 charclass scan to find a valid start char. This is based on the principle
2442 that theres a good chance the string being searched contains lots of stuff
2443 that cant be a start char.
2445 fail[ 0 ] = fail[ 1 ] = 0;
2446 DEBUG_TRIE_COMPILE_r({
2447 PerlIO_printf(Perl_debug_log,
2448 "%*sStclass Failtable (%"UVuf" states): 0",
2449 (int)(depth * 2), "", (UV)numstates
2451 for( q_read=1; q_read<numstates; q_read++ ) {
2452 PerlIO_printf(Perl_debug_log, ", %"UVuf, (UV)fail[q_read]);
2454 PerlIO_printf(Perl_debug_log, "\n");
2457 /*RExC_seen |= REG_SEEN_TRIEDFA;*/
2462 * There are strange code-generation bugs caused on sparc64 by gcc-2.95.2.
2463 * These need to be revisited when a newer toolchain becomes available.
2465 #if defined(__sparc64__) && defined(__GNUC__)
2466 # if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 96)
2467 # undef SPARC64_GCC_WORKAROUND
2468 # define SPARC64_GCC_WORKAROUND 1
2472 #define DEBUG_PEEP(str,scan,depth) \
2473 DEBUG_OPTIMISE_r({if (scan){ \
2474 SV * const mysv=sv_newmortal(); \
2475 regnode *Next = regnext(scan); \
2476 regprop(RExC_rx, mysv, scan); \
2477 PerlIO_printf(Perl_debug_log, "%*s" str ">%3d: %s (%d)\n", \
2478 (int)depth*2, "", REG_NODE_NUM(scan), SvPV_nolen_const(mysv),\
2479 Next ? (REG_NODE_NUM(Next)) : 0 ); \
2486 #define JOIN_EXACT(scan,min,flags) \
2487 if (PL_regkind[OP(scan)] == EXACT) \
2488 join_exact(pRExC_state,(scan),(min),(flags),NULL,depth+1)
2491 S_join_exact(pTHX_ RExC_state_t *pRExC_state, regnode *scan, I32 *min, U32 flags,regnode *val, U32 depth) {
2492 /* Merge several consecutive EXACTish nodes into one. */
2493 regnode *n = regnext(scan);
2495 regnode *next = scan + NODE_SZ_STR(scan);
2499 regnode *stop = scan;
2500 GET_RE_DEBUG_FLAGS_DECL;
2502 PERL_UNUSED_ARG(depth);
2505 PERL_ARGS_ASSERT_JOIN_EXACT;
2506 #ifndef EXPERIMENTAL_INPLACESCAN
2507 PERL_UNUSED_ARG(flags);
2508 PERL_UNUSED_ARG(val);
2510 DEBUG_PEEP("join",scan,depth);
2512 /* Skip NOTHING, merge EXACT*. */
2514 ( PL_regkind[OP(n)] == NOTHING ||
2515 (stringok && (OP(n) == OP(scan))))
2517 && NEXT_OFF(scan) + NEXT_OFF(n) < I16_MAX) {
2519 if (OP(n) == TAIL || n > next)
2521 if (PL_regkind[OP(n)] == NOTHING) {
2522 DEBUG_PEEP("skip:",n,depth);
2523 NEXT_OFF(scan) += NEXT_OFF(n);
2524 next = n + NODE_STEP_REGNODE;
2531 else if (stringok) {
2532 const unsigned int oldl = STR_LEN(scan);
2533 regnode * const nnext = regnext(n);
2535 DEBUG_PEEP("merg",n,depth);
2538 if (oldl + STR_LEN(n) > U8_MAX)
2540 NEXT_OFF(scan) += NEXT_OFF(n);
2541 STR_LEN(scan) += STR_LEN(n);
2542 next = n + NODE_SZ_STR(n);
2543 /* Now we can overwrite *n : */
2544 Move(STRING(n), STRING(scan) + oldl, STR_LEN(n), char);
2552 #ifdef EXPERIMENTAL_INPLACESCAN
2553 if (flags && !NEXT_OFF(n)) {
2554 DEBUG_PEEP("atch", val, depth);
2555 if (reg_off_by_arg[OP(n)]) {
2556 ARG_SET(n, val - n);
2559 NEXT_OFF(n) = val - n;
2565 #define GREEK_SMALL_LETTER_IOTA_WITH_DIALYTIKA_AND_TONOS 0x0390
2566 #define IOTA_D_T GREEK_SMALL_LETTER_IOTA_WITH_DIALYTIKA_AND_TONOS
2567 #define GREEK_SMALL_LETTER_UPSILON_WITH_DIALYTIKA_AND_TONOS 0x03B0
2568 #define UPSILON_D_T GREEK_SMALL_LETTER_UPSILON_WITH_DIALYTIKA_AND_TONOS
2571 && ( OP(scan) == EXACTF || OP(scan) == EXACTFU || OP(scan) == EXACTFA)
2572 && ( STR_LEN(scan) >= 6 ) )
2575 Two problematic code points in Unicode casefolding of EXACT nodes:
2577 U+0390 - GREEK SMALL LETTER IOTA WITH DIALYTIKA AND TONOS
2578 U+03B0 - GREEK SMALL LETTER UPSILON WITH DIALYTIKA AND TONOS
2584 U+03B9 U+0308 U+0301 0xCE 0xB9 0xCC 0x88 0xCC 0x81
2585 U+03C5 U+0308 U+0301 0xCF 0x85 0xCC 0x88 0xCC 0x81
2587 This means that in case-insensitive matching (or "loose matching",
2588 as Unicode calls it), an EXACTF of length six (the UTF-8 encoded byte
2589 length of the above casefolded versions) can match a target string
2590 of length two (the byte length of UTF-8 encoded U+0390 or U+03B0).
2591 This would rather mess up the minimum length computation.
2593 What we'll do is to look for the tail four bytes, and then peek
2594 at the preceding two bytes to see whether we need to decrease
2595 the minimum length by four (six minus two).
2597 Thanks to the design of UTF-8, there cannot be false matches:
2598 A sequence of valid UTF-8 bytes cannot be a subsequence of
2599 another valid sequence of UTF-8 bytes.
2602 char * const s0 = STRING(scan), *s, *t;
2603 char * const s1 = s0 + STR_LEN(scan) - 1;
2604 char * const s2 = s1 - 4;
2605 #ifdef EBCDIC /* RD tunifold greek 0390 and 03B0 */
2606 const char t0[] = "\xaf\x49\xaf\x42";
2608 const char t0[] = "\xcc\x88\xcc\x81";
2610 const char * const t1 = t0 + 3;
2613 s < s2 && (t = ninstr(s, s1, t0, t1));
2616 if (((U8)t[-1] == 0x68 && (U8)t[-2] == 0xB4) ||
2617 ((U8)t[-1] == 0x46 && (U8)t[-2] == 0xB5))
2619 if (((U8)t[-1] == 0xB9 && (U8)t[-2] == 0xCE) ||
2620 ((U8)t[-1] == 0x85 && (U8)t[-2] == 0xCF))
2628 n = scan + NODE_SZ_STR(scan);
2630 if (PL_regkind[OP(n)] != NOTHING || OP(n) == NOTHING) {
2637 DEBUG_OPTIMISE_r(if (merged){DEBUG_PEEP("finl",scan,depth)});
2641 /* REx optimizer. Converts nodes into quicker variants "in place".
2642 Finds fixed substrings. */
2644 /* Stops at toplevel WHILEM as well as at "last". At end *scanp is set
2645 to the position after last scanned or to NULL. */
2647 #define INIT_AND_WITHP \
2648 assert(!and_withp); \
2649 Newx(and_withp,1,struct regnode_charclass_class); \
2650 SAVEFREEPV(and_withp)
2652 /* this is a chain of data about sub patterns we are processing that
2653 need to be handled separately/specially in study_chunk. Its so
2654 we can simulate recursion without losing state. */
2656 typedef struct scan_frame {
2657 regnode *last; /* last node to process in this frame */
2658 regnode *next; /* next node to process when last is reached */
2659 struct scan_frame *prev; /*previous frame*/
2660 I32 stop; /* what stopparen do we use */
2664 #define SCAN_COMMIT(s, data, m) scan_commit(s, data, m, is_inf)
2666 #define CASE_SYNST_FNC(nAmE) \
2668 if (flags & SCF_DO_STCLASS_AND) { \
2669 for (value = 0; value < 256; value++) \
2670 if (!is_ ## nAmE ## _cp(value)) \
2671 ANYOF_BITMAP_CLEAR(data->start_class, value); \
2674 for (value = 0; value < 256; value++) \
2675 if (is_ ## nAmE ## _cp(value)) \
2676 ANYOF_BITMAP_SET(data->start_class, value); \
2680 if (flags & SCF_DO_STCLASS_AND) { \
2681 for (value = 0; value < 256; value++) \
2682 if (is_ ## nAmE ## _cp(value)) \
2683 ANYOF_BITMAP_CLEAR(data->start_class, value); \
2686 for (value = 0; value < 256; value++) \
2687 if (!is_ ## nAmE ## _cp(value)) \
2688 ANYOF_BITMAP_SET(data->start_class, value); \
2695 S_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp,
2696 I32 *minlenp, I32 *deltap,
2701 struct regnode_charclass_class *and_withp,
2702 U32 flags, U32 depth)
2703 /* scanp: Start here (read-write). */
2704 /* deltap: Write maxlen-minlen here. */
2705 /* last: Stop before this one. */
2706 /* data: string data about the pattern */
2707 /* stopparen: treat close N as END */
2708 /* recursed: which subroutines have we recursed into */
2709 /* and_withp: Valid if flags & SCF_DO_STCLASS_OR */
2712 I32 min = 0, pars = 0, code;
2713 regnode *scan = *scanp, *next;
2715 int is_inf = (flags & SCF_DO_SUBSTR) && (data->flags & SF_IS_INF);
2716 int is_inf_internal = 0; /* The studied chunk is infinite */
2717 I32 is_par = OP(scan) == OPEN ? ARG(scan) : 0;
2718 scan_data_t data_fake;
2719 SV *re_trie_maxbuff = NULL;
2720 regnode *first_non_open = scan;
2721 I32 stopmin = I32_MAX;
2722 scan_frame *frame = NULL;
2723 GET_RE_DEBUG_FLAGS_DECL;
2725 PERL_ARGS_ASSERT_STUDY_CHUNK;
2728 StructCopy(&zero_scan_data, &data_fake, scan_data_t);
2732 while (first_non_open && OP(first_non_open) == OPEN)
2733 first_non_open=regnext(first_non_open);
2738 while ( scan && OP(scan) != END && scan < last ){
2739 /* Peephole optimizer: */
2740 DEBUG_STUDYDATA("Peep:", data,depth);
2741 DEBUG_PEEP("Peep",scan,depth);
2742 JOIN_EXACT(scan,&min,0);
2744 /* Follow the next-chain of the current node and optimize
2745 away all the NOTHINGs from it. */
2746 if (OP(scan) != CURLYX) {
2747 const int max = (reg_off_by_arg[OP(scan)]
2749 /* I32 may be smaller than U16 on CRAYs! */
2750 : (I32_MAX < U16_MAX ? I32_MAX : U16_MAX));
2751 int off = (reg_off_by_arg[OP(scan)] ? ARG(scan) : NEXT_OFF(scan));
2755 /* Skip NOTHING and LONGJMP. */
2756 while ((n = regnext(n))
2757 && ((PL_regkind[OP(n)] == NOTHING && (noff = NEXT_OFF(n)))
2758 || ((OP(n) == LONGJMP) && (noff = ARG(n))))
2759 && off + noff < max)
2761 if (reg_off_by_arg[OP(scan)])
2764 NEXT_OFF(scan) = off;
2769 /* The principal pseudo-switch. Cannot be a switch, since we
2770 look into several different things. */
2771 if (OP(scan) == BRANCH || OP(scan) == BRANCHJ
2772 || OP(scan) == IFTHEN) {
2773 next = regnext(scan);
2775 /* demq: the op(next)==code check is to see if we have "branch-branch" AFAICT */
2777 if (OP(next) == code || code == IFTHEN) {
2778 /* NOTE - There is similar code to this block below for handling
2779 TRIE nodes on a re-study. If you change stuff here check there
2781 I32 max1 = 0, min1 = I32_MAX, num = 0;
2782 struct regnode_charclass_class accum;
2783 regnode * const startbranch=scan;
2785 if (flags & SCF_DO_SUBSTR)
2786 SCAN_COMMIT(pRExC_state, data, minlenp); /* Cannot merge strings after this. */
2787 if (flags & SCF_DO_STCLASS)
2788 cl_init_zero(pRExC_state, &accum);
2790 while (OP(scan) == code) {
2791 I32 deltanext, minnext, f = 0, fake;
2792 struct regnode_charclass_class this_class;
2795 data_fake.flags = 0;
2797 data_fake.whilem_c = data->whilem_c;
2798 data_fake.last_closep = data->last_closep;
2801 data_fake.last_closep = &fake;
2803 data_fake.pos_delta = delta;
2804 next = regnext(scan);
2805 scan = NEXTOPER(scan);
2807 scan = NEXTOPER(scan);
2808 if (flags & SCF_DO_STCLASS) {
2809 cl_init(pRExC_state, &this_class);
2810 data_fake.start_class = &this_class;
2811 f = SCF_DO_STCLASS_AND;
2813 if (flags & SCF_WHILEM_VISITED_POS)
2814 f |= SCF_WHILEM_VISITED_POS;
2816 /* we suppose the run is continuous, last=next...*/
2817 minnext = study_chunk(pRExC_state, &scan, minlenp, &deltanext,
2819 stopparen, recursed, NULL, f,depth+1);
2822 if (max1 < minnext + deltanext)
2823 max1 = minnext + deltanext;
2824 if (deltanext == I32_MAX)
2825 is_inf = is_inf_internal = 1;
2827 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
2829 if (data_fake.flags & SCF_SEEN_ACCEPT) {
2830 if ( stopmin > minnext)
2831 stopmin = min + min1;
2832 flags &= ~SCF_DO_SUBSTR;
2834 data->flags |= SCF_SEEN_ACCEPT;
2837 if (data_fake.flags & SF_HAS_EVAL)
2838 data->flags |= SF_HAS_EVAL;
2839 data->whilem_c = data_fake.whilem_c;
2841 if (flags & SCF_DO_STCLASS)
2842 cl_or(pRExC_state, &accum, &this_class);
2844 if (code == IFTHEN && num < 2) /* Empty ELSE branch */
2846 if (flags & SCF_DO_SUBSTR) {
2847 data->pos_min += min1;
2848 data->pos_delta += max1 - min1;
2849 if (max1 != min1 || is_inf)
2850 data->longest = &(data->longest_float);
2853 delta += max1 - min1;
2854 if (flags & SCF_DO_STCLASS_OR) {
2855 cl_or(pRExC_state, data->start_class, &accum);
2857 cl_and(data->start_class, and_withp);
2858 flags &= ~SCF_DO_STCLASS;
2861 else if (flags & SCF_DO_STCLASS_AND) {
2863 cl_and(data->start_class, &accum);
2864 flags &= ~SCF_DO_STCLASS;
2867 /* Switch to OR mode: cache the old value of
2868 * data->start_class */
2870 StructCopy(data->start_class, and_withp,
2871 struct regnode_charclass_class);
2872 flags &= ~SCF_DO_STCLASS_AND;
2873 StructCopy(&accum, data->start_class,
2874 struct regnode_charclass_class);
2875 flags |= SCF_DO_STCLASS_OR;
2876 data->start_class->flags |= ANYOF_EOS;
2880 if (PERL_ENABLE_TRIE_OPTIMISATION && OP( startbranch ) == BRANCH ) {
2883 Assuming this was/is a branch we are dealing with: 'scan' now
2884 points at the item that follows the branch sequence, whatever
2885 it is. We now start at the beginning of the sequence and look
2892 which would be constructed from a pattern like /A|LIST|OF|WORDS/
2894 If we can find such a subsequence we need to turn the first
2895 element into a trie and then add the subsequent branch exact
2896 strings to the trie.
2900 1. patterns where the whole set of branches can be converted.
2902 2. patterns where only a subset can be converted.
2904 In case 1 we can replace the whole set with a single regop
2905 for the trie. In case 2 we need to keep the start and end
2908 'BRANCH EXACT; BRANCH EXACT; BRANCH X'
2909 becomes BRANCH TRIE; BRANCH X;
2911 There is an additional case, that being where there is a
2912 common prefix, which gets split out into an EXACT like node
2913 preceding the TRIE node.
2915 If x(1..n)==tail then we can do a simple trie, if not we make
2916 a "jump" trie, such that when we match the appropriate word
2917 we "jump" to the appropriate tail node. Essentially we turn
2918 a nested if into a case structure of sorts.
2923 if (!re_trie_maxbuff) {
2924 re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, 1);
2925 if (!SvIOK(re_trie_maxbuff))
2926 sv_setiv(re_trie_maxbuff, RE_TRIE_MAXBUF_INIT);
2928 if ( SvIV(re_trie_maxbuff)>=0 ) {
2930 regnode *first = (regnode *)NULL;
2931 regnode *last = (regnode *)NULL;
2932 regnode *tail = scan;
2937 SV * const mysv = sv_newmortal(); /* for dumping */
2939 /* var tail is used because there may be a TAIL
2940 regop in the way. Ie, the exacts will point to the
2941 thing following the TAIL, but the last branch will
2942 point at the TAIL. So we advance tail. If we
2943 have nested (?:) we may have to move through several
2947 while ( OP( tail ) == TAIL ) {
2948 /* this is the TAIL generated by (?:) */
2949 tail = regnext( tail );
2954 regprop(RExC_rx, mysv, tail );
2955 PerlIO_printf( Perl_debug_log, "%*s%s%s\n",
2956 (int)depth * 2 + 2, "",
2957 "Looking for TRIE'able sequences. Tail node is: ",
2958 SvPV_nolen_const( mysv )
2964 step through the branches, cur represents each
2965 branch, noper is the first thing to be matched
2966 as part of that branch and noper_next is the
2967 regnext() of that node. if noper is an EXACT
2968 and noper_next is the same as scan (our current
2969 position in the regex) then the EXACT branch is
2970 a possible optimization target. Once we have
2971 two or more consecutive such branches we can
2972 create a trie of the EXACT's contents and stich
2973 it in place. If the sequence represents all of
2974 the branches we eliminate the whole thing and
2975 replace it with a single TRIE. If it is a
2976 subsequence then we need to stitch it in. This
2977 means the first branch has to remain, and needs
2978 to be repointed at the item on the branch chain
2979 following the last branch optimized. This could
2980 be either a BRANCH, in which case the
2981 subsequence is internal, or it could be the
2982 item following the branch sequence in which
2983 case the subsequence is at the end.
2987 /* dont use tail as the end marker for this traverse */
2988 for ( cur = startbranch ; cur != scan ; cur = regnext( cur ) ) {
2989 regnode * const noper = NEXTOPER( cur );
2990 #if defined(DEBUGGING) || defined(NOJUMPTRIE)
2991 regnode * const noper_next = regnext( noper );
2995 regprop(RExC_rx, mysv, cur);
2996 PerlIO_printf( Perl_debug_log, "%*s- %s (%d)",
2997 (int)depth * 2 + 2,"", SvPV_nolen_const( mysv ), REG_NODE_NUM(cur) );
2999 regprop(RExC_rx, mysv, noper);
3000 PerlIO_printf( Perl_debug_log, " -> %s",
3001 SvPV_nolen_const(mysv));
3004 regprop(RExC_rx, mysv, noper_next );
3005 PerlIO_printf( Perl_debug_log,"\t=> %s\t",
3006 SvPV_nolen_const(mysv));
3008 PerlIO_printf( Perl_debug_log, "(First==%d,Last==%d,Cur==%d)\n",
3009 REG_NODE_NUM(first), REG_NODE_NUM(last), REG_NODE_NUM(cur) );
3011 if ( (((first && optype!=NOTHING) ? OP( noper ) == optype
3012 : PL_regkind[ OP( noper ) ] == EXACT )
3013 || OP(noper) == NOTHING )
3015 && noper_next == tail
3020 if ( !first || optype == NOTHING ) {
3021 if (!first) first = cur;
3022 optype = OP( noper );
3028 Currently we do not believe that the trie logic can
3029 handle case insensitive matching properly when the
3030 pattern is not unicode (thus forcing unicode semantics).
3032 If/when this is fixed the following define can be swapped
3033 in below to fully enable trie logic.
3035 XXX It may work if not UTF and/or /a (AT_LEAST_UNI_SEMANTICS) but perhaps
3038 #define TRIE_TYPE_IS_SAFE 1
3041 #define TRIE_TYPE_IS_SAFE ((UTF && UNI_SEMANTICS) || optype==EXACT)
3043 if ( last && TRIE_TYPE_IS_SAFE ) {
3044 make_trie( pRExC_state,
3045 startbranch, first, cur, tail, count,
3048 if ( PL_regkind[ OP( noper ) ] == EXACT
3050 && noper_next == tail
3055 optype = OP( noper );
3065 regprop(RExC_rx, mysv, cur);
3066 PerlIO_printf( Perl_debug_log,
3067 "%*s- %s (%d) <SCAN FINISHED>\n", (int)depth * 2 + 2,
3068 "", SvPV_nolen_const( mysv ),REG_NODE_NUM(cur));
3072 if ( last && TRIE_TYPE_IS_SAFE ) {
3073 made= make_trie( pRExC_state, startbranch, first, scan, tail, count, optype, depth+1 );
3074 #ifdef TRIE_STUDY_OPT
3075 if ( ((made == MADE_EXACT_TRIE &&
3076 startbranch == first)
3077 || ( first_non_open == first )) &&
3079 flags |= SCF_TRIE_RESTUDY;
3080 if ( startbranch == first
3083 RExC_seen &=~REG_TOP_LEVEL_BRANCHES;
3093 else if ( code == BRANCHJ ) { /* single branch is optimized. */
3094 scan = NEXTOPER(NEXTOPER(scan));
3095 } else /* single branch is optimized. */
3096 scan = NEXTOPER(scan);
3098 } else if (OP(scan) == SUSPEND || OP(scan) == GOSUB || OP(scan) == GOSTART) {
3099 scan_frame *newframe = NULL;
3104 if (OP(scan) != SUSPEND) {
3105 /* set the pointer */
3106 if (OP(scan) == GOSUB) {
3108 RExC_recurse[ARG2L(scan)] = scan;
3109 start = RExC_open_parens[paren-1];
3110 end = RExC_close_parens[paren-1];
3113 start = RExC_rxi->program + 1;
3117 Newxz(recursed, (((RExC_npar)>>3) +1), U8);
3118 SAVEFREEPV(recursed);
3120 if (!PAREN_TEST(recursed,paren+1)) {
3121 PAREN_SET(recursed,paren+1);
3122 Newx(newframe,1,scan_frame);
3124 if (flags & SCF_DO_SUBSTR) {
3125 SCAN_COMMIT(pRExC_state,data,minlenp);
3126 data->longest = &(data->longest_float);
3128 is_inf = is_inf_internal = 1;
3129 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
3130 cl_anything(pRExC_state, data->start_class);
3131 flags &= ~SCF_DO_STCLASS;
3134 Newx(newframe,1,scan_frame);
3137 end = regnext(scan);
3142 SAVEFREEPV(newframe);
3143 newframe->next = regnext(scan);
3144 newframe->last = last;
3145 newframe->stop = stopparen;
3146 newframe->prev = frame;
3156 else if (OP(scan) == EXACT) {
3157 I32 l = STR_LEN(scan);
3160 const U8 * const s = (U8*)STRING(scan);
3161 l = utf8_length(s, s + l);
3162 uc = utf8_to_uvchr(s, NULL);
3164 uc = *((U8*)STRING(scan));
3167 if (flags & SCF_DO_SUBSTR) { /* Update longest substr. */
3168 /* The code below prefers earlier match for fixed
3169 offset, later match for variable offset. */
3170 if (data->last_end == -1) { /* Update the start info. */
3171 data->last_start_min = data->pos_min;
3172 data->last_start_max = is_inf
3173 ? I32_MAX : data->pos_min + data->pos_delta;
3175 sv_catpvn(data->last_found, STRING(scan), STR_LEN(scan));
3177 SvUTF8_on(data->last_found);
3179 SV * const sv = data->last_found;
3180 MAGIC * const mg = SvUTF8(sv) && SvMAGICAL(sv) ?
3181 mg_find(sv, PERL_MAGIC_utf8) : NULL;
3182 if (mg && mg->mg_len >= 0)
3183 mg->mg_len += utf8_length((U8*)STRING(scan),
3184 (U8*)STRING(scan)+STR_LEN(scan));
3186 data->last_end = data->pos_min + l;
3187 data->pos_min += l; /* As in the first entry. */
3188 data->flags &= ~SF_BEFORE_EOL;
3190 if (flags & SCF_DO_STCLASS_AND) {
3191 /* Check whether it is compatible with what we know already! */
3195 /* If compatible, we or it in below. It is compatible if is
3196 * in the bitmp and either 1) its bit or its fold is set, or 2)
3197 * it's for a locale. Even if there isn't unicode semantics
3198 * here, at runtime there may be because of matching against a
3199 * utf8 string, so accept a possible false positive for
3200 * latin1-range folds */
3202 (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
3203 && !ANYOF_BITMAP_TEST(data->start_class, uc)
3204 && (!(data->start_class->flags & ANYOF_LOC_NONBITMAP_FOLD)
3205 || !ANYOF_BITMAP_TEST(data->start_class, PL_fold_latin1[uc])))
3210 ANYOF_CLASS_ZERO(data->start_class);
3211 ANYOF_BITMAP_ZERO(data->start_class);
3213 ANYOF_BITMAP_SET(data->start_class, uc);
3214 else if (uc >= 0x100) {
3217 /* Some Unicode code points fold to the Latin1 range; as
3218 * XXX temporary code, instead of figuring out if this is
3219 * one, just assume it is and set all the start class bits
3220 * that could be some such above 255 code point's fold
3221 * which will generate fals positives. As the code
3222 * elsewhere that does compute the fold settles down, it
3223 * can be extracted out and re-used here */
3224 for (i = 0; i < 256; i++){
3225 if (_HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(i)) {
3226 ANYOF_BITMAP_SET(data->start_class, i);
3230 data->start_class->flags &= ~ANYOF_EOS;
3232 data->start_class->flags &= ~ANYOF_UNICODE_ALL;
3234 else if (flags & SCF_DO_STCLASS_OR) {
3235 /* false positive possible if the class is case-folded */
3237 ANYOF_BITMAP_SET(data->start_class, uc);
3239 data->start_class->flags |= ANYOF_UNICODE_ALL;
3240 data->start_class->flags &= ~ANYOF_EOS;
3241 cl_and(data->start_class, and_withp);
3243 flags &= ~SCF_DO_STCLASS;
3245 else if (PL_regkind[OP(scan)] == EXACT) { /* But OP != EXACT! */
3246 I32 l = STR_LEN(scan);
3247 UV uc = *((U8*)STRING(scan));
3249 /* Search for fixed substrings supports EXACT only. */
3250 if (flags & SCF_DO_SUBSTR) {
3252 SCAN_COMMIT(pRExC_state, data, minlenp);
3255 const U8 * const s = (U8 *)STRING(scan);
3256 l = utf8_length(s, s + l);
3257 uc = utf8_to_uvchr(s, NULL);
3260 if (flags & SCF_DO_SUBSTR)
3262 if (flags & SCF_DO_STCLASS_AND) {
3263 /* Check whether it is compatible with what we know already! */
3266 (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
3267 && !ANYOF_BITMAP_TEST(data->start_class, uc)
3268 && !ANYOF_BITMAP_TEST(data->start_class, PL_fold_latin1[uc])))
3272 ANYOF_CLASS_ZERO(data->start_class);
3273 ANYOF_BITMAP_ZERO(data->start_class);
3275 ANYOF_BITMAP_SET(data->start_class, uc);
3276 data->start_class->flags &= ~ANYOF_EOS;
3277 data->start_class->flags |= ANYOF_LOC_NONBITMAP_FOLD;
3278 if (OP(scan) == EXACTFL) {
3279 /* XXX This set is probably no longer necessary, and
3280 * probably wrong as LOCALE now is on in the initial
3282 data->start_class->flags |= ANYOF_LOCALE;
3286 /* Also set the other member of the fold pair. In case
3287 * that unicode semantics is called for at runtime, use
3288 * the full latin1 fold. (Can't do this for locale,
3289 * because not known until runtime */
3290 ANYOF_BITMAP_SET(data->start_class, PL_fold_latin1[uc]);
3293 else if (uc >= 0x100) {
3295 for (i = 0; i < 256; i++){
3296 if (_HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(i)) {
3297 ANYOF_BITMAP_SET(data->start_class, i);
3302 else if (flags & SCF_DO_STCLASS_OR) {
3303 if (data->start_class->flags & ANYOF_LOC_NONBITMAP_FOLD) {
3304 /* false positive possible if the class is case-folded.
3305 Assume that the locale settings are the same... */
3307 ANYOF_BITMAP_SET(data->start_class, uc);
3308 if (OP(scan) != EXACTFL) {
3310 /* And set the other member of the fold pair, but
3311 * can't do that in locale because not known until
3313 ANYOF_BITMAP_SET(data->start_class,
3314 PL_fold_latin1[uc]);
3317 data->start_class->flags &= ~ANYOF_EOS;
3319 cl_and(data->start_class, and_withp);
3321 flags &= ~SCF_DO_STCLASS;
3323 else if (REGNODE_VARIES(OP(scan))) {
3324 I32 mincount, maxcount, minnext, deltanext, fl = 0;
3325 I32 f = flags, pos_before = 0;
3326 regnode * const oscan = scan;
3327 struct regnode_charclass_class this_class;
3328 struct regnode_charclass_class *oclass = NULL;
3329 I32 next_is_eval = 0;
3331 switch (PL_regkind[OP(scan)]) {
3332 case WHILEM: /* End of (?:...)* . */
3333 scan = NEXTOPER(scan);
3336 if (flags & (SCF_DO_SUBSTR | SCF_DO_STCLASS)) {
3337 next = NEXTOPER(scan);
3338 if (OP(next) == EXACT || (flags & SCF_DO_STCLASS)) {
3340 maxcount = REG_INFTY;
3341 next = regnext(scan);
3342 scan = NEXTOPER(scan);
3346 if (flags & SCF_DO_SUBSTR)
3351 if (flags & SCF_DO_STCLASS) {
3353 maxcount = REG_INFTY;
3354 next = regnext(scan);
3355 scan = NEXTOPER(scan);
3358 is_inf = is_inf_internal = 1;
3359 scan = regnext(scan);
3360 if (flags & SCF_DO_SUBSTR) {
3361 SCAN_COMMIT(pRExC_state, data, minlenp); /* Cannot extend fixed substrings */
3362 data->longest = &(data->longest_float);
3364 goto optimize_curly_tail;
3366 if (stopparen>0 && (OP(scan)==CURLYN || OP(scan)==CURLYM)
3367 && (scan->flags == stopparen))
3372 mincount = ARG1(scan);
3373 maxcount = ARG2(scan);
3375 next = regnext(scan);
3376 if (OP(scan) == CURLYX) {
3377 I32 lp = (data ? *(data->last_closep) : 0);
3378 scan->flags = ((lp <= (I32)U8_MAX) ? (U8)lp : U8_MAX);
3380 scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3381 next_is_eval = (OP(scan) == EVAL);
3383 if (flags & SCF_DO_SUBSTR) {
3384 if (mincount == 0) SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot extend fixed substrings */
3385 pos_before = data->pos_min;
3389 data->flags &= ~(SF_HAS_PAR|SF_IN_PAR|SF_HAS_EVAL);
3391 data->flags |= SF_IS_INF;
3393 if (flags & SCF_DO_STCLASS) {
3394 cl_init(pRExC_state, &this_class);
3395 oclass = data->start_class;
3396 data->start_class = &this_class;
3397 f |= SCF_DO_STCLASS_AND;
3398 f &= ~SCF_DO_STCLASS_OR;
3400 /* Exclude from super-linear cache processing any {n,m}
3401 regops for which the combination of input pos and regex
3402 pos is not enough information to determine if a match
3405 For example, in the regex /foo(bar\s*){4,8}baz/ with the
3406 regex pos at the \s*, the prospects for a match depend not
3407 only on the input position but also on how many (bar\s*)
3408 repeats into the {4,8} we are. */
3409 if ((mincount > 1) || (maxcount > 1 && maxcount != REG_INFTY))
3410 f &= ~SCF_WHILEM_VISITED_POS;
3412 /* This will finish on WHILEM, setting scan, or on NULL: */
3413 minnext = study_chunk(pRExC_state, &scan, minlenp, &deltanext,
3414 last, data, stopparen, recursed, NULL,
3416 ? (f & ~SCF_DO_SUBSTR) : f),depth+1);
3418 if (flags & SCF_DO_STCLASS)
3419 data->start_class = oclass;
3420 if (mincount == 0 || minnext == 0) {
3421 if (flags & SCF_DO_STCLASS_OR) {
3422 cl_or(pRExC_state, data->start_class, &this_class);
3424 else if (flags & SCF_DO_STCLASS_AND) {
3425 /* Switch to OR mode: cache the old value of
3426 * data->start_class */
3428 StructCopy(data->start_class, and_withp,
3429 struct regnode_charclass_class);
3430 flags &= ~SCF_DO_STCLASS_AND;
3431 StructCopy(&this_class, data->start_class,
3432 struct regnode_charclass_class);
3433 flags |= SCF_DO_STCLASS_OR;
3434 data->start_class->flags |= ANYOF_EOS;
3436 } else { /* Non-zero len */
3437 if (flags & SCF_DO_STCLASS_OR) {
3438 cl_or(pRExC_state, data->start_class, &this_class);
3439 cl_and(data->start_class, and_withp);
3441 else if (flags & SCF_DO_STCLASS_AND)
3442 cl_and(data->start_class, &this_class);
3443 flags &= ~SCF_DO_STCLASS;
3445 if (!scan) /* It was not CURLYX, but CURLY. */
3447 if ( /* ? quantifier ok, except for (?{ ... }) */
3448 (next_is_eval || !(mincount == 0 && maxcount == 1))
3449 && (minnext == 0) && (deltanext == 0)
3450 && data && !(data->flags & (SF_HAS_PAR|SF_IN_PAR))
3451 && maxcount <= REG_INFTY/3) /* Complement check for big count */
3453 ckWARNreg(RExC_parse,
3454 "Quantifier unexpected on zero-length expression");
3457 min += minnext * mincount;
3458 is_inf_internal |= ((maxcount == REG_INFTY
3459 && (minnext + deltanext) > 0)
3460 || deltanext == I32_MAX);
3461 is_inf |= is_inf_internal;
3462 delta += (minnext + deltanext) * maxcount - minnext * mincount;
3464 /* Try powerful optimization CURLYX => CURLYN. */
3465 if ( OP(oscan) == CURLYX && data
3466 && data->flags & SF_IN_PAR
3467 && !(data->flags & SF_HAS_EVAL)
3468 && !deltanext && minnext == 1 ) {
3469 /* Try to optimize to CURLYN. */
3470 regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS;
3471 regnode * const nxt1 = nxt;
3478 if (!REGNODE_SIMPLE(OP(nxt))
3479 && !(PL_regkind[OP(nxt)] == EXACT
3480 && STR_LEN(nxt) == 1))
3486 if (OP(nxt) != CLOSE)
3488 if (RExC_open_parens) {
3489 RExC_open_parens[ARG(nxt1)-1]=oscan; /*open->CURLYM*/
3490 RExC_close_parens[ARG(nxt1)-1]=nxt+2; /*close->while*/
3492 /* Now we know that nxt2 is the only contents: */
3493 oscan->flags = (U8)ARG(nxt);
3495 OP(nxt1) = NOTHING; /* was OPEN. */
3498 OP(nxt1 + 1) = OPTIMIZED; /* was count. */
3499 NEXT_OFF(nxt1+ 1) = 0; /* just for consistency. */
3500 NEXT_OFF(nxt2) = 0; /* just for consistency with CURLY. */
3501 OP(nxt) = OPTIMIZED; /* was CLOSE. */
3502 OP(nxt + 1) = OPTIMIZED; /* was count. */
3503 NEXT_OFF(nxt+ 1) = 0; /* just for consistency. */
3508 /* Try optimization CURLYX => CURLYM. */
3509 if ( OP(oscan) == CURLYX && data
3510 && !(data->flags & SF_HAS_PAR)
3511 && !(data->flags & SF_HAS_EVAL)
3512 && !deltanext /* atom is fixed width */
3513 && minnext != 0 /* CURLYM can't handle zero width */
3515 /* XXXX How to optimize if data == 0? */
3516 /* Optimize to a simpler form. */
3517 regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN */
3521 while ( (nxt2 = regnext(nxt)) /* skip over embedded stuff*/
3522 && (OP(nxt2) != WHILEM))
3524 OP(nxt2) = SUCCEED; /* Whas WHILEM */
3525 /* Need to optimize away parenths. */
3526 if ((data->flags & SF_IN_PAR) && OP(nxt) == CLOSE) {
3527 /* Set the parenth number. */
3528 regnode *nxt1 = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN*/
3530 oscan->flags = (U8)ARG(nxt);
3531 if (RExC_open_parens) {
3532 RExC_open_parens[ARG(nxt1)-1]=oscan; /*open->CURLYM*/
3533 RExC_close_parens[ARG(nxt1)-1]=nxt2+1; /*close->NOTHING*/
3535 OP(nxt1) = OPTIMIZED; /* was OPEN. */
3536 OP(nxt) = OPTIMIZED; /* was CLOSE. */
3539 OP(nxt1 + 1) = OPTIMIZED; /* was count. */
3540 OP(nxt + 1) = OPTIMIZED; /* was count. */
3541 NEXT_OFF(nxt1 + 1) = 0; /* just for consistency. */
3542 NEXT_OFF(nxt + 1) = 0; /* just for consistency. */
3545 while ( nxt1 && (OP(nxt1) != WHILEM)) {
3546 regnode *nnxt = regnext(nxt1);
3548 if (reg_off_by_arg[OP(nxt1)])
3549 ARG_SET(nxt1, nxt2 - nxt1);
3550 else if (nxt2 - nxt1 < U16_MAX)
3551 NEXT_OFF(nxt1) = nxt2 - nxt1;
3553 OP(nxt) = NOTHING; /* Cannot beautify */
3558 /* Optimize again: */
3559 study_chunk(pRExC_state, &nxt1, minlenp, &deltanext, nxt,
3560 NULL, stopparen, recursed, NULL, 0,depth+1);
3565 else if ((OP(oscan) == CURLYX)
3566 && (flags & SCF_WHILEM_VISITED_POS)
3567 /* See the comment on a similar expression above.
3568 However, this time it's not a subexpression
3569 we care about, but the expression itself. */
3570 && (maxcount == REG_INFTY)
3571 && data && ++data->whilem_c < 16) {
3572 /* This stays as CURLYX, we can put the count/of pair. */
3573 /* Find WHILEM (as in regexec.c) */
3574 regnode *nxt = oscan + NEXT_OFF(oscan);
3576 if (OP(PREVOPER(nxt)) == NOTHING) /* LONGJMP */
3578 PREVOPER(nxt)->flags = (U8)(data->whilem_c
3579 | (RExC_whilem_seen << 4)); /* On WHILEM */
3581 if (data && fl & (SF_HAS_PAR|SF_IN_PAR))
3583 if (flags & SCF_DO_SUBSTR) {
3584 SV *last_str = NULL;
3585 int counted = mincount != 0;
3587 if (data->last_end > 0 && mincount != 0) { /* Ends with a string. */
3588 #if defined(SPARC64_GCC_WORKAROUND)
3591 const char *s = NULL;
3594 if (pos_before >= data->last_start_min)
3597 b = data->last_start_min;
3600 s = SvPV_const(data->last_found, l);
3601 old = b - data->last_start_min;
3604 I32 b = pos_before >= data->last_start_min
3605 ? pos_before : data->last_start_min;
3607 const char * const s = SvPV_const(data->last_found, l);
3608 I32 old = b - data->last_start_min;
3612 old = utf8_hop((U8*)s, old) - (U8*)s;
3614 /* Get the added string: */
3615 last_str = newSVpvn_utf8(s + old, l, UTF);
3616 if (deltanext == 0 && pos_before == b) {
3617 /* What was added is a constant string */
3619 SvGROW(last_str, (mincount * l) + 1);
3620 repeatcpy(SvPVX(last_str) + l,
3621 SvPVX_const(last_str), l, mincount - 1);
3622 SvCUR_set(last_str, SvCUR(last_str) * mincount);
3623 /* Add additional parts. */
3624 SvCUR_set(data->last_found,
3625 SvCUR(data->last_found) - l);
3626 sv_catsv(data->last_found, last_str);
3628 SV * sv = data->last_found;
3630 SvUTF8(sv) && SvMAGICAL(sv) ?
3631 mg_find(sv, PERL_MAGIC_utf8) : NULL;
3632 if (mg && mg->mg_len >= 0)
3633 mg->mg_len += CHR_SVLEN(last_str) - l;
3635 data->last_end += l * (mincount - 1);
3638 /* start offset must point into the last copy */
3639 data->last_start_min += minnext * (mincount - 1);
3640 data->last_start_max += is_inf ? I32_MAX
3641 : (maxcount - 1) * (minnext + data->pos_delta);
3644 /* It is counted once already... */
3645 data->pos_min += minnext * (mincount - counted);
3646 data->pos_delta += - counted * deltanext +
3647 (minnext + deltanext) * maxcount - minnext * mincount;
3648 if (mincount != maxcount) {
3649 /* Cannot extend fixed substrings found inside
3651 SCAN_COMMIT(pRExC_state,data,minlenp);
3652 if (mincount && last_str) {
3653 SV * const sv = data->last_found;
3654 MAGIC * const mg = SvUTF8(sv) && SvMAGICAL(sv) ?
3655 mg_find(sv, PERL_MAGIC_utf8) : NULL;
3659 sv_setsv(sv, last_str);
3660 data->last_end = data->pos_min;
3661 data->last_start_min =
3662 data->pos_min - CHR_SVLEN(last_str);
3663 data->last_start_max = is_inf
3665 : data->pos_min + data->pos_delta
3666 - CHR_SVLEN(last_str);
3668 data->longest = &(data->longest_float);
3670 SvREFCNT_dec(last_str);
3672 if (data && (fl & SF_HAS_EVAL))
3673 data->flags |= SF_HAS_EVAL;
3674 optimize_curly_tail:
3675 if (OP(oscan) != CURLYX) {
3676 while (PL_regkind[OP(next = regnext(oscan))] == NOTHING
3678 NEXT_OFF(oscan) += NEXT_OFF(next);
3681 default: /* REF, ANYOFV, and CLUMP only? */
3682 if (flags & SCF_DO_SUBSTR) {
3683 SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot expect anything... */
3684 data->longest = &(data->longest_float);
3686 is_inf = is_inf_internal = 1;
3687 if (flags & SCF_DO_STCLASS_OR)
3688 cl_anything(pRExC_state, data->start_class);
3689 flags &= ~SCF_DO_STCLASS;
3693 else if (OP(scan) == LNBREAK) {
3694 if (flags & SCF_DO_STCLASS) {
3696 data->start_class->flags &= ~ANYOF_EOS; /* No match on empty */
3697 if (flags & SCF_DO_STCLASS_AND) {
3698 for (value = 0; value < 256; value++)
3699 if (!is_VERTWS_cp(value))
3700 ANYOF_BITMAP_CLEAR(data->start_class, value);
3703 for (value = 0; value < 256; value++)
3704 if (is_VERTWS_cp(value))
3705 ANYOF_BITMAP_SET(data->start_class, value);
3707 if (flags & SCF_DO_STCLASS_OR)
3708 cl_and(data->start_class, and_withp);
3709 flags &= ~SCF_DO_STCLASS;
3713 if (flags & SCF_DO_SUBSTR) {
3714 SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot expect anything... */
3716 data->pos_delta += 1;
3717 data->longest = &(data->longest_float);
3720 else if (OP(scan) == FOLDCHAR) {
3721 int d = ARG(scan) == LATIN_SMALL_LETTER_SHARP_S ? 1 : 2;
3722 flags &= ~SCF_DO_STCLASS;
3725 if (flags & SCF_DO_SUBSTR) {
3726 SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot expect anything... */
3728 data->pos_delta += d;
3729 data->longest = &(data->longest_float);
3732 else if (REGNODE_SIMPLE(OP(scan))) {
3735 if (flags & SCF_DO_SUBSTR) {
3736 SCAN_COMMIT(pRExC_state,data,minlenp);
3740 if (flags & SCF_DO_STCLASS) {
3741 data->start_class->flags &= ~ANYOF_EOS; /* No match on empty */
3743 /* Some of the logic below assumes that switching
3744 locale on will only add false positives. */
3745 switch (PL_regkind[OP(scan)]) {
3749 /* Perl_croak(aTHX_ "panic: unexpected simple REx opcode %d", OP(scan)); */
3750 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
3751 cl_anything(pRExC_state, data->start_class);
3754 if (OP(scan) == SANY)
3756 if (flags & SCF_DO_STCLASS_OR) { /* Everything but \n */
3757 value = (ANYOF_BITMAP_TEST(data->start_class,'\n')
3758 || ANYOF_CLASS_TEST_ANY_SET(data->start_class));
3759 cl_anything(pRExC_state, data->start_class);
3761 if (flags & SCF_DO_STCLASS_AND || !value)
3762 ANYOF_BITMAP_CLEAR(data->start_class,'\n');
3765 if (flags & SCF_DO_STCLASS_AND)
3766 cl_and(data->start_class,
3767 (struct regnode_charclass_class*)scan);
3769 cl_or(pRExC_state, data->start_class,
3770 (struct regnode_charclass_class*)scan);
3773 if (flags & SCF_DO_STCLASS_AND) {
3774 if (!(data->start_class->flags & ANYOF_LOCALE)) {
3775 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM);
3776 if (OP(scan) == ALNUMU) {
3777 for (value = 0; value < 256; value++) {
3778 if (!isWORDCHAR_L1(value)) {
3779 ANYOF_BITMAP_CLEAR(data->start_class, value);
3783 for (value = 0; value < 256; value++) {
3784 if (!isALNUM(value)) {
3785 ANYOF_BITMAP_CLEAR(data->start_class, value);
3792 if (data->start_class->flags & ANYOF_LOCALE)
3793 ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM);
3795 /* Even if under locale, set the bits for non-locale
3796 * in case it isn't a true locale-node. This will
3797 * create false positives if it truly is locale */
3798 if (OP(scan) == ALNUMU) {
3799 for (value = 0; value < 256; value++) {
3800 if (isWORDCHAR_L1(value)) {
3801 ANYOF_BITMAP_SET(data->start_class, value);
3805 for (value = 0; value < 256; value++) {
3806 if (isALNUM(value)) {
3807 ANYOF_BITMAP_SET(data->start_class, value);
3814 if (flags & SCF_DO_STCLASS_AND) {
3815 if (!(data->start_class->flags & ANYOF_LOCALE)) {
3816 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM);
3817 if (OP(scan) == NALNUMU) {
3818 for (value = 0; value < 256; value++) {
3819 if (isWORDCHAR_L1(value)) {
3820 ANYOF_BITMAP_CLEAR(data->start_class, value);
3824 for (value = 0; value < 256; value++) {
3825 if (isALNUM(value)) {
3826 ANYOF_BITMAP_CLEAR(data->start_class, value);
3833 if (data->start_class->flags & ANYOF_LOCALE)
3834 ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM);
3836 /* Even if under locale, set the bits for non-locale in
3837 * case it isn't a true locale-node. This will create
3838 * false positives if it truly is locale */
3839 if (OP(scan) == NALNUMU) {
3840 for (value = 0; value < 256; value++) {
3841 if (! isWORDCHAR_L1(value)) {
3842 ANYOF_BITMAP_SET(data->start_class, value);
3846 for (value = 0; value < 256; value++) {
3847 if (! isALNUM(value)) {
3848 ANYOF_BITMAP_SET(data->start_class, value);
3855 if (flags & SCF_DO_STCLASS_AND) {
3856 if (!(data->start_class->flags & ANYOF_LOCALE)) {
3857 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE);
3858 if (OP(scan) == SPACEU) {
3859 for (value = 0; value < 256; value++) {
3860 if (!isSPACE_L1(value)) {
3861 ANYOF_BITMAP_CLEAR(data->start_class, value);
3865 for (value = 0; value < 256; value++) {
3866 if (!isSPACE(value)) {
3867 ANYOF_BITMAP_CLEAR(data->start_class, value);
3874 if (data->start_class->flags & ANYOF_LOCALE) {
3875 ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE);
3877 if (OP(scan) == SPACEU) {
3878 for (value = 0; value < 256; value++) {
3879 if (isSPACE_L1(value)) {
3880 ANYOF_BITMAP_SET(data->start_class, value);
3884 for (value = 0; value < 256; value++) {
3885 if (isSPACE(value)) {
3886 ANYOF_BITMAP_SET(data->start_class, value);
3893 if (flags & SCF_DO_STCLASS_AND) {
3894 if (!(data->start_class->flags & ANYOF_LOCALE)) {
3895 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE);
3896 if (OP(scan) == NSPACEU) {
3897 for (value = 0; value < 256; value++) {
3898 if (isSPACE_L1(value)) {
3899 ANYOF_BITMAP_CLEAR(data->start_class, value);
3903 for (value = 0; value < 256; value++) {
3904 if (isSPACE(value)) {
3905 ANYOF_BITMAP_CLEAR(data->start_class, value);
3912 if (data->start_class->flags & ANYOF_LOCALE)
3913 ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE);
3914 if (OP(scan) == NSPACEU) {
3915 for (value = 0; value < 256; value++) {
3916 if (!isSPACE_L1(value)) {
3917 ANYOF_BITMAP_SET(data->start_class, value);
3922 for (value = 0; value < 256; value++) {
3923 if (!isSPACE(value)) {
3924 ANYOF_BITMAP_SET(data->start_class, value);
3931 if (flags & SCF_DO_STCLASS_AND) {
3932 if (!(data->start_class->flags & ANYOF_LOCALE)) {
3933 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NDIGIT);
3934 for (value = 0; value < 256; value++)
3935 if (!isDIGIT(value))
3936 ANYOF_BITMAP_CLEAR(data->start_class, value);
3940 if (data->start_class->flags & ANYOF_LOCALE)
3941 ANYOF_CLASS_SET(data->start_class,ANYOF_DIGIT);
3942 for (value = 0; value < 256; value++)
3944 ANYOF_BITMAP_SET(data->start_class, value);
3948 if (flags & SCF_DO_STCLASS_AND) {
3949 if (!(data->start_class->flags & ANYOF_LOCALE))
3950 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_DIGIT);
3951 for (value = 0; value < 256; value++)
3953 ANYOF_BITMAP_CLEAR(data->start_class, value);
3956 if (data->start_class->flags & ANYOF_LOCALE)
3957 ANYOF_CLASS_SET(data->start_class,ANYOF_NDIGIT);
3958 for (value = 0; value < 256; value++)
3959 if (!isDIGIT(value))
3960 ANYOF_BITMAP_SET(data->start_class, value);
3963 CASE_SYNST_FNC(VERTWS);
3964 CASE_SYNST_FNC(HORIZWS);
3967 if (flags & SCF_DO_STCLASS_OR)
3968 cl_and(data->start_class, and_withp);
3969 flags &= ~SCF_DO_STCLASS;
3972 else if (PL_regkind[OP(scan)] == EOL && flags & SCF_DO_SUBSTR) {
3973 data->flags |= (OP(scan) == MEOL
3977 else if ( PL_regkind[OP(scan)] == BRANCHJ
3978 /* Lookbehind, or need to calculate parens/evals/stclass: */
3979 && (scan->flags || data || (flags & SCF_DO_STCLASS))
3980 && (OP(scan) == IFMATCH || OP(scan) == UNLESSM)) {
3981 if ( !PERL_ENABLE_POSITIVE_ASSERTION_STUDY
3982 || OP(scan) == UNLESSM )
3984 /* Negative Lookahead/lookbehind
3985 In this case we can't do fixed string optimisation.
3988 I32 deltanext, minnext, fake = 0;
3990 struct regnode_charclass_class intrnl;
3993 data_fake.flags = 0;
3995 data_fake.whilem_c = data->whilem_c;
3996 data_fake.last_closep = data->last_closep;
3999 data_fake.last_closep = &fake;
4000 data_fake.pos_delta = delta;
4001 if ( flags & SCF_DO_STCLASS && !scan->flags
4002 && OP(scan) == IFMATCH ) { /* Lookahead */
4003 cl_init(pRExC_state, &intrnl);
4004 data_fake.start_class = &intrnl;
4005 f |= SCF_DO_STCLASS_AND;
4007 if (flags & SCF_WHILEM_VISITED_POS)
4008 f |= SCF_WHILEM_VISITED_POS;
4009 next = regnext(scan);
4010 nscan = NEXTOPER(NEXTOPER(scan));
4011 minnext = study_chunk(pRExC_state, &nscan, minlenp, &deltanext,
4012 last, &data_fake, stopparen, recursed, NULL, f, depth+1);
4015 FAIL("Variable length lookbehind not implemented");
4017 else if (minnext > (I32)U8_MAX) {
4018 FAIL2("Lookbehind longer than %"UVuf" not implemented", (UV)U8_MAX);
4020 scan->flags = (U8)minnext;
4023 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
4025 if (data_fake.flags & SF_HAS_EVAL)
4026 data->flags |= SF_HAS_EVAL;
4027 data->whilem_c = data_fake.whilem_c;
4029 if (f & SCF_DO_STCLASS_AND) {
4030 if (flags & SCF_DO_STCLASS_OR) {
4031 /* OR before, AND after: ideally we would recurse with
4032 * data_fake to get the AND applied by study of the
4033 * remainder of the pattern, and then derecurse;
4034 * *** HACK *** for now just treat as "no information".
4035 * See [perl #56690].
4037 cl_init(pRExC_state, data->start_class);
4039 /* AND before and after: combine and continue */
4040 const int was = (data->start_class->flags & ANYOF_EOS);
4042 cl_and(data->start_class, &intrnl);
4044 data->start_class->flags |= ANYOF_EOS;
4048 #if PERL_ENABLE_POSITIVE_ASSERTION_STUDY
4050 /* Positive Lookahead/lookbehind
4051 In this case we can do fixed string optimisation,
4052 but we must be careful about it. Note in the case of
4053 lookbehind the positions will be offset by the minimum
4054 length of the pattern, something we won't know about
4055 until after the recurse.
4057 I32 deltanext, fake = 0;
4059 struct regnode_charclass_class intrnl;
4061 /* We use SAVEFREEPV so that when the full compile
4062 is finished perl will clean up the allocated
4063 minlens when it's all done. This way we don't
4064 have to worry about freeing them when we know
4065 they wont be used, which would be a pain.
4068 Newx( minnextp, 1, I32 );
4069 SAVEFREEPV(minnextp);
4072 StructCopy(data, &data_fake, scan_data_t);
4073 if ((flags & SCF_DO_SUBSTR) && data->last_found) {
4076 SCAN_COMMIT(pRExC_state, &data_fake,minlenp);
4077 data_fake.last_found=newSVsv(data->last_found);
4081 data_fake.last_closep = &fake;
4082 data_fake.flags = 0;
4083 data_fake.pos_delta = delta;
4085 data_fake.flags |= SF_IS_INF;
4086 if ( flags & SCF_DO_STCLASS && !scan->flags
4087 && OP(scan) == IFMATCH ) { /* Lookahead */
4088 cl_init(pRExC_state, &intrnl);
4089 data_fake.start_class = &intrnl;
4090 f |= SCF_DO_STCLASS_AND;
4092 if (flags & SCF_WHILEM_VISITED_POS)
4093 f |= SCF_WHILEM_VISITED_POS;
4094 next = regnext(scan);
4095 nscan = NEXTOPER(NEXTOPER(scan));
4097 *minnextp = study_chunk(pRExC_state, &nscan, minnextp, &deltanext,
4098 last, &data_fake, stopparen, recursed, NULL, f,depth+1);
4101 FAIL("Variable length lookbehind not implemented");
4103 else if (*minnextp > (I32)U8_MAX) {
4104 FAIL2("Lookbehind longer than %"UVuf" not implemented", (UV)U8_MAX);
4106 scan->flags = (U8)*minnextp;
4111 if (f & SCF_DO_STCLASS_AND) {
4112 const int was = (data->start_class->flags & ANYOF_EOS);
4114 cl_and(data->start_class, &intrnl);
4116 data->start_class->flags |= ANYOF_EOS;
4119 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
4121 if (data_fake.flags & SF_HAS_EVAL)
4122 data->flags |= SF_HAS_EVAL;
4123 data->whilem_c = data_fake.whilem_c;
4124 if ((flags & SCF_DO_SUBSTR) && data_fake.last_found) {
4125 if (RExC_rx->minlen<*minnextp)
4126 RExC_rx->minlen=*minnextp;
4127 SCAN_COMMIT(pRExC_state, &data_fake, minnextp);
4128 SvREFCNT_dec(data_fake.last_found);
4130 if ( data_fake.minlen_fixed != minlenp )
4132 data->offset_fixed= data_fake.offset_fixed;
4133 data->minlen_fixed= data_fake.minlen_fixed;
4134 data->lookbehind_fixed+= scan->flags;
4136 if ( data_fake.minlen_float != minlenp )
4138 data->minlen_float= data_fake.minlen_float;
4139 data->offset_float_min=data_fake.offset_float_min;
4140 data->offset_float_max=data_fake.offset_float_max;
4141 data->lookbehind_float+= scan->flags;
4150 else if (OP(scan) == OPEN) {
4151 if (stopparen != (I32)ARG(scan))
4154 else if (OP(scan) == CLOSE) {
4155 if (stopparen == (I32)ARG(scan)) {
4158 if ((I32)ARG(scan) == is_par) {
4159 next = regnext(scan);
4161 if ( next && (OP(next) != WHILEM) && next < last)
4162 is_par = 0; /* Disable optimization */
4165 *(data->last_closep) = ARG(scan);
4167 else if (OP(scan) == EVAL) {
4169 data->flags |= SF_HAS_EVAL;
4171 else if ( PL_regkind[OP(scan)] == ENDLIKE ) {
4172 if (flags & SCF_DO_SUBSTR) {
4173 SCAN_COMMIT(pRExC_state,data,minlenp);
4174 flags &= ~SCF_DO_SUBSTR;
4176 if (data && OP(scan)==ACCEPT) {
4177 data->flags |= SCF_SEEN_ACCEPT;
4182 else if (OP(scan) == LOGICAL && scan->flags == 2) /* Embedded follows */
4184 if (flags & SCF_DO_SUBSTR) {
4185 SCAN_COMMIT(pRExC_state,data,minlenp);
4186 data->longest = &(data->longest_float);
4188 is_inf = is_inf_internal = 1;
4189 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
4190 cl_anything(pRExC_state, data->start_class);
4191 flags &= ~SCF_DO_STCLASS;
4193 else if (OP(scan) == GPOS) {
4194 if (!(RExC_rx->extflags & RXf_GPOS_FLOAT) &&
4195 !(delta || is_inf || (data && data->pos_delta)))
4197 if (!(RExC_rx->extflags & RXf_ANCH) && (flags & SCF_DO_SUBSTR))
4198 RExC_rx->extflags |= RXf_ANCH_GPOS;
4199 if (RExC_rx->gofs < (U32)min)
4200 RExC_rx->gofs = min;
4202 RExC_rx->extflags |= RXf_GPOS_FLOAT;
4206 #ifdef TRIE_STUDY_OPT
4207 #ifdef FULL_TRIE_STUDY
4208 else if (PL_regkind[OP(scan)] == TRIE) {
4209 /* NOTE - There is similar code to this block above for handling
4210 BRANCH nodes on the initial study. If you change stuff here
4212 regnode *trie_node= scan;
4213 regnode *tail= regnext(scan);
4214 reg_trie_data *trie = (reg_trie_data*)RExC_rxi->data->data[ ARG(scan) ];
4215 I32 max1 = 0, min1 = I32_MAX;
4216 struct regnode_charclass_class accum;
4218 if (flags & SCF_DO_SUBSTR) /* XXXX Add !SUSPEND? */
4219 SCAN_COMMIT(pRExC_state, data,minlenp); /* Cannot merge strings after this. */
4220 if (flags & SCF_DO_STCLASS)
4221 cl_init_zero(pRExC_state, &accum);
4227 const regnode *nextbranch= NULL;
4230 for ( word=1 ; word <= trie->wordcount ; word++)
4232 I32 deltanext=0, minnext=0, f = 0, fake;
4233 struct regnode_charclass_class this_class;
4235 data_fake.flags = 0;
4237 data_fake.whilem_c = data->whilem_c;
4238 data_fake.last_closep = data->last_closep;
4241 data_fake.last_closep = &fake;
4242 data_fake.pos_delta = delta;
4243 if (flags & SCF_DO_STCLASS) {
4244 cl_init(pRExC_state, &this_class);
4245 data_fake.start_class = &this_class;
4246 f = SCF_DO_STCLASS_AND;
4248 if (flags & SCF_WHILEM_VISITED_POS)
4249 f |= SCF_WHILEM_VISITED_POS;
4251 if (trie->jump[word]) {
4253 nextbranch = trie_node + trie->jump[0];
4254 scan= trie_node + trie->jump[word];
4255 /* We go from the jump point to the branch that follows
4256 it. Note this means we need the vestigal unused branches
4257 even though they arent otherwise used.
4259 minnext = study_chunk(pRExC_state, &scan, minlenp,
4260 &deltanext, (regnode *)nextbranch, &data_fake,
4261 stopparen, recursed, NULL, f,depth+1);
4263 if (nextbranch && PL_regkind[OP(nextbranch)]==BRANCH)
4264 nextbranch= regnext((regnode*)nextbranch);
4266 if (min1 > (I32)(minnext + trie->minlen))
4267 min1 = minnext + trie->minlen;
4268 if (max1 < (I32)(minnext + deltanext + trie->maxlen))
4269 max1 = minnext + deltanext + trie->maxlen;
4270 if (deltanext == I32_MAX)
4271 is_inf = is_inf_internal = 1;
4273 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
4275 if (data_fake.flags & SCF_SEEN_ACCEPT) {
4276 if ( stopmin > min + min1)
4277 stopmin = min + min1;
4278 flags &= ~SCF_DO_SUBSTR;
4280 data->flags |= SCF_SEEN_ACCEPT;
4283 if (data_fake.flags & SF_HAS_EVAL)
4284 data->flags |= SF_HAS_EVAL;
4285 data->whilem_c = data_fake.whilem_c;
4287 if (flags & SCF_DO_STCLASS)
4288 cl_or(pRExC_state, &accum, &this_class);
4291 if (flags & SCF_DO_SUBSTR) {
4292 data->pos_min += min1;
4293 data->pos_delta += max1 - min1;
4294 if (max1 != min1 || is_inf)
4295 data->longest = &(data->longest_float);
4298 delta += max1 - min1;
4299 if (flags & SCF_DO_STCLASS_OR) {
4300 cl_or(pRExC_state, data->start_class, &accum);
4302 cl_and(data->start_class, and_withp);
4303 flags &= ~SCF_DO_STCLASS;
4306 else if (flags & SCF_DO_STCLASS_AND) {
4308 cl_and(data->start_class, &accum);
4309 flags &= ~SCF_DO_STCLASS;
4312 /* Switch to OR mode: cache the old value of
4313 * data->start_class */
4315 StructCopy(data->start_class, and_withp,
4316 struct regnode_charclass_class);
4317 flags &= ~SCF_DO_STCLASS_AND;
4318 StructCopy(&accum, data->start_class,
4319 struct regnode_charclass_class);
4320 flags |= SCF_DO_STCLASS_OR;
4321 data->start_class->flags |= ANYOF_EOS;
4328 else if (PL_regkind[OP(scan)] == TRIE) {
4329 reg_trie_data *trie = (reg_trie_data*)RExC_rxi->data->data[ ARG(scan) ];
4332 min += trie->minlen;
4333 delta += (trie->maxlen - trie->minlen);
4334 flags &= ~SCF_DO_STCLASS; /* xxx */
4335 if (flags & SCF_DO_SUBSTR) {
4336 SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot expect anything... */
4337 data->pos_min += trie->minlen;
4338 data->pos_delta += (trie->maxlen - trie->minlen);
4339 if (trie->maxlen != trie->minlen)
4340 data->longest = &(data->longest_float);
4342 if (trie->jump) /* no more substrings -- for now /grr*/
4343 flags &= ~SCF_DO_SUBSTR;
4345 #endif /* old or new */
4346 #endif /* TRIE_STUDY_OPT */
4348 /* Else: zero-length, ignore. */
4349 scan = regnext(scan);
4354 stopparen = frame->stop;
4355 frame = frame->prev;
4356 goto fake_study_recurse;
4361 DEBUG_STUDYDATA("pre-fin:",data,depth);
4364 *deltap = is_inf_internal ? I32_MAX : delta;
4365 if (flags & SCF_DO_SUBSTR && is_inf)
4366 data->pos_delta = I32_MAX - data->pos_min;
4367 if (is_par > (I32)U8_MAX)
4369 if (is_par && pars==1 && data) {
4370 data->flags |= SF_IN_PAR;
4371 data->flags &= ~SF_HAS_PAR;
4373 else if (pars && data) {
4374 data->flags |= SF_HAS_PAR;
4375 data->flags &= ~SF_IN_PAR;
4377 if (flags & SCF_DO_STCLASS_OR)
4378 cl_and(data->start_class, and_withp);
4379 if (flags & SCF_TRIE_RESTUDY)
4380 data->flags |= SCF_TRIE_RESTUDY;
4382 DEBUG_STUDYDATA("post-fin:",data,depth);
4384 return min < stopmin ? min : stopmin;
4388 S_add_data(RExC_state_t *pRExC_state, U32 n, const char *s)
4390 U32 count = RExC_rxi->data ? RExC_rxi->data->count : 0;
4392 PERL_ARGS_ASSERT_ADD_DATA;
4394 Renewc(RExC_rxi->data,
4395 sizeof(*RExC_rxi->data) + sizeof(void*) * (count + n - 1),
4396 char, struct reg_data);
4398 Renew(RExC_rxi->data->what, count + n, U8);
4400 Newx(RExC_rxi->data->what, n, U8);
4401 RExC_rxi->data->count = count + n;
4402 Copy(s, RExC_rxi->data->what + count, n, U8);
4406 /*XXX: todo make this not included in a non debugging perl */
4407 #ifndef PERL_IN_XSUB_RE
4409 Perl_reginitcolors(pTHX)
4412 const char * const s = PerlEnv_getenv("PERL_RE_COLORS");
4414 char *t = savepv(s);
4418 t = strchr(t, '\t');
4424 PL_colors[i] = t = (char *)"";
4429 PL_colors[i++] = (char *)"";
4436 #ifdef TRIE_STUDY_OPT
4437 #define CHECK_RESTUDY_GOTO \
4439 (data.flags & SCF_TRIE_RESTUDY) \
4443 #define CHECK_RESTUDY_GOTO
4447 - pregcomp - compile a regular expression into internal code
4449 * We can't allocate space until we know how big the compiled form will be,
4450 * but we can't compile it (and thus know how big it is) until we've got a
4451 * place to put the code. So we cheat: we compile it twice, once with code
4452 * generation turned off and size counting turned on, and once "for real".
4453 * This also means that we don't allocate space until we are sure that the
4454 * thing really will compile successfully, and we never have to move the
4455 * code and thus invalidate pointers into it. (Note that it has to be in
4456 * one piece because free() must be able to free it all.) [NB: not true in perl]
4458 * Beware that the optimization-preparation code in here knows about some
4459 * of the structure of the compiled regexp. [I'll say.]
4464 #ifndef PERL_IN_XSUB_RE
4465 #define RE_ENGINE_PTR &PL_core_reg_engine
4467 extern const struct regexp_engine my_reg_engine;
4468 #define RE_ENGINE_PTR &my_reg_engine
4471 #ifndef PERL_IN_XSUB_RE
4473 Perl_pregcomp(pTHX_ SV * const pattern, const U32 flags)
4476 HV * const table = GvHV(PL_hintgv);
4478 PERL_ARGS_ASSERT_PREGCOMP;
4480 /* Dispatch a request to compile a regexp to correct
4483 SV **ptr= hv_fetchs(table, "regcomp", FALSE);
4484 GET_RE_DEBUG_FLAGS_DECL;
4485 if (ptr && SvIOK(*ptr) && SvIV(*ptr)) {
4486 const regexp_engine *eng=INT2PTR(regexp_engine*,SvIV(*ptr));
4488 PerlIO_printf(Perl_debug_log, "Using engine %"UVxf"\n",
4491 return CALLREGCOMP_ENG(eng, pattern, flags);
4494 return Perl_re_compile(aTHX_ pattern, flags);
4499 Perl_re_compile(pTHX_ SV * const pattern, U32 orig_pm_flags)
4504 register regexp_internal *ri;
4513 /* these are all flags - maybe they should be turned
4514 * into a single int with different bit masks */
4515 I32 sawlookahead = 0;
4518 bool used_setjump = FALSE;
4519 regex_charset initial_charset = get_regex_charset(orig_pm_flags);
4524 RExC_state_t RExC_state;
4525 RExC_state_t * const pRExC_state = &RExC_state;
4526 #ifdef TRIE_STUDY_OPT
4528 RExC_state_t copyRExC_state;
4530 GET_RE_DEBUG_FLAGS_DECL;
4532 PERL_ARGS_ASSERT_RE_COMPILE;
4534 DEBUG_r(if (!PL_colorset) reginitcolors());
4536 RExC_utf8 = RExC_orig_utf8 = SvUTF8(pattern);
4537 RExC_uni_semantics = 0;
4538 RExC_contains_locale = 0;
4540 /****************** LONG JUMP TARGET HERE***********************/
4541 /* Longjmp back to here if have to switch in midstream to utf8 */
4542 if (! RExC_orig_utf8) {
4543 JMPENV_PUSH(jump_ret);
4544 used_setjump = TRUE;
4547 if (jump_ret == 0) { /* First time through */
4548 exp = SvPV(pattern, plen);
4550 /* ignore the utf8ness if the pattern is 0 length */
4552 RExC_utf8 = RExC_orig_utf8 = 0;
4556 SV *dsv= sv_newmortal();
4557 RE_PV_QUOTED_DECL(s, RExC_utf8,
4558 dsv, exp, plen, 60);
4559 PerlIO_printf(Perl_debug_log, "%sCompiling REx%s %s\n",
4560 PL_colors[4],PL_colors[5],s);
4563 else { /* longjumped back */
4566 /* If the cause for the longjmp was other than changing to utf8, pop
4567 * our own setjmp, and longjmp to the correct handler */
4568 if (jump_ret != UTF8_LONGJMP) {
4570 JMPENV_JUMP(jump_ret);
4575 /* It's possible to write a regexp in ascii that represents Unicode
4576 codepoints outside of the byte range, such as via \x{100}. If we
4577 detect such a sequence we have to convert the entire pattern to utf8
4578 and then recompile, as our sizing calculation will have been based
4579 on 1 byte == 1 character, but we will need to use utf8 to encode
4580 at least some part of the pattern, and therefore must convert the whole
4583 DEBUG_PARSE_r(PerlIO_printf(Perl_debug_log,
4584 "UTF8 mismatch! Converting to utf8 for resizing and compile\n"));
4585 exp = (char*)Perl_bytes_to_utf8(aTHX_ (U8*)SvPV(pattern, plen), &len);
4587 RExC_orig_utf8 = RExC_utf8 = 1;
4591 #ifdef TRIE_STUDY_OPT
4595 pm_flags = orig_pm_flags;
4597 if (initial_charset == REGEX_LOCALE_CHARSET) {
4598 RExC_contains_locale = 1;
4600 else if (RExC_utf8 && initial_charset == REGEX_DEPENDS_CHARSET) {
4602 /* Set to use unicode semantics if the pattern is in utf8 and has the
4603 * 'depends' charset specified, as it means unicode when utf8 */
4604 set_regex_charset(&pm_flags, REGEX_UNICODE_CHARSET);
4608 RExC_flags = pm_flags;
4612 RExC_in_lookbehind = 0;
4613 RExC_seen_zerolen = *exp == '^' ? -1 : 0;
4614 RExC_seen_evals = 0;
4616 RExC_override_recoding = 0;
4618 /* First pass: determine size, legality. */
4626 RExC_emit = &PL_regdummy;
4627 RExC_whilem_seen = 0;
4628 RExC_open_parens = NULL;
4629 RExC_close_parens = NULL;
4631 RExC_paren_names = NULL;
4633 RExC_paren_name_list = NULL;
4635 RExC_recurse = NULL;
4636 RExC_recurse_count = 0;
4638 #if 0 /* REGC() is (currently) a NOP at the first pass.
4639 * Clever compilers notice this and complain. --jhi */
4640 REGC((U8)REG_MAGIC, (char*)RExC_emit);
4642 DEBUG_PARSE_r(PerlIO_printf(Perl_debug_log, "Starting first pass (sizing)\n"));
4643 if (reg(pRExC_state, 0, &flags,1) == NULL) {
4644 RExC_precomp = NULL;
4648 /* Here, finished first pass. Get rid of any added setjmp */
4654 PerlIO_printf(Perl_debug_log,
4655 "Required size %"IVdf" nodes\n"
4656 "Starting second pass (creation)\n",
4659 RExC_lastparse=NULL;
4662 /* The first pass could have found things that force Unicode semantics */
4663 if ((RExC_utf8 || RExC_uni_semantics)
4664 && get_regex_charset(pm_flags) == REGEX_DEPENDS_CHARSET)
4666 set_regex_charset(&pm_flags, REGEX_UNICODE_CHARSET);
4669 /* Small enough for pointer-storage convention?
4670 If extralen==0, this means that we will not need long jumps. */
4671 if (RExC_size >= 0x10000L && RExC_extralen)
4672 RExC_size += RExC_extralen;
4675 if (RExC_whilem_seen > 15)
4676 RExC_whilem_seen = 15;
4678 /* Allocate space and zero-initialize. Note, the two step process
4679 of zeroing when in debug mode, thus anything assigned has to
4680 happen after that */
4681 rx = (REGEXP*) newSV_type(SVt_REGEXP);
4682 r = (struct regexp*)SvANY(rx);
4683 Newxc(ri, sizeof(regexp_internal) + (unsigned)RExC_size * sizeof(regnode),
4684 char, regexp_internal);
4685 if ( r == NULL || ri == NULL )
4686 FAIL("Regexp out of space");
4688 /* avoid reading uninitialized memory in DEBUGGING code in study_chunk() */
4689 Zero(ri, sizeof(regexp_internal) + (unsigned)RExC_size * sizeof(regnode), char);
4691 /* bulk initialize base fields with 0. */
4692 Zero(ri, sizeof(regexp_internal), char);
4695 /* non-zero initialization begins here */
4697 r->engine= RE_ENGINE_PTR;
4698 r->extflags = pm_flags;
4700 bool has_p = ((r->extflags & RXf_PMf_KEEPCOPY) == RXf_PMf_KEEPCOPY);
4701 bool has_charset = (get_regex_charset(r->extflags) != REGEX_DEPENDS_CHARSET);
4703 /* The caret is output if there are any defaults: if not all the STD
4704 * flags are set, or if no character set specifier is needed */
4706 (((r->extflags & RXf_PMf_STD_PMMOD) != RXf_PMf_STD_PMMOD)
4708 bool has_runon = ((RExC_seen & REG_SEEN_RUN_ON_COMMENT)==REG_SEEN_RUN_ON_COMMENT);
4709 U16 reganch = (U16)((r->extflags & RXf_PMf_STD_PMMOD)
4710 >> RXf_PMf_STD_PMMOD_SHIFT);
4711 const char *fptr = STD_PAT_MODS; /*"msix"*/
4713 /* Allocate for the worst case, which is all the std flags are turned
4714 * on. If more precision is desired, we could do a population count of
4715 * the flags set. This could be done with a small lookup table, or by
4716 * shifting, masking and adding, or even, when available, assembly
4717 * language for a machine-language population count.
4718 * We never output a minus, as all those are defaults, so are
4719 * covered by the caret */
4720 const STRLEN wraplen = plen + has_p + has_runon
4721 + has_default /* If needs a caret */
4723 /* If needs a character set specifier */
4724 + ((has_charset) ? MAX_CHARSET_NAME_LENGTH : 0)
4725 + (sizeof(STD_PAT_MODS) - 1)
4726 + (sizeof("(?:)") - 1);
4728 p = sv_grow(MUTABLE_SV(rx), wraplen + 1); /* +1 for the ending NUL */
4730 SvFLAGS(rx) |= SvUTF8(pattern);
4733 /* If a default, cover it using the caret */
4735 *p++= DEFAULT_PAT_MOD;
4739 const char* const name = get_regex_charset_name(r->extflags, &len);
4740 Copy(name, p, len, char);
4744 *p++ = KEEPCOPY_PAT_MOD; /*'p'*/
4747 while((ch = *fptr++)) {
4755 Copy(RExC_precomp, p, plen, char);
4756 assert ((RX_WRAPPED(rx) - p) < 16);
4757 r->pre_prefix = p - RX_WRAPPED(rx);
4763 SvCUR_set(rx, p - SvPVX_const(rx));
4767 r->nparens = RExC_npar - 1; /* set early to validate backrefs */
4769 if (RExC_seen & REG_SEEN_RECURSE) {
4770 Newxz(RExC_open_parens, RExC_npar,regnode *);
4771 SAVEFREEPV(RExC_open_parens);
4772 Newxz(RExC_close_parens,RExC_npar,regnode *);
4773 SAVEFREEPV(RExC_close_parens);
4776 /* Useful during FAIL. */
4777 #ifdef RE_TRACK_PATTERN_OFFSETS
4778 Newxz(ri->u.offsets, 2*RExC_size+1, U32); /* MJD 20001228 */
4779 DEBUG_OFFSETS_r(PerlIO_printf(Perl_debug_log,
4780 "%s %"UVuf" bytes for offset annotations.\n",
4781 ri->u.offsets ? "Got" : "Couldn't get",
4782 (UV)((2*RExC_size+1) * sizeof(U32))));
4784 SetProgLen(ri,RExC_size);
4789 /* Second pass: emit code. */
4790 RExC_flags = pm_flags; /* don't let top level (?i) bleed */
4795 RExC_emit_start = ri->program;
4796 RExC_emit = ri->program;
4797 RExC_emit_bound = ri->program + RExC_size + 1;
4799 /* Store the count of eval-groups for security checks: */
4800 RExC_rx->seen_evals = RExC_seen_evals;
4801 REGC((U8)REG_MAGIC, (char*) RExC_emit++);
4802 if (reg(pRExC_state, 0, &flags,1) == NULL) {
4806 /* XXXX To minimize changes to RE engine we always allocate
4807 3-units-long substrs field. */
4808 Newx(r->substrs, 1, struct reg_substr_data);
4809 if (RExC_recurse_count) {
4810 Newxz(RExC_recurse,RExC_recurse_count,regnode *);
4811 SAVEFREEPV(RExC_recurse);
4815 r->minlen = minlen = sawlookahead = sawplus = sawopen = 0;
4816 Zero(r->substrs, 1, struct reg_substr_data);
4818 #ifdef TRIE_STUDY_OPT
4820 StructCopy(&zero_scan_data, &data, scan_data_t);
4821 copyRExC_state = RExC_state;
4824 DEBUG_OPTIMISE_r(PerlIO_printf(Perl_debug_log,"Restudying\n"));
4826 RExC_state = copyRExC_state;
4827 if (seen & REG_TOP_LEVEL_BRANCHES)
4828 RExC_seen |= REG_TOP_LEVEL_BRANCHES;
4830 RExC_seen &= ~REG_TOP_LEVEL_BRANCHES;
4831 if (data.last_found) {
4832 SvREFCNT_dec(data.longest_fixed);
4833 SvREFCNT_dec(data.longest_float);
4834 SvREFCNT_dec(data.last_found);
4836 StructCopy(&zero_scan_data, &data, scan_data_t);
4839 StructCopy(&zero_scan_data, &data, scan_data_t);
4842 /* Dig out information for optimizations. */
4843 r->extflags = RExC_flags; /* was pm_op */
4844 /*dmq: removed as part of de-PMOP: pm->op_pmflags = RExC_flags; */
4847 SvUTF8_on(rx); /* Unicode in it? */
4848 ri->regstclass = NULL;
4849 if (RExC_naughty >= 10) /* Probably an expensive pattern. */
4850 r->intflags |= PREGf_NAUGHTY;
4851 scan = ri->program + 1; /* First BRANCH. */
4853 /* testing for BRANCH here tells us whether there is "must appear"
4854 data in the pattern. If there is then we can use it for optimisations */
4855 if (!(RExC_seen & REG_TOP_LEVEL_BRANCHES)) { /* Only one top-level choice. */
4857 STRLEN longest_float_length, longest_fixed_length;
4858 struct regnode_charclass_class ch_class; /* pointed to by data */
4860 I32 last_close = 0; /* pointed to by data */
4861 regnode *first= scan;
4862 regnode *first_next= regnext(first);
4864 * Skip introductions and multiplicators >= 1
4865 * so that we can extract the 'meat' of the pattern that must
4866 * match in the large if() sequence following.
4867 * NOTE that EXACT is NOT covered here, as it is normally
4868 * picked up by the optimiser separately.
4870 * This is unfortunate as the optimiser isnt handling lookahead
4871 * properly currently.
4874 while ((OP(first) == OPEN && (sawopen = 1)) ||
4875 /* An OR of *one* alternative - should not happen now. */
4876 (OP(first) == BRANCH && OP(first_next) != BRANCH) ||
4877 /* for now we can't handle lookbehind IFMATCH*/
4878 (OP(first) == IFMATCH && !first->flags && (sawlookahead = 1)) ||
4879 (OP(first) == PLUS) ||
4880 (OP(first) == MINMOD) ||
4881 /* An {n,m} with n>0 */
4882 (PL_regkind[OP(first)] == CURLY && ARG1(first) > 0) ||
4883 (OP(first) == NOTHING && PL_regkind[OP(first_next)] != END ))
4886 * the only op that could be a regnode is PLUS, all the rest
4887 * will be regnode_1 or regnode_2.
4890 if (OP(first) == PLUS)
4893 first += regarglen[OP(first)];
4895 first = NEXTOPER(first);
4896 first_next= regnext(first);
4899 /* Starting-point info. */
4901 DEBUG_PEEP("first:",first,0);
4902 /* Ignore EXACT as we deal with it later. */
4903 if (PL_regkind[OP(first)] == EXACT) {
4904 if (OP(first) == EXACT)
4905 NOOP; /* Empty, get anchored substr later. */
4907 ri->regstclass = first;
4910 else if (PL_regkind[OP(first)] == TRIE &&
4911 ((reg_trie_data *)ri->data->data[ ARG(first) ])->minlen>0)
4914 /* this can happen only on restudy */
4915 if ( OP(first) == TRIE ) {
4916 struct regnode_1 *trieop = (struct regnode_1 *)
4917 PerlMemShared_calloc(1, sizeof(struct regnode_1));
4918 StructCopy(first,trieop,struct regnode_1);
4919 trie_op=(regnode *)trieop;
4921 struct regnode_charclass *trieop = (struct regnode_charclass *)
4922 PerlMemShared_calloc(1, sizeof(struct regnode_charclass));
4923 StructCopy(first,trieop,struct regnode_charclass);
4924 trie_op=(regnode *)trieop;
4927 make_trie_failtable(pRExC_state, (regnode *)first, trie_op, 0);
4928 ri->regstclass = trie_op;
4931 else if (REGNODE_SIMPLE(OP(first)))
4932 ri->regstclass = first;
4933 else if (PL_regkind[OP(first)] == BOUND ||
4934 PL_regkind[OP(first)] == NBOUND)
4935 ri->regstclass = first;
4936 else if (PL_regkind[OP(first)] == BOL) {
4937 r->extflags |= (OP(first) == MBOL
4939 : (OP(first) == SBOL
4942 first = NEXTOPER(first);
4945 else if (OP(first) == GPOS) {
4946 r->extflags |= RXf_ANCH_GPOS;
4947 first = NEXTOPER(first);
4950 else if ((!sawopen || !RExC_sawback) &&
4951 (OP(first) == STAR &&
4952 PL_regkind[OP(NEXTOPER(first))] == REG_ANY) &&
4953 !(r->extflags & RXf_ANCH) && !(RExC_seen & REG_SEEN_EVAL))
4955 /* turn .* into ^.* with an implied $*=1 */
4957 (OP(NEXTOPER(first)) == REG_ANY)
4960 r->extflags |= type;
4961 r->intflags |= PREGf_IMPLICIT;
4962 first = NEXTOPER(first);
4965 if (sawplus && !sawlookahead && (!sawopen || !RExC_sawback)
4966 && !(RExC_seen & REG_SEEN_EVAL)) /* May examine pos and $& */
4967 /* x+ must match at the 1st pos of run of x's */
4968 r->intflags |= PREGf_SKIP;
4970 /* Scan is after the zeroth branch, first is atomic matcher. */
4971 #ifdef TRIE_STUDY_OPT
4974 PerlIO_printf(Perl_debug_log, "first at %"IVdf"\n",
4975 (IV)(first - scan + 1))
4979 PerlIO_printf(Perl_debug_log, "first at %"IVdf"\n",
4980 (IV)(first - scan + 1))
4986 * If there's something expensive in the r.e., find the
4987 * longest literal string that must appear and make it the
4988 * regmust. Resolve ties in favor of later strings, since
4989 * the regstart check works with the beginning of the r.e.
4990 * and avoiding duplication strengthens checking. Not a
4991 * strong reason, but sufficient in the absence of others.
4992 * [Now we resolve ties in favor of the earlier string if
4993 * it happens that c_offset_min has been invalidated, since the
4994 * earlier string may buy us something the later one won't.]
4997 data.longest_fixed = newSVpvs("");
4998 data.longest_float = newSVpvs("");
4999 data.last_found = newSVpvs("");
5000 data.longest = &(data.longest_fixed);
5002 if (!ri->regstclass) {
5003 cl_init(pRExC_state, &ch_class);
5004 data.start_class = &ch_class;
5005 stclass_flag = SCF_DO_STCLASS_AND;
5006 } else /* XXXX Check for BOUND? */
5008 data.last_closep = &last_close;
5010 minlen = study_chunk(pRExC_state, &first, &minlen, &fake, scan + RExC_size, /* Up to end */
5011 &data, -1, NULL, NULL,
5012 SCF_DO_SUBSTR | SCF_WHILEM_VISITED_POS | stclass_flag,0);
5018 if ( RExC_npar == 1 && data.longest == &(data.longest_fixed)
5019 && data.last_start_min == 0 && data.last_end > 0
5020 && !RExC_seen_zerolen
5021 && !(RExC_seen & REG_SEEN_VERBARG)
5022 && (!(RExC_seen & REG_SEEN_GPOS) || (r->extflags & RXf_ANCH_GPOS)))
5023 r->extflags |= RXf_CHECK_ALL;
5024 scan_commit(pRExC_state, &data,&minlen,0);
5025 SvREFCNT_dec(data.last_found);
5027 /* Note that code very similar to this but for anchored string
5028 follows immediately below, changes may need to be made to both.
5031 longest_float_length = CHR_SVLEN(data.longest_float);
5032 if (longest_float_length
5033 || (data.flags & SF_FL_BEFORE_EOL
5034 && (!(data.flags & SF_FL_BEFORE_MEOL)
5035 || (RExC_flags & RXf_PMf_MULTILINE))))
5039 if (SvCUR(data.longest_fixed) /* ok to leave SvCUR */
5040 && data.offset_fixed == data.offset_float_min
5041 && SvCUR(data.longest_fixed) == SvCUR(data.longest_float))
5042 goto remove_float; /* As in (a)+. */
5044 /* copy the information about the longest float from the reg_scan_data
5045 over to the program. */
5046 if (SvUTF8(data.longest_float)) {
5047 r->float_utf8 = data.longest_float;
5048 r->float_substr = NULL;
5050 r->float_substr = data.longest_float;
5051 r->float_utf8 = NULL;
5053 /* float_end_shift is how many chars that must be matched that
5054 follow this item. We calculate it ahead of time as once the
5055 lookbehind offset is added in we lose the ability to correctly
5057 ml = data.minlen_float ? *(data.minlen_float)
5058 : (I32)longest_float_length;
5059 r->float_end_shift = ml - data.offset_float_min
5060 - longest_float_length + (SvTAIL(data.longest_float) != 0)
5061 + data.lookbehind_float;
5062 r->float_min_offset = data.offset_float_min - data.lookbehind_float;
5063 r->float_max_offset = data.offset_float_max;
5064 if (data.offset_float_max < I32_MAX) /* Don't offset infinity */
5065 r->float_max_offset -= data.lookbehind_float;
5067 t = (data.flags & SF_FL_BEFORE_EOL /* Can't have SEOL and MULTI */
5068 && (!(data.flags & SF_FL_BEFORE_MEOL)
5069 || (RExC_flags & RXf_PMf_MULTILINE)));
5070 fbm_compile(data.longest_float, t ? FBMcf_TAIL : 0);
5074 r->float_substr = r->float_utf8 = NULL;
5075 SvREFCNT_dec(data.longest_float);
5076 longest_float_length = 0;
5079 /* Note that code very similar to this but for floating string
5080 is immediately above, changes may need to be made to both.
5083 longest_fixed_length = CHR_SVLEN(data.longest_fixed);
5084 if (longest_fixed_length
5085 || (data.flags & SF_FIX_BEFORE_EOL /* Cannot have SEOL and MULTI */
5086 && (!(data.flags & SF_FIX_BEFORE_MEOL)
5087 || (RExC_flags & RXf_PMf_MULTILINE))))
5091 /* copy the information about the longest fixed
5092 from the reg_scan_data over to the program. */
5093 if (SvUTF8(data.longest_fixed)) {
5094 r->anchored_utf8 = data.longest_fixed;
5095 r->anchored_substr = NULL;
5097 r->anchored_substr = data.longest_fixed;
5098 r->anchored_utf8 = NULL;
5100 /* fixed_end_shift is how many chars that must be matched that
5101 follow this item. We calculate it ahead of time as once the
5102 lookbehind offset is added in we lose the ability to correctly
5104 ml = data.minlen_fixed ? *(data.minlen_fixed)
5105 : (I32)longest_fixed_length;
5106 r->anchored_end_shift = ml - data.offset_fixed
5107 - longest_fixed_length + (SvTAIL(data.longest_fixed) != 0)
5108 + data.lookbehind_fixed;
5109 r->anchored_offset = data.offset_fixed - data.lookbehind_fixed;
5111 t = (data.flags & SF_FIX_BEFORE_EOL /* Can't have SEOL and MULTI */
5112 && (!(data.flags & SF_FIX_BEFORE_MEOL)
5113 || (RExC_flags & RXf_PMf_MULTILINE)));
5114 fbm_compile(data.longest_fixed, t ? FBMcf_TAIL : 0);
5117 r->anchored_substr = r->anchored_utf8 = NULL;
5118 SvREFCNT_dec(data.longest_fixed);
5119 longest_fixed_length = 0;
5122 && (OP(ri->regstclass) == REG_ANY || OP(ri->regstclass) == SANY))
5123 ri->regstclass = NULL;
5125 if ((!(r->anchored_substr || r->anchored_utf8) || r->anchored_offset)
5127 && !(data.start_class->flags & ANYOF_EOS)
5128 && !cl_is_anything(data.start_class))
5130 const U32 n = add_data(pRExC_state, 1, "f");
5131 data.start_class->flags |= ANYOF_IS_SYNTHETIC;
5133 Newx(RExC_rxi->data->data[n], 1,
5134 struct regnode_charclass_class);
5135 StructCopy(data.start_class,
5136 (struct regnode_charclass_class*)RExC_rxi->data->data[n],
5137 struct regnode_charclass_class);
5138 ri->regstclass = (regnode*)RExC_rxi->data->data[n];
5139 r->intflags &= ~PREGf_SKIP; /* Used in find_byclass(). */
5140 DEBUG_COMPILE_r({ SV *sv = sv_newmortal();
5141 regprop(r, sv, (regnode*)data.start_class);
5142 PerlIO_printf(Perl_debug_log,
5143 "synthetic stclass \"%s\".\n",
5144 SvPVX_const(sv));});
5147 /* A temporary algorithm prefers floated substr to fixed one to dig more info. */
5148 if (longest_fixed_length > longest_float_length) {
5149 r->check_end_shift = r->anchored_end_shift;
5150 r->check_substr = r->anchored_substr;
5151 r->check_utf8 = r->anchored_utf8;
5152 r->check_offset_min = r->check_offset_max = r->anchored_offset;
5153 if (r->extflags & RXf_ANCH_SINGLE)
5154 r->extflags |= RXf_NOSCAN;
5157 r->check_end_shift = r->float_end_shift;
5158 r->check_substr = r->float_substr;
5159 r->check_utf8 = r->float_utf8;
5160 r->check_offset_min = r->float_min_offset;
5161 r->check_offset_max = r->float_max_offset;
5163 /* XXXX Currently intuiting is not compatible with ANCH_GPOS.
5164 This should be changed ASAP! */
5165 if ((r->check_substr || r->check_utf8) && !(r->extflags & RXf_ANCH_GPOS)) {
5166 r->extflags |= RXf_USE_INTUIT;
5167 if (SvTAIL(r->check_substr ? r->check_substr : r->check_utf8))
5168 r->extflags |= RXf_INTUIT_TAIL;
5170 /* XXX Unneeded? dmq (shouldn't as this is handled elsewhere)
5171 if ( (STRLEN)minlen < longest_float_length )
5172 minlen= longest_float_length;
5173 if ( (STRLEN)minlen < longest_fixed_length )
5174 minlen= longest_fixed_length;
5178 /* Several toplevels. Best we can is to set minlen. */
5180 struct regnode_charclass_class ch_class;
5183 DEBUG_PARSE_r(PerlIO_printf(Perl_debug_log, "\nMulti Top Level\n"));
5185 scan = ri->program + 1;
5186 cl_init(pRExC_state, &ch_class);
5187 data.start_class = &ch_class;
5188 data.last_closep = &last_close;
5191 minlen = study_chunk(pRExC_state, &scan, &minlen, &fake, scan + RExC_size,
5192 &data, -1, NULL, NULL, SCF_DO_STCLASS_AND|SCF_WHILEM_VISITED_POS,0);
5196 r->check_substr = r->check_utf8 = r->anchored_substr = r->anchored_utf8
5197 = r->float_substr = r->float_utf8 = NULL;
5199 if (!(data.start_class->flags & ANYOF_EOS)
5200 && !cl_is_anything(data.start_class))
5202 const U32 n = add_data(pRExC_state, 1, "f");
5203 data.start_class->flags |= ANYOF_IS_SYNTHETIC;
5205 Newx(RExC_rxi->data->data[n], 1,
5206 struct regnode_charclass_class);
5207 StructCopy(data.start_class,
5208 (struct regnode_charclass_class*)RExC_rxi->data->data[n],
5209 struct regnode_charclass_class);
5210 ri->regstclass = (regnode*)RExC_rxi->data->data[n];
5211 r->intflags &= ~PREGf_SKIP; /* Used in find_byclass(). */
5212 DEBUG_COMPILE_r({ SV* sv = sv_newmortal();
5213 regprop(r, sv, (regnode*)data.start_class);
5214 PerlIO_printf(Perl_debug_log,
5215 "synthetic stclass \"%s\".\n",
5216 SvPVX_const(sv));});
5220 /* Guard against an embedded (?=) or (?<=) with a longer minlen than
5221 the "real" pattern. */
5223 PerlIO_printf(Perl_debug_log,"minlen: %"IVdf" r->minlen:%"IVdf"\n",
5224 (IV)minlen, (IV)r->minlen);
5226 r->minlenret = minlen;
5227 if (r->minlen < minlen)
5230 if (RExC_seen & REG_SEEN_GPOS)
5231 r->extflags |= RXf_GPOS_SEEN;
5232 if (RExC_seen & REG_SEEN_LOOKBEHIND)
5233 r->extflags |= RXf_LOOKBEHIND_SEEN;
5234 if (RExC_seen & REG_SEEN_EVAL)
5235 r->extflags |= RXf_EVAL_SEEN;
5236 if (RExC_seen & REG_SEEN_CANY)
5237 r->extflags |= RXf_CANY_SEEN;
5238 if (RExC_seen & REG_SEEN_VERBARG)
5239 r->intflags |= PREGf_VERBARG_SEEN;
5240 if (RExC_seen & REG_SEEN_CUTGROUP)
5241 r->intflags |= PREGf_CUTGROUP_SEEN;
5242 if (RExC_paren_names)
5243 RXp_PAREN_NAMES(r) = MUTABLE_HV(SvREFCNT_inc(RExC_paren_names));
5245 RXp_PAREN_NAMES(r) = NULL;
5247 #ifdef STUPID_PATTERN_CHECKS
5248 if (RX_PRELEN(rx) == 0)
5249 r->extflags |= RXf_NULL;
5250 if (r->extflags & RXf_SPLIT && RX_PRELEN(rx) == 1 && RX_PRECOMP(rx)[0] == ' ')
5251 /* XXX: this should happen BEFORE we compile */
5252 r->extflags |= (RXf_SKIPWHITE|RXf_WHITE);
5253 else if (RX_PRELEN(rx) == 3 && memEQ("\\s+", RX_PRECOMP(rx), 3))
5254 r->extflags |= RXf_WHITE;
5255 else if (RX_PRELEN(rx) == 1 && RXp_PRECOMP(rx)[0] == '^')
5256 r->extflags |= RXf_START_ONLY;
5258 if (r->extflags & RXf_SPLIT && RX_PRELEN(rx) == 1 && RX_PRECOMP(rx)[0] == ' ')
5259 /* XXX: this should happen BEFORE we compile */
5260 r->extflags |= (RXf_SKIPWHITE|RXf_WHITE);
5262 regnode *first = ri->program + 1;
5265 if (PL_regkind[fop] == NOTHING && OP(NEXTOPER(first)) == END)
5266 r->extflags |= RXf_NULL;
5267 else if (PL_regkind[fop] == BOL && OP(NEXTOPER(first)) == END)
5268 r->extflags |= RXf_START_ONLY;
5269 else if (fop == PLUS && OP(NEXTOPER(first)) == SPACE
5270 && OP(regnext(first)) == END)
5271 r->extflags |= RXf_WHITE;
5275 if (RExC_paren_names) {
5276 ri->name_list_idx = add_data( pRExC_state, 1, "a" );
5277 ri->data->data[ri->name_list_idx] = (void*)SvREFCNT_inc(RExC_paren_name_list);
5280 ri->name_list_idx = 0;
5282 if (RExC_recurse_count) {
5283 for ( ; RExC_recurse_count ; RExC_recurse_count-- ) {
5284 const regnode *scan = RExC_recurse[RExC_recurse_count-1];
5285 ARG2L_SET( scan, RExC_open_parens[ARG(scan)-1] - scan );
5288 Newxz(r->offs, RExC_npar, regexp_paren_pair);
5289 /* assume we don't need to swap parens around before we match */
5292 PerlIO_printf(Perl_debug_log,"Final program:\n");
5295 #ifdef RE_TRACK_PATTERN_OFFSETS
5296 DEBUG_OFFSETS_r(if (ri->u.offsets) {
5297 const U32 len = ri->u.offsets[0];
5299 GET_RE_DEBUG_FLAGS_DECL;
5300 PerlIO_printf(Perl_debug_log, "Offsets: [%"UVuf"]\n\t", (UV)ri->u.offsets[0]);
5301 for (i = 1; i <= len; i++) {
5302 if (ri->u.offsets[i*2-1] || ri->u.offsets[i*2])
5303 PerlIO_printf(Perl_debug_log, "%"UVuf":%"UVuf"[%"UVuf"] ",
5304 (UV)i, (UV)ri->u.offsets[i*2-1], (UV)ri->u.offsets[i*2]);
5306 PerlIO_printf(Perl_debug_log, "\n");
5312 #undef RE_ENGINE_PTR
5316 Perl_reg_named_buff(pTHX_ REGEXP * const rx, SV * const key, SV * const value,
5319 PERL_ARGS_ASSERT_REG_NAMED_BUFF;
5321 PERL_UNUSED_ARG(value);
5323 if (flags & RXapif_FETCH) {
5324 return reg_named_buff_fetch(rx, key, flags);
5325 } else if (flags & (RXapif_STORE | RXapif_DELETE | RXapif_CLEAR)) {
5326 Perl_croak_no_modify(aTHX);
5328 } else if (flags & RXapif_EXISTS) {
5329 return reg_named_buff_exists(rx, key, flags)
5332 } else if (flags & RXapif_REGNAMES) {
5333 return reg_named_buff_all(rx, flags);
5334 } else if (flags & (RXapif_SCALAR | RXapif_REGNAMES_COUNT)) {
5335 return reg_named_buff_scalar(rx, flags);
5337 Perl_croak(aTHX_ "panic: Unknown flags %d in named_buff", (int)flags);
5343 Perl_reg_named_buff_iter(pTHX_ REGEXP * const rx, const SV * const lastkey,
5346 PERL_ARGS_ASSERT_REG_NAMED_BUFF_ITER;
5347 PERL_UNUSED_ARG(lastkey);
5349 if (flags & RXapif_FIRSTKEY)
5350 return reg_named_buff_firstkey(rx, flags);
5351 else if (flags & RXapif_NEXTKEY)
5352 return reg_named_buff_nextkey(rx, flags);
5354 Perl_croak(aTHX_ "panic: Unknown flags %d in named_buff_iter", (int)flags);
5360 Perl_reg_named_buff_fetch(pTHX_ REGEXP * const r, SV * const namesv,
5363 AV *retarray = NULL;
5365 struct regexp *const rx = (struct regexp *)SvANY(r);
5367 PERL_ARGS_ASSERT_REG_NAMED_BUFF_FETCH;
5369 if (flags & RXapif_ALL)
5372 if (rx && RXp_PAREN_NAMES(rx)) {
5373 HE *he_str = hv_fetch_ent( RXp_PAREN_NAMES(rx), namesv, 0, 0 );
5376 SV* sv_dat=HeVAL(he_str);
5377 I32 *nums=(I32*)SvPVX(sv_dat);
5378 for ( i=0; i<SvIVX(sv_dat); i++ ) {
5379 if ((I32)(rx->nparens) >= nums[i]
5380 && rx->offs[nums[i]].start != -1
5381 && rx->offs[nums[i]].end != -1)
5384 CALLREG_NUMBUF_FETCH(r,nums[i],ret);
5388 ret = newSVsv(&PL_sv_undef);
5391 av_push(retarray, ret);
5394 return newRV_noinc(MUTABLE_SV(retarray));
5401 Perl_reg_named_buff_exists(pTHX_ REGEXP * const r, SV * const key,
5404 struct regexp *const rx = (struct regexp *)SvANY(r);
5406 PERL_ARGS_ASSERT_REG_NAMED_BUFF_EXISTS;
5408 if (rx && RXp_PAREN_NAMES(rx)) {
5409 if (flags & RXapif_ALL) {
5410 return hv_exists_ent(RXp_PAREN_NAMES(rx), key, 0);
5412 SV *sv = CALLREG_NAMED_BUFF_FETCH(r, key, flags);
5426 Perl_reg_named_buff_firstkey(pTHX_ REGEXP * const r, const U32 flags)
5428 struct regexp *const rx = (struct regexp *)SvANY(r);
5430 PERL_ARGS_ASSERT_REG_NAMED_BUFF_FIRSTKEY;
5432 if ( rx && RXp_PAREN_NAMES(rx) ) {
5433 (void)hv_iterinit(RXp_PAREN_NAMES(rx));
5435 return CALLREG_NAMED_BUFF_NEXTKEY(r, NULL, flags & ~RXapif_FIRSTKEY);
5442 Perl_reg_named_buff_nextkey(pTHX_ REGEXP * const r, const U32 flags)
5444 struct regexp *const rx = (struct regexp *)SvANY(r);
5445 GET_RE_DEBUG_FLAGS_DECL;
5447 PERL_ARGS_ASSERT_REG_NAMED_BUFF_NEXTKEY;
5449 if (rx && RXp_PAREN_NAMES(rx)) {
5450 HV *hv = RXp_PAREN_NAMES(rx);
5452 while ( (temphe = hv_iternext_flags(hv,0)) ) {
5455 SV* sv_dat = HeVAL(temphe);
5456 I32 *nums = (I32*)SvPVX(sv_dat);
5457 for ( i = 0; i < SvIVX(sv_dat); i++ ) {
5458 if ((I32)(rx->lastparen) >= nums[i] &&
5459 rx->offs[nums[i]].start != -1 &&
5460 rx->offs[nums[i]].end != -1)
5466 if (parno || flags & RXapif_ALL) {
5467 return newSVhek(HeKEY_hek(temphe));
5475 Perl_reg_named_buff_scalar(pTHX_ REGEXP * const r, const U32 flags)
5480 struct regexp *const rx = (struct regexp *)SvANY(r);
5482 PERL_ARGS_ASSERT_REG_NAMED_BUFF_SCALAR;
5484 if (rx && RXp_PAREN_NAMES(rx)) {
5485 if (flags & (RXapif_ALL | RXapif_REGNAMES_COUNT)) {
5486 return newSViv(HvTOTALKEYS(RXp_PAREN_NAMES(rx)));
5487 } else if (flags & RXapif_ONE) {
5488 ret = CALLREG_NAMED_BUFF_ALL(r, (flags | RXapif_REGNAMES));
5489 av = MUTABLE_AV(SvRV(ret));
5490 length = av_len(av);
5492 return newSViv(length + 1);
5494 Perl_croak(aTHX_ "panic: Unknown flags %d in named_buff_scalar", (int)flags);
5498 return &PL_sv_undef;
5502 Perl_reg_named_buff_all(pTHX_ REGEXP * const r, const U32 flags)
5504 struct regexp *const rx = (struct regexp *)SvANY(r);
5507 PERL_ARGS_ASSERT_REG_NAMED_BUFF_ALL;
5509 if (rx && RXp_PAREN_NAMES(rx)) {
5510 HV *hv= RXp_PAREN_NAMES(rx);
5512 (void)hv_iterinit(hv);
5513 while ( (temphe = hv_iternext_flags(hv,0)) ) {
5516 SV* sv_dat = HeVAL(temphe);
5517 I32 *nums = (I32*)SvPVX(sv_dat);
5518 for ( i = 0; i < SvIVX(sv_dat); i++ ) {
5519 if ((I32)(rx->lastparen) >= nums[i] &&
5520 rx->offs[nums[i]].start != -1 &&
5521 rx->offs[nums[i]].end != -1)
5527 if (parno || flags & RXapif_ALL) {
5528 av_push(av, newSVhek(HeKEY_hek(temphe)));
5533 return newRV_noinc(MUTABLE_SV(av));
5537 Perl_reg_numbered_buff_fetch(pTHX_ REGEXP * const r, const I32 paren,
5540 struct regexp *const rx = (struct regexp *)SvANY(r);
5545 PERL_ARGS_ASSERT_REG_NUMBERED_BUFF_FETCH;
5548 sv_setsv(sv,&PL_sv_undef);
5552 if (paren == RX_BUFF_IDX_PREMATCH && rx->offs[0].start != -1) {
5554 i = rx->offs[0].start;
5558 if (paren == RX_BUFF_IDX_POSTMATCH && rx->offs[0].end != -1) {
5560 s = rx->subbeg + rx->offs[0].end;
5561 i = rx->sublen - rx->offs[0].end;
5564 if ( 0 <= paren && paren <= (I32)rx->nparens &&
5565 (s1 = rx->offs[paren].start) != -1 &&
5566 (t1 = rx->offs[paren].end) != -1)
5570 s = rx->subbeg + s1;
5572 sv_setsv(sv,&PL_sv_undef);
5575 assert(rx->sublen >= (s - rx->subbeg) + i );
5577 const int oldtainted = PL_tainted;
5579 sv_setpvn(sv, s, i);
5580 PL_tainted = oldtainted;
5581 if ( (rx->extflags & RXf_CANY_SEEN)
5582 ? (RXp_MATCH_UTF8(rx)
5583 && (!i || is_utf8_string((U8*)s, i)))
5584 : (RXp_MATCH_UTF8(rx)) )
5591 if (RXp_MATCH_TAINTED(rx)) {
5592 if (SvTYPE(sv) >= SVt_PVMG) {
5593 MAGIC* const mg = SvMAGIC(sv);
5596 SvMAGIC_set(sv, mg->mg_moremagic);
5598 if ((mgt = SvMAGIC(sv))) {
5599 mg->mg_moremagic = mgt;
5600 SvMAGIC_set(sv, mg);
5610 sv_setsv(sv,&PL_sv_undef);
5616 Perl_reg_numbered_buff_store(pTHX_ REGEXP * const rx, const I32 paren,
5617 SV const * const value)
5619 PERL_ARGS_ASSERT_REG_NUMBERED_BUFF_STORE;
5621 PERL_UNUSED_ARG(rx);
5622 PERL_UNUSED_ARG(paren);
5623 PERL_UNUSED_ARG(value);
5626 Perl_croak_no_modify(aTHX);
5630 Perl_reg_numbered_buff_length(pTHX_ REGEXP * const r, const SV * const sv,
5633 struct regexp *const rx = (struct regexp *)SvANY(r);
5637 PERL_ARGS_ASSERT_REG_NUMBERED_BUFF_LENGTH;
5639 /* Some of this code was originally in C<Perl_magic_len> in F<mg.c> */
5641 /* $` / ${^PREMATCH} */
5642 case RX_BUFF_IDX_PREMATCH:
5643 if (rx->offs[0].start != -1) {
5644 i = rx->offs[0].start;
5652 /* $' / ${^POSTMATCH} */
5653 case RX_BUFF_IDX_POSTMATCH:
5654 if (rx->offs[0].end != -1) {
5655 i = rx->sublen - rx->offs[0].end;
5657 s1 = rx->offs[0].end;
5663 /* $& / ${^MATCH}, $1, $2, ... */
5665 if (paren <= (I32)rx->nparens &&
5666 (s1 = rx->offs[paren].start) != -1 &&
5667 (t1 = rx->offs[paren].end) != -1)
5672 if (ckWARN(WARN_UNINITIALIZED))
5673 report_uninit((const SV *)sv);
5678 if (i > 0 && RXp_MATCH_UTF8(rx)) {
5679 const char * const s = rx->subbeg + s1;
5684 if (is_utf8_string_loclen((U8*)s, i, &ep, &el))
5691 Perl_reg_qr_package(pTHX_ REGEXP * const rx)
5693 PERL_ARGS_ASSERT_REG_QR_PACKAGE;
5694 PERL_UNUSED_ARG(rx);
5698 return newSVpvs("Regexp");
5701 /* Scans the name of a named buffer from the pattern.
5702 * If flags is REG_RSN_RETURN_NULL returns null.
5703 * If flags is REG_RSN_RETURN_NAME returns an SV* containing the name
5704 * If flags is REG_RSN_RETURN_DATA returns the data SV* corresponding
5705 * to the parsed name as looked up in the RExC_paren_names hash.
5706 * If there is an error throws a vFAIL().. type exception.
5709 #define REG_RSN_RETURN_NULL 0
5710 #define REG_RSN_RETURN_NAME 1
5711 #define REG_RSN_RETURN_DATA 2
5714 S_reg_scan_name(pTHX_ RExC_state_t *pRExC_state, U32 flags)
5716 char *name_start = RExC_parse;
5718 PERL_ARGS_ASSERT_REG_SCAN_NAME;
5720 if (isIDFIRST_lazy_if(RExC_parse, UTF)) {
5721 /* skip IDFIRST by using do...while */
5724 RExC_parse += UTF8SKIP(RExC_parse);
5725 } while (isALNUM_utf8((U8*)RExC_parse));
5729 } while (isALNUM(*RExC_parse));
5734 = newSVpvn_flags(name_start, (int)(RExC_parse - name_start),
5735 SVs_TEMP | (UTF ? SVf_UTF8 : 0));
5736 if ( flags == REG_RSN_RETURN_NAME)
5738 else if (flags==REG_RSN_RETURN_DATA) {
5741 if ( ! sv_name ) /* should not happen*/
5742 Perl_croak(aTHX_ "panic: no svname in reg_scan_name");
5743 if (RExC_paren_names)
5744 he_str = hv_fetch_ent( RExC_paren_names, sv_name, 0, 0 );
5746 sv_dat = HeVAL(he_str);
5748 vFAIL("Reference to nonexistent named group");
5752 Perl_croak(aTHX_ "panic: bad flag in reg_scan_name");
5759 #define DEBUG_PARSE_MSG(funcname) DEBUG_PARSE_r({ \
5760 int rem=(int)(RExC_end - RExC_parse); \
5769 if (RExC_lastparse!=RExC_parse) \
5770 PerlIO_printf(Perl_debug_log," >%.*s%-*s", \
5773 iscut ? "..." : "<" \
5776 PerlIO_printf(Perl_debug_log,"%16s",""); \
5779 num = RExC_size + 1; \
5781 num=REG_NODE_NUM(RExC_emit); \
5782 if (RExC_lastnum!=num) \
5783 PerlIO_printf(Perl_debug_log,"|%4d",num); \
5785 PerlIO_printf(Perl_debug_log,"|%4s",""); \
5786 PerlIO_printf(Perl_debug_log,"|%*s%-4s", \
5787 (int)((depth*2)), "", \
5791 RExC_lastparse=RExC_parse; \
5796 #define DEBUG_PARSE(funcname) DEBUG_PARSE_r({ \
5797 DEBUG_PARSE_MSG((funcname)); \
5798 PerlIO_printf(Perl_debug_log,"%4s","\n"); \
5800 #define DEBUG_PARSE_FMT(funcname,fmt,args) DEBUG_PARSE_r({ \
5801 DEBUG_PARSE_MSG((funcname)); \
5802 PerlIO_printf(Perl_debug_log,fmt "\n",args); \
5805 /* This section of code defines the inversion list object and its methods. The
5806 * interfaces are highly subject to change, so as much as possible is static to
5807 * this file. An inversion list is here implemented as a malloc'd C array with
5808 * some added info. More will be coming when functionality is added later.
5810 * Some of the methods should always be private to the implementation, and some
5811 * should eventually be made public */
5813 #define INVLIST_INITIAL_LEN 10
5814 #define INVLIST_ARRAY_KEY "array"
5815 #define INVLIST_MAX_KEY "max"
5816 #define INVLIST_LEN_KEY "len"
5818 PERL_STATIC_INLINE UV*
5819 S_invlist_array(pTHX_ HV* const invlist)
5821 /* Returns the pointer to the inversion list's array. Every time the
5822 * length changes, this needs to be called in case malloc or realloc moved
5825 SV** list_ptr = hv_fetchs(invlist, INVLIST_ARRAY_KEY, FALSE);
5827 PERL_ARGS_ASSERT_INVLIST_ARRAY;
5829 if (list_ptr == NULL) {
5830 Perl_croak(aTHX_ "panic: inversion list without a '%s' element",
5834 return INT2PTR(UV *, SvUV(*list_ptr));
5837 PERL_STATIC_INLINE void
5838 S_invlist_set_array(pTHX_ HV* const invlist, const UV* const array)
5840 PERL_ARGS_ASSERT_INVLIST_SET_ARRAY;
5842 /* Sets the array stored in the inversion list to the memory beginning with
5845 if (hv_stores(invlist, INVLIST_ARRAY_KEY, newSVuv(PTR2UV(array))) == NULL) {
5846 Perl_croak(aTHX_ "panic: can't store '%s' entry in inversion list",
5851 PERL_STATIC_INLINE UV
5852 S_invlist_len(pTHX_ HV* const invlist)
5854 /* Returns the current number of elements in the inversion list's array */
5856 SV** len_ptr = hv_fetchs(invlist, INVLIST_LEN_KEY, FALSE);
5858 PERL_ARGS_ASSERT_INVLIST_LEN;
5860 if (len_ptr == NULL) {
5861 Perl_croak(aTHX_ "panic: inversion list without a '%s' element",
5865 return SvUV(*len_ptr);
5868 PERL_STATIC_INLINE UV
5869 S_invlist_max(pTHX_ HV* const invlist)
5871 /* Returns the maximum number of elements storable in the inversion list's
5872 * array, without having to realloc() */
5874 SV** max_ptr = hv_fetchs(invlist, INVLIST_MAX_KEY, FALSE);
5876 PERL_ARGS_ASSERT_INVLIST_MAX;
5878 if (max_ptr == NULL) {
5879 Perl_croak(aTHX_ "panic: inversion list without a '%s' element",
5883 return SvUV(*max_ptr);
5886 PERL_STATIC_INLINE void
5887 S_invlist_set_len(pTHX_ HV* const invlist, const UV len)
5889 /* Sets the current number of elements stored in the inversion list */
5891 PERL_ARGS_ASSERT_INVLIST_SET_LEN;
5893 if (len != 0 && len > invlist_max(invlist)) {
5894 Perl_croak(aTHX_ "panic: Can't make '%s=%"UVuf"' more than %s=%"UVuf" in inversion list", INVLIST_LEN_KEY, len, INVLIST_MAX_KEY, invlist_max(invlist));
5897 if (hv_stores(invlist, INVLIST_LEN_KEY, newSVuv(len)) == NULL) {
5898 Perl_croak(aTHX_ "panic: can't store '%s' entry in inversion list",
5903 PERL_STATIC_INLINE void
5904 S_invlist_set_max(pTHX_ HV* const invlist, const UV max)
5907 /* Sets the maximum number of elements storable in the inversion list
5908 * without having to realloc() */
5910 PERL_ARGS_ASSERT_INVLIST_SET_MAX;
5912 if (max < invlist_len(invlist)) {
5913 Perl_croak(aTHX_ "panic: Can't make '%s=%"UVuf"' less than %s=%"UVuf" in inversion list", INVLIST_MAX_KEY, invlist_len(invlist), INVLIST_LEN_KEY, invlist_max(invlist));
5916 if (hv_stores(invlist, INVLIST_MAX_KEY, newSVuv(max)) == NULL) {
5917 Perl_croak(aTHX_ "panic: can't store '%s' entry in inversion list",
5922 #ifndef PERL_IN_XSUB_RE
5924 Perl__new_invlist(pTHX_ IV initial_size)
5927 /* Return a pointer to a newly constructed inversion list, with enough
5928 * space to store 'initial_size' elements. If that number is negative, a
5929 * system default is used instead */
5931 HV* invlist = newHV();
5934 if (initial_size < 0) {
5935 initial_size = INVLIST_INITIAL_LEN;
5938 /* Allocate the initial space */
5939 Newx(list, initial_size, UV);
5940 invlist_set_array(invlist, list);
5942 /* set_len has to come before set_max, as the latter inspects the len */
5943 invlist_set_len(invlist, 0);
5944 invlist_set_max(invlist, initial_size);
5950 PERL_STATIC_INLINE void
5951 S_invlist_destroy(pTHX_ HV* const invlist)
5953 /* Inversion list destructor */
5955 SV** list_ptr = hv_fetchs(invlist, INVLIST_ARRAY_KEY, FALSE);
5957 PERL_ARGS_ASSERT_INVLIST_DESTROY;
5959 if (list_ptr != NULL) {
5960 UV *list = INT2PTR(UV *, SvUV(*list_ptr)); /* PERL_POISON needs lvalue */
5966 S_invlist_extend(pTHX_ HV* const invlist, const UV new_max)
5968 /* Change the maximum size of an inversion list (up or down) */
5972 const UV old_max = invlist_max(invlist);
5974 PERL_ARGS_ASSERT_INVLIST_EXTEND;
5976 if (old_max == new_max) { /* If a no-op */
5980 array = orig_array = invlist_array(invlist);
5981 Renew(array, new_max, UV);
5983 /* If the size change moved the list in memory, set the new one */
5984 if (array != orig_array) {
5985 invlist_set_array(invlist, array);
5988 invlist_set_max(invlist, new_max);
5992 PERL_STATIC_INLINE void
5993 S_invlist_trim(pTHX_ HV* const invlist)
5995 PERL_ARGS_ASSERT_INVLIST_TRIM;
5997 /* Change the length of the inversion list to how many entries it currently
6000 invlist_extend(invlist, invlist_len(invlist));
6003 /* An element is in an inversion list iff its index is even numbered: 0, 2, 4,
6006 #define ELEMENT_IN_INVLIST_SET(i) (! ((i) & 1))
6008 #ifndef PERL_IN_XSUB_RE
6010 Perl__append_range_to_invlist(pTHX_ HV* const invlist, const UV start, const UV end)
6012 /* Subject to change or removal. Append the range from 'start' to 'end' at
6013 * the end of the inversion list. The range must be above any existing
6016 UV* array = invlist_array(invlist);
6017 UV max = invlist_max(invlist);
6018 UV len = invlist_len(invlist);
6020 PERL_ARGS_ASSERT__APPEND_RANGE_TO_INVLIST;
6024 /* Here, the existing list is non-empty. The current max entry in the
6025 * list is generally the first value not in the set, except when the
6026 * set extends to the end of permissible values, in which case it is
6027 * the first entry in that final set, and so this call is an attempt to
6028 * append out-of-order */
6030 UV final_element = len - 1;
6031 if (array[final_element] > start
6032 || ELEMENT_IN_INVLIST_SET(final_element))
6034 Perl_croak(aTHX_ "panic: attempting to append to an inversion list, but wasn't at the end of the list");
6037 /* Here, it is a legal append. If the new range begins with the first
6038 * value not in the set, it is extending the set, so the new first
6039 * value not in the set is one greater than the newly extended range.
6041 if (array[final_element] == start) {
6042 if (end != UV_MAX) {
6043 array[final_element] = end + 1;
6046 /* But if the end is the maximum representable on the machine,
6047 * just let the range that this would extend have no end */
6048 invlist_set_len(invlist, len - 1);
6054 /* Here the new range doesn't extend any existing set. Add it */
6056 len += 2; /* Includes an element each for the start and end of range */
6058 /* If overflows the existing space, extend, which may cause the array to be
6061 invlist_extend(invlist, len);
6062 array = invlist_array(invlist);
6065 invlist_set_len(invlist, len);
6067 /* The next item on the list starts the range, the one after that is
6068 * one past the new range. */
6069 array[len - 2] = start;
6070 if (end != UV_MAX) {
6071 array[len - 1] = end + 1;
6074 /* But if the end is the maximum representable on the machine, just let
6075 * the range have no end */
6076 invlist_set_len(invlist, len - 1);
6082 S_invlist_union(pTHX_ HV* const a, HV* const b)
6084 /* Return a new inversion list which is the union of two inversion lists.
6085 * The basis for this comes from "Unicode Demystified" Chapter 13 by
6086 * Richard Gillam, published by Addison-Wesley, and explained at some
6087 * length there. The preface says to incorporate its examples into your
6088 * code at your own risk.
6090 * The algorithm is like a merge sort.
6092 * XXX A potential performance improvement is to keep track as we go along
6093 * if only one of the inputs contributes to the result, meaning the other
6094 * is a subset of that one. In that case, we can skip the final copy and
6095 * return the larger of the input lists */
6097 UV* array_a = invlist_array(a); /* a's array */
6098 UV* array_b = invlist_array(b);
6099 UV len_a = invlist_len(a); /* length of a's array */
6100 UV len_b = invlist_len(b);
6102 HV* u; /* the resulting union */
6106 UV i_a = 0; /* current index into a's array */
6110 /* running count, as explained in the algorithm source book; items are
6111 * stopped accumulating and are output when the count changes to/from 0.
6112 * The count is incremented when we start a range that's in the set, and
6113 * decremented when we start a range that's not in the set. So its range
6114 * is 0 to 2. Only when the count is zero is something not in the set.
6118 PERL_ARGS_ASSERT_INVLIST_UNION;
6120 /* Size the union for the worst case: that the sets are completely
6122 u = _new_invlist(len_a + len_b);
6123 array_u = invlist_array(u);
6125 /* Go through each list item by item, stopping when exhausted one of
6127 while (i_a < len_a && i_b < len_b) {
6128 UV cp; /* The element to potentially add to the union's array */
6129 bool cp_in_set; /* is it in the the input list's set or not */
6131 /* We need to take one or the other of the two inputs for the union.
6132 * Since we are merging two sorted lists, we take the smaller of the
6133 * next items. In case of a tie, we take the one that is in its set
6134 * first. If we took one not in the set first, it would decrement the
6135 * count, possibly to 0 which would cause it to be output as ending the
6136 * range, and the next time through we would take the same number, and
6137 * output it again as beginning the next range. By doing it the
6138 * opposite way, there is no possibility that the count will be
6139 * momentarily decremented to 0, and thus the two adjoining ranges will
6140 * be seamlessly merged. (In a tie and both are in the set or both not
6141 * in the set, it doesn't matter which we take first.) */
6142 if (array_a[i_a] < array_b[i_b]
6143 || (array_a[i_a] == array_b[i_b] && ELEMENT_IN_INVLIST_SET(i_a)))
6145 cp_in_set = ELEMENT_IN_INVLIST_SET(i_a);
6149 cp_in_set = ELEMENT_IN_INVLIST_SET(i_b);
6153 /* Here, have chosen which of the two inputs to look at. Only output
6154 * if the running count changes to/from 0, which marks the
6155 * beginning/end of a range in that's in the set */
6158 array_u[i_u++] = cp;
6165 array_u[i_u++] = cp;
6170 /* Here, we are finished going through at least one of the lists, which
6171 * means there is something remaining in at most one. We check if the list
6172 * that hasn't been exhausted is positioned such that we are in the middle
6173 * of a range in its set or not. (We are in the set if the next item in
6174 * the array marks the beginning of something not in the set) If in the
6175 * set, we decrement 'count'; if 0, there is potentially more to output.
6176 * There are four cases:
6177 * 1) Both weren't in their sets, count is 0, and remains 0. What's left
6178 * in the union is entirely from the non-exhausted set.
6179 * 2) Both were in their sets, count is 2. Nothing further should
6180 * be output, as everything that remains will be in the exhausted
6181 * list's set, hence in the union; decrementing to 1 but not 0 insures
6183 * 3) the exhausted was in its set, non-exhausted isn't, count is 1.
6184 * Nothing further should be output because the union includes
6185 * everything from the exhausted set. Not decrementing insures that.
6186 * 4) the exhausted wasn't in its set, non-exhausted is, count is 1;
6187 * decrementing to 0 insures that we look at the remainder of the
6188 * non-exhausted set */
6189 if ((i_a != len_a && ! ELEMENT_IN_INVLIST_SET(i_a))
6190 || (i_b != len_b && ! ELEMENT_IN_INVLIST_SET(i_b)))
6195 /* The final length is what we've output so far, plus what else is about to
6196 * be output. (If 'count' is non-zero, then the input list we exhausted
6197 * has everything remaining up to the machine's limit in its set, and hence
6198 * in the union, so there will be no further output. */
6201 /* At most one of the subexpressions will be non-zero */
6202 len_u += (len_a - i_a) + (len_b - i_b);
6205 /* Set result to final length, which can change the pointer to array_u, so
6207 if (len_u != invlist_len(u)) {
6208 invlist_set_len(u, len_u);
6210 array_u = invlist_array(u);
6213 /* When 'count' is 0, the list that was exhausted (if one was shorter than
6214 * the other) ended with everything above it not in its set. That means
6215 * that the remaining part of the union is precisely the same as the
6216 * non-exhausted list, so can just copy it unchanged. (If both list were
6217 * exhausted at the same time, then the operations below will be both 0.)
6220 IV copy_count; /* At most one will have a non-zero copy count */
6221 if ((copy_count = len_a - i_a) > 0) {
6222 Copy(array_a + i_a, array_u + i_u, copy_count, UV);
6224 else if ((copy_count = len_b - i_b) > 0) {
6225 Copy(array_b + i_b, array_u + i_u, copy_count, UV);
6233 S_invlist_intersection(pTHX_ HV* const a, HV* const b)
6235 /* Return the intersection of two inversion lists. The basis for this
6236 * comes from "Unicode Demystified" Chapter 13 by Richard Gillam, published
6237 * by Addison-Wesley, and explained at some length there. The preface says
6238 * to incorporate its examples into your code at your own risk.
6240 * The algorithm is like a merge sort, and is essentially the same as the
6244 UV* array_a = invlist_array(a); /* a's array */
6245 UV* array_b = invlist_array(b);
6246 UV len_a = invlist_len(a); /* length of a's array */
6247 UV len_b = invlist_len(b);
6249 HV* r; /* the resulting intersection */
6253 UV i_a = 0; /* current index into a's array */
6257 /* running count, as explained in the algorithm source book; items are
6258 * stopped accumulating and are output when the count changes to/from 2.
6259 * The count is incremented when we start a range that's in the set, and
6260 * decremented when we start a range that's not in the set. So its range
6261 * is 0 to 2. Only when the count is 2 is something in the intersection.
6265 PERL_ARGS_ASSERT_INVLIST_INTERSECTION;
6267 /* Size the intersection for the worst case: that the intersection ends up
6268 * fragmenting everything to be completely disjoint */
6269 r= _new_invlist(len_a + len_b);
6270 array_r = invlist_array(r);
6272 /* Go through each list item by item, stopping when exhausted one of
6274 while (i_a < len_a && i_b < len_b) {
6275 UV cp; /* The element to potentially add to the intersection's
6277 bool cp_in_set; /* Is it in the input list's set or not */
6279 /* We need to take one or the other of the two inputs for the union.
6280 * Since we are merging two sorted lists, we take the smaller of the
6281 * next items. In case of a tie, we take the one that is not in its
6282 * set first (a difference from the union algorithm). If we took one
6283 * in the set first, it would increment the count, possibly to 2 which
6284 * would cause it to be output as starting a range in the intersection,
6285 * and the next time through we would take that same number, and output
6286 * it again as ending the set. By doing it the opposite of this, we
6287 * there is no possibility that the count will be momentarily
6288 * incremented to 2. (In a tie and both are in the set or both not in
6289 * the set, it doesn't matter which we take first.) */
6290 if (array_a[i_a] < array_b[i_b]
6291 || (array_a[i_a] == array_b[i_b] && ! ELEMENT_IN_INVLIST_SET(i_a)))
6293 cp_in_set = ELEMENT_IN_INVLIST_SET(i_a);
6297 cp_in_set = ELEMENT_IN_INVLIST_SET(i_b);
6301 /* Here, have chosen which of the two inputs to look at. Only output
6302 * if the running count changes to/from 2, which marks the
6303 * beginning/end of a range that's in the intersection */
6307 array_r[i_r++] = cp;
6312 array_r[i_r++] = cp;
6318 /* Here, we are finished going through at least one of the sets, which
6319 * means there is something remaining in at most one. See the comments in
6321 if ((i_a != len_a && ! ELEMENT_IN_INVLIST_SET(i_a))
6322 || (i_b != len_b && ! ELEMENT_IN_INVLIST_SET(i_b)))
6327 /* The final length is what we've output so far plus what else is in the
6328 * intersection. Only one of the subexpressions below will be non-zero */
6331 len_r += (len_a - i_a) + (len_b - i_b);
6334 /* Set result to final length, which can change the pointer to array_r, so
6336 if (len_r != invlist_len(r)) {
6337 invlist_set_len(r, len_r);
6339 array_r = invlist_array(r);
6342 /* Finish outputting any remaining */
6343 if (count == 2) { /* Only one of will have a non-zero copy count */
6345 if ((copy_count = len_a - i_a) > 0) {
6346 Copy(array_a + i_a, array_r + i_r, copy_count, UV);
6348 else if ((copy_count = len_b - i_b) > 0) {
6349 Copy(array_b + i_b, array_r + i_r, copy_count, UV);
6357 S_add_range_to_invlist(pTHX_ HV* invlist, const UV start, const UV end)
6359 /* Add the range from 'start' to 'end' inclusive to the inversion list's
6360 * set. A pointer to the inversion list is returned. This may actually be
6361 * a new list, in which case the passed in one has been destroyed. The
6362 * passed in inversion list can be NULL, in which case a new one is created
6363 * with just the one range in it */
6369 if (invlist == NULL) {
6370 invlist = _new_invlist(2);
6374 len = invlist_len(invlist);
6377 /* If comes after the final entry, can just append it to the end */
6379 || start >= invlist_array(invlist)
6380 [invlist_len(invlist) - 1])
6382 _append_range_to_invlist(invlist, start, end);
6386 /* Here, can't just append things, create and return a new inversion list
6387 * which is the union of this range and the existing inversion list */
6388 range_invlist = _new_invlist(2);
6389 _append_range_to_invlist(range_invlist, start, end);
6391 added_invlist = invlist_union(invlist, range_invlist);
6393 /* The passed in list can be freed, as well as our temporary */
6394 invlist_destroy(range_invlist);
6395 if (invlist != added_invlist) {
6396 invlist_destroy(invlist);
6399 return added_invlist;
6402 PERL_STATIC_INLINE HV*
6403 S_add_cp_to_invlist(pTHX_ HV* invlist, const UV cp) {
6404 return add_range_to_invlist(invlist, cp, cp);
6407 /* End of inversion list object */
6410 - reg - regular expression, i.e. main body or parenthesized thing
6412 * Caller must absorb opening parenthesis.
6414 * Combining parenthesis handling with the base level of regular expression
6415 * is a trifle forced, but the need to tie the tails of the branches to what
6416 * follows makes it hard to avoid.
6418 #define REGTAIL(x,y,z) regtail((x),(y),(z),depth+1)
6420 #define REGTAIL_STUDY(x,y,z) regtail_study((x),(y),(z),depth+1)
6422 #define REGTAIL_STUDY(x,y,z) regtail((x),(y),(z),depth+1)
6426 S_reg(pTHX_ RExC_state_t *pRExC_state, I32 paren, I32 *flagp,U32 depth)
6427 /* paren: Parenthesized? 0=top, 1=(, inside: changed to letter. */
6430 register regnode *ret; /* Will be the head of the group. */
6431 register regnode *br;
6432 register regnode *lastbr;
6433 register regnode *ender = NULL;
6434 register I32 parno = 0;
6436 U32 oregflags = RExC_flags;
6437 bool have_branch = 0;
6439 I32 freeze_paren = 0;
6440 I32 after_freeze = 0;
6442 /* for (?g), (?gc), and (?o) warnings; warning
6443 about (?c) will warn about (?g) -- japhy */
6445 #define WASTED_O 0x01
6446 #define WASTED_G 0x02
6447 #define WASTED_C 0x04
6448 #define WASTED_GC (0x02|0x04)
6449 I32 wastedflags = 0x00;
6451 char * parse_start = RExC_parse; /* MJD */
6452 char * const oregcomp_parse = RExC_parse;
6454 GET_RE_DEBUG_FLAGS_DECL;
6456 PERL_ARGS_ASSERT_REG;
6457 DEBUG_PARSE("reg ");
6459 *flagp = 0; /* Tentatively. */
6462 /* Make an OPEN node, if parenthesized. */
6464 if ( *RExC_parse == '*') { /* (*VERB:ARG) */
6465 char *start_verb = RExC_parse;
6466 STRLEN verb_len = 0;
6467 char *start_arg = NULL;
6468 unsigned char op = 0;
6470 int internal_argval = 0; /* internal_argval is only useful if !argok */
6471 while ( *RExC_parse && *RExC_parse != ')' ) {
6472 if ( *RExC_parse == ':' ) {
6473 start_arg = RExC_parse + 1;
6479 verb_len = RExC_parse - start_verb;
6482 while ( *RExC_parse && *RExC_parse != ')' )
6484 if ( *RExC_parse != ')' )
6485 vFAIL("Unterminated verb pattern argument");
6486 if ( RExC_parse == start_arg )
6489 if ( *RExC_parse != ')' )
6490 vFAIL("Unterminated verb pattern");
6493 switch ( *start_verb ) {
6494 case 'A': /* (*ACCEPT) */
6495 if ( memEQs(start_verb,verb_len,"ACCEPT") ) {
6497 internal_argval = RExC_nestroot;
6500 case 'C': /* (*COMMIT) */
6501 if ( memEQs(start_verb,verb_len,"COMMIT") )
6504 case 'F': /* (*FAIL) */
6505 if ( verb_len==1 || memEQs(start_verb,verb_len,"FAIL") ) {
6510 case ':': /* (*:NAME) */
6511 case 'M': /* (*MARK:NAME) */
6512 if ( verb_len==0 || memEQs(start_verb,verb_len,"MARK") ) {
6517 case 'P': /* (*PRUNE) */
6518 if ( memEQs(start_verb,verb_len,"PRUNE") )
6521 case 'S': /* (*SKIP) */
6522 if ( memEQs(start_verb,verb_len,"SKIP") )
6525 case 'T': /* (*THEN) */
6526 /* [19:06] <TimToady> :: is then */
6527 if ( memEQs(start_verb,verb_len,"THEN") ) {
6529 RExC_seen |= REG_SEEN_CUTGROUP;
6535 vFAIL3("Unknown verb pattern '%.*s'",
6536 verb_len, start_verb);
6539 if ( start_arg && internal_argval ) {
6540 vFAIL3("Verb pattern '%.*s' may not have an argument",
6541 verb_len, start_verb);
6542 } else if ( argok < 0 && !start_arg ) {
6543 vFAIL3("Verb pattern '%.*s' has a mandatory argument",
6544 verb_len, start_verb);
6546 ret = reganode(pRExC_state, op, internal_argval);
6547 if ( ! internal_argval && ! SIZE_ONLY ) {
6549 SV *sv = newSVpvn( start_arg, RExC_parse - start_arg);
6550 ARG(ret) = add_data( pRExC_state, 1, "S" );
6551 RExC_rxi->data->data[ARG(ret)]=(void*)sv;
6558 if (!internal_argval)
6559 RExC_seen |= REG_SEEN_VERBARG;
6560 } else if ( start_arg ) {
6561 vFAIL3("Verb pattern '%.*s' may not have an argument",
6562 verb_len, start_verb);
6564 ret = reg_node(pRExC_state, op);
6566 nextchar(pRExC_state);
6569 if (*RExC_parse == '?') { /* (?...) */
6570 bool is_logical = 0;
6571 const char * const seqstart = RExC_parse;
6572 bool has_use_defaults = FALSE;
6575 paren = *RExC_parse++;
6576 ret = NULL; /* For look-ahead/behind. */
6579 case 'P': /* (?P...) variants for those used to PCRE/Python */
6580 paren = *RExC_parse++;
6581 if ( paren == '<') /* (?P<...>) named capture */
6583 else if (paren == '>') { /* (?P>name) named recursion */
6584 goto named_recursion;
6586 else if (paren == '=') { /* (?P=...) named backref */
6587 /* this pretty much dupes the code for \k<NAME> in regatom(), if
6588 you change this make sure you change that */
6589 char* name_start = RExC_parse;
6591 SV *sv_dat = reg_scan_name(pRExC_state,
6592 SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
6593 if (RExC_parse == name_start || *RExC_parse != ')')
6594 vFAIL2("Sequence %.3s... not terminated",parse_start);
6597 num = add_data( pRExC_state, 1, "S" );
6598 RExC_rxi->data->data[num]=(void*)sv_dat;
6599 SvREFCNT_inc_simple_void(sv_dat);
6602 ret = reganode(pRExC_state,
6605 : (MORE_ASCII_RESTRICTED)
6607 : (AT_LEAST_UNI_SEMANTICS)
6615 Set_Node_Offset(ret, parse_start+1);
6616 Set_Node_Cur_Length(ret); /* MJD */
6618 nextchar(pRExC_state);
6622 vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
6624 case '<': /* (?<...) */
6625 if (*RExC_parse == '!')
6627 else if (*RExC_parse != '=')
6633 case '\'': /* (?'...') */
6634 name_start= RExC_parse;
6635 svname = reg_scan_name(pRExC_state,
6636 SIZE_ONLY ? /* reverse test from the others */
6637 REG_RSN_RETURN_NAME :
6638 REG_RSN_RETURN_NULL);
6639 if (RExC_parse == name_start) {
6641 vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
6644 if (*RExC_parse != paren)
6645 vFAIL2("Sequence (?%c... not terminated",
6646 paren=='>' ? '<' : paren);
6650 if (!svname) /* shouldn't happen */
6652 "panic: reg_scan_name returned NULL");
6653 if (!RExC_paren_names) {
6654 RExC_paren_names= newHV();
6655 sv_2mortal(MUTABLE_SV(RExC_paren_names));
6657 RExC_paren_name_list= newAV();
6658 sv_2mortal(MUTABLE_SV(RExC_paren_name_list));
6661 he_str = hv_fetch_ent( RExC_paren_names, svname, 1, 0 );
6663 sv_dat = HeVAL(he_str);
6665 /* croak baby croak */
6667 "panic: paren_name hash element allocation failed");
6668 } else if ( SvPOK(sv_dat) ) {
6669 /* (?|...) can mean we have dupes so scan to check
6670 its already been stored. Maybe a flag indicating
6671 we are inside such a construct would be useful,
6672 but the arrays are likely to be quite small, so
6673 for now we punt -- dmq */
6674 IV count = SvIV(sv_dat);
6675 I32 *pv = (I32*)SvPVX(sv_dat);
6677 for ( i = 0 ; i < count ; i++ ) {
6678 if ( pv[i] == RExC_npar ) {
6684 pv = (I32*)SvGROW(sv_dat, SvCUR(sv_dat) + sizeof(I32)+1);
6685 SvCUR_set(sv_dat, SvCUR(sv_dat) + sizeof(I32));
6686 pv[count] = RExC_npar;
6687 SvIV_set(sv_dat, SvIVX(sv_dat) + 1);
6690 (void)SvUPGRADE(sv_dat,SVt_PVNV);
6691 sv_setpvn(sv_dat, (char *)&(RExC_npar), sizeof(I32));
6693 SvIV_set(sv_dat, 1);
6696 if (!av_store(RExC_paren_name_list, RExC_npar, SvREFCNT_inc(svname)))
6697 SvREFCNT_dec(svname);
6700 /*sv_dump(sv_dat);*/
6702 nextchar(pRExC_state);
6704 goto capturing_parens;
6706 RExC_seen |= REG_SEEN_LOOKBEHIND;
6707 RExC_in_lookbehind++;
6709 case '=': /* (?=...) */
6710 RExC_seen_zerolen++;
6712 case '!': /* (?!...) */
6713 RExC_seen_zerolen++;
6714 if (*RExC_parse == ')') {
6715 ret=reg_node(pRExC_state, OPFAIL);
6716 nextchar(pRExC_state);
6720 case '|': /* (?|...) */
6721 /* branch reset, behave like a (?:...) except that
6722 buffers in alternations share the same numbers */
6724 after_freeze = freeze_paren = RExC_npar;
6726 case ':': /* (?:...) */
6727 case '>': /* (?>...) */
6729 case '$': /* (?$...) */
6730 case '@': /* (?@...) */
6731 vFAIL2("Sequence (?%c...) not implemented", (int)paren);
6733 case '#': /* (?#...) */
6734 while (*RExC_parse && *RExC_parse != ')')
6736 if (*RExC_parse != ')')
6737 FAIL("Sequence (?#... not terminated");
6738 nextchar(pRExC_state);
6741 case '0' : /* (?0) */
6742 case 'R' : /* (?R) */
6743 if (*RExC_parse != ')')
6744 FAIL("Sequence (?R) not terminated");
6745 ret = reg_node(pRExC_state, GOSTART);
6746 *flagp |= POSTPONED;
6747 nextchar(pRExC_state);
6750 { /* named and numeric backreferences */
6752 case '&': /* (?&NAME) */
6753 parse_start = RExC_parse - 1;
6756 SV *sv_dat = reg_scan_name(pRExC_state,
6757 SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
6758 num = sv_dat ? *((I32 *)SvPVX(sv_dat)) : 0;
6760 goto gen_recurse_regop;
6763 if (!(RExC_parse[0] >= '1' && RExC_parse[0] <= '9')) {
6765 vFAIL("Illegal pattern");
6767 goto parse_recursion;
6769 case '-': /* (?-1) */
6770 if (!(RExC_parse[0] >= '1' && RExC_parse[0] <= '9')) {
6771 RExC_parse--; /* rewind to let it be handled later */
6775 case '1': case '2': case '3': case '4': /* (?1) */
6776 case '5': case '6': case '7': case '8': case '9':
6779 num = atoi(RExC_parse);
6780 parse_start = RExC_parse - 1; /* MJD */
6781 if (*RExC_parse == '-')
6783 while (isDIGIT(*RExC_parse))
6785 if (*RExC_parse!=')')
6786 vFAIL("Expecting close bracket");
6789 if ( paren == '-' ) {
6791 Diagram of capture buffer numbering.
6792 Top line is the normal capture buffer numbers
6793 Bottom line is the negative indexing as from
6797 /(a(x)y)(a(b(c(?-2)d)e)f)(g(h))/
6801 num = RExC_npar + num;
6804 vFAIL("Reference to nonexistent group");
6806 } else if ( paren == '+' ) {
6807 num = RExC_npar + num - 1;
6810 ret = reganode(pRExC_state, GOSUB, num);
6812 if (num > (I32)RExC_rx->nparens) {
6814 vFAIL("Reference to nonexistent group");
6816 ARG2L_SET( ret, RExC_recurse_count++);
6818 DEBUG_OPTIMISE_MORE_r(PerlIO_printf(Perl_debug_log,
6819 "Recurse #%"UVuf" to %"IVdf"\n", (UV)ARG(ret), (IV)ARG2L(ret)));
6823 RExC_seen |= REG_SEEN_RECURSE;
6824 Set_Node_Length(ret, 1 + regarglen[OP(ret)]); /* MJD */
6825 Set_Node_Offset(ret, parse_start); /* MJD */
6827 *flagp |= POSTPONED;
6828 nextchar(pRExC_state);
6830 } /* named and numeric backreferences */
6833 case '?': /* (??...) */
6835 if (*RExC_parse != '{') {
6837 vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
6840 *flagp |= POSTPONED;
6841 paren = *RExC_parse++;
6843 case '{': /* (?{...}) */
6848 char *s = RExC_parse;
6850 RExC_seen_zerolen++;
6851 RExC_seen |= REG_SEEN_EVAL;
6852 while (count && (c = *RExC_parse)) {
6863 if (*RExC_parse != ')') {
6865 vFAIL("Sequence (?{...}) not terminated or not {}-balanced");
6869 OP_4tree *sop, *rop;
6870 SV * const sv = newSVpvn(s, RExC_parse - 1 - s);
6873 Perl_save_re_context(aTHX);
6874 rop = Perl_sv_compile_2op_is_broken(aTHX_ sv, &sop, "re", &pad);
6875 sop->op_private |= OPpREFCOUNTED;
6876 /* re_dup will OpREFCNT_inc */
6877 OpREFCNT_set(sop, 1);
6880 n = add_data(pRExC_state, 3, "nop");
6881 RExC_rxi->data->data[n] = (void*)rop;
6882 RExC_rxi->data->data[n+1] = (void*)sop;
6883 RExC_rxi->data->data[n+2] = (void*)pad;
6886 else { /* First pass */
6887 if (PL_reginterp_cnt < ++RExC_seen_evals
6889 /* No compiled RE interpolated, has runtime
6890 components ===> unsafe. */
6891 FAIL("Eval-group not allowed at runtime, use re 'eval'");
6892 if (PL_tainting && PL_tainted)
6893 FAIL("Eval-group in insecure regular expression");
6894 #if PERL_VERSION > 8
6895 if (IN_PERL_COMPILETIME)
6900 nextchar(pRExC_state);
6902 ret = reg_node(pRExC_state, LOGICAL);
6905 REGTAIL(pRExC_state, ret, reganode(pRExC_state, EVAL, n));
6906 /* deal with the length of this later - MJD */
6909 ret = reganode(pRExC_state, EVAL, n);
6910 Set_Node_Length(ret, RExC_parse - parse_start + 1);
6911 Set_Node_Offset(ret, parse_start);
6914 case '(': /* (?(?{...})...) and (?(?=...)...) */
6917 if (RExC_parse[0] == '?') { /* (?(?...)) */
6918 if (RExC_parse[1] == '=' || RExC_parse[1] == '!'
6919 || RExC_parse[1] == '<'
6920 || RExC_parse[1] == '{') { /* Lookahead or eval. */
6923 ret = reg_node(pRExC_state, LOGICAL);
6926 REGTAIL(pRExC_state, ret, reg(pRExC_state, 1, &flag,depth+1));
6930 else if ( RExC_parse[0] == '<' /* (?(<NAME>)...) */
6931 || RExC_parse[0] == '\'' ) /* (?('NAME')...) */
6933 char ch = RExC_parse[0] == '<' ? '>' : '\'';
6934 char *name_start= RExC_parse++;
6936 SV *sv_dat=reg_scan_name(pRExC_state,
6937 SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
6938 if (RExC_parse == name_start || *RExC_parse != ch)
6939 vFAIL2("Sequence (?(%c... not terminated",
6940 (ch == '>' ? '<' : ch));
6943 num = add_data( pRExC_state, 1, "S" );
6944 RExC_rxi->data->data[num]=(void*)sv_dat;
6945 SvREFCNT_inc_simple_void(sv_dat);
6947 ret = reganode(pRExC_state,NGROUPP,num);
6948 goto insert_if_check_paren;
6950 else if (RExC_parse[0] == 'D' &&
6951 RExC_parse[1] == 'E' &&
6952 RExC_parse[2] == 'F' &&
6953 RExC_parse[3] == 'I' &&
6954 RExC_parse[4] == 'N' &&
6955 RExC_parse[5] == 'E')
6957 ret = reganode(pRExC_state,DEFINEP,0);
6960 goto insert_if_check_paren;
6962 else if (RExC_parse[0] == 'R') {
6965 if (RExC_parse[0] >= '1' && RExC_parse[0] <= '9' ) {
6966 parno = atoi(RExC_parse++);
6967 while (isDIGIT(*RExC_parse))
6969 } else if (RExC_parse[0] == '&') {
6972 sv_dat = reg_scan_name(pRExC_state,
6973 SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
6974 parno = sv_dat ? *((I32 *)SvPVX(sv_dat)) : 0;
6976 ret = reganode(pRExC_state,INSUBP,parno);
6977 goto insert_if_check_paren;
6979 else if (RExC_parse[0] >= '1' && RExC_parse[0] <= '9' ) {
6982 parno = atoi(RExC_parse++);
6984 while (isDIGIT(*RExC_parse))
6986 ret = reganode(pRExC_state, GROUPP, parno);
6988 insert_if_check_paren:
6989 if ((c = *nextchar(pRExC_state)) != ')')
6990 vFAIL("Switch condition not recognized");
6992 REGTAIL(pRExC_state, ret, reganode(pRExC_state, IFTHEN, 0));
6993 br = regbranch(pRExC_state, &flags, 1,depth+1);
6995 br = reganode(pRExC_state, LONGJMP, 0);
6997 REGTAIL(pRExC_state, br, reganode(pRExC_state, LONGJMP, 0));
6998 c = *nextchar(pRExC_state);
7003 vFAIL("(?(DEFINE)....) does not allow branches");
7004 lastbr = reganode(pRExC_state, IFTHEN, 0); /* Fake one for optimizer. */
7005 regbranch(pRExC_state, &flags, 1,depth+1);
7006 REGTAIL(pRExC_state, ret, lastbr);
7009 c = *nextchar(pRExC_state);
7014 vFAIL("Switch (?(condition)... contains too many branches");
7015 ender = reg_node(pRExC_state, TAIL);
7016 REGTAIL(pRExC_state, br, ender);
7018 REGTAIL(pRExC_state, lastbr, ender);
7019 REGTAIL(pRExC_state, NEXTOPER(NEXTOPER(lastbr)), ender);
7022 REGTAIL(pRExC_state, ret, ender);
7023 RExC_size++; /* XXX WHY do we need this?!!
7024 For large programs it seems to be required
7025 but I can't figure out why. -- dmq*/
7029 vFAIL2("Unknown switch condition (?(%.2s", RExC_parse);
7033 RExC_parse--; /* for vFAIL to print correctly */
7034 vFAIL("Sequence (? incomplete");
7036 case DEFAULT_PAT_MOD: /* Use default flags with the exceptions
7038 has_use_defaults = TRUE;
7039 STD_PMMOD_FLAGS_CLEAR(&RExC_flags);
7040 set_regex_charset(&RExC_flags, (RExC_utf8 || RExC_uni_semantics)
7041 ? REGEX_UNICODE_CHARSET
7042 : REGEX_DEPENDS_CHARSET);
7046 parse_flags: /* (?i) */
7048 U32 posflags = 0, negflags = 0;
7049 U32 *flagsp = &posflags;
7050 bool has_charset_modifier = 0;
7051 regex_charset cs = (RExC_utf8 || RExC_uni_semantics)
7052 ? REGEX_UNICODE_CHARSET
7053 : REGEX_DEPENDS_CHARSET;
7055 while (*RExC_parse) {
7056 /* && strchr("iogcmsx", *RExC_parse) */
7057 /* (?g), (?gc) and (?o) are useless here
7058 and must be globally applied -- japhy */
7059 switch (*RExC_parse) {
7060 CASE_STD_PMMOD_FLAGS_PARSE_SET(flagsp);
7061 case LOCALE_PAT_MOD:
7062 if (has_charset_modifier || flagsp == &negflags) {
7063 goto fail_modifiers;
7065 cs = REGEX_LOCALE_CHARSET;
7066 has_charset_modifier = 1;
7067 RExC_contains_locale = 1;
7069 case UNICODE_PAT_MOD:
7070 if (has_charset_modifier || flagsp == &negflags) {
7071 goto fail_modifiers;
7073 cs = REGEX_UNICODE_CHARSET;
7074 has_charset_modifier = 1;
7076 case ASCII_RESTRICT_PAT_MOD:
7077 if (has_charset_modifier || flagsp == &negflags) {
7078 goto fail_modifiers;
7080 if (*(RExC_parse + 1) == ASCII_RESTRICT_PAT_MOD) {
7081 /* Doubled modifier implies more restricted */
7082 cs = REGEX_ASCII_MORE_RESTRICTED_CHARSET;
7086 cs = REGEX_ASCII_RESTRICTED_CHARSET;
7088 has_charset_modifier = 1;
7090 case DEPENDS_PAT_MOD:
7091 if (has_use_defaults
7092 || has_charset_modifier
7093 || flagsp == &negflags)
7095 goto fail_modifiers;
7098 /* The dual charset means unicode semantics if the
7099 * pattern (or target, not known until runtime) are
7100 * utf8, or something in the pattern indicates unicode
7102 cs = (RExC_utf8 || RExC_uni_semantics)
7103 ? REGEX_UNICODE_CHARSET
7104 : REGEX_DEPENDS_CHARSET;
7105 has_charset_modifier = 1;
7107 case ONCE_PAT_MOD: /* 'o' */
7108 case GLOBAL_PAT_MOD: /* 'g' */
7109 if (SIZE_ONLY && ckWARN(WARN_REGEXP)) {
7110 const I32 wflagbit = *RExC_parse == 'o' ? WASTED_O : WASTED_G;
7111 if (! (wastedflags & wflagbit) ) {
7112 wastedflags |= wflagbit;
7115 "Useless (%s%c) - %suse /%c modifier",
7116 flagsp == &negflags ? "?-" : "?",
7118 flagsp == &negflags ? "don't " : "",
7125 case CONTINUE_PAT_MOD: /* 'c' */
7126 if (SIZE_ONLY && ckWARN(WARN_REGEXP)) {
7127 if (! (wastedflags & WASTED_C) ) {
7128 wastedflags |= WASTED_GC;
7131 "Useless (%sc) - %suse /gc modifier",
7132 flagsp == &negflags ? "?-" : "?",
7133 flagsp == &negflags ? "don't " : ""
7138 case KEEPCOPY_PAT_MOD: /* 'p' */
7139 if (flagsp == &negflags) {
7141 ckWARNreg(RExC_parse + 1,"Useless use of (?-p)");
7143 *flagsp |= RXf_PMf_KEEPCOPY;
7147 /* A flag is a default iff it is following a minus, so
7148 * if there is a minus, it means will be trying to
7149 * re-specify a default which is an error */
7150 if (has_use_defaults || flagsp == &negflags) {
7153 vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
7157 wastedflags = 0; /* reset so (?g-c) warns twice */
7163 RExC_flags |= posflags;
7164 RExC_flags &= ~negflags;
7165 set_regex_charset(&RExC_flags, cs);
7167 oregflags |= posflags;
7168 oregflags &= ~negflags;
7169 set_regex_charset(&oregflags, cs);
7171 nextchar(pRExC_state);
7182 vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
7187 }} /* one for the default block, one for the switch */
7194 ret = reganode(pRExC_state, OPEN, parno);
7197 RExC_nestroot = parno;
7198 if (RExC_seen & REG_SEEN_RECURSE
7199 && !RExC_open_parens[parno-1])
7201 DEBUG_OPTIMISE_MORE_r(PerlIO_printf(Perl_debug_log,
7202 "Setting open paren #%"IVdf" to %d\n",
7203 (IV)parno, REG_NODE_NUM(ret)));
7204 RExC_open_parens[parno-1]= ret;
7207 Set_Node_Length(ret, 1); /* MJD */
7208 Set_Node_Offset(ret, RExC_parse); /* MJD */
7216 /* Pick up the branches, linking them together. */
7217 parse_start = RExC_parse; /* MJD */
7218 br = regbranch(pRExC_state, &flags, 1,depth+1);
7220 /* branch_len = (paren != 0); */
7224 if (*RExC_parse == '|') {
7225 if (!SIZE_ONLY && RExC_extralen) {
7226 reginsert(pRExC_state, BRANCHJ, br, depth+1);
7229 reginsert(pRExC_state, BRANCH, br, depth+1);
7230 Set_Node_Length(br, paren != 0);
7231 Set_Node_Offset_To_R(br-RExC_emit_start, parse_start-RExC_start);
7235 RExC_extralen += 1; /* For BRANCHJ-BRANCH. */
7237 else if (paren == ':') {
7238 *flagp |= flags&SIMPLE;
7240 if (is_open) { /* Starts with OPEN. */
7241 REGTAIL(pRExC_state, ret, br); /* OPEN -> first. */
7243 else if (paren != '?') /* Not Conditional */
7245 *flagp |= flags & (SPSTART | HASWIDTH | POSTPONED);
7247 while (*RExC_parse == '|') {
7248 if (!SIZE_ONLY && RExC_extralen) {
7249 ender = reganode(pRExC_state, LONGJMP,0);
7250 REGTAIL(pRExC_state, NEXTOPER(NEXTOPER(lastbr)), ender); /* Append to the previous. */
7253 RExC_extralen += 2; /* Account for LONGJMP. */
7254 nextchar(pRExC_state);
7256 if (RExC_npar > after_freeze)
7257 after_freeze = RExC_npar;
7258 RExC_npar = freeze_paren;
7260 br = regbranch(pRExC_state, &flags, 0, depth+1);
7264 REGTAIL(pRExC_state, lastbr, br); /* BRANCH -> BRANCH. */
7266 *flagp |= flags & (SPSTART | HASWIDTH | POSTPONED);
7269 if (have_branch || paren != ':') {
7270 /* Make a closing node, and hook it on the end. */
7273 ender = reg_node(pRExC_state, TAIL);
7276 ender = reganode(pRExC_state, CLOSE, parno);
7277 if (!SIZE_ONLY && RExC_seen & REG_SEEN_RECURSE) {
7278 DEBUG_OPTIMISE_MORE_r(PerlIO_printf(Perl_debug_log,
7279 "Setting close paren #%"IVdf" to %d\n",
7280 (IV)parno, REG_NODE_NUM(ender)));
7281 RExC_close_parens[parno-1]= ender;
7282 if (RExC_nestroot == parno)
7285 Set_Node_Offset(ender,RExC_parse+1); /* MJD */
7286 Set_Node_Length(ender,1); /* MJD */
7292 *flagp &= ~HASWIDTH;
7295 ender = reg_node(pRExC_state, SUCCEED);
7298 ender = reg_node(pRExC_state, END);
7300 assert(!RExC_opend); /* there can only be one! */
7305 REGTAIL(pRExC_state, lastbr, ender);
7307 if (have_branch && !SIZE_ONLY) {
7309 RExC_seen |= REG_TOP_LEVEL_BRANCHES;
7311 /* Hook the tails of the branches to the closing node. */
7312 for (br = ret; br; br = regnext(br)) {
7313 const U8 op = PL_regkind[OP(br)];
7315 REGTAIL_STUDY(pRExC_state, NEXTOPER(br), ender);
7317 else if (op == BRANCHJ) {
7318 REGTAIL_STUDY(pRExC_state, NEXTOPER(NEXTOPER(br)), ender);
7326 static const char parens[] = "=!<,>";
7328 if (paren && (p = strchr(parens, paren))) {
7329 U8 node = ((p - parens) % 2) ? UNLESSM : IFMATCH;
7330 int flag = (p - parens) > 1;
7333 node = SUSPEND, flag = 0;
7334 reginsert(pRExC_state, node,ret, depth+1);
7335 Set_Node_Cur_Length(ret);
7336 Set_Node_Offset(ret, parse_start + 1);
7338 REGTAIL_STUDY(pRExC_state, ret, reg_node(pRExC_state, TAIL));
7342 /* Check for proper termination. */
7344 RExC_flags = oregflags;
7345 if (RExC_parse >= RExC_end || *nextchar(pRExC_state) != ')') {
7346 RExC_parse = oregcomp_parse;
7347 vFAIL("Unmatched (");
7350 else if (!paren && RExC_parse < RExC_end) {
7351 if (*RExC_parse == ')') {
7353 vFAIL("Unmatched )");
7356 FAIL("Junk on end of regexp"); /* "Can't happen". */
7360 if (RExC_in_lookbehind) {
7361 RExC_in_lookbehind--;
7363 if (after_freeze > RExC_npar)
7364 RExC_npar = after_freeze;
7369 - regbranch - one alternative of an | operator
7371 * Implements the concatenation operator.
7374 S_regbranch(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, I32 first, U32 depth)
7377 register regnode *ret;
7378 register regnode *chain = NULL;
7379 register regnode *latest;
7380 I32 flags = 0, c = 0;
7381 GET_RE_DEBUG_FLAGS_DECL;
7383 PERL_ARGS_ASSERT_REGBRANCH;
7385 DEBUG_PARSE("brnc");
7390 if (!SIZE_ONLY && RExC_extralen)
7391 ret = reganode(pRExC_state, BRANCHJ,0);
7393 ret = reg_node(pRExC_state, BRANCH);
7394 Set_Node_Length(ret, 1);
7398 if (!first && SIZE_ONLY)
7399 RExC_extralen += 1; /* BRANCHJ */
7401 *flagp = WORST; /* Tentatively. */
7404 nextchar(pRExC_state);
7405 while (RExC_parse < RExC_end && *RExC_parse != '|' && *RExC_parse != ')') {
7407 latest = regpiece(pRExC_state, &flags,depth+1);
7408 if (latest == NULL) {
7409 if (flags & TRYAGAIN)
7413 else if (ret == NULL)
7415 *flagp |= flags&(HASWIDTH|POSTPONED);
7416 if (chain == NULL) /* First piece. */
7417 *flagp |= flags&SPSTART;
7420 REGTAIL(pRExC_state, chain, latest);
7425 if (chain == NULL) { /* Loop ran zero times. */
7426 chain = reg_node(pRExC_state, NOTHING);
7431 *flagp |= flags&SIMPLE;
7438 - regpiece - something followed by possible [*+?]
7440 * Note that the branching code sequences used for ? and the general cases
7441 * of * and + are somewhat optimized: they use the same NOTHING node as
7442 * both the endmarker for their branch list and the body of the last branch.
7443 * It might seem that this node could be dispensed with entirely, but the
7444 * endmarker role is not redundant.
7447 S_regpiece(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, U32 depth)
7450 register regnode *ret;
7452 register char *next;
7454 const char * const origparse = RExC_parse;
7456 I32 max = REG_INFTY;
7458 const char *maxpos = NULL;
7459 GET_RE_DEBUG_FLAGS_DECL;
7461 PERL_ARGS_ASSERT_REGPIECE;
7463 DEBUG_PARSE("piec");
7465 ret = regatom(pRExC_state, &flags,depth+1);
7467 if (flags & TRYAGAIN)
7474 if (op == '{' && regcurly(RExC_parse)) {
7476 parse_start = RExC_parse; /* MJD */
7477 next = RExC_parse + 1;
7478 while (isDIGIT(*next) || *next == ',') {
7487 if (*next == '}') { /* got one */
7491 min = atoi(RExC_parse);
7495 maxpos = RExC_parse;
7497 if (!max && *maxpos != '0')
7498 max = REG_INFTY; /* meaning "infinity" */
7499 else if (max >= REG_INFTY)
7500 vFAIL2("Quantifier in {,} bigger than %d", REG_INFTY - 1);
7502 nextchar(pRExC_state);
7505 if ((flags&SIMPLE)) {
7506 RExC_naughty += 2 + RExC_naughty / 2;
7507 reginsert(pRExC_state, CURLY, ret, depth+1);
7508 Set_Node_Offset(ret, parse_start+1); /* MJD */
7509 Set_Node_Cur_Length(ret);
7512 regnode * const w = reg_node(pRExC_state, WHILEM);
7515 REGTAIL(pRExC_state, ret, w);
7516 if (!SIZE_ONLY && RExC_extralen) {
7517 reginsert(pRExC_state, LONGJMP,ret, depth+1);
7518 reginsert(pRExC_state, NOTHING,ret, depth+1);
7519 NEXT_OFF(ret) = 3; /* Go over LONGJMP. */
7521 reginsert(pRExC_state, CURLYX,ret, depth+1);
7523 Set_Node_Offset(ret, parse_start+1);
7524 Set_Node_Length(ret,
7525 op == '{' ? (RExC_parse - parse_start) : 1);
7527 if (!SIZE_ONLY && RExC_extralen)
7528 NEXT_OFF(ret) = 3; /* Go over NOTHING to LONGJMP. */
7529 REGTAIL(pRExC_state, ret, reg_node(pRExC_state, NOTHING));
7531 RExC_whilem_seen++, RExC_extralen += 3;
7532 RExC_naughty += 4 + RExC_naughty; /* compound interest */
7541 vFAIL("Can't do {n,m} with n > m");
7543 ARG1_SET(ret, (U16)min);
7544 ARG2_SET(ret, (U16)max);
7556 #if 0 /* Now runtime fix should be reliable. */
7558 /* if this is reinstated, don't forget to put this back into perldiag:
7560 =item Regexp *+ operand could be empty at {#} in regex m/%s/
7562 (F) The part of the regexp subject to either the * or + quantifier
7563 could match an empty string. The {#} shows in the regular
7564 expression about where the problem was discovered.
7568 if (!(flags&HASWIDTH) && op != '?')
7569 vFAIL("Regexp *+ operand could be empty");
7572 parse_start = RExC_parse;
7573 nextchar(pRExC_state);
7575 *flagp = (op != '+') ? (WORST|SPSTART|HASWIDTH) : (WORST|HASWIDTH);
7577 if (op == '*' && (flags&SIMPLE)) {
7578 reginsert(pRExC_state, STAR, ret, depth+1);
7582 else if (op == '*') {
7586 else if (op == '+' && (flags&SIMPLE)) {
7587 reginsert(pRExC_state, PLUS, ret, depth+1);
7591 else if (op == '+') {
7595 else if (op == '?') {
7600 if (!SIZE_ONLY && !(flags&(HASWIDTH|POSTPONED)) && max > REG_INFTY/3) {
7601 ckWARN3reg(RExC_parse,
7602 "%.*s matches null string many times",
7603 (int)(RExC_parse >= origparse ? RExC_parse - origparse : 0),
7607 if (RExC_parse < RExC_end && *RExC_parse == '?') {
7608 nextchar(pRExC_state);
7609 reginsert(pRExC_state, MINMOD, ret, depth+1);
7610 REGTAIL(pRExC_state, ret, ret + NODE_STEP_REGNODE);
7612 #ifndef REG_ALLOW_MINMOD_SUSPEND
7615 if (RExC_parse < RExC_end && *RExC_parse == '+') {
7617 nextchar(pRExC_state);
7618 ender = reg_node(pRExC_state, SUCCEED);
7619 REGTAIL(pRExC_state, ret, ender);
7620 reginsert(pRExC_state, SUSPEND, ret, depth+1);
7622 ender = reg_node(pRExC_state, TAIL);
7623 REGTAIL(pRExC_state, ret, ender);
7627 if (RExC_parse < RExC_end && ISMULT2(RExC_parse)) {
7629 vFAIL("Nested quantifiers");
7636 /* reg_namedseq(pRExC_state,UVp, UV depth)
7638 This is expected to be called by a parser routine that has
7639 recognized '\N' and needs to handle the rest. RExC_parse is
7640 expected to point at the first char following the N at the time
7643 The \N may be inside (indicated by valuep not being NULL) or outside a
7646 \N may begin either a named sequence, or if outside a character class, mean
7647 to match a non-newline. For non single-quoted regexes, the tokenizer has
7648 attempted to decide which, and in the case of a named sequence converted it
7649 into one of the forms: \N{} (if the sequence is null), or \N{U+c1.c2...},
7650 where c1... are the characters in the sequence. For single-quoted regexes,
7651 the tokenizer passes the \N sequence through unchanged; this code will not
7652 attempt to determine this nor expand those. The net effect is that if the
7653 beginning of the passed-in pattern isn't '{U+' or there is no '}', it
7654 signals that this \N occurrence means to match a non-newline.
7656 Only the \N{U+...} form should occur in a character class, for the same
7657 reason that '.' inside a character class means to just match a period: it
7658 just doesn't make sense.
7660 If valuep is non-null then it is assumed that we are parsing inside
7661 of a charclass definition and the first codepoint in the resolved
7662 string is returned via *valuep and the routine will return NULL.
7663 In this mode if a multichar string is returned from the charnames
7664 handler, a warning will be issued, and only the first char in the
7665 sequence will be examined. If the string returned is zero length
7666 then the value of *valuep is undefined and NON-NULL will
7667 be returned to indicate failure. (This will NOT be a valid pointer
7670 If valuep is null then it is assumed that we are parsing normal text and a
7671 new EXACT node is inserted into the program containing the resolved string,
7672 and a pointer to the new node is returned. But if the string is zero length
7673 a NOTHING node is emitted instead.
7675 On success RExC_parse is set to the char following the endbrace.
7676 Parsing failures will generate a fatal error via vFAIL(...)
7679 S_reg_namedseq(pTHX_ RExC_state_t *pRExC_state, UV *valuep, I32 *flagp, U32 depth)
7681 char * endbrace; /* '}' following the name */
7682 regnode *ret = NULL;
7685 GET_RE_DEBUG_FLAGS_DECL;
7687 PERL_ARGS_ASSERT_REG_NAMEDSEQ;
7691 /* The [^\n] meaning of \N ignores spaces and comments under the /x
7692 * modifier. The other meaning does not */
7693 p = (RExC_flags & RXf_PMf_EXTENDED)
7694 ? regwhite( pRExC_state, RExC_parse )
7697 /* Disambiguate between \N meaning a named character versus \N meaning
7698 * [^\n]. The former is assumed when it can't be the latter. */
7699 if (*p != '{' || regcurly(p)) {
7702 /* no bare \N in a charclass */
7703 vFAIL("\\N in a character class must be a named character: \\N{...}");
7705 nextchar(pRExC_state);
7706 ret = reg_node(pRExC_state, REG_ANY);
7707 *flagp |= HASWIDTH|SIMPLE;
7710 Set_Node_Length(ret, 1); /* MJD */
7714 /* Here, we have decided it should be a named sequence */
7716 /* The test above made sure that the next real character is a '{', but
7717 * under the /x modifier, it could be separated by space (or a comment and
7718 * \n) and this is not allowed (for consistency with \x{...} and the
7719 * tokenizer handling of \N{NAME}). */
7720 if (*RExC_parse != '{') {
7721 vFAIL("Missing braces on \\N{}");
7724 RExC_parse++; /* Skip past the '{' */
7726 if (! (endbrace = strchr(RExC_parse, '}')) /* no trailing brace */
7727 || ! (endbrace == RExC_parse /* nothing between the {} */
7728 || (endbrace - RExC_parse >= 2 /* U+ (bad hex is checked below */
7729 && strnEQ(RExC_parse, "U+", 2)))) /* for a better error msg) */
7731 if (endbrace) RExC_parse = endbrace; /* position msg's '<--HERE' */
7732 vFAIL("\\N{NAME} must be resolved by the lexer");
7735 if (endbrace == RExC_parse) { /* empty: \N{} */
7737 RExC_parse = endbrace + 1;
7738 return reg_node(pRExC_state,NOTHING);
7742 ckWARNreg(RExC_parse,
7743 "Ignoring zero length \\N{} in character class"
7745 RExC_parse = endbrace + 1;
7748 return (regnode *) &RExC_parse; /* Invalid regnode pointer */
7751 REQUIRE_UTF8; /* named sequences imply Unicode semantics */
7752 RExC_parse += 2; /* Skip past the 'U+' */
7754 if (valuep) { /* In a bracketed char class */
7755 /* We only pay attention to the first char of
7756 multichar strings being returned. I kinda wonder
7757 if this makes sense as it does change the behaviour
7758 from earlier versions, OTOH that behaviour was broken
7759 as well. XXX Solution is to recharacterize as
7760 [rest-of-class]|multi1|multi2... */
7762 STRLEN length_of_hex;
7763 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
7764 | PERL_SCAN_DISALLOW_PREFIX
7765 | (SIZE_ONLY ? PERL_SCAN_SILENT_ILLDIGIT : 0);
7767 char * endchar = RExC_parse + strcspn(RExC_parse, ".}");
7768 if (endchar < endbrace) {
7769 ckWARNreg(endchar, "Using just the first character returned by \\N{} in character class");
7772 length_of_hex = (STRLEN)(endchar - RExC_parse);
7773 *valuep = grok_hex(RExC_parse, &length_of_hex, &flags, NULL);
7775 /* The tokenizer should have guaranteed validity, but it's possible to
7776 * bypass it by using single quoting, so check */
7777 if (length_of_hex == 0
7778 || length_of_hex != (STRLEN)(endchar - RExC_parse) )
7780 RExC_parse += length_of_hex; /* Includes all the valid */
7781 RExC_parse += (RExC_orig_utf8) /* point to after 1st invalid */
7782 ? UTF8SKIP(RExC_parse)
7784 /* Guard against malformed utf8 */
7785 if (RExC_parse >= endchar) RExC_parse = endchar;
7786 vFAIL("Invalid hexadecimal number in \\N{U+...}");
7789 RExC_parse = endbrace + 1;
7790 if (endchar == endbrace) return NULL;
7792 ret = (regnode *) &RExC_parse; /* Invalid regnode pointer */
7794 else { /* Not a char class */
7796 /* What is done here is to convert this to a sub-pattern of the form
7797 * (?:\x{char1}\x{char2}...)
7798 * and then call reg recursively. That way, it retains its atomicness,
7799 * while not having to worry about special handling that some code
7800 * points may have. toke.c has converted the original Unicode values
7801 * to native, so that we can just pass on the hex values unchanged. We
7802 * do have to set a flag to keep recoding from happening in the
7805 SV * substitute_parse = newSVpvn_flags("?:", 2, SVf_UTF8|SVs_TEMP);
7807 char *endchar; /* Points to '.' or '}' ending cur char in the input
7809 char *orig_end = RExC_end;
7811 while (RExC_parse < endbrace) {
7813 /* Code points are separated by dots. If none, there is only one
7814 * code point, and is terminated by the brace */
7815 endchar = RExC_parse + strcspn(RExC_parse, ".}");
7817 /* Convert to notation the rest of the code understands */
7818 sv_catpv(substitute_parse, "\\x{");
7819 sv_catpvn(substitute_parse, RExC_parse, endchar - RExC_parse);
7820 sv_catpv(substitute_parse, "}");
7822 /* Point to the beginning of the next character in the sequence. */
7823 RExC_parse = endchar + 1;
7825 sv_catpv(substitute_parse, ")");
7827 RExC_parse = SvPV(substitute_parse, len);
7829 /* Don't allow empty number */
7831 vFAIL("Invalid hexadecimal number in \\N{U+...}");
7833 RExC_end = RExC_parse + len;
7835 /* The values are Unicode, and therefore not subject to recoding */
7836 RExC_override_recoding = 1;
7838 ret = reg(pRExC_state, 1, flagp, depth+1);
7840 RExC_parse = endbrace;
7841 RExC_end = orig_end;
7842 RExC_override_recoding = 0;
7844 nextchar(pRExC_state);
7854 * It returns the code point in utf8 for the value in *encp.
7855 * value: a code value in the source encoding
7856 * encp: a pointer to an Encode object
7858 * If the result from Encode is not a single character,
7859 * it returns U+FFFD (Replacement character) and sets *encp to NULL.
7862 S_reg_recode(pTHX_ const char value, SV **encp)
7865 SV * const sv = newSVpvn_flags(&value, numlen, SVs_TEMP);
7866 const char * const s = *encp ? sv_recode_to_utf8(sv, *encp) : SvPVX(sv);
7867 const STRLEN newlen = SvCUR(sv);
7868 UV uv = UNICODE_REPLACEMENT;
7870 PERL_ARGS_ASSERT_REG_RECODE;
7874 ? utf8n_to_uvchr((U8*)s, newlen, &numlen, UTF8_ALLOW_DEFAULT)
7877 if (!newlen || numlen != newlen) {
7878 uv = UNICODE_REPLACEMENT;
7886 - regatom - the lowest level
7888 Try to identify anything special at the start of the pattern. If there
7889 is, then handle it as required. This may involve generating a single regop,
7890 such as for an assertion; or it may involve recursing, such as to
7891 handle a () structure.
7893 If the string doesn't start with something special then we gobble up
7894 as much literal text as we can.
7896 Once we have been able to handle whatever type of thing started the
7897 sequence, we return.
7899 Note: we have to be careful with escapes, as they can be both literal
7900 and special, and in the case of \10 and friends can either, depending
7901 on context. Specifically there are two separate switches for handling
7902 escape sequences, with the one for handling literal escapes requiring
7903 a dummy entry for all of the special escapes that are actually handled
7908 S_regatom(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, U32 depth)
7911 register regnode *ret = NULL;
7913 char *parse_start = RExC_parse;
7915 GET_RE_DEBUG_FLAGS_DECL;
7916 DEBUG_PARSE("atom");
7917 *flagp = WORST; /* Tentatively. */
7919 PERL_ARGS_ASSERT_REGATOM;
7922 switch ((U8)*RExC_parse) {
7924 RExC_seen_zerolen++;
7925 nextchar(pRExC_state);
7926 if (RExC_flags & RXf_PMf_MULTILINE)
7927 ret = reg_node(pRExC_state, MBOL);
7928 else if (RExC_flags & RXf_PMf_SINGLELINE)
7929 ret = reg_node(pRExC_state, SBOL);
7931 ret = reg_node(pRExC_state, BOL);
7932 Set_Node_Length(ret, 1); /* MJD */
7935 nextchar(pRExC_state);
7937 RExC_seen_zerolen++;
7938 if (RExC_flags & RXf_PMf_MULTILINE)
7939 ret = reg_node(pRExC_state, MEOL);
7940 else if (RExC_flags & RXf_PMf_SINGLELINE)
7941 ret = reg_node(pRExC_state, SEOL);
7943 ret = reg_node(pRExC_state, EOL);
7944 Set_Node_Length(ret, 1); /* MJD */
7947 nextchar(pRExC_state);
7948 if (RExC_flags & RXf_PMf_SINGLELINE)
7949 ret = reg_node(pRExC_state, SANY);
7951 ret = reg_node(pRExC_state, REG_ANY);
7952 *flagp |= HASWIDTH|SIMPLE;
7954 Set_Node_Length(ret, 1); /* MJD */
7958 char * const oregcomp_parse = ++RExC_parse;
7959 ret = regclass(pRExC_state,depth+1);
7960 if (*RExC_parse != ']') {
7961 RExC_parse = oregcomp_parse;
7962 vFAIL("Unmatched [");
7964 nextchar(pRExC_state);
7965 *flagp |= HASWIDTH|SIMPLE;
7966 Set_Node_Length(ret, RExC_parse - oregcomp_parse + 1); /* MJD */
7970 nextchar(pRExC_state);
7971 ret = reg(pRExC_state, 1, &flags,depth+1);
7973 if (flags & TRYAGAIN) {
7974 if (RExC_parse == RExC_end) {
7975 /* Make parent create an empty node if needed. */
7983 *flagp |= flags&(HASWIDTH|SPSTART|SIMPLE|POSTPONED);
7987 if (flags & TRYAGAIN) {
7991 vFAIL("Internal urp");
7992 /* Supposed to be caught earlier. */
7995 if (!regcurly(RExC_parse)) {
8004 vFAIL("Quantifier follows nothing");
8009 This switch handles escape sequences that resolve to some kind
8010 of special regop and not to literal text. Escape sequnces that
8011 resolve to literal text are handled below in the switch marked
8014 Every entry in this switch *must* have a corresponding entry
8015 in the literal escape switch. However, the opposite is not
8016 required, as the default for this switch is to jump to the
8017 literal text handling code.
8019 switch ((U8)*++RExC_parse) {
8020 /* Special Escapes */
8022 RExC_seen_zerolen++;
8023 ret = reg_node(pRExC_state, SBOL);
8025 goto finish_meta_pat;
8027 ret = reg_node(pRExC_state, GPOS);
8028 RExC_seen |= REG_SEEN_GPOS;
8030 goto finish_meta_pat;
8032 RExC_seen_zerolen++;
8033 ret = reg_node(pRExC_state, KEEPS);
8035 /* XXX:dmq : disabling in-place substitution seems to
8036 * be necessary here to avoid cases of memory corruption, as
8037 * with: C<$_="x" x 80; s/x\K/y/> -- rgs
8039 RExC_seen |= REG_SEEN_LOOKBEHIND;
8040 goto finish_meta_pat;
8042 ret = reg_node(pRExC_state, SEOL);
8044 RExC_seen_zerolen++; /* Do not optimize RE away */
8045 goto finish_meta_pat;
8047 ret = reg_node(pRExC_state, EOS);
8049 RExC_seen_zerolen++; /* Do not optimize RE away */
8050 goto finish_meta_pat;
8052 ret = reg_node(pRExC_state, CANY);
8053 RExC_seen |= REG_SEEN_CANY;
8054 *flagp |= HASWIDTH|SIMPLE;
8055 goto finish_meta_pat;
8057 ret = reg_node(pRExC_state, CLUMP);
8059 goto finish_meta_pat;
8061 switch (get_regex_charset(RExC_flags)) {
8062 case REGEX_LOCALE_CHARSET:
8065 case REGEX_UNICODE_CHARSET:
8068 case REGEX_ASCII_RESTRICTED_CHARSET:
8069 case REGEX_ASCII_MORE_RESTRICTED_CHARSET:
8072 case REGEX_DEPENDS_CHARSET:
8078 ret = reg_node(pRExC_state, op);
8079 *flagp |= HASWIDTH|SIMPLE;
8080 goto finish_meta_pat;
8082 switch (get_regex_charset(RExC_flags)) {
8083 case REGEX_LOCALE_CHARSET:
8086 case REGEX_UNICODE_CHARSET:
8089 case REGEX_ASCII_RESTRICTED_CHARSET:
8090 case REGEX_ASCII_MORE_RESTRICTED_CHARSET:
8093 case REGEX_DEPENDS_CHARSET:
8099 ret = reg_node(pRExC_state, op);
8100 *flagp |= HASWIDTH|SIMPLE;
8101 goto finish_meta_pat;
8103 RExC_seen_zerolen++;
8104 RExC_seen |= REG_SEEN_LOOKBEHIND;
8105 switch (get_regex_charset(RExC_flags)) {
8106 case REGEX_LOCALE_CHARSET:
8109 case REGEX_UNICODE_CHARSET:
8112 case REGEX_ASCII_RESTRICTED_CHARSET:
8113 case REGEX_ASCII_MORE_RESTRICTED_CHARSET:
8116 case REGEX_DEPENDS_CHARSET:
8122 ret = reg_node(pRExC_state, op);
8123 FLAGS(ret) = get_regex_charset(RExC_flags);
8125 if (! SIZE_ONLY && (U8) *(RExC_parse + 1) == '{') {
8126 ckWARNregdep(RExC_parse, "\"\\b{\" is deprecated; use \"\\b\\{\" instead");
8128 goto finish_meta_pat;
8130 RExC_seen_zerolen++;
8131 RExC_seen |= REG_SEEN_LOOKBEHIND;
8132 switch (get_regex_charset(RExC_flags)) {
8133 case REGEX_LOCALE_CHARSET:
8136 case REGEX_UNICODE_CHARSET:
8139 case REGEX_ASCII_RESTRICTED_CHARSET:
8140 case REGEX_ASCII_MORE_RESTRICTED_CHARSET:
8143 case REGEX_DEPENDS_CHARSET:
8149 ret = reg_node(pRExC_state, op);
8150 FLAGS(ret) = get_regex_charset(RExC_flags);
8152 if (! SIZE_ONLY && (U8) *(RExC_parse + 1) == '{') {
8153 ckWARNregdep(RExC_parse, "\"\\B{\" is deprecated; use \"\\B\\{\" instead");
8155 goto finish_meta_pat;
8157 switch (get_regex_charset(RExC_flags)) {
8158 case REGEX_LOCALE_CHARSET:
8161 case REGEX_UNICODE_CHARSET:
8164 case REGEX_ASCII_RESTRICTED_CHARSET:
8165 case REGEX_ASCII_MORE_RESTRICTED_CHARSET:
8168 case REGEX_DEPENDS_CHARSET:
8174 ret = reg_node(pRExC_state, op);
8175 *flagp |= HASWIDTH|SIMPLE;
8176 goto finish_meta_pat;
8178 switch (get_regex_charset(RExC_flags)) {
8179 case REGEX_LOCALE_CHARSET:
8182 case REGEX_UNICODE_CHARSET:
8185 case REGEX_ASCII_RESTRICTED_CHARSET:
8186 case REGEX_ASCII_MORE_RESTRICTED_CHARSET:
8189 case REGEX_DEPENDS_CHARSET:
8195 ret = reg_node(pRExC_state, op);
8196 *flagp |= HASWIDTH|SIMPLE;
8197 goto finish_meta_pat;
8199 switch (get_regex_charset(RExC_flags)) {
8200 case REGEX_LOCALE_CHARSET:
8203 case REGEX_ASCII_RESTRICTED_CHARSET:
8204 case REGEX_ASCII_MORE_RESTRICTED_CHARSET:
8207 case REGEX_DEPENDS_CHARSET: /* No difference between these */
8208 case REGEX_UNICODE_CHARSET:
8214 ret = reg_node(pRExC_state, op);
8215 *flagp |= HASWIDTH|SIMPLE;
8216 goto finish_meta_pat;
8218 switch (get_regex_charset(RExC_flags)) {
8219 case REGEX_LOCALE_CHARSET:
8222 case REGEX_ASCII_RESTRICTED_CHARSET:
8223 case REGEX_ASCII_MORE_RESTRICTED_CHARSET:
8226 case REGEX_DEPENDS_CHARSET: /* No difference between these */
8227 case REGEX_UNICODE_CHARSET:
8233 ret = reg_node(pRExC_state, op);
8234 *flagp |= HASWIDTH|SIMPLE;
8235 goto finish_meta_pat;
8237 ret = reg_node(pRExC_state, LNBREAK);
8238 *flagp |= HASWIDTH|SIMPLE;
8239 goto finish_meta_pat;
8241 ret = reg_node(pRExC_state, HORIZWS);
8242 *flagp |= HASWIDTH|SIMPLE;
8243 goto finish_meta_pat;
8245 ret = reg_node(pRExC_state, NHORIZWS);
8246 *flagp |= HASWIDTH|SIMPLE;
8247 goto finish_meta_pat;
8249 ret = reg_node(pRExC_state, VERTWS);
8250 *flagp |= HASWIDTH|SIMPLE;
8251 goto finish_meta_pat;
8253 ret = reg_node(pRExC_state, NVERTWS);
8254 *flagp |= HASWIDTH|SIMPLE;
8256 nextchar(pRExC_state);
8257 Set_Node_Length(ret, 2); /* MJD */
8262 char* const oldregxend = RExC_end;
8264 char* parse_start = RExC_parse - 2;
8267 if (RExC_parse[1] == '{') {
8268 /* a lovely hack--pretend we saw [\pX] instead */
8269 RExC_end = strchr(RExC_parse, '}');
8271 const U8 c = (U8)*RExC_parse;
8273 RExC_end = oldregxend;
8274 vFAIL2("Missing right brace on \\%c{}", c);
8279 RExC_end = RExC_parse + 2;
8280 if (RExC_end > oldregxend)
8281 RExC_end = oldregxend;
8285 ret = regclass(pRExC_state,depth+1);
8287 RExC_end = oldregxend;
8290 Set_Node_Offset(ret, parse_start + 2);
8291 Set_Node_Cur_Length(ret);
8292 nextchar(pRExC_state);
8293 *flagp |= HASWIDTH|SIMPLE;
8297 /* Handle \N and \N{NAME} here and not below because it can be
8298 multicharacter. join_exact() will join them up later on.
8299 Also this makes sure that things like /\N{BLAH}+/ and
8300 \N{BLAH} being multi char Just Happen. dmq*/
8302 ret= reg_namedseq(pRExC_state, NULL, flagp, depth);
8304 case 'k': /* Handle \k<NAME> and \k'NAME' */
8307 char ch= RExC_parse[1];
8308 if (ch != '<' && ch != '\'' && ch != '{') {
8310 vFAIL2("Sequence %.2s... not terminated",parse_start);
8312 /* this pretty much dupes the code for (?P=...) in reg(), if
8313 you change this make sure you change that */
8314 char* name_start = (RExC_parse += 2);
8316 SV *sv_dat = reg_scan_name(pRExC_state,
8317 SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
8318 ch= (ch == '<') ? '>' : (ch == '{') ? '}' : '\'';
8319 if (RExC_parse == name_start || *RExC_parse != ch)
8320 vFAIL2("Sequence %.3s... not terminated",parse_start);
8323 num = add_data( pRExC_state, 1, "S" );
8324 RExC_rxi->data->data[num]=(void*)sv_dat;
8325 SvREFCNT_inc_simple_void(sv_dat);
8329 ret = reganode(pRExC_state,
8332 : (MORE_ASCII_RESTRICTED)
8334 : (AT_LEAST_UNI_SEMANTICS)
8342 /* override incorrect value set in reganode MJD */
8343 Set_Node_Offset(ret, parse_start+1);
8344 Set_Node_Cur_Length(ret); /* MJD */
8345 nextchar(pRExC_state);
8351 case '1': case '2': case '3': case '4':
8352 case '5': case '6': case '7': case '8': case '9':
8355 bool isg = *RExC_parse == 'g';
8360 if (*RExC_parse == '{') {
8364 if (*RExC_parse == '-') {
8368 if (hasbrace && !isDIGIT(*RExC_parse)) {
8369 if (isrel) RExC_parse--;
8371 goto parse_named_seq;
8373 num = atoi(RExC_parse);
8374 if (isg && num == 0)
8375 vFAIL("Reference to invalid group 0");
8377 num = RExC_npar - num;
8379 vFAIL("Reference to nonexistent or unclosed group");
8381 if (!isg && num > 9 && num >= RExC_npar)
8384 char * const parse_start = RExC_parse - 1; /* MJD */
8385 while (isDIGIT(*RExC_parse))
8387 if (parse_start == RExC_parse - 1)
8388 vFAIL("Unterminated \\g... pattern");
8390 if (*RExC_parse != '}')
8391 vFAIL("Unterminated \\g{...} pattern");
8395 if (num > (I32)RExC_rx->nparens)
8396 vFAIL("Reference to nonexistent group");
8399 ret = reganode(pRExC_state,
8402 : (MORE_ASCII_RESTRICTED)
8404 : (AT_LEAST_UNI_SEMANTICS)
8412 /* override incorrect value set in reganode MJD */
8413 Set_Node_Offset(ret, parse_start+1);
8414 Set_Node_Cur_Length(ret); /* MJD */
8416 nextchar(pRExC_state);
8421 if (RExC_parse >= RExC_end)
8422 FAIL("Trailing \\");
8425 /* Do not generate "unrecognized" warnings here, we fall
8426 back into the quick-grab loop below */
8433 if (RExC_flags & RXf_PMf_EXTENDED) {
8434 if ( reg_skipcomment( pRExC_state ) )
8441 parse_start = RExC_parse - 1;
8454 char_state latest_char_state = generic_char;
8455 register STRLEN len;
8460 U8 tmpbuf[UTF8_MAXBYTES_CASE+1], *foldbuf;
8461 regnode * orig_emit;
8464 orig_emit = RExC_emit; /* Save the original output node position in
8465 case we need to output a different node
8467 ret = reg_node(pRExC_state,
8468 (U8) ((! FOLD) ? EXACT
8471 : (MORE_ASCII_RESTRICTED)
8473 : (AT_LEAST_UNI_SEMANTICS)
8478 for (len = 0, p = RExC_parse - 1;
8479 len < 127 && p < RExC_end;
8482 char * const oldp = p;
8484 if (RExC_flags & RXf_PMf_EXTENDED)
8485 p = regwhite( pRExC_state, p );
8496 /* Literal Escapes Switch
8498 This switch is meant to handle escape sequences that
8499 resolve to a literal character.
8501 Every escape sequence that represents something
8502 else, like an assertion or a char class, is handled
8503 in the switch marked 'Special Escapes' above in this
8504 routine, but also has an entry here as anything that
8505 isn't explicitly mentioned here will be treated as
8506 an unescaped equivalent literal.
8510 /* These are all the special escapes. */
8511 case 'A': /* Start assertion */
8512 case 'b': case 'B': /* Word-boundary assertion*/
8513 case 'C': /* Single char !DANGEROUS! */
8514 case 'd': case 'D': /* digit class */
8515 case 'g': case 'G': /* generic-backref, pos assertion */
8516 case 'h': case 'H': /* HORIZWS */
8517 case 'k': case 'K': /* named backref, keep marker */
8518 case 'N': /* named char sequence */
8519 case 'p': case 'P': /* Unicode property */
8520 case 'R': /* LNBREAK */
8521 case 's': case 'S': /* space class */
8522 case 'v': case 'V': /* VERTWS */
8523 case 'w': case 'W': /* word class */
8524 case 'X': /* eXtended Unicode "combining character sequence" */
8525 case 'z': case 'Z': /* End of line/string assertion */
8529 /* Anything after here is an escape that resolves to a
8530 literal. (Except digits, which may or may not)
8549 ender = ASCII_TO_NATIVE('\033');
8553 ender = ASCII_TO_NATIVE('\007');
8558 STRLEN brace_len = len;
8560 const char* error_msg;
8562 bool valid = grok_bslash_o(p,
8569 RExC_parse = p; /* going to die anyway; point
8570 to exact spot of failure */
8577 if (PL_encoding && ender < 0x100) {
8578 goto recode_encoding;
8587 char* const e = strchr(p, '}');
8591 vFAIL("Missing right brace on \\x{}");
8594 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
8595 | PERL_SCAN_DISALLOW_PREFIX;
8596 STRLEN numlen = e - p - 1;
8597 ender = grok_hex(p + 1, &numlen, &flags, NULL);
8604 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
8606 ender = grok_hex(p, &numlen, &flags, NULL);
8609 if (PL_encoding && ender < 0x100)
8610 goto recode_encoding;
8614 ender = grok_bslash_c(*p++, UTF, SIZE_ONLY);
8616 case '0': case '1': case '2': case '3':case '4':
8617 case '5': case '6': case '7': case '8':case '9':
8619 (isDIGIT(p[1]) && atoi(p) >= RExC_npar))
8621 I32 flags = PERL_SCAN_SILENT_ILLDIGIT;
8623 ender = grok_oct(p, &numlen, &flags, NULL);
8633 if (PL_encoding && ender < 0x100)
8634 goto recode_encoding;
8637 if (! RExC_override_recoding) {
8638 SV* enc = PL_encoding;
8639 ender = reg_recode((const char)(U8)ender, &enc);
8640 if (!enc && SIZE_ONLY)
8641 ckWARNreg(p, "Invalid escape in the specified encoding");
8647 FAIL("Trailing \\");
8650 if (!SIZE_ONLY&& isALPHA(*p)) {
8651 /* Include any { following the alpha to emphasize
8652 * that it could be part of an escape at some point
8654 int len = (*(p + 1) == '{') ? 2 : 1;
8655 ckWARN3reg(p + len, "Unrecognized escape \\%.*s passed through", len, p);
8657 goto normal_default;
8662 if (UTF8_IS_START(*p) && UTF) {
8664 ender = utf8n_to_uvchr((U8*)p, RExC_end - p,
8665 &numlen, UTF8_ALLOW_DEFAULT);
8671 } /* End of switch on the literal */
8673 /* Certain characters are problematic because their folded
8674 * length is so different from their original length that it
8675 * isn't handleable by the optimizer. They are therefore not
8676 * placed in an EXACTish node; and are here handled specially.
8677 * (Even if the optimizer handled LATIN_SMALL_LETTER_SHARP_S,
8678 * putting it in a special node keeps regexec from having to
8679 * deal with a non-utf8 multi-char fold */
8681 && (ender > 255 || (! MORE_ASCII_RESTRICTED && ! LOC)))
8683 /* We look for either side of the fold. For example \xDF
8684 * folds to 'ss'. We look for both the single character
8685 * \xDF and the sequence 'ss'. When we find something that
8686 * could be one of those, we stop and flush whatever we
8687 * have output so far into the EXACTish node that was being
8688 * built. Then restore the input pointer to what it was.
8689 * regatom will return that EXACT node, and will be called
8690 * again, positioned so the first character is the one in
8691 * question, which we return in a different node type.
8692 * The multi-char folds are a sequence, so the occurrence
8693 * of the first character in that sequence doesn't
8694 * necessarily mean that what follows is the rest of the
8695 * sequence. We keep track of that with a state machine,
8696 * with the state being set to the latest character
8697 * processed before the current one. Most characters will
8698 * set the state to 0, but if one occurs that is part of a
8699 * potential tricky fold sequence, the state is set to that
8700 * character, and the next loop iteration sees if the state
8701 * should progress towards the final folded-from character,
8702 * or if it was a false alarm. If it turns out to be a
8703 * false alarm, the character(s) will be output in a new
8704 * EXACTish node, and join_exact() will later combine them.
8705 * In the case of the 'ss' sequence, which is more common
8706 * and more easily checked, some look-ahead is done to
8707 * save time by ruling-out some false alarms */
8710 latest_char_state = generic_char;
8714 if (AT_LEAST_UNI_SEMANTICS) {
8715 if (latest_char_state == char_s) { /* 'ss' */
8716 ender = LATIN_SMALL_LETTER_SHARP_S;
8719 else if (p < RExC_end) {
8721 /* Look-ahead at the next character. If it
8722 * is also an s, we handle as a sharp s
8723 * tricky regnode. */
8724 if (*p == 's' || *p == 'S') {
8726 /* But first flush anything in the
8727 * EXACTish buffer */
8732 p++; /* Account for swallowing this
8734 ender = LATIN_SMALL_LETTER_SHARP_S;
8737 /* Here, the next character is not a
8738 * literal 's', but still could
8739 * evaluate to one if part of a \o{},
8740 * \x or \OCTAL-DIGIT. The minimum
8741 * length required for that is 4, eg
8745 && (isDIGIT(*(p + 1))
8747 || *(p + 1) == 'o' ))
8750 /* Here, it could be an 's', too much
8751 * bother to figure it out here. Flush
8752 * the buffer if any; when come back
8753 * here, set the state so know that the
8754 * previous char was an 's' */
8756 latest_char_state = generic_char;
8760 latest_char_state = char_s;
8766 /* Here, can't be an 'ss' sequence, or at least not
8767 * one that could fold to/from the sharp ss */
8768 latest_char_state = generic_char;
8770 case 0x03C5: /* First char in upsilon series */
8771 if (p < RExC_end - 4) { /* Need >= 4 bytes left */
8772 latest_char_state = upsilon_1;
8779 latest_char_state = generic_char;
8782 case 0x03B9: /* First char in iota series */
8783 if (p < RExC_end - 4) {
8784 latest_char_state = iota_1;
8791 latest_char_state = generic_char;
8795 if (latest_char_state == upsilon_1) {
8796 latest_char_state = upsilon_2;
8798 else if (latest_char_state == iota_1) {
8799 latest_char_state = iota_2;
8802 latest_char_state = generic_char;
8806 if (latest_char_state == upsilon_2) {
8807 ender = GREEK_SMALL_LETTER_UPSILON_WITH_DIALYTIKA_AND_TONOS;
8810 else if (latest_char_state == iota_2) {
8811 ender = GREEK_SMALL_LETTER_IOTA_WITH_DIALYTIKA_AND_TONOS;
8814 latest_char_state = generic_char;
8817 /* These are the tricky fold characters. Flush any
8819 case GREEK_SMALL_LETTER_UPSILON_WITH_DIALYTIKA_AND_TONOS:
8820 case GREEK_SMALL_LETTER_IOTA_WITH_DIALYTIKA_AND_TONOS:
8821 case LATIN_SMALL_LETTER_SHARP_S:
8822 case LATIN_CAPITAL_LETTER_SHARP_S:
8831 char* const oldregxend = RExC_end;
8832 U8 tmpbuf[UTF8_MAXBYTES+1];
8834 /* Here, we know we need to generate a special
8835 * regnode, and 'ender' contains the tricky
8836 * character. What's done is to pretend it's in a
8837 * [bracketed] class, and let the code that deals
8838 * with those handle it, as that code has all the
8839 * intelligence necessary. First save the current
8840 * parse state, get rid of the already allocated
8841 * but empty EXACT node that the ANYOFV node will
8842 * replace, and point the parse to a buffer which
8843 * we fill with the character we want the regclass
8844 * code to think is being parsed */
8845 RExC_emit = orig_emit;
8846 RExC_parse = (char *) tmpbuf;
8848 U8 *d = uvchr_to_utf8(tmpbuf, ender);
8850 RExC_end = (char *) d;
8852 else { /* ender above 255 already excluded */
8853 tmpbuf[0] = (U8) ender;
8855 RExC_end = RExC_parse + 1;
8858 ret = regclass(pRExC_state,depth+1);
8860 /* Here, have parsed the buffer. Reset the parse to
8861 * the actual input, and return */
8862 RExC_end = oldregxend;
8865 Set_Node_Offset(ret, RExC_parse);
8866 Set_Node_Cur_Length(ret);
8867 nextchar(pRExC_state);
8868 *flagp |= HASWIDTH|SIMPLE;
8874 if ( RExC_flags & RXf_PMf_EXTENDED)
8875 p = regwhite( pRExC_state, p );
8877 /* Prime the casefolded buffer. Locale rules, which apply
8878 * only to code points < 256, aren't known until execution,
8879 * so for them, just output the original character using
8881 if (LOC && ender < 256) {
8882 if (UNI_IS_INVARIANT(ender)) {
8883 *tmpbuf = (U8) ender;
8886 *tmpbuf = UTF8_TWO_BYTE_HI(ender);
8887 *(tmpbuf + 1) = UTF8_TWO_BYTE_LO(ender);
8891 else if (isASCII(ender)) { /* Note: Here can't also be LOC
8893 ender = toLOWER(ender);
8894 *tmpbuf = (U8) ender;
8897 else if (! MORE_ASCII_RESTRICTED && ! LOC) {
8899 /* Locale and /aa require more selectivity about the
8900 * fold, so are handled below. Otherwise, here, just
8902 ender = toFOLD_uni(ender, tmpbuf, &foldlen);
8905 /* Under locale rules or /aa we are not to mix,
8906 * respectively, ords < 256 or ASCII with non-. So
8907 * reject folds that mix them, using only the
8908 * non-folded code point. So do the fold to a
8909 * temporary, and inspect each character in it. */
8910 U8 trialbuf[UTF8_MAXBYTES_CASE+1];
8912 UV tmpender = toFOLD_uni(ender, trialbuf, &foldlen);
8913 U8* e = s + foldlen;
8914 bool fold_ok = TRUE;
8918 || (LOC && (UTF8_IS_INVARIANT(*s)
8919 || UTF8_IS_DOWNGRADEABLE_START(*s))))
8927 Copy(trialbuf, tmpbuf, foldlen, U8);
8931 uvuni_to_utf8(tmpbuf, ender);
8932 foldlen = UNISKIP(ender);
8936 if (p < RExC_end && ISMULT2(p)) { /* Back off on ?+*. */
8941 /* Emit all the Unicode characters. */
8943 for (foldbuf = tmpbuf;
8945 foldlen -= numlen) {
8946 ender = utf8_to_uvchr(foldbuf, &numlen);
8948 const STRLEN unilen = reguni(pRExC_state, ender, s);
8951 /* In EBCDIC the numlen
8952 * and unilen can differ. */
8954 if (numlen >= foldlen)
8958 break; /* "Can't happen." */
8962 const STRLEN unilen = reguni(pRExC_state, ender, s);
8971 REGC((char)ender, s++);
8977 /* Emit all the Unicode characters. */
8979 for (foldbuf = tmpbuf;
8981 foldlen -= numlen) {
8982 ender = utf8_to_uvchr(foldbuf, &numlen);
8984 const STRLEN unilen = reguni(pRExC_state, ender, s);
8987 /* In EBCDIC the numlen
8988 * and unilen can differ. */
8990 if (numlen >= foldlen)
8998 const STRLEN unilen = reguni(pRExC_state, ender, s);
9007 REGC((char)ender, s++);
9010 loopdone: /* Jumped to when encounters something that shouldn't be in
9013 Set_Node_Cur_Length(ret); /* MJD */
9014 nextchar(pRExC_state);
9016 /* len is STRLEN which is unsigned, need to copy to signed */
9019 vFAIL("Internal disaster");
9023 if (len == 1 && UNI_IS_INVARIANT(ender))
9027 RExC_size += STR_SZ(len);
9030 RExC_emit += STR_SZ(len);
9038 /* Jumped to when an unrecognized character set is encountered */
9040 Perl_croak(aTHX_ "panic: Unknown regex character set encoding: %u", get_regex_charset(RExC_flags));
9045 S_regwhite( RExC_state_t *pRExC_state, char *p )
9047 const char *e = RExC_end;
9049 PERL_ARGS_ASSERT_REGWHITE;
9054 else if (*p == '#') {
9063 RExC_seen |= REG_SEEN_RUN_ON_COMMENT;
9071 /* Parse POSIX character classes: [[:foo:]], [[=foo=]], [[.foo.]].
9072 Character classes ([:foo:]) can also be negated ([:^foo:]).
9073 Returns a named class id (ANYOF_XXX) if successful, -1 otherwise.
9074 Equivalence classes ([=foo=]) and composites ([.foo.]) are parsed,
9075 but trigger failures because they are currently unimplemented. */
9077 #define POSIXCC_DONE(c) ((c) == ':')
9078 #define POSIXCC_NOTYET(c) ((c) == '=' || (c) == '.')
9079 #define POSIXCC(c) (POSIXCC_DONE(c) || POSIXCC_NOTYET(c))
9082 S_regpposixcc(pTHX_ RExC_state_t *pRExC_state, I32 value)
9085 I32 namedclass = OOB_NAMEDCLASS;
9087 PERL_ARGS_ASSERT_REGPPOSIXCC;
9089 if (value == '[' && RExC_parse + 1 < RExC_end &&
9090 /* I smell either [: or [= or [. -- POSIX has been here, right? */
9091 POSIXCC(UCHARAT(RExC_parse))) {
9092 const char c = UCHARAT(RExC_parse);
9093 char* const s = RExC_parse++;
9095 while (RExC_parse < RExC_end && UCHARAT(RExC_parse) != c)
9097 if (RExC_parse == RExC_end)
9098 /* Grandfather lone [:, [=, [. */
9101 const char* const t = RExC_parse++; /* skip over the c */
9104 if (UCHARAT(RExC_parse) == ']') {
9105 const char *posixcc = s + 1;
9106 RExC_parse++; /* skip over the ending ] */
9109 const I32 complement = *posixcc == '^' ? *posixcc++ : 0;
9110 const I32 skip = t - posixcc;
9112 /* Initially switch on the length of the name. */
9115 if (memEQ(posixcc, "word", 4)) /* this is not POSIX, this is the Perl \w */
9116 namedclass = complement ? ANYOF_NALNUM : ANYOF_ALNUM;
9119 /* Names all of length 5. */
9120 /* alnum alpha ascii blank cntrl digit graph lower
9121 print punct space upper */
9122 /* Offset 4 gives the best switch position. */
9123 switch (posixcc[4]) {
9125 if (memEQ(posixcc, "alph", 4)) /* alpha */
9126 namedclass = complement ? ANYOF_NALPHA : ANYOF_ALPHA;
9129 if (memEQ(posixcc, "spac", 4)) /* space */
9130 namedclass = complement ? ANYOF_NPSXSPC : ANYOF_PSXSPC;
9133 if (memEQ(posixcc, "grap", 4)) /* graph */
9134 namedclass = complement ? ANYOF_NGRAPH : ANYOF_GRAPH;
9137 if (memEQ(posixcc, "asci", 4)) /* ascii */
9138 namedclass = complement ? ANYOF_NASCII : ANYOF_ASCII;
9141 if (memEQ(posixcc, "blan", 4)) /* blank */
9142 namedclass = complement ? ANYOF_NBLANK : ANYOF_BLANK;
9145 if (memEQ(posixcc, "cntr", 4)) /* cntrl */
9146 namedclass = complement ? ANYOF_NCNTRL : ANYOF_CNTRL;
9149 if (memEQ(posixcc, "alnu", 4)) /* alnum */
9150 namedclass = complement ? ANYOF_NALNUMC : ANYOF_ALNUMC;
9153 if (memEQ(posixcc, "lowe", 4)) /* lower */
9154 namedclass = complement ? ANYOF_NLOWER : ANYOF_LOWER;
9155 else if (memEQ(posixcc, "uppe", 4)) /* upper */
9156 namedclass = complement ? ANYOF_NUPPER : ANYOF_UPPER;
9159 if (memEQ(posixcc, "digi", 4)) /* digit */
9160 namedclass = complement ? ANYOF_NDIGIT : ANYOF_DIGIT;
9161 else if (memEQ(posixcc, "prin", 4)) /* print */
9162 namedclass = complement ? ANYOF_NPRINT : ANYOF_PRINT;
9163 else if (memEQ(posixcc, "punc", 4)) /* punct */
9164 namedclass = complement ? ANYOF_NPUNCT : ANYOF_PUNCT;
9169 if (memEQ(posixcc, "xdigit", 6))
9170 namedclass = complement ? ANYOF_NXDIGIT : ANYOF_XDIGIT;
9174 if (namedclass == OOB_NAMEDCLASS)
9175 Simple_vFAIL3("POSIX class [:%.*s:] unknown",
9177 assert (posixcc[skip] == ':');
9178 assert (posixcc[skip+1] == ']');
9179 } else if (!SIZE_ONLY) {
9180 /* [[=foo=]] and [[.foo.]] are still future. */
9182 /* adjust RExC_parse so the warning shows after
9184 while (UCHARAT(RExC_parse) && UCHARAT(RExC_parse) != ']')
9186 Simple_vFAIL3("POSIX syntax [%c %c] is reserved for future extensions", c, c);
9189 /* Maternal grandfather:
9190 * "[:" ending in ":" but not in ":]" */
9200 S_checkposixcc(pTHX_ RExC_state_t *pRExC_state)
9204 PERL_ARGS_ASSERT_CHECKPOSIXCC;
9206 if (POSIXCC(UCHARAT(RExC_parse))) {
9207 const char *s = RExC_parse;
9208 const char c = *s++;
9212 if (*s && c == *s && s[1] == ']') {
9214 "POSIX syntax [%c %c] belongs inside character classes",
9217 /* [[=foo=]] and [[.foo.]] are still future. */
9218 if (POSIXCC_NOTYET(c)) {
9219 /* adjust RExC_parse so the error shows after
9221 while (UCHARAT(RExC_parse) && UCHARAT(RExC_parse++) != ']')
9223 Simple_vFAIL3("POSIX syntax [%c %c] is reserved for future extensions", c, c);
9229 /* No locale test, and always Unicode semantics */
9230 #define _C_C_T_NOLOC_(NAME,TEST,WORD) \
9232 for (value = 0; value < 256; value++) \
9234 stored += set_regclass_bit(pRExC_state, ret, (U8) value, &l1_fold_invlist, &unicode_alternate); \
9238 case ANYOF_N##NAME: \
9239 for (value = 0; value < 256; value++) \
9241 stored += set_regclass_bit(pRExC_state, ret, (U8) value, &l1_fold_invlist, &unicode_alternate); \
9246 /* Like the above, but there are differences if we are in uni-8-bit or not, so
9247 * there are two tests passed in, to use depending on that. There aren't any
9248 * cases where the label is different from the name, so no need for that
9250 #define _C_C_T_(NAME, TEST_8, TEST_7, WORD) \
9252 if (LOC) ANYOF_CLASS_SET(ret, ANYOF_##NAME); \
9253 else if (UNI_SEMANTICS) { \
9254 for (value = 0; value < 256; value++) { \
9255 if (TEST_8(value)) stored += \
9256 set_regclass_bit(pRExC_state, ret, (U8) value, &l1_fold_invlist, &unicode_alternate); \
9260 for (value = 0; value < 128; value++) { \
9261 if (TEST_7(UNI_TO_NATIVE(value))) stored += \
9262 set_regclass_bit(pRExC_state, ret, \
9263 (U8) UNI_TO_NATIVE(value), &l1_fold_invlist, &unicode_alternate); \
9269 case ANYOF_N##NAME: \
9270 if (LOC) ANYOF_CLASS_SET(ret, ANYOF_N##NAME); \
9271 else if (UNI_SEMANTICS) { \
9272 for (value = 0; value < 256; value++) { \
9273 if (! TEST_8(value)) stored += \
9274 set_regclass_bit(pRExC_state, ret, (U8) value, &l1_fold_invlist, &unicode_alternate); \
9278 for (value = 0; value < 128; value++) { \
9279 if (! TEST_7(UNI_TO_NATIVE(value))) stored += set_regclass_bit( \
9280 pRExC_state, ret, (U8) UNI_TO_NATIVE(value), &l1_fold_invlist, &unicode_alternate); \
9282 if (AT_LEAST_ASCII_RESTRICTED) { \
9283 for (value = 128; value < 256; value++) { \
9284 stored += set_regclass_bit( \
9285 pRExC_state, ret, (U8) UNI_TO_NATIVE(value), &l1_fold_invlist, &unicode_alternate); \
9287 ANYOF_FLAGS(ret) |= ANYOF_UNICODE_ALL; \
9290 /* For a non-ut8 target string with DEPENDS semantics, all above \
9291 * ASCII Latin1 code points match the complement of any of the \
9292 * classes. But in utf8, they have their Unicode semantics, so \
9293 * can't just set them in the bitmap, or else regexec.c will think \
9294 * they matched when they shouldn't. */ \
9295 ANYOF_FLAGS(ret) |= ANYOF_NON_UTF8_LATIN1_ALL; \
9303 S_set_regclass_bit_fold(pTHX_ RExC_state_t *pRExC_state, regnode* node, const U8 value, HV** invlist_ptr, AV** alternate_ptr)
9306 /* Handle the setting of folds in the bitmap for non-locale ANYOF nodes.
9307 * Locale folding is done at run-time, so this function should not be
9308 * called for nodes that are for locales.
9310 * This function sets the bit corresponding to the fold of the input
9311 * 'value', if not already set. The fold of 'f' is 'F', and the fold of
9314 * It also knows about the characters that are in the bitmap that have
9315 * folds that are matchable only outside it, and sets the appropriate lists
9318 * It returns the number of bits that actually changed from 0 to 1 */
9323 PERL_ARGS_ASSERT_SET_REGCLASS_BIT_FOLD;
9325 fold = (AT_LEAST_UNI_SEMANTICS) ? PL_fold_latin1[value]
9328 /* It assumes the bit for 'value' has already been set */
9329 if (fold != value && ! ANYOF_BITMAP_TEST(node, fold)) {
9330 ANYOF_BITMAP_SET(node, fold);
9333 if (_HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(value) && (! isASCII(value) || ! MORE_ASCII_RESTRICTED)) {
9334 /* Certain Latin1 characters have matches outside the bitmap. To get
9335 * here, 'value' is one of those characters. None of these matches is
9336 * valid for ASCII characters under /aa, which have been excluded by
9337 * the 'if' above. The matches fall into three categories:
9338 * 1) They are singly folded-to or -from an above 255 character, as
9339 * LATIN SMALL LETTER Y WITH DIAERESIS and LATIN CAPITAL LETTER Y
9341 * 2) They are part of a multi-char fold with another character in the
9342 * bitmap, only LATIN SMALL LETTER SHARP S => "ss" fits that bill;
9343 * 3) They are part of a multi-char fold with a character not in the
9344 * bitmap, such as various ligatures.
9345 * We aren't dealing fully with multi-char folds, except we do deal
9346 * with the pattern containing a character that has a multi-char fold
9347 * (not so much the inverse).
9348 * For types 1) and 3), the matches only happen when the target string
9349 * is utf8; that's not true for 2), and we set a flag for it.
9351 * The code below adds to the passed in inversion list the single fold
9352 * closures for 'value'. The values are hard-coded here so that an
9353 * innocent-looking character class, like /[ks]/i won't have to go out
9354 * to disk to find the possible matches. XXX It would be better to
9355 * generate these via regen, in case a new version of the Unicode
9356 * standard adds new mappings, though that is not really likely. */
9361 *invlist_ptr = add_cp_to_invlist(*invlist_ptr, 0x212A);
9365 /* LATIN SMALL LETTER LONG S */
9366 *invlist_ptr = add_cp_to_invlist(*invlist_ptr, 0x017F);
9369 *invlist_ptr = add_cp_to_invlist(*invlist_ptr,
9370 GREEK_SMALL_LETTER_MU);
9371 *invlist_ptr = add_cp_to_invlist(*invlist_ptr,
9372 GREEK_CAPITAL_LETTER_MU);
9374 case LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE:
9375 case LATIN_SMALL_LETTER_A_WITH_RING_ABOVE:
9377 *invlist_ptr = add_cp_to_invlist(*invlist_ptr, 0x212B);
9378 if (DEPENDS_SEMANTICS) { /* See DEPENDS comment below */
9379 *invlist_ptr = add_cp_to_invlist(*invlist_ptr,
9380 PL_fold_latin1[value]);
9383 case LATIN_SMALL_LETTER_Y_WITH_DIAERESIS:
9384 *invlist_ptr = add_cp_to_invlist(*invlist_ptr,
9385 LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS);
9387 case LATIN_SMALL_LETTER_SHARP_S:
9388 *invlist_ptr = add_cp_to_invlist(*invlist_ptr,
9389 LATIN_CAPITAL_LETTER_SHARP_S);
9391 /* Under /a, /d, and /u, this can match the two chars "ss" */
9392 if (! MORE_ASCII_RESTRICTED) {
9393 add_alternate(alternate_ptr, (U8 *) "ss", 2);
9395 /* And under /u or /a, it can match even if the target is
9397 if (AT_LEAST_UNI_SEMANTICS) {
9398 ANYOF_FLAGS(node) |= ANYOF_NONBITMAP_NON_UTF8;
9406 /* These all are targets of multi-character folds, which can
9407 * occur with only non-Latin1 characters in the fold, so they
9408 * can match if the target string isn't UTF-8 */
9409 ANYOF_FLAGS(node) |= ANYOF_NONBITMAP_NON_UTF8;
9417 /* These all are targets of multi-character folds, which occur
9418 * only with a non-Latin1 character as part of the fold, so
9419 * they can't match unless the target string is in UTF-8, so no
9420 * action here is necessary */
9423 /* Use deprecated warning to increase the chances of this
9425 ckWARN2regdep(RExC_parse, "Perl folding rules are not up-to-date for 0x%x; please use the perlbug utility to report;", value);
9429 else if (DEPENDS_SEMANTICS
9431 && PL_fold_latin1[value] != value)
9433 /* Under DEPENDS rules, non-ASCII Latin1 characters match their
9434 * folds only when the target string is in UTF-8. We add the fold
9435 * here to the list of things to match outside the bitmap, which
9436 * won't be looked at unless it is UTF8 (or else if something else
9437 * says to look even if not utf8, but those things better not happen
9438 * under DEPENDS semantics. */
9439 *invlist_ptr = add_cp_to_invlist(*invlist_ptr, PL_fold_latin1[value]);
9446 PERL_STATIC_INLINE U8
9447 S_set_regclass_bit(pTHX_ RExC_state_t *pRExC_state, regnode* node, const U8 value, HV** invlist_ptr, AV** alternate_ptr)
9449 /* This inline function sets a bit in the bitmap if not already set, and if
9450 * appropriate, its fold, returning the number of bits that actually
9451 * changed from 0 to 1 */
9455 PERL_ARGS_ASSERT_SET_REGCLASS_BIT;
9457 if (ANYOF_BITMAP_TEST(node, value)) { /* Already set */
9461 ANYOF_BITMAP_SET(node, value);
9464 if (FOLD && ! LOC) { /* Locale folds aren't known until runtime */
9465 stored += set_regclass_bit_fold(pRExC_state, node, value, invlist_ptr, alternate_ptr);
9472 S_add_alternate(pTHX_ AV** alternate_ptr, U8* string, STRLEN len)
9474 /* Adds input 'string' with length 'len' to the ANYOF node's unicode
9475 * alternate list, pointed to by 'alternate_ptr'. This is an array of
9476 * the multi-character folds of characters in the node */
9479 PERL_ARGS_ASSERT_ADD_ALTERNATE;
9481 if (! *alternate_ptr) {
9482 *alternate_ptr = newAV();
9484 sv = newSVpvn_utf8((char*)string, len, TRUE);
9485 av_push(*alternate_ptr, sv);
9490 parse a class specification and produce either an ANYOF node that
9491 matches the pattern or perhaps will be optimized into an EXACTish node
9492 instead. The node contains a bit map for the first 256 characters, with the
9493 corresponding bit set if that character is in the list. For characters
9494 above 255, a range list is used */
9497 S_regclass(pTHX_ RExC_state_t *pRExC_state, U32 depth)
9500 register UV nextvalue;
9501 register IV prevvalue = OOB_UNICODE;
9502 register IV range = 0;
9503 UV value = 0; /* XXX:dmq: needs to be referenceable (unfortunately) */
9504 register regnode *ret;
9507 char *rangebegin = NULL;
9508 bool need_class = 0;
9510 STRLEN initial_listsv_len = 0; /* Kind of a kludge to see if it is more
9511 than just initialized. */
9514 /* code points this node matches that can't be stored in the bitmap */
9515 HV* nonbitmap = NULL;
9517 /* The items that are to match that aren't stored in the bitmap, but are a
9518 * result of things that are stored there. This is the fold closure of
9519 * such a character, either because it has DEPENDS semantics and shouldn't
9520 * be matched unless the target string is utf8, or is a code point that is
9521 * too large for the bit map, as for example, the fold of the MICRO SIGN is
9522 * above 255. This all is solely for performance reasons. By having this
9523 * code know the outside-the-bitmap folds that the bitmapped characters are
9524 * involved with, we don't have to go out to disk to find the list of
9525 * matches, unless the character class includes code points that aren't
9526 * storable in the bit map. That means that a character class with an 's'
9527 * in it, for example, doesn't need to go out to disk to find everything
9528 * that matches. A 2nd list is used so that the 'nonbitmap' list is kept
9529 * empty unless there is something whose fold we don't know about, and will
9530 * have to go out to the disk to find. */
9531 HV* l1_fold_invlist = NULL;
9533 /* List of multi-character folds that are matched by this node */
9534 AV* unicode_alternate = NULL;
9536 UV literal_endpoint = 0;
9538 UV stored = 0; /* how many chars stored in the bitmap */
9540 regnode * const orig_emit = RExC_emit; /* Save the original RExC_emit in
9541 case we need to change the emitted regop to an EXACT. */
9542 const char * orig_parse = RExC_parse;
9543 GET_RE_DEBUG_FLAGS_DECL;
9545 PERL_ARGS_ASSERT_REGCLASS;
9547 PERL_UNUSED_ARG(depth);
9550 DEBUG_PARSE("clas");
9552 /* Assume we are going to generate an ANYOF node. */
9553 ret = reganode(pRExC_state, ANYOF, 0);
9557 ANYOF_FLAGS(ret) = 0;
9560 if (UCHARAT(RExC_parse) == '^') { /* Complement of range. */
9564 ANYOF_FLAGS(ret) |= ANYOF_INVERT;
9568 RExC_size += ANYOF_SKIP;
9569 listsv = &PL_sv_undef; /* For code scanners: listsv always non-NULL. */
9572 RExC_emit += ANYOF_SKIP;
9574 ANYOF_FLAGS(ret) |= ANYOF_LOCALE;
9576 ANYOF_BITMAP_ZERO(ret);
9577 listsv = newSVpvs("# comment\n");
9578 initial_listsv_len = SvCUR(listsv);
9581 nextvalue = RExC_parse < RExC_end ? UCHARAT(RExC_parse) : 0;
9583 if (!SIZE_ONLY && POSIXCC(nextvalue))
9584 checkposixcc(pRExC_state);
9586 /* allow 1st char to be ] (allowing it to be - is dealt with later) */
9587 if (UCHARAT(RExC_parse) == ']')
9591 while (RExC_parse < RExC_end && UCHARAT(RExC_parse) != ']') {
9595 namedclass = OOB_NAMEDCLASS; /* initialize as illegal */
9598 rangebegin = RExC_parse;
9600 value = utf8n_to_uvchr((U8*)RExC_parse,
9601 RExC_end - RExC_parse,
9602 &numlen, UTF8_ALLOW_DEFAULT);
9603 RExC_parse += numlen;
9606 value = UCHARAT(RExC_parse++);
9608 nextvalue = RExC_parse < RExC_end ? UCHARAT(RExC_parse) : 0;
9609 if (value == '[' && POSIXCC(nextvalue))
9610 namedclass = regpposixcc(pRExC_state, value);
9611 else if (value == '\\') {
9613 value = utf8n_to_uvchr((U8*)RExC_parse,
9614 RExC_end - RExC_parse,
9615 &numlen, UTF8_ALLOW_DEFAULT);
9616 RExC_parse += numlen;
9619 value = UCHARAT(RExC_parse++);
9620 /* Some compilers cannot handle switching on 64-bit integer
9621 * values, therefore value cannot be an UV. Yes, this will
9622 * be a problem later if we want switch on Unicode.
9623 * A similar issue a little bit later when switching on
9624 * namedclass. --jhi */
9625 switch ((I32)value) {
9626 case 'w': namedclass = ANYOF_ALNUM; break;
9627 case 'W': namedclass = ANYOF_NALNUM; break;
9628 case 's': namedclass = ANYOF_SPACE; break;
9629 case 'S': namedclass = ANYOF_NSPACE; break;
9630 case 'd': namedclass = ANYOF_DIGIT; break;
9631 case 'D': namedclass = ANYOF_NDIGIT; break;
9632 case 'v': namedclass = ANYOF_VERTWS; break;
9633 case 'V': namedclass = ANYOF_NVERTWS; break;
9634 case 'h': namedclass = ANYOF_HORIZWS; break;
9635 case 'H': namedclass = ANYOF_NHORIZWS; break;
9636 case 'N': /* Handle \N{NAME} in class */
9638 /* We only pay attention to the first char of
9639 multichar strings being returned. I kinda wonder
9640 if this makes sense as it does change the behaviour
9641 from earlier versions, OTOH that behaviour was broken
9643 UV v; /* value is register so we cant & it /grrr */
9644 if (reg_namedseq(pRExC_state, &v, NULL, depth)) {
9654 if (RExC_parse >= RExC_end)
9655 vFAIL2("Empty \\%c{}", (U8)value);
9656 if (*RExC_parse == '{') {
9657 const U8 c = (U8)value;
9658 e = strchr(RExC_parse++, '}');
9660 vFAIL2("Missing right brace on \\%c{}", c);
9661 while (isSPACE(UCHARAT(RExC_parse)))
9663 if (e == RExC_parse)
9664 vFAIL2("Empty \\%c{}", c);
9666 while (isSPACE(UCHARAT(RExC_parse + n - 1)))
9674 if (UCHARAT(RExC_parse) == '^') {
9677 value = value == 'p' ? 'P' : 'p'; /* toggle */
9678 while (isSPACE(UCHARAT(RExC_parse))) {
9684 /* Add the property name to the list. If /i matching, give
9685 * a different name which consists of the normal name
9686 * sandwiched between two underscores and '_i'. The design
9687 * is discussed in the commit message for this. */
9688 Perl_sv_catpvf(aTHX_ listsv, "%cutf8::%s%.*s%s\n",
9689 (value=='p' ? '+' : '!'),
9698 /* The \p could match something in the Latin1 range, hence
9699 * something that isn't utf8 */
9700 ANYOF_FLAGS(ret) |= ANYOF_NONBITMAP_NON_UTF8;
9701 namedclass = ANYOF_MAX; /* no official name, but it's named */
9703 /* \p means they want Unicode semantics */
9704 RExC_uni_semantics = 1;
9707 case 'n': value = '\n'; break;
9708 case 'r': value = '\r'; break;
9709 case 't': value = '\t'; break;
9710 case 'f': value = '\f'; break;
9711 case 'b': value = '\b'; break;
9712 case 'e': value = ASCII_TO_NATIVE('\033');break;
9713 case 'a': value = ASCII_TO_NATIVE('\007');break;
9715 RExC_parse--; /* function expects to be pointed at the 'o' */
9717 const char* error_msg;
9718 bool valid = grok_bslash_o(RExC_parse,
9723 RExC_parse += numlen;
9728 if (PL_encoding && value < 0x100) {
9729 goto recode_encoding;
9733 if (*RExC_parse == '{') {
9734 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
9735 | PERL_SCAN_DISALLOW_PREFIX;
9736 char * const e = strchr(RExC_parse++, '}');
9738 vFAIL("Missing right brace on \\x{}");
9740 numlen = e - RExC_parse;
9741 value = grok_hex(RExC_parse, &numlen, &flags, NULL);
9745 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
9747 value = grok_hex(RExC_parse, &numlen, &flags, NULL);
9748 RExC_parse += numlen;
9750 if (PL_encoding && value < 0x100)
9751 goto recode_encoding;
9754 value = grok_bslash_c(*RExC_parse++, UTF, SIZE_ONLY);
9756 case '0': case '1': case '2': case '3': case '4':
9757 case '5': case '6': case '7':
9759 /* Take 1-3 octal digits */
9760 I32 flags = PERL_SCAN_SILENT_ILLDIGIT;
9762 value = grok_oct(--RExC_parse, &numlen, &flags, NULL);
9763 RExC_parse += numlen;
9764 if (PL_encoding && value < 0x100)
9765 goto recode_encoding;
9769 if (! RExC_override_recoding) {
9770 SV* enc = PL_encoding;
9771 value = reg_recode((const char)(U8)value, &enc);
9772 if (!enc && SIZE_ONLY)
9773 ckWARNreg(RExC_parse,
9774 "Invalid escape in the specified encoding");
9778 /* Allow \_ to not give an error */
9779 if (!SIZE_ONLY && isALNUM(value) && value != '_') {
9780 ckWARN2reg(RExC_parse,
9781 "Unrecognized escape \\%c in character class passed through",
9786 } /* end of \blah */
9792 if (namedclass > OOB_NAMEDCLASS) { /* this is a named class \blah */
9794 /* What matches in a locale is not known until runtime, so need to
9795 * (one time per class) allocate extra space to pass to regexec.
9796 * The space will contain a bit for each named class that is to be
9797 * matched against. This isn't needed for \p{} and pseudo-classes,
9798 * as they are not affected by locale, and hence are dealt with
9800 if (LOC && namedclass < ANYOF_MAX && ! need_class) {
9803 RExC_size += ANYOF_CLASS_SKIP - ANYOF_SKIP;
9806 RExC_emit += ANYOF_CLASS_SKIP - ANYOF_SKIP;
9807 ANYOF_CLASS_ZERO(ret);
9809 ANYOF_FLAGS(ret) |= ANYOF_CLASS;
9812 /* a bad range like a-\d, a-[:digit:]. The '-' is taken as a
9813 * literal, as is the character that began the false range, i.e.
9814 * the 'a' in the examples */
9818 RExC_parse >= rangebegin ?
9819 RExC_parse - rangebegin : 0;
9820 ckWARN4reg(RExC_parse,
9821 "False [] range \"%*.*s\"",
9825 set_regclass_bit(pRExC_state, ret, '-', &l1_fold_invlist, &unicode_alternate);
9826 if (prevvalue < 256) {
9828 set_regclass_bit(pRExC_state, ret, (U8) prevvalue, &l1_fold_invlist, &unicode_alternate);
9831 nonbitmap = add_cp_to_invlist(nonbitmap, prevvalue);
9835 range = 0; /* this was not a true range */
9841 const char *what = NULL;
9844 /* Possible truncation here but in some 64-bit environments
9845 * the compiler gets heartburn about switch on 64-bit values.
9846 * A similar issue a little earlier when switching on value.
9848 switch ((I32)namedclass) {
9850 case _C_C_T_(ALNUMC, isALNUMC_L1, isALNUMC, "XPosixAlnum");
9851 case _C_C_T_(ALPHA, isALPHA_L1, isALPHA, "XPosixAlpha");
9852 case _C_C_T_(BLANK, isBLANK_L1, isBLANK, "XPosixBlank");
9853 case _C_C_T_(CNTRL, isCNTRL_L1, isCNTRL, "XPosixCntrl");
9854 case _C_C_T_(GRAPH, isGRAPH_L1, isGRAPH, "XPosixGraph");
9855 case _C_C_T_(LOWER, isLOWER_L1, isLOWER, "XPosixLower");
9856 case _C_C_T_(PRINT, isPRINT_L1, isPRINT, "XPosixPrint");
9857 case _C_C_T_(PSXSPC, isPSXSPC_L1, isPSXSPC, "XPosixSpace");
9858 case _C_C_T_(PUNCT, isPUNCT_L1, isPUNCT, "XPosixPunct");
9859 case _C_C_T_(UPPER, isUPPER_L1, isUPPER, "XPosixUpper");
9860 /* \s, \w match all unicode if utf8. */
9861 case _C_C_T_(SPACE, isSPACE_L1, isSPACE, "SpacePerl");
9862 case _C_C_T_(ALNUM, isWORDCHAR_L1, isALNUM, "Word");
9863 case _C_C_T_(XDIGIT, isXDIGIT_L1, isXDIGIT, "XPosixXDigit");
9864 case _C_C_T_NOLOC_(VERTWS, is_VERTWS_latin1(&value), "VertSpace");
9865 case _C_C_T_NOLOC_(HORIZWS, is_HORIZWS_latin1(&value), "HorizSpace");
9868 ANYOF_CLASS_SET(ret, ANYOF_ASCII);
9870 for (value = 0; value < 128; value++)
9872 set_regclass_bit(pRExC_state, ret, (U8) ASCII_TO_NATIVE(value), &l1_fold_invlist, &unicode_alternate);
9875 what = NULL; /* Doesn't match outside ascii, so
9876 don't want to add +utf8:: */
9880 ANYOF_CLASS_SET(ret, ANYOF_NASCII);
9882 for (value = 128; value < 256; value++)
9884 set_regclass_bit(pRExC_state, ret, (U8) ASCII_TO_NATIVE(value), &l1_fold_invlist, &unicode_alternate);
9886 ANYOF_FLAGS(ret) |= ANYOF_UNICODE_ALL;
9892 ANYOF_CLASS_SET(ret, ANYOF_DIGIT);
9894 /* consecutive digits assumed */
9895 for (value = '0'; value <= '9'; value++)
9897 set_regclass_bit(pRExC_state, ret, (U8) value, &l1_fold_invlist, &unicode_alternate);
9904 ANYOF_CLASS_SET(ret, ANYOF_NDIGIT);
9906 /* consecutive digits assumed */
9907 for (value = 0; value < '0'; value++)
9909 set_regclass_bit(pRExC_state, ret, (U8) value, &l1_fold_invlist, &unicode_alternate);
9910 for (value = '9' + 1; value < 256; value++)
9912 set_regclass_bit(pRExC_state, ret, (U8) value, &l1_fold_invlist, &unicode_alternate);
9916 if (AT_LEAST_ASCII_RESTRICTED ) {
9917 ANYOF_FLAGS(ret) |= ANYOF_UNICODE_ALL;
9921 /* this is to handle \p and \P */
9924 vFAIL("Invalid [::] class");
9927 if (what && ! (AT_LEAST_ASCII_RESTRICTED)) {
9928 /* Strings such as "+utf8::isWord\n" */
9929 Perl_sv_catpvf(aTHX_ listsv, "%cutf8::Is%s\n", yesno, what);
9934 } /* end of namedclass \blah */
9937 if (prevvalue > (IV)value) /* b-a */ {
9938 const int w = RExC_parse - rangebegin;
9939 Simple_vFAIL4("Invalid [] range \"%*.*s\"", w, w, rangebegin);
9940 range = 0; /* not a valid range */
9944 prevvalue = value; /* save the beginning of the range */
9945 if (RExC_parse+1 < RExC_end
9946 && *RExC_parse == '-'
9947 && RExC_parse[1] != ']')
9951 /* a bad range like \w-, [:word:]- ? */
9952 if (namedclass > OOB_NAMEDCLASS) {
9953 if (ckWARN(WARN_REGEXP)) {
9955 RExC_parse >= rangebegin ?
9956 RExC_parse - rangebegin : 0;
9958 "False [] range \"%*.*s\"",
9963 set_regclass_bit(pRExC_state, ret, '-', &l1_fold_invlist, &unicode_alternate);
9965 range = 1; /* yeah, it's a range! */
9966 continue; /* but do it the next time */
9970 /* non-Latin1 code point implies unicode semantics. Must be set in
9971 * pass1 so is there for the whole of pass 2 */
9973 RExC_uni_semantics = 1;
9976 /* now is the next time */
9978 if (prevvalue < 256) {
9979 const IV ceilvalue = value < 256 ? value : 255;
9982 /* In EBCDIC [\x89-\x91] should include
9983 * the \x8e but [i-j] should not. */
9984 if (literal_endpoint == 2 &&
9985 ((isLOWER(prevvalue) && isLOWER(ceilvalue)) ||
9986 (isUPPER(prevvalue) && isUPPER(ceilvalue))))
9988 if (isLOWER(prevvalue)) {
9989 for (i = prevvalue; i <= ceilvalue; i++)
9990 if (isLOWER(i) && !ANYOF_BITMAP_TEST(ret,i)) {
9992 set_regclass_bit(pRExC_state, ret, (U8) i, &l1_fold_invlist, &unicode_alternate);
9995 for (i = prevvalue; i <= ceilvalue; i++)
9996 if (isUPPER(i) && !ANYOF_BITMAP_TEST(ret,i)) {
9998 set_regclass_bit(pRExC_state, ret, (U8) i, &l1_fold_invlist, &unicode_alternate);
10004 for (i = prevvalue; i <= ceilvalue; i++) {
10005 stored += set_regclass_bit(pRExC_state, ret, (U8) i, &l1_fold_invlist, &unicode_alternate);
10009 const UV prevnatvalue = NATIVE_TO_UNI(prevvalue);
10010 const UV natvalue = NATIVE_TO_UNI(value);
10011 nonbitmap = add_range_to_invlist(nonbitmap, prevnatvalue, natvalue);
10014 literal_endpoint = 0;
10018 range = 0; /* this range (if it was one) is done now */
10025 /****** !SIZE_ONLY AFTER HERE *********/
10027 /* If folding and there are code points above 255, we calculate all
10028 * characters that could fold to or from the ones already on the list */
10029 if (FOLD && nonbitmap) {
10032 HV* fold_intersection;
10035 /* This is a list of all the characters that participate in folds
10036 * (except marks, etc in multi-char folds */
10037 if (! PL_utf8_foldable) {
10038 SV* swash = swash_init("utf8", "Cased", &PL_sv_undef, 1, 0);
10039 PL_utf8_foldable = _swash_to_invlist(swash);
10042 /* This is a hash that for a particular fold gives all characters
10043 * that are involved in it */
10044 if (! PL_utf8_foldclosures) {
10046 /* If we were unable to find any folds, then we likely won't be
10047 * able to find the closures. So just create an empty list.
10048 * Folding will effectively be restricted to the non-Unicode rules
10049 * hard-coded into Perl. (This case happens legitimately during
10050 * compilation of Perl itself before the Unicode tables are
10052 if (invlist_len(PL_utf8_foldable) == 0) {
10053 PL_utf8_foldclosures = _new_invlist(0);
10055 /* If the folds haven't been read in, call a fold function
10057 if (! PL_utf8_tofold) {
10058 U8 dummy[UTF8_MAXBYTES+1];
10060 to_utf8_fold((U8*) "A", dummy, &dummy_len);
10062 PL_utf8_foldclosures = _swash_inversion_hash(PL_utf8_tofold);
10066 /* Only the characters in this class that participate in folds need
10067 * be checked. Get the intersection of this class and all the
10068 * possible characters that are foldable. This can quickly narrow
10069 * down a large class */
10070 fold_intersection = invlist_intersection(PL_utf8_foldable, nonbitmap);
10072 /* Now look at the foldable characters in this class individually */
10073 fold_list = invlist_array(fold_intersection);
10074 for (i = 0; i < invlist_len(fold_intersection); i++) {
10077 /* The next entry is the beginning of the range that is in the
10079 UV start = fold_list[i++];
10082 /* The next entry is the beginning of the next range, which
10083 * isn't in the class, so the end of the current range is one
10084 * less than that */
10085 UV end = fold_list[i] - 1;
10087 /* Look at every character in the range */
10088 for (j = start; j <= end; j++) {
10091 U8 foldbuf[UTF8_MAXBYTES_CASE+1];
10093 const UV f = to_uni_fold(j, foldbuf, &foldlen);
10095 if (foldlen > (STRLEN)UNISKIP(f)) {
10097 /* Any multicharacter foldings (disallowed in
10098 * lookbehind patterns) require the following
10099 * transform: [ABCDEF] -> (?:[ABCabcDEFd]|pq|rst) where
10100 * E folds into "pq" and F folds into "rst", all other
10101 * characters fold to single characters. We save away
10102 * these multicharacter foldings, to be later saved as
10103 * part of the additional "s" data. */
10104 if (! RExC_in_lookbehind) {
10106 U8* e = foldbuf + foldlen;
10108 /* If any of the folded characters of this are in
10109 * the Latin1 range, tell the regex engine that
10110 * this can match a non-utf8 target string. The
10111 * only multi-byte fold whose source is in the
10112 * Latin1 range (U+00DF) applies only when the
10113 * target string is utf8, or under unicode rules */
10114 if (j > 255 || AT_LEAST_UNI_SEMANTICS) {
10117 /* Can't mix ascii with non- under /aa */
10118 if (MORE_ASCII_RESTRICTED
10119 && (isASCII(*loc) != isASCII(j)))
10121 goto end_multi_fold;
10123 if (UTF8_IS_INVARIANT(*loc)
10124 || UTF8_IS_DOWNGRADEABLE_START(*loc))
10126 /* Can't mix above and below 256 under
10129 goto end_multi_fold;
10132 |= ANYOF_NONBITMAP_NON_UTF8;
10135 loc += UTF8SKIP(loc);
10139 add_alternate(&unicode_alternate, foldbuf, foldlen);
10143 /* This is special-cased, as it is the only letter which
10144 * has both a multi-fold and single-fold in Latin1. All
10145 * the other chars that have single and multi-folds are
10146 * always in utf8, and the utf8 folding algorithm catches
10148 if (! LOC && j == LATIN_CAPITAL_LETTER_SHARP_S) {
10149 stored += set_regclass_bit(pRExC_state,
10151 LATIN_SMALL_LETTER_SHARP_S,
10152 &l1_fold_invlist, &unicode_alternate);
10156 /* Single character fold. Add everything in its fold
10157 * closure to the list that this node should match */
10160 /* The fold closures data structure is a hash with the
10161 * keys being every character that is folded to, like
10162 * 'k', and the values each an array of everything that
10163 * folds to its key. e.g. [ 'k', 'K', KELVIN_SIGN ] */
10164 if ((listp = hv_fetch(PL_utf8_foldclosures,
10165 (char *) foldbuf, foldlen, FALSE)))
10167 AV* list = (AV*) *listp;
10169 for (k = 0; k <= av_len(list); k++) {
10170 SV** c_p = av_fetch(list, k, FALSE);
10173 Perl_croak(aTHX_ "panic: invalid PL_utf8_foldclosures structure");
10177 /* /aa doesn't allow folds between ASCII and
10178 * non-; /l doesn't allow them between above
10180 if ((MORE_ASCII_RESTRICTED
10181 && (isASCII(c) != isASCII(j)))
10182 || (LOC && ((c < 256) != (j < 256))))
10187 if (c < 256 && AT_LEAST_UNI_SEMANTICS) {
10188 stored += set_regclass_bit(pRExC_state,
10191 &l1_fold_invlist, &unicode_alternate);
10193 /* It may be that the code point is already
10194 * in this range or already in the bitmap,
10195 * in which case we need do nothing */
10196 else if ((c < start || c > end)
10198 || ! ANYOF_BITMAP_TEST(ret, c)))
10200 nonbitmap = add_cp_to_invlist(nonbitmap, c);
10207 invlist_destroy(fold_intersection);
10210 /* Combine the two lists into one. */
10211 if (l1_fold_invlist) {
10213 nonbitmap = invlist_union(nonbitmap, l1_fold_invlist);
10216 nonbitmap = l1_fold_invlist;
10220 /* Here, we have calculated what code points should be in the character
10221 * class. Now we can see about various optimizations. Fold calculation
10222 * needs to take place before inversion. Otherwise /[^k]/i would invert to
10223 * include K, which under /i would match k. */
10225 /* Optimize inverted simple patterns (e.g. [^a-z]). Note that we haven't
10226 * set the FOLD flag yet, so this this does optimize those. It doesn't
10227 * optimize locale. Doing so perhaps could be done as long as there is
10228 * nothing like \w in it; some thought also would have to be given to the
10229 * interaction with above 0x100 chars */
10231 && (ANYOF_FLAGS(ret) & ANYOF_FLAGS_ALL) == ANYOF_INVERT
10232 && ! unicode_alternate
10234 && SvCUR(listsv) == initial_listsv_len)
10236 for (value = 0; value < ANYOF_BITMAP_SIZE; ++value)
10237 ANYOF_BITMAP(ret)[value] ^= 0xFF;
10238 stored = 256 - stored;
10240 /* The inversion means that everything above 255 is matched; and at the
10241 * same time we clear the invert flag */
10242 ANYOF_FLAGS(ret) = ANYOF_UNICODE_ALL;
10245 /* Folding in the bitmap is taken care of above, but not for locale (for
10246 * which we have to wait to see what folding is in effect at runtime), and
10247 * for things not in the bitmap. Set run-time fold flag for these */
10248 if (FOLD && (LOC || nonbitmap || unicode_alternate)) {
10249 ANYOF_FLAGS(ret) |= ANYOF_LOC_NONBITMAP_FOLD;
10252 /* A single character class can be "optimized" into an EXACTish node.
10253 * Note that since we don't currently count how many characters there are
10254 * outside the bitmap, we are XXX missing optimization possibilities for
10255 * them. This optimization can't happen unless this is a truly single
10256 * character class, which means that it can't be an inversion into a
10257 * many-character class, and there must be no possibility of there being
10258 * things outside the bitmap. 'stored' (only) for locales doesn't include
10259 * \w, etc, so have to make a special test that they aren't present
10261 * Similarly A 2-character class of the very special form like [bB] can be
10262 * optimized into an EXACTFish node, but only for non-locales, and for
10263 * characters which only have the two folds; so things like 'fF' and 'Ii'
10264 * wouldn't work because they are part of the fold of 'LATIN SMALL LIGATURE
10267 && ! unicode_alternate
10268 && SvCUR(listsv) == initial_listsv_len
10269 && ! (ANYOF_FLAGS(ret) & (ANYOF_INVERT|ANYOF_UNICODE_ALL))
10270 && (((stored == 1 && ((! (ANYOF_FLAGS(ret) & ANYOF_LOCALE))
10271 || (! ANYOF_CLASS_TEST_ANY_SET(ret)))))
10272 || (stored == 2 && ((! (ANYOF_FLAGS(ret) & ANYOF_LOCALE))
10273 && (! _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(value))
10274 /* If the latest code point has a fold whose
10275 * bit is set, it must be the only other one */
10276 && ((prevvalue = PL_fold_latin1[value]) != (IV)value)
10277 && ANYOF_BITMAP_TEST(ret, prevvalue)))))
10279 /* Note that the information needed to decide to do this optimization
10280 * is not currently available until the 2nd pass, and that the actually
10281 * used EXACTish node takes less space than the calculated ANYOF node,
10282 * and hence the amount of space calculated in the first pass is larger
10283 * than actually used, so this optimization doesn't gain us any space.
10284 * But an EXACT node is faster than an ANYOF node, and can be combined
10285 * with any adjacent EXACT nodes later by the optimizer for further
10286 * gains. The speed of executing an EXACTF is similar to an ANYOF
10287 * node, so the optimization advantage comes from the ability to join
10288 * it to adjacent EXACT nodes */
10290 const char * cur_parse= RExC_parse;
10292 RExC_emit = (regnode *)orig_emit;
10293 RExC_parse = (char *)orig_parse;
10297 /* A locale node with one point can be folded; all the other cases
10298 * with folding will have two points, since we calculate them above
10300 if (ANYOF_FLAGS(ret) & ANYOF_LOC_NONBITMAP_FOLD) {
10306 } /* else 2 chars in the bit map: the folds of each other */
10307 else if (AT_LEAST_UNI_SEMANTICS || !isASCII(value)) {
10309 /* To join adjacent nodes, they must be the exact EXACTish type.
10310 * Try to use the most likely type, by using EXACTFU if the regex
10311 * calls for them, or is required because the character is
10315 else { /* Otherwise, more likely to be EXACTF type */
10319 ret = reg_node(pRExC_state, op);
10320 RExC_parse = (char *)cur_parse;
10321 if (UTF && ! NATIVE_IS_INVARIANT(value)) {
10322 *STRING(ret)= UTF8_EIGHT_BIT_HI((U8) value);
10323 *(STRING(ret) + 1)= UTF8_EIGHT_BIT_LO((U8) value);
10325 RExC_emit += STR_SZ(2);
10328 *STRING(ret)= (char)value;
10330 RExC_emit += STR_SZ(1);
10332 SvREFCNT_dec(listsv);
10337 UV* nonbitmap_array = invlist_array(nonbitmap);
10338 UV nonbitmap_len = invlist_len(nonbitmap);
10341 /* Here have the full list of items to match that aren't in the
10342 * bitmap. Convert to the structure that the rest of the code is
10343 * expecting. XXX That rest of the code should convert to this
10345 for (i = 0; i < nonbitmap_len; i++) {
10347 /* The next entry is the beginning of the range that is in the
10349 UV start = nonbitmap_array[i++];
10352 /* The next entry is the beginning of the next range, which isn't
10353 * in the class, so the end of the current range is one less than
10354 * that. But if there is no next range, it means that the range
10355 * begun by 'start' extends to infinity, which for this platform
10356 * ends at UV_MAX */
10357 if (i == nonbitmap_len) {
10361 end = nonbitmap_array[i] - 1;
10364 if (start == end) {
10365 Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\n", start);
10368 /* The \t sets the whole range */
10369 Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\t%04"UVxf"\n",
10374 invlist_destroy(nonbitmap);
10377 if (SvCUR(listsv) == initial_listsv_len && ! unicode_alternate) {
10378 ARG_SET(ret, ANYOF_NONBITMAP_EMPTY);
10379 SvREFCNT_dec(listsv);
10380 SvREFCNT_dec(unicode_alternate);
10384 AV * const av = newAV();
10386 /* The 0th element stores the character class description
10387 * in its textual form: used later (regexec.c:Perl_regclass_swash())
10388 * to initialize the appropriate swash (which gets stored in
10389 * the 1st element), and also useful for dumping the regnode.
10390 * The 2nd element stores the multicharacter foldings,
10391 * used later (regexec.c:S_reginclass()). */
10392 av_store(av, 0, listsv);
10393 av_store(av, 1, NULL);
10394 av_store(av, 2, MUTABLE_SV(unicode_alternate));
10395 if (unicode_alternate) { /* This node is variable length */
10398 rv = newRV_noinc(MUTABLE_SV(av));
10399 n = add_data(pRExC_state, 1, "s");
10400 RExC_rxi->data->data[n] = (void*)rv;
10408 /* reg_skipcomment()
10410 Absorbs an /x style # comments from the input stream.
10411 Returns true if there is more text remaining in the stream.
10412 Will set the REG_SEEN_RUN_ON_COMMENT flag if the comment
10413 terminates the pattern without including a newline.
10415 Note its the callers responsibility to ensure that we are
10416 actually in /x mode
10421 S_reg_skipcomment(pTHX_ RExC_state_t *pRExC_state)
10425 PERL_ARGS_ASSERT_REG_SKIPCOMMENT;
10427 while (RExC_parse < RExC_end)
10428 if (*RExC_parse++ == '\n') {
10433 /* we ran off the end of the pattern without ending
10434 the comment, so we have to add an \n when wrapping */
10435 RExC_seen |= REG_SEEN_RUN_ON_COMMENT;
10443 Advances the parse position, and optionally absorbs
10444 "whitespace" from the inputstream.
10446 Without /x "whitespace" means (?#...) style comments only,
10447 with /x this means (?#...) and # comments and whitespace proper.
10449 Returns the RExC_parse point from BEFORE the scan occurs.
10451 This is the /x friendly way of saying RExC_parse++.
10455 S_nextchar(pTHX_ RExC_state_t *pRExC_state)
10457 char* const retval = RExC_parse++;
10459 PERL_ARGS_ASSERT_NEXTCHAR;
10462 if (*RExC_parse == '(' && RExC_parse[1] == '?' &&
10463 RExC_parse[2] == '#') {
10464 while (*RExC_parse != ')') {
10465 if (RExC_parse == RExC_end)
10466 FAIL("Sequence (?#... not terminated");
10472 if (RExC_flags & RXf_PMf_EXTENDED) {
10473 if (isSPACE(*RExC_parse)) {
10477 else if (*RExC_parse == '#') {
10478 if ( reg_skipcomment( pRExC_state ) )
10487 - reg_node - emit a node
10489 STATIC regnode * /* Location. */
10490 S_reg_node(pTHX_ RExC_state_t *pRExC_state, U8 op)
10493 register regnode *ptr;
10494 regnode * const ret = RExC_emit;
10495 GET_RE_DEBUG_FLAGS_DECL;
10497 PERL_ARGS_ASSERT_REG_NODE;
10500 SIZE_ALIGN(RExC_size);
10504 if (RExC_emit >= RExC_emit_bound)
10505 Perl_croak(aTHX_ "panic: reg_node overrun trying to emit %d", op);
10507 NODE_ALIGN_FILL(ret);
10509 FILL_ADVANCE_NODE(ptr, op);
10510 #ifdef RE_TRACK_PATTERN_OFFSETS
10511 if (RExC_offsets) { /* MJD */
10512 MJD_OFFSET_DEBUG(("%s:%d: (op %s) %s %"UVuf" (len %"UVuf") (max %"UVuf").\n",
10513 "reg_node", __LINE__,
10515 (UV)(RExC_emit - RExC_emit_start) > RExC_offsets[0]
10516 ? "Overwriting end of array!\n" : "OK",
10517 (UV)(RExC_emit - RExC_emit_start),
10518 (UV)(RExC_parse - RExC_start),
10519 (UV)RExC_offsets[0]));
10520 Set_Node_Offset(RExC_emit, RExC_parse + (op == END));
10528 - reganode - emit a node with an argument
10530 STATIC regnode * /* Location. */
10531 S_reganode(pTHX_ RExC_state_t *pRExC_state, U8 op, U32 arg)
10534 register regnode *ptr;
10535 regnode * const ret = RExC_emit;
10536 GET_RE_DEBUG_FLAGS_DECL;
10538 PERL_ARGS_ASSERT_REGANODE;
10541 SIZE_ALIGN(RExC_size);
10546 assert(2==regarglen[op]+1);
10548 Anything larger than this has to allocate the extra amount.
10549 If we changed this to be:
10551 RExC_size += (1 + regarglen[op]);
10553 then it wouldn't matter. Its not clear what side effect
10554 might come from that so its not done so far.
10559 if (RExC_emit >= RExC_emit_bound)
10560 Perl_croak(aTHX_ "panic: reg_node overrun trying to emit %d", op);
10562 NODE_ALIGN_FILL(ret);
10564 FILL_ADVANCE_NODE_ARG(ptr, op, arg);
10565 #ifdef RE_TRACK_PATTERN_OFFSETS
10566 if (RExC_offsets) { /* MJD */
10567 MJD_OFFSET_DEBUG(("%s(%d): (op %s) %s %"UVuf" <- %"UVuf" (max %"UVuf").\n",
10571 (UV)(RExC_emit - RExC_emit_start) > RExC_offsets[0] ?
10572 "Overwriting end of array!\n" : "OK",
10573 (UV)(RExC_emit - RExC_emit_start),
10574 (UV)(RExC_parse - RExC_start),
10575 (UV)RExC_offsets[0]));
10576 Set_Cur_Node_Offset;
10584 - reguni - emit (if appropriate) a Unicode character
10587 S_reguni(pTHX_ const RExC_state_t *pRExC_state, UV uv, char* s)
10591 PERL_ARGS_ASSERT_REGUNI;
10593 return SIZE_ONLY ? UNISKIP(uv) : (uvchr_to_utf8((U8*)s, uv) - (U8*)s);
10597 - reginsert - insert an operator in front of already-emitted operand
10599 * Means relocating the operand.
10602 S_reginsert(pTHX_ RExC_state_t *pRExC_state, U8 op, regnode *opnd, U32 depth)
10605 register regnode *src;
10606 register regnode *dst;
10607 register regnode *place;
10608 const int offset = regarglen[(U8)op];
10609 const int size = NODE_STEP_REGNODE + offset;
10610 GET_RE_DEBUG_FLAGS_DECL;
10612 PERL_ARGS_ASSERT_REGINSERT;
10613 PERL_UNUSED_ARG(depth);
10614 /* (PL_regkind[(U8)op] == CURLY ? EXTRA_STEP_2ARGS : 0); */
10615 DEBUG_PARSE_FMT("inst"," - %s",PL_reg_name[op]);
10624 if (RExC_open_parens) {
10626 /*DEBUG_PARSE_FMT("inst"," - %"IVdf, (IV)RExC_npar);*/
10627 for ( paren=0 ; paren < RExC_npar ; paren++ ) {
10628 if ( RExC_open_parens[paren] >= opnd ) {
10629 /*DEBUG_PARSE_FMT("open"," - %d",size);*/
10630 RExC_open_parens[paren] += size;
10632 /*DEBUG_PARSE_FMT("open"," - %s","ok");*/
10634 if ( RExC_close_parens[paren] >= opnd ) {
10635 /*DEBUG_PARSE_FMT("close"," - %d",size);*/
10636 RExC_close_parens[paren] += size;
10638 /*DEBUG_PARSE_FMT("close"," - %s","ok");*/
10643 while (src > opnd) {
10644 StructCopy(--src, --dst, regnode);
10645 #ifdef RE_TRACK_PATTERN_OFFSETS
10646 if (RExC_offsets) { /* MJD 20010112 */
10647 MJD_OFFSET_DEBUG(("%s(%d): (op %s) %s copy %"UVuf" -> %"UVuf" (max %"UVuf").\n",
10651 (UV)(dst - RExC_emit_start) > RExC_offsets[0]
10652 ? "Overwriting end of array!\n" : "OK",
10653 (UV)(src - RExC_emit_start),
10654 (UV)(dst - RExC_emit_start),
10655 (UV)RExC_offsets[0]));
10656 Set_Node_Offset_To_R(dst-RExC_emit_start, Node_Offset(src));
10657 Set_Node_Length_To_R(dst-RExC_emit_start, Node_Length(src));
10663 place = opnd; /* Op node, where operand used to be. */
10664 #ifdef RE_TRACK_PATTERN_OFFSETS
10665 if (RExC_offsets) { /* MJD */
10666 MJD_OFFSET_DEBUG(("%s(%d): (op %s) %s %"UVuf" <- %"UVuf" (max %"UVuf").\n",
10670 (UV)(place - RExC_emit_start) > RExC_offsets[0]
10671 ? "Overwriting end of array!\n" : "OK",
10672 (UV)(place - RExC_emit_start),
10673 (UV)(RExC_parse - RExC_start),
10674 (UV)RExC_offsets[0]));
10675 Set_Node_Offset(place, RExC_parse);
10676 Set_Node_Length(place, 1);
10679 src = NEXTOPER(place);
10680 FILL_ADVANCE_NODE(place, op);
10681 Zero(src, offset, regnode);
10685 - regtail - set the next-pointer at the end of a node chain of p to val.
10686 - SEE ALSO: regtail_study
10688 /* TODO: All three parms should be const */
10690 S_regtail(pTHX_ RExC_state_t *pRExC_state, regnode *p, const regnode *val,U32 depth)
10693 register regnode *scan;
10694 GET_RE_DEBUG_FLAGS_DECL;
10696 PERL_ARGS_ASSERT_REGTAIL;
10698 PERL_UNUSED_ARG(depth);
10704 /* Find last node. */
10707 regnode * const temp = regnext(scan);
10709 SV * const mysv=sv_newmortal();
10710 DEBUG_PARSE_MSG((scan==p ? "tail" : ""));
10711 regprop(RExC_rx, mysv, scan);
10712 PerlIO_printf(Perl_debug_log, "~ %s (%d) %s %s\n",
10713 SvPV_nolen_const(mysv), REG_NODE_NUM(scan),
10714 (temp == NULL ? "->" : ""),
10715 (temp == NULL ? PL_reg_name[OP(val)] : "")
10723 if (reg_off_by_arg[OP(scan)]) {
10724 ARG_SET(scan, val - scan);
10727 NEXT_OFF(scan) = val - scan;
10733 - regtail_study - set the next-pointer at the end of a node chain of p to val.
10734 - Look for optimizable sequences at the same time.
10735 - currently only looks for EXACT chains.
10737 This is experimental code. The idea is to use this routine to perform
10738 in place optimizations on branches and groups as they are constructed,
10739 with the long term intention of removing optimization from study_chunk so
10740 that it is purely analytical.
10742 Currently only used when in DEBUG mode. The macro REGTAIL_STUDY() is used
10743 to control which is which.
10746 /* TODO: All four parms should be const */
10749 S_regtail_study(pTHX_ RExC_state_t *pRExC_state, regnode *p, const regnode *val,U32 depth)
10752 register regnode *scan;
10754 #ifdef EXPERIMENTAL_INPLACESCAN
10757 GET_RE_DEBUG_FLAGS_DECL;
10759 PERL_ARGS_ASSERT_REGTAIL_STUDY;
10765 /* Find last node. */
10769 regnode * const temp = regnext(scan);
10770 #ifdef EXPERIMENTAL_INPLACESCAN
10771 if (PL_regkind[OP(scan)] == EXACT)
10772 if (join_exact(pRExC_state,scan,&min,1,val,depth+1))
10776 switch (OP(scan)) {
10782 if( exact == PSEUDO )
10784 else if ( exact != OP(scan) )
10793 SV * const mysv=sv_newmortal();
10794 DEBUG_PARSE_MSG((scan==p ? "tsdy" : ""));
10795 regprop(RExC_rx, mysv, scan);
10796 PerlIO_printf(Perl_debug_log, "~ %s (%d) -> %s\n",
10797 SvPV_nolen_const(mysv),
10798 REG_NODE_NUM(scan),
10799 PL_reg_name[exact]);
10806 SV * const mysv_val=sv_newmortal();
10807 DEBUG_PARSE_MSG("");
10808 regprop(RExC_rx, mysv_val, val);
10809 PerlIO_printf(Perl_debug_log, "~ attach to %s (%"IVdf") offset to %"IVdf"\n",
10810 SvPV_nolen_const(mysv_val),
10811 (IV)REG_NODE_NUM(val),
10815 if (reg_off_by_arg[OP(scan)]) {
10816 ARG_SET(scan, val - scan);
10819 NEXT_OFF(scan) = val - scan;
10827 - regdump - dump a regexp onto Perl_debug_log in vaguely comprehensible form
10831 S_regdump_extflags(pTHX_ const char *lead, const U32 flags)
10837 for (bit=0; bit<32; bit++) {
10838 if (flags & (1<<bit)) {
10839 if ((1<<bit) & RXf_PMf_CHARSET) { /* Output separately, below */
10842 if (!set++ && lead)
10843 PerlIO_printf(Perl_debug_log, "%s",lead);
10844 PerlIO_printf(Perl_debug_log, "%s ",PL_reg_extflags_name[bit]);
10847 if ((cs = get_regex_charset(flags)) != REGEX_DEPENDS_CHARSET) {
10848 if (!set++ && lead) {
10849 PerlIO_printf(Perl_debug_log, "%s",lead);
10852 case REGEX_UNICODE_CHARSET:
10853 PerlIO_printf(Perl_debug_log, "UNICODE");
10855 case REGEX_LOCALE_CHARSET:
10856 PerlIO_printf(Perl_debug_log, "LOCALE");
10858 case REGEX_ASCII_RESTRICTED_CHARSET:
10859 PerlIO_printf(Perl_debug_log, "ASCII-RESTRICTED");
10861 case REGEX_ASCII_MORE_RESTRICTED_CHARSET:
10862 PerlIO_printf(Perl_debug_log, "ASCII-MORE_RESTRICTED");
10865 PerlIO_printf(Perl_debug_log, "UNKNOWN CHARACTER SET");
10871 PerlIO_printf(Perl_debug_log, "\n");
10873 PerlIO_printf(Perl_debug_log, "%s[none-set]\n",lead);
10879 Perl_regdump(pTHX_ const regexp *r)
10883 SV * const sv = sv_newmortal();
10884 SV *dsv= sv_newmortal();
10885 RXi_GET_DECL(r,ri);
10886 GET_RE_DEBUG_FLAGS_DECL;
10888 PERL_ARGS_ASSERT_REGDUMP;
10890 (void)dumpuntil(r, ri->program, ri->program + 1, NULL, NULL, sv, 0, 0);
10892 /* Header fields of interest. */
10893 if (r->anchored_substr) {
10894 RE_PV_QUOTED_DECL(s, 0, dsv, SvPVX_const(r->anchored_substr),
10895 RE_SV_DUMPLEN(r->anchored_substr), 30);
10896 PerlIO_printf(Perl_debug_log,
10897 "anchored %s%s at %"IVdf" ",
10898 s, RE_SV_TAIL(r->anchored_substr),
10899 (IV)r->anchored_offset);
10900 } else if (r->anchored_utf8) {
10901 RE_PV_QUOTED_DECL(s, 1, dsv, SvPVX_const(r->anchored_utf8),
10902 RE_SV_DUMPLEN(r->anchored_utf8), 30);
10903 PerlIO_printf(Perl_debug_log,
10904 "anchored utf8 %s%s at %"IVdf" ",
10905 s, RE_SV_TAIL(r->anchored_utf8),
10906 (IV)r->anchored_offset);
10908 if (r->float_substr) {
10909 RE_PV_QUOTED_DECL(s, 0, dsv, SvPVX_const(r->float_substr),
10910 RE_SV_DUMPLEN(r->float_substr), 30);
10911 PerlIO_printf(Perl_debug_log,
10912 "floating %s%s at %"IVdf"..%"UVuf" ",
10913 s, RE_SV_TAIL(r->float_substr),
10914 (IV)r->float_min_offset, (UV)r->float_max_offset);
10915 } else if (r->float_utf8) {
10916 RE_PV_QUOTED_DECL(s, 1, dsv, SvPVX_const(r->float_utf8),
10917 RE_SV_DUMPLEN(r->float_utf8), 30);
10918 PerlIO_printf(Perl_debug_log,
10919 "floating utf8 %s%s at %"IVdf"..%"UVuf" ",
10920 s, RE_SV_TAIL(r->float_utf8),
10921 (IV)r->float_min_offset, (UV)r->float_max_offset);
10923 if (r->check_substr || r->check_utf8)
10924 PerlIO_printf(Perl_debug_log,
10926 (r->check_substr == r->float_substr
10927 && r->check_utf8 == r->float_utf8
10928 ? "(checking floating" : "(checking anchored"));
10929 if (r->extflags & RXf_NOSCAN)
10930 PerlIO_printf(Perl_debug_log, " noscan");
10931 if (r->extflags & RXf_CHECK_ALL)
10932 PerlIO_printf(Perl_debug_log, " isall");
10933 if (r->check_substr || r->check_utf8)
10934 PerlIO_printf(Perl_debug_log, ") ");
10936 if (ri->regstclass) {
10937 regprop(r, sv, ri->regstclass);
10938 PerlIO_printf(Perl_debug_log, "stclass %s ", SvPVX_const(sv));
10940 if (r->extflags & RXf_ANCH) {
10941 PerlIO_printf(Perl_debug_log, "anchored");
10942 if (r->extflags & RXf_ANCH_BOL)
10943 PerlIO_printf(Perl_debug_log, "(BOL)");
10944 if (r->extflags & RXf_ANCH_MBOL)
10945 PerlIO_printf(Perl_debug_log, "(MBOL)");
10946 if (r->extflags & RXf_ANCH_SBOL)
10947 PerlIO_printf(Perl_debug_log, "(SBOL)");
10948 if (r->extflags & RXf_ANCH_GPOS)
10949 PerlIO_printf(Perl_debug_log, "(GPOS)");
10950 PerlIO_putc(Perl_debug_log, ' ');
10952 if (r->extflags & RXf_GPOS_SEEN)
10953 PerlIO_printf(Perl_debug_log, "GPOS:%"UVuf" ", (UV)r->gofs);
10954 if (r->intflags & PREGf_SKIP)
10955 PerlIO_printf(Perl_debug_log, "plus ");
10956 if (r->intflags & PREGf_IMPLICIT)
10957 PerlIO_printf(Perl_debug_log, "implicit ");
10958 PerlIO_printf(Perl_debug_log, "minlen %"IVdf" ", (IV)r->minlen);
10959 if (r->extflags & RXf_EVAL_SEEN)
10960 PerlIO_printf(Perl_debug_log, "with eval ");
10961 PerlIO_printf(Perl_debug_log, "\n");
10962 DEBUG_FLAGS_r(regdump_extflags("r->extflags: ",r->extflags));
10964 PERL_ARGS_ASSERT_REGDUMP;
10965 PERL_UNUSED_CONTEXT;
10966 PERL_UNUSED_ARG(r);
10967 #endif /* DEBUGGING */
10971 - regprop - printable representation of opcode
10973 #define EMIT_ANYOF_TEST_SEPARATOR(do_sep,sv,flags) \
10976 Perl_sv_catpvf(aTHX_ sv,"%s][%s",PL_colors[1],PL_colors[0]); \
10977 if (flags & ANYOF_INVERT) \
10978 /*make sure the invert info is in each */ \
10979 sv_catpvs(sv, "^"); \
10985 Perl_regprop(pTHX_ const regexp *prog, SV *sv, const regnode *o)
10990 RXi_GET_DECL(prog,progi);
10991 GET_RE_DEBUG_FLAGS_DECL;
10993 PERL_ARGS_ASSERT_REGPROP;
10997 if (OP(o) > REGNODE_MAX) /* regnode.type is unsigned */
10998 /* It would be nice to FAIL() here, but this may be called from
10999 regexec.c, and it would be hard to supply pRExC_state. */
11000 Perl_croak(aTHX_ "Corrupted regexp opcode %d > %d", (int)OP(o), (int)REGNODE_MAX);
11001 sv_catpv(sv, PL_reg_name[OP(o)]); /* Take off const! */
11003 k = PL_regkind[OP(o)];
11006 sv_catpvs(sv, " ");
11007 /* Using is_utf8_string() (via PERL_PV_UNI_DETECT)
11008 * is a crude hack but it may be the best for now since
11009 * we have no flag "this EXACTish node was UTF-8"
11011 pv_pretty(sv, STRING(o), STR_LEN(o), 60, PL_colors[0], PL_colors[1],
11012 PERL_PV_ESCAPE_UNI_DETECT |
11013 PERL_PV_ESCAPE_NONASCII |
11014 PERL_PV_PRETTY_ELLIPSES |
11015 PERL_PV_PRETTY_LTGT |
11016 PERL_PV_PRETTY_NOCLEAR
11018 } else if (k == TRIE) {
11019 /* print the details of the trie in dumpuntil instead, as
11020 * progi->data isn't available here */
11021 const char op = OP(o);
11022 const U32 n = ARG(o);
11023 const reg_ac_data * const ac = IS_TRIE_AC(op) ?
11024 (reg_ac_data *)progi->data->data[n] :
11026 const reg_trie_data * const trie
11027 = (reg_trie_data*)progi->data->data[!IS_TRIE_AC(op) ? n : ac->trie];
11029 Perl_sv_catpvf(aTHX_ sv, "-%s",PL_reg_name[o->flags]);
11030 DEBUG_TRIE_COMPILE_r(
11031 Perl_sv_catpvf(aTHX_ sv,
11032 "<S:%"UVuf"/%"IVdf" W:%"UVuf" L:%"UVuf"/%"UVuf" C:%"UVuf"/%"UVuf">",
11033 (UV)trie->startstate,
11034 (IV)trie->statecount-1, /* -1 because of the unused 0 element */
11035 (UV)trie->wordcount,
11038 (UV)TRIE_CHARCOUNT(trie),
11039 (UV)trie->uniquecharcount
11042 if ( IS_ANYOF_TRIE(op) || trie->bitmap ) {
11044 int rangestart = -1;
11045 U8* bitmap = IS_ANYOF_TRIE(op) ? (U8*)ANYOF_BITMAP(o) : (U8*)TRIE_BITMAP(trie);
11046 sv_catpvs(sv, "[");
11047 for (i = 0; i <= 256; i++) {
11048 if (i < 256 && BITMAP_TEST(bitmap,i)) {
11049 if (rangestart == -1)
11051 } else if (rangestart != -1) {
11052 if (i <= rangestart + 3)
11053 for (; rangestart < i; rangestart++)
11054 put_byte(sv, rangestart);
11056 put_byte(sv, rangestart);
11057 sv_catpvs(sv, "-");
11058 put_byte(sv, i - 1);
11063 sv_catpvs(sv, "]");
11066 } else if (k == CURLY) {
11067 if (OP(o) == CURLYM || OP(o) == CURLYN || OP(o) == CURLYX)
11068 Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* Parenth number */
11069 Perl_sv_catpvf(aTHX_ sv, " {%d,%d}", ARG1(o), ARG2(o));
11071 else if (k == WHILEM && o->flags) /* Ordinal/of */
11072 Perl_sv_catpvf(aTHX_ sv, "[%d/%d]", o->flags & 0xf, o->flags>>4);
11073 else if (k == REF || k == OPEN || k == CLOSE || k == GROUPP || OP(o)==ACCEPT) {
11074 Perl_sv_catpvf(aTHX_ sv, "%d", (int)ARG(o)); /* Parenth number */
11075 if ( RXp_PAREN_NAMES(prog) ) {
11076 if ( k != REF || (OP(o) < NREF)) {
11077 AV *list= MUTABLE_AV(progi->data->data[progi->name_list_idx]);
11078 SV **name= av_fetch(list, ARG(o), 0 );
11080 Perl_sv_catpvf(aTHX_ sv, " '%"SVf"'", SVfARG(*name));
11083 AV *list= MUTABLE_AV(progi->data->data[ progi->name_list_idx ]);
11084 SV *sv_dat= MUTABLE_SV(progi->data->data[ ARG( o ) ]);
11085 I32 *nums=(I32*)SvPVX(sv_dat);
11086 SV **name= av_fetch(list, nums[0], 0 );
11089 for ( n=0; n<SvIVX(sv_dat); n++ ) {
11090 Perl_sv_catpvf(aTHX_ sv, "%s%"IVdf,
11091 (n ? "," : ""), (IV)nums[n]);
11093 Perl_sv_catpvf(aTHX_ sv, " '%"SVf"'", SVfARG(*name));
11097 } else if (k == GOSUB)
11098 Perl_sv_catpvf(aTHX_ sv, "%d[%+d]", (int)ARG(o),(int)ARG2L(o)); /* Paren and offset */
11099 else if (k == VERB) {
11101 Perl_sv_catpvf(aTHX_ sv, ":%"SVf,
11102 SVfARG((MUTABLE_SV(progi->data->data[ ARG( o ) ]))));
11103 } else if (k == LOGICAL)
11104 Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* 2: embedded, otherwise 1 */
11105 else if (k == FOLDCHAR)
11106 Perl_sv_catpvf(aTHX_ sv, "[0x%"UVXf"]", PTR2UV(ARG(o)) );
11107 else if (k == ANYOF) {
11108 int i, rangestart = -1;
11109 const U8 flags = ANYOF_FLAGS(o);
11112 /* Should be synchronized with * ANYOF_ #xdefines in regcomp.h */
11113 static const char * const anyofs[] = {
11146 if (flags & ANYOF_LOCALE)
11147 sv_catpvs(sv, "{loc}");
11148 if (flags & ANYOF_LOC_NONBITMAP_FOLD)
11149 sv_catpvs(sv, "{i}");
11150 Perl_sv_catpvf(aTHX_ sv, "[%s", PL_colors[0]);
11151 if (flags & ANYOF_INVERT)
11152 sv_catpvs(sv, "^");
11154 /* output what the standard cp 0-255 bitmap matches */
11155 for (i = 0; i <= 256; i++) {
11156 if (i < 256 && ANYOF_BITMAP_TEST(o,i)) {
11157 if (rangestart == -1)
11159 } else if (rangestart != -1) {
11160 if (i <= rangestart + 3)
11161 for (; rangestart < i; rangestart++)
11162 put_byte(sv, rangestart);
11164 put_byte(sv, rangestart);
11165 sv_catpvs(sv, "-");
11166 put_byte(sv, i - 1);
11173 EMIT_ANYOF_TEST_SEPARATOR(do_sep,sv,flags);
11174 /* output any special charclass tests (used entirely under use locale) */
11175 if (ANYOF_CLASS_TEST_ANY_SET(o))
11176 for (i = 0; i < (int)(sizeof(anyofs)/sizeof(char*)); i++)
11177 if (ANYOF_CLASS_TEST(o,i)) {
11178 sv_catpv(sv, anyofs[i]);
11182 EMIT_ANYOF_TEST_SEPARATOR(do_sep,sv,flags);
11184 if (flags & ANYOF_NON_UTF8_LATIN1_ALL) {
11185 sv_catpvs(sv, "{non-utf8-latin1-all}");
11188 /* output information about the unicode matching */
11189 if (flags & ANYOF_UNICODE_ALL)
11190 sv_catpvs(sv, "{unicode_all}");
11191 else if (ANYOF_NONBITMAP(o))
11192 sv_catpvs(sv, "{unicode}");
11193 if (flags & ANYOF_NONBITMAP_NON_UTF8)
11194 sv_catpvs(sv, "{outside bitmap}");
11196 if (ANYOF_NONBITMAP(o)) {
11198 SV * const sw = regclass_swash(prog, o, FALSE, &lv, 0);
11202 U8 s[UTF8_MAXBYTES_CASE+1];
11204 for (i = 0; i <= 256; i++) { /* just the first 256 */
11205 uvchr_to_utf8(s, i);
11207 if (i < 256 && swash_fetch(sw, s, TRUE)) {
11208 if (rangestart == -1)
11210 } else if (rangestart != -1) {
11211 if (i <= rangestart + 3)
11212 for (; rangestart < i; rangestart++) {
11213 const U8 * const e = uvchr_to_utf8(s,rangestart);
11215 for(p = s; p < e; p++)
11219 const U8 *e = uvchr_to_utf8(s,rangestart);
11221 for (p = s; p < e; p++)
11223 sv_catpvs(sv, "-");
11224 e = uvchr_to_utf8(s, i-1);
11225 for (p = s; p < e; p++)
11232 sv_catpvs(sv, "..."); /* et cetera */
11236 char *s = savesvpv(lv);
11237 char * const origs = s;
11239 while (*s && *s != '\n')
11243 const char * const t = ++s;
11261 Perl_sv_catpvf(aTHX_ sv, "%s]", PL_colors[1]);
11263 else if (k == BRANCHJ && (OP(o) == UNLESSM || OP(o) == IFMATCH))
11264 Perl_sv_catpvf(aTHX_ sv, "[%d]", -(o->flags));
11266 PERL_UNUSED_CONTEXT;
11267 PERL_UNUSED_ARG(sv);
11268 PERL_UNUSED_ARG(o);
11269 PERL_UNUSED_ARG(prog);
11270 #endif /* DEBUGGING */
11274 Perl_re_intuit_string(pTHX_ REGEXP * const r)
11275 { /* Assume that RE_INTUIT is set */
11277 struct regexp *const prog = (struct regexp *)SvANY(r);
11278 GET_RE_DEBUG_FLAGS_DECL;
11280 PERL_ARGS_ASSERT_RE_INTUIT_STRING;
11281 PERL_UNUSED_CONTEXT;
11285 const char * const s = SvPV_nolen_const(prog->check_substr
11286 ? prog->check_substr : prog->check_utf8);
11288 if (!PL_colorset) reginitcolors();
11289 PerlIO_printf(Perl_debug_log,
11290 "%sUsing REx %ssubstr:%s \"%s%.60s%s%s\"\n",
11292 prog->check_substr ? "" : "utf8 ",
11293 PL_colors[5],PL_colors[0],
11296 (strlen(s) > 60 ? "..." : ""));
11299 return prog->check_substr ? prog->check_substr : prog->check_utf8;
11305 handles refcounting and freeing the perl core regexp structure. When
11306 it is necessary to actually free the structure the first thing it
11307 does is call the 'free' method of the regexp_engine associated to
11308 the regexp, allowing the handling of the void *pprivate; member
11309 first. (This routine is not overridable by extensions, which is why
11310 the extensions free is called first.)
11312 See regdupe and regdupe_internal if you change anything here.
11314 #ifndef PERL_IN_XSUB_RE
11316 Perl_pregfree(pTHX_ REGEXP *r)
11322 Perl_pregfree2(pTHX_ REGEXP *rx)
11325 struct regexp *const r = (struct regexp *)SvANY(rx);
11326 GET_RE_DEBUG_FLAGS_DECL;
11328 PERL_ARGS_ASSERT_PREGFREE2;
11330 if (r->mother_re) {
11331 ReREFCNT_dec(r->mother_re);
11333 CALLREGFREE_PVT(rx); /* free the private data */
11334 SvREFCNT_dec(RXp_PAREN_NAMES(r));
11337 SvREFCNT_dec(r->anchored_substr);
11338 SvREFCNT_dec(r->anchored_utf8);
11339 SvREFCNT_dec(r->float_substr);
11340 SvREFCNT_dec(r->float_utf8);
11341 Safefree(r->substrs);
11343 RX_MATCH_COPY_FREE(rx);
11344 #ifdef PERL_OLD_COPY_ON_WRITE
11345 SvREFCNT_dec(r->saved_copy);
11352 This is a hacky workaround to the structural issue of match results
11353 being stored in the regexp structure which is in turn stored in
11354 PL_curpm/PL_reg_curpm. The problem is that due to qr// the pattern
11355 could be PL_curpm in multiple contexts, and could require multiple
11356 result sets being associated with the pattern simultaneously, such
11357 as when doing a recursive match with (??{$qr})
11359 The solution is to make a lightweight copy of the regexp structure
11360 when a qr// is returned from the code executed by (??{$qr}) this
11361 lightweight copy doesn't actually own any of its data except for
11362 the starp/end and the actual regexp structure itself.
11368 Perl_reg_temp_copy (pTHX_ REGEXP *ret_x, REGEXP *rx)
11370 struct regexp *ret;
11371 struct regexp *const r = (struct regexp *)SvANY(rx);
11372 register const I32 npar = r->nparens+1;
11374 PERL_ARGS_ASSERT_REG_TEMP_COPY;
11377 ret_x = (REGEXP*) newSV_type(SVt_REGEXP);
11378 ret = (struct regexp *)SvANY(ret_x);
11380 (void)ReREFCNT_inc(rx);
11381 /* We can take advantage of the existing "copied buffer" mechanism in SVs
11382 by pointing directly at the buffer, but flagging that the allocated
11383 space in the copy is zero. As we've just done a struct copy, it's now
11384 a case of zero-ing that, rather than copying the current length. */
11385 SvPV_set(ret_x, RX_WRAPPED(rx));
11386 SvFLAGS(ret_x) |= SvFLAGS(rx) & (SVf_POK|SVp_POK|SVf_UTF8);
11387 memcpy(&(ret->xpv_cur), &(r->xpv_cur),
11388 sizeof(regexp) - STRUCT_OFFSET(regexp, xpv_cur));
11389 SvLEN_set(ret_x, 0);
11390 SvSTASH_set(ret_x, NULL);
11391 SvMAGIC_set(ret_x, NULL);
11392 Newx(ret->offs, npar, regexp_paren_pair);
11393 Copy(r->offs, ret->offs, npar, regexp_paren_pair);
11395 Newx(ret->substrs, 1, struct reg_substr_data);
11396 StructCopy(r->substrs, ret->substrs, struct reg_substr_data);
11398 SvREFCNT_inc_void(ret->anchored_substr);
11399 SvREFCNT_inc_void(ret->anchored_utf8);
11400 SvREFCNT_inc_void(ret->float_substr);
11401 SvREFCNT_inc_void(ret->float_utf8);
11403 /* check_substr and check_utf8, if non-NULL, point to either their
11404 anchored or float namesakes, and don't hold a second reference. */
11406 RX_MATCH_COPIED_off(ret_x);
11407 #ifdef PERL_OLD_COPY_ON_WRITE
11408 ret->saved_copy = NULL;
11410 ret->mother_re = rx;
11416 /* regfree_internal()
11418 Free the private data in a regexp. This is overloadable by
11419 extensions. Perl takes care of the regexp structure in pregfree(),
11420 this covers the *pprivate pointer which technically perl doesn't
11421 know about, however of course we have to handle the
11422 regexp_internal structure when no extension is in use.
11424 Note this is called before freeing anything in the regexp
11429 Perl_regfree_internal(pTHX_ REGEXP * const rx)
11432 struct regexp *const r = (struct regexp *)SvANY(rx);
11433 RXi_GET_DECL(r,ri);
11434 GET_RE_DEBUG_FLAGS_DECL;
11436 PERL_ARGS_ASSERT_REGFREE_INTERNAL;
11442 SV *dsv= sv_newmortal();
11443 RE_PV_QUOTED_DECL(s, RX_UTF8(rx),
11444 dsv, RX_PRECOMP(rx), RX_PRELEN(rx), 60);
11445 PerlIO_printf(Perl_debug_log,"%sFreeing REx:%s %s\n",
11446 PL_colors[4],PL_colors[5],s);
11449 #ifdef RE_TRACK_PATTERN_OFFSETS
11451 Safefree(ri->u.offsets); /* 20010421 MJD */
11454 int n = ri->data->count;
11455 PAD* new_comppad = NULL;
11460 /* If you add a ->what type here, update the comment in regcomp.h */
11461 switch (ri->data->what[n]) {
11466 SvREFCNT_dec(MUTABLE_SV(ri->data->data[n]));
11469 Safefree(ri->data->data[n]);
11472 new_comppad = MUTABLE_AV(ri->data->data[n]);
11475 if (new_comppad == NULL)
11476 Perl_croak(aTHX_ "panic: pregfree comppad");
11477 PAD_SAVE_LOCAL(old_comppad,
11478 /* Watch out for global destruction's random ordering. */
11479 (SvTYPE(new_comppad) == SVt_PVAV) ? new_comppad : NULL
11482 refcnt = OpREFCNT_dec((OP_4tree*)ri->data->data[n]);
11485 op_free((OP_4tree*)ri->data->data[n]);
11487 PAD_RESTORE_LOCAL(old_comppad);
11488 SvREFCNT_dec(MUTABLE_SV(new_comppad));
11489 new_comppad = NULL;
11494 { /* Aho Corasick add-on structure for a trie node.
11495 Used in stclass optimization only */
11497 reg_ac_data *aho=(reg_ac_data*)ri->data->data[n];
11499 refcount = --aho->refcount;
11502 PerlMemShared_free(aho->states);
11503 PerlMemShared_free(aho->fail);
11504 /* do this last!!!! */
11505 PerlMemShared_free(ri->data->data[n]);
11506 PerlMemShared_free(ri->regstclass);
11512 /* trie structure. */
11514 reg_trie_data *trie=(reg_trie_data*)ri->data->data[n];
11516 refcount = --trie->refcount;
11519 PerlMemShared_free(trie->charmap);
11520 PerlMemShared_free(trie->states);
11521 PerlMemShared_free(trie->trans);
11523 PerlMemShared_free(trie->bitmap);
11525 PerlMemShared_free(trie->jump);
11526 PerlMemShared_free(trie->wordinfo);
11527 /* do this last!!!! */
11528 PerlMemShared_free(ri->data->data[n]);
11533 Perl_croak(aTHX_ "panic: regfree data code '%c'", ri->data->what[n]);
11536 Safefree(ri->data->what);
11537 Safefree(ri->data);
11543 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
11544 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
11545 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
11548 re_dup - duplicate a regexp.
11550 This routine is expected to clone a given regexp structure. It is only
11551 compiled under USE_ITHREADS.
11553 After all of the core data stored in struct regexp is duplicated
11554 the regexp_engine.dupe method is used to copy any private data
11555 stored in the *pprivate pointer. This allows extensions to handle
11556 any duplication it needs to do.
11558 See pregfree() and regfree_internal() if you change anything here.
11560 #if defined(USE_ITHREADS)
11561 #ifndef PERL_IN_XSUB_RE
11563 Perl_re_dup_guts(pTHX_ const REGEXP *sstr, REGEXP *dstr, CLONE_PARAMS *param)
11567 const struct regexp *r = (const struct regexp *)SvANY(sstr);
11568 struct regexp *ret = (struct regexp *)SvANY(dstr);
11570 PERL_ARGS_ASSERT_RE_DUP_GUTS;
11572 npar = r->nparens+1;
11573 Newx(ret->offs, npar, regexp_paren_pair);
11574 Copy(r->offs, ret->offs, npar, regexp_paren_pair);
11576 /* no need to copy these */
11577 Newx(ret->swap, npar, regexp_paren_pair);
11580 if (ret->substrs) {
11581 /* Do it this way to avoid reading from *r after the StructCopy().
11582 That way, if any of the sv_dup_inc()s dislodge *r from the L1
11583 cache, it doesn't matter. */
11584 const bool anchored = r->check_substr
11585 ? r->check_substr == r->anchored_substr
11586 : r->check_utf8 == r->anchored_utf8;
11587 Newx(ret->substrs, 1, struct reg_substr_data);
11588 StructCopy(r->substrs, ret->substrs, struct reg_substr_data);
11590 ret->anchored_substr = sv_dup_inc(ret->anchored_substr, param);
11591 ret->anchored_utf8 = sv_dup_inc(ret->anchored_utf8, param);
11592 ret->float_substr = sv_dup_inc(ret->float_substr, param);
11593 ret->float_utf8 = sv_dup_inc(ret->float_utf8, param);
11595 /* check_substr and check_utf8, if non-NULL, point to either their
11596 anchored or float namesakes, and don't hold a second reference. */
11598 if (ret->check_substr) {
11600 assert(r->check_utf8 == r->anchored_utf8);
11601 ret->check_substr = ret->anchored_substr;
11602 ret->check_utf8 = ret->anchored_utf8;
11604 assert(r->check_substr == r->float_substr);
11605 assert(r->check_utf8 == r->float_utf8);
11606 ret->check_substr = ret->float_substr;
11607 ret->check_utf8 = ret->float_utf8;
11609 } else if (ret->check_utf8) {
11611 ret->check_utf8 = ret->anchored_utf8;
11613 ret->check_utf8 = ret->float_utf8;
11618 RXp_PAREN_NAMES(ret) = hv_dup_inc(RXp_PAREN_NAMES(ret), param);
11621 RXi_SET(ret,CALLREGDUPE_PVT(dstr,param));
11623 if (RX_MATCH_COPIED(dstr))
11624 ret->subbeg = SAVEPVN(ret->subbeg, ret->sublen);
11626 ret->subbeg = NULL;
11627 #ifdef PERL_OLD_COPY_ON_WRITE
11628 ret->saved_copy = NULL;
11631 if (ret->mother_re) {
11632 if (SvPVX_const(dstr) == SvPVX_const(ret->mother_re)) {
11633 /* Our storage points directly to our mother regexp, but that's
11634 1: a buffer in a different thread
11635 2: something we no longer hold a reference on
11636 so we need to copy it locally. */
11637 /* Note we need to sue SvCUR() on our mother_re, because it, in
11638 turn, may well be pointing to its own mother_re. */
11639 SvPV_set(dstr, SAVEPVN(SvPVX_const(ret->mother_re),
11640 SvCUR(ret->mother_re)+1));
11641 SvLEN_set(dstr, SvCUR(ret->mother_re)+1);
11643 ret->mother_re = NULL;
11647 #endif /* PERL_IN_XSUB_RE */
11652 This is the internal complement to regdupe() which is used to copy
11653 the structure pointed to by the *pprivate pointer in the regexp.
11654 This is the core version of the extension overridable cloning hook.
11655 The regexp structure being duplicated will be copied by perl prior
11656 to this and will be provided as the regexp *r argument, however
11657 with the /old/ structures pprivate pointer value. Thus this routine
11658 may override any copying normally done by perl.
11660 It returns a pointer to the new regexp_internal structure.
11664 Perl_regdupe_internal(pTHX_ REGEXP * const rx, CLONE_PARAMS *param)
11667 struct regexp *const r = (struct regexp *)SvANY(rx);
11668 regexp_internal *reti;
11670 RXi_GET_DECL(r,ri);
11672 PERL_ARGS_ASSERT_REGDUPE_INTERNAL;
11674 npar = r->nparens+1;
11677 Newxc(reti, sizeof(regexp_internal) + len*sizeof(regnode), char, regexp_internal);
11678 Copy(ri->program, reti->program, len+1, regnode);
11681 reti->regstclass = NULL;
11684 struct reg_data *d;
11685 const int count = ri->data->count;
11688 Newxc(d, sizeof(struct reg_data) + count*sizeof(void *),
11689 char, struct reg_data);
11690 Newx(d->what, count, U8);
11693 for (i = 0; i < count; i++) {
11694 d->what[i] = ri->data->what[i];
11695 switch (d->what[i]) {
11696 /* legal options are one of: sSfpontTua
11697 see also regcomp.h and pregfree() */
11698 case 'a': /* actually an AV, but the dup function is identical. */
11701 case 'p': /* actually an AV, but the dup function is identical. */
11702 case 'u': /* actually an HV, but the dup function is identical. */
11703 d->data[i] = sv_dup_inc((const SV *)ri->data->data[i], param);
11706 /* This is cheating. */
11707 Newx(d->data[i], 1, struct regnode_charclass_class);
11708 StructCopy(ri->data->data[i], d->data[i],
11709 struct regnode_charclass_class);
11710 reti->regstclass = (regnode*)d->data[i];
11713 /* Compiled op trees are readonly and in shared memory,
11714 and can thus be shared without duplication. */
11716 d->data[i] = (void*)OpREFCNT_inc((OP*)ri->data->data[i]);
11720 /* Trie stclasses are readonly and can thus be shared
11721 * without duplication. We free the stclass in pregfree
11722 * when the corresponding reg_ac_data struct is freed.
11724 reti->regstclass= ri->regstclass;
11728 ((reg_trie_data*)ri->data->data[i])->refcount++;
11732 d->data[i] = ri->data->data[i];
11735 Perl_croak(aTHX_ "panic: re_dup unknown data code '%c'", ri->data->what[i]);
11744 reti->name_list_idx = ri->name_list_idx;
11746 #ifdef RE_TRACK_PATTERN_OFFSETS
11747 if (ri->u.offsets) {
11748 Newx(reti->u.offsets, 2*len+1, U32);
11749 Copy(ri->u.offsets, reti->u.offsets, 2*len+1, U32);
11752 SetProgLen(reti,len);
11755 return (void*)reti;
11758 #endif /* USE_ITHREADS */
11760 #ifndef PERL_IN_XSUB_RE
11763 - regnext - dig the "next" pointer out of a node
11766 Perl_regnext(pTHX_ register regnode *p)
11769 register I32 offset;
11774 if (OP(p) > REGNODE_MAX) { /* regnode.type is unsigned */
11775 Perl_croak(aTHX_ "Corrupted regexp opcode %d > %d", (int)OP(p), (int)REGNODE_MAX);
11778 offset = (reg_off_by_arg[OP(p)] ? ARG(p) : NEXT_OFF(p));
11787 S_re_croak2(pTHX_ const char* pat1,const char* pat2,...)
11790 STRLEN l1 = strlen(pat1);
11791 STRLEN l2 = strlen(pat2);
11794 const char *message;
11796 PERL_ARGS_ASSERT_RE_CROAK2;
11802 Copy(pat1, buf, l1 , char);
11803 Copy(pat2, buf + l1, l2 , char);
11804 buf[l1 + l2] = '\n';
11805 buf[l1 + l2 + 1] = '\0';
11807 /* ANSI variant takes additional second argument */
11808 va_start(args, pat2);
11812 msv = vmess(buf, &args);
11814 message = SvPV_const(msv,l1);
11817 Copy(message, buf, l1 , char);
11818 buf[l1-1] = '\0'; /* Overwrite \n */
11819 Perl_croak(aTHX_ "%s", buf);
11822 /* XXX Here's a total kludge. But we need to re-enter for swash routines. */
11824 #ifndef PERL_IN_XSUB_RE
11826 Perl_save_re_context(pTHX)
11830 struct re_save_state *state;
11832 SAVEVPTR(PL_curcop);
11833 SSGROW(SAVESTACK_ALLOC_FOR_RE_SAVE_STATE + 1);
11835 state = (struct re_save_state *)(PL_savestack + PL_savestack_ix);
11836 PL_savestack_ix += SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
11837 SSPUSHUV(SAVEt_RE_STATE);
11839 Copy(&PL_reg_state, state, 1, struct re_save_state);
11841 PL_reg_start_tmp = 0;
11842 PL_reg_start_tmpl = 0;
11843 PL_reg_oldsaved = NULL;
11844 PL_reg_oldsavedlen = 0;
11845 PL_reg_maxiter = 0;
11846 PL_reg_leftiter = 0;
11847 PL_reg_poscache = NULL;
11848 PL_reg_poscache_size = 0;
11849 #ifdef PERL_OLD_COPY_ON_WRITE
11853 /* Save $1..$n (#18107: UTF-8 s/(\w+)/uc($1)/e); AMS 20021106. */
11855 const REGEXP * const rx = PM_GETRE(PL_curpm);
11858 for (i = 1; i <= RX_NPARENS(rx); i++) {
11859 char digits[TYPE_CHARS(long)];
11860 const STRLEN len = my_snprintf(digits, sizeof(digits), "%lu", (long)i);
11861 GV *const *const gvp
11862 = (GV**)hv_fetch(PL_defstash, digits, len, 0);
11865 GV * const gv = *gvp;
11866 if (SvTYPE(gv) == SVt_PVGV && GvSV(gv))
11876 clear_re(pTHX_ void *r)
11879 ReREFCNT_dec((REGEXP *)r);
11885 S_put_byte(pTHX_ SV *sv, int c)
11887 PERL_ARGS_ASSERT_PUT_BYTE;
11889 /* Our definition of isPRINT() ignores locales, so only bytes that are
11890 not part of UTF-8 are considered printable. I assume that the same
11891 holds for UTF-EBCDIC.
11892 Also, code point 255 is not printable in either (it's E0 in EBCDIC,
11893 which Wikipedia says:
11895 EO, or Eight Ones, is an 8-bit EBCDIC character code represented as all
11896 ones (binary 1111 1111, hexadecimal FF). It is similar, but not
11897 identical, to the ASCII delete (DEL) or rubout control character.
11898 ) So the old condition can be simplified to !isPRINT(c) */
11901 Perl_sv_catpvf(aTHX_ sv, "\\x%02x", c);
11904 Perl_sv_catpvf(aTHX_ sv, "\\x{%x}", c);
11908 const char string = c;
11909 if (c == '-' || c == ']' || c == '\\' || c == '^')
11910 sv_catpvs(sv, "\\");
11911 sv_catpvn(sv, &string, 1);
11916 #define CLEAR_OPTSTART \
11917 if (optstart) STMT_START { \
11918 DEBUG_OPTIMISE_r(PerlIO_printf(Perl_debug_log, " (%"IVdf" nodes)\n", (IV)(node - optstart))); \
11922 #define DUMPUNTIL(b,e) CLEAR_OPTSTART; node=dumpuntil(r,start,(b),(e),last,sv,indent+1,depth+1);
11924 STATIC const regnode *
11925 S_dumpuntil(pTHX_ const regexp *r, const regnode *start, const regnode *node,
11926 const regnode *last, const regnode *plast,
11927 SV* sv, I32 indent, U32 depth)
11930 register U8 op = PSEUDO; /* Arbitrary non-END op. */
11931 register const regnode *next;
11932 const regnode *optstart= NULL;
11934 RXi_GET_DECL(r,ri);
11935 GET_RE_DEBUG_FLAGS_DECL;
11937 PERL_ARGS_ASSERT_DUMPUNTIL;
11939 #ifdef DEBUG_DUMPUNTIL
11940 PerlIO_printf(Perl_debug_log, "--- %d : %d - %d - %d\n",indent,node-start,
11941 last ? last-start : 0,plast ? plast-start : 0);
11944 if (plast && plast < last)
11947 while (PL_regkind[op] != END && (!last || node < last)) {
11948 /* While that wasn't END last time... */
11951 if (op == CLOSE || op == WHILEM)
11953 next = regnext((regnode *)node);
11956 if (OP(node) == OPTIMIZED) {
11957 if (!optstart && RE_DEBUG_FLAG(RE_DEBUG_COMPILE_OPTIMISE))
11964 regprop(r, sv, node);
11965 PerlIO_printf(Perl_debug_log, "%4"IVdf":%*s%s", (IV)(node - start),
11966 (int)(2*indent + 1), "", SvPVX_const(sv));
11968 if (OP(node) != OPTIMIZED) {
11969 if (next == NULL) /* Next ptr. */
11970 PerlIO_printf(Perl_debug_log, " (0)");
11971 else if (PL_regkind[(U8)op] == BRANCH && PL_regkind[OP(next)] != BRANCH )
11972 PerlIO_printf(Perl_debug_log, " (FAIL)");
11974 PerlIO_printf(Perl_debug_log, " (%"IVdf")", (IV)(next - start));
11975 (void)PerlIO_putc(Perl_debug_log, '\n');
11979 if (PL_regkind[(U8)op] == BRANCHJ) {
11982 register const regnode *nnode = (OP(next) == LONGJMP
11983 ? regnext((regnode *)next)
11985 if (last && nnode > last)
11987 DUMPUNTIL(NEXTOPER(NEXTOPER(node)), nnode);
11990 else if (PL_regkind[(U8)op] == BRANCH) {
11992 DUMPUNTIL(NEXTOPER(node), next);
11994 else if ( PL_regkind[(U8)op] == TRIE ) {
11995 const regnode *this_trie = node;
11996 const char op = OP(node);
11997 const U32 n = ARG(node);
11998 const reg_ac_data * const ac = op>=AHOCORASICK ?
11999 (reg_ac_data *)ri->data->data[n] :
12001 const reg_trie_data * const trie =
12002 (reg_trie_data*)ri->data->data[op<AHOCORASICK ? n : ac->trie];
12004 AV *const trie_words = MUTABLE_AV(ri->data->data[n + TRIE_WORDS_OFFSET]);
12006 const regnode *nextbranch= NULL;
12009 for (word_idx= 0; word_idx < (I32)trie->wordcount; word_idx++) {
12010 SV ** const elem_ptr = av_fetch(trie_words,word_idx,0);
12012 PerlIO_printf(Perl_debug_log, "%*s%s ",
12013 (int)(2*(indent+3)), "",
12014 elem_ptr ? pv_pretty(sv, SvPV_nolen_const(*elem_ptr), SvCUR(*elem_ptr), 60,
12015 PL_colors[0], PL_colors[1],
12016 (SvUTF8(*elem_ptr) ? PERL_PV_ESCAPE_UNI : 0) |
12017 PERL_PV_PRETTY_ELLIPSES |
12018 PERL_PV_PRETTY_LTGT
12023 U16 dist= trie->jump[word_idx+1];
12024 PerlIO_printf(Perl_debug_log, "(%"UVuf")\n",
12025 (UV)((dist ? this_trie + dist : next) - start));
12028 nextbranch= this_trie + trie->jump[0];
12029 DUMPUNTIL(this_trie + dist, nextbranch);
12031 if (nextbranch && PL_regkind[OP(nextbranch)]==BRANCH)
12032 nextbranch= regnext((regnode *)nextbranch);
12034 PerlIO_printf(Perl_debug_log, "\n");
12037 if (last && next > last)
12042 else if ( op == CURLY ) { /* "next" might be very big: optimizer */
12043 DUMPUNTIL(NEXTOPER(node) + EXTRA_STEP_2ARGS,
12044 NEXTOPER(node) + EXTRA_STEP_2ARGS + 1);
12046 else if (PL_regkind[(U8)op] == CURLY && op != CURLYX) {
12048 DUMPUNTIL(NEXTOPER(node) + EXTRA_STEP_2ARGS, next);
12050 else if ( op == PLUS || op == STAR) {
12051 DUMPUNTIL(NEXTOPER(node), NEXTOPER(node) + 1);
12053 else if (PL_regkind[(U8)op] == ANYOF) {
12054 /* arglen 1 + class block */
12055 node += 1 + ((ANYOF_FLAGS(node) & ANYOF_CLASS)
12056 ? ANYOF_CLASS_SKIP : ANYOF_SKIP);
12057 node = NEXTOPER(node);
12059 else if (PL_regkind[(U8)op] == EXACT) {
12060 /* Literal string, where present. */
12061 node += NODE_SZ_STR(node) - 1;
12062 node = NEXTOPER(node);
12065 node = NEXTOPER(node);
12066 node += regarglen[(U8)op];
12068 if (op == CURLYX || op == OPEN)
12072 #ifdef DEBUG_DUMPUNTIL
12073 PerlIO_printf(Perl_debug_log, "--- %d\n", (int)indent);
12078 #endif /* DEBUGGING */
12082 * c-indentation-style: bsd
12083 * c-basic-offset: 4
12084 * indent-tabs-mode: t
12087 * ex: set ts=8 sts=4 sw=4 noet:
https://perl5.git.perl.org / perl5.git / blob