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
84 EXTERN_C const struct regexp_engine my_reg_engine;
89 #include "dquote_inline.h"
90 #include "invlist_inline.h"
91 #include "unicode_constants.h"
93 #define HAS_NONLATIN1_FOLD_CLOSURE(i) \
94 _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(i)
95 #define HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE(i) \
96 _HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(i)
97 #define IS_NON_FINAL_FOLD(c) _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c)
98 #define IS_IN_SOME_FOLD_L1(c) _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c)
101 #define STATIC static
104 /* this is a chain of data about sub patterns we are processing that
105 need to be handled separately/specially in study_chunk. Its so
106 we can simulate recursion without losing state. */
108 typedef struct scan_frame {
109 regnode *last_regnode; /* last node to process in this frame */
110 regnode *next_regnode; /* next node to process when last is reached */
111 U32 prev_recursed_depth;
112 I32 stopparen; /* what stopparen do we use */
113 U32 is_top_frame; /* what flags do we use? */
115 struct scan_frame *this_prev_frame; /* this previous frame */
116 struct scan_frame *prev_frame; /* previous frame */
117 struct scan_frame *next_frame; /* next frame */
120 /* Certain characters are output as a sequence with the first being a
122 #define isBACKSLASHED_PUNCT(c) \
123 ((c) == '-' || (c) == ']' || (c) == '\\' || (c) == '^')
126 struct RExC_state_t {
127 U32 flags; /* RXf_* are we folding, multilining? */
128 U32 pm_flags; /* PMf_* stuff from the calling PMOP */
129 char *precomp; /* uncompiled string. */
130 char *precomp_end; /* pointer to end of uncompiled string. */
131 REGEXP *rx_sv; /* The SV that is the regexp. */
132 regexp *rx; /* perl core regexp structure */
133 regexp_internal *rxi; /* internal data for regexp object
135 char *start; /* Start of input for compile */
136 char *end; /* End of input for compile */
137 char *parse; /* Input-scan pointer. */
138 char *adjusted_start; /* 'start', adjusted. See code use */
139 STRLEN precomp_adj; /* an offset beyond precomp. See code use */
140 SSize_t whilem_seen; /* number of WHILEM in this expr */
141 regnode *emit_start; /* Start of emitted-code area */
142 regnode *emit_bound; /* First regnode outside of the
144 regnode *emit; /* Code-emit pointer; if = &emit_dummy,
145 implies compiling, so don't emit */
146 regnode_ssc emit_dummy; /* placeholder for emit to point to;
147 large enough for the largest
148 non-EXACTish node, so can use it as
150 I32 naughty; /* How bad is this pattern? */
151 I32 sawback; /* Did we see \1, ...? */
153 SSize_t size; /* Code size. */
154 I32 npar; /* Capture buffer count, (OPEN) plus
155 one. ("par" 0 is the whole
157 I32 nestroot; /* root parens we are in - used by
161 regnode **open_parens; /* pointers to open parens */
162 regnode **close_parens; /* pointers to close parens */
163 regnode *end_op; /* END node in program */
164 I32 utf8; /* whether the pattern is utf8 or not */
165 I32 orig_utf8; /* whether the pattern was originally in utf8 */
166 /* XXX use this for future optimisation of case
167 * where pattern must be upgraded to utf8. */
168 I32 uni_semantics; /* If a d charset modifier should use unicode
169 rules, even if the pattern is not in
171 HV *paren_names; /* Paren names */
173 regnode **recurse; /* Recurse regops */
174 I32 recurse_count; /* Number of recurse regops we have generated */
175 U8 *study_chunk_recursed; /* bitmap of which subs we have moved
177 U32 study_chunk_recursed_bytes; /* bytes in bitmap */
180 I32 override_recoding;
182 I32 recode_x_to_native;
184 I32 in_multi_char_class;
185 struct reg_code_blocks *code_blocks;/* positions of literal (?{})
187 int code_index; /* next code_blocks[] slot */
188 SSize_t maxlen; /* mininum possible number of chars in string to match */
189 scan_frame *frame_head;
190 scan_frame *frame_last;
193 #ifdef ADD_TO_REGEXEC
194 char *starttry; /* -Dr: where regtry was called. */
195 #define RExC_starttry (pRExC_state->starttry)
197 SV *runtime_code_qr; /* qr with the runtime code blocks */
199 const char *lastparse;
201 AV *paren_name_list; /* idx -> name */
202 U32 study_chunk_recursed_count;
205 #define RExC_lastparse (pRExC_state->lastparse)
206 #define RExC_lastnum (pRExC_state->lastnum)
207 #define RExC_paren_name_list (pRExC_state->paren_name_list)
208 #define RExC_study_chunk_recursed_count (pRExC_state->study_chunk_recursed_count)
209 #define RExC_mysv (pRExC_state->mysv1)
210 #define RExC_mysv1 (pRExC_state->mysv1)
211 #define RExC_mysv2 (pRExC_state->mysv2)
214 bool seen_unfolded_sharp_s;
219 #define RExC_flags (pRExC_state->flags)
220 #define RExC_pm_flags (pRExC_state->pm_flags)
221 #define RExC_precomp (pRExC_state->precomp)
222 #define RExC_precomp_adj (pRExC_state->precomp_adj)
223 #define RExC_adjusted_start (pRExC_state->adjusted_start)
224 #define RExC_precomp_end (pRExC_state->precomp_end)
225 #define RExC_rx_sv (pRExC_state->rx_sv)
226 #define RExC_rx (pRExC_state->rx)
227 #define RExC_rxi (pRExC_state->rxi)
228 #define RExC_start (pRExC_state->start)
229 #define RExC_end (pRExC_state->end)
230 #define RExC_parse (pRExC_state->parse)
231 #define RExC_whilem_seen (pRExC_state->whilem_seen)
233 /* Set during the sizing pass when there is a LATIN SMALL LETTER SHARP S in any
234 * EXACTF node, hence was parsed under /di rules. If later in the parse,
235 * something forces the pattern into using /ui rules, the sharp s should be
236 * folded into the sequence 'ss', which takes up more space than previously
237 * calculated. This means that the sizing pass needs to be restarted. (The
238 * node also becomes an EXACTFU_SS.) For all other characters, an EXACTF node
239 * that gets converted to /ui (and EXACTFU) occupies the same amount of space,
240 * so there is no need to resize [perl #125990]. */
241 #define RExC_seen_unfolded_sharp_s (pRExC_state->seen_unfolded_sharp_s)
243 #ifdef RE_TRACK_PATTERN_OFFSETS
244 #define RExC_offsets (pRExC_state->rxi->u.offsets) /* I am not like the
247 #define RExC_emit (pRExC_state->emit)
248 #define RExC_emit_dummy (pRExC_state->emit_dummy)
249 #define RExC_emit_start (pRExC_state->emit_start)
250 #define RExC_emit_bound (pRExC_state->emit_bound)
251 #define RExC_sawback (pRExC_state->sawback)
252 #define RExC_seen (pRExC_state->seen)
253 #define RExC_size (pRExC_state->size)
254 #define RExC_maxlen (pRExC_state->maxlen)
255 #define RExC_npar (pRExC_state->npar)
256 #define RExC_nestroot (pRExC_state->nestroot)
257 #define RExC_extralen (pRExC_state->extralen)
258 #define RExC_seen_zerolen (pRExC_state->seen_zerolen)
259 #define RExC_utf8 (pRExC_state->utf8)
260 #define RExC_uni_semantics (pRExC_state->uni_semantics)
261 #define RExC_orig_utf8 (pRExC_state->orig_utf8)
262 #define RExC_open_parens (pRExC_state->open_parens)
263 #define RExC_close_parens (pRExC_state->close_parens)
264 #define RExC_end_op (pRExC_state->end_op)
265 #define RExC_paren_names (pRExC_state->paren_names)
266 #define RExC_recurse (pRExC_state->recurse)
267 #define RExC_recurse_count (pRExC_state->recurse_count)
268 #define RExC_study_chunk_recursed (pRExC_state->study_chunk_recursed)
269 #define RExC_study_chunk_recursed_bytes \
270 (pRExC_state->study_chunk_recursed_bytes)
271 #define RExC_in_lookbehind (pRExC_state->in_lookbehind)
272 #define RExC_contains_locale (pRExC_state->contains_locale)
274 # define RExC_recode_x_to_native (pRExC_state->recode_x_to_native)
276 #define RExC_in_multi_char_class (pRExC_state->in_multi_char_class)
277 #define RExC_frame_head (pRExC_state->frame_head)
278 #define RExC_frame_last (pRExC_state->frame_last)
279 #define RExC_frame_count (pRExC_state->frame_count)
280 #define RExC_strict (pRExC_state->strict)
281 #define RExC_study_started (pRExC_state->study_started)
282 #define RExC_warn_text (pRExC_state->warn_text)
284 /* Heuristic check on the complexity of the pattern: if TOO_NAUGHTY, we set
285 * a flag to disable back-off on the fixed/floating substrings - if it's
286 * a high complexity pattern we assume the benefit of avoiding a full match
287 * is worth the cost of checking for the substrings even if they rarely help.
289 #define RExC_naughty (pRExC_state->naughty)
290 #define TOO_NAUGHTY (10)
291 #define MARK_NAUGHTY(add) \
292 if (RExC_naughty < TOO_NAUGHTY) \
293 RExC_naughty += (add)
294 #define MARK_NAUGHTY_EXP(exp, add) \
295 if (RExC_naughty < TOO_NAUGHTY) \
296 RExC_naughty += RExC_naughty / (exp) + (add)
298 #define ISMULT1(c) ((c) == '*' || (c) == '+' || (c) == '?')
299 #define ISMULT2(s) ((*s) == '*' || (*s) == '+' || (*s) == '?' || \
300 ((*s) == '{' && regcurly(s)))
303 * Flags to be passed up and down.
305 #define WORST 0 /* Worst case. */
306 #define HASWIDTH 0x01 /* Known to match non-null strings. */
308 /* Simple enough to be STAR/PLUS operand; in an EXACTish node must be a single
309 * character. (There needs to be a case: in the switch statement in regexec.c
310 * for any node marked SIMPLE.) Note that this is not the same thing as
313 #define SPSTART 0x04 /* Starts with * or + */
314 #define POSTPONED 0x08 /* (?1),(?&name), (??{...}) or similar */
315 #define TRYAGAIN 0x10 /* Weeded out a declaration. */
316 #define RESTART_PASS1 0x20 /* Need to restart sizing pass */
317 #define NEED_UTF8 0x40 /* In conjunction with RESTART_PASS1, need to
318 calcuate sizes as UTF-8 */
320 #define REG_NODE_NUM(x) ((x) ? (int)((x)-RExC_emit_start) : -1)
322 /* whether trie related optimizations are enabled */
323 #if PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION
324 #define TRIE_STUDY_OPT
325 #define FULL_TRIE_STUDY
331 #define PBYTE(u8str,paren) ((U8*)(u8str))[(paren) >> 3]
332 #define PBITVAL(paren) (1 << ((paren) & 7))
333 #define PAREN_TEST(u8str,paren) ( PBYTE(u8str,paren) & PBITVAL(paren))
334 #define PAREN_SET(u8str,paren) PBYTE(u8str,paren) |= PBITVAL(paren)
335 #define PAREN_UNSET(u8str,paren) PBYTE(u8str,paren) &= (~PBITVAL(paren))
337 #define REQUIRE_UTF8(flagp) STMT_START { \
340 *flagp = RESTART_PASS1|NEED_UTF8; \
345 /* Change from /d into /u rules, and restart the parse if we've already seen
346 * something whose size would increase as a result, by setting *flagp and
347 * returning 'restart_retval'. RExC_uni_semantics is a flag that indicates
348 * we've change to /u during the parse. */
349 #define REQUIRE_UNI_RULES(flagp, restart_retval) \
351 if (DEPENDS_SEMANTICS) { \
353 set_regex_charset(&RExC_flags, REGEX_UNICODE_CHARSET); \
354 RExC_uni_semantics = 1; \
355 if (RExC_seen_unfolded_sharp_s) { \
356 *flagp |= RESTART_PASS1; \
357 return restart_retval; \
362 /* This converts the named class defined in regcomp.h to its equivalent class
363 * number defined in handy.h. */
364 #define namedclass_to_classnum(class) ((int) ((class) / 2))
365 #define classnum_to_namedclass(classnum) ((classnum) * 2)
367 #define _invlist_union_complement_2nd(a, b, output) \
368 _invlist_union_maybe_complement_2nd(a, b, TRUE, output)
369 #define _invlist_intersection_complement_2nd(a, b, output) \
370 _invlist_intersection_maybe_complement_2nd(a, b, TRUE, output)
372 /* About scan_data_t.
374 During optimisation we recurse through the regexp program performing
375 various inplace (keyhole style) optimisations. In addition study_chunk
376 and scan_commit populate this data structure with information about
377 what strings MUST appear in the pattern. We look for the longest
378 string that must appear at a fixed location, and we look for the
379 longest string that may appear at a floating location. So for instance
384 Both 'FOO' and 'A' are fixed strings. Both 'B' and 'BAR' are floating
385 strings (because they follow a .* construct). study_chunk will identify
386 both FOO and BAR as being the longest fixed and floating strings respectively.
388 The strings can be composites, for instance
392 will result in a composite fixed substring 'foo'.
394 For each string some basic information is maintained:
396 - offset or min_offset
397 This is the position the string must appear at, or not before.
398 It also implicitly (when combined with minlenp) tells us how many
399 characters must match before the string we are searching for.
400 Likewise when combined with minlenp and the length of the string it
401 tells us how many characters must appear after the string we have
405 Only used for floating strings. This is the rightmost point that
406 the string can appear at. If set to SSize_t_MAX it indicates that the
407 string can occur infinitely far to the right.
410 A pointer to the minimum number of characters of the pattern that the
411 string was found inside. This is important as in the case of positive
412 lookahead or positive lookbehind we can have multiple patterns
417 The minimum length of the pattern overall is 3, the minimum length
418 of the lookahead part is 3, but the minimum length of the part that
419 will actually match is 1. So 'FOO's minimum length is 3, but the
420 minimum length for the F is 1. This is important as the minimum length
421 is used to determine offsets in front of and behind the string being
422 looked for. Since strings can be composites this is the length of the
423 pattern at the time it was committed with a scan_commit. Note that
424 the length is calculated by study_chunk, so that the minimum lengths
425 are not known until the full pattern has been compiled, thus the
426 pointer to the value.
430 In the case of lookbehind the string being searched for can be
431 offset past the start point of the final matching string.
432 If this value was just blithely removed from the min_offset it would
433 invalidate some of the calculations for how many chars must match
434 before or after (as they are derived from min_offset and minlen and
435 the length of the string being searched for).
436 When the final pattern is compiled and the data is moved from the
437 scan_data_t structure into the regexp structure the information
438 about lookbehind is factored in, with the information that would
439 have been lost precalculated in the end_shift field for the
442 The fields pos_min and pos_delta are used to store the minimum offset
443 and the delta to the maximum offset at the current point in the pattern.
447 typedef struct scan_data_t {
448 /*I32 len_min; unused */
449 /*I32 len_delta; unused */
453 SSize_t last_end; /* min value, <0 unless valid. */
454 SSize_t last_start_min;
455 SSize_t last_start_max;
456 SV **longest; /* Either &l_fixed, or &l_float. */
457 SV *longest_fixed; /* longest fixed string found in pattern */
458 SSize_t offset_fixed; /* offset where it starts */
459 SSize_t *minlen_fixed; /* pointer to the minlen relevant to the string */
460 I32 lookbehind_fixed; /* is the position of the string modfied by LB */
461 SV *longest_float; /* longest floating string found in pattern */
462 SSize_t offset_float_min; /* earliest point in string it can appear */
463 SSize_t offset_float_max; /* latest point in string it can appear */
464 SSize_t *minlen_float; /* pointer to the minlen relevant to the string */
465 SSize_t lookbehind_float; /* is the pos of the string modified by LB */
468 SSize_t *last_closep;
469 regnode_ssc *start_class;
473 * Forward declarations for pregcomp()'s friends.
476 static const scan_data_t zero_scan_data =
477 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ,0};
479 #define SF_BEFORE_EOL (SF_BEFORE_SEOL|SF_BEFORE_MEOL)
480 #define SF_BEFORE_SEOL 0x0001
481 #define SF_BEFORE_MEOL 0x0002
482 #define SF_FIX_BEFORE_EOL (SF_FIX_BEFORE_SEOL|SF_FIX_BEFORE_MEOL)
483 #define SF_FL_BEFORE_EOL (SF_FL_BEFORE_SEOL|SF_FL_BEFORE_MEOL)
485 #define SF_FIX_SHIFT_EOL (+2)
486 #define SF_FL_SHIFT_EOL (+4)
488 #define SF_FIX_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FIX_SHIFT_EOL)
489 #define SF_FIX_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FIX_SHIFT_EOL)
491 #define SF_FL_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FL_SHIFT_EOL)
492 #define SF_FL_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FL_SHIFT_EOL) /* 0x20 */
493 #define SF_IS_INF 0x0040
494 #define SF_HAS_PAR 0x0080
495 #define SF_IN_PAR 0x0100
496 #define SF_HAS_EVAL 0x0200
499 /* SCF_DO_SUBSTR is the flag that tells the regexp analyzer to track the
500 * longest substring in the pattern. When it is not set the optimiser keeps
501 * track of position, but does not keep track of the actual strings seen,
503 * So for instance /foo/ will be parsed with SCF_DO_SUBSTR being true, but
506 * Similarly, /foo.*(blah|erm|huh).*fnorble/ will have "foo" and "fnorble"
507 * parsed with SCF_DO_SUBSTR on, but while processing the (...) it will be
508 * turned off because of the alternation (BRANCH). */
509 #define SCF_DO_SUBSTR 0x0400
511 #define SCF_DO_STCLASS_AND 0x0800
512 #define SCF_DO_STCLASS_OR 0x1000
513 #define SCF_DO_STCLASS (SCF_DO_STCLASS_AND|SCF_DO_STCLASS_OR)
514 #define SCF_WHILEM_VISITED_POS 0x2000
516 #define SCF_TRIE_RESTUDY 0x4000 /* Do restudy? */
517 #define SCF_SEEN_ACCEPT 0x8000
518 #define SCF_TRIE_DOING_RESTUDY 0x10000
519 #define SCF_IN_DEFINE 0x20000
524 #define UTF cBOOL(RExC_utf8)
526 /* The enums for all these are ordered so things work out correctly */
527 #define LOC (get_regex_charset(RExC_flags) == REGEX_LOCALE_CHARSET)
528 #define DEPENDS_SEMANTICS (get_regex_charset(RExC_flags) \
529 == REGEX_DEPENDS_CHARSET)
530 #define UNI_SEMANTICS (get_regex_charset(RExC_flags) == REGEX_UNICODE_CHARSET)
531 #define AT_LEAST_UNI_SEMANTICS (get_regex_charset(RExC_flags) \
532 >= REGEX_UNICODE_CHARSET)
533 #define ASCII_RESTRICTED (get_regex_charset(RExC_flags) \
534 == REGEX_ASCII_RESTRICTED_CHARSET)
535 #define AT_LEAST_ASCII_RESTRICTED (get_regex_charset(RExC_flags) \
536 >= REGEX_ASCII_RESTRICTED_CHARSET)
537 #define ASCII_FOLD_RESTRICTED (get_regex_charset(RExC_flags) \
538 == REGEX_ASCII_MORE_RESTRICTED_CHARSET)
540 #define FOLD cBOOL(RExC_flags & RXf_PMf_FOLD)
542 /* For programs that want to be strictly Unicode compatible by dying if any
543 * attempt is made to match a non-Unicode code point against a Unicode
545 #define ALWAYS_WARN_SUPER ckDEAD(packWARN(WARN_NON_UNICODE))
547 #define OOB_NAMEDCLASS -1
549 /* There is no code point that is out-of-bounds, so this is problematic. But
550 * its only current use is to initialize a variable that is always set before
552 #define OOB_UNICODE 0xDEADBEEF
554 #define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv))
557 /* length of regex to show in messages that don't mark a position within */
558 #define RegexLengthToShowInErrorMessages 127
561 * If MARKER[12] are adjusted, be sure to adjust the constants at the top
562 * of t/op/regmesg.t, the tests in t/op/re_tests, and those in
563 * op/pragma/warn/regcomp.
565 #define MARKER1 "<-- HERE" /* marker as it appears in the description */
566 #define MARKER2 " <-- HERE " /* marker as it appears within the regex */
568 #define REPORT_LOCATION " in regex; marked by " MARKER1 \
569 " in m/%" UTF8f MARKER2 "%" UTF8f "/"
571 /* The code in this file in places uses one level of recursion with parsing
572 * rebased to an alternate string constructed by us in memory. This can take
573 * the form of something that is completely different from the input, or
574 * something that uses the input as part of the alternate. In the first case,
575 * there should be no possibility of an error, as we are in complete control of
576 * the alternate string. But in the second case we don't control the input
577 * portion, so there may be errors in that. Here's an example:
579 * is handled specially because \x{df} folds to a sequence of more than one
580 * character, 'ss'. What is done is to create and parse an alternate string,
581 * which looks like this:
582 * /(?:\x{DF}|[abc\x{DF}def])/ui
583 * where it uses the input unchanged in the middle of something it constructs,
584 * which is a branch for the DF outside the character class, and clustering
585 * parens around the whole thing. (It knows enough to skip the DF inside the
586 * class while in this substitute parse.) 'abc' and 'def' may have errors that
587 * need to be reported. The general situation looks like this:
590 * Input: ----------------------------------------------------
591 * Constructed: ---------------------------------------------------
594 * The input string sI..eI is the input pattern. The string sC..EC is the
595 * constructed substitute parse string. The portions sC..tC and eC..EC are
596 * constructed by us. The portion tC..eC is an exact duplicate of the input
597 * pattern tI..eI. In the diagram, these are vertically aligned. Suppose that
598 * while parsing, we find an error at xC. We want to display a message showing
599 * the real input string. Thus we need to find the point xI in it which
600 * corresponds to xC. xC >= tC, since the portion of the string sC..tC has
601 * been constructed by us, and so shouldn't have errors. We get:
603 * xI = sI + (tI - sI) + (xC - tC)
605 * and, the offset into sI is:
607 * (xI - sI) = (tI - sI) + (xC - tC)
609 * When the substitute is constructed, we save (tI -sI) as RExC_precomp_adj,
610 * and we save tC as RExC_adjusted_start.
612 * During normal processing of the input pattern, everything points to that,
613 * with RExC_precomp_adj set to 0, and RExC_adjusted_start set to sI.
616 #define tI_sI RExC_precomp_adj
617 #define tC RExC_adjusted_start
618 #define sC RExC_precomp
619 #define xI_offset(xC) ((IV) (tI_sI + (xC - tC)))
620 #define xI(xC) (sC + xI_offset(xC))
621 #define eC RExC_precomp_end
623 #define REPORT_LOCATION_ARGS(xC) \
625 (xI(xC) > eC) /* Don't run off end */ \
626 ? eC - sC /* Length before the <--HERE */ \
628 sC), /* The input pattern printed up to the <--HERE */ \
630 (xI(xC) > eC) ? 0 : eC - xI(xC), /* Length after <--HERE */ \
631 (xI(xC) > eC) ? eC : xI(xC)) /* pattern after <--HERE */
633 /* Used to point after bad bytes for an error message, but avoid skipping
634 * past a nul byte. */
635 #define SKIP_IF_CHAR(s) (!*(s) ? 0 : UTF ? UTF8SKIP(s) : 1)
638 * Calls SAVEDESTRUCTOR_X if needed, then calls Perl_croak with the given
639 * arg. Show regex, up to a maximum length. If it's too long, chop and add
642 #define _FAIL(code) STMT_START { \
643 const char *ellipses = ""; \
644 IV len = RExC_precomp_end - RExC_precomp; \
647 SAVEFREESV(RExC_rx_sv); \
648 if (len > RegexLengthToShowInErrorMessages) { \
649 /* chop 10 shorter than the max, to ensure meaning of "..." */ \
650 len = RegexLengthToShowInErrorMessages - 10; \
656 #define FAIL(msg) _FAIL( \
657 Perl_croak(aTHX_ "%s in regex m/%" UTF8f "%s/", \
658 msg, UTF8fARG(UTF, len, RExC_precomp), ellipses))
660 #define FAIL2(msg,arg) _FAIL( \
661 Perl_croak(aTHX_ msg " in regex m/%" UTF8f "%s/", \
662 arg, UTF8fARG(UTF, len, RExC_precomp), ellipses))
665 * Simple_vFAIL -- like FAIL, but marks the current location in the scan
667 #define Simple_vFAIL(m) STMT_START { \
668 Perl_croak(aTHX_ "%s" REPORT_LOCATION, \
669 m, REPORT_LOCATION_ARGS(RExC_parse)); \
673 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL()
675 #define vFAIL(m) STMT_START { \
677 SAVEFREESV(RExC_rx_sv); \
682 * Like Simple_vFAIL(), but accepts two arguments.
684 #define Simple_vFAIL2(m,a1) STMT_START { \
685 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, \
686 REPORT_LOCATION_ARGS(RExC_parse)); \
690 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL2().
692 #define vFAIL2(m,a1) STMT_START { \
694 SAVEFREESV(RExC_rx_sv); \
695 Simple_vFAIL2(m, a1); \
700 * Like Simple_vFAIL(), but accepts three arguments.
702 #define Simple_vFAIL3(m, a1, a2) STMT_START { \
703 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, a2, \
704 REPORT_LOCATION_ARGS(RExC_parse)); \
708 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL3().
710 #define vFAIL3(m,a1,a2) STMT_START { \
712 SAVEFREESV(RExC_rx_sv); \
713 Simple_vFAIL3(m, a1, a2); \
717 * Like Simple_vFAIL(), but accepts four arguments.
719 #define Simple_vFAIL4(m, a1, a2, a3) STMT_START { \
720 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, a2, a3, \
721 REPORT_LOCATION_ARGS(RExC_parse)); \
724 #define vFAIL4(m,a1,a2,a3) STMT_START { \
726 SAVEFREESV(RExC_rx_sv); \
727 Simple_vFAIL4(m, a1, a2, a3); \
730 /* A specialized version of vFAIL2 that works with UTF8f */
731 #define vFAIL2utf8f(m, a1) STMT_START { \
733 SAVEFREESV(RExC_rx_sv); \
734 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, \
735 REPORT_LOCATION_ARGS(RExC_parse)); \
738 #define vFAIL3utf8f(m, a1, a2) STMT_START { \
740 SAVEFREESV(RExC_rx_sv); \
741 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, a2, \
742 REPORT_LOCATION_ARGS(RExC_parse)); \
745 /* These have asserts in them because of [perl #122671] Many warnings in
746 * regcomp.c can occur twice. If they get output in pass1 and later in that
747 * pass, the pattern has to be converted to UTF-8 and the pass restarted, they
748 * would get output again. So they should be output in pass2, and these
749 * asserts make sure new warnings follow that paradigm. */
751 /* m is not necessarily a "literal string", in this macro */
752 #define reg_warn_non_literal_string(loc, m) STMT_START { \
753 __ASSERT_(PASS2) Perl_warner(aTHX_ packWARN(WARN_REGEXP), \
754 "%s" REPORT_LOCATION, \
755 m, REPORT_LOCATION_ARGS(loc)); \
758 #define ckWARNreg(loc,m) STMT_START { \
759 __ASSERT_(PASS2) Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), \
761 REPORT_LOCATION_ARGS(loc)); \
764 #define vWARN(loc, m) STMT_START { \
765 __ASSERT_(PASS2) Perl_warner(aTHX_ packWARN(WARN_REGEXP), \
767 REPORT_LOCATION_ARGS(loc)); \
770 #define vWARN_dep(loc, m) STMT_START { \
771 __ASSERT_(PASS2) Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), \
773 REPORT_LOCATION_ARGS(loc)); \
776 #define ckWARNdep(loc,m) STMT_START { \
777 __ASSERT_(PASS2) Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
779 REPORT_LOCATION_ARGS(loc)); \
782 #define ckWARNregdep(loc,m) STMT_START { \
783 __ASSERT_(PASS2) Perl_ck_warner_d(aTHX_ packWARN2(WARN_DEPRECATED, \
786 REPORT_LOCATION_ARGS(loc)); \
789 #define ckWARN2reg_d(loc,m, a1) STMT_START { \
790 __ASSERT_(PASS2) Perl_ck_warner_d(aTHX_ packWARN(WARN_REGEXP), \
792 a1, REPORT_LOCATION_ARGS(loc)); \
795 #define ckWARN2reg(loc, m, a1) STMT_START { \
796 __ASSERT_(PASS2) Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), \
798 a1, REPORT_LOCATION_ARGS(loc)); \
801 #define vWARN3(loc, m, a1, a2) STMT_START { \
802 __ASSERT_(PASS2) Perl_warner(aTHX_ packWARN(WARN_REGEXP), \
804 a1, a2, REPORT_LOCATION_ARGS(loc)); \
807 #define ckWARN3reg(loc, m, a1, a2) STMT_START { \
808 __ASSERT_(PASS2) Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), \
811 REPORT_LOCATION_ARGS(loc)); \
814 #define vWARN4(loc, m, a1, a2, a3) STMT_START { \
815 __ASSERT_(PASS2) Perl_warner(aTHX_ packWARN(WARN_REGEXP), \
818 REPORT_LOCATION_ARGS(loc)); \
821 #define ckWARN4reg(loc, m, a1, a2, a3) STMT_START { \
822 __ASSERT_(PASS2) Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), \
825 REPORT_LOCATION_ARGS(loc)); \
828 #define vWARN5(loc, m, a1, a2, a3, a4) STMT_START { \
829 __ASSERT_(PASS2) Perl_warner(aTHX_ packWARN(WARN_REGEXP), \
832 REPORT_LOCATION_ARGS(loc)); \
835 /* Macros for recording node offsets. 20001227 mjd@plover.com
836 * Nodes are numbered 1, 2, 3, 4. Node #n's position is recorded in
837 * element 2*n-1 of the array. Element #2n holds the byte length node #n.
838 * Element 0 holds the number n.
839 * Position is 1 indexed.
841 #ifndef RE_TRACK_PATTERN_OFFSETS
842 #define Set_Node_Offset_To_R(node,byte)
843 #define Set_Node_Offset(node,byte)
844 #define Set_Cur_Node_Offset
845 #define Set_Node_Length_To_R(node,len)
846 #define Set_Node_Length(node,len)
847 #define Set_Node_Cur_Length(node,start)
848 #define Node_Offset(n)
849 #define Node_Length(n)
850 #define Set_Node_Offset_Length(node,offset,len)
851 #define ProgLen(ri) ri->u.proglen
852 #define SetProgLen(ri,x) ri->u.proglen = x
854 #define ProgLen(ri) ri->u.offsets[0]
855 #define SetProgLen(ri,x) ri->u.offsets[0] = x
856 #define Set_Node_Offset_To_R(node,byte) STMT_START { \
858 MJD_OFFSET_DEBUG(("** (%d) offset of node %d is %d.\n", \
859 __LINE__, (int)(node), (int)(byte))); \
861 Perl_croak(aTHX_ "value of node is %d in Offset macro", \
864 RExC_offsets[2*(node)-1] = (byte); \
869 #define Set_Node_Offset(node,byte) \
870 Set_Node_Offset_To_R((node)-RExC_emit_start, (byte)-RExC_start)
871 #define Set_Cur_Node_Offset Set_Node_Offset(RExC_emit, RExC_parse)
873 #define Set_Node_Length_To_R(node,len) STMT_START { \
875 MJD_OFFSET_DEBUG(("** (%d) size of node %d is %d.\n", \
876 __LINE__, (int)(node), (int)(len))); \
878 Perl_croak(aTHX_ "value of node is %d in Length macro", \
881 RExC_offsets[2*(node)] = (len); \
886 #define Set_Node_Length(node,len) \
887 Set_Node_Length_To_R((node)-RExC_emit_start, len)
888 #define Set_Node_Cur_Length(node, start) \
889 Set_Node_Length(node, RExC_parse - start)
891 /* Get offsets and lengths */
892 #define Node_Offset(n) (RExC_offsets[2*((n)-RExC_emit_start)-1])
893 #define Node_Length(n) (RExC_offsets[2*((n)-RExC_emit_start)])
895 #define Set_Node_Offset_Length(node,offset,len) STMT_START { \
896 Set_Node_Offset_To_R((node)-RExC_emit_start, (offset)); \
897 Set_Node_Length_To_R((node)-RExC_emit_start, (len)); \
901 #if PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS
902 #define EXPERIMENTAL_INPLACESCAN
903 #endif /*PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS*/
907 Perl_re_printf(pTHX_ const char *fmt, ...)
911 PerlIO *f= Perl_debug_log;
912 PERL_ARGS_ASSERT_RE_PRINTF;
914 result = PerlIO_vprintf(f, fmt, ap);
920 Perl_re_indentf(pTHX_ const char *fmt, U32 depth, ...)
924 PerlIO *f= Perl_debug_log;
925 PERL_ARGS_ASSERT_RE_INDENTF;
927 PerlIO_printf(f, "%*s", ( (int)depth % 20 ) * 2, "");
928 result = PerlIO_vprintf(f, fmt, ap);
932 #endif /* DEBUGGING */
934 #define DEBUG_RExC_seen() \
935 DEBUG_OPTIMISE_MORE_r({ \
936 Perl_re_printf( aTHX_ "RExC_seen: "); \
938 if (RExC_seen & REG_ZERO_LEN_SEEN) \
939 Perl_re_printf( aTHX_ "REG_ZERO_LEN_SEEN "); \
941 if (RExC_seen & REG_LOOKBEHIND_SEEN) \
942 Perl_re_printf( aTHX_ "REG_LOOKBEHIND_SEEN "); \
944 if (RExC_seen & REG_GPOS_SEEN) \
945 Perl_re_printf( aTHX_ "REG_GPOS_SEEN "); \
947 if (RExC_seen & REG_RECURSE_SEEN) \
948 Perl_re_printf( aTHX_ "REG_RECURSE_SEEN "); \
950 if (RExC_seen & REG_TOP_LEVEL_BRANCHES_SEEN) \
951 Perl_re_printf( aTHX_ "REG_TOP_LEVEL_BRANCHES_SEEN "); \
953 if (RExC_seen & REG_VERBARG_SEEN) \
954 Perl_re_printf( aTHX_ "REG_VERBARG_SEEN "); \
956 if (RExC_seen & REG_CUTGROUP_SEEN) \
957 Perl_re_printf( aTHX_ "REG_CUTGROUP_SEEN "); \
959 if (RExC_seen & REG_RUN_ON_COMMENT_SEEN) \
960 Perl_re_printf( aTHX_ "REG_RUN_ON_COMMENT_SEEN "); \
962 if (RExC_seen & REG_UNFOLDED_MULTI_SEEN) \
963 Perl_re_printf( aTHX_ "REG_UNFOLDED_MULTI_SEEN "); \
965 if (RExC_seen & REG_UNBOUNDED_QUANTIFIER_SEEN) \
966 Perl_re_printf( aTHX_ "REG_UNBOUNDED_QUANTIFIER_SEEN "); \
968 Perl_re_printf( aTHX_ "\n"); \
971 #define DEBUG_SHOW_STUDY_FLAG(flags,flag) \
972 if ((flags) & flag) Perl_re_printf( aTHX_ "%s ", #flag)
974 #define DEBUG_SHOW_STUDY_FLAGS(flags,open_str,close_str) \
976 Perl_re_printf( aTHX_ "%s", open_str); \
977 DEBUG_SHOW_STUDY_FLAG(flags,SF_FL_BEFORE_SEOL); \
978 DEBUG_SHOW_STUDY_FLAG(flags,SF_FL_BEFORE_MEOL); \
979 DEBUG_SHOW_STUDY_FLAG(flags,SF_IS_INF); \
980 DEBUG_SHOW_STUDY_FLAG(flags,SF_HAS_PAR); \
981 DEBUG_SHOW_STUDY_FLAG(flags,SF_IN_PAR); \
982 DEBUG_SHOW_STUDY_FLAG(flags,SF_HAS_EVAL); \
983 DEBUG_SHOW_STUDY_FLAG(flags,SCF_DO_SUBSTR); \
984 DEBUG_SHOW_STUDY_FLAG(flags,SCF_DO_STCLASS_AND); \
985 DEBUG_SHOW_STUDY_FLAG(flags,SCF_DO_STCLASS_OR); \
986 DEBUG_SHOW_STUDY_FLAG(flags,SCF_DO_STCLASS); \
987 DEBUG_SHOW_STUDY_FLAG(flags,SCF_WHILEM_VISITED_POS); \
988 DEBUG_SHOW_STUDY_FLAG(flags,SCF_TRIE_RESTUDY); \
989 DEBUG_SHOW_STUDY_FLAG(flags,SCF_SEEN_ACCEPT); \
990 DEBUG_SHOW_STUDY_FLAG(flags,SCF_TRIE_DOING_RESTUDY); \
991 DEBUG_SHOW_STUDY_FLAG(flags,SCF_IN_DEFINE); \
992 Perl_re_printf( aTHX_ "%s", close_str); \
996 #define DEBUG_STUDYDATA(str,data,depth) \
997 DEBUG_OPTIMISE_MORE_r(if(data){ \
998 Perl_re_indentf( aTHX_ "" str "Pos:%" IVdf "/%" IVdf \
999 " Flags: 0x%" UVXf, \
1001 (IV)((data)->pos_min), \
1002 (IV)((data)->pos_delta), \
1003 (UV)((data)->flags) \
1005 DEBUG_SHOW_STUDY_FLAGS((data)->flags," [ ","]"); \
1006 Perl_re_printf( aTHX_ \
1007 " Whilem_c: %" IVdf " Lcp: %" IVdf " %s", \
1008 (IV)((data)->whilem_c), \
1009 (IV)((data)->last_closep ? *((data)->last_closep) : -1), \
1010 is_inf ? "INF " : "" \
1012 if ((data)->last_found) \
1013 Perl_re_printf( aTHX_ \
1014 "Last:'%s' %" IVdf ":%" IVdf "/%" IVdf \
1015 " %sFixed:'%s' @ %" IVdf \
1016 " %sFloat: '%s' @ %" IVdf "/%" IVdf, \
1017 SvPVX_const((data)->last_found), \
1018 (IV)((data)->last_end), \
1019 (IV)((data)->last_start_min), \
1020 (IV)((data)->last_start_max), \
1021 ((data)->longest && \
1022 (data)->longest==&((data)->longest_fixed)) ? "*" : "", \
1023 SvPVX_const((data)->longest_fixed), \
1024 (IV)((data)->offset_fixed), \
1025 ((data)->longest && \
1026 (data)->longest==&((data)->longest_float)) ? "*" : "", \
1027 SvPVX_const((data)->longest_float), \
1028 (IV)((data)->offset_float_min), \
1029 (IV)((data)->offset_float_max) \
1031 Perl_re_printf( aTHX_ "\n"); \
1035 /* =========================================================
1036 * BEGIN edit_distance stuff.
1038 * This calculates how many single character changes of any type are needed to
1039 * transform a string into another one. It is taken from version 3.1 of
1041 * https://metacpan.org/pod/Text::Levenshtein::Damerau::XS
1044 /* Our unsorted dictionary linked list. */
1045 /* Note we use UVs, not chars. */
1050 struct dictionary* next;
1052 typedef struct dictionary item;
1055 PERL_STATIC_INLINE item*
1056 push(UV key,item* curr)
1059 Newxz(head, 1, item);
1067 PERL_STATIC_INLINE item*
1068 find(item* head, UV key)
1070 item* iterator = head;
1072 if (iterator->key == key){
1075 iterator = iterator->next;
1081 PERL_STATIC_INLINE item*
1082 uniquePush(item* head,UV key)
1084 item* iterator = head;
1087 if (iterator->key == key) {
1090 iterator = iterator->next;
1093 return push(key,head);
1096 PERL_STATIC_INLINE void
1097 dict_free(item* head)
1099 item* iterator = head;
1102 item* temp = iterator;
1103 iterator = iterator->next;
1110 /* End of Dictionary Stuff */
1112 /* All calculations/work are done here */
1114 S_edit_distance(const UV* src,
1116 const STRLEN x, /* length of src[] */
1117 const STRLEN y, /* length of tgt[] */
1118 const SSize_t maxDistance
1122 UV swapCount,swapScore,targetCharCount,i,j;
1124 UV score_ceil = x + y;
1126 PERL_ARGS_ASSERT_EDIT_DISTANCE;
1128 /* intialize matrix start values */
1129 Newxz(scores, ( (x + 2) * (y + 2)), UV);
1130 scores[0] = score_ceil;
1131 scores[1 * (y + 2) + 0] = score_ceil;
1132 scores[0 * (y + 2) + 1] = score_ceil;
1133 scores[1 * (y + 2) + 1] = 0;
1134 head = uniquePush(uniquePush(head,src[0]),tgt[0]);
1139 for (i=1;i<=x;i++) {
1141 head = uniquePush(head,src[i]);
1142 scores[(i+1) * (y + 2) + 1] = i;
1143 scores[(i+1) * (y + 2) + 0] = score_ceil;
1146 for (j=1;j<=y;j++) {
1149 head = uniquePush(head,tgt[j]);
1150 scores[1 * (y + 2) + (j + 1)] = j;
1151 scores[0 * (y + 2) + (j + 1)] = score_ceil;
1154 targetCharCount = find(head,tgt[j-1])->value;
1155 swapScore = scores[targetCharCount * (y + 2) + swapCount] + i - targetCharCount - 1 + j - swapCount;
1157 if (src[i-1] != tgt[j-1]){
1158 scores[(i+1) * (y + 2) + (j + 1)] = MIN(swapScore,(MIN(scores[i * (y + 2) + j], MIN(scores[(i+1) * (y + 2) + j], scores[i * (y + 2) + (j + 1)])) + 1));
1162 scores[(i+1) * (y + 2) + (j + 1)] = MIN(scores[i * (y + 2) + j], swapScore);
1166 find(head,src[i-1])->value = i;
1170 IV score = scores[(x+1) * (y + 2) + (y + 1)];
1173 return (maxDistance != 0 && maxDistance < score)?(-1):score;
1177 /* END of edit_distance() stuff
1178 * ========================================================= */
1180 /* is c a control character for which we have a mnemonic? */
1181 #define isMNEMONIC_CNTRL(c) _IS_MNEMONIC_CNTRL_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c)
1184 S_cntrl_to_mnemonic(const U8 c)
1186 /* Returns the mnemonic string that represents character 'c', if one
1187 * exists; NULL otherwise. The only ones that exist for the purposes of
1188 * this routine are a few control characters */
1191 case '\a': return "\\a";
1192 case '\b': return "\\b";
1193 case ESC_NATIVE: return "\\e";
1194 case '\f': return "\\f";
1195 case '\n': return "\\n";
1196 case '\r': return "\\r";
1197 case '\t': return "\\t";
1203 /* Mark that we cannot extend a found fixed substring at this point.
1204 Update the longest found anchored substring and the longest found
1205 floating substrings if needed. */
1208 S_scan_commit(pTHX_ const RExC_state_t *pRExC_state, scan_data_t *data,
1209 SSize_t *minlenp, int is_inf)
1211 const STRLEN l = CHR_SVLEN(data->last_found);
1212 const STRLEN old_l = CHR_SVLEN(*data->longest);
1213 GET_RE_DEBUG_FLAGS_DECL;
1215 PERL_ARGS_ASSERT_SCAN_COMMIT;
1217 if ((l >= old_l) && ((l > old_l) || (data->flags & SF_BEFORE_EOL))) {
1218 SvSetMagicSV(*data->longest, data->last_found);
1219 if (*data->longest == data->longest_fixed) {
1220 data->offset_fixed = l ? data->last_start_min : data->pos_min;
1221 if (data->flags & SF_BEFORE_EOL)
1223 |= ((data->flags & SF_BEFORE_EOL) << SF_FIX_SHIFT_EOL);
1225 data->flags &= ~SF_FIX_BEFORE_EOL;
1226 data->minlen_fixed=minlenp;
1227 data->lookbehind_fixed=0;
1229 else { /* *data->longest == data->longest_float */
1230 data->offset_float_min = l ? data->last_start_min : data->pos_min;
1231 data->offset_float_max = (l
1232 ? data->last_start_max
1233 : (data->pos_delta > SSize_t_MAX - data->pos_min
1235 : data->pos_min + data->pos_delta));
1237 || (STRLEN)data->offset_float_max > (STRLEN)SSize_t_MAX)
1238 data->offset_float_max = SSize_t_MAX;
1239 if (data->flags & SF_BEFORE_EOL)
1241 |= ((data->flags & SF_BEFORE_EOL) << SF_FL_SHIFT_EOL);
1243 data->flags &= ~SF_FL_BEFORE_EOL;
1244 data->minlen_float=minlenp;
1245 data->lookbehind_float=0;
1248 SvCUR_set(data->last_found, 0);
1250 SV * const sv = data->last_found;
1251 if (SvUTF8(sv) && SvMAGICAL(sv)) {
1252 MAGIC * const mg = mg_find(sv, PERL_MAGIC_utf8);
1257 data->last_end = -1;
1258 data->flags &= ~SF_BEFORE_EOL;
1259 DEBUG_STUDYDATA("commit: ",data,0);
1262 /* An SSC is just a regnode_charclass_posix with an extra field: the inversion
1263 * list that describes which code points it matches */
1266 S_ssc_anything(pTHX_ regnode_ssc *ssc)
1268 /* Set the SSC 'ssc' to match an empty string or any code point */
1270 PERL_ARGS_ASSERT_SSC_ANYTHING;
1272 assert(is_ANYOF_SYNTHETIC(ssc));
1274 /* mortalize so won't leak */
1275 ssc->invlist = sv_2mortal(_add_range_to_invlist(NULL, 0, UV_MAX));
1276 ANYOF_FLAGS(ssc) |= SSC_MATCHES_EMPTY_STRING; /* Plus matches empty */
1280 S_ssc_is_anything(const regnode_ssc *ssc)
1282 /* Returns TRUE if the SSC 'ssc' can match the empty string and any code
1283 * point; FALSE otherwise. Thus, this is used to see if using 'ssc' buys
1284 * us anything: if the function returns TRUE, 'ssc' hasn't been restricted
1285 * in any way, so there's no point in using it */
1290 PERL_ARGS_ASSERT_SSC_IS_ANYTHING;
1292 assert(is_ANYOF_SYNTHETIC(ssc));
1294 if (! (ANYOF_FLAGS(ssc) & SSC_MATCHES_EMPTY_STRING)) {
1298 /* See if the list consists solely of the range 0 - Infinity */
1299 invlist_iterinit(ssc->invlist);
1300 ret = invlist_iternext(ssc->invlist, &start, &end)
1304 invlist_iterfinish(ssc->invlist);
1310 /* If e.g., both \w and \W are set, matches everything */
1311 if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)) {
1313 for (i = 0; i < ANYOF_POSIXL_MAX; i += 2) {
1314 if (ANYOF_POSIXL_TEST(ssc, i) && ANYOF_POSIXL_TEST(ssc, i+1)) {
1324 S_ssc_init(pTHX_ const RExC_state_t *pRExC_state, regnode_ssc *ssc)
1326 /* Initializes the SSC 'ssc'. This includes setting it to match an empty
1327 * string, any code point, or any posix class under locale */
1329 PERL_ARGS_ASSERT_SSC_INIT;
1331 Zero(ssc, 1, regnode_ssc);
1332 set_ANYOF_SYNTHETIC(ssc);
1333 ARG_SET(ssc, ANYOF_ONLY_HAS_BITMAP);
1336 /* If any portion of the regex is to operate under locale rules that aren't
1337 * fully known at compile time, initialization includes it. The reason
1338 * this isn't done for all regexes is that the optimizer was written under
1339 * the assumption that locale was all-or-nothing. Given the complexity and
1340 * lack of documentation in the optimizer, and that there are inadequate
1341 * test cases for locale, many parts of it may not work properly, it is
1342 * safest to avoid locale unless necessary. */
1343 if (RExC_contains_locale) {
1344 ANYOF_POSIXL_SETALL(ssc);
1347 ANYOF_POSIXL_ZERO(ssc);
1352 S_ssc_is_cp_posixl_init(const RExC_state_t *pRExC_state,
1353 const regnode_ssc *ssc)
1355 /* Returns TRUE if the SSC 'ssc' is in its initial state with regard only
1356 * to the list of code points matched, and locale posix classes; hence does
1357 * not check its flags) */
1362 PERL_ARGS_ASSERT_SSC_IS_CP_POSIXL_INIT;
1364 assert(is_ANYOF_SYNTHETIC(ssc));
1366 invlist_iterinit(ssc->invlist);
1367 ret = invlist_iternext(ssc->invlist, &start, &end)
1371 invlist_iterfinish(ssc->invlist);
1377 if (RExC_contains_locale && ! ANYOF_POSIXL_SSC_TEST_ALL_SET(ssc)) {
1385 S_get_ANYOF_cp_list_for_ssc(pTHX_ const RExC_state_t *pRExC_state,
1386 const regnode_charclass* const node)
1388 /* Returns a mortal inversion list defining which code points are matched
1389 * by 'node', which is of type ANYOF. Handles complementing the result if
1390 * appropriate. If some code points aren't knowable at this time, the
1391 * returned list must, and will, contain every code point that is a
1395 SV* only_utf8_locale_invlist = NULL;
1397 const U32 n = ARG(node);
1398 bool new_node_has_latin1 = FALSE;
1400 PERL_ARGS_ASSERT_GET_ANYOF_CP_LIST_FOR_SSC;
1402 /* Look at the data structure created by S_set_ANYOF_arg() */
1403 if (n != ANYOF_ONLY_HAS_BITMAP) {
1404 SV * const rv = MUTABLE_SV(RExC_rxi->data->data[n]);
1405 AV * const av = MUTABLE_AV(SvRV(rv));
1406 SV **const ary = AvARRAY(av);
1407 assert(RExC_rxi->data->what[n] == 's');
1409 if (ary[1] && ary[1] != &PL_sv_undef) { /* Has compile-time swash */
1410 invlist = sv_2mortal(invlist_clone(_get_swash_invlist(ary[1])));
1412 else if (ary[0] && ary[0] != &PL_sv_undef) {
1414 /* Here, no compile-time swash, and there are things that won't be
1415 * known until runtime -- we have to assume it could be anything */
1416 invlist = sv_2mortal(_new_invlist(1));
1417 return _add_range_to_invlist(invlist, 0, UV_MAX);
1419 else if (ary[3] && ary[3] != &PL_sv_undef) {
1421 /* Here no compile-time swash, and no run-time only data. Use the
1422 * node's inversion list */
1423 invlist = sv_2mortal(invlist_clone(ary[3]));
1426 /* Get the code points valid only under UTF-8 locales */
1427 if ((ANYOF_FLAGS(node) & ANYOFL_FOLD)
1428 && ary[2] && ary[2] != &PL_sv_undef)
1430 only_utf8_locale_invlist = ary[2];
1435 invlist = sv_2mortal(_new_invlist(0));
1438 /* An ANYOF node contains a bitmap for the first NUM_ANYOF_CODE_POINTS
1439 * code points, and an inversion list for the others, but if there are code
1440 * points that should match only conditionally on the target string being
1441 * UTF-8, those are placed in the inversion list, and not the bitmap.
1442 * Since there are circumstances under which they could match, they are
1443 * included in the SSC. But if the ANYOF node is to be inverted, we have
1444 * to exclude them here, so that when we invert below, the end result
1445 * actually does include them. (Think about "\xe0" =~ /[^\xc0]/di;). We
1446 * have to do this here before we add the unconditionally matched code
1448 if (ANYOF_FLAGS(node) & ANYOF_INVERT) {
1449 _invlist_intersection_complement_2nd(invlist,
1454 /* Add in the points from the bit map */
1455 for (i = 0; i < NUM_ANYOF_CODE_POINTS; i++) {
1456 if (ANYOF_BITMAP_TEST(node, i)) {
1457 unsigned int start = i++;
1459 for (; i < NUM_ANYOF_CODE_POINTS && ANYOF_BITMAP_TEST(node, i); ++i) {
1462 invlist = _add_range_to_invlist(invlist, start, i-1);
1463 new_node_has_latin1 = TRUE;
1467 /* If this can match all upper Latin1 code points, have to add them
1468 * as well. But don't add them if inverting, as when that gets done below,
1469 * it would exclude all these characters, including the ones it shouldn't
1470 * that were added just above */
1471 if (! (ANYOF_FLAGS(node) & ANYOF_INVERT) && OP(node) == ANYOFD
1472 && (ANYOF_FLAGS(node) & ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER))
1474 _invlist_union(invlist, PL_UpperLatin1, &invlist);
1477 /* Similarly for these */
1478 if (ANYOF_FLAGS(node) & ANYOF_MATCHES_ALL_ABOVE_BITMAP) {
1479 _invlist_union_complement_2nd(invlist, PL_InBitmap, &invlist);
1482 if (ANYOF_FLAGS(node) & ANYOF_INVERT) {
1483 _invlist_invert(invlist);
1485 else if (new_node_has_latin1 && ANYOF_FLAGS(node) & ANYOFL_FOLD) {
1487 /* Under /li, any 0-255 could fold to any other 0-255, depending on the
1488 * locale. We can skip this if there are no 0-255 at all. */
1489 _invlist_union(invlist, PL_Latin1, &invlist);
1492 /* Similarly add the UTF-8 locale possible matches. These have to be
1493 * deferred until after the non-UTF-8 locale ones are taken care of just
1494 * above, or it leads to wrong results under ANYOF_INVERT */
1495 if (only_utf8_locale_invlist) {
1496 _invlist_union_maybe_complement_2nd(invlist,
1497 only_utf8_locale_invlist,
1498 ANYOF_FLAGS(node) & ANYOF_INVERT,
1505 /* These two functions currently do the exact same thing */
1506 #define ssc_init_zero ssc_init
1508 #define ssc_add_cp(ssc, cp) ssc_add_range((ssc), (cp), (cp))
1509 #define ssc_match_all_cp(ssc) ssc_add_range(ssc, 0, UV_MAX)
1511 /* 'AND' a given class with another one. Can create false positives. 'ssc'
1512 * should not be inverted. 'and_with->flags & ANYOF_MATCHES_POSIXL' should be
1513 * 0 if 'and_with' is a regnode_charclass instead of a regnode_ssc. */
1516 S_ssc_and(pTHX_ const RExC_state_t *pRExC_state, regnode_ssc *ssc,
1517 const regnode_charclass *and_with)
1519 /* Accumulate into SSC 'ssc' its 'AND' with 'and_with', which is either
1520 * another SSC or a regular ANYOF class. Can create false positives. */
1525 PERL_ARGS_ASSERT_SSC_AND;
1527 assert(is_ANYOF_SYNTHETIC(ssc));
1529 /* 'and_with' is used as-is if it too is an SSC; otherwise have to extract
1530 * the code point inversion list and just the relevant flags */
1531 if (is_ANYOF_SYNTHETIC(and_with)) {
1532 anded_cp_list = ((regnode_ssc *)and_with)->invlist;
1533 anded_flags = ANYOF_FLAGS(and_with);
1535 /* XXX This is a kludge around what appears to be deficiencies in the
1536 * optimizer. If we make S_ssc_anything() add in the WARN_SUPER flag,
1537 * there are paths through the optimizer where it doesn't get weeded
1538 * out when it should. And if we don't make some extra provision for
1539 * it like the code just below, it doesn't get added when it should.
1540 * This solution is to add it only when AND'ing, which is here, and
1541 * only when what is being AND'ed is the pristine, original node
1542 * matching anything. Thus it is like adding it to ssc_anything() but
1543 * only when the result is to be AND'ed. Probably the same solution
1544 * could be adopted for the same problem we have with /l matching,
1545 * which is solved differently in S_ssc_init(), and that would lead to
1546 * fewer false positives than that solution has. But if this solution
1547 * creates bugs, the consequences are only that a warning isn't raised
1548 * that should be; while the consequences for having /l bugs is
1549 * incorrect matches */
1550 if (ssc_is_anything((regnode_ssc *)and_with)) {
1551 anded_flags |= ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER;
1555 anded_cp_list = get_ANYOF_cp_list_for_ssc(pRExC_state, and_with);
1556 if (OP(and_with) == ANYOFD) {
1557 anded_flags = ANYOF_FLAGS(and_with) & ANYOF_COMMON_FLAGS;
1560 anded_flags = ANYOF_FLAGS(and_with)
1561 &( ANYOF_COMMON_FLAGS
1562 |ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER
1563 |ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP);
1564 if (ANYOFL_UTF8_LOCALE_REQD(ANYOF_FLAGS(and_with))) {
1566 ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD;
1571 ANYOF_FLAGS(ssc) &= anded_flags;
1573 /* Below, C1 is the list of code points in 'ssc'; P1, its posix classes.
1574 * C2 is the list of code points in 'and-with'; P2, its posix classes.
1575 * 'and_with' may be inverted. When not inverted, we have the situation of
1577 * (C1 | P1) & (C2 | P2)
1578 * = (C1 & (C2 | P2)) | (P1 & (C2 | P2))
1579 * = ((C1 & C2) | (C1 & P2)) | ((P1 & C2) | (P1 & P2))
1580 * <= ((C1 & C2) | P2)) | ( P1 | (P1 & P2))
1581 * <= ((C1 & C2) | P1 | P2)
1582 * Alternatively, the last few steps could be:
1583 * = ((C1 & C2) | (C1 & P2)) | ((P1 & C2) | (P1 & P2))
1584 * <= ((C1 & C2) | C1 ) | ( C2 | (P1 & P2))
1585 * <= (C1 | C2 | (P1 & P2))
1586 * We favor the second approach if either P1 or P2 is non-empty. This is
1587 * because these components are a barrier to doing optimizations, as what
1588 * they match cannot be known until the moment of matching as they are
1589 * dependent on the current locale, 'AND"ing them likely will reduce or
1591 * But we can do better if we know that C1,P1 are in their initial state (a
1592 * frequent occurrence), each matching everything:
1593 * (<everything>) & (C2 | P2) = C2 | P2
1594 * Similarly, if C2,P2 are in their initial state (again a frequent
1595 * occurrence), the result is a no-op
1596 * (C1 | P1) & (<everything>) = C1 | P1
1599 * (C1 | P1) & ~(C2 | P2) = (C1 | P1) & (~C2 & ~P2)
1600 * = (C1 & (~C2 & ~P2)) | (P1 & (~C2 & ~P2))
1601 * <= (C1 & ~C2) | (P1 & ~P2)
1604 if ((ANYOF_FLAGS(and_with) & ANYOF_INVERT)
1605 && ! is_ANYOF_SYNTHETIC(and_with))
1609 ssc_intersection(ssc,
1611 FALSE /* Has already been inverted */
1614 /* If either P1 or P2 is empty, the intersection will be also; can skip
1616 if (! (ANYOF_FLAGS(and_with) & ANYOF_MATCHES_POSIXL)) {
1617 ANYOF_POSIXL_ZERO(ssc);
1619 else if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)) {
1621 /* Note that the Posix class component P from 'and_with' actually
1623 * P = Pa | Pb | ... | Pn
1624 * where each component is one posix class, such as in [\w\s].
1626 * ~P = ~(Pa | Pb | ... | Pn)
1627 * = ~Pa & ~Pb & ... & ~Pn
1628 * <= ~Pa | ~Pb | ... | ~Pn
1629 * The last is something we can easily calculate, but unfortunately
1630 * is likely to have many false positives. We could do better
1631 * in some (but certainly not all) instances if two classes in
1632 * P have known relationships. For example
1633 * :lower: <= :alpha: <= :alnum: <= \w <= :graph: <= :print:
1635 * :lower: & :print: = :lower:
1636 * And similarly for classes that must be disjoint. For example,
1637 * since \s and \w can have no elements in common based on rules in
1638 * the POSIX standard,
1639 * \w & ^\S = nothing
1640 * Unfortunately, some vendor locales do not meet the Posix
1641 * standard, in particular almost everything by Microsoft.
1642 * The loop below just changes e.g., \w into \W and vice versa */
1644 regnode_charclass_posixl temp;
1645 int add = 1; /* To calculate the index of the complement */
1647 ANYOF_POSIXL_ZERO(&temp);
1648 for (i = 0; i < ANYOF_MAX; i++) {
1650 || ! ANYOF_POSIXL_TEST((regnode_charclass_posixl*) and_with, i)
1651 || ! ANYOF_POSIXL_TEST((regnode_charclass_posixl*) and_with, i + 1));
1653 if (ANYOF_POSIXL_TEST((regnode_charclass_posixl*) and_with, i)) {
1654 ANYOF_POSIXL_SET(&temp, i + add);
1656 add = 0 - add; /* 1 goes to -1; -1 goes to 1 */
1658 ANYOF_POSIXL_AND(&temp, ssc);
1660 } /* else ssc already has no posixes */
1661 } /* else: Not inverted. This routine is a no-op if 'and_with' is an SSC
1662 in its initial state */
1663 else if (! is_ANYOF_SYNTHETIC(and_with)
1664 || ! ssc_is_cp_posixl_init(pRExC_state, (regnode_ssc *)and_with))
1666 /* But if 'ssc' is in its initial state, the result is just 'and_with';
1667 * copy it over 'ssc' */
1668 if (ssc_is_cp_posixl_init(pRExC_state, ssc)) {
1669 if (is_ANYOF_SYNTHETIC(and_with)) {
1670 StructCopy(and_with, ssc, regnode_ssc);
1673 ssc->invlist = anded_cp_list;
1674 ANYOF_POSIXL_ZERO(ssc);
1675 if (ANYOF_FLAGS(and_with) & ANYOF_MATCHES_POSIXL) {
1676 ANYOF_POSIXL_OR((regnode_charclass_posixl*) and_with, ssc);
1680 else if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)
1681 || (ANYOF_FLAGS(and_with) & ANYOF_MATCHES_POSIXL))
1683 /* One or the other of P1, P2 is non-empty. */
1684 if (ANYOF_FLAGS(and_with) & ANYOF_MATCHES_POSIXL) {
1685 ANYOF_POSIXL_AND((regnode_charclass_posixl*) and_with, ssc);
1687 ssc_union(ssc, anded_cp_list, FALSE);
1689 else { /* P1 = P2 = empty */
1690 ssc_intersection(ssc, anded_cp_list, FALSE);
1696 S_ssc_or(pTHX_ const RExC_state_t *pRExC_state, regnode_ssc *ssc,
1697 const regnode_charclass *or_with)
1699 /* Accumulate into SSC 'ssc' its 'OR' with 'or_with', which is either
1700 * another SSC or a regular ANYOF class. Can create false positives if
1701 * 'or_with' is to be inverted. */
1706 PERL_ARGS_ASSERT_SSC_OR;
1708 assert(is_ANYOF_SYNTHETIC(ssc));
1710 /* 'or_with' is used as-is if it too is an SSC; otherwise have to extract
1711 * the code point inversion list and just the relevant flags */
1712 if (is_ANYOF_SYNTHETIC(or_with)) {
1713 ored_cp_list = ((regnode_ssc*) or_with)->invlist;
1714 ored_flags = ANYOF_FLAGS(or_with);
1717 ored_cp_list = get_ANYOF_cp_list_for_ssc(pRExC_state, or_with);
1718 ored_flags = ANYOF_FLAGS(or_with) & ANYOF_COMMON_FLAGS;
1719 if (OP(or_with) != ANYOFD) {
1721 |= ANYOF_FLAGS(or_with)
1722 & ( ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER
1723 |ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP);
1724 if (ANYOFL_UTF8_LOCALE_REQD(ANYOF_FLAGS(or_with))) {
1726 ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD;
1731 ANYOF_FLAGS(ssc) |= ored_flags;
1733 /* Below, C1 is the list of code points in 'ssc'; P1, its posix classes.
1734 * C2 is the list of code points in 'or-with'; P2, its posix classes.
1735 * 'or_with' may be inverted. When not inverted, we have the simple
1736 * situation of computing:
1737 * (C1 | P1) | (C2 | P2) = (C1 | C2) | (P1 | P2)
1738 * If P1|P2 yields a situation with both a class and its complement are
1739 * set, like having both \w and \W, this matches all code points, and we
1740 * can delete these from the P component of the ssc going forward. XXX We
1741 * might be able to delete all the P components, but I (khw) am not certain
1742 * about this, and it is better to be safe.
1745 * (C1 | P1) | ~(C2 | P2) = (C1 | P1) | (~C2 & ~P2)
1746 * <= (C1 | P1) | ~C2
1747 * <= (C1 | ~C2) | P1
1748 * (which results in actually simpler code than the non-inverted case)
1751 if ((ANYOF_FLAGS(or_with) & ANYOF_INVERT)
1752 && ! is_ANYOF_SYNTHETIC(or_with))
1754 /* We ignore P2, leaving P1 going forward */
1755 } /* else Not inverted */
1756 else if (ANYOF_FLAGS(or_with) & ANYOF_MATCHES_POSIXL) {
1757 ANYOF_POSIXL_OR((regnode_charclass_posixl*)or_with, ssc);
1758 if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)) {
1760 for (i = 0; i < ANYOF_MAX; i += 2) {
1761 if (ANYOF_POSIXL_TEST(ssc, i) && ANYOF_POSIXL_TEST(ssc, i + 1))
1763 ssc_match_all_cp(ssc);
1764 ANYOF_POSIXL_CLEAR(ssc, i);
1765 ANYOF_POSIXL_CLEAR(ssc, i+1);
1773 FALSE /* Already has been inverted */
1777 PERL_STATIC_INLINE void
1778 S_ssc_union(pTHX_ regnode_ssc *ssc, SV* const invlist, const bool invert2nd)
1780 PERL_ARGS_ASSERT_SSC_UNION;
1782 assert(is_ANYOF_SYNTHETIC(ssc));
1784 _invlist_union_maybe_complement_2nd(ssc->invlist,
1790 PERL_STATIC_INLINE void
1791 S_ssc_intersection(pTHX_ regnode_ssc *ssc,
1793 const bool invert2nd)
1795 PERL_ARGS_ASSERT_SSC_INTERSECTION;
1797 assert(is_ANYOF_SYNTHETIC(ssc));
1799 _invlist_intersection_maybe_complement_2nd(ssc->invlist,
1805 PERL_STATIC_INLINE void
1806 S_ssc_add_range(pTHX_ regnode_ssc *ssc, const UV start, const UV end)
1808 PERL_ARGS_ASSERT_SSC_ADD_RANGE;
1810 assert(is_ANYOF_SYNTHETIC(ssc));
1812 ssc->invlist = _add_range_to_invlist(ssc->invlist, start, end);
1815 PERL_STATIC_INLINE void
1816 S_ssc_cp_and(pTHX_ regnode_ssc *ssc, const UV cp)
1818 /* AND just the single code point 'cp' into the SSC 'ssc' */
1820 SV* cp_list = _new_invlist(2);
1822 PERL_ARGS_ASSERT_SSC_CP_AND;
1824 assert(is_ANYOF_SYNTHETIC(ssc));
1826 cp_list = add_cp_to_invlist(cp_list, cp);
1827 ssc_intersection(ssc, cp_list,
1828 FALSE /* Not inverted */
1830 SvREFCNT_dec_NN(cp_list);
1833 PERL_STATIC_INLINE void
1834 S_ssc_clear_locale(regnode_ssc *ssc)
1836 /* Set the SSC 'ssc' to not match any locale things */
1837 PERL_ARGS_ASSERT_SSC_CLEAR_LOCALE;
1839 assert(is_ANYOF_SYNTHETIC(ssc));
1841 ANYOF_POSIXL_ZERO(ssc);
1842 ANYOF_FLAGS(ssc) &= ~ANYOF_LOCALE_FLAGS;
1845 #define NON_OTHER_COUNT NON_OTHER_COUNT_FOR_USE_ONLY_BY_REGCOMP_DOT_C
1848 S_is_ssc_worth_it(const RExC_state_t * pRExC_state, const regnode_ssc * ssc)
1850 /* The synthetic start class is used to hopefully quickly winnow down
1851 * places where a pattern could start a match in the target string. If it
1852 * doesn't really narrow things down that much, there isn't much point to
1853 * having the overhead of using it. This function uses some very crude
1854 * heuristics to decide if to use the ssc or not.
1856 * It returns TRUE if 'ssc' rules out more than half what it considers to
1857 * be the "likely" possible matches, but of course it doesn't know what the
1858 * actual things being matched are going to be; these are only guesses
1860 * For /l matches, it assumes that the only likely matches are going to be
1861 * in the 0-255 range, uniformly distributed, so half of that is 127
1862 * For /a and /d matches, it assumes that the likely matches will be just
1863 * the ASCII range, so half of that is 63
1864 * For /u and there isn't anything matching above the Latin1 range, it
1865 * assumes that that is the only range likely to be matched, and uses
1866 * half that as the cut-off: 127. If anything matches above Latin1,
1867 * it assumes that all of Unicode could match (uniformly), except for
1868 * non-Unicode code points and things in the General Category "Other"
1869 * (unassigned, private use, surrogates, controls and formats). This
1870 * is a much large number. */
1872 U32 count = 0; /* Running total of number of code points matched by
1874 UV start, end; /* Start and end points of current range in inversion
1876 const U32 max_code_points = (LOC)
1878 : (( ! UNI_SEMANTICS
1879 || invlist_highest(ssc->invlist) < 256)
1882 const U32 max_match = max_code_points / 2;
1884 PERL_ARGS_ASSERT_IS_SSC_WORTH_IT;
1886 invlist_iterinit(ssc->invlist);
1887 while (invlist_iternext(ssc->invlist, &start, &end)) {
1888 if (start >= max_code_points) {
1891 end = MIN(end, max_code_points - 1);
1892 count += end - start + 1;
1893 if (count >= max_match) {
1894 invlist_iterfinish(ssc->invlist);
1904 S_ssc_finalize(pTHX_ RExC_state_t *pRExC_state, regnode_ssc *ssc)
1906 /* The inversion list in the SSC is marked mortal; now we need a more
1907 * permanent copy, which is stored the same way that is done in a regular
1908 * ANYOF node, with the first NUM_ANYOF_CODE_POINTS code points in a bit
1911 SV* invlist = invlist_clone(ssc->invlist);
1913 PERL_ARGS_ASSERT_SSC_FINALIZE;
1915 assert(is_ANYOF_SYNTHETIC(ssc));
1917 /* The code in this file assumes that all but these flags aren't relevant
1918 * to the SSC, except SSC_MATCHES_EMPTY_STRING, which should be cleared
1919 * by the time we reach here */
1920 assert(! (ANYOF_FLAGS(ssc)
1921 & ~( ANYOF_COMMON_FLAGS
1922 |ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER
1923 |ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP)));
1925 populate_ANYOF_from_invlist( (regnode *) ssc, &invlist);
1927 set_ANYOF_arg(pRExC_state, (regnode *) ssc, invlist,
1928 NULL, NULL, NULL, FALSE);
1930 /* Make sure is clone-safe */
1931 ssc->invlist = NULL;
1933 if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)) {
1934 ANYOF_FLAGS(ssc) |= ANYOF_MATCHES_POSIXL;
1937 if (RExC_contains_locale) {
1941 assert(! (ANYOF_FLAGS(ssc) & ANYOF_LOCALE_FLAGS) || RExC_contains_locale);
1944 #define TRIE_LIST_ITEM(state,idx) (trie->states[state].trans.list)[ idx ]
1945 #define TRIE_LIST_CUR(state) ( TRIE_LIST_ITEM( state, 0 ).forid )
1946 #define TRIE_LIST_LEN(state) ( TRIE_LIST_ITEM( state, 0 ).newstate )
1947 #define TRIE_LIST_USED(idx) ( trie->states[state].trans.list \
1948 ? (TRIE_LIST_CUR( idx ) - 1) \
1954 dump_trie(trie,widecharmap,revcharmap)
1955 dump_trie_interim_list(trie,widecharmap,revcharmap,next_alloc)
1956 dump_trie_interim_table(trie,widecharmap,revcharmap,next_alloc)
1958 These routines dump out a trie in a somewhat readable format.
1959 The _interim_ variants are used for debugging the interim
1960 tables that are used to generate the final compressed
1961 representation which is what dump_trie expects.
1963 Part of the reason for their existence is to provide a form
1964 of documentation as to how the different representations function.
1969 Dumps the final compressed table form of the trie to Perl_debug_log.
1970 Used for debugging make_trie().
1974 S_dump_trie(pTHX_ const struct _reg_trie_data *trie, HV *widecharmap,
1975 AV *revcharmap, U32 depth)
1978 SV *sv=sv_newmortal();
1979 int colwidth= widecharmap ? 6 : 4;
1981 GET_RE_DEBUG_FLAGS_DECL;
1983 PERL_ARGS_ASSERT_DUMP_TRIE;
1985 Perl_re_indentf( aTHX_ "Char : %-6s%-6s%-4s ",
1986 depth+1, "Match","Base","Ofs" );
1988 for( state = 0 ; state < trie->uniquecharcount ; state++ ) {
1989 SV ** const tmp = av_fetch( revcharmap, state, 0);
1991 Perl_re_printf( aTHX_ "%*s",
1993 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
1994 PL_colors[0], PL_colors[1],
1995 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
1996 PERL_PV_ESCAPE_FIRSTCHAR
2001 Perl_re_printf( aTHX_ "\n");
2002 Perl_re_indentf( aTHX_ "State|-----------------------", depth+1);
2004 for( state = 0 ; state < trie->uniquecharcount ; state++ )
2005 Perl_re_printf( aTHX_ "%.*s", colwidth, "--------");
2006 Perl_re_printf( aTHX_ "\n");
2008 for( state = 1 ; state < trie->statecount ; state++ ) {
2009 const U32 base = trie->states[ state ].trans.base;
2011 Perl_re_indentf( aTHX_ "#%4" UVXf "|", depth+1, (UV)state);
2013 if ( trie->states[ state ].wordnum ) {
2014 Perl_re_printf( aTHX_ " W%4X", trie->states[ state ].wordnum );
2016 Perl_re_printf( aTHX_ "%6s", "" );
2019 Perl_re_printf( aTHX_ " @%4" UVXf " ", (UV)base );
2024 while( ( base + ofs < trie->uniquecharcount ) ||
2025 ( base + ofs - trie->uniquecharcount < trie->lasttrans
2026 && trie->trans[ base + ofs - trie->uniquecharcount ].check
2030 Perl_re_printf( aTHX_ "+%2" UVXf "[ ", (UV)ofs);
2032 for ( ofs = 0 ; ofs < trie->uniquecharcount ; ofs++ ) {
2033 if ( ( base + ofs >= trie->uniquecharcount )
2034 && ( base + ofs - trie->uniquecharcount
2036 && trie->trans[ base + ofs
2037 - trie->uniquecharcount ].check == state )
2039 Perl_re_printf( aTHX_ "%*" UVXf, colwidth,
2040 (UV)trie->trans[ base + ofs - trie->uniquecharcount ].next
2043 Perl_re_printf( aTHX_ "%*s",colwidth," ." );
2047 Perl_re_printf( aTHX_ "]");
2050 Perl_re_printf( aTHX_ "\n" );
2052 Perl_re_indentf( aTHX_ "word_info N:(prev,len)=",
2054 for (word=1; word <= trie->wordcount; word++) {
2055 Perl_re_printf( aTHX_ " %d:(%d,%d)",
2056 (int)word, (int)(trie->wordinfo[word].prev),
2057 (int)(trie->wordinfo[word].len));
2059 Perl_re_printf( aTHX_ "\n" );
2062 Dumps a fully constructed but uncompressed trie in list form.
2063 List tries normally only are used for construction when the number of
2064 possible chars (trie->uniquecharcount) is very high.
2065 Used for debugging make_trie().
2068 S_dump_trie_interim_list(pTHX_ const struct _reg_trie_data *trie,
2069 HV *widecharmap, AV *revcharmap, U32 next_alloc,
2073 SV *sv=sv_newmortal();
2074 int colwidth= widecharmap ? 6 : 4;
2075 GET_RE_DEBUG_FLAGS_DECL;
2077 PERL_ARGS_ASSERT_DUMP_TRIE_INTERIM_LIST;
2079 /* print out the table precompression. */
2080 Perl_re_indentf( aTHX_ "State :Word | Transition Data\n",
2082 Perl_re_indentf( aTHX_ "%s",
2083 depth+1, "------:-----+-----------------\n" );
2085 for( state=1 ; state < next_alloc ; state ++ ) {
2088 Perl_re_indentf( aTHX_ " %4" UVXf " :",
2089 depth+1, (UV)state );
2090 if ( ! trie->states[ state ].wordnum ) {
2091 Perl_re_printf( aTHX_ "%5s| ","");
2093 Perl_re_printf( aTHX_ "W%4x| ",
2094 trie->states[ state ].wordnum
2097 for( charid = 1 ; charid <= TRIE_LIST_USED( state ) ; charid++ ) {
2098 SV ** const tmp = av_fetch( revcharmap,
2099 TRIE_LIST_ITEM(state,charid).forid, 0);
2101 Perl_re_printf( aTHX_ "%*s:%3X=%4" UVXf " | ",
2103 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp),
2105 PL_colors[0], PL_colors[1],
2106 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)
2107 | PERL_PV_ESCAPE_FIRSTCHAR
2109 TRIE_LIST_ITEM(state,charid).forid,
2110 (UV)TRIE_LIST_ITEM(state,charid).newstate
2113 Perl_re_printf( aTHX_ "\n%*s| ",
2114 (int)((depth * 2) + 14), "");
2117 Perl_re_printf( aTHX_ "\n");
2122 Dumps a fully constructed but uncompressed trie in table form.
2123 This is the normal DFA style state transition table, with a few
2124 twists to facilitate compression later.
2125 Used for debugging make_trie().
2128 S_dump_trie_interim_table(pTHX_ const struct _reg_trie_data *trie,
2129 HV *widecharmap, AV *revcharmap, U32 next_alloc,
2134 SV *sv=sv_newmortal();
2135 int colwidth= widecharmap ? 6 : 4;
2136 GET_RE_DEBUG_FLAGS_DECL;
2138 PERL_ARGS_ASSERT_DUMP_TRIE_INTERIM_TABLE;
2141 print out the table precompression so that we can do a visual check
2142 that they are identical.
2145 Perl_re_indentf( aTHX_ "Char : ", depth+1 );
2147 for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) {
2148 SV ** const tmp = av_fetch( revcharmap, charid, 0);
2150 Perl_re_printf( aTHX_ "%*s",
2152 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
2153 PL_colors[0], PL_colors[1],
2154 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
2155 PERL_PV_ESCAPE_FIRSTCHAR
2161 Perl_re_printf( aTHX_ "\n");
2162 Perl_re_indentf( aTHX_ "State+-", depth+1 );
2164 for( charid=0 ; charid < trie->uniquecharcount ; charid++ ) {
2165 Perl_re_printf( aTHX_ "%.*s", colwidth,"--------");
2168 Perl_re_printf( aTHX_ "\n" );
2170 for( state=1 ; state < next_alloc ; state += trie->uniquecharcount ) {
2172 Perl_re_indentf( aTHX_ "%4" UVXf " : ",
2174 (UV)TRIE_NODENUM( state ) );
2176 for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) {
2177 UV v=(UV)SAFE_TRIE_NODENUM( trie->trans[ state + charid ].next );
2179 Perl_re_printf( aTHX_ "%*" UVXf, colwidth, v );
2181 Perl_re_printf( aTHX_ "%*s", colwidth, "." );
2183 if ( ! trie->states[ TRIE_NODENUM( state ) ].wordnum ) {
2184 Perl_re_printf( aTHX_ " (%4" UVXf ")\n",
2185 (UV)trie->trans[ state ].check );
2187 Perl_re_printf( aTHX_ " (%4" UVXf ") W%4X\n",
2188 (UV)trie->trans[ state ].check,
2189 trie->states[ TRIE_NODENUM( state ) ].wordnum );
2197 /* make_trie(startbranch,first,last,tail,word_count,flags,depth)
2198 startbranch: the first branch in the whole branch sequence
2199 first : start branch of sequence of branch-exact nodes.
2200 May be the same as startbranch
2201 last : Thing following the last branch.
2202 May be the same as tail.
2203 tail : item following the branch sequence
2204 count : words in the sequence
2205 flags : currently the OP() type we will be building one of /EXACT(|F|FA|FU|FU_SS|L|FLU8)/
2206 depth : indent depth
2208 Inplace optimizes a sequence of 2 or more Branch-Exact nodes into a TRIE node.
2210 A trie is an N'ary tree where the branches are determined by digital
2211 decomposition of the key. IE, at the root node you look up the 1st character and
2212 follow that branch repeat until you find the end of the branches. Nodes can be
2213 marked as "accepting" meaning they represent a complete word. Eg:
2217 would convert into the following structure. Numbers represent states, letters
2218 following numbers represent valid transitions on the letter from that state, if
2219 the number is in square brackets it represents an accepting state, otherwise it
2220 will be in parenthesis.
2222 +-h->+-e->[3]-+-r->(8)-+-s->[9]
2226 (1) +-i->(6)-+-s->[7]
2228 +-s->(3)-+-h->(4)-+-e->[5]
2230 Accept Word Mapping: 3=>1 (he),5=>2 (she), 7=>3 (his), 9=>4 (hers)
2232 This shows that when matching against the string 'hers' we will begin at state 1
2233 read 'h' and move to state 2, read 'e' and move to state 3 which is accepting,
2234 then read 'r' and go to state 8 followed by 's' which takes us to state 9 which
2235 is also accepting. Thus we know that we can match both 'he' and 'hers' with a
2236 single traverse. We store a mapping from accepting to state to which word was
2237 matched, and then when we have multiple possibilities we try to complete the
2238 rest of the regex in the order in which they occurred in the alternation.
2240 The only prior NFA like behaviour that would be changed by the TRIE support is
2241 the silent ignoring of duplicate alternations which are of the form:
2243 / (DUPE|DUPE) X? (?{ ... }) Y /x
2245 Thus EVAL blocks following a trie may be called a different number of times with
2246 and without the optimisation. With the optimisations dupes will be silently
2247 ignored. This inconsistent behaviour of EVAL type nodes is well established as
2248 the following demonstrates:
2250 'words'=~/(word|word|word)(?{ print $1 })[xyz]/
2252 which prints out 'word' three times, but
2254 'words'=~/(word|word|word)(?{ print $1 })S/
2256 which doesnt print it out at all. This is due to other optimisations kicking in.
2258 Example of what happens on a structural level:
2260 The regexp /(ac|ad|ab)+/ will produce the following debug output:
2262 1: CURLYM[1] {1,32767}(18)
2273 This would be optimizable with startbranch=5, first=5, last=16, tail=16
2274 and should turn into:
2276 1: CURLYM[1] {1,32767}(18)
2278 [Words:3 Chars Stored:6 Unique Chars:4 States:5 NCP:1]
2286 Cases where tail != last would be like /(?foo|bar)baz/:
2296 which would be optimizable with startbranch=1, first=1, last=7, tail=8
2297 and would end up looking like:
2300 [Words:2 Chars Stored:6 Unique Chars:5 States:7 NCP:1]
2307 d = uvchr_to_utf8_flags(d, uv, 0);
2309 is the recommended Unicode-aware way of saying
2314 #define TRIE_STORE_REVCHAR(val) \
2317 SV *zlopp = newSV(UTF8_MAXBYTES); \
2318 unsigned char *flrbbbbb = (unsigned char *) SvPVX(zlopp); \
2319 unsigned const char *const kapow = uvchr_to_utf8(flrbbbbb, val); \
2320 SvCUR_set(zlopp, kapow - flrbbbbb); \
2323 av_push(revcharmap, zlopp); \
2325 char ooooff = (char)val; \
2326 av_push(revcharmap, newSVpvn(&ooooff, 1)); \
2330 /* This gets the next character from the input, folding it if not already
2332 #define TRIE_READ_CHAR STMT_START { \
2335 /* if it is UTF then it is either already folded, or does not need \
2337 uvc = valid_utf8_to_uvchr( (const U8*) uc, &len); \
2339 else if (folder == PL_fold_latin1) { \
2340 /* This folder implies Unicode rules, which in the range expressible \
2341 * by not UTF is the lower case, with the two exceptions, one of \
2342 * which should have been taken care of before calling this */ \
2343 assert(*uc != LATIN_SMALL_LETTER_SHARP_S); \
2344 uvc = toLOWER_L1(*uc); \
2345 if (UNLIKELY(uvc == MICRO_SIGN)) uvc = GREEK_SMALL_LETTER_MU; \
2348 /* raw data, will be folded later if needed */ \
2356 #define TRIE_LIST_PUSH(state,fid,ns) STMT_START { \
2357 if ( TRIE_LIST_CUR( state ) >=TRIE_LIST_LEN( state ) ) { \
2358 U32 ging = TRIE_LIST_LEN( state ) *= 2; \
2359 Renew( trie->states[ state ].trans.list, ging, reg_trie_trans_le ); \
2361 TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).forid = fid; \
2362 TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).newstate = ns; \
2363 TRIE_LIST_CUR( state )++; \
2366 #define TRIE_LIST_NEW(state) STMT_START { \
2367 Newxz( trie->states[ state ].trans.list, \
2368 4, reg_trie_trans_le ); \
2369 TRIE_LIST_CUR( state ) = 1; \
2370 TRIE_LIST_LEN( state ) = 4; \
2373 #define TRIE_HANDLE_WORD(state) STMT_START { \
2374 U16 dupe= trie->states[ state ].wordnum; \
2375 regnode * const noper_next = regnext( noper ); \
2378 /* store the word for dumping */ \
2380 if (OP(noper) != NOTHING) \
2381 tmp = newSVpvn_utf8(STRING(noper), STR_LEN(noper), UTF); \
2383 tmp = newSVpvn_utf8( "", 0, UTF ); \
2384 av_push( trie_words, tmp ); \
2388 trie->wordinfo[curword].prev = 0; \
2389 trie->wordinfo[curword].len = wordlen; \
2390 trie->wordinfo[curword].accept = state; \
2392 if ( noper_next < tail ) { \
2394 trie->jump = (U16 *) PerlMemShared_calloc( word_count + 1, \
2396 trie->jump[curword] = (U16)(noper_next - convert); \
2398 jumper = noper_next; \
2400 nextbranch= regnext(cur); \
2404 /* It's a dupe. Pre-insert into the wordinfo[].prev */\
2405 /* chain, so that when the bits of chain are later */\
2406 /* linked together, the dups appear in the chain */\
2407 trie->wordinfo[curword].prev = trie->wordinfo[dupe].prev; \
2408 trie->wordinfo[dupe].prev = curword; \
2410 /* we haven't inserted this word yet. */ \
2411 trie->states[ state ].wordnum = curword; \
2416 #define TRIE_TRANS_STATE(state,base,ucharcount,charid,special) \
2417 ( ( base + charid >= ucharcount \
2418 && base + charid < ubound \
2419 && state == trie->trans[ base - ucharcount + charid ].check \
2420 && trie->trans[ base - ucharcount + charid ].next ) \
2421 ? trie->trans[ base - ucharcount + charid ].next \
2422 : ( state==1 ? special : 0 ) \
2425 #define TRIE_BITMAP_SET_FOLDED(trie, uvc, folder) \
2427 TRIE_BITMAP_SET(trie, uvc); \
2428 /* store the folded codepoint */ \
2430 TRIE_BITMAP_SET(trie, folder[(U8) uvc ]); \
2433 /* store first byte of utf8 representation of */ \
2434 /* variant codepoints */ \
2435 if (! UVCHR_IS_INVARIANT(uvc)) { \
2436 TRIE_BITMAP_SET(trie, UTF8_TWO_BYTE_HI(uvc)); \
2441 #define MADE_JUMP_TRIE 2
2442 #define MADE_EXACT_TRIE 4
2445 S_make_trie(pTHX_ RExC_state_t *pRExC_state, regnode *startbranch,
2446 regnode *first, regnode *last, regnode *tail,
2447 U32 word_count, U32 flags, U32 depth)
2449 /* first pass, loop through and scan words */
2450 reg_trie_data *trie;
2451 HV *widecharmap = NULL;
2452 AV *revcharmap = newAV();
2458 regnode *jumper = NULL;
2459 regnode *nextbranch = NULL;
2460 regnode *convert = NULL;
2461 U32 *prev_states; /* temp array mapping each state to previous one */
2462 /* we just use folder as a flag in utf8 */
2463 const U8 * folder = NULL;
2466 const U32 data_slot = add_data( pRExC_state, STR_WITH_LEN("tuuu"));
2467 AV *trie_words = NULL;
2468 /* along with revcharmap, this only used during construction but both are
2469 * useful during debugging so we store them in the struct when debugging.
2472 const U32 data_slot = add_data( pRExC_state, STR_WITH_LEN("tu"));
2473 STRLEN trie_charcount=0;
2475 SV *re_trie_maxbuff;
2476 GET_RE_DEBUG_FLAGS_DECL;
2478 PERL_ARGS_ASSERT_MAKE_TRIE;
2480 PERL_UNUSED_ARG(depth);
2484 case EXACT: case EXACTL: break;
2488 case EXACTFLU8: folder = PL_fold_latin1; break;
2489 case EXACTF: folder = PL_fold; break;
2490 default: Perl_croak( aTHX_ "panic! In trie construction, unknown node type %u %s", (unsigned) flags, PL_reg_name[flags] );
2493 trie = (reg_trie_data *) PerlMemShared_calloc( 1, sizeof(reg_trie_data) );
2495 trie->startstate = 1;
2496 trie->wordcount = word_count;
2497 RExC_rxi->data->data[ data_slot ] = (void*)trie;
2498 trie->charmap = (U16 *) PerlMemShared_calloc( 256, sizeof(U16) );
2499 if (flags == EXACT || flags == EXACTL)
2500 trie->bitmap = (char *) PerlMemShared_calloc( ANYOF_BITMAP_SIZE, 1 );
2501 trie->wordinfo = (reg_trie_wordinfo *) PerlMemShared_calloc(
2502 trie->wordcount+1, sizeof(reg_trie_wordinfo));
2505 trie_words = newAV();
2508 re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, 1);
2509 assert(re_trie_maxbuff);
2510 if (!SvIOK(re_trie_maxbuff)) {
2511 sv_setiv(re_trie_maxbuff, RE_TRIE_MAXBUF_INIT);
2513 DEBUG_TRIE_COMPILE_r({
2514 Perl_re_indentf( aTHX_
2515 "make_trie start==%d, first==%d, last==%d, tail==%d depth=%d\n",
2517 REG_NODE_NUM(startbranch),REG_NODE_NUM(first),
2518 REG_NODE_NUM(last), REG_NODE_NUM(tail), (int)depth);
2521 /* Find the node we are going to overwrite */
2522 if ( first == startbranch && OP( last ) != BRANCH ) {
2523 /* whole branch chain */
2526 /* branch sub-chain */
2527 convert = NEXTOPER( first );
2530 /* -- First loop and Setup --
2532 We first traverse the branches and scan each word to determine if it
2533 contains widechars, and how many unique chars there are, this is
2534 important as we have to build a table with at least as many columns as we
2537 We use an array of integers to represent the character codes 0..255
2538 (trie->charmap) and we use a an HV* to store Unicode characters. We use
2539 the native representation of the character value as the key and IV's for
2542 *TODO* If we keep track of how many times each character is used we can
2543 remap the columns so that the table compression later on is more
2544 efficient in terms of memory by ensuring the most common value is in the
2545 middle and the least common are on the outside. IMO this would be better
2546 than a most to least common mapping as theres a decent chance the most
2547 common letter will share a node with the least common, meaning the node
2548 will not be compressible. With a middle is most common approach the worst
2549 case is when we have the least common nodes twice.
2553 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
2554 regnode *noper = NEXTOPER( cur );
2558 U32 wordlen = 0; /* required init */
2559 STRLEN minchars = 0;
2560 STRLEN maxchars = 0;
2561 bool set_bit = trie->bitmap ? 1 : 0; /*store the first char in the
2564 if (OP(noper) == NOTHING) {
2565 /* skip past a NOTHING at the start of an alternation
2566 * eg, /(?:)a|(?:b)/ should be the same as /a|b/
2568 regnode *noper_next= regnext(noper);
2569 if (noper_next < tail)
2573 if ( noper < tail &&
2575 OP(noper) == flags ||
2578 OP(noper) == EXACTFU_SS
2582 uc= (U8*)STRING(noper);
2583 e= uc + STR_LEN(noper);
2590 if ( set_bit ) { /* bitmap only alloced when !(UTF&&Folding) */
2591 TRIE_BITMAP_SET(trie,*uc); /* store the raw first byte
2592 regardless of encoding */
2593 if (OP( noper ) == EXACTFU_SS) {
2594 /* false positives are ok, so just set this */
2595 TRIE_BITMAP_SET(trie, LATIN_SMALL_LETTER_SHARP_S);
2599 for ( ; uc < e ; uc += len ) { /* Look at each char in the current
2601 TRIE_CHARCOUNT(trie)++;
2604 /* TRIE_READ_CHAR returns the current character, or its fold if /i
2605 * is in effect. Under /i, this character can match itself, or
2606 * anything that folds to it. If not under /i, it can match just
2607 * itself. Most folds are 1-1, for example k, K, and KELVIN SIGN
2608 * all fold to k, and all are single characters. But some folds
2609 * expand to more than one character, so for example LATIN SMALL
2610 * LIGATURE FFI folds to the three character sequence 'ffi'. If
2611 * the string beginning at 'uc' is 'ffi', it could be matched by
2612 * three characters, or just by the one ligature character. (It
2613 * could also be matched by two characters: LATIN SMALL LIGATURE FF
2614 * followed by 'i', or by 'f' followed by LATIN SMALL LIGATURE FI).
2615 * (Of course 'I' and/or 'F' instead of 'i' and 'f' can also
2616 * match.) The trie needs to know the minimum and maximum number
2617 * of characters that could match so that it can use size alone to
2618 * quickly reject many match attempts. The max is simple: it is
2619 * the number of folded characters in this branch (since a fold is
2620 * never shorter than what folds to it. */
2624 /* And the min is equal to the max if not under /i (indicated by
2625 * 'folder' being NULL), or there are no multi-character folds. If
2626 * there is a multi-character fold, the min is incremented just
2627 * once, for the character that folds to the sequence. Each
2628 * character in the sequence needs to be added to the list below of
2629 * characters in the trie, but we count only the first towards the
2630 * min number of characters needed. This is done through the
2631 * variable 'foldlen', which is returned by the macros that look
2632 * for these sequences as the number of bytes the sequence
2633 * occupies. Each time through the loop, we decrement 'foldlen' by
2634 * how many bytes the current char occupies. Only when it reaches
2635 * 0 do we increment 'minchars' or look for another multi-character
2637 if (folder == NULL) {
2640 else if (foldlen > 0) {
2641 foldlen -= (UTF) ? UTF8SKIP(uc) : 1;
2646 /* See if *uc is the beginning of a multi-character fold. If
2647 * so, we decrement the length remaining to look at, to account
2648 * for the current character this iteration. (We can use 'uc'
2649 * instead of the fold returned by TRIE_READ_CHAR because for
2650 * non-UTF, the latin1_safe macro is smart enough to account
2651 * for all the unfolded characters, and because for UTF, the
2652 * string will already have been folded earlier in the
2653 * compilation process */
2655 if ((foldlen = is_MULTI_CHAR_FOLD_utf8_safe(uc, e))) {
2656 foldlen -= UTF8SKIP(uc);
2659 else if ((foldlen = is_MULTI_CHAR_FOLD_latin1_safe(uc, e))) {
2664 /* The current character (and any potential folds) should be added
2665 * to the possible matching characters for this position in this
2669 U8 folded= folder[ (U8) uvc ];
2670 if ( !trie->charmap[ folded ] ) {
2671 trie->charmap[ folded ]=( ++trie->uniquecharcount );
2672 TRIE_STORE_REVCHAR( folded );
2675 if ( !trie->charmap[ uvc ] ) {
2676 trie->charmap[ uvc ]=( ++trie->uniquecharcount );
2677 TRIE_STORE_REVCHAR( uvc );
2680 /* store the codepoint in the bitmap, and its folded
2682 TRIE_BITMAP_SET_FOLDED(trie, uvc, folder);
2683 set_bit = 0; /* We've done our bit :-) */
2687 /* XXX We could come up with the list of code points that fold
2688 * to this using PL_utf8_foldclosures, except not for
2689 * multi-char folds, as there may be multiple combinations
2690 * there that could work, which needs to wait until runtime to
2691 * resolve (The comment about LIGATURE FFI above is such an
2696 widecharmap = newHV();
2698 svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 1 );
2701 Perl_croak( aTHX_ "error creating/fetching widecharmap entry for 0x%" UVXf, uvc );
2703 if ( !SvTRUE( *svpp ) ) {
2704 sv_setiv( *svpp, ++trie->uniquecharcount );
2705 TRIE_STORE_REVCHAR(uvc);
2708 } /* end loop through characters in this branch of the trie */
2710 /* We take the min and max for this branch and combine to find the min
2711 * and max for all branches processed so far */
2712 if( cur == first ) {
2713 trie->minlen = minchars;
2714 trie->maxlen = maxchars;
2715 } else if (minchars < trie->minlen) {
2716 trie->minlen = minchars;
2717 } else if (maxchars > trie->maxlen) {
2718 trie->maxlen = maxchars;
2720 } /* end first pass */
2721 DEBUG_TRIE_COMPILE_r(
2722 Perl_re_indentf( aTHX_
2723 "TRIE(%s): W:%d C:%d Uq:%d Min:%d Max:%d\n",
2725 ( widecharmap ? "UTF8" : "NATIVE" ), (int)word_count,
2726 (int)TRIE_CHARCOUNT(trie), trie->uniquecharcount,
2727 (int)trie->minlen, (int)trie->maxlen )
2731 We now know what we are dealing with in terms of unique chars and
2732 string sizes so we can calculate how much memory a naive
2733 representation using a flat table will take. If it's over a reasonable
2734 limit (as specified by ${^RE_TRIE_MAXBUF}) we use a more memory
2735 conservative but potentially much slower representation using an array
2738 At the end we convert both representations into the same compressed
2739 form that will be used in regexec.c for matching with. The latter
2740 is a form that cannot be used to construct with but has memory
2741 properties similar to the list form and access properties similar
2742 to the table form making it both suitable for fast searches and
2743 small enough that its feasable to store for the duration of a program.
2745 See the comment in the code where the compressed table is produced
2746 inplace from the flat tabe representation for an explanation of how
2747 the compression works.
2752 Newx(prev_states, TRIE_CHARCOUNT(trie) + 2, U32);
2755 if ( (IV)( ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount + 1)
2756 > SvIV(re_trie_maxbuff) )
2759 Second Pass -- Array Of Lists Representation
2761 Each state will be represented by a list of charid:state records
2762 (reg_trie_trans_le) the first such element holds the CUR and LEN
2763 points of the allocated array. (See defines above).
2765 We build the initial structure using the lists, and then convert
2766 it into the compressed table form which allows faster lookups
2767 (but cant be modified once converted).
2770 STRLEN transcount = 1;
2772 DEBUG_TRIE_COMPILE_MORE_r( Perl_re_indentf( aTHX_ "Compiling trie using list compiler\n",
2775 trie->states = (reg_trie_state *)
2776 PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2,
2777 sizeof(reg_trie_state) );
2781 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
2783 regnode *noper = NEXTOPER( cur );
2784 U32 state = 1; /* required init */
2785 U16 charid = 0; /* sanity init */
2786 U32 wordlen = 0; /* required init */
2788 if (OP(noper) == NOTHING) {
2789 regnode *noper_next= regnext(noper);
2790 if (noper_next < tail)
2794 if ( noper < tail && ( OP(noper) == flags || ( flags == EXACTFU && OP(noper) == EXACTFU_SS ) ) ) {
2795 const U8 *uc= (U8*)STRING(noper);
2796 const U8 *e= uc + STR_LEN(noper);
2798 for ( ; uc < e ; uc += len ) {
2803 charid = trie->charmap[ uvc ];
2805 SV** const svpp = hv_fetch( widecharmap,
2812 charid=(U16)SvIV( *svpp );
2815 /* charid is now 0 if we dont know the char read, or
2816 * nonzero if we do */
2823 if ( !trie->states[ state ].trans.list ) {
2824 TRIE_LIST_NEW( state );
2827 check <= TRIE_LIST_USED( state );
2830 if ( TRIE_LIST_ITEM( state, check ).forid
2833 newstate = TRIE_LIST_ITEM( state, check ).newstate;
2838 newstate = next_alloc++;
2839 prev_states[newstate] = state;
2840 TRIE_LIST_PUSH( state, charid, newstate );
2845 Perl_croak( aTHX_ "panic! In trie construction, no char mapping for %" IVdf, uvc );
2849 TRIE_HANDLE_WORD(state);
2851 } /* end second pass */
2853 /* next alloc is the NEXT state to be allocated */
2854 trie->statecount = next_alloc;
2855 trie->states = (reg_trie_state *)
2856 PerlMemShared_realloc( trie->states,
2858 * sizeof(reg_trie_state) );
2860 /* and now dump it out before we compress it */
2861 DEBUG_TRIE_COMPILE_MORE_r(dump_trie_interim_list(trie, widecharmap,
2862 revcharmap, next_alloc,
2866 trie->trans = (reg_trie_trans *)
2867 PerlMemShared_calloc( transcount, sizeof(reg_trie_trans) );
2874 for( state=1 ; state < next_alloc ; state ++ ) {
2878 DEBUG_TRIE_COMPILE_MORE_r(
2879 Perl_re_printf( aTHX_ "tp: %d zp: %d ",tp,zp)
2883 if (trie->states[state].trans.list) {
2884 U16 minid=TRIE_LIST_ITEM( state, 1).forid;
2888 for( idx = 2 ; idx <= TRIE_LIST_USED( state ) ; idx++ ) {
2889 const U16 forid = TRIE_LIST_ITEM( state, idx).forid;
2890 if ( forid < minid ) {
2892 } else if ( forid > maxid ) {
2896 if ( transcount < tp + maxid - minid + 1) {
2898 trie->trans = (reg_trie_trans *)
2899 PerlMemShared_realloc( trie->trans,
2901 * sizeof(reg_trie_trans) );
2902 Zero( trie->trans + (transcount / 2),
2906 base = trie->uniquecharcount + tp - minid;
2907 if ( maxid == minid ) {
2909 for ( ; zp < tp ; zp++ ) {
2910 if ( ! trie->trans[ zp ].next ) {
2911 base = trie->uniquecharcount + zp - minid;
2912 trie->trans[ zp ].next = TRIE_LIST_ITEM( state,
2914 trie->trans[ zp ].check = state;
2920 trie->trans[ tp ].next = TRIE_LIST_ITEM( state,
2922 trie->trans[ tp ].check = state;
2927 for ( idx=1; idx <= TRIE_LIST_USED( state ) ; idx++ ) {
2928 const U32 tid = base
2929 - trie->uniquecharcount
2930 + TRIE_LIST_ITEM( state, idx ).forid;
2931 trie->trans[ tid ].next = TRIE_LIST_ITEM( state,
2933 trie->trans[ tid ].check = state;
2935 tp += ( maxid - minid + 1 );
2937 Safefree(trie->states[ state ].trans.list);
2940 DEBUG_TRIE_COMPILE_MORE_r(
2941 Perl_re_printf( aTHX_ " base: %d\n",base);
2944 trie->states[ state ].trans.base=base;
2946 trie->lasttrans = tp + 1;
2950 Second Pass -- Flat Table Representation.
2952 we dont use the 0 slot of either trans[] or states[] so we add 1 to
2953 each. We know that we will need Charcount+1 trans at most to store
2954 the data (one row per char at worst case) So we preallocate both
2955 structures assuming worst case.
2957 We then construct the trie using only the .next slots of the entry
2960 We use the .check field of the first entry of the node temporarily
2961 to make compression both faster and easier by keeping track of how
2962 many non zero fields are in the node.
2964 Since trans are numbered from 1 any 0 pointer in the table is a FAIL
2967 There are two terms at use here: state as a TRIE_NODEIDX() which is
2968 a number representing the first entry of the node, and state as a
2969 TRIE_NODENUM() which is the trans number. state 1 is TRIE_NODEIDX(1)
2970 and TRIE_NODENUM(1), state 2 is TRIE_NODEIDX(2) and TRIE_NODENUM(3)
2971 if there are 2 entrys per node. eg:
2979 The table is internally in the right hand, idx form. However as we
2980 also have to deal with the states array which is indexed by nodenum
2981 we have to use TRIE_NODENUM() to convert.
2984 DEBUG_TRIE_COMPILE_MORE_r( Perl_re_indentf( aTHX_ "Compiling trie using table compiler\n",
2987 trie->trans = (reg_trie_trans *)
2988 PerlMemShared_calloc( ( TRIE_CHARCOUNT(trie) + 1 )
2989 * trie->uniquecharcount + 1,
2990 sizeof(reg_trie_trans) );
2991 trie->states = (reg_trie_state *)
2992 PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2,
2993 sizeof(reg_trie_state) );
2994 next_alloc = trie->uniquecharcount + 1;
2997 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
2999 regnode *noper = NEXTOPER( cur );
3001 U32 state = 1; /* required init */
3003 U16 charid = 0; /* sanity init */
3004 U32 accept_state = 0; /* sanity init */
3006 U32 wordlen = 0; /* required init */
3008 if (OP(noper) == NOTHING) {
3009 regnode *noper_next= regnext(noper);
3010 if (noper_next < tail)
3014 if ( noper < tail && ( OP(noper) == flags || ( flags == EXACTFU && OP(noper) == EXACTFU_SS ) ) ) {
3015 const U8 *uc= (U8*)STRING(noper);
3016 const U8 *e= uc + STR_LEN(noper);
3018 for ( ; uc < e ; uc += len ) {
3023 charid = trie->charmap[ uvc ];
3025 SV* const * const svpp = hv_fetch( widecharmap,
3029 charid = svpp ? (U16)SvIV(*svpp) : 0;
3033 if ( !trie->trans[ state + charid ].next ) {
3034 trie->trans[ state + charid ].next = next_alloc;
3035 trie->trans[ state ].check++;
3036 prev_states[TRIE_NODENUM(next_alloc)]
3037 = TRIE_NODENUM(state);
3038 next_alloc += trie->uniquecharcount;
3040 state = trie->trans[ state + charid ].next;
3042 Perl_croak( aTHX_ "panic! In trie construction, no char mapping for %" IVdf, uvc );
3044 /* charid is now 0 if we dont know the char read, or
3045 * nonzero if we do */
3048 accept_state = TRIE_NODENUM( state );
3049 TRIE_HANDLE_WORD(accept_state);
3051 } /* end second pass */
3053 /* and now dump it out before we compress it */
3054 DEBUG_TRIE_COMPILE_MORE_r(dump_trie_interim_table(trie, widecharmap,
3056 next_alloc, depth+1));
3060 * Inplace compress the table.*
3062 For sparse data sets the table constructed by the trie algorithm will
3063 be mostly 0/FAIL transitions or to put it another way mostly empty.
3064 (Note that leaf nodes will not contain any transitions.)
3066 This algorithm compresses the tables by eliminating most such
3067 transitions, at the cost of a modest bit of extra work during lookup:
3069 - Each states[] entry contains a .base field which indicates the
3070 index in the state[] array wheres its transition data is stored.
3072 - If .base is 0 there are no valid transitions from that node.
3074 - If .base is nonzero then charid is added to it to find an entry in
3077 -If trans[states[state].base+charid].check!=state then the
3078 transition is taken to be a 0/Fail transition. Thus if there are fail
3079 transitions at the front of the node then the .base offset will point
3080 somewhere inside the previous nodes data (or maybe even into a node
3081 even earlier), but the .check field determines if the transition is
3085 The following process inplace converts the table to the compressed
3086 table: We first do not compress the root node 1,and mark all its
3087 .check pointers as 1 and set its .base pointer as 1 as well. This
3088 allows us to do a DFA construction from the compressed table later,
3089 and ensures that any .base pointers we calculate later are greater
3092 - We set 'pos' to indicate the first entry of the second node.
3094 - We then iterate over the columns of the node, finding the first and
3095 last used entry at l and m. We then copy l..m into pos..(pos+m-l),
3096 and set the .check pointers accordingly, and advance pos
3097 appropriately and repreat for the next node. Note that when we copy
3098 the next pointers we have to convert them from the original
3099 NODEIDX form to NODENUM form as the former is not valid post
3102 - If a node has no transitions used we mark its base as 0 and do not
3103 advance the pos pointer.
3105 - If a node only has one transition we use a second pointer into the
3106 structure to fill in allocated fail transitions from other states.
3107 This pointer is independent of the main pointer and scans forward
3108 looking for null transitions that are allocated to a state. When it
3109 finds one it writes the single transition into the "hole". If the
3110 pointer doesnt find one the single transition is appended as normal.
3112 - Once compressed we can Renew/realloc the structures to release the
3115 See "Table-Compression Methods" in sec 3.9 of the Red Dragon,
3116 specifically Fig 3.47 and the associated pseudocode.
3120 const U32 laststate = TRIE_NODENUM( next_alloc );
3123 trie->statecount = laststate;
3125 for ( state = 1 ; state < laststate ; state++ ) {
3127 const U32 stateidx = TRIE_NODEIDX( state );
3128 const U32 o_used = trie->trans[ stateidx ].check;
3129 U32 used = trie->trans[ stateidx ].check;
3130 trie->trans[ stateidx ].check = 0;
3133 used && charid < trie->uniquecharcount;
3136 if ( flag || trie->trans[ stateidx + charid ].next ) {
3137 if ( trie->trans[ stateidx + charid ].next ) {
3139 for ( ; zp < pos ; zp++ ) {
3140 if ( ! trie->trans[ zp ].next ) {
3144 trie->states[ state ].trans.base
3146 + trie->uniquecharcount
3148 trie->trans[ zp ].next
3149 = SAFE_TRIE_NODENUM( trie->trans[ stateidx
3151 trie->trans[ zp ].check = state;
3152 if ( ++zp > pos ) pos = zp;
3159 trie->states[ state ].trans.base
3160 = pos + trie->uniquecharcount - charid ;
3162 trie->trans[ pos ].next
3163 = SAFE_TRIE_NODENUM(
3164 trie->trans[ stateidx + charid ].next );
3165 trie->trans[ pos ].check = state;
3170 trie->lasttrans = pos + 1;
3171 trie->states = (reg_trie_state *)
3172 PerlMemShared_realloc( trie->states, laststate
3173 * sizeof(reg_trie_state) );
3174 DEBUG_TRIE_COMPILE_MORE_r(
3175 Perl_re_indentf( aTHX_ "Alloc: %d Orig: %" IVdf " elements, Final:%" IVdf ". Savings of %%%5.2f\n",
3177 (int)( ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount
3181 ( ( next_alloc - pos ) * 100 ) / (double)next_alloc );
3184 } /* end table compress */
3186 DEBUG_TRIE_COMPILE_MORE_r(
3187 Perl_re_indentf( aTHX_ "Statecount:%" UVxf " Lasttrans:%" UVxf "\n",
3189 (UV)trie->statecount,
3190 (UV)trie->lasttrans)
3192 /* resize the trans array to remove unused space */
3193 trie->trans = (reg_trie_trans *)
3194 PerlMemShared_realloc( trie->trans, trie->lasttrans
3195 * sizeof(reg_trie_trans) );
3197 { /* Modify the program and insert the new TRIE node */
3198 U8 nodetype =(U8)(flags & 0xFF);
3202 regnode *optimize = NULL;
3203 #ifdef RE_TRACK_PATTERN_OFFSETS
3206 U32 mjd_nodelen = 0;
3207 #endif /* RE_TRACK_PATTERN_OFFSETS */
3208 #endif /* DEBUGGING */
3210 This means we convert either the first branch or the first Exact,
3211 depending on whether the thing following (in 'last') is a branch
3212 or not and whther first is the startbranch (ie is it a sub part of
3213 the alternation or is it the whole thing.)
3214 Assuming its a sub part we convert the EXACT otherwise we convert
3215 the whole branch sequence, including the first.
3217 /* Find the node we are going to overwrite */
3218 if ( first != startbranch || OP( last ) == BRANCH ) {
3219 /* branch sub-chain */
3220 NEXT_OFF( first ) = (U16)(last - first);
3221 #ifdef RE_TRACK_PATTERN_OFFSETS
3223 mjd_offset= Node_Offset((convert));
3224 mjd_nodelen= Node_Length((convert));
3227 /* whole branch chain */
3229 #ifdef RE_TRACK_PATTERN_OFFSETS
3232 const regnode *nop = NEXTOPER( convert );
3233 mjd_offset= Node_Offset((nop));
3234 mjd_nodelen= Node_Length((nop));
3238 Perl_re_indentf( aTHX_ "MJD offset:%" UVuf " MJD length:%" UVuf "\n",
3240 (UV)mjd_offset, (UV)mjd_nodelen)
3243 /* But first we check to see if there is a common prefix we can
3244 split out as an EXACT and put in front of the TRIE node. */
3245 trie->startstate= 1;
3246 if ( trie->bitmap && !widecharmap && !trie->jump ) {
3247 /* we want to find the first state that has more than
3248 * one transition, if that state is not the first state
3249 * then we have a common prefix which we can remove.
3252 for ( state = 1 ; state < trie->statecount-1 ; state++ ) {
3254 I32 first_ofs = -1; /* keeps track of the ofs of the first
3255 transition, -1 means none */
3257 const U32 base = trie->states[ state ].trans.base;
3259 /* does this state terminate an alternation? */
3260 if ( trie->states[state].wordnum )
3263 for ( ofs = 0 ; ofs < trie->uniquecharcount ; ofs++ ) {
3264 if ( ( base + ofs >= trie->uniquecharcount ) &&
3265 ( base + ofs - trie->uniquecharcount < trie->lasttrans ) &&
3266 trie->trans[ base + ofs - trie->uniquecharcount ].check == state )
3268 if ( ++count > 1 ) {
3269 /* we have more than one transition */
3272 /* if this is the first state there is no common prefix
3273 * to extract, so we can exit */
3274 if ( state == 1 ) break;
3275 tmp = av_fetch( revcharmap, ofs, 0);
3276 ch = (U8*)SvPV_nolen_const( *tmp );
3278 /* if we are on count 2 then we need to initialize the
3279 * bitmap, and store the previous char if there was one
3282 /* clear the bitmap */
3283 Zero(trie->bitmap, ANYOF_BITMAP_SIZE, char);
3285 Perl_re_indentf( aTHX_ "New Start State=%" UVuf " Class: [",
3288 if (first_ofs >= 0) {
3289 SV ** const tmp = av_fetch( revcharmap, first_ofs, 0);
3290 const U8 * const ch = (U8*)SvPV_nolen_const( *tmp );
3292 TRIE_BITMAP_SET_FOLDED(trie,*ch,folder);
3294 Perl_re_printf( aTHX_ "%s", (char*)ch)
3298 /* store the current firstchar in the bitmap */
3299 TRIE_BITMAP_SET_FOLDED(trie,*ch,folder);
3300 DEBUG_OPTIMISE_r(Perl_re_printf( aTHX_ "%s", ch));
3306 /* This state has only one transition, its transition is part
3307 * of a common prefix - we need to concatenate the char it
3308 * represents to what we have so far. */
3309 SV **tmp = av_fetch( revcharmap, first_ofs, 0);
3311 char *ch = SvPV( *tmp, len );
3313 SV *sv=sv_newmortal();
3314 Perl_re_indentf( aTHX_ "Prefix State: %" UVuf " Ofs:%" UVuf " Char='%s'\n",
3316 (UV)state, (UV)first_ofs,
3317 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 6,
3318 PL_colors[0], PL_colors[1],
3319 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
3320 PERL_PV_ESCAPE_FIRSTCHAR
3325 OP( convert ) = nodetype;
3326 str=STRING(convert);
3329 STR_LEN(convert) += len;
3335 DEBUG_OPTIMISE_r(Perl_re_printf( aTHX_ "]\n"));
3340 trie->prefixlen = (state-1);
3342 regnode *n = convert+NODE_SZ_STR(convert);
3343 NEXT_OFF(convert) = NODE_SZ_STR(convert);
3344 trie->startstate = state;
3345 trie->minlen -= (state - 1);
3346 trie->maxlen -= (state - 1);
3348 /* At least the UNICOS C compiler choked on this
3349 * being argument to DEBUG_r(), so let's just have
3352 #ifdef PERL_EXT_RE_BUILD
3358 regnode *fix = convert;
3359 U32 word = trie->wordcount;
3361 Set_Node_Offset_Length(convert, mjd_offset, state - 1);
3362 while( ++fix < n ) {
3363 Set_Node_Offset_Length(fix, 0, 0);
3366 SV ** const tmp = av_fetch( trie_words, word, 0 );
3368 if ( STR_LEN(convert) <= SvCUR(*tmp) )
3369 sv_chop(*tmp, SvPV_nolen(*tmp) + STR_LEN(convert));
3371 sv_chop(*tmp, SvPV_nolen(*tmp) + SvCUR(*tmp));
3379 NEXT_OFF(convert) = (U16)(tail - convert);
3380 DEBUG_r(optimize= n);
3386 if ( trie->maxlen ) {
3387 NEXT_OFF( convert ) = (U16)(tail - convert);
3388 ARG_SET( convert, data_slot );
3389 /* Store the offset to the first unabsorbed branch in
3390 jump[0], which is otherwise unused by the jump logic.
3391 We use this when dumping a trie and during optimisation. */
3393 trie->jump[0] = (U16)(nextbranch - convert);
3395 /* If the start state is not accepting (meaning there is no empty string/NOTHING)
3396 * and there is a bitmap
3397 * and the first "jump target" node we found leaves enough room
3398 * then convert the TRIE node into a TRIEC node, with the bitmap
3399 * embedded inline in the opcode - this is hypothetically faster.
3401 if ( !trie->states[trie->startstate].wordnum
3403 && ( (char *)jumper - (char *)convert) >= (int)sizeof(struct regnode_charclass) )
3405 OP( convert ) = TRIEC;
3406 Copy(trie->bitmap, ((struct regnode_charclass *)convert)->bitmap, ANYOF_BITMAP_SIZE, char);
3407 PerlMemShared_free(trie->bitmap);
3410 OP( convert ) = TRIE;
3412 /* store the type in the flags */
3413 convert->flags = nodetype;
3417 + regarglen[ OP( convert ) ];
3419 /* XXX We really should free up the resource in trie now,
3420 as we won't use them - (which resources?) dmq */
3422 /* needed for dumping*/
3423 DEBUG_r(if (optimize) {
3424 regnode *opt = convert;
3426 while ( ++opt < optimize) {
3427 Set_Node_Offset_Length(opt,0,0);
3430 Try to clean up some of the debris left after the
3433 while( optimize < jumper ) {
3434 mjd_nodelen += Node_Length((optimize));
3435 OP( optimize ) = OPTIMIZED;
3436 Set_Node_Offset_Length(optimize,0,0);
3439 Set_Node_Offset_Length(convert,mjd_offset,mjd_nodelen);
3441 } /* end node insert */
3443 /* Finish populating the prev field of the wordinfo array. Walk back
3444 * from each accept state until we find another accept state, and if
3445 * so, point the first word's .prev field at the second word. If the
3446 * second already has a .prev field set, stop now. This will be the
3447 * case either if we've already processed that word's accept state,
3448 * or that state had multiple words, and the overspill words were
3449 * already linked up earlier.
3456 for (word=1; word <= trie->wordcount; word++) {
3458 if (trie->wordinfo[word].prev)
3460 state = trie->wordinfo[word].accept;
3462 state = prev_states[state];
3465 prev = trie->states[state].wordnum;
3469 trie->wordinfo[word].prev = prev;
3471 Safefree(prev_states);
3475 /* and now dump out the compressed format */
3476 DEBUG_TRIE_COMPILE_r(dump_trie(trie, widecharmap, revcharmap, depth+1));
3478 RExC_rxi->data->data[ data_slot + 1 ] = (void*)widecharmap;
3480 RExC_rxi->data->data[ data_slot + TRIE_WORDS_OFFSET ] = (void*)trie_words;
3481 RExC_rxi->data->data[ data_slot + 3 ] = (void*)revcharmap;
3483 SvREFCNT_dec_NN(revcharmap);
3487 : trie->startstate>1
3493 S_construct_ahocorasick_from_trie(pTHX_ RExC_state_t *pRExC_state, regnode *source, U32 depth)
3495 /* The Trie is constructed and compressed now so we can build a fail array if
3498 This is basically the Aho-Corasick algorithm. Its from exercise 3.31 and
3500 "Red Dragon" -- Compilers, principles, techniques, and tools. Aho, Sethi,
3504 We find the fail state for each state in the trie, this state is the longest
3505 proper suffix of the current state's 'word' that is also a proper prefix of
3506 another word in our trie. State 1 represents the word '' and is thus the
3507 default fail state. This allows the DFA not to have to restart after its
3508 tried and failed a word at a given point, it simply continues as though it
3509 had been matching the other word in the first place.
3511 'abcdgu'=~/abcdefg|cdgu/
3512 When we get to 'd' we are still matching the first word, we would encounter
3513 'g' which would fail, which would bring us to the state representing 'd' in
3514 the second word where we would try 'g' and succeed, proceeding to match
3517 /* add a fail transition */
3518 const U32 trie_offset = ARG(source);
3519 reg_trie_data *trie=(reg_trie_data *)RExC_rxi->data->data[trie_offset];
3521 const U32 ucharcount = trie->uniquecharcount;
3522 const U32 numstates = trie->statecount;
3523 const U32 ubound = trie->lasttrans + ucharcount;
3527 U32 base = trie->states[ 1 ].trans.base;
3530 const U32 data_slot = add_data( pRExC_state, STR_WITH_LEN("T"));
3532 GET_RE_DEBUG_FLAGS_DECL;
3534 PERL_ARGS_ASSERT_CONSTRUCT_AHOCORASICK_FROM_TRIE;
3535 PERL_UNUSED_CONTEXT;
3537 PERL_UNUSED_ARG(depth);
3540 if ( OP(source) == TRIE ) {
3541 struct regnode_1 *op = (struct regnode_1 *)
3542 PerlMemShared_calloc(1, sizeof(struct regnode_1));
3543 StructCopy(source,op,struct regnode_1);
3544 stclass = (regnode *)op;
3546 struct regnode_charclass *op = (struct regnode_charclass *)
3547 PerlMemShared_calloc(1, sizeof(struct regnode_charclass));
3548 StructCopy(source,op,struct regnode_charclass);
3549 stclass = (regnode *)op;
3551 OP(stclass)+=2; /* convert the TRIE type to its AHO-CORASICK equivalent */
3553 ARG_SET( stclass, data_slot );
3554 aho = (reg_ac_data *) PerlMemShared_calloc( 1, sizeof(reg_ac_data) );
3555 RExC_rxi->data->data[ data_slot ] = (void*)aho;
3556 aho->trie=trie_offset;
3557 aho->states=(reg_trie_state *)PerlMemShared_malloc( numstates * sizeof(reg_trie_state) );
3558 Copy( trie->states, aho->states, numstates, reg_trie_state );
3559 Newxz( q, numstates, U32);
3560 aho->fail = (U32 *) PerlMemShared_calloc( numstates, sizeof(U32) );
3563 /* initialize fail[0..1] to be 1 so that we always have
3564 a valid final fail state */
3565 fail[ 0 ] = fail[ 1 ] = 1;
3567 for ( charid = 0; charid < ucharcount ; charid++ ) {
3568 const U32 newstate = TRIE_TRANS_STATE( 1, base, ucharcount, charid, 0 );
3570 q[ q_write ] = newstate;
3571 /* set to point at the root */
3572 fail[ q[ q_write++ ] ]=1;
3575 while ( q_read < q_write) {
3576 const U32 cur = q[ q_read++ % numstates ];
3577 base = trie->states[ cur ].trans.base;
3579 for ( charid = 0 ; charid < ucharcount ; charid++ ) {
3580 const U32 ch_state = TRIE_TRANS_STATE( cur, base, ucharcount, charid, 1 );
3582 U32 fail_state = cur;
3585 fail_state = fail[ fail_state ];
3586 fail_base = aho->states[ fail_state ].trans.base;
3587 } while ( !TRIE_TRANS_STATE( fail_state, fail_base, ucharcount, charid, 1 ) );
3589 fail_state = TRIE_TRANS_STATE( fail_state, fail_base, ucharcount, charid, 1 );
3590 fail[ ch_state ] = fail_state;
3591 if ( !aho->states[ ch_state ].wordnum && aho->states[ fail_state ].wordnum )
3593 aho->states[ ch_state ].wordnum = aho->states[ fail_state ].wordnum;
3595 q[ q_write++ % numstates] = ch_state;
3599 /* restore fail[0..1] to 0 so that we "fall out" of the AC loop
3600 when we fail in state 1, this allows us to use the
3601 charclass scan to find a valid start char. This is based on the principle
3602 that theres a good chance the string being searched contains lots of stuff
3603 that cant be a start char.
3605 fail[ 0 ] = fail[ 1 ] = 0;
3606 DEBUG_TRIE_COMPILE_r({
3607 Perl_re_indentf( aTHX_ "Stclass Failtable (%" UVuf " states): 0",
3608 depth, (UV)numstates
3610 for( q_read=1; q_read<numstates; q_read++ ) {
3611 Perl_re_printf( aTHX_ ", %" UVuf, (UV)fail[q_read]);
3613 Perl_re_printf( aTHX_ "\n");
3616 /*RExC_seen |= REG_TRIEDFA_SEEN;*/
3621 #define DEBUG_PEEP(str,scan,depth) \
3622 DEBUG_OPTIMISE_r({if (scan){ \
3623 regnode *Next = regnext(scan); \
3624 regprop(RExC_rx, RExC_mysv, scan, NULL, pRExC_state);\
3625 Perl_re_indentf( aTHX_ "" str ">%3d: %s (%d)", \
3626 depth, REG_NODE_NUM(scan), SvPV_nolen_const(RExC_mysv),\
3627 Next ? (REG_NODE_NUM(Next)) : 0 );\
3628 DEBUG_SHOW_STUDY_FLAGS(flags," [ ","]");\
3629 Perl_re_printf( aTHX_ "\n"); \
3632 /* The below joins as many adjacent EXACTish nodes as possible into a single
3633 * one. The regop may be changed if the node(s) contain certain sequences that
3634 * require special handling. The joining is only done if:
3635 * 1) there is room in the current conglomerated node to entirely contain the
3637 * 2) they are the exact same node type
3639 * The adjacent nodes actually may be separated by NOTHING-kind nodes, and
3640 * these get optimized out
3642 * XXX khw thinks this should be enhanced to fill EXACT (at least) nodes as full
3643 * as possible, even if that means splitting an existing node so that its first
3644 * part is moved to the preceeding node. This would maximise the efficiency of
3645 * memEQ during matching. Elsewhere in this file, khw proposes splitting
3646 * EXACTFish nodes into portions that don't change under folding vs those that
3647 * do. Those portions that don't change may be the only things in the pattern that
3648 * could be used to find fixed and floating strings.
3650 * If a node is to match under /i (folded), the number of characters it matches
3651 * can be different than its character length if it contains a multi-character
3652 * fold. *min_subtract is set to the total delta number of characters of the
3655 * And *unfolded_multi_char is set to indicate whether or not the node contains
3656 * an unfolded multi-char fold. This happens when whether the fold is valid or
3657 * not won't be known until runtime; namely for EXACTF nodes that contain LATIN
3658 * SMALL LETTER SHARP S, as only if the target string being matched against
3659 * turns out to be UTF-8 is that fold valid; and also for EXACTFL nodes whose
3660 * folding rules depend on the locale in force at runtime. (Multi-char folds
3661 * whose components are all above the Latin1 range are not run-time locale
3662 * dependent, and have already been folded by the time this function is
3665 * This is as good a place as any to discuss the design of handling these
3666 * multi-character fold sequences. It's been wrong in Perl for a very long
3667 * time. There are three code points in Unicode whose multi-character folds
3668 * were long ago discovered to mess things up. The previous designs for
3669 * dealing with these involved assigning a special node for them. This
3670 * approach doesn't always work, as evidenced by this example:
3671 * "\xDFs" =~ /s\xDF/ui # Used to fail before these patches
3672 * Both sides fold to "sss", but if the pattern is parsed to create a node that
3673 * would match just the \xDF, it won't be able to handle the case where a
3674 * successful match would have to cross the node's boundary. The new approach
3675 * that hopefully generally solves the problem generates an EXACTFU_SS node
3676 * that is "sss" in this case.
3678 * It turns out that there are problems with all multi-character folds, and not
3679 * just these three. Now the code is general, for all such cases. The
3680 * approach taken is:
3681 * 1) This routine examines each EXACTFish node that could contain multi-
3682 * character folded sequences. Since a single character can fold into
3683 * such a sequence, the minimum match length for this node is less than
3684 * the number of characters in the node. This routine returns in
3685 * *min_subtract how many characters to subtract from the the actual
3686 * length of the string to get a real minimum match length; it is 0 if
3687 * there are no multi-char foldeds. This delta is used by the caller to
3688 * adjust the min length of the match, and the delta between min and max,
3689 * so that the optimizer doesn't reject these possibilities based on size
3691 * 2) For the sequence involving the Sharp s (\xDF), the node type EXACTFU_SS
3692 * is used for an EXACTFU node that contains at least one "ss" sequence in
3693 * it. For non-UTF-8 patterns and strings, this is the only case where
3694 * there is a possible fold length change. That means that a regular
3695 * EXACTFU node without UTF-8 involvement doesn't have to concern itself
3696 * with length changes, and so can be processed faster. regexec.c takes
3697 * advantage of this. Generally, an EXACTFish node that is in UTF-8 is
3698 * pre-folded by regcomp.c (except EXACTFL, some of whose folds aren't
3699 * known until runtime). This saves effort in regex matching. However,
3700 * the pre-folding isn't done for non-UTF8 patterns because the fold of
3701 * the MICRO SIGN requires UTF-8, and we don't want to slow things down by
3702 * forcing the pattern into UTF8 unless necessary. Also what EXACTF (and,
3703 * again, EXACTFL) nodes fold to isn't known until runtime. The fold
3704 * possibilities for the non-UTF8 patterns are quite simple, except for
3705 * the sharp s. All the ones that don't involve a UTF-8 target string are
3706 * members of a fold-pair, and arrays are set up for all of them so that
3707 * the other member of the pair can be found quickly. Code elsewhere in
3708 * this file makes sure that in EXACTFU nodes, the sharp s gets folded to
3709 * 'ss', even if the pattern isn't UTF-8. This avoids the issues
3710 * described in the next item.
3711 * 3) A problem remains for unfolded multi-char folds. (These occur when the
3712 * validity of the fold won't be known until runtime, and so must remain
3713 * unfolded for now. This happens for the sharp s in EXACTF and EXACTFA
3714 * nodes when the pattern isn't in UTF-8. (Note, BTW, that there cannot
3715 * be an EXACTF node with a UTF-8 pattern.) They also occur for various
3716 * folds in EXACTFL nodes, regardless of the UTF-ness of the pattern.)
3717 * The reason this is a problem is that the optimizer part of regexec.c
3718 * (probably unwittingly, in Perl_regexec_flags()) makes an assumption
3719 * that a character in the pattern corresponds to at most a single
3720 * character in the target string. (And I do mean character, and not byte
3721 * here, unlike other parts of the documentation that have never been
3722 * updated to account for multibyte Unicode.) sharp s in EXACTF and
3723 * EXACTFL nodes can match the two character string 'ss'; in EXACTFA nodes
3724 * it can match "\x{17F}\x{17F}". These, along with other ones in EXACTFL
3725 * nodes, violate the assumption, and they are the only instances where it
3726 * is violated. I'm reluctant to try to change the assumption, as the
3727 * code involved is impenetrable to me (khw), so instead the code here
3728 * punts. This routine examines EXACTFL nodes, and (when the pattern
3729 * isn't UTF-8) EXACTF and EXACTFA for such unfolded folds, and returns a
3730 * boolean indicating whether or not the node contains such a fold. When
3731 * it is true, the caller sets a flag that later causes the optimizer in
3732 * this file to not set values for the floating and fixed string lengths,
3733 * and thus avoids the optimizer code in regexec.c that makes the invalid
3734 * assumption. Thus, there is no optimization based on string lengths for
3735 * EXACTFL nodes that contain these few folds, nor for non-UTF8-pattern
3736 * EXACTF and EXACTFA nodes that contain the sharp s. (The reason the
3737 * assumption is wrong only in these cases is that all other non-UTF-8
3738 * folds are 1-1; and, for UTF-8 patterns, we pre-fold all other folds to
3739 * their expanded versions. (Again, we can't prefold sharp s to 'ss' in
3740 * EXACTF nodes because we don't know at compile time if it actually
3741 * matches 'ss' or not. For EXACTF nodes it will match iff the target
3742 * string is in UTF-8. This is in contrast to EXACTFU nodes, where it
3743 * always matches; and EXACTFA where it never does. In an EXACTFA node in
3744 * a UTF-8 pattern, sharp s is folded to "\x{17F}\x{17F}, avoiding the
3745 * problem; but in a non-UTF8 pattern, folding it to that above-Latin1
3746 * string would require the pattern to be forced into UTF-8, the overhead
3747 * of which we want to avoid. Similarly the unfolded multi-char folds in
3748 * EXACTFL nodes will match iff the locale at the time of match is a UTF-8
3751 * Similarly, the code that generates tries doesn't currently handle
3752 * not-already-folded multi-char folds, and it looks like a pain to change
3753 * that. Therefore, trie generation of EXACTFA nodes with the sharp s
3754 * doesn't work. Instead, such an EXACTFA is turned into a new regnode,
3755 * EXACTFA_NO_TRIE, which the trie code knows not to handle. Most people
3756 * using /iaa matching will be doing so almost entirely with ASCII
3757 * strings, so this should rarely be encountered in practice */
3759 #define JOIN_EXACT(scan,min_subtract,unfolded_multi_char, flags) \
3760 if (PL_regkind[OP(scan)] == EXACT) \
3761 join_exact(pRExC_state,(scan),(min_subtract),unfolded_multi_char, (flags),NULL,depth+1)
3764 S_join_exact(pTHX_ RExC_state_t *pRExC_state, regnode *scan,
3765 UV *min_subtract, bool *unfolded_multi_char,
3766 U32 flags,regnode *val, U32 depth)
3768 /* Merge several consecutive EXACTish nodes into one. */
3769 regnode *n = regnext(scan);
3771 regnode *next = scan + NODE_SZ_STR(scan);
3775 regnode *stop = scan;
3776 GET_RE_DEBUG_FLAGS_DECL;
3778 PERL_UNUSED_ARG(depth);
3781 PERL_ARGS_ASSERT_JOIN_EXACT;
3782 #ifndef EXPERIMENTAL_INPLACESCAN
3783 PERL_UNUSED_ARG(flags);
3784 PERL_UNUSED_ARG(val);
3786 DEBUG_PEEP("join",scan,depth);
3788 /* Look through the subsequent nodes in the chain. Skip NOTHING, merge
3789 * EXACT ones that are mergeable to the current one. */
3791 && (PL_regkind[OP(n)] == NOTHING
3792 || (stringok && OP(n) == OP(scan)))
3794 && NEXT_OFF(scan) + NEXT_OFF(n) < I16_MAX)
3797 if (OP(n) == TAIL || n > next)
3799 if (PL_regkind[OP(n)] == NOTHING) {
3800 DEBUG_PEEP("skip:",n,depth);
3801 NEXT_OFF(scan) += NEXT_OFF(n);
3802 next = n + NODE_STEP_REGNODE;
3809 else if (stringok) {
3810 const unsigned int oldl = STR_LEN(scan);
3811 regnode * const nnext = regnext(n);
3813 /* XXX I (khw) kind of doubt that this works on platforms (should
3814 * Perl ever run on one) where U8_MAX is above 255 because of lots
3815 * of other assumptions */
3816 /* Don't join if the sum can't fit into a single node */
3817 if (oldl + STR_LEN(n) > U8_MAX)
3820 DEBUG_PEEP("merg",n,depth);
3823 NEXT_OFF(scan) += NEXT_OFF(n);
3824 STR_LEN(scan) += STR_LEN(n);
3825 next = n + NODE_SZ_STR(n);
3826 /* Now we can overwrite *n : */
3827 Move(STRING(n), STRING(scan) + oldl, STR_LEN(n), char);
3835 #ifdef EXPERIMENTAL_INPLACESCAN
3836 if (flags && !NEXT_OFF(n)) {
3837 DEBUG_PEEP("atch", val, depth);
3838 if (reg_off_by_arg[OP(n)]) {
3839 ARG_SET(n, val - n);
3842 NEXT_OFF(n) = val - n;
3850 *unfolded_multi_char = FALSE;
3852 /* Here, all the adjacent mergeable EXACTish nodes have been merged. We
3853 * can now analyze for sequences of problematic code points. (Prior to
3854 * this final joining, sequences could have been split over boundaries, and
3855 * hence missed). The sequences only happen in folding, hence for any
3856 * non-EXACT EXACTish node */
3857 if (OP(scan) != EXACT && OP(scan) != EXACTL) {
3858 U8* s0 = (U8*) STRING(scan);
3860 U8* s_end = s0 + STR_LEN(scan);
3862 int total_count_delta = 0; /* Total delta number of characters that
3863 multi-char folds expand to */
3865 /* One pass is made over the node's string looking for all the
3866 * possibilities. To avoid some tests in the loop, there are two main
3867 * cases, for UTF-8 patterns (which can't have EXACTF nodes) and
3872 if (OP(scan) == EXACTFL) {
3875 /* An EXACTFL node would already have been changed to another
3876 * node type unless there is at least one character in it that
3877 * is problematic; likely a character whose fold definition
3878 * won't be known until runtime, and so has yet to be folded.
3879 * For all but the UTF-8 locale, folds are 1-1 in length, but
3880 * to handle the UTF-8 case, we need to create a temporary
3881 * folded copy using UTF-8 locale rules in order to analyze it.
3882 * This is because our macros that look to see if a sequence is
3883 * a multi-char fold assume everything is folded (otherwise the
3884 * tests in those macros would be too complicated and slow).
3885 * Note that here, the non-problematic folds will have already
3886 * been done, so we can just copy such characters. We actually
3887 * don't completely fold the EXACTFL string. We skip the
3888 * unfolded multi-char folds, as that would just create work
3889 * below to figure out the size they already are */
3891 Newx(folded, UTF8_MAX_FOLD_CHAR_EXPAND * STR_LEN(scan) + 1, U8);
3894 STRLEN s_len = UTF8SKIP(s);
3895 if (! is_PROBLEMATIC_LOCALE_FOLD_utf8(s)) {
3896 Copy(s, d, s_len, U8);
3899 else if (is_FOLDS_TO_MULTI_utf8(s)) {
3900 *unfolded_multi_char = TRUE;
3901 Copy(s, d, s_len, U8);
3904 else if (isASCII(*s)) {
3905 *(d++) = toFOLD(*s);
3909 _toFOLD_utf8_flags(s, s_end, d, &len, FOLD_FLAGS_FULL);
3915 /* Point the remainder of the routine to look at our temporary
3919 } /* End of creating folded copy of EXACTFL string */
3921 /* Examine the string for a multi-character fold sequence. UTF-8
3922 * patterns have all characters pre-folded by the time this code is
3924 while (s < s_end - 1) /* Can stop 1 before the end, as minimum
3925 length sequence we are looking for is 2 */
3927 int count = 0; /* How many characters in a multi-char fold */
3928 int len = is_MULTI_CHAR_FOLD_utf8_safe(s, s_end);
3929 if (! len) { /* Not a multi-char fold: get next char */
3934 /* Nodes with 'ss' require special handling, except for
3935 * EXACTFA-ish for which there is no multi-char fold to this */
3936 if (len == 2 && *s == 's' && *(s+1) == 's'
3937 && OP(scan) != EXACTFA
3938 && OP(scan) != EXACTFA_NO_TRIE)
3941 if (OP(scan) != EXACTFL) {
3942 OP(scan) = EXACTFU_SS;
3946 else { /* Here is a generic multi-char fold. */
3947 U8* multi_end = s + len;
3949 /* Count how many characters are in it. In the case of
3950 * /aa, no folds which contain ASCII code points are
3951 * allowed, so check for those, and skip if found. */
3952 if (OP(scan) != EXACTFA && OP(scan) != EXACTFA_NO_TRIE) {
3953 count = utf8_length(s, multi_end);
3957 while (s < multi_end) {
3960 goto next_iteration;
3970 /* The delta is how long the sequence is minus 1 (1 is how long
3971 * the character that folds to the sequence is) */
3972 total_count_delta += count - 1;
3976 /* We created a temporary folded copy of the string in EXACTFL
3977 * nodes. Therefore we need to be sure it doesn't go below zero,
3978 * as the real string could be shorter */
3979 if (OP(scan) == EXACTFL) {
3980 int total_chars = utf8_length((U8*) STRING(scan),
3981 (U8*) STRING(scan) + STR_LEN(scan));
3982 if (total_count_delta > total_chars) {
3983 total_count_delta = total_chars;
3987 *min_subtract += total_count_delta;
3990 else if (OP(scan) == EXACTFA) {
3992 /* Non-UTF-8 pattern, EXACTFA node. There can't be a multi-char
3993 * fold to the ASCII range (and there are no existing ones in the
3994 * upper latin1 range). But, as outlined in the comments preceding
3995 * this function, we need to flag any occurrences of the sharp s.
3996 * This character forbids trie formation (because of added
3998 #if UNICODE_MAJOR_VERSION > 3 /* no multifolds in early Unicode */ \
3999 || (UNICODE_MAJOR_VERSION == 3 && ( UNICODE_DOT_VERSION > 0) \
4000 || UNICODE_DOT_DOT_VERSION > 0)
4002 if (*s == LATIN_SMALL_LETTER_SHARP_S) {
4003 OP(scan) = EXACTFA_NO_TRIE;
4004 *unfolded_multi_char = TRUE;
4012 /* Non-UTF-8 pattern, not EXACTFA node. Look for the multi-char
4013 * folds that are all Latin1. As explained in the comments
4014 * preceding this function, we look also for the sharp s in EXACTF
4015 * and EXACTFL nodes; it can be in the final position. Otherwise
4016 * we can stop looking 1 byte earlier because have to find at least
4017 * two characters for a multi-fold */
4018 const U8* upper = (OP(scan) == EXACTF || OP(scan) == EXACTFL)
4023 int len = is_MULTI_CHAR_FOLD_latin1_safe(s, s_end);
4024 if (! len) { /* Not a multi-char fold. */
4025 if (*s == LATIN_SMALL_LETTER_SHARP_S
4026 && (OP(scan) == EXACTF || OP(scan) == EXACTFL))
4028 *unfolded_multi_char = TRUE;
4035 && isALPHA_FOLD_EQ(*s, 's')
4036 && isALPHA_FOLD_EQ(*(s+1), 's'))
4039 /* EXACTF nodes need to know that the minimum length
4040 * changed so that a sharp s in the string can match this
4041 * ss in the pattern, but they remain EXACTF nodes, as they
4042 * won't match this unless the target string is is UTF-8,
4043 * which we don't know until runtime. EXACTFL nodes can't
4044 * transform into EXACTFU nodes */
4045 if (OP(scan) != EXACTF && OP(scan) != EXACTFL) {
4046 OP(scan) = EXACTFU_SS;
4050 *min_subtract += len - 1;
4058 /* Allow dumping but overwriting the collection of skipped
4059 * ops and/or strings with fake optimized ops */
4060 n = scan + NODE_SZ_STR(scan);
4068 DEBUG_OPTIMISE_r(if (merged){DEBUG_PEEP("finl",scan,depth)});
4072 /* REx optimizer. Converts nodes into quicker variants "in place".
4073 Finds fixed substrings. */
4075 /* Stops at toplevel WHILEM as well as at "last". At end *scanp is set
4076 to the position after last scanned or to NULL. */
4078 #define INIT_AND_WITHP \
4079 assert(!and_withp); \
4080 Newx(and_withp,1, regnode_ssc); \
4081 SAVEFREEPV(and_withp)
4085 S_unwind_scan_frames(pTHX_ const void *p)
4087 scan_frame *f= (scan_frame *)p;
4089 scan_frame *n= f->next_frame;
4097 S_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp,
4098 SSize_t *minlenp, SSize_t *deltap,
4103 regnode_ssc *and_withp,
4104 U32 flags, U32 depth)
4105 /* scanp: Start here (read-write). */
4106 /* deltap: Write maxlen-minlen here. */
4107 /* last: Stop before this one. */
4108 /* data: string data about the pattern */
4109 /* stopparen: treat close N as END */
4110 /* recursed: which subroutines have we recursed into */
4111 /* and_withp: Valid if flags & SCF_DO_STCLASS_OR */
4113 /* There must be at least this number of characters to match */
4116 regnode *scan = *scanp, *next;
4118 int is_inf = (flags & SCF_DO_SUBSTR) && (data->flags & SF_IS_INF);
4119 int is_inf_internal = 0; /* The studied chunk is infinite */
4120 I32 is_par = OP(scan) == OPEN ? ARG(scan) : 0;
4121 scan_data_t data_fake;
4122 SV *re_trie_maxbuff = NULL;
4123 regnode *first_non_open = scan;
4124 SSize_t stopmin = SSize_t_MAX;
4125 scan_frame *frame = NULL;
4126 GET_RE_DEBUG_FLAGS_DECL;
4128 PERL_ARGS_ASSERT_STUDY_CHUNK;
4129 RExC_study_started= 1;
4133 while (first_non_open && OP(first_non_open) == OPEN)
4134 first_non_open=regnext(first_non_open);
4140 RExC_study_chunk_recursed_count++;
4142 DEBUG_OPTIMISE_MORE_r(
4144 Perl_re_indentf( aTHX_ "study_chunk stopparen=%ld recursed_count=%lu depth=%lu recursed_depth=%lu scan=%p last=%p",
4145 depth, (long)stopparen,
4146 (unsigned long)RExC_study_chunk_recursed_count,
4147 (unsigned long)depth, (unsigned long)recursed_depth,
4150 if (recursed_depth) {
4153 for ( j = 0 ; j < recursed_depth ; j++ ) {
4154 for ( i = 0 ; i < (U32)RExC_npar ; i++ ) {
4156 PAREN_TEST(RExC_study_chunk_recursed +
4157 ( j * RExC_study_chunk_recursed_bytes), i )
4160 !PAREN_TEST(RExC_study_chunk_recursed +
4161 (( j - 1 ) * RExC_study_chunk_recursed_bytes), i)
4164 Perl_re_printf( aTHX_ " %d",(int)i);
4168 if ( j + 1 < recursed_depth ) {
4169 Perl_re_printf( aTHX_ ",");
4173 Perl_re_printf( aTHX_ "\n");
4176 while ( scan && OP(scan) != END && scan < last ){
4177 UV min_subtract = 0; /* How mmany chars to subtract from the minimum
4178 node length to get a real minimum (because
4179 the folded version may be shorter) */
4180 bool unfolded_multi_char = FALSE;
4181 /* Peephole optimizer: */
4182 DEBUG_STUDYDATA("Peep:", data, depth);
4183 DEBUG_PEEP("Peep", scan, depth);
4186 /* The reason we do this here is that we need to deal with things like
4187 * /(?:f)(?:o)(?:o)/ which cant be dealt with by the normal EXACT
4188 * parsing code, as each (?:..) is handled by a different invocation of
4191 JOIN_EXACT(scan,&min_subtract, &unfolded_multi_char, 0);
4193 /* Follow the next-chain of the current node and optimize
4194 away all the NOTHINGs from it. */
4195 if (OP(scan) != CURLYX) {
4196 const int max = (reg_off_by_arg[OP(scan)]
4198 /* I32 may be smaller than U16 on CRAYs! */
4199 : (I32_MAX < U16_MAX ? I32_MAX : U16_MAX));
4200 int off = (reg_off_by_arg[OP(scan)] ? ARG(scan) : NEXT_OFF(scan));
4204 /* Skip NOTHING and LONGJMP. */
4205 while ((n = regnext(n))
4206 && ((PL_regkind[OP(n)] == NOTHING && (noff = NEXT_OFF(n)))
4207 || ((OP(n) == LONGJMP) && (noff = ARG(n))))
4208 && off + noff < max)
4210 if (reg_off_by_arg[OP(scan)])
4213 NEXT_OFF(scan) = off;
4216 /* The principal pseudo-switch. Cannot be a switch, since we
4217 look into several different things. */
4218 if ( OP(scan) == DEFINEP ) {
4220 SSize_t deltanext = 0;
4221 SSize_t fake_last_close = 0;
4222 I32 f = SCF_IN_DEFINE;
4224 StructCopy(&zero_scan_data, &data_fake, scan_data_t);
4225 scan = regnext(scan);
4226 assert( OP(scan) == IFTHEN );
4227 DEBUG_PEEP("expect IFTHEN", scan, depth);
4229 data_fake.last_closep= &fake_last_close;
4231 next = regnext(scan);
4232 scan = NEXTOPER(NEXTOPER(scan));
4233 DEBUG_PEEP("scan", scan, depth);
4234 DEBUG_PEEP("next", next, depth);
4236 /* we suppose the run is continuous, last=next...
4237 * NOTE we dont use the return here! */
4238 (void)study_chunk(pRExC_state, &scan, &minlen,
4239 &deltanext, next, &data_fake, stopparen,
4240 recursed_depth, NULL, f, depth+1);
4245 OP(scan) == BRANCH ||
4246 OP(scan) == BRANCHJ ||
4249 next = regnext(scan);
4252 /* The op(next)==code check below is to see if we
4253 * have "BRANCH-BRANCH", "BRANCHJ-BRANCHJ", "IFTHEN-IFTHEN"
4254 * IFTHEN is special as it might not appear in pairs.
4255 * Not sure whether BRANCH-BRANCHJ is possible, regardless
4256 * we dont handle it cleanly. */
4257 if (OP(next) == code || code == IFTHEN) {
4258 /* NOTE - There is similar code to this block below for
4259 * handling TRIE nodes on a re-study. If you change stuff here
4260 * check there too. */
4261 SSize_t max1 = 0, min1 = SSize_t_MAX, num = 0;
4263 regnode * const startbranch=scan;
4265 if (flags & SCF_DO_SUBSTR) {
4266 /* Cannot merge strings after this. */
4267 scan_commit(pRExC_state, data, minlenp, is_inf);
4270 if (flags & SCF_DO_STCLASS)
4271 ssc_init_zero(pRExC_state, &accum);
4273 while (OP(scan) == code) {
4274 SSize_t deltanext, minnext, fake;
4276 regnode_ssc this_class;
4278 DEBUG_PEEP("Branch", scan, depth);
4281 StructCopy(&zero_scan_data, &data_fake, scan_data_t);
4283 data_fake.whilem_c = data->whilem_c;
4284 data_fake.last_closep = data->last_closep;
4287 data_fake.last_closep = &fake;
4289 data_fake.pos_delta = delta;
4290 next = regnext(scan);
4292 scan = NEXTOPER(scan); /* everything */
4293 if (code != BRANCH) /* everything but BRANCH */
4294 scan = NEXTOPER(scan);
4296 if (flags & SCF_DO_STCLASS) {
4297 ssc_init(pRExC_state, &this_class);
4298 data_fake.start_class = &this_class;
4299 f = SCF_DO_STCLASS_AND;
4301 if (flags & SCF_WHILEM_VISITED_POS)
4302 f |= SCF_WHILEM_VISITED_POS;
4304 /* we suppose the run is continuous, last=next...*/
4305 minnext = study_chunk(pRExC_state, &scan, minlenp,
4306 &deltanext, next, &data_fake, stopparen,
4307 recursed_depth, NULL, f,depth+1);
4311 if (deltanext == SSize_t_MAX) {
4312 is_inf = is_inf_internal = 1;
4314 } else if (max1 < minnext + deltanext)
4315 max1 = minnext + deltanext;
4317 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
4319 if (data_fake.flags & SCF_SEEN_ACCEPT) {
4320 if ( stopmin > minnext)
4321 stopmin = min + min1;
4322 flags &= ~SCF_DO_SUBSTR;
4324 data->flags |= SCF_SEEN_ACCEPT;
4327 if (data_fake.flags & SF_HAS_EVAL)
4328 data->flags |= SF_HAS_EVAL;
4329 data->whilem_c = data_fake.whilem_c;
4331 if (flags & SCF_DO_STCLASS)
4332 ssc_or(pRExC_state, &accum, (regnode_charclass*)&this_class);
4334 if (code == IFTHEN && num < 2) /* Empty ELSE branch */
4336 if (flags & SCF_DO_SUBSTR) {
4337 data->pos_min += min1;
4338 if (data->pos_delta >= SSize_t_MAX - (max1 - min1))
4339 data->pos_delta = SSize_t_MAX;
4341 data->pos_delta += max1 - min1;
4342 if (max1 != min1 || is_inf)
4343 data->longest = &(data->longest_float);
4346 if (delta == SSize_t_MAX
4347 || SSize_t_MAX - delta - (max1 - min1) < 0)
4348 delta = SSize_t_MAX;
4350 delta += max1 - min1;
4351 if (flags & SCF_DO_STCLASS_OR) {
4352 ssc_or(pRExC_state, data->start_class, (regnode_charclass*) &accum);
4354 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
4355 flags &= ~SCF_DO_STCLASS;
4358 else if (flags & SCF_DO_STCLASS_AND) {
4360 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) &accum);
4361 flags &= ~SCF_DO_STCLASS;
4364 /* Switch to OR mode: cache the old value of
4365 * data->start_class */
4367 StructCopy(data->start_class, and_withp, regnode_ssc);
4368 flags &= ~SCF_DO_STCLASS_AND;
4369 StructCopy(&accum, data->start_class, regnode_ssc);
4370 flags |= SCF_DO_STCLASS_OR;
4374 if (PERL_ENABLE_TRIE_OPTIMISATION &&
4375 OP( startbranch ) == BRANCH )
4379 Assuming this was/is a branch we are dealing with: 'scan'
4380 now points at the item that follows the branch sequence,
4381 whatever it is. We now start at the beginning of the
4382 sequence and look for subsequences of
4388 which would be constructed from a pattern like
4391 If we can find such a subsequence we need to turn the first
4392 element into a trie and then add the subsequent branch exact
4393 strings to the trie.
4397 1. patterns where the whole set of branches can be
4400 2. patterns where only a subset can be converted.
4402 In case 1 we can replace the whole set with a single regop
4403 for the trie. In case 2 we need to keep the start and end
4406 'BRANCH EXACT; BRANCH EXACT; BRANCH X'
4407 becomes BRANCH TRIE; BRANCH X;
4409 There is an additional case, that being where there is a
4410 common prefix, which gets split out into an EXACT like node
4411 preceding the TRIE node.
4413 If x(1..n)==tail then we can do a simple trie, if not we make
4414 a "jump" trie, such that when we match the appropriate word
4415 we "jump" to the appropriate tail node. Essentially we turn
4416 a nested if into a case structure of sorts.
4421 if (!re_trie_maxbuff) {
4422 re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, 1);
4423 if (!SvIOK(re_trie_maxbuff))
4424 sv_setiv(re_trie_maxbuff, RE_TRIE_MAXBUF_INIT);
4426 if ( SvIV(re_trie_maxbuff)>=0 ) {
4428 regnode *first = (regnode *)NULL;
4429 regnode *last = (regnode *)NULL;
4430 regnode *tail = scan;
4434 /* var tail is used because there may be a TAIL
4435 regop in the way. Ie, the exacts will point to the
4436 thing following the TAIL, but the last branch will
4437 point at the TAIL. So we advance tail. If we
4438 have nested (?:) we may have to move through several
4442 while ( OP( tail ) == TAIL ) {
4443 /* this is the TAIL generated by (?:) */
4444 tail = regnext( tail );
4448 DEBUG_TRIE_COMPILE_r({
4449 regprop(RExC_rx, RExC_mysv, tail, NULL, pRExC_state);
4450 Perl_re_indentf( aTHX_ "%s %" UVuf ":%s\n",
4452 "Looking for TRIE'able sequences. Tail node is ",
4453 (UV)(tail - RExC_emit_start),
4454 SvPV_nolen_const( RExC_mysv )
4460 Step through the branches
4461 cur represents each branch,
4462 noper is the first thing to be matched as part
4464 noper_next is the regnext() of that node.
4466 We normally handle a case like this
4467 /FOO[xyz]|BAR[pqr]/ via a "jump trie" but we also
4468 support building with NOJUMPTRIE, which restricts
4469 the trie logic to structures like /FOO|BAR/.
4471 If noper is a trieable nodetype then the branch is
4472 a possible optimization target. If we are building
4473 under NOJUMPTRIE then we require that noper_next is
4474 the same as scan (our current position in the regex
4477 Once we have two or more consecutive such branches
4478 we can create a trie of the EXACT's contents and
4479 stitch it in place into the program.
4481 If the sequence represents all of the branches in
4482 the alternation we replace the entire thing with a
4485 Otherwise when it is a subsequence we need to
4486 stitch it in place and replace only the relevant
4487 branches. This means the first branch has to remain
4488 as it is used by the alternation logic, and its
4489 next pointer, and needs to be repointed at the item
4490 on the branch chain following the last branch we
4491 have optimized away.
4493 This could be either a BRANCH, in which case the
4494 subsequence is internal, or it could be the item
4495 following the branch sequence in which case the
4496 subsequence is at the end (which does not
4497 necessarily mean the first node is the start of the
4500 TRIE_TYPE(X) is a define which maps the optype to a
4504 ----------------+-----------
4508 EXACTFU_SS | EXACTFU
4511 EXACTFLU8 | EXACTFLU8
4515 #define TRIE_TYPE(X) ( ( NOTHING == (X) ) \
4517 : ( EXACT == (X) ) \
4519 : ( EXACTFU == (X) || EXACTFU_SS == (X) ) \
4521 : ( EXACTFA == (X) ) \
4523 : ( EXACTL == (X) ) \
4525 : ( EXACTFLU8 == (X) ) \
4529 /* dont use tail as the end marker for this traverse */
4530 for ( cur = startbranch ; cur != scan ; cur = regnext( cur ) ) {
4531 regnode * const noper = NEXTOPER( cur );
4532 U8 noper_type = OP( noper );
4533 U8 noper_trietype = TRIE_TYPE( noper_type );
4534 #if defined(DEBUGGING) || defined(NOJUMPTRIE)
4535 regnode * const noper_next = regnext( noper );
4536 U8 noper_next_type = (noper_next && noper_next < tail) ? OP(noper_next) : 0;
4537 U8 noper_next_trietype = (noper_next && noper_next < tail) ? TRIE_TYPE( noper_next_type ) :0;
4540 DEBUG_TRIE_COMPILE_r({
4541 regprop(RExC_rx, RExC_mysv, cur, NULL, pRExC_state);
4542 Perl_re_indentf( aTHX_ "- %d:%s (%d)",
4544 REG_NODE_NUM(cur), SvPV_nolen_const( RExC_mysv ), REG_NODE_NUM(cur) );
4546 regprop(RExC_rx, RExC_mysv, noper, NULL, pRExC_state);
4547 Perl_re_printf( aTHX_ " -> %d:%s",
4548 REG_NODE_NUM(noper), SvPV_nolen_const(RExC_mysv));
4551 regprop(RExC_rx, RExC_mysv, noper_next, NULL, pRExC_state);
4552 Perl_re_printf( aTHX_ "\t=> %d:%s\t",
4553 REG_NODE_NUM(noper_next), SvPV_nolen_const(RExC_mysv));
4555 Perl_re_printf( aTHX_ "(First==%d,Last==%d,Cur==%d,tt==%s,ntt==%s,nntt==%s)\n",
4556 REG_NODE_NUM(first), REG_NODE_NUM(last), REG_NODE_NUM(cur),
4557 PL_reg_name[trietype], PL_reg_name[noper_trietype], PL_reg_name[noper_next_trietype]
4561 /* Is noper a trieable nodetype that can be merged
4562 * with the current trie (if there is one)? */
4566 ( noper_trietype == NOTHING )
4567 || ( trietype == NOTHING )
4568 || ( trietype == noper_trietype )
4571 && noper_next >= tail
4575 /* Handle mergable triable node Either we are
4576 * the first node in a new trieable sequence,
4577 * in which case we do some bookkeeping,
4578 * otherwise we update the end pointer. */
4581 if ( noper_trietype == NOTHING ) {
4582 #if !defined(DEBUGGING) && !defined(NOJUMPTRIE)
4583 regnode * const noper_next = regnext( noper );
4584 U8 noper_next_type = (noper_next && noper_next < tail) ? OP(noper_next) : 0;
4585 U8 noper_next_trietype = noper_next_type ? TRIE_TYPE( noper_next_type ) :0;
4588 if ( noper_next_trietype ) {
4589 trietype = noper_next_trietype;
4590 } else if (noper_next_type) {
4591 /* a NOTHING regop is 1 regop wide.
4592 * We need at least two for a trie
4593 * so we can't merge this in */
4597 trietype = noper_trietype;
4600 if ( trietype == NOTHING )
4601 trietype = noper_trietype;
4606 } /* end handle mergable triable node */
4608 /* handle unmergable node -
4609 * noper may either be a triable node which can
4610 * not be tried together with the current trie,
4611 * or a non triable node */
4613 /* If last is set and trietype is not
4614 * NOTHING then we have found at least two
4615 * triable branch sequences in a row of a
4616 * similar trietype so we can turn them
4617 * into a trie. If/when we allow NOTHING to
4618 * start a trie sequence this condition
4619 * will be required, and it isn't expensive
4620 * so we leave it in for now. */
4621 if ( trietype && trietype != NOTHING )
4622 make_trie( pRExC_state,
4623 startbranch, first, cur, tail,
4624 count, trietype, depth+1 );
4625 last = NULL; /* note: we clear/update
4626 first, trietype etc below,
4627 so we dont do it here */
4631 && noper_next >= tail
4634 /* noper is triable, so we can start a new
4638 trietype = noper_trietype;
4640 /* if we already saw a first but the
4641 * current node is not triable then we have
4642 * to reset the first information. */
4647 } /* end handle unmergable node */
4648 } /* loop over branches */
4649 DEBUG_TRIE_COMPILE_r({
4650 regprop(RExC_rx, RExC_mysv, cur, NULL, pRExC_state);
4651 Perl_re_indentf( aTHX_ "- %s (%d) <SCAN FINISHED> ",
4652 depth+1, SvPV_nolen_const( RExC_mysv ),REG_NODE_NUM(cur));
4653 Perl_re_printf( aTHX_ "(First==%d, Last==%d, Cur==%d, tt==%s)\n",
4654 REG_NODE_NUM(first), REG_NODE_NUM(last), REG_NODE_NUM(cur),
4655 PL_reg_name[trietype]
4659 if ( last && trietype ) {
4660 if ( trietype != NOTHING ) {
4661 /* the last branch of the sequence was part of
4662 * a trie, so we have to construct it here
4663 * outside of the loop */
4664 made= make_trie( pRExC_state, startbranch,
4665 first, scan, tail, count,
4666 trietype, depth+1 );
4667 #ifdef TRIE_STUDY_OPT
4668 if ( ((made == MADE_EXACT_TRIE &&
4669 startbranch == first)
4670 || ( first_non_open == first )) &&
4672 flags |= SCF_TRIE_RESTUDY;
4673 if ( startbranch == first
4676 RExC_seen &=~REG_TOP_LEVEL_BRANCHES_SEEN;
4681 /* at this point we know whatever we have is a
4682 * NOTHING sequence/branch AND if 'startbranch'
4683 * is 'first' then we can turn the whole thing
4686 if ( startbranch == first ) {
4688 /* the entire thing is a NOTHING sequence,
4689 * something like this: (?:|) So we can
4690 * turn it into a plain NOTHING op. */
4691 DEBUG_TRIE_COMPILE_r({
4692 regprop(RExC_rx, RExC_mysv, cur, NULL, pRExC_state);
4693 Perl_re_indentf( aTHX_ "- %s (%d) <NOTHING BRANCH SEQUENCE>\n",
4695 SvPV_nolen_const( RExC_mysv ),REG_NODE_NUM(cur));
4698 OP(startbranch)= NOTHING;
4699 NEXT_OFF(startbranch)= tail - startbranch;
4700 for ( opt= startbranch + 1; opt < tail ; opt++ )
4704 } /* end if ( last) */
4705 } /* TRIE_MAXBUF is non zero */
4710 else if ( code == BRANCHJ ) { /* single branch is optimized. */
4711 scan = NEXTOPER(NEXTOPER(scan));
4712 } else /* single branch is optimized. */
4713 scan = NEXTOPER(scan);
4715 } else if (OP(scan) == SUSPEND || OP(scan) == GOSUB) {
4717 regnode *start = NULL;
4718 regnode *end = NULL;
4719 U32 my_recursed_depth= recursed_depth;
4721 if (OP(scan) != SUSPEND) { /* GOSUB */
4722 /* Do setup, note this code has side effects beyond
4723 * the rest of this block. Specifically setting
4724 * RExC_recurse[] must happen at least once during
4727 RExC_recurse[ARG2L(scan)] = scan;
4728 start = RExC_open_parens[paren];
4729 end = RExC_close_parens[paren];
4731 /* NOTE we MUST always execute the above code, even
4732 * if we do nothing with a GOSUB */
4734 ( flags & SCF_IN_DEFINE )
4737 (is_inf_internal || is_inf || (data && data->flags & SF_IS_INF))
4739 ( (flags & (SCF_DO_STCLASS | SCF_DO_SUBSTR)) == 0 )
4742 /* no need to do anything here if we are in a define. */
4743 /* or we are after some kind of infinite construct
4744 * so we can skip recursing into this item.
4745 * Since it is infinite we will not change the maxlen
4746 * or delta, and if we miss something that might raise
4747 * the minlen it will merely pessimise a little.
4749 * Iow /(?(DEFINE)(?<foo>foo|food))a+(?&foo)/
4750 * might result in a minlen of 1 and not of 4,
4751 * but this doesn't make us mismatch, just try a bit
4752 * harder than we should.
4754 scan= regnext(scan);
4761 !PAREN_TEST(RExC_study_chunk_recursed + ((recursed_depth-1) * RExC_study_chunk_recursed_bytes), paren)
4763 /* it is quite possible that there are more efficient ways
4764 * to do this. We maintain a bitmap per level of recursion
4765 * of which patterns we have entered so we can detect if a
4766 * pattern creates a possible infinite loop. When we
4767 * recurse down a level we copy the previous levels bitmap
4768 * down. When we are at recursion level 0 we zero the top
4769 * level bitmap. It would be nice to implement a different
4770 * more efficient way of doing this. In particular the top
4771 * level bitmap may be unnecessary.
4773 if (!recursed_depth) {
4774 Zero(RExC_study_chunk_recursed, RExC_study_chunk_recursed_bytes, U8);
4776 Copy(RExC_study_chunk_recursed + ((recursed_depth-1) * RExC_study_chunk_recursed_bytes),
4777 RExC_study_chunk_recursed + (recursed_depth * RExC_study_chunk_recursed_bytes),
4778 RExC_study_chunk_recursed_bytes, U8);
4780 /* we havent recursed into this paren yet, so recurse into it */
4781 DEBUG_STUDYDATA("gosub-set:", data,depth);
4782 PAREN_SET(RExC_study_chunk_recursed + (recursed_depth * RExC_study_chunk_recursed_bytes), paren);
4783 my_recursed_depth= recursed_depth + 1;
4785 DEBUG_STUDYDATA("gosub-inf:", data,depth);
4786 /* some form of infinite recursion, assume infinite length
4788 if (flags & SCF_DO_SUBSTR) {
4789 scan_commit(pRExC_state, data, minlenp, is_inf);
4790 data->longest = &(data->longest_float);
4792 is_inf = is_inf_internal = 1;
4793 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
4794 ssc_anything(data->start_class);
4795 flags &= ~SCF_DO_STCLASS;
4797 start= NULL; /* reset start so we dont recurse later on. */
4802 end = regnext(scan);
4805 scan_frame *newframe;
4807 if (!RExC_frame_last) {
4808 Newxz(newframe, 1, scan_frame);
4809 SAVEDESTRUCTOR_X(S_unwind_scan_frames, newframe);
4810 RExC_frame_head= newframe;
4812 } else if (!RExC_frame_last->next_frame) {
4813 Newxz(newframe,1,scan_frame);
4814 RExC_frame_last->next_frame= newframe;
4815 newframe->prev_frame= RExC_frame_last;
4818 newframe= RExC_frame_last->next_frame;
4820 RExC_frame_last= newframe;
4822 newframe->next_regnode = regnext(scan);
4823 newframe->last_regnode = last;
4824 newframe->stopparen = stopparen;
4825 newframe->prev_recursed_depth = recursed_depth;
4826 newframe->this_prev_frame= frame;
4828 DEBUG_STUDYDATA("frame-new:",data,depth);
4829 DEBUG_PEEP("fnew", scan, depth);
4836 recursed_depth= my_recursed_depth;
4841 else if (OP(scan) == EXACT || OP(scan) == EXACTL) {
4842 SSize_t l = STR_LEN(scan);
4845 const U8 * const s = (U8*)STRING(scan);
4846 uc = utf8_to_uvchr_buf(s, s + l, NULL);
4847 l = utf8_length(s, s + l);
4849 uc = *((U8*)STRING(scan));
4852 if (flags & SCF_DO_SUBSTR) { /* Update longest substr. */
4853 /* The code below prefers earlier match for fixed
4854 offset, later match for variable offset. */
4855 if (data->last_end == -1) { /* Update the start info. */
4856 data->last_start_min = data->pos_min;
4857 data->last_start_max = is_inf
4858 ? SSize_t_MAX : data->pos_min + data->pos_delta;
4860 sv_catpvn(data->last_found, STRING(scan), STR_LEN(scan));
4862 SvUTF8_on(data->last_found);
4864 SV * const sv = data->last_found;
4865 MAGIC * const mg = SvUTF8(sv) && SvMAGICAL(sv) ?
4866 mg_find(sv, PERL_MAGIC_utf8) : NULL;
4867 if (mg && mg->mg_len >= 0)
4868 mg->mg_len += utf8_length((U8*)STRING(scan),
4869 (U8*)STRING(scan)+STR_LEN(scan));
4871 data->last_end = data->pos_min + l;
4872 data->pos_min += l; /* As in the first entry. */
4873 data->flags &= ~SF_BEFORE_EOL;
4876 /* ANDing the code point leaves at most it, and not in locale, and
4877 * can't match null string */
4878 if (flags & SCF_DO_STCLASS_AND) {
4879 ssc_cp_and(data->start_class, uc);
4880 ANYOF_FLAGS(data->start_class) &= ~SSC_MATCHES_EMPTY_STRING;
4881 ssc_clear_locale(data->start_class);
4883 else if (flags & SCF_DO_STCLASS_OR) {
4884 ssc_add_cp(data->start_class, uc);
4885 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
4887 /* See commit msg 749e076fceedeb708a624933726e7989f2302f6a */
4888 ANYOF_FLAGS(data->start_class) &= ~SSC_MATCHES_EMPTY_STRING;
4890 flags &= ~SCF_DO_STCLASS;
4892 else if (PL_regkind[OP(scan)] == EXACT) {
4893 /* But OP != EXACT!, so is EXACTFish */
4894 SSize_t l = STR_LEN(scan);
4895 const U8 * s = (U8*)STRING(scan);
4897 /* Search for fixed substrings supports EXACT only. */
4898 if (flags & SCF_DO_SUBSTR) {
4900 scan_commit(pRExC_state, data, minlenp, is_inf);
4903 l = utf8_length(s, s + l);
4905 if (unfolded_multi_char) {
4906 RExC_seen |= REG_UNFOLDED_MULTI_SEEN;
4908 min += l - min_subtract;
4910 delta += min_subtract;
4911 if (flags & SCF_DO_SUBSTR) {
4912 data->pos_min += l - min_subtract;
4913 if (data->pos_min < 0) {
4916 data->pos_delta += min_subtract;
4918 data->longest = &(data->longest_float);
4922 if (flags & SCF_DO_STCLASS) {
4923 SV* EXACTF_invlist = _make_exactf_invlist(pRExC_state, scan);
4925 assert(EXACTF_invlist);
4926 if (flags & SCF_DO_STCLASS_AND) {
4927 if (OP(scan) != EXACTFL)
4928 ssc_clear_locale(data->start_class);
4929 ANYOF_FLAGS(data->start_class) &= ~SSC_MATCHES_EMPTY_STRING;
4930 ANYOF_POSIXL_ZERO(data->start_class);
4931 ssc_intersection(data->start_class, EXACTF_invlist, FALSE);
4933 else { /* SCF_DO_STCLASS_OR */
4934 ssc_union(data->start_class, EXACTF_invlist, FALSE);
4935 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
4937 /* See commit msg 749e076fceedeb708a624933726e7989f2302f6a */
4938 ANYOF_FLAGS(data->start_class) &= ~SSC_MATCHES_EMPTY_STRING;
4940 flags &= ~SCF_DO_STCLASS;
4941 SvREFCNT_dec(EXACTF_invlist);
4944 else if (REGNODE_VARIES(OP(scan))) {
4945 SSize_t mincount, maxcount, minnext, deltanext, pos_before = 0;
4946 I32 fl = 0, f = flags;
4947 regnode * const oscan = scan;
4948 regnode_ssc this_class;
4949 regnode_ssc *oclass = NULL;
4950 I32 next_is_eval = 0;
4952 switch (PL_regkind[OP(scan)]) {
4953 case WHILEM: /* End of (?:...)* . */
4954 scan = NEXTOPER(scan);
4957 if (flags & (SCF_DO_SUBSTR | SCF_DO_STCLASS)) {
4958 next = NEXTOPER(scan);
4959 if (OP(next) == EXACT
4960 || OP(next) == EXACTL
4961 || (flags & SCF_DO_STCLASS))
4964 maxcount = REG_INFTY;
4965 next = regnext(scan);
4966 scan = NEXTOPER(scan);
4970 if (flags & SCF_DO_SUBSTR)
4975 if (flags & SCF_DO_STCLASS) {
4977 maxcount = REG_INFTY;
4978 next = regnext(scan);
4979 scan = NEXTOPER(scan);
4982 if (flags & SCF_DO_SUBSTR) {
4983 scan_commit(pRExC_state, data, minlenp, is_inf);
4984 /* Cannot extend fixed substrings */
4985 data->longest = &(data->longest_float);
4987 is_inf = is_inf_internal = 1;
4988 scan = regnext(scan);
4989 goto optimize_curly_tail;
4991 if (stopparen>0 && (OP(scan)==CURLYN || OP(scan)==CURLYM)
4992 && (scan->flags == stopparen))
4997 mincount = ARG1(scan);
4998 maxcount = ARG2(scan);
5000 next = regnext(scan);
5001 if (OP(scan) == CURLYX) {
5002 I32 lp = (data ? *(data->last_closep) : 0);
5003 scan->flags = ((lp <= (I32)U8_MAX) ? (U8)lp : U8_MAX);
5005 scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
5006 next_is_eval = (OP(scan) == EVAL);
5008 if (flags & SCF_DO_SUBSTR) {
5010 scan_commit(pRExC_state, data, minlenp, is_inf);
5011 /* Cannot extend fixed substrings */
5012 pos_before = data->pos_min;
5016 data->flags &= ~(SF_HAS_PAR|SF_IN_PAR|SF_HAS_EVAL);
5018 data->flags |= SF_IS_INF;
5020 if (flags & SCF_DO_STCLASS) {
5021 ssc_init(pRExC_state, &this_class);
5022 oclass = data->start_class;
5023 data->start_class = &this_class;
5024 f |= SCF_DO_STCLASS_AND;
5025 f &= ~SCF_DO_STCLASS_OR;
5027 /* Exclude from super-linear cache processing any {n,m}
5028 regops for which the combination of input pos and regex
5029 pos is not enough information to determine if a match
5032 For example, in the regex /foo(bar\s*){4,8}baz/ with the
5033 regex pos at the \s*, the prospects for a match depend not
5034 only on the input position but also on how many (bar\s*)
5035 repeats into the {4,8} we are. */
5036 if ((mincount > 1) || (maxcount > 1 && maxcount != REG_INFTY))
5037 f &= ~SCF_WHILEM_VISITED_POS;
5039 /* This will finish on WHILEM, setting scan, or on NULL: */
5040 minnext = study_chunk(pRExC_state, &scan, minlenp, &deltanext,
5041 last, data, stopparen, recursed_depth, NULL,
5043 ? (f & ~SCF_DO_SUBSTR)
5047 if (flags & SCF_DO_STCLASS)
5048 data->start_class = oclass;
5049 if (mincount == 0 || minnext == 0) {
5050 if (flags & SCF_DO_STCLASS_OR) {
5051 ssc_or(pRExC_state, data->start_class, (regnode_charclass *) &this_class);
5053 else if (flags & SCF_DO_STCLASS_AND) {
5054 /* Switch to OR mode: cache the old value of
5055 * data->start_class */
5057 StructCopy(data->start_class, and_withp, regnode_ssc);
5058 flags &= ~SCF_DO_STCLASS_AND;
5059 StructCopy(&this_class, data->start_class, regnode_ssc);
5060 flags |= SCF_DO_STCLASS_OR;
5061 ANYOF_FLAGS(data->start_class)
5062 |= SSC_MATCHES_EMPTY_STRING;
5064 } else { /* Non-zero len */
5065 if (flags & SCF_DO_STCLASS_OR) {
5066 ssc_or(pRExC_state, data->start_class, (regnode_charclass *) &this_class);
5067 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
5069 else if (flags & SCF_DO_STCLASS_AND)
5070 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) &this_class);
5071 flags &= ~SCF_DO_STCLASS;
5073 if (!scan) /* It was not CURLYX, but CURLY. */
5075 if (!(flags & SCF_TRIE_DOING_RESTUDY)
5076 /* ? quantifier ok, except for (?{ ... }) */
5077 && (next_is_eval || !(mincount == 0 && maxcount == 1))
5078 && (minnext == 0) && (deltanext == 0)
5079 && data && !(data->flags & (SF_HAS_PAR|SF_IN_PAR))
5080 && maxcount <= REG_INFTY/3) /* Complement check for big
5083 /* Fatal warnings may leak the regexp without this: */
5084 SAVEFREESV(RExC_rx_sv);
5085 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
5086 "Quantifier unexpected on zero-length expression "
5087 "in regex m/%" UTF8f "/",
5088 UTF8fARG(UTF, RExC_precomp_end - RExC_precomp,
5090 (void)ReREFCNT_inc(RExC_rx_sv);
5093 min += minnext * mincount;
5094 is_inf_internal |= deltanext == SSize_t_MAX
5095 || (maxcount == REG_INFTY && minnext + deltanext > 0);
5096 is_inf |= is_inf_internal;
5098 delta = SSize_t_MAX;
5100 delta += (minnext + deltanext) * maxcount
5101 - minnext * mincount;
5103 /* Try powerful optimization CURLYX => CURLYN. */
5104 if ( OP(oscan) == CURLYX && data
5105 && data->flags & SF_IN_PAR
5106 && !(data->flags & SF_HAS_EVAL)
5107 && !deltanext && minnext == 1 ) {
5108 /* Try to optimize to CURLYN. */
5109 regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS;
5110 regnode * const nxt1 = nxt;
5117 if (!REGNODE_SIMPLE(OP(nxt))
5118 && !(PL_regkind[OP(nxt)] == EXACT
5119 && STR_LEN(nxt) == 1))
5125 if (OP(nxt) != CLOSE)
5127 if (RExC_open_parens) {
5128 RExC_open_parens[ARG(nxt1)]=oscan; /*open->CURLYM*/
5129 RExC_close_parens[ARG(nxt1)]=nxt+2; /*close->while*/
5131 /* Now we know that nxt2 is the only contents: */
5132 oscan->flags = (U8)ARG(nxt);
5134 OP(nxt1) = NOTHING; /* was OPEN. */
5137 OP(nxt1 + 1) = OPTIMIZED; /* was count. */
5138 NEXT_OFF(nxt1+ 1) = 0; /* just for consistency. */
5139 NEXT_OFF(nxt2) = 0; /* just for consistency with CURLY. */
5140 OP(nxt) = OPTIMIZED; /* was CLOSE. */
5141 OP(nxt + 1) = OPTIMIZED; /* was count. */
5142 NEXT_OFF(nxt+ 1) = 0; /* just for consistency. */
5147 /* Try optimization CURLYX => CURLYM. */
5148 if ( OP(oscan) == CURLYX && data
5149 && !(data->flags & SF_HAS_PAR)
5150 && !(data->flags & SF_HAS_EVAL)
5151 && !deltanext /* atom is fixed width */
5152 && minnext != 0 /* CURLYM can't handle zero width */
5154 /* Nor characters whose fold at run-time may be
5155 * multi-character */
5156 && ! (RExC_seen & REG_UNFOLDED_MULTI_SEEN)
5158 /* XXXX How to optimize if data == 0? */
5159 /* Optimize to a simpler form. */
5160 regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN */
5164 while ( (nxt2 = regnext(nxt)) /* skip over embedded stuff*/
5165 && (OP(nxt2) != WHILEM))
5167 OP(nxt2) = SUCCEED; /* Whas WHILEM */
5168 /* Need to optimize away parenths. */
5169 if ((data->flags & SF_IN_PAR) && OP(nxt) == CLOSE) {
5170 /* Set the parenth number. */
5171 regnode *nxt1 = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN*/
5173 oscan->flags = (U8)ARG(nxt);
5174 if (RExC_open_parens) {
5175 RExC_open_parens[ARG(nxt1)]=oscan; /*open->CURLYM*/
5176 RExC_close_parens[ARG(nxt1)]=nxt2+1; /*close->NOTHING*/
5178 OP(nxt1) = OPTIMIZED; /* was OPEN. */
5179 OP(nxt) = OPTIMIZED; /* was CLOSE. */
5182 OP(nxt1 + 1) = OPTIMIZED; /* was count. */
5183 OP(nxt + 1) = OPTIMIZED; /* was count. */
5184 NEXT_OFF(nxt1 + 1) = 0; /* just for consistency. */
5185 NEXT_OFF(nxt + 1) = 0; /* just for consistency. */
5188 while ( nxt1 && (OP(nxt1) != WHILEM)) {
5189 regnode *nnxt = regnext(nxt1);
5191 if (reg_off_by_arg[OP(nxt1)])
5192 ARG_SET(nxt1, nxt2 - nxt1);
5193 else if (nxt2 - nxt1 < U16_MAX)
5194 NEXT_OFF(nxt1) = nxt2 - nxt1;
5196 OP(nxt) = NOTHING; /* Cannot beautify */
5201 /* Optimize again: */
5202 study_chunk(pRExC_state, &nxt1, minlenp, &deltanext, nxt,
5203 NULL, stopparen, recursed_depth, NULL, 0,depth+1);
5208 else if ((OP(oscan) == CURLYX)
5209 && (flags & SCF_WHILEM_VISITED_POS)
5210 /* See the comment on a similar expression above.
5211 However, this time it's not a subexpression
5212 we care about, but the expression itself. */
5213 && (maxcount == REG_INFTY)
5215 /* This stays as CURLYX, we can put the count/of pair. */
5216 /* Find WHILEM (as in regexec.c) */
5217 regnode *nxt = oscan + NEXT_OFF(oscan);
5219 if (OP(PREVOPER(nxt)) == NOTHING) /* LONGJMP */
5221 nxt = PREVOPER(nxt);
5222 if (nxt->flags & 0xf) {
5223 /* we've already set whilem count on this node */
5224 } else if (++data->whilem_c < 16) {
5225 assert(data->whilem_c <= RExC_whilem_seen);
5226 nxt->flags = (U8)(data->whilem_c
5227 | (RExC_whilem_seen << 4)); /* On WHILEM */
5230 if (data && fl & (SF_HAS_PAR|SF_IN_PAR))
5232 if (flags & SCF_DO_SUBSTR) {
5233 SV *last_str = NULL;
5234 STRLEN last_chrs = 0;
5235 int counted = mincount != 0;
5237 if (data->last_end > 0 && mincount != 0) { /* Ends with a
5239 SSize_t b = pos_before >= data->last_start_min
5240 ? pos_before : data->last_start_min;
5242 const char * const s = SvPV_const(data->last_found, l);
5243 SSize_t old = b - data->last_start_min;
5246 old = utf8_hop((U8*)s, old) - (U8*)s;
5248 /* Get the added string: */
5249 last_str = newSVpvn_utf8(s + old, l, UTF);
5250 last_chrs = UTF ? utf8_length((U8*)(s + old),
5251 (U8*)(s + old + l)) : l;
5252 if (deltanext == 0 && pos_before == b) {
5253 /* What was added is a constant string */
5256 SvGROW(last_str, (mincount * l) + 1);
5257 repeatcpy(SvPVX(last_str) + l,
5258 SvPVX_const(last_str), l,
5260 SvCUR_set(last_str, SvCUR(last_str) * mincount);
5261 /* Add additional parts. */
5262 SvCUR_set(data->last_found,
5263 SvCUR(data->last_found) - l);
5264 sv_catsv(data->last_found, last_str);
5266 SV * sv = data->last_found;
5268 SvUTF8(sv) && SvMAGICAL(sv) ?
5269 mg_find(sv, PERL_MAGIC_utf8) : NULL;
5270 if (mg && mg->mg_len >= 0)
5271 mg->mg_len += last_chrs * (mincount-1);
5273 last_chrs *= mincount;
5274 data->last_end += l * (mincount - 1);
5277 /* start offset must point into the last copy */
5278 data->last_start_min += minnext * (mincount - 1);
5279 data->last_start_max =
5282 : data->last_start_max +
5283 (maxcount - 1) * (minnext + data->pos_delta);
5286 /* It is counted once already... */
5287 data->pos_min += minnext * (mincount - counted);
5289 Perl_re_printf( aTHX_ "counted=%" UVuf " deltanext=%" UVuf
5290 " SSize_t_MAX=%" UVuf " minnext=%" UVuf
5291 " maxcount=%" UVuf " mincount=%" UVuf "\n",
5292 (UV)counted, (UV)deltanext, (UV)SSize_t_MAX, (UV)minnext, (UV)maxcount,
5294 if (deltanext != SSize_t_MAX)
5295 Perl_re_printf( aTHX_ "LHS=%" UVuf " RHS=%" UVuf "\n",
5296 (UV)(-counted * deltanext + (minnext + deltanext) * maxcount
5297 - minnext * mincount), (UV)(SSize_t_MAX - data->pos_delta));
5299 if (deltanext == SSize_t_MAX
5300 || -counted * deltanext + (minnext + deltanext) * maxcount - minnext * mincount >= SSize_t_MAX - data->pos_delta)
5301 data->pos_delta = SSize_t_MAX;
5303 data->pos_delta += - counted * deltanext +
5304 (minnext + deltanext) * maxcount - minnext * mincount;
5305 if (mincount != maxcount) {
5306 /* Cannot extend fixed substrings found inside
5308 scan_commit(pRExC_state, data, minlenp, is_inf);
5309 if (mincount && last_str) {
5310 SV * const sv = data->last_found;
5311 MAGIC * const mg = SvUTF8(sv) && SvMAGICAL(sv) ?
5312 mg_find(sv, PERL_MAGIC_utf8) : NULL;
5316 sv_setsv(sv, last_str);
5317 data->last_end = data->pos_min;
5318 data->last_start_min = data->pos_min - last_chrs;
5319 data->last_start_max = is_inf
5321 : data->pos_min + data->pos_delta - last_chrs;
5323 data->longest = &(data->longest_float);
5325 SvREFCNT_dec(last_str);
5327 if (data && (fl & SF_HAS_EVAL))
5328 data->flags |= SF_HAS_EVAL;
5329 optimize_curly_tail:
5330 if (OP(oscan) != CURLYX) {
5331 while (PL_regkind[OP(next = regnext(oscan))] == NOTHING
5333 NEXT_OFF(oscan) += NEXT_OFF(next);
5339 Perl_croak(aTHX_ "panic: unexpected varying REx opcode %d",
5344 if (flags & SCF_DO_SUBSTR) {
5345 /* Cannot expect anything... */
5346 scan_commit(pRExC_state, data, minlenp, is_inf);
5347 data->longest = &(data->longest_float);
5349 is_inf = is_inf_internal = 1;
5350 if (flags & SCF_DO_STCLASS_OR) {
5351 if (OP(scan) == CLUMP) {
5352 /* Actually is any start char, but very few code points
5353 * aren't start characters */
5354 ssc_match_all_cp(data->start_class);
5357 ssc_anything(data->start_class);
5360 flags &= ~SCF_DO_STCLASS;
5364 else if (OP(scan) == LNBREAK) {
5365 if (flags & SCF_DO_STCLASS) {
5366 if (flags & SCF_DO_STCLASS_AND) {
5367 ssc_intersection(data->start_class,
5368 PL_XPosix_ptrs[_CC_VERTSPACE], FALSE);
5369 ssc_clear_locale(data->start_class);
5370 ANYOF_FLAGS(data->start_class)
5371 &= ~SSC_MATCHES_EMPTY_STRING;
5373 else if (flags & SCF_DO_STCLASS_OR) {
5374 ssc_union(data->start_class,
5375 PL_XPosix_ptrs[_CC_VERTSPACE],
5377 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
5379 /* See commit msg for
5380 * 749e076fceedeb708a624933726e7989f2302f6a */
5381 ANYOF_FLAGS(data->start_class)
5382 &= ~SSC_MATCHES_EMPTY_STRING;
5384 flags &= ~SCF_DO_STCLASS;
5387 if (delta != SSize_t_MAX)
5388 delta++; /* Because of the 2 char string cr-lf */
5389 if (flags & SCF_DO_SUBSTR) {
5390 /* Cannot expect anything... */
5391 scan_commit(pRExC_state, data, minlenp, is_inf);
5393 data->pos_delta += 1;
5394 data->longest = &(data->longest_float);
5397 else if (REGNODE_SIMPLE(OP(scan))) {
5399 if (flags & SCF_DO_SUBSTR) {
5400 scan_commit(pRExC_state, data, minlenp, is_inf);
5404 if (flags & SCF_DO_STCLASS) {
5406 SV* my_invlist = NULL;
5409 /* See commit msg 749e076fceedeb708a624933726e7989f2302f6a */
5410 ANYOF_FLAGS(data->start_class) &= ~SSC_MATCHES_EMPTY_STRING;
5412 /* Some of the logic below assumes that switching
5413 locale on will only add false positives. */
5418 Perl_croak(aTHX_ "panic: unexpected simple REx opcode %d",
5422 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
5423 ssc_match_all_cp(data->start_class);
5428 SV* REG_ANY_invlist = _new_invlist(2);
5429 REG_ANY_invlist = add_cp_to_invlist(REG_ANY_invlist,
5431 if (flags & SCF_DO_STCLASS_OR) {
5432 ssc_union(data->start_class,
5434 TRUE /* TRUE => invert, hence all but \n
5438 else if (flags & SCF_DO_STCLASS_AND) {
5439 ssc_intersection(data->start_class,
5441 TRUE /* TRUE => invert */
5443 ssc_clear_locale(data->start_class);
5445 SvREFCNT_dec_NN(REG_ANY_invlist);
5452 if (flags & SCF_DO_STCLASS_AND)
5453 ssc_and(pRExC_state, data->start_class,
5454 (regnode_charclass *) scan);
5456 ssc_or(pRExC_state, data->start_class,
5457 (regnode_charclass *) scan);
5465 namedclass = classnum_to_namedclass(FLAGS(scan)) + invert;
5466 if (flags & SCF_DO_STCLASS_AND) {
5467 bool was_there = cBOOL(
5468 ANYOF_POSIXL_TEST(data->start_class,
5470 ANYOF_POSIXL_ZERO(data->start_class);
5471 if (was_there) { /* Do an AND */
5472 ANYOF_POSIXL_SET(data->start_class, namedclass);
5474 /* No individual code points can now match */
5475 data->start_class->invlist
5476 = sv_2mortal(_new_invlist(0));
5479 int complement = namedclass + ((invert) ? -1 : 1);
5481 assert(flags & SCF_DO_STCLASS_OR);
5483 /* If the complement of this class was already there,
5484 * the result is that they match all code points,
5485 * (\d + \D == everything). Remove the classes from
5486 * future consideration. Locale is not relevant in
5488 if (ANYOF_POSIXL_TEST(data->start_class, complement)) {
5489 ssc_match_all_cp(data->start_class);
5490 ANYOF_POSIXL_CLEAR(data->start_class, namedclass);
5491 ANYOF_POSIXL_CLEAR(data->start_class, complement);
5493 else { /* The usual case; just add this class to the
5495 ANYOF_POSIXL_SET(data->start_class, namedclass);
5500 case NPOSIXA: /* For these, we always know the exact set of
5505 if (FLAGS(scan) == _CC_ASCII) {
5506 my_invlist = invlist_clone(PL_XPosix_ptrs[_CC_ASCII]);
5509 _invlist_intersection(PL_XPosix_ptrs[FLAGS(scan)],
5510 PL_XPosix_ptrs[_CC_ASCII],
5521 my_invlist = invlist_clone(PL_XPosix_ptrs[FLAGS(scan)]);
5523 /* NPOSIXD matches all upper Latin1 code points unless the
5524 * target string being matched is UTF-8, which is
5525 * unknowable until match time. Since we are going to
5526 * invert, we want to get rid of all of them so that the
5527 * inversion will match all */
5528 if (OP(scan) == NPOSIXD) {
5529 _invlist_subtract(my_invlist, PL_UpperLatin1,
5535 if (flags & SCF_DO_STCLASS_AND) {
5536 ssc_intersection(data->start_class, my_invlist, invert);
5537 ssc_clear_locale(data->start_class);
5540 assert(flags & SCF_DO_STCLASS_OR);
5541 ssc_union(data->start_class, my_invlist, invert);
5543 SvREFCNT_dec(my_invlist);
5545 if (flags & SCF_DO_STCLASS_OR)
5546 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
5547 flags &= ~SCF_DO_STCLASS;
5550 else if (PL_regkind[OP(scan)] == EOL && flags & SCF_DO_SUBSTR) {
5551 data->flags |= (OP(scan) == MEOL
5554 scan_commit(pRExC_state, data, minlenp, is_inf);
5557 else if ( PL_regkind[OP(scan)] == BRANCHJ
5558 /* Lookbehind, or need to calculate parens/evals/stclass: */
5559 && (scan->flags || data || (flags & SCF_DO_STCLASS))
5560 && (OP(scan) == IFMATCH || OP(scan) == UNLESSM))
5562 if ( !PERL_ENABLE_POSITIVE_ASSERTION_STUDY
5563 || OP(scan) == UNLESSM )
5565 /* Negative Lookahead/lookbehind
5566 In this case we can't do fixed string optimisation.
5569 SSize_t deltanext, minnext, fake = 0;
5574 StructCopy(&zero_scan_data, &data_fake, scan_data_t);
5576 data_fake.whilem_c = data->whilem_c;
5577 data_fake.last_closep = data->last_closep;
5580 data_fake.last_closep = &fake;
5581 data_fake.pos_delta = delta;
5582 if ( flags & SCF_DO_STCLASS && !scan->flags
5583 && OP(scan) == IFMATCH ) { /* Lookahead */
5584 ssc_init(pRExC_state, &intrnl);
5585 data_fake.start_class = &intrnl;
5586 f |= SCF_DO_STCLASS_AND;
5588 if (flags & SCF_WHILEM_VISITED_POS)
5589 f |= SCF_WHILEM_VISITED_POS;
5590 next = regnext(scan);
5591 nscan = NEXTOPER(NEXTOPER(scan));
5592 minnext = study_chunk(pRExC_state, &nscan, minlenp, &deltanext,
5593 last, &data_fake, stopparen,
5594 recursed_depth, NULL, f, depth+1);
5597 FAIL("Variable length lookbehind not implemented");
5599 else if (minnext > (I32)U8_MAX) {
5600 FAIL2("Lookbehind longer than %" UVuf " not implemented",
5603 scan->flags = (U8)minnext;
5606 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
5608 if (data_fake.flags & SF_HAS_EVAL)
5609 data->flags |= SF_HAS_EVAL;
5610 data->whilem_c = data_fake.whilem_c;
5612 if (f & SCF_DO_STCLASS_AND) {
5613 if (flags & SCF_DO_STCLASS_OR) {
5614 /* OR before, AND after: ideally we would recurse with
5615 * data_fake to get the AND applied by study of the
5616 * remainder of the pattern, and then derecurse;
5617 * *** HACK *** for now just treat as "no information".
5618 * See [perl #56690].
5620 ssc_init(pRExC_state, data->start_class);
5622 /* AND before and after: combine and continue. These
5623 * assertions are zero-length, so can match an EMPTY
5625 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) &intrnl);
5626 ANYOF_FLAGS(data->start_class)
5627 |= SSC_MATCHES_EMPTY_STRING;
5631 #if PERL_ENABLE_POSITIVE_ASSERTION_STUDY
5633 /* Positive Lookahead/lookbehind
5634 In this case we can do fixed string optimisation,
5635 but we must be careful about it. Note in the case of
5636 lookbehind the positions will be offset by the minimum
5637 length of the pattern, something we won't know about
5638 until after the recurse.
5640 SSize_t deltanext, fake = 0;
5644 /* We use SAVEFREEPV so that when the full compile
5645 is finished perl will clean up the allocated
5646 minlens when it's all done. This way we don't
5647 have to worry about freeing them when we know
5648 they wont be used, which would be a pain.
5651 Newx( minnextp, 1, SSize_t );
5652 SAVEFREEPV(minnextp);
5655 StructCopy(data, &data_fake, scan_data_t);
5656 if ((flags & SCF_DO_SUBSTR) && data->last_found) {
5659 scan_commit(pRExC_state, &data_fake, minlenp, is_inf);
5660 data_fake.last_found=newSVsv(data->last_found);
5664 data_fake.last_closep = &fake;
5665 data_fake.flags = 0;
5666 data_fake.pos_delta = delta;
5668 data_fake.flags |= SF_IS_INF;
5669 if ( flags & SCF_DO_STCLASS && !scan->flags
5670 && OP(scan) == IFMATCH ) { /* Lookahead */
5671 ssc_init(pRExC_state, &intrnl);
5672 data_fake.start_class = &intrnl;
5673 f |= SCF_DO_STCLASS_AND;
5675 if (flags & SCF_WHILEM_VISITED_POS)
5676 f |= SCF_WHILEM_VISITED_POS;
5677 next = regnext(scan);
5678 nscan = NEXTOPER(NEXTOPER(scan));
5680 *minnextp = study_chunk(pRExC_state, &nscan, minnextp,
5681 &deltanext, last, &data_fake,
5682 stopparen, recursed_depth, NULL,
5686 FAIL("Variable length lookbehind not implemented");
5688 else if (*minnextp > (I32)U8_MAX) {
5689 FAIL2("Lookbehind longer than %" UVuf " not implemented",
5692 scan->flags = (U8)*minnextp;
5697 if (f & SCF_DO_STCLASS_AND) {
5698 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) &intrnl);
5699 ANYOF_FLAGS(data->start_class) |= SSC_MATCHES_EMPTY_STRING;
5702 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
5704 if (data_fake.flags & SF_HAS_EVAL)
5705 data->flags |= SF_HAS_EVAL;
5706 data->whilem_c = data_fake.whilem_c;
5707 if ((flags & SCF_DO_SUBSTR) && data_fake.last_found) {
5708 if (RExC_rx->minlen<*minnextp)
5709 RExC_rx->minlen=*minnextp;
5710 scan_commit(pRExC_state, &data_fake, minnextp, is_inf);
5711 SvREFCNT_dec_NN(data_fake.last_found);
5713 if ( data_fake.minlen_fixed != minlenp )
5715 data->offset_fixed= data_fake.offset_fixed;
5716 data->minlen_fixed= data_fake.minlen_fixed;
5717 data->lookbehind_fixed+= scan->flags;
5719 if ( data_fake.minlen_float != minlenp )
5721 data->minlen_float= data_fake.minlen_float;
5722 data->offset_float_min=data_fake.offset_float_min;
5723 data->offset_float_max=data_fake.offset_float_max;
5724 data->lookbehind_float+= scan->flags;
5731 else if (OP(scan) == OPEN) {
5732 if (stopparen != (I32)ARG(scan))
5735 else if (OP(scan) == CLOSE) {
5736 if (stopparen == (I32)ARG(scan)) {
5739 if ((I32)ARG(scan) == is_par) {
5740 next = regnext(scan);
5742 if ( next && (OP(next) != WHILEM) && next < last)
5743 is_par = 0; /* Disable optimization */
5746 *(data->last_closep) = ARG(scan);
5748 else if (OP(scan) == EVAL) {
5750 data->flags |= SF_HAS_EVAL;
5752 else if ( PL_regkind[OP(scan)] == ENDLIKE ) {
5753 if (flags & SCF_DO_SUBSTR) {
5754 scan_commit(pRExC_state, data, minlenp, is_inf);
5755 flags &= ~SCF_DO_SUBSTR;
5757 if (data && OP(scan)==ACCEPT) {
5758 data->flags |= SCF_SEEN_ACCEPT;
5763 else if (OP(scan) == LOGICAL && scan->flags == 2) /* Embedded follows */
5765 if (flags & SCF_DO_SUBSTR) {
5766 scan_commit(pRExC_state, data, minlenp, is_inf);
5767 data->longest = &(data->longest_float);
5769 is_inf = is_inf_internal = 1;
5770 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
5771 ssc_anything(data->start_class);
5772 flags &= ~SCF_DO_STCLASS;
5774 else if (OP(scan) == GPOS) {
5775 if (!(RExC_rx->intflags & PREGf_GPOS_FLOAT) &&
5776 !(delta || is_inf || (data && data->pos_delta)))
5778 if (!(RExC_rx->intflags & PREGf_ANCH) && (flags & SCF_DO_SUBSTR))
5779 RExC_rx->intflags |= PREGf_ANCH_GPOS;
5780 if (RExC_rx->gofs < (STRLEN)min)
5781 RExC_rx->gofs = min;
5783 RExC_rx->intflags |= PREGf_GPOS_FLOAT;
5787 #ifdef TRIE_STUDY_OPT
5788 #ifdef FULL_TRIE_STUDY
5789 else if (PL_regkind[OP(scan)] == TRIE) {
5790 /* NOTE - There is similar code to this block above for handling
5791 BRANCH nodes on the initial study. If you change stuff here
5793 regnode *trie_node= scan;
5794 regnode *tail= regnext(scan);
5795 reg_trie_data *trie = (reg_trie_data*)RExC_rxi->data->data[ ARG(scan) ];
5796 SSize_t max1 = 0, min1 = SSize_t_MAX;
5799 if (flags & SCF_DO_SUBSTR) { /* XXXX Add !SUSPEND? */
5800 /* Cannot merge strings after this. */
5801 scan_commit(pRExC_state, data, minlenp, is_inf);
5803 if (flags & SCF_DO_STCLASS)
5804 ssc_init_zero(pRExC_state, &accum);
5810 const regnode *nextbranch= NULL;
5813 for ( word=1 ; word <= trie->wordcount ; word++)
5815 SSize_t deltanext=0, minnext=0, f = 0, fake;
5816 regnode_ssc this_class;
5818 StructCopy(&zero_scan_data, &data_fake, scan_data_t);
5820 data_fake.whilem_c = data->whilem_c;
5821 data_fake.last_closep = data->last_closep;
5824 data_fake.last_closep = &fake;
5825 data_fake.pos_delta = delta;
5826 if (flags & SCF_DO_STCLASS) {
5827 ssc_init(pRExC_state, &this_class);
5828 data_fake.start_class = &this_class;
5829 f = SCF_DO_STCLASS_AND;
5831 if (flags & SCF_WHILEM_VISITED_POS)
5832 f |= SCF_WHILEM_VISITED_POS;
5834 if (trie->jump[word]) {
5836 nextbranch = trie_node + trie->jump[0];
5837 scan= trie_node + trie->jump[word];
5838 /* We go from the jump point to the branch that follows
5839 it. Note this means we need the vestigal unused
5840 branches even though they arent otherwise used. */
5841 minnext = study_chunk(pRExC_state, &scan, minlenp,
5842 &deltanext, (regnode *)nextbranch, &data_fake,
5843 stopparen, recursed_depth, NULL, f,depth+1);
5845 if (nextbranch && PL_regkind[OP(nextbranch)]==BRANCH)
5846 nextbranch= regnext((regnode*)nextbranch);
5848 if (min1 > (SSize_t)(minnext + trie->minlen))
5849 min1 = minnext + trie->minlen;
5850 if (deltanext == SSize_t_MAX) {
5851 is_inf = is_inf_internal = 1;
5853 } else if (max1 < (SSize_t)(minnext + deltanext + trie->maxlen))
5854 max1 = minnext + deltanext + trie->maxlen;
5856 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
5858 if (data_fake.flags & SCF_SEEN_ACCEPT) {
5859 if ( stopmin > min + min1)
5860 stopmin = min + min1;
5861 flags &= ~SCF_DO_SUBSTR;
5863 data->flags |= SCF_SEEN_ACCEPT;
5866 if (data_fake.flags & SF_HAS_EVAL)
5867 data->flags |= SF_HAS_EVAL;
5868 data->whilem_c = data_fake.whilem_c;
5870 if (flags & SCF_DO_STCLASS)
5871 ssc_or(pRExC_state, &accum, (regnode_charclass *) &this_class);
5874 if (flags & SCF_DO_SUBSTR) {
5875 data->pos_min += min1;
5876 data->pos_delta += max1 - min1;
5877 if (max1 != min1 || is_inf)
5878 data->longest = &(data->longest_float);
5881 if (delta != SSize_t_MAX)
5882 delta += max1 - min1;
5883 if (flags & SCF_DO_STCLASS_OR) {
5884 ssc_or(pRExC_state, data->start_class, (regnode_charclass *) &accum);
5886 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
5887 flags &= ~SCF_DO_STCLASS;
5890 else if (flags & SCF_DO_STCLASS_AND) {
5892 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) &accum);
5893 flags &= ~SCF_DO_STCLASS;
5896 /* Switch to OR mode: cache the old value of
5897 * data->start_class */
5899 StructCopy(data->start_class, and_withp, regnode_ssc);
5900 flags &= ~SCF_DO_STCLASS_AND;
5901 StructCopy(&accum, data->start_class, regnode_ssc);
5902 flags |= SCF_DO_STCLASS_OR;
5909 else if (PL_regkind[OP(scan)] == TRIE) {
5910 reg_trie_data *trie = (reg_trie_data*)RExC_rxi->data->data[ ARG(scan) ];
5913 min += trie->minlen;
5914 delta += (trie->maxlen - trie->minlen);
5915 flags &= ~SCF_DO_STCLASS; /* xxx */
5916 if (flags & SCF_DO_SUBSTR) {
5917 /* Cannot expect anything... */
5918 scan_commit(pRExC_state, data, minlenp, is_inf);
5919 data->pos_min += trie->minlen;
5920 data->pos_delta += (trie->maxlen - trie->minlen);
5921 if (trie->maxlen != trie->minlen)
5922 data->longest = &(data->longest_float);
5924 if (trie->jump) /* no more substrings -- for now /grr*/
5925 flags &= ~SCF_DO_SUBSTR;
5927 #endif /* old or new */
5928 #endif /* TRIE_STUDY_OPT */
5930 /* Else: zero-length, ignore. */
5931 scan = regnext(scan);
5936 /* we need to unwind recursion. */
5939 DEBUG_STUDYDATA("frame-end:",data,depth);
5940 DEBUG_PEEP("fend", scan, depth);
5942 /* restore previous context */
5943 last = frame->last_regnode;
5944 scan = frame->next_regnode;
5945 stopparen = frame->stopparen;
5946 recursed_depth = frame->prev_recursed_depth;
5948 RExC_frame_last = frame->prev_frame;
5949 frame = frame->this_prev_frame;
5950 goto fake_study_recurse;
5954 DEBUG_STUDYDATA("pre-fin:",data,depth);
5957 *deltap = is_inf_internal ? SSize_t_MAX : delta;
5959 if (flags & SCF_DO_SUBSTR && is_inf)
5960 data->pos_delta = SSize_t_MAX - data->pos_min;
5961 if (is_par > (I32)U8_MAX)
5963 if (is_par && pars==1 && data) {
5964 data->flags |= SF_IN_PAR;
5965 data->flags &= ~SF_HAS_PAR;
5967 else if (pars && data) {
5968 data->flags |= SF_HAS_PAR;
5969 data->flags &= ~SF_IN_PAR;
5971 if (flags & SCF_DO_STCLASS_OR)
5972 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
5973 if (flags & SCF_TRIE_RESTUDY)
5974 data->flags |= SCF_TRIE_RESTUDY;
5976 DEBUG_STUDYDATA("post-fin:",data,depth);
5979 SSize_t final_minlen= min < stopmin ? min : stopmin;
5981 if (!(RExC_seen & REG_UNBOUNDED_QUANTIFIER_SEEN)) {
5982 if (final_minlen > SSize_t_MAX - delta)
5983 RExC_maxlen = SSize_t_MAX;
5984 else if (RExC_maxlen < final_minlen + delta)
5985 RExC_maxlen = final_minlen + delta;
5987 return final_minlen;
5989 NOT_REACHED; /* NOTREACHED */
5993 S_add_data(RExC_state_t* const pRExC_state, const char* const s, const U32 n)
5995 U32 count = RExC_rxi->data ? RExC_rxi->data->count : 0;
5997 PERL_ARGS_ASSERT_ADD_DATA;
5999 Renewc(RExC_rxi->data,
6000 sizeof(*RExC_rxi->data) + sizeof(void*) * (count + n - 1),
6001 char, struct reg_data);
6003 Renew(RExC_rxi->data->what, count + n, U8);
6005 Newx(RExC_rxi->data->what, n, U8);
6006 RExC_rxi->data->count = count + n;
6007 Copy(s, RExC_rxi->data->what + count, n, U8);
6011 /*XXX: todo make this not included in a non debugging perl, but appears to be
6012 * used anyway there, in 'use re' */
6013 #ifndef PERL_IN_XSUB_RE
6015 Perl_reginitcolors(pTHX)
6017 const char * const s = PerlEnv_getenv("PERL_RE_COLORS");
6019 char *t = savepv(s);
6023 t = strchr(t, '\t');
6029 PL_colors[i] = t = (char *)"";
6034 PL_colors[i++] = (char *)"";
6041 #ifdef TRIE_STUDY_OPT
6042 #define CHECK_RESTUDY_GOTO_butfirst(dOsomething) \
6045 (data.flags & SCF_TRIE_RESTUDY) \
6053 #define CHECK_RESTUDY_GOTO_butfirst
6057 * pregcomp - compile a regular expression into internal code
6059 * Decides which engine's compiler to call based on the hint currently in
6063 #ifndef PERL_IN_XSUB_RE
6065 /* return the currently in-scope regex engine (or the default if none) */
6067 regexp_engine const *
6068 Perl_current_re_engine(pTHX)
6070 if (IN_PERL_COMPILETIME) {
6071 HV * const table = GvHV(PL_hintgv);
6074 if (!table || !(PL_hints & HINT_LOCALIZE_HH))
6075 return &PL_core_reg_engine;
6076 ptr = hv_fetchs(table, "regcomp", FALSE);
6077 if ( !(ptr && SvIOK(*ptr) && SvIV(*ptr)))
6078 return &PL_core_reg_engine;
6079 return INT2PTR(regexp_engine*,SvIV(*ptr));
6083 if (!PL_curcop->cop_hints_hash)
6084 return &PL_core_reg_engine;
6085 ptr = cop_hints_fetch_pvs(PL_curcop, "regcomp", 0);
6086 if ( !(ptr && SvIOK(ptr) && SvIV(ptr)))
6087 return &PL_core_reg_engine;
6088 return INT2PTR(regexp_engine*,SvIV(ptr));
6094 Perl_pregcomp(pTHX_ SV * const pattern, const U32 flags)
6096 regexp_engine const *eng = current_re_engine();
6097 GET_RE_DEBUG_FLAGS_DECL;
6099 PERL_ARGS_ASSERT_PREGCOMP;
6101 /* Dispatch a request to compile a regexp to correct regexp engine. */
6103 Perl_re_printf( aTHX_ "Using engine %" UVxf "\n",
6106 return CALLREGCOMP_ENG(eng, pattern, flags);
6110 /* public(ish) entry point for the perl core's own regex compiling code.
6111 * It's actually a wrapper for Perl_re_op_compile that only takes an SV
6112 * pattern rather than a list of OPs, and uses the internal engine rather
6113 * than the current one */
6116 Perl_re_compile(pTHX_ SV * const pattern, U32 rx_flags)
6118 SV *pat = pattern; /* defeat constness! */
6119 PERL_ARGS_ASSERT_RE_COMPILE;
6120 return Perl_re_op_compile(aTHX_ &pat, 1, NULL,
6121 #ifdef PERL_IN_XSUB_RE
6124 &PL_core_reg_engine,
6126 NULL, NULL, rx_flags, 0);
6131 S_free_codeblocks(pTHX_ struct reg_code_blocks *cbs)
6135 if (--cbs->refcnt > 0)
6137 for (n = 0; n < cbs->count; n++) {
6138 REGEXP *rx = cbs->cb[n].src_regex;
6139 cbs->cb[n].src_regex = NULL;
6147 static struct reg_code_blocks *
6148 S_alloc_code_blocks(pTHX_ int ncode)
6150 struct reg_code_blocks *cbs;
6151 Newx(cbs, 1, struct reg_code_blocks);
6154 SAVEDESTRUCTOR_X(S_free_codeblocks, cbs);
6156 Newx(cbs->cb, ncode, struct reg_code_block);
6163 /* upgrade pattern pat_p of length plen_p to UTF8, and if there are code
6164 * blocks, recalculate the indices. Update pat_p and plen_p in-place to
6165 * point to the realloced string and length.
6167 * This is essentially a copy of Perl_bytes_to_utf8() with the code index
6171 S_pat_upgrade_to_utf8(pTHX_ RExC_state_t * const pRExC_state,
6172 char **pat_p, STRLEN *plen_p, int num_code_blocks)
6174 U8 *const src = (U8*)*pat_p;
6179 GET_RE_DEBUG_FLAGS_DECL;
6181 DEBUG_PARSE_r(Perl_re_printf( aTHX_
6182 "UTF8 mismatch! Converting to utf8 for resizing and compile\n"));
6184 Newx(dst, *plen_p * 2 + 1, U8);
6187 while (s < *plen_p) {
6188 append_utf8_from_native_byte(src[s], &d);
6190 if (n < num_code_blocks) {
6191 assert(pRExC_state->code_blocks);
6192 if (!do_end && pRExC_state->code_blocks->cb[n].start == s) {
6193 pRExC_state->code_blocks->cb[n].start = d - dst - 1;
6194 assert(*(d - 1) == '(');
6197 else if (do_end && pRExC_state->code_blocks->cb[n].end == s) {
6198 pRExC_state->code_blocks->cb[n].end = d - dst - 1;
6199 assert(*(d - 1) == ')');
6208 *pat_p = (char*) dst;
6210 RExC_orig_utf8 = RExC_utf8 = 1;
6215 /* S_concat_pat(): concatenate a list of args to the pattern string pat,
6216 * while recording any code block indices, and handling overloading,
6217 * nested qr// objects etc. If pat is null, it will allocate a new
6218 * string, or just return the first arg, if there's only one.
6220 * Returns the malloced/updated pat.
6221 * patternp and pat_count is the array of SVs to be concatted;
6222 * oplist is the optional list of ops that generated the SVs;
6223 * recompile_p is a pointer to a boolean that will be set if
6224 * the regex will need to be recompiled.
6225 * delim, if non-null is an SV that will be inserted between each element
6229 S_concat_pat(pTHX_ RExC_state_t * const pRExC_state,
6230 SV *pat, SV ** const patternp, int pat_count,
6231 OP *oplist, bool *recompile_p, SV *delim)
6235 bool use_delim = FALSE;
6236 bool alloced = FALSE;
6238 /* if we know we have at least two args, create an empty string,
6239 * then concatenate args to that. For no args, return an empty string */
6240 if (!pat && pat_count != 1) {
6246 for (svp = patternp; svp < patternp + pat_count; svp++) {
6249 STRLEN orig_patlen = 0;
6251 SV *msv = use_delim ? delim : *svp;
6252 if (!msv) msv = &PL_sv_undef;
6254 /* if we've got a delimiter, we go round the loop twice for each
6255 * svp slot (except the last), using the delimiter the second
6264 if (SvTYPE(msv) == SVt_PVAV) {
6265 /* we've encountered an interpolated array within
6266 * the pattern, e.g. /...@a..../. Expand the list of elements,
6267 * then recursively append elements.
6268 * The code in this block is based on S_pushav() */
6270 AV *const av = (AV*)msv;
6271 const SSize_t maxarg = AvFILL(av) + 1;
6275 assert(oplist->op_type == OP_PADAV
6276 || oplist->op_type == OP_RV2AV);
6277 oplist = OpSIBLING(oplist);
6280 if (SvRMAGICAL(av)) {
6283 Newx(array, maxarg, SV*);
6285 for (i=0; i < maxarg; i++) {
6286 SV ** const svp = av_fetch(av, i, FALSE);
6287 array[i] = svp ? *svp : &PL_sv_undef;
6291 array = AvARRAY(av);
6293 pat = S_concat_pat(aTHX_ pRExC_state, pat,
6294 array, maxarg, NULL, recompile_p,
6296 GvSV((gv_fetchpvs("\"", GV_ADDMULTI, SVt_PV))));
6302 /* we make the assumption here that each op in the list of
6303 * op_siblings maps to one SV pushed onto the stack,
6304 * except for code blocks, with have both an OP_NULL and
6306 * This allows us to match up the list of SVs against the
6307 * list of OPs to find the next code block.
6309 * Note that PUSHMARK PADSV PADSV ..
6311 * PADRANGE PADSV PADSV ..
6312 * so the alignment still works. */
6315 if (oplist->op_type == OP_NULL
6316 && (oplist->op_flags & OPf_SPECIAL))
6318 assert(n < pRExC_state->code_blocks->count);
6319 pRExC_state->code_blocks->cb[n].start = pat ? SvCUR(pat) : 0;
6320 pRExC_state->code_blocks->cb[n].block = oplist;
6321 pRExC_state->code_blocks->cb[n].src_regex = NULL;
6324 oplist = OpSIBLING(oplist); /* skip CONST */
6327 oplist = OpSIBLING(oplist);;
6330 /* apply magic and QR overloading to arg */
6333 if (SvROK(msv) && SvAMAGIC(msv)) {
6334 SV *sv = AMG_CALLunary(msv, regexp_amg);
6338 if (SvTYPE(sv) != SVt_REGEXP)
6339 Perl_croak(aTHX_ "Overloaded qr did not return a REGEXP");
6344 /* try concatenation overload ... */
6345 if (pat && (SvAMAGIC(pat) || SvAMAGIC(msv)) &&
6346 (sv = amagic_call(pat, msv, concat_amg, AMGf_assign)))
6349 /* overloading involved: all bets are off over literal
6350 * code. Pretend we haven't seen it */
6352 pRExC_state->code_blocks->count -= n;
6356 /* ... or failing that, try "" overload */
6357 while (SvAMAGIC(msv)
6358 && (sv = AMG_CALLunary(msv, string_amg))
6362 && SvRV(msv) == SvRV(sv))
6367 if (SvROK(msv) && SvTYPE(SvRV(msv)) == SVt_REGEXP)
6371 /* this is a partially unrolled
6372 * sv_catsv_nomg(pat, msv);
6373 * that allows us to adjust code block indices if
6376 char *dst = SvPV_force_nomg(pat, dlen);
6378 if (SvUTF8(msv) && !SvUTF8(pat)) {
6379 S_pat_upgrade_to_utf8(aTHX_ pRExC_state, &dst, &dlen, n);
6380 sv_setpvn(pat, dst, dlen);
6383 sv_catsv_nomg(pat, msv);
6387 /* We have only one SV to process, but we need to verify
6388 * it is properly null terminated or we will fail asserts
6389 * later. In theory we probably shouldn't get such SV's,
6390 * but if we do we should handle it gracefully. */
6391 if ( SvTYPE(msv) != SVt_PV || (SvLEN(msv) > SvCUR(msv) && *(SvEND(msv)) == 0) ) {
6392 /* not a string, or a string with a trailing null */
6395 /* a string with no trailing null, we need to copy it
6396 * so it we have a trailing null */
6402 pRExC_state->code_blocks->cb[n-1].end = SvCUR(pat)-1;
6405 /* extract any code blocks within any embedded qr//'s */
6406 if (rx && SvTYPE(rx) == SVt_REGEXP
6407 && RX_ENGINE((REGEXP*)rx)->op_comp)
6410 RXi_GET_DECL(ReANY((REGEXP *)rx), ri);
6411 if (ri->code_blocks && ri->code_blocks->count) {
6413 /* the presence of an embedded qr// with code means
6414 * we should always recompile: the text of the
6415 * qr// may not have changed, but it may be a
6416 * different closure than last time */
6418 if (pRExC_state->code_blocks) {
6419 pRExC_state->code_blocks->count += ri->code_blocks->count;
6420 Renew(pRExC_state->code_blocks->cb,
6421 pRExC_state->code_blocks->count,
6422 struct reg_code_block);
6425 pRExC_state->code_blocks = S_alloc_code_blocks(aTHX_
6426 ri->code_blocks->count);
6428 for (i=0; i < ri->code_blocks->count; i++) {
6429 struct reg_code_block *src, *dst;
6430 STRLEN offset = orig_patlen
6431 + ReANY((REGEXP *)rx)->pre_prefix;
6432 assert(n < pRExC_state->code_blocks->count);
6433 src = &ri->code_blocks->cb[i];
6434 dst = &pRExC_state->code_blocks->cb[n];
6435 dst->start = src->start + offset;
6436 dst->end = src->end + offset;
6437 dst->block = src->block;
6438 dst->src_regex = (REGEXP*) SvREFCNT_inc( (SV*)
6447 /* avoid calling magic multiple times on a single element e.g. =~ $qr */
6456 /* see if there are any run-time code blocks in the pattern.
6457 * False positives are allowed */
6460 S_has_runtime_code(pTHX_ RExC_state_t * const pRExC_state,
6461 char *pat, STRLEN plen)
6466 PERL_UNUSED_CONTEXT;
6468 for (s = 0; s < plen; s++) {
6469 if ( pRExC_state->code_blocks
6470 && n < pRExC_state->code_blocks->count
6471 && s == pRExC_state->code_blocks->cb[n].start)
6473 s = pRExC_state->code_blocks->cb[n].end;
6477 /* TODO ideally should handle [..], (#..), /#.../x to reduce false
6479 if (pat[s] == '(' && s+2 <= plen && pat[s+1] == '?' &&
6481 || (s + 2 <= plen && pat[s+2] == '?' && pat[s+3] == '{'))
6488 /* Handle run-time code blocks. We will already have compiled any direct
6489 * or indirect literal code blocks. Now, take the pattern 'pat' and make a
6490 * copy of it, but with any literal code blocks blanked out and
6491 * appropriate chars escaped; then feed it into
6493 * eval "qr'modified_pattern'"
6497 * a\bc(?{"this was literal"})def'ghi\\jkl(?{"this is runtime"})mno
6501 * qr'a\\bc_______________________def\'ghi\\\\jkl(?{"this is runtime"})mno'
6503 * After eval_sv()-ing that, grab any new code blocks from the returned qr
6504 * and merge them with any code blocks of the original regexp.
6506 * If the pat is non-UTF8, while the evalled qr is UTF8, don't merge;
6507 * instead, just save the qr and return FALSE; this tells our caller that
6508 * the original pattern needs upgrading to utf8.
6512 S_compile_runtime_code(pTHX_ RExC_state_t * const pRExC_state,
6513 char *pat, STRLEN plen)
6517 GET_RE_DEBUG_FLAGS_DECL;
6519 if (pRExC_state->runtime_code_qr) {
6520 /* this is the second time we've been called; this should
6521 * only happen if the main pattern got upgraded to utf8
6522 * during compilation; re-use the qr we compiled first time
6523 * round (which should be utf8 too)
6525 qr = pRExC_state->runtime_code_qr;
6526 pRExC_state->runtime_code_qr = NULL;
6527 assert(RExC_utf8 && SvUTF8(qr));
6533 int newlen = plen + 7; /* allow for "qr''xx\0" extra chars */
6537 /* determine how many extra chars we need for ' and \ escaping */
6538 for (s = 0; s < plen; s++) {
6539 if (pat[s] == '\'' || pat[s] == '\\')
6543 Newx(newpat, newlen, char);
6545 *p++ = 'q'; *p++ = 'r'; *p++ = '\'';
6547 for (s = 0; s < plen; s++) {
6548 if ( pRExC_state->code_blocks
6549 && n < pRExC_state->code_blocks->count
6550 && s == pRExC_state->code_blocks->cb[n].start)
6552 /* blank out literal code block */
6553 assert(pat[s] == '(');
6554 while (s <= pRExC_state->code_blocks->cb[n].end) {
6562 if (pat[s] == '\'' || pat[s] == '\\')
6567 if (pRExC_state->pm_flags & RXf_PMf_EXTENDED) {
6569 if (pRExC_state->pm_flags & RXf_PMf_EXTENDED_MORE) {
6575 Perl_re_printf( aTHX_
6576 "%sre-parsing pattern for runtime code:%s %s\n",
6577 PL_colors[4],PL_colors[5],newpat);
6580 sv = newSVpvn_flags(newpat, p-newpat-1, RExC_utf8 ? SVf_UTF8 : 0);
6586 PUSHSTACKi(PERLSI_REQUIRE);
6587 /* G_RE_REPARSING causes the toker to collapse \\ into \ when
6588 * parsing qr''; normally only q'' does this. It also alters
6590 eval_sv(sv, G_SCALAR|G_RE_REPARSING);
6591 SvREFCNT_dec_NN(sv);
6596 SV * const errsv = ERRSV;
6597 if (SvTRUE_NN(errsv))
6598 /* use croak_sv ? */
6599 Perl_croak_nocontext("%" SVf, SVfARG(errsv));
6601 assert(SvROK(qr_ref));
6603 assert(SvTYPE(qr) == SVt_REGEXP && RX_ENGINE((REGEXP*)qr)->op_comp);
6604 /* the leaving below frees the tmp qr_ref.
6605 * Give qr a life of its own */
6613 if (!RExC_utf8 && SvUTF8(qr)) {
6614 /* first time through; the pattern got upgraded; save the
6615 * qr for the next time through */
6616 assert(!pRExC_state->runtime_code_qr);
6617 pRExC_state->runtime_code_qr = qr;
6622 /* extract any code blocks within the returned qr// */
6625 /* merge the main (r1) and run-time (r2) code blocks into one */
6627 RXi_GET_DECL(ReANY((REGEXP *)qr), r2);
6628 struct reg_code_block *new_block, *dst;
6629 RExC_state_t * const r1 = pRExC_state; /* convenient alias */
6633 if (!r2->code_blocks || !r2->code_blocks->count) /* we guessed wrong */
6635 SvREFCNT_dec_NN(qr);
6639 if (!r1->code_blocks)
6640 r1->code_blocks = S_alloc_code_blocks(aTHX_ 0);
6642 r1c = r1->code_blocks->count;
6643 r2c = r2->code_blocks->count;
6645 Newx(new_block, r1c + r2c, struct reg_code_block);
6649 while (i1 < r1c || i2 < r2c) {
6650 struct reg_code_block *src;
6654 src = &r2->code_blocks->cb[i2++];
6658 src = &r1->code_blocks->cb[i1++];
6659 else if ( r1->code_blocks->cb[i1].start
6660 < r2->code_blocks->cb[i2].start)
6662 src = &r1->code_blocks->cb[i1++];
6663 assert(src->end < r2->code_blocks->cb[i2].start);
6666 assert( r1->code_blocks->cb[i1].start
6667 > r2->code_blocks->cb[i2].start);
6668 src = &r2->code_blocks->cb[i2++];
6670 assert(src->end < r1->code_blocks->cb[i1].start);
6673 assert(pat[src->start] == '(');
6674 assert(pat[src->end] == ')');
6675 dst->start = src->start;
6676 dst->end = src->end;
6677 dst->block = src->block;
6678 dst->src_regex = is_qr ? (REGEXP*) SvREFCNT_inc( (SV*) qr)
6682 r1->code_blocks->count += r2c;
6683 Safefree(r1->code_blocks->cb);
6684 r1->code_blocks->cb = new_block;
6687 SvREFCNT_dec_NN(qr);
6693 S_setup_longest(pTHX_ RExC_state_t *pRExC_state, SV* sv_longest,
6694 SV** rx_utf8, SV** rx_substr, SSize_t* rx_end_shift,
6695 SSize_t lookbehind, SSize_t offset, SSize_t *minlen,
6696 STRLEN longest_length, bool eol, bool meol)
6698 /* This is the common code for setting up the floating and fixed length
6699 * string data extracted from Perl_re_op_compile() below. Returns a boolean
6700 * as to whether succeeded or not */
6705 if (! (longest_length
6706 || (eol /* Can't have SEOL and MULTI */
6707 && (! meol || (RExC_flags & RXf_PMf_MULTILINE)))
6709 /* See comments for join_exact for why REG_UNFOLDED_MULTI_SEEN */
6710 || (RExC_seen & REG_UNFOLDED_MULTI_SEEN))
6715 /* copy the information about the longest from the reg_scan_data
6716 over to the program. */
6717 if (SvUTF8(sv_longest)) {
6718 *rx_utf8 = sv_longest;
6721 *rx_substr = sv_longest;
6724 /* end_shift is how many chars that must be matched that
6725 follow this item. We calculate it ahead of time as once the
6726 lookbehind offset is added in we lose the ability to correctly
6728 ml = minlen ? *(minlen) : (SSize_t)longest_length;
6729 *rx_end_shift = ml - offset
6731 /* XXX SvTAIL is always false here - did you mean FBMcf_TAIL
6733 + (SvTAIL(sv_longest) != 0)
6737 t = (eol/* Can't have SEOL and MULTI */
6738 && (! meol || (RExC_flags & RXf_PMf_MULTILINE)));
6739 fbm_compile(sv_longest, t ? FBMcf_TAIL : 0);
6745 * Perl_re_op_compile - the perl internal RE engine's function to compile a
6746 * regular expression into internal code.
6747 * The pattern may be passed either as:
6748 * a list of SVs (patternp plus pat_count)
6749 * a list of OPs (expr)
6750 * If both are passed, the SV list is used, but the OP list indicates
6751 * which SVs are actually pre-compiled code blocks
6753 * The SVs in the list have magic and qr overloading applied to them (and
6754 * the list may be modified in-place with replacement SVs in the latter
6757 * If the pattern hasn't changed from old_re, then old_re will be
6760 * eng is the current engine. If that engine has an op_comp method, then
6761 * handle directly (i.e. we assume that op_comp was us); otherwise, just
6762 * do the initial concatenation of arguments and pass on to the external
6765 * If is_bare_re is not null, set it to a boolean indicating whether the
6766 * arg list reduced (after overloading) to a single bare regex which has
6767 * been returned (i.e. /$qr/).
6769 * orig_rx_flags contains RXf_* flags. See perlreapi.pod for more details.
6771 * pm_flags contains the PMf_* flags, typically based on those from the
6772 * pm_flags field of the related PMOP. Currently we're only interested in
6773 * PMf_HAS_CV, PMf_IS_QR, PMf_USE_RE_EVAL.
6775 * We can't allocate space until we know how big the compiled form will be,
6776 * but we can't compile it (and thus know how big it is) until we've got a
6777 * place to put the code. So we cheat: we compile it twice, once with code
6778 * generation turned off and size counting turned on, and once "for real".
6779 * This also means that we don't allocate space until we are sure that the
6780 * thing really will compile successfully, and we never have to move the
6781 * code and thus invalidate pointers into it. (Note that it has to be in
6782 * one piece because free() must be able to free it all.) [NB: not true in perl]
6784 * Beware that the optimization-preparation code in here knows about some
6785 * of the structure of the compiled regexp. [I'll say.]
6789 Perl_re_op_compile(pTHX_ SV ** const patternp, int pat_count,
6790 OP *expr, const regexp_engine* eng, REGEXP *old_re,
6791 bool *is_bare_re, U32 orig_rx_flags, U32 pm_flags)
6795 regexp_internal *ri;
6803 SV** new_patternp = patternp;
6805 /* these are all flags - maybe they should be turned
6806 * into a single int with different bit masks */
6807 I32 sawlookahead = 0;
6812 regex_charset initial_charset = get_regex_charset(orig_rx_flags);
6814 bool runtime_code = 0;
6816 RExC_state_t RExC_state;
6817 RExC_state_t * const pRExC_state = &RExC_state;
6818 #ifdef TRIE_STUDY_OPT
6820 RExC_state_t copyRExC_state;
6822 GET_RE_DEBUG_FLAGS_DECL;
6824 PERL_ARGS_ASSERT_RE_OP_COMPILE;
6826 DEBUG_r(if (!PL_colorset) reginitcolors());
6828 /* Initialize these here instead of as-needed, as is quick and avoids
6829 * having to test them each time otherwise */
6830 if (! PL_AboveLatin1) {
6832 char * dump_len_string;
6835 PL_AboveLatin1 = _new_invlist_C_array(AboveLatin1_invlist);
6836 PL_Latin1 = _new_invlist_C_array(Latin1_invlist);
6837 PL_UpperLatin1 = _new_invlist_C_array(UpperLatin1_invlist);
6838 PL_utf8_foldable = _new_invlist_C_array(_Perl_Any_Folds_invlist);
6839 PL_HasMultiCharFold =
6840 _new_invlist_C_array(_Perl_Folds_To_Multi_Char_invlist);
6842 /* This is calculated here, because the Perl program that generates the
6843 * static global ones doesn't currently have access to
6844 * NUM_ANYOF_CODE_POINTS */
6845 PL_InBitmap = _new_invlist(2);
6846 PL_InBitmap = _add_range_to_invlist(PL_InBitmap, 0,
6847 NUM_ANYOF_CODE_POINTS - 1);
6849 dump_len_string = PerlEnv_getenv("PERL_DUMP_RE_MAX_LEN");
6850 if ( ! dump_len_string
6851 || ! grok_atoUV(dump_len_string, (UV *)&PL_dump_re_max_len, NULL))
6853 PL_dump_re_max_len = 0;
6858 pRExC_state->warn_text = NULL;
6859 pRExC_state->code_blocks = NULL;
6862 *is_bare_re = FALSE;
6864 if (expr && (expr->op_type == OP_LIST ||
6865 (expr->op_type == OP_NULL && expr->op_targ == OP_LIST))) {
6866 /* allocate code_blocks if needed */
6870 for (o = cLISTOPx(expr)->op_first; o; o = OpSIBLING(o))
6871 if (o->op_type == OP_NULL && (o->op_flags & OPf_SPECIAL))
6872 ncode++; /* count of DO blocks */
6875 pRExC_state->code_blocks = S_alloc_code_blocks(aTHX_ ncode);
6879 /* compile-time pattern with just OP_CONSTs and DO blocks */
6884 /* find how many CONSTs there are */
6887 if (expr->op_type == OP_CONST)
6890 for (o = cLISTOPx(expr)->op_first; o; o = OpSIBLING(o)) {
6891 if (o->op_type == OP_CONST)
6895 /* fake up an SV array */
6897 assert(!new_patternp);
6898 Newx(new_patternp, n, SV*);
6899 SAVEFREEPV(new_patternp);
6903 if (expr->op_type == OP_CONST)
6904 new_patternp[n] = cSVOPx_sv(expr);
6906 for (o = cLISTOPx(expr)->op_first; o; o = OpSIBLING(o)) {
6907 if (o->op_type == OP_CONST)
6908 new_patternp[n++] = cSVOPo_sv;
6913 DEBUG_PARSE_r(Perl_re_printf( aTHX_
6914 "Assembling pattern from %d elements%s\n", pat_count,
6915 orig_rx_flags & RXf_SPLIT ? " for split" : ""));
6917 /* set expr to the first arg op */
6919 if (pRExC_state->code_blocks && pRExC_state->code_blocks->count
6920 && expr->op_type != OP_CONST)
6922 expr = cLISTOPx(expr)->op_first;
6923 assert( expr->op_type == OP_PUSHMARK
6924 || (expr->op_type == OP_NULL && expr->op_targ == OP_PUSHMARK)
6925 || expr->op_type == OP_PADRANGE);
6926 expr = OpSIBLING(expr);
6929 pat = S_concat_pat(aTHX_ pRExC_state, NULL, new_patternp, pat_count,
6930 expr, &recompile, NULL);
6932 /* handle bare (possibly after overloading) regex: foo =~ $re */
6937 if (SvTYPE(re) == SVt_REGEXP) {
6941 DEBUG_PARSE_r(Perl_re_printf( aTHX_
6942 "Precompiled pattern%s\n",
6943 orig_rx_flags & RXf_SPLIT ? " for split" : ""));
6949 exp = SvPV_nomg(pat, plen);
6951 if (!eng->op_comp) {
6952 if ((SvUTF8(pat) && IN_BYTES)
6953 || SvGMAGICAL(pat) || SvAMAGIC(pat))
6955 /* make a temporary copy; either to convert to bytes,
6956 * or to avoid repeating get-magic / overloaded stringify */
6957 pat = newSVpvn_flags(exp, plen, SVs_TEMP |
6958 (IN_BYTES ? 0 : SvUTF8(pat)));
6960 return CALLREGCOMP_ENG(eng, pat, orig_rx_flags);
6963 /* ignore the utf8ness if the pattern is 0 length */
6964 RExC_utf8 = RExC_orig_utf8 = (plen == 0 || IN_BYTES) ? 0 : SvUTF8(pat);
6966 RExC_uni_semantics = 0;
6967 RExC_seen_unfolded_sharp_s = 0;
6968 RExC_contains_locale = 0;
6969 RExC_strict = cBOOL(pm_flags & RXf_PMf_STRICT);
6970 RExC_study_started = 0;
6971 pRExC_state->runtime_code_qr = NULL;
6972 RExC_frame_head= NULL;
6973 RExC_frame_last= NULL;
6974 RExC_frame_count= 0;
6977 RExC_mysv1= sv_newmortal();
6978 RExC_mysv2= sv_newmortal();
6981 SV *dsv= sv_newmortal();
6982 RE_PV_QUOTED_DECL(s, RExC_utf8, dsv, exp, plen, 60);
6983 Perl_re_printf( aTHX_ "%sCompiling REx%s %s\n",
6984 PL_colors[4],PL_colors[5],s);
6988 /* we jump here if we have to recompile, e.g., from upgrading the pattern
6991 if ((pm_flags & PMf_USE_RE_EVAL)
6992 /* this second condition covers the non-regex literal case,
6993 * i.e. $foo =~ '(?{})'. */
6994 || (IN_PERL_COMPILETIME && (PL_hints & HINT_RE_EVAL))
6996 runtime_code = S_has_runtime_code(aTHX_ pRExC_state, exp, plen);
6998 /* return old regex if pattern hasn't changed */
6999 /* XXX: note in the below we have to check the flags as well as the
7002 * Things get a touch tricky as we have to compare the utf8 flag
7003 * independently from the compile flags. */
7007 && !!RX_UTF8(old_re) == !!RExC_utf8
7008 && ( RX_COMPFLAGS(old_re) == ( orig_rx_flags & RXf_PMf_FLAGCOPYMASK ) )
7009 && RX_PRECOMP(old_re)
7010 && RX_PRELEN(old_re) == plen
7011 && memEQ(RX_PRECOMP(old_re), exp, plen)
7012 && !runtime_code /* with runtime code, always recompile */ )
7017 rx_flags = orig_rx_flags;
7019 if ( initial_charset == REGEX_DEPENDS_CHARSET
7020 && (RExC_utf8 ||RExC_uni_semantics))
7023 /* Set to use unicode semantics if the pattern is in utf8 and has the
7024 * 'depends' charset specified, as it means unicode when utf8 */
7025 set_regex_charset(&rx_flags, REGEX_UNICODE_CHARSET);
7029 RExC_precomp_adj = 0;
7030 RExC_flags = rx_flags;
7031 RExC_pm_flags = pm_flags;
7034 assert(TAINTING_get || !TAINT_get);
7036 Perl_croak(aTHX_ "Eval-group in insecure regular expression");
7038 if (!S_compile_runtime_code(aTHX_ pRExC_state, exp, plen)) {
7039 /* whoops, we have a non-utf8 pattern, whilst run-time code
7040 * got compiled as utf8. Try again with a utf8 pattern */
7041 S_pat_upgrade_to_utf8(aTHX_ pRExC_state, &exp, &plen,
7042 pRExC_state->code_blocks ? pRExC_state->code_blocks->count : 0);
7043 goto redo_first_pass;
7046 assert(!pRExC_state->runtime_code_qr);
7052 RExC_in_lookbehind = 0;
7053 RExC_seen_zerolen = *exp == '^' ? -1 : 0;
7056 RExC_recode_x_to_native = 0;
7058 RExC_in_multi_char_class = 0;
7060 /* First pass: determine size, legality. */
7062 RExC_start = RExC_adjusted_start = exp;
7063 RExC_end = exp + plen;
7064 RExC_precomp_end = RExC_end;
7069 RExC_emit = (regnode *) &RExC_emit_dummy;
7070 RExC_whilem_seen = 0;
7071 RExC_open_parens = NULL;
7072 RExC_close_parens = NULL;
7074 RExC_paren_names = NULL;
7076 RExC_paren_name_list = NULL;
7078 RExC_recurse = NULL;
7079 RExC_study_chunk_recursed = NULL;
7080 RExC_study_chunk_recursed_bytes= 0;
7081 RExC_recurse_count = 0;
7082 pRExC_state->code_index = 0;
7084 /* This NUL is guaranteed because the pattern comes from an SV*, and the sv
7085 * code makes sure the final byte is an uncounted NUL. But should this
7086 * ever not be the case, lots of things could read beyond the end of the
7087 * buffer: loops like
7088 * while(isFOO(*RExC_parse)) RExC_parse++;
7089 * strchr(RExC_parse, "foo");
7090 * etc. So it is worth noting. */
7091 assert(*RExC_end == '\0');
7094 Perl_re_printf( aTHX_ "Starting first pass (sizing)\n");
7096 RExC_lastparse=NULL;
7099 if (reg(pRExC_state, 0, &flags,1) == NULL) {
7100 /* It's possible to write a regexp in ascii that represents Unicode
7101 codepoints outside of the byte range, such as via \x{100}. If we
7102 detect such a sequence we have to convert the entire pattern to utf8
7103 and then recompile, as our sizing calculation will have been based
7104 on 1 byte == 1 character, but we will need to use utf8 to encode
7105 at least some part of the pattern, and therefore must convert the whole
7108 if (flags & RESTART_PASS1) {
7109 if (flags & NEED_UTF8) {
7110 S_pat_upgrade_to_utf8(aTHX_ pRExC_state, &exp, &plen,
7111 pRExC_state->code_blocks ? pRExC_state->code_blocks->count : 0);
7114 DEBUG_PARSE_r(Perl_re_printf( aTHX_
7115 "Need to redo pass 1\n"));
7118 goto redo_first_pass;
7120 Perl_croak(aTHX_ "panic: reg returned NULL to re_op_compile for sizing pass, flags=%#" UVxf, (UV) flags);
7124 Perl_re_printf( aTHX_
7125 "Required size %" IVdf " nodes\n"
7126 "Starting second pass (creation)\n",
7129 RExC_lastparse=NULL;
7132 /* The first pass could have found things that force Unicode semantics */
7133 if ((RExC_utf8 || RExC_uni_semantics)
7134 && get_regex_charset(rx_flags) == REGEX_DEPENDS_CHARSET)
7136 set_regex_charset(&rx_flags, REGEX_UNICODE_CHARSET);
7139 /* Small enough for pointer-storage convention?
7140 If extralen==0, this means that we will not need long jumps. */
7141 if (RExC_size >= 0x10000L && RExC_extralen)
7142 RExC_size += RExC_extralen;
7145 if (RExC_whilem_seen > 15)
7146 RExC_whilem_seen = 15;
7148 /* Allocate space and zero-initialize. Note, the two step process
7149 of zeroing when in debug mode, thus anything assigned has to
7150 happen after that */
7151 rx = (REGEXP*) newSV_type(SVt_REGEXP);
7153 Newxc(ri, sizeof(regexp_internal) + (unsigned)RExC_size * sizeof(regnode),
7154 char, regexp_internal);
7155 if ( r == NULL || ri == NULL )
7156 FAIL("Regexp out of space");
7158 /* avoid reading uninitialized memory in DEBUGGING code in study_chunk() */
7159 Zero(ri, sizeof(regexp_internal) + (unsigned)RExC_size * sizeof(regnode),
7162 /* bulk initialize base fields with 0. */
7163 Zero(ri, sizeof(regexp_internal), char);
7166 /* non-zero initialization begins here */
7169 r->extflags = rx_flags;
7170 RXp_COMPFLAGS(r) = orig_rx_flags & RXf_PMf_FLAGCOPYMASK;
7172 if (pm_flags & PMf_IS_QR) {
7173 ri->code_blocks = pRExC_state->code_blocks;
7174 if (ri->code_blocks)
7175 ri->code_blocks->refcnt++;
7179 bool has_p = ((r->extflags & RXf_PMf_KEEPCOPY) == RXf_PMf_KEEPCOPY);
7180 bool has_charset = (get_regex_charset(r->extflags)
7181 != REGEX_DEPENDS_CHARSET);
7183 /* The caret is output if there are any defaults: if not all the STD
7184 * flags are set, or if no character set specifier is needed */
7186 (((r->extflags & RXf_PMf_STD_PMMOD) != RXf_PMf_STD_PMMOD)
7188 bool has_runon = ((RExC_seen & REG_RUN_ON_COMMENT_SEEN)
7189 == REG_RUN_ON_COMMENT_SEEN);
7190 U8 reganch = (U8)((r->extflags & RXf_PMf_STD_PMMOD)
7191 >> RXf_PMf_STD_PMMOD_SHIFT);
7192 const char *fptr = STD_PAT_MODS; /*"msixxn"*/
7195 /* We output all the necessary flags; we never output a minus, as all
7196 * those are defaults, so are
7197 * covered by the caret */
7198 const STRLEN wraplen = plen + has_p + has_runon
7199 + has_default /* If needs a caret */
7200 + PL_bitcount[reganch] /* 1 char for each set standard flag */
7202 /* If needs a character set specifier */
7203 + ((has_charset) ? MAX_CHARSET_NAME_LENGTH : 0)
7204 + (sizeof("(?:)") - 1);
7206 /* make sure PL_bitcount bounds not exceeded */
7207 assert(sizeof(STD_PAT_MODS) <= 8);
7209 Newx(p, wraplen + 1, char); /* +1 for the ending NUL */
7210 r->xpv_len_u.xpvlenu_pv = p;
7212 SvFLAGS(rx) |= SVf_UTF8;
7215 /* If a default, cover it using the caret */
7217 *p++= DEFAULT_PAT_MOD;
7221 const char* const name = get_regex_charset_name(r->extflags, &len);
7222 Copy(name, p, len, char);
7226 *p++ = KEEPCOPY_PAT_MOD; /*'p'*/
7229 while((ch = *fptr++)) {
7237 Copy(RExC_precomp, p, plen, char);
7238 assert ((RX_WRAPPED(rx) - p) < 16);
7239 r->pre_prefix = p - RX_WRAPPED(rx);
7245 SvCUR_set(rx, p - RX_WRAPPED(rx));
7249 r->nparens = RExC_npar - 1; /* set early to validate backrefs */
7251 /* Useful during FAIL. */
7252 #ifdef RE_TRACK_PATTERN_OFFSETS
7253 Newxz(ri->u.offsets, 2*RExC_size+1, U32); /* MJD 20001228 */
7254 DEBUG_OFFSETS_r(Perl_re_printf( aTHX_
7255 "%s %" UVuf " bytes for offset annotations.\n",
7256 ri->u.offsets ? "Got" : "Couldn't get",
7257 (UV)((2*RExC_size+1) * sizeof(U32))));
7259 SetProgLen(ri,RExC_size);
7264 /* Second pass: emit code. */
7265 RExC_flags = rx_flags; /* don't let top level (?i) bleed */
7266 RExC_pm_flags = pm_flags;
7268 RExC_end = exp + plen;
7270 RExC_emit_start = ri->program;
7271 RExC_emit = ri->program;
7272 RExC_emit_bound = ri->program + RExC_size + 1;
7273 pRExC_state->code_index = 0;
7275 *((char*) RExC_emit++) = (char) REG_MAGIC;
7276 /* setup various meta data about recursion, this all requires
7277 * RExC_npar to be correctly set, and a bit later on we clear it */
7278 if (RExC_seen & REG_RECURSE_SEEN) {
7279 DEBUG_OPTIMISE_MORE_r(Perl_re_printf( aTHX_
7280 "%*s%*s Setting up open/close parens\n",
7281 22, "| |", (int)(0 * 2 + 1), ""));
7283 /* setup RExC_open_parens, which holds the address of each
7284 * OPEN tag, and to make things simpler for the 0 index
7285 * the start of the program - this is used later for offsets */
7286 Newxz(RExC_open_parens, RExC_npar,regnode *);
7287 SAVEFREEPV(RExC_open_parens);
7288 RExC_open_parens[0] = RExC_emit;
7290 /* setup RExC_close_parens, which holds the address of each
7291 * CLOSE tag, and to make things simpler for the 0 index
7292 * the end of the program - this is used later for offsets */
7293 Newxz(RExC_close_parens, RExC_npar,regnode *);
7294 SAVEFREEPV(RExC_close_parens);
7295 /* we dont know where end op starts yet, so we dont
7296 * need to set RExC_close_parens[0] like we do RExC_open_parens[0] above */
7298 /* Note, RExC_npar is 1 + the number of parens in a pattern.
7299 * So its 1 if there are no parens. */
7300 RExC_study_chunk_recursed_bytes= (RExC_npar >> 3) +
7301 ((RExC_npar & 0x07) != 0);
7302 Newx(RExC_study_chunk_recursed,
7303 RExC_study_chunk_recursed_bytes * RExC_npar, U8);
7304 SAVEFREEPV(RExC_study_chunk_recursed);
7307 if (reg(pRExC_state, 0, &flags,1) == NULL) {
7309 Perl_croak(aTHX_ "panic: reg returned NULL to re_op_compile for generation pass, flags=%#" UVxf, (UV) flags);
7312 Perl_re_printf( aTHX_ "Starting post parse optimization\n");
7315 /* XXXX To minimize changes to RE engine we always allocate
7316 3-units-long substrs field. */
7317 Newx(r->substrs, 1, struct reg_substr_data);
7318 if (RExC_recurse_count) {
7319 Newxz(RExC_recurse,RExC_recurse_count,regnode *);
7320 SAVEFREEPV(RExC_recurse);
7324 r->minlen = minlen = sawlookahead = sawplus = sawopen = sawminmod = 0;
7326 RExC_study_chunk_recursed_count= 0;
7328 Zero(r->substrs, 1, struct reg_substr_data);
7329 if (RExC_study_chunk_recursed) {
7330 Zero(RExC_study_chunk_recursed,
7331 RExC_study_chunk_recursed_bytes * RExC_npar, U8);
7335 #ifdef TRIE_STUDY_OPT
7337 StructCopy(&zero_scan_data, &data, scan_data_t);
7338 copyRExC_state = RExC_state;
7341 DEBUG_OPTIMISE_r(Perl_re_printf( aTHX_ "Restudying\n"));
7343 RExC_state = copyRExC_state;
7344 if (seen & REG_TOP_LEVEL_BRANCHES_SEEN)
7345 RExC_seen |= REG_TOP_LEVEL_BRANCHES_SEEN;
7347 RExC_seen &= ~REG_TOP_LEVEL_BRANCHES_SEEN;
7348 StructCopy(&zero_scan_data, &data, scan_data_t);
7351 StructCopy(&zero_scan_data, &data, scan_data_t);
7354 /* Dig out information for optimizations. */
7355 r->extflags = RExC_flags; /* was pm_op */
7356 /*dmq: removed as part of de-PMOP: pm->op_pmflags = RExC_flags; */
7359 SvUTF8_on(rx); /* Unicode in it? */
7360 ri->regstclass = NULL;
7361 if (RExC_naughty >= TOO_NAUGHTY) /* Probably an expensive pattern. */
7362 r->intflags |= PREGf_NAUGHTY;
7363 scan = ri->program + 1; /* First BRANCH. */
7365 /* testing for BRANCH here tells us whether there is "must appear"
7366 data in the pattern. If there is then we can use it for optimisations */
7367 if (!(RExC_seen & REG_TOP_LEVEL_BRANCHES_SEEN)) { /* Only one top-level choice.
7370 STRLEN longest_float_length, longest_fixed_length;
7371 regnode_ssc ch_class; /* pointed to by data */
7373 SSize_t last_close = 0; /* pointed to by data */
7374 regnode *first= scan;
7375 regnode *first_next= regnext(first);
7377 * Skip introductions and multiplicators >= 1
7378 * so that we can extract the 'meat' of the pattern that must
7379 * match in the large if() sequence following.
7380 * NOTE that EXACT is NOT covered here, as it is normally
7381 * picked up by the optimiser separately.
7383 * This is unfortunate as the optimiser isnt handling lookahead
7384 * properly currently.
7387 while ((OP(first) == OPEN && (sawopen = 1)) ||
7388 /* An OR of *one* alternative - should not happen now. */
7389 (OP(first) == BRANCH && OP(first_next) != BRANCH) ||
7390 /* for now we can't handle lookbehind IFMATCH*/
7391 (OP(first) == IFMATCH && !first->flags && (sawlookahead = 1)) ||
7392 (OP(first) == PLUS) ||
7393 (OP(first) == MINMOD) ||
7394 /* An {n,m} with n>0 */
7395 (PL_regkind[OP(first)] == CURLY && ARG1(first) > 0) ||
7396 (OP(first) == NOTHING && PL_regkind[OP(first_next)] != END ))
7399 * the only op that could be a regnode is PLUS, all the rest
7400 * will be regnode_1 or regnode_2.
7402 * (yves doesn't think this is true)
7404 if (OP(first) == PLUS)
7407 if (OP(first) == MINMOD)
7409 first += regarglen[OP(first)];
7411 first = NEXTOPER(first);
7412 first_next= regnext(first);
7415 /* Starting-point info. */
7417 DEBUG_PEEP("first:",first,0);
7418 /* Ignore EXACT as we deal with it later. */
7419 if (PL_regkind[OP(first)] == EXACT) {
7420 if (OP(first) == EXACT || OP(first) == EXACTL)
7421 NOOP; /* Empty, get anchored substr later. */
7423 ri->regstclass = first;
7426 else if (PL_regkind[OP(first)] == TRIE &&
7427 ((reg_trie_data *)ri->data->data[ ARG(first) ])->minlen>0)
7429 /* this can happen only on restudy */
7430 ri->regstclass = construct_ahocorasick_from_trie(pRExC_state, (regnode *)first, 0);
7433 else if (REGNODE_SIMPLE(OP(first)))
7434 ri->regstclass = first;
7435 else if (PL_regkind[OP(first)] == BOUND ||
7436 PL_regkind[OP(first)] == NBOUND)
7437 ri->regstclass = first;
7438 else if (PL_regkind[OP(first)] == BOL) {
7439 r->intflags |= (OP(first) == MBOL
7442 first = NEXTOPER(first);
7445 else if (OP(first) == GPOS) {
7446 r->intflags |= PREGf_ANCH_GPOS;
7447 first = NEXTOPER(first);
7450 else if ((!sawopen || !RExC_sawback) &&
7452 (OP(first) == STAR &&
7453 PL_regkind[OP(NEXTOPER(first))] == REG_ANY) &&
7454 !(r->intflags & PREGf_ANCH) && !pRExC_state->code_blocks)
7456 /* turn .* into ^.* with an implied $*=1 */
7458 (OP(NEXTOPER(first)) == REG_ANY)
7461 r->intflags |= (type | PREGf_IMPLICIT);
7462 first = NEXTOPER(first);
7465 if (sawplus && !sawminmod && !sawlookahead
7466 && (!sawopen || !RExC_sawback)
7467 && !pRExC_state->code_blocks) /* May examine pos and $& */
7468 /* x+ must match at the 1st pos of run of x's */
7469 r->intflags |= PREGf_SKIP;
7471 /* Scan is after the zeroth branch, first is atomic matcher. */
7472 #ifdef TRIE_STUDY_OPT
7475 Perl_re_printf( aTHX_ "first at %" IVdf "\n",
7476 (IV)(first - scan + 1))
7480 Perl_re_printf( aTHX_ "first at %" IVdf "\n",
7481 (IV)(first - scan + 1))
7487 * If there's something expensive in the r.e., find the
7488 * longest literal string that must appear and make it the
7489 * regmust. Resolve ties in favor of later strings, since
7490 * the regstart check works with the beginning of the r.e.
7491 * and avoiding duplication strengthens checking. Not a
7492 * strong reason, but sufficient in the absence of others.
7493 * [Now we resolve ties in favor of the earlier string if
7494 * it happens that c_offset_min has been invalidated, since the
7495 * earlier string may buy us something the later one won't.]
7498 data.longest_fixed = newSVpvs("");
7499 data.longest_float = newSVpvs("");
7500 data.last_found = newSVpvs("");
7501 data.longest = &(data.longest_fixed);
7502 ENTER_with_name("study_chunk");
7503 SAVEFREESV(data.longest_fixed);
7504 SAVEFREESV(data.longest_float);
7505 SAVEFREESV(data.last_found);
7507 if (!ri->regstclass) {
7508 ssc_init(pRExC_state, &ch_class);
7509 data.start_class = &ch_class;
7510 stclass_flag = SCF_DO_STCLASS_AND;
7511 } else /* XXXX Check for BOUND? */
7513 data.last_closep = &last_close;
7516 minlen = study_chunk(pRExC_state, &first, &minlen, &fake,
7517 scan + RExC_size, /* Up to end */
7519 SCF_DO_SUBSTR | SCF_WHILEM_VISITED_POS | stclass_flag
7520 | (restudied ? SCF_TRIE_DOING_RESTUDY : 0),
7524 CHECK_RESTUDY_GOTO_butfirst(LEAVE_with_name("study_chunk"));
7527 if ( RExC_npar == 1 && data.longest == &(data.longest_fixed)
7528 && data.last_start_min == 0 && data.last_end > 0
7529 && !RExC_seen_zerolen
7530 && !(RExC_seen & REG_VERBARG_SEEN)
7531 && !(RExC_seen & REG_GPOS_SEEN)
7533 r->extflags |= RXf_CHECK_ALL;
7535 scan_commit(pRExC_state, &data,&minlen,0);
7537 longest_float_length = CHR_SVLEN(data.longest_float);
7539 if (! ((SvCUR(data.longest_fixed) /* ok to leave SvCUR */
7540 && data.offset_fixed == data.offset_float_min
7541 && SvCUR(data.longest_fixed) == SvCUR(data.longest_float)))
7542 && S_setup_longest (aTHX_ pRExC_state,
7546 &(r->float_end_shift),
7547 data.lookbehind_float,
7548 data.offset_float_min,
7550 longest_float_length,
7551 cBOOL(data.flags & SF_FL_BEFORE_EOL),
7552 cBOOL(data.flags & SF_FL_BEFORE_MEOL)))
7554 r->float_min_offset = data.offset_float_min - data.lookbehind_float;
7555 r->float_max_offset = data.offset_float_max;
7556 if (data.offset_float_max < SSize_t_MAX) /* Don't offset infinity */
7557 r->float_max_offset -= data.lookbehind_float;
7558 SvREFCNT_inc_simple_void_NN(data.longest_float);
7561 r->float_substr = r->float_utf8 = NULL;
7562 longest_float_length = 0;
7565 longest_fixed_length = CHR_SVLEN(data.longest_fixed);
7567 if (S_setup_longest (aTHX_ pRExC_state,
7569 &(r->anchored_utf8),
7570 &(r->anchored_substr),
7571 &(r->anchored_end_shift),
7572 data.lookbehind_fixed,
7575 longest_fixed_length,
7576 cBOOL(data.flags & SF_FIX_BEFORE_EOL),
7577 cBOOL(data.flags & SF_FIX_BEFORE_MEOL)))
7579 r->anchored_offset = data.offset_fixed - data.lookbehind_fixed;
7580 SvREFCNT_inc_simple_void_NN(data.longest_fixed);
7583 r->anchored_substr = r->anchored_utf8 = NULL;
7584 longest_fixed_length = 0;
7586 LEAVE_with_name("study_chunk");
7589 && (OP(ri->regstclass) == REG_ANY || OP(ri->regstclass) == SANY))
7590 ri->regstclass = NULL;
7592 if ((!(r->anchored_substr || r->anchored_utf8) || r->anchored_offset)
7594 && ! (ANYOF_FLAGS(data.start_class) & SSC_MATCHES_EMPTY_STRING)
7595 && is_ssc_worth_it(pRExC_state, data.start_class))
7597 const U32 n = add_data(pRExC_state, STR_WITH_LEN("f"));
7599 ssc_finalize(pRExC_state, data.start_class);
7601 Newx(RExC_rxi->data->data[n], 1, regnode_ssc);
7602 StructCopy(data.start_class,
7603 (regnode_ssc*)RExC_rxi->data->data[n],
7605 ri->regstclass = (regnode*)RExC_rxi->data->data[n];
7606 r->intflags &= ~PREGf_SKIP; /* Used in find_byclass(). */
7607 DEBUG_COMPILE_r({ SV *sv = sv_newmortal();
7608 regprop(r, sv, (regnode*)data.start_class, NULL, pRExC_state);
7609 Perl_re_printf( aTHX_
7610 "synthetic stclass \"%s\".\n",
7611 SvPVX_const(sv));});
7612 data.start_class = NULL;
7615 /* A temporary algorithm prefers floated substr to fixed one to dig
7617 if (longest_fixed_length > longest_float_length) {
7618 r->substrs->check_ix = 0;
7619 r->check_end_shift = r->anchored_end_shift;
7620 r->check_substr = r->anchored_substr;
7621 r->check_utf8 = r->anchored_utf8;
7622 r->check_offset_min = r->check_offset_max = r->anchored_offset;
7623 if (r->intflags & (PREGf_ANCH_SBOL|PREGf_ANCH_GPOS))
7624 r->intflags |= PREGf_NOSCAN;
7627 r->substrs->check_ix = 1;
7628 r->check_end_shift = r->float_end_shift;
7629 r->check_substr = r->float_substr;
7630 r->check_utf8 = r->float_utf8;
7631 r->check_offset_min = r->float_min_offset;
7632 r->check_offset_max = r->float_max_offset;
7634 if ((r->check_substr || r->check_utf8) ) {
7635 r->extflags |= RXf_USE_INTUIT;
7636 if (SvTAIL(r->check_substr ? r->check_substr : r->check_utf8))
7637 r->extflags |= RXf_INTUIT_TAIL;
7639 r->substrs->data[0].max_offset = r->substrs->data[0].min_offset;
7641 /* XXX Unneeded? dmq (shouldn't as this is handled elsewhere)
7642 if ( (STRLEN)minlen < longest_float_length )
7643 minlen= longest_float_length;
7644 if ( (STRLEN)minlen < longest_fixed_length )
7645 minlen= longest_fixed_length;
7649 /* Several toplevels. Best we can is to set minlen. */
7651 regnode_ssc ch_class;
7652 SSize_t last_close = 0;
7654 DEBUG_PARSE_r(Perl_re_printf( aTHX_ "\nMulti Top Level\n"));
7656 scan = ri->program + 1;
7657 ssc_init(pRExC_state, &ch_class);
7658 data.start_class = &ch_class;
7659 data.last_closep = &last_close;
7662 minlen = study_chunk(pRExC_state,
7663 &scan, &minlen, &fake, scan + RExC_size, &data, -1, 0, NULL,
7664 SCF_DO_STCLASS_AND|SCF_WHILEM_VISITED_POS|(restudied
7665 ? SCF_TRIE_DOING_RESTUDY
7669 CHECK_RESTUDY_GOTO_butfirst(NOOP);
7671 r->check_substr = r->check_utf8 = r->anchored_substr = r->anchored_utf8
7672 = r->float_substr = r->float_utf8 = NULL;
7674 if (! (ANYOF_FLAGS(data.start_class) & SSC_MATCHES_EMPTY_STRING)
7675 && is_ssc_worth_it(pRExC_state, data.start_class))
7677 const U32 n = add_data(pRExC_state, STR_WITH_LEN("f"));
7679 ssc_finalize(pRExC_state, data.start_class);
7681 Newx(RExC_rxi->data->data[n], 1, regnode_ssc);
7682 StructCopy(data.start_class,
7683 (regnode_ssc*)RExC_rxi->data->data[n],
7685 ri->regstclass = (regnode*)RExC_rxi->data->data[n];
7686 r->intflags &= ~PREGf_SKIP; /* Used in find_byclass(). */
7687 DEBUG_COMPILE_r({ SV* sv = sv_newmortal();
7688 regprop(r, sv, (regnode*)data.start_class, NULL, pRExC_state);
7689 Perl_re_printf( aTHX_
7690 "synthetic stclass \"%s\".\n",
7691 SvPVX_const(sv));});
7692 data.start_class = NULL;
7696 if (RExC_seen & REG_UNBOUNDED_QUANTIFIER_SEEN) {
7697 r->extflags |= RXf_UNBOUNDED_QUANTIFIER_SEEN;
7698 r->maxlen = REG_INFTY;
7701 r->maxlen = RExC_maxlen;
7704 /* Guard against an embedded (?=) or (?<=) with a longer minlen than
7705 the "real" pattern. */
7707 Perl_re_printf( aTHX_ "minlen: %" IVdf " r->minlen:%" IVdf " maxlen:%" IVdf "\n",
7708 (IV)minlen, (IV)r->minlen, (IV)RExC_maxlen);
7710 r->minlenret = minlen;
7711 if (r->minlen < minlen)
7714 if (RExC_seen & REG_RECURSE_SEEN ) {
7715 r->intflags |= PREGf_RECURSE_SEEN;
7716 Newxz(r->recurse_locinput, r->nparens + 1, char *);
7718 if (RExC_seen & REG_GPOS_SEEN)
7719 r->intflags |= PREGf_GPOS_SEEN;
7720 if (RExC_seen & REG_LOOKBEHIND_SEEN)
7721 r->extflags |= RXf_NO_INPLACE_SUBST; /* inplace might break the
7723 if (pRExC_state->code_blocks)
7724 r->extflags |= RXf_EVAL_SEEN;
7725 if (RExC_seen & REG_VERBARG_SEEN)
7727 r->intflags |= PREGf_VERBARG_SEEN;
7728 r->extflags |= RXf_NO_INPLACE_SUBST; /* don't understand this! Yves */
7730 if (RExC_seen & REG_CUTGROUP_SEEN)
7731 r->intflags |= PREGf_CUTGROUP_SEEN;
7732 if (pm_flags & PMf_USE_RE_EVAL)
7733 r->intflags |= PREGf_USE_RE_EVAL;
7734 if (RExC_paren_names)
7735 RXp_PAREN_NAMES(r) = MUTABLE_HV(SvREFCNT_inc(RExC_paren_names));
7737 RXp_PAREN_NAMES(r) = NULL;
7739 /* If we have seen an anchor in our pattern then we set the extflag RXf_IS_ANCHORED
7740 * so it can be used in pp.c */
7741 if (r->intflags & PREGf_ANCH)
7742 r->extflags |= RXf_IS_ANCHORED;
7746 /* this is used to identify "special" patterns that might result
7747 * in Perl NOT calling the regex engine and instead doing the match "itself",
7748 * particularly special cases in split//. By having the regex compiler
7749 * do this pattern matching at a regop level (instead of by inspecting the pattern)
7750 * we avoid weird issues with equivalent patterns resulting in different behavior,
7751 * AND we allow non Perl engines to get the same optimizations by the setting the
7752 * flags appropriately - Yves */
7753 regnode *first = ri->program + 1;
7755 regnode *next = regnext(first);
7758 if (PL_regkind[fop] == NOTHING && nop == END)
7759 r->extflags |= RXf_NULL;
7760 else if ((fop == MBOL || (fop == SBOL && !first->flags)) && nop == END)
7761 /* when fop is SBOL first->flags will be true only when it was
7762 * produced by parsing /\A/, and not when parsing /^/. This is
7763 * very important for the split code as there we want to
7764 * treat /^/ as /^/m, but we do not want to treat /\A/ as /^/m.
7765 * See rt #122761 for more details. -- Yves */
7766 r->extflags |= RXf_START_ONLY;
7767 else if (fop == PLUS
7768 && PL_regkind[nop] == POSIXD && FLAGS(next) == _CC_SPACE
7770 r->extflags |= RXf_WHITE;
7771 else if ( r->extflags & RXf_SPLIT
7772 && (fop == EXACT || fop == EXACTL)
7773 && STR_LEN(first) == 1
7774 && *(STRING(first)) == ' '
7776 r->extflags |= (RXf_SKIPWHITE|RXf_WHITE);
7780 if (RExC_contains_locale) {
7781 RXp_EXTFLAGS(r) |= RXf_TAINTED;
7785 if (RExC_paren_names) {
7786 ri->name_list_idx = add_data( pRExC_state, STR_WITH_LEN("a"));
7787 ri->data->data[ri->name_list_idx]
7788 = (void*)SvREFCNT_inc(RExC_paren_name_list);
7791 ri->name_list_idx = 0;
7793 while ( RExC_recurse_count > 0 ) {
7794 const regnode *scan = RExC_recurse[ --RExC_recurse_count ];
7796 * This data structure is set up in study_chunk() and is used
7797 * to calculate the distance between a GOSUB regopcode and
7798 * the OPEN/CURLYM (CURLYM's are special and can act like OPEN's)
7801 * If for some reason someone writes code that optimises
7802 * away a GOSUB opcode then the assert should be changed to
7803 * an if(scan) to guard the ARG2L_SET() - Yves
7806 assert(scan && OP(scan) == GOSUB);
7807 ARG2L_SET( scan, RExC_open_parens[ARG(scan)] - scan );
7810 Newxz(r->offs, RExC_npar, regexp_paren_pair);
7811 /* assume we don't need to swap parens around before we match */
7813 Perl_re_printf( aTHX_ "study_chunk_recursed_count: %lu\n",
7814 (unsigned long)RExC_study_chunk_recursed_count);
7818 Perl_re_printf( aTHX_ "Final program:\n");
7821 #ifdef RE_TRACK_PATTERN_OFFSETS
7822 DEBUG_OFFSETS_r(if (ri->u.offsets) {
7823 const STRLEN len = ri->u.offsets[0];
7825 GET_RE_DEBUG_FLAGS_DECL;
7826 Perl_re_printf( aTHX_
7827 "Offsets: [%" UVuf "]\n\t", (UV)ri->u.offsets[0]);
7828 for (i = 1; i <= len; i++) {
7829 if (ri->u.offsets[i*2-1] || ri->u.offsets[i*2])
7830 Perl_re_printf( aTHX_ "%" UVuf ":%" UVuf "[%" UVuf "] ",
7831 (UV)i, (UV)ri->u.offsets[i*2-1], (UV)ri->u.offsets[i*2]);
7833 Perl_re_printf( aTHX_ "\n");
7838 /* under ithreads the ?pat? PMf_USED flag on the pmop is simulated
7839 * by setting the regexp SV to readonly-only instead. If the
7840 * pattern's been recompiled, the USEDness should remain. */
7841 if (old_re && SvREADONLY(old_re))
7849 Perl_reg_named_buff(pTHX_ REGEXP * const rx, SV * const key, SV * const value,
7852 PERL_ARGS_ASSERT_REG_NAMED_BUFF;
7854 PERL_UNUSED_ARG(value);
7856 if (flags & RXapif_FETCH) {
7857 return reg_named_buff_fetch(rx, key, flags);
7858 } else if (flags & (RXapif_STORE | RXapif_DELETE | RXapif_CLEAR)) {
7859 Perl_croak_no_modify();
7861 } else if (flags & RXapif_EXISTS) {
7862 return reg_named_buff_exists(rx, key, flags)
7865 } else if (flags & RXapif_REGNAMES) {
7866 return reg_named_buff_all(rx, flags);
7867 } else if (flags & (RXapif_SCALAR | RXapif_REGNAMES_COUNT)) {
7868 return reg_named_buff_scalar(rx, flags);
7870 Perl_croak(aTHX_ "panic: Unknown flags %d in named_buff", (int)flags);
7876 Perl_reg_named_buff_iter(pTHX_ REGEXP * const rx, const SV * const lastkey,
7879 PERL_ARGS_ASSERT_REG_NAMED_BUFF_ITER;
7880 PERL_UNUSED_ARG(lastkey);
7882 if (flags & RXapif_FIRSTKEY)
7883 return reg_named_buff_firstkey(rx, flags);
7884 else if (flags & RXapif_NEXTKEY)
7885 return reg_named_buff_nextkey(rx, flags);
7887 Perl_croak(aTHX_ "panic: Unknown flags %d in named_buff_iter",
7894 Perl_reg_named_buff_fetch(pTHX_ REGEXP * const r, SV * const namesv,
7897 AV *retarray = NULL;
7899 struct regexp *const rx = ReANY(r);
7901 PERL_ARGS_ASSERT_REG_NAMED_BUFF_FETCH;
7903 if (flags & RXapif_ALL)
7906 if (rx && RXp_PAREN_NAMES(rx)) {
7907 HE *he_str = hv_fetch_ent( RXp_PAREN_NAMES(rx), namesv, 0, 0 );
7910 SV* sv_dat=HeVAL(he_str);
7911 I32 *nums=(I32*)SvPVX(sv_dat);
7912 for ( i=0; i<SvIVX(sv_dat); i++ ) {
7913 if ((I32)(rx->nparens) >= nums[i]
7914 && rx->offs[nums[i]].start != -1
7915 && rx->offs[nums[i]].end != -1)
7918 CALLREG_NUMBUF_FETCH(r,nums[i],ret);
7923 ret = newSVsv(&PL_sv_undef);
7926 av_push(retarray, ret);
7929 return newRV_noinc(MUTABLE_SV(retarray));
7936 Perl_reg_named_buff_exists(pTHX_ REGEXP * const r, SV * const key,
7939 struct regexp *const rx = ReANY(r);
7941 PERL_ARGS_ASSERT_REG_NAMED_BUFF_EXISTS;
7943 if (rx && RXp_PAREN_NAMES(rx)) {
7944 if (flags & RXapif_ALL) {
7945 return hv_exists_ent(RXp_PAREN_NAMES(rx), key, 0);
7947 SV *sv = CALLREG_NAMED_BUFF_FETCH(r, key, flags);
7949 SvREFCNT_dec_NN(sv);
7961 Perl_reg_named_buff_firstkey(pTHX_ REGEXP * const r, const U32 flags)
7963 struct regexp *const rx = ReANY(r);
7965 PERL_ARGS_ASSERT_REG_NAMED_BUFF_FIRSTKEY;
7967 if ( rx && RXp_PAREN_NAMES(rx) ) {
7968 (void)hv_iterinit(RXp_PAREN_NAMES(rx));
7970 return CALLREG_NAMED_BUFF_NEXTKEY(r, NULL, flags & ~RXapif_FIRSTKEY);
7977 Perl_reg_named_buff_nextkey(pTHX_ REGEXP * const r, const U32 flags)
7979 struct regexp *const rx = ReANY(r);
7980 GET_RE_DEBUG_FLAGS_DECL;
7982 PERL_ARGS_ASSERT_REG_NAMED_BUFF_NEXTKEY;
7984 if (rx && RXp_PAREN_NAMES(rx)) {
7985 HV *hv = RXp_PAREN_NAMES(rx);
7987 while ( (temphe = hv_iternext_flags(hv,0)) ) {
7990 SV* sv_dat = HeVAL(temphe);
7991 I32 *nums = (I32*)SvPVX(sv_dat);
7992 for ( i = 0; i < SvIVX(sv_dat); i++ ) {
7993 if ((I32)(rx->lastparen) >= nums[i] &&
7994 rx->offs[nums[i]].start != -1 &&
7995 rx->offs[nums[i]].end != -1)
8001 if (parno || flags & RXapif_ALL) {
8002 return newSVhek(HeKEY_hek(temphe));
8010 Perl_reg_named_buff_scalar(pTHX_ REGEXP * const r, const U32 flags)
8015 struct regexp *const rx = ReANY(r);
8017 PERL_ARGS_ASSERT_REG_NAMED_BUFF_SCALAR;
8019 if (rx && RXp_PAREN_NAMES(rx)) {
8020 if (flags & (RXapif_ALL | RXapif_REGNAMES_COUNT)) {
8021 return newSViv(HvTOTALKEYS(RXp_PAREN_NAMES(rx)));
8022 } else if (flags & RXapif_ONE) {
8023 ret = CALLREG_NAMED_BUFF_ALL(r, (flags | RXapif_REGNAMES));
8024 av = MUTABLE_AV(SvRV(ret));
8025 length = av_tindex(av);
8026 SvREFCNT_dec_NN(ret);
8027 return newSViv(length + 1);
8029 Perl_croak(aTHX_ "panic: Unknown flags %d in named_buff_scalar",
8034 return &PL_sv_undef;
8038 Perl_reg_named_buff_all(pTHX_ REGEXP * const r, const U32 flags)
8040 struct regexp *const rx = ReANY(r);
8043 PERL_ARGS_ASSERT_REG_NAMED_BUFF_ALL;
8045 if (rx && RXp_PAREN_NAMES(rx)) {
8046 HV *hv= RXp_PAREN_NAMES(rx);
8048 (void)hv_iterinit(hv);
8049 while ( (temphe = hv_iternext_flags(hv,0)) ) {
8052 SV* sv_dat = HeVAL(temphe);
8053 I32 *nums = (I32*)SvPVX(sv_dat);
8054 for ( i = 0; i < SvIVX(sv_dat); i++ ) {
8055 if ((I32)(rx->lastparen) >= nums[i] &&
8056 rx->offs[nums[i]].start != -1 &&
8057 rx->offs[nums[i]].end != -1)
8063 if (parno || flags & RXapif_ALL) {
8064 av_push(av, newSVhek(HeKEY_hek(temphe)));
8069 return newRV_noinc(MUTABLE_SV(av));
8073 Perl_reg_numbered_buff_fetch(pTHX_ REGEXP * const r, const I32 paren,
8076 struct regexp *const rx = ReANY(r);
8082 PERL_ARGS_ASSERT_REG_NUMBERED_BUFF_FETCH;
8084 if ( n == RX_BUFF_IDX_CARET_PREMATCH
8085 || n == RX_BUFF_IDX_CARET_FULLMATCH
8086 || n == RX_BUFF_IDX_CARET_POSTMATCH
8089 bool keepcopy = cBOOL(rx->extflags & RXf_PMf_KEEPCOPY);
8091 /* on something like
8094 * the KEEPCOPY is set on the PMOP rather than the regex */
8095 if (PL_curpm && r == PM_GETRE(PL_curpm))
8096 keepcopy = cBOOL(PL_curpm->op_pmflags & PMf_KEEPCOPY);
8105 if (n == RX_BUFF_IDX_CARET_FULLMATCH)
8106 /* no need to distinguish between them any more */
8107 n = RX_BUFF_IDX_FULLMATCH;
8109 if ((n == RX_BUFF_IDX_PREMATCH || n == RX_BUFF_IDX_CARET_PREMATCH)
8110 && rx->offs[0].start != -1)
8112 /* $`, ${^PREMATCH} */
8113 i = rx->offs[0].start;
8117 if ((n == RX_BUFF_IDX_POSTMATCH || n == RX_BUFF_IDX_CARET_POSTMATCH)
8118 && rx->offs[0].end != -1)
8120 /* $', ${^POSTMATCH} */
8121 s = rx->subbeg - rx->suboffset + rx->offs[0].end;
8122 i = rx->sublen + rx->suboffset - rx->offs[0].end;
8125 if ( 0 <= n && n <= (I32)rx->nparens &&
8126 (s1 = rx->offs[n].start) != -1 &&
8127 (t1 = rx->offs[n].end) != -1)
8129 /* $&, ${^MATCH}, $1 ... */
8131 s = rx->subbeg + s1 - rx->suboffset;
8136 assert(s >= rx->subbeg);
8137 assert((STRLEN)rx->sublen >= (STRLEN)((s - rx->subbeg) + i) );
8139 #ifdef NO_TAINT_SUPPORT
8140 sv_setpvn(sv, s, i);
8142 const int oldtainted = TAINT_get;
8144 sv_setpvn(sv, s, i);
8145 TAINT_set(oldtainted);
8147 if (RXp_MATCH_UTF8(rx))
8152 if (RXp_MATCH_TAINTED(rx)) {
8153 if (SvTYPE(sv) >= SVt_PVMG) {
8154 MAGIC* const mg = SvMAGIC(sv);
8157 SvMAGIC_set(sv, mg->mg_moremagic);
8159 if ((mgt = SvMAGIC(sv))) {
8160 mg->mg_moremagic = mgt;
8161 SvMAGIC_set(sv, mg);
8178 Perl_reg_numbered_buff_store(pTHX_ REGEXP * const rx, const I32 paren,
8179 SV const * const value)
8181 PERL_ARGS_ASSERT_REG_NUMBERED_BUFF_STORE;
8183 PERL_UNUSED_ARG(rx);
8184 PERL_UNUSED_ARG(paren);
8185 PERL_UNUSED_ARG(value);
8188 Perl_croak_no_modify();
8192 Perl_reg_numbered_buff_length(pTHX_ REGEXP * const r, const SV * const sv,
8195 struct regexp *const rx = ReANY(r);
8199 PERL_ARGS_ASSERT_REG_NUMBERED_BUFF_LENGTH;
8201 if ( paren == RX_BUFF_IDX_CARET_PREMATCH
8202 || paren == RX_BUFF_IDX_CARET_FULLMATCH
8203 || paren == RX_BUFF_IDX_CARET_POSTMATCH
8206 bool keepcopy = cBOOL(rx->extflags & RXf_PMf_KEEPCOPY);
8208 /* on something like
8211 * the KEEPCOPY is set on the PMOP rather than the regex */
8212 if (PL_curpm && r == PM_GETRE(PL_curpm))
8213 keepcopy = cBOOL(PL_curpm->op_pmflags & PMf_KEEPCOPY);
8219 /* Some of this code was originally in C<Perl_magic_len> in F<mg.c> */
8221 case RX_BUFF_IDX_CARET_PREMATCH: /* ${^PREMATCH} */
8222 case RX_BUFF_IDX_PREMATCH: /* $` */
8223 if (rx->offs[0].start != -1) {
8224 i = rx->offs[0].start;
8233 case RX_BUFF_IDX_CARET_POSTMATCH: /* ${^POSTMATCH} */
8234 case RX_BUFF_IDX_POSTMATCH: /* $' */
8235 if (rx->offs[0].end != -1) {
8236 i = rx->sublen - rx->offs[0].end;
8238 s1 = rx->offs[0].end;
8245 default: /* $& / ${^MATCH}, $1, $2, ... */
8246 if (paren <= (I32)rx->nparens &&
8247 (s1 = rx->offs[paren].start) != -1 &&
8248 (t1 = rx->offs[paren].end) != -1)
8254 if (ckWARN(WARN_UNINITIALIZED))
8255 report_uninit((const SV *)sv);
8260 if (i > 0 && RXp_MATCH_UTF8(rx)) {
8261 const char * const s = rx->subbeg - rx->suboffset + s1;
8266 if (is_utf8_string_loclen((U8*)s, i, &ep, &el))
8273 Perl_reg_qr_package(pTHX_ REGEXP * const rx)
8275 PERL_ARGS_ASSERT_REG_QR_PACKAGE;
8276 PERL_UNUSED_ARG(rx);
8280 return newSVpvs("Regexp");
8283 /* Scans the name of a named buffer from the pattern.
8284 * If flags is REG_RSN_RETURN_NULL returns null.
8285 * If flags is REG_RSN_RETURN_NAME returns an SV* containing the name
8286 * If flags is REG_RSN_RETURN_DATA returns the data SV* corresponding
8287 * to the parsed name as looked up in the RExC_paren_names hash.
8288 * If there is an error throws a vFAIL().. type exception.
8291 #define REG_RSN_RETURN_NULL 0
8292 #define REG_RSN_RETURN_NAME 1
8293 #define REG_RSN_RETURN_DATA 2
8296 S_reg_scan_name(pTHX_ RExC_state_t *pRExC_state, U32 flags)
8298 char *name_start = RExC_parse;
8300 PERL_ARGS_ASSERT_REG_SCAN_NAME;
8302 assert (RExC_parse <= RExC_end);
8303 if (RExC_parse == RExC_end) NOOP;
8304 else if (isIDFIRST_lazy_if_safe(RExC_parse, RExC_end, UTF)) {
8305 /* Note that the code here assumes well-formed UTF-8. Skip IDFIRST by
8306 * using do...while */
8309 RExC_parse += UTF8SKIP(RExC_parse);
8310 } while ( RExC_parse < RExC_end
8311 && isWORDCHAR_utf8_safe((U8*)RExC_parse, (U8*) RExC_end));
8315 } while (RExC_parse < RExC_end && isWORDCHAR(*RExC_parse));
8317 RExC_parse++; /* so the <- from the vFAIL is after the offending
8319 vFAIL("Group name must start with a non-digit word character");
8323 = newSVpvn_flags(name_start, (int)(RExC_parse - name_start),
8324 SVs_TEMP | (UTF ? SVf_UTF8 : 0));
8325 if ( flags == REG_RSN_RETURN_NAME)
8327 else if (flags==REG_RSN_RETURN_DATA) {
8330 if ( ! sv_name ) /* should not happen*/
8331 Perl_croak(aTHX_ "panic: no svname in reg_scan_name");
8332 if (RExC_paren_names)
8333 he_str = hv_fetch_ent( RExC_paren_names, sv_name, 0, 0 );
8335 sv_dat = HeVAL(he_str);
8337 vFAIL("Reference to nonexistent named group");
8341 Perl_croak(aTHX_ "panic: bad flag %lx in reg_scan_name",
8342 (unsigned long) flags);
8344 NOT_REACHED; /* NOTREACHED */
8349 #define DEBUG_PARSE_MSG(funcname) DEBUG_PARSE_r({ \
8351 if (RExC_lastparse!=RExC_parse) { \
8352 Perl_re_printf( aTHX_ "%s", \
8353 Perl_pv_pretty(aTHX_ RExC_mysv1, RExC_parse, \
8354 RExC_end - RExC_parse, 16, \
8356 PERL_PV_ESCAPE_UNI_DETECT | \
8357 PERL_PV_PRETTY_ELLIPSES | \
8358 PERL_PV_PRETTY_LTGT | \
8359 PERL_PV_ESCAPE_RE | \
8360 PERL_PV_PRETTY_EXACTSIZE \
8364 Perl_re_printf( aTHX_ "%16s",""); \
8367 num = RExC_size + 1; \
8369 num=REG_NODE_NUM(RExC_emit); \
8370 if (RExC_lastnum!=num) \
8371 Perl_re_printf( aTHX_ "|%4d",num); \
8373 Perl_re_printf( aTHX_ "|%4s",""); \
8374 Perl_re_printf( aTHX_ "|%*s%-4s", \
8375 (int)((depth*2)), "", \
8379 RExC_lastparse=RExC_parse; \
8384 #define DEBUG_PARSE(funcname) DEBUG_PARSE_r({ \
8385 DEBUG_PARSE_MSG((funcname)); \
8386 Perl_re_printf( aTHX_ "%4s","\n"); \
8388 #define DEBUG_PARSE_FMT(funcname,fmt,args) DEBUG_PARSE_r({\
8389 DEBUG_PARSE_MSG((funcname)); \
8390 Perl_re_printf( aTHX_ fmt "\n",args); \
8393 /* This section of code defines the inversion list object and its methods. The
8394 * interfaces are highly subject to change, so as much as possible is static to
8395 * this file. An inversion list is here implemented as a malloc'd C UV array
8396 * as an SVt_INVLIST scalar.
8398 * An inversion list for Unicode is an array of code points, sorted by ordinal
8399 * number. Each element gives the code point that begins a range that extends
8400 * up-to but not including the code point given by the next element. The final
8401 * element gives the first code point of a range that extends to the platform's
8402 * infinity. The even-numbered elements (invlist[0], invlist[2], invlist[4],
8403 * ...) give ranges whose code points are all in the inversion list. We say
8404 * that those ranges are in the set. The odd-numbered elements give ranges
8405 * whose code points are not in the inversion list, and hence not in the set.
8406 * Thus, element [0] is the first code point in the list. Element [1]
8407 * is the first code point beyond that not in the list; and element [2] is the
8408 * first code point beyond that that is in the list. In other words, the first
8409 * range is invlist[0]..(invlist[1]-1), and all code points in that range are
8410 * in the inversion list. The second range is invlist[1]..(invlist[2]-1), and
8411 * all code points in that range are not in the inversion list. The third
8412 * range invlist[2]..(invlist[3]-1) gives code points that are in the inversion
8413 * list, and so forth. Thus every element whose index is divisible by two
8414 * gives the beginning of a range that is in the list, and every element whose
8415 * index is not divisible by two gives the beginning of a range not in the
8416 * list. If the final element's index is divisible by two, the inversion list
8417 * extends to the platform's infinity; otherwise the highest code point in the
8418 * inversion list is the contents of that element minus 1.
8420 * A range that contains just a single code point N will look like
8422 * invlist[i+1] == N+1
8424 * If N is UV_MAX (the highest representable code point on the machine), N+1 is
8425 * impossible to represent, so element [i+1] is omitted. The single element
8427 * invlist[0] == UV_MAX
8428 * contains just UV_MAX, but is interpreted as matching to infinity.
8430 * Taking the complement (inverting) an inversion list is quite simple, if the
8431 * first element is 0, remove it; otherwise add a 0 element at the beginning.
8432 * This implementation reserves an element at the beginning of each inversion
8433 * list to always contain 0; there is an additional flag in the header which
8434 * indicates if the list begins at the 0, or is offset to begin at the next
8435 * element. This means that the inversion list can be inverted without any
8436 * copying; just flip the flag.
8438 * More about inversion lists can be found in "Unicode Demystified"
8439 * Chapter 13 by Richard Gillam, published by Addison-Wesley.
8441 * The inversion list data structure is currently implemented as an SV pointing
8442 * to an array of UVs that the SV thinks are bytes. This allows us to have an
8443 * array of UV whose memory management is automatically handled by the existing
8444 * facilities for SV's.
8446 * Some of the methods should always be private to the implementation, and some
8447 * should eventually be made public */
8449 /* The header definitions are in F<invlist_inline.h> */
8451 #ifndef PERL_IN_XSUB_RE
8453 PERL_STATIC_INLINE UV*
8454 S__invlist_array_init(SV* const invlist, const bool will_have_0)
8456 /* Returns a pointer to the first element in the inversion list's array.
8457 * This is called upon initialization of an inversion list. Where the
8458 * array begins depends on whether the list has the code point U+0000 in it
8459 * or not. The other parameter tells it whether the code that follows this
8460 * call is about to put a 0 in the inversion list or not. The first
8461 * element is either the element reserved for 0, if TRUE, or the element
8462 * after it, if FALSE */
8464 bool* offset = get_invlist_offset_addr(invlist);
8465 UV* zero_addr = (UV *) SvPVX(invlist);
8467 PERL_ARGS_ASSERT__INVLIST_ARRAY_INIT;
8470 assert(! _invlist_len(invlist));
8474 /* 1^1 = 0; 1^0 = 1 */
8475 *offset = 1 ^ will_have_0;
8476 return zero_addr + *offset;
8481 PERL_STATIC_INLINE void
8482 S_invlist_set_len(pTHX_ SV* const invlist, const UV len, const bool offset)
8484 /* Sets the current number of elements stored in the inversion list.
8485 * Updates SvCUR correspondingly */
8486 PERL_UNUSED_CONTEXT;
8487 PERL_ARGS_ASSERT_INVLIST_SET_LEN;
8489 assert(SvTYPE(invlist) == SVt_INVLIST);
8494 : TO_INTERNAL_SIZE(len + offset));
8495 assert(SvLEN(invlist) == 0 || SvCUR(invlist) <= SvLEN(invlist));
8498 #ifndef PERL_IN_XSUB_RE
8501 S_invlist_replace_list_destroys_src(pTHX_ SV * dest, SV * src)
8503 /* Replaces the inversion list in 'dest' with the one from 'src'. It
8504 * steals the list from 'src', so 'src' is made to have a NULL list. This
8505 * is similar to what SvSetMagicSV() would do, if it were implemented on
8506 * inversion lists, though this routine avoids a copy */
8508 const UV src_len = _invlist_len(src);
8509 const bool src_offset = *get_invlist_offset_addr(src);
8510 const STRLEN src_byte_len = SvLEN(src);
8511 char * array = SvPVX(src);
8513 const int oldtainted = TAINT_get;
8515 PERL_ARGS_ASSERT_INVLIST_REPLACE_LIST_DESTROYS_SRC;
8517 assert(SvTYPE(src) == SVt_INVLIST);
8518 assert(SvTYPE(dest) == SVt_INVLIST);
8519 assert(! invlist_is_iterating(src));
8520 assert(SvCUR(src) == 0 || SvCUR(src) < SvLEN(src));
8522 /* Make sure it ends in the right place with a NUL, as our inversion list
8523 * manipulations aren't careful to keep this true, but sv_usepvn_flags()
8525 array[src_byte_len - 1] = '\0';
8527 TAINT_NOT; /* Otherwise it breaks */
8528 sv_usepvn_flags(dest,
8532 /* This flag is documented to cause a copy to be avoided */
8533 SV_HAS_TRAILING_NUL);
8534 TAINT_set(oldtainted);
8539 /* Finish up copying over the other fields in an inversion list */
8540 *get_invlist_offset_addr(dest) = src_offset;
8541 invlist_set_len(dest, src_len, src_offset);
8542 *get_invlist_previous_index_addr(dest) = 0;
8543 invlist_iterfinish(dest);
8546 PERL_STATIC_INLINE IV*
8547 S_get_invlist_previous_index_addr(SV* invlist)
8549 /* Return the address of the IV that is reserved to hold the cached index
8551 PERL_ARGS_ASSERT_GET_INVLIST_PREVIOUS_INDEX_ADDR;
8553 assert(SvTYPE(invlist) == SVt_INVLIST);
8555 return &(((XINVLIST*) SvANY(invlist))->prev_index);
8558 PERL_STATIC_INLINE IV
8559 S_invlist_previous_index(SV* const invlist)
8561 /* Returns cached index of previous search */
8563 PERL_ARGS_ASSERT_INVLIST_PREVIOUS_INDEX;
8565 return *get_invlist_previous_index_addr(invlist);
8568 PERL_STATIC_INLINE void
8569 S_invlist_set_previous_index(SV* const invlist, const IV index)
8571 /* Caches <index> for later retrieval */
8573 PERL_ARGS_ASSERT_INVLIST_SET_PREVIOUS_INDEX;
8575 assert(index == 0 || index < (int) _invlist_len(invlist));
8577 *get_invlist_previous_index_addr(invlist) = index;
8580 PERL_STATIC_INLINE void
8581 S_invlist_trim(SV* invlist)
8583 /* Free the not currently-being-used space in an inversion list */
8585 /* But don't free up the space needed for the 0 UV that is always at the
8586 * beginning of the list, nor the trailing NUL */
8587 const UV min_size = TO_INTERNAL_SIZE(1) + 1;
8589 PERL_ARGS_ASSERT_INVLIST_TRIM;
8591 assert(SvTYPE(invlist) == SVt_INVLIST);
8593 SvPV_renew(invlist, MAX(min_size, SvCUR(invlist) + 1));
8596 PERL_STATIC_INLINE void
8597 S_invlist_clear(pTHX_ SV* invlist) /* Empty the inversion list */
8599 PERL_ARGS_ASSERT_INVLIST_CLEAR;
8601 assert(SvTYPE(invlist) == SVt_INVLIST);
8603 invlist_set_len(invlist, 0, 0);
8604 invlist_trim(invlist);
8607 #endif /* ifndef PERL_IN_XSUB_RE */
8609 PERL_STATIC_INLINE bool
8610 S_invlist_is_iterating(SV* const invlist)
8612 PERL_ARGS_ASSERT_INVLIST_IS_ITERATING;
8614 return *(get_invlist_iter_addr(invlist)) < (STRLEN) UV_MAX;
8617 #ifndef PERL_IN_XSUB_RE
8619 PERL_STATIC_INLINE UV
8620 S_invlist_max(SV* const invlist)
8622 /* Returns the maximum number of elements storable in the inversion list's
8623 * array, without having to realloc() */
8625 PERL_ARGS_ASSERT_INVLIST_MAX;
8627 assert(SvTYPE(invlist) == SVt_INVLIST);
8629 /* Assumes worst case, in which the 0 element is not counted in the
8630 * inversion list, so subtracts 1 for that */
8631 return SvLEN(invlist) == 0 /* This happens under _new_invlist_C_array */
8632 ? FROM_INTERNAL_SIZE(SvCUR(invlist)) - 1
8633 : FROM_INTERNAL_SIZE(SvLEN(invlist)) - 1;
8636 Perl__new_invlist(pTHX_ IV initial_size)
8639 /* Return a pointer to a newly constructed inversion list, with enough
8640 * space to store 'initial_size' elements. If that number is negative, a
8641 * system default is used instead */
8645 if (initial_size < 0) {
8649 /* Allocate the initial space */
8650 new_list = newSV_type(SVt_INVLIST);
8652 /* First 1 is in case the zero element isn't in the list; second 1 is for
8654 SvGROW(new_list, TO_INTERNAL_SIZE(initial_size + 1) + 1);
8655 invlist_set_len(new_list, 0, 0);
8657 /* Force iterinit() to be used to get iteration to work */
8658 *get_invlist_iter_addr(new_list) = (STRLEN) UV_MAX;
8660 *get_invlist_previous_index_addr(new_list) = 0;
8666 Perl__new_invlist_C_array(pTHX_ const UV* const list)
8668 /* Return a pointer to a newly constructed inversion list, initialized to
8669 * point to <list>, which has to be in the exact correct inversion list
8670 * form, including internal fields. Thus this is a dangerous routine that
8671 * should not be used in the wrong hands. The passed in 'list' contains
8672 * several header fields at the beginning that are not part of the
8673 * inversion list body proper */
8675 const STRLEN length = (STRLEN) list[0];
8676 const UV version_id = list[1];
8677 const bool offset = cBOOL(list[2]);
8678 #define HEADER_LENGTH 3
8679 /* If any of the above changes in any way, you must change HEADER_LENGTH
8680 * (if appropriate) and regenerate INVLIST_VERSION_ID by running
8681 * perl -E 'say int(rand 2**31-1)'
8683 #define INVLIST_VERSION_ID 148565664 /* This is a combination of a version and
8684 data structure type, so that one being
8685 passed in can be validated to be an
8686 inversion list of the correct vintage.
8689 SV* invlist = newSV_type(SVt_INVLIST);
8691 PERL_ARGS_ASSERT__NEW_INVLIST_C_ARRAY;
8693 if (version_id != INVLIST_VERSION_ID) {
8694 Perl_croak(aTHX_ "panic: Incorrect version for previously generated inversion list");
8697 /* The generated array passed in includes header elements that aren't part
8698 * of the list proper, so start it just after them */
8699 SvPV_set(invlist, (char *) (list + HEADER_LENGTH));
8701 SvLEN_set(invlist, 0); /* Means we own the contents, and the system
8702 shouldn't touch it */
8704 *(get_invlist_offset_addr(invlist)) = offset;
8706 /* The 'length' passed to us is the physical number of elements in the
8707 * inversion list. But if there is an offset the logical number is one
8709 invlist_set_len(invlist, length - offset, offset);
8711 invlist_set_previous_index(invlist, 0);
8713 /* Initialize the iteration pointer. */
8714 invlist_iterfinish(invlist);
8716 SvREADONLY_on(invlist);
8722 S_invlist_extend(pTHX_ SV* const invlist, const UV new_max)
8724 /* Grow the maximum size of an inversion list */
8726 PERL_ARGS_ASSERT_INVLIST_EXTEND;
8728 assert(SvTYPE(invlist) == SVt_INVLIST);
8730 /* Add one to account for the zero element at the beginning which may not
8731 * be counted by the calling parameters */
8732 SvGROW((SV *)invlist, TO_INTERNAL_SIZE(new_max + 1));
8736 S__append_range_to_invlist(pTHX_ SV* const invlist,
8737 const UV start, const UV end)
8739 /* Subject to change or removal. Append the range from 'start' to 'end' at
8740 * the end of the inversion list. The range must be above any existing
8744 UV max = invlist_max(invlist);
8745 UV len = _invlist_len(invlist);
8748 PERL_ARGS_ASSERT__APPEND_RANGE_TO_INVLIST;
8750 if (len == 0) { /* Empty lists must be initialized */
8751 offset = start != 0;
8752 array = _invlist_array_init(invlist, ! offset);
8755 /* Here, the existing list is non-empty. The current max entry in the
8756 * list is generally the first value not in the set, except when the
8757 * set extends to the end of permissible values, in which case it is
8758 * the first entry in that final set, and so this call is an attempt to
8759 * append out-of-order */
8761 UV final_element = len - 1;
8762 array = invlist_array(invlist);
8763 if ( array[final_element] > start
8764 || ELEMENT_RANGE_MATCHES_INVLIST(final_element))
8766 Perl_croak(aTHX_ "panic: attempting to append to an inversion list, but wasn't at the end of the list, final=%" UVuf ", start=%" UVuf ", match=%c",
8767 array[final_element], start,
8768 ELEMENT_RANGE_MATCHES_INVLIST(final_element) ? 't' : 'f');
8771 /* Here, it is a legal append. If the new range begins 1 above the end
8772 * of the range below it, it is extending the range below it, so the
8773 * new first value not in the set is one greater than the newly
8774 * extended range. */
8775 offset = *get_invlist_offset_addr(invlist);
8776 if (array[final_element] == start) {
8777 if (end != UV_MAX) {
8778 array[final_element] = end + 1;
8781 /* But if the end is the maximum representable on the machine,
8782 * assume that infinity was actually what was meant. Just let
8783 * the range that this would extend to have no end */
8784 invlist_set_len(invlist, len - 1, offset);
8790 /* Here the new range doesn't extend any existing set. Add it */
8792 len += 2; /* Includes an element each for the start and end of range */
8794 /* If wll overflow the existing space, extend, which may cause the array to
8797 invlist_extend(invlist, len);
8799 /* Have to set len here to avoid assert failure in invlist_array() */
8800 invlist_set_len(invlist, len, offset);
8802 array = invlist_array(invlist);
8805 invlist_set_len(invlist, len, offset);
8808 /* The next item on the list starts the range, the one after that is
8809 * one past the new range. */
8810 array[len - 2] = start;
8811 if (end != UV_MAX) {
8812 array[len - 1] = end + 1;
8815 /* But if the end is the maximum representable on the machine, just let
8816 * the range have no end */
8817 invlist_set_len(invlist, len - 1, offset);
8822 Perl__invlist_search(SV* const invlist, const UV cp)
8824 /* Searches the inversion list for the entry that contains the input code
8825 * point <cp>. If <cp> is not in the list, -1 is returned. Otherwise, the
8826 * return value is the index into the list's array of the range that
8827 * contains <cp>, that is, 'i' such that
8828 * array[i] <= cp < array[i+1]
8833 IV high = _invlist_len(invlist);
8834 const IV highest_element = high - 1;
8837 PERL_ARGS_ASSERT__INVLIST_SEARCH;
8839 /* If list is empty, return failure. */
8844 /* (We can't get the array unless we know the list is non-empty) */
8845 array = invlist_array(invlist);
8847 mid = invlist_previous_index(invlist);
8849 if (mid > highest_element) {
8850 mid = highest_element;
8853 /* <mid> contains the cache of the result of the previous call to this
8854 * function (0 the first time). See if this call is for the same result,
8855 * or if it is for mid-1. This is under the theory that calls to this
8856 * function will often be for related code points that are near each other.
8857 * And benchmarks show that caching gives better results. We also test
8858 * here if the code point is within the bounds of the list. These tests
8859 * replace others that would have had to be made anyway to make sure that
8860 * the array bounds were not exceeded, and these give us extra information
8861 * at the same time */
8862 if (cp >= array[mid]) {
8863 if (cp >= array[highest_element]) {
8864 return highest_element;
8867 /* Here, array[mid] <= cp < array[highest_element]. This means that
8868 * the final element is not the answer, so can exclude it; it also
8869 * means that <mid> is not the final element, so can refer to 'mid + 1'
8871 if (cp < array[mid + 1]) {
8877 else { /* cp < aray[mid] */
8878 if (cp < array[0]) { /* Fail if outside the array */
8882 if (cp >= array[mid - 1]) {
8887 /* Binary search. What we are looking for is <i> such that
8888 * array[i] <= cp < array[i+1]
8889 * The loop below converges on the i+1. Note that there may not be an
8890 * (i+1)th element in the array, and things work nonetheless */
8891 while (low < high) {
8892 mid = (low + high) / 2;
8893 assert(mid <= highest_element);
8894 if (array[mid] <= cp) { /* cp >= array[mid] */
8897 /* We could do this extra test to exit the loop early.
8898 if (cp < array[low]) {
8903 else { /* cp < array[mid] */
8910 invlist_set_previous_index(invlist, high);
8915 Perl__invlist_populate_swatch(SV* const invlist,
8916 const UV start, const UV end, U8* swatch)
8918 /* populates a swatch of a swash the same way swatch_get() does in utf8.c,
8919 * but is used when the swash has an inversion list. This makes this much
8920 * faster, as it uses a binary search instead of a linear one. This is
8921 * intimately tied to that function, and perhaps should be in utf8.c,
8922 * except it is intimately tied to inversion lists as well. It assumes
8923 * that <swatch> is all 0's on input */
8926 const IV len = _invlist_len(invlist);
8930 PERL_ARGS_ASSERT__INVLIST_POPULATE_SWATCH;
8932 if (len == 0) { /* Empty inversion list */
8936 array = invlist_array(invlist);
8938 /* Find which element it is */
8939 i = _invlist_search(invlist, start);
8941 /* We populate from <start> to <end> */
8942 while (current < end) {
8945 /* The inversion list gives the results for every possible code point
8946 * after the first one in the list. Only those ranges whose index is
8947 * even are ones that the inversion list matches. For the odd ones,
8948 * and if the initial code point is not in the list, we have to skip
8949 * forward to the next element */
8950 if (i == -1 || ! ELEMENT_RANGE_MATCHES_INVLIST(i)) {
8952 if (i >= len) { /* Finished if beyond the end of the array */
8956 if (current >= end) { /* Finished if beyond the end of what we
8958 if (LIKELY(end < UV_MAX)) {
8962 /* We get here when the upper bound is the maximum
8963 * representable on the machine, and we are looking for just
8964 * that code point. Have to special case it */
8966 goto join_end_of_list;
8969 assert(current >= start);
8971 /* The current range ends one below the next one, except don't go past
8974 upper = (i < len && array[i] < end) ? array[i] : end;
8976 /* Here we are in a range that matches. Populate a bit in the 3-bit U8
8977 * for each code point in it */
8978 for (; current < upper; current++) {
8979 const STRLEN offset = (STRLEN)(current - start);
8980 swatch[offset >> 3] |= 1 << (offset & 7);
8985 /* Quit if at the end of the list */
8988 /* But first, have to deal with the highest possible code point on
8989 * the platform. The previous code assumes that <end> is one
8990 * beyond where we want to populate, but that is impossible at the
8991 * platform's infinity, so have to handle it specially */
8992 if (UNLIKELY(end == UV_MAX && ELEMENT_RANGE_MATCHES_INVLIST(len-1)))
8994 const STRLEN offset = (STRLEN)(end - start);
8995 swatch[offset >> 3] |= 1 << (offset & 7);
9000 /* Advance to the next range, which will be for code points not in the
9009 Perl__invlist_union_maybe_complement_2nd(pTHX_ SV* const a, SV* const b,
9010 const bool complement_b, SV** output)
9012 /* Take the union of two inversion lists and point '*output' to it. On
9013 * input, '*output' MUST POINT TO NULL OR TO AN SV* INVERSION LIST (possibly
9014 * even 'a' or 'b'). If to an inversion list, the contents of the original
9015 * list will be replaced by the union. The first list, 'a', may be
9016 * NULL, in which case a copy of the second list is placed in '*output'.
9017 * If 'complement_b' is TRUE, the union is taken of the complement
9018 * (inversion) of 'b' instead of b itself.
9020 * The basis for this comes from "Unicode Demystified" Chapter 13 by
9021 * Richard Gillam, published by Addison-Wesley, and explained at some
9022 * length there. The preface says to incorporate its examples into your
9023 * code at your own risk.
9025 * The algorithm is like a merge sort. */
9027 const UV* array_a; /* a's array */
9029 UV len_a; /* length of a's array */
9032 SV* u; /* the resulting union */
9036 UV i_a = 0; /* current index into a's array */
9040 /* running count, as explained in the algorithm source book; items are
9041 * stopped accumulating and are output when the count changes to/from 0.
9042 * The count is incremented when we start a range that's in an input's set,
9043 * and decremented when we start a range that's not in a set. So this
9044 * variable can be 0, 1, or 2. When it is 0 neither input is in their set,
9045 * and hence nothing goes into the union; 1, just one of the inputs is in
9046 * its set (and its current range gets added to the union); and 2 when both
9047 * inputs are in their sets. */
9050 PERL_ARGS_ASSERT__INVLIST_UNION_MAYBE_COMPLEMENT_2ND;
9052 assert(*output == NULL || SvTYPE(*output) == SVt_INVLIST);
9054 len_b = _invlist_len(b);
9057 /* Here, 'b' is empty, hence it's complement is all possible code
9058 * points. So if the union includes the complement of 'b', it includes
9059 * everything, and we need not even look at 'a'. It's easiest to
9060 * create a new inversion list that matches everything. */
9062 SV* everything = _add_range_to_invlist(NULL, 0, UV_MAX);
9064 if (*output == NULL) { /* If the output didn't exist, just point it
9066 *output = everything;
9068 else { /* Otherwise, replace its contents with the new list */
9069 invlist_replace_list_destroys_src(*output, everything);
9070 SvREFCNT_dec_NN(everything);
9076 /* Here, we don't want the complement of 'b', and since 'b' is empty,
9077 * the union will come entirely from 'a'. If 'a' is NULL or empty, the
9078 * output will be empty */
9080 if (a == NULL || _invlist_len(a) == 0) {
9081 if (*output == NULL) {
9082 *output = _new_invlist(0);
9085 invlist_clear(*output);
9090 /* Here, 'a' is not empty, but 'b' is, so 'a' entirely determines the
9091 * union. We can just return a copy of 'a' if '*output' doesn't point
9092 * to an existing list */
9093 if (*output == NULL) {
9094 *output = invlist_clone(a);
9098 /* If the output is to overwrite 'a', we have a no-op, as it's
9104 /* Here, '*output' is to be overwritten by 'a' */
9105 u = invlist_clone(a);
9106 invlist_replace_list_destroys_src(*output, u);
9112 /* Here 'b' is not empty. See about 'a' */
9114 if (a == NULL || ((len_a = _invlist_len(a)) == 0)) {
9116 /* Here, 'a' is empty (and b is not). That means the union will come
9117 * entirely from 'b'. If '*output' is NULL, we can directly return a
9118 * clone of 'b'. Otherwise, we replace the contents of '*output' with
9121 SV ** dest = (*output == NULL) ? output : &u;
9122 *dest = invlist_clone(b);
9124 _invlist_invert(*dest);
9128 invlist_replace_list_destroys_src(*output, u);
9135 /* Here both lists exist and are non-empty */
9136 array_a = invlist_array(a);
9137 array_b = invlist_array(b);
9139 /* If are to take the union of 'a' with the complement of b, set it
9140 * up so are looking at b's complement. */
9143 /* To complement, we invert: if the first element is 0, remove it. To
9144 * do this, we just pretend the array starts one later */
9145 if (array_b[0] == 0) {
9151 /* But if the first element is not zero, we pretend the list starts
9152 * at the 0 that is always stored immediately before the array. */
9158 /* Size the union for the worst case: that the sets are completely
9160 u = _new_invlist(len_a + len_b);
9162 /* Will contain U+0000 if either component does */
9163 array_u = _invlist_array_init(u, ( len_a > 0 && array_a[0] == 0)
9164 || (len_b > 0 && array_b[0] == 0));
9166 /* Go through each input list item by item, stopping when have exhausted
9168 while (i_a < len_a && i_b < len_b) {
9169 UV cp; /* The element to potentially add to the union's array */
9170 bool cp_in_set; /* is it in the the input list's set or not */
9172 /* We need to take one or the other of the two inputs for the union.
9173 * Since we are merging two sorted lists, we take the smaller of the
9174 * next items. In case of a tie, we take first the one that is in its
9175 * set. If we first took the one not in its set, it would decrement
9176 * the count, possibly to 0 which would cause it to be output as ending
9177 * the range, and the next time through we would take the same number,
9178 * and output it again as beginning the next range. By doing it the
9179 * opposite way, there is no possibility that the count will be
9180 * momentarily decremented to 0, and thus the two adjoining ranges will
9181 * be seamlessly merged. (In a tie and both are in the set or both not
9182 * in the set, it doesn't matter which we take first.) */
9183 if ( array_a[i_a] < array_b[i_b]
9184 || ( array_a[i_a] == array_b[i_b]
9185 && ELEMENT_RANGE_MATCHES_INVLIST(i_a)))
9187 cp_in_set = ELEMENT_RANGE_MATCHES_INVLIST(i_a);
9188 cp = array_a[i_a++];
9191 cp_in_set = ELEMENT_RANGE_MATCHES_INVLIST(i_b);
9192 cp = array_b[i_b++];
9195 /* Here, have chosen which of the two inputs to look at. Only output
9196 * if the running count changes to/from 0, which marks the
9197 * beginning/end of a range that's in the set */
9200 array_u[i_u++] = cp;
9207 array_u[i_u++] = cp;
9213 /* The loop above increments the index into exactly one of the input lists
9214 * each iteration, and ends when either index gets to its list end. That
9215 * means the other index is lower than its end, and so something is
9216 * remaining in that one. We decrement 'count', as explained below, if
9217 * that list is in its set. (i_a and i_b each currently index the element
9218 * beyond the one we care about.) */
9219 if ( (i_a != len_a && PREV_RANGE_MATCHES_INVLIST(i_a))
9220 || (i_b != len_b && PREV_RANGE_MATCHES_INVLIST(i_b)))
9225 /* Above we decremented 'count' if the list that had unexamined elements in
9226 * it was in its set. This has made it so that 'count' being non-zero
9227 * means there isn't anything left to output; and 'count' equal to 0 means
9228 * that what is left to output is precisely that which is left in the
9229 * non-exhausted input list.
9231 * To see why, note first that the exhausted input obviously has nothing
9232 * left to add to the union. If it was in its set at its end, that means
9233 * the set extends from here to the platform's infinity, and hence so does
9234 * the union and the non-exhausted set is irrelevant. The exhausted set
9235 * also contributed 1 to 'count'. If 'count' was 2, it got decremented to
9236 * 1, but if it was 1, the non-exhausted set wasn't in its set, and so
9237 * 'count' remains at 1. This is consistent with the decremented 'count'
9238 * != 0 meaning there's nothing left to add to the union.
9240 * But if the exhausted input wasn't in its set, it contributed 0 to
9241 * 'count', and the rest of the union will be whatever the other input is.
9242 * If 'count' was 0, neither list was in its set, and 'count' remains 0;
9243 * otherwise it gets decremented to 0. This is consistent with 'count'
9244 * == 0 meaning the remainder of the union is whatever is left in the
9245 * non-exhausted list. */
9250 IV copy_count = len_a - i_a;
9251 if (copy_count > 0) { /* The non-exhausted input is 'a' */
9252 Copy(array_a + i_a, array_u + i_u, copy_count, UV);
9254 else { /* The non-exhausted input is b */
9255 copy_count = len_b - i_b;
9256 Copy(array_b + i_b, array_u + i_u, copy_count, UV);
9258 len_u = i_u + copy_count;
9261 /* Set the result to the final length, which can change the pointer to
9262 * array_u, so re-find it. (Note that it is unlikely that this will
9263 * change, as we are shrinking the space, not enlarging it) */
9264 if (len_u != _invlist_len(u)) {
9265 invlist_set_len(u, len_u, *get_invlist_offset_addr(u));
9267 array_u = invlist_array(u);
9270 if (*output == NULL) { /* Simply return the new inversion list */
9274 /* Otherwise, overwrite the inversion list that was in '*output'. We
9275 * could instead free '*output', and then set it to 'u', but experience
9276 * has shown [perl #127392] that if the input is a mortal, we can get a
9277 * huge build-up of these during regex compilation before they get
9279 invlist_replace_list_destroys_src(*output, u);
9287 Perl__invlist_intersection_maybe_complement_2nd(pTHX_ SV* const a, SV* const b,
9288 const bool complement_b, SV** i)
9290 /* Take the intersection of two inversion lists and point '*i' to it. On
9291 * input, '*i' MUST POINT TO NULL OR TO AN SV* INVERSION LIST (possibly
9292 * even 'a' or 'b'). If to an inversion list, the contents of the original
9293 * list will be replaced by the intersection. The first list, 'a', may be
9294 * NULL, in which case '*i' will be an empty list. If 'complement_b' is
9295 * TRUE, the result will be the intersection of 'a' and the complement (or
9296 * inversion) of 'b' instead of 'b' directly.
9298 * The basis for this comes from "Unicode Demystified" Chapter 13 by
9299 * Richard Gillam, published by Addison-Wesley, and explained at some
9300 * length there. The preface says to incorporate its examples into your
9301 * code at your own risk. In fact, it had bugs
9303 * The algorithm is like a merge sort, and is essentially the same as the
9307 const UV* array_a; /* a's array */
9309 UV len_a; /* length of a's array */
9312 SV* r; /* the resulting intersection */
9316 UV i_a = 0; /* current index into a's array */
9320 /* running count of how many of the two inputs are postitioned at ranges
9321 * that are in their sets. As explained in the algorithm source book,
9322 * items are stopped accumulating and are output when the count changes
9323 * to/from 2. The count is incremented when we start a range that's in an
9324 * input's set, and decremented when we start a range that's not in a set.
9325 * Only when it is 2 are we in the intersection. */
9328 PERL_ARGS_ASSERT__INVLIST_INTERSECTION_MAYBE_COMPLEMENT_2ND;
9330 assert(*i == NULL || SvTYPE(*i) == SVt_INVLIST);
9332 /* Special case if either one is empty */
9333 len_a = (a == NULL) ? 0 : _invlist_len(a);
9334 if ((len_a == 0) || ((len_b = _invlist_len(b)) == 0)) {
9335 if (len_a != 0 && complement_b) {
9337 /* Here, 'a' is not empty, therefore from the enclosing 'if', 'b'
9338 * must be empty. Here, also we are using 'b's complement, which
9339 * hence must be every possible code point. Thus the intersection
9342 if (*i == a) { /* No-op */
9347 *i = invlist_clone(a);
9351 r = invlist_clone(a);
9352 invlist_replace_list_destroys_src(*i, r);
9357 /* Here, 'a' or 'b' is empty and not using the complement of 'b'. The
9358 * intersection must be empty */
9360 *i = _new_invlist(0);
9368 /* Here both lists exist and are non-empty */
9369 array_a = invlist_array(a);
9370 array_b = invlist_array(b);
9372 /* If are to take the intersection of 'a' with the complement of b, set it
9373 * up so are looking at b's complement. */
9376 /* To complement, we invert: if the first element is 0, remove it. To
9377 * do this, we just pretend the array starts one later */
9378 if (array_b[0] == 0) {
9384 /* But if the first element is not zero, we pretend the list starts
9385 * at the 0 that is always stored immediately before the array. */
9391 /* Size the intersection for the worst case: that the intersection ends up
9392 * fragmenting everything to be completely disjoint */
9393 r= _new_invlist(len_a + len_b);
9395 /* Will contain U+0000 iff both components do */
9396 array_r = _invlist_array_init(r, len_a > 0 && array_a[0] == 0
9397 && len_b > 0 && array_b[0] == 0);
9399 /* Go through each list item by item, stopping when have exhausted one of
9401 while (i_a < len_a && i_b < len_b) {
9402 UV cp; /* The element to potentially add to the intersection's
9404 bool cp_in_set; /* Is it in the input list's set or not */
9406 /* We need to take one or the other of the two inputs for the
9407 * intersection. Since we are merging two sorted lists, we take the
9408 * smaller of the next items. In case of a tie, we take first the one
9409 * that is not in its set (a difference from the union algorithm). If
9410 * we first took the one in its set, it would increment the count,
9411 * possibly to 2 which would cause it to be output as starting a range
9412 * in the intersection, and the next time through we would take that
9413 * same number, and output it again as ending the set. By doing the
9414 * opposite of this, there is no possibility that the count will be
9415 * momentarily incremented to 2. (In a tie and both are in the set or
9416 * both not in the set, it doesn't matter which we take first.) */
9417 if ( array_a[i_a] < array_b[i_b]
9418 || ( array_a[i_a] == array_b[i_b]
9419 && ! ELEMENT_RANGE_MATCHES_INVLIST(i_a)))
9421 cp_in_set = ELEMENT_RANGE_MATCHES_INVLIST(i_a);
9422 cp = array_a[i_a++];
9425 cp_in_set = ELEMENT_RANGE_MATCHES_INVLIST(i_b);
9429 /* Here, have chosen which of the two inputs to look at. Only output
9430 * if the running count changes to/from 2, which marks the
9431 * beginning/end of a range that's in the intersection */
9435 array_r[i_r++] = cp;
9440 array_r[i_r++] = cp;
9447 /* The loop above increments the index into exactly one of the input lists
9448 * each iteration, and ends when either index gets to its list end. That
9449 * means the other index is lower than its end, and so something is
9450 * remaining in that one. We increment 'count', as explained below, if the
9451 * exhausted list was in its set. (i_a and i_b each currently index the
9452 * element beyond the one we care about.) */
9453 if ( (i_a == len_a && PREV_RANGE_MATCHES_INVLIST(i_a))
9454 || (i_b == len_b && PREV_RANGE_MATCHES_INVLIST(i_b)))
9459 /* Above we incremented 'count' if the exhausted list was in its set. This
9460 * has made it so that 'count' being below 2 means there is nothing left to
9461 * output; otheriwse what's left to add to the intersection is precisely
9462 * that which is left in the non-exhausted input list.
9464 * To see why, note first that the exhausted input obviously has nothing
9465 * left to affect the intersection. If it was in its set at its end, that
9466 * means the set extends from here to the platform's infinity, and hence
9467 * anything in the non-exhausted's list will be in the intersection, and
9468 * anything not in it won't be. Hence, the rest of the intersection is
9469 * precisely what's in the non-exhausted list The exhausted set also
9470 * contributed 1 to 'count', meaning 'count' was at least 1. Incrementing
9471 * it means 'count' is now at least 2. This is consistent with the
9472 * incremented 'count' being >= 2 means to add the non-exhausted list to
9475 * But if the exhausted input wasn't in its set, it contributed 0 to
9476 * 'count', and the intersection can't include anything further; the
9477 * non-exhausted set is irrelevant. 'count' was at most 1, and doesn't get
9478 * incremented. This is consistent with 'count' being < 2 meaning nothing
9479 * further to add to the intersection. */
9480 if (count < 2) { /* Nothing left to put in the intersection. */
9483 else { /* copy the non-exhausted list, unchanged. */
9484 IV copy_count = len_a - i_a;
9485 if (copy_count > 0) { /* a is the one with stuff left */
9486 Copy(array_a + i_a, array_r + i_r, copy_count, UV);
9488 else { /* b is the one with stuff left */
9489 copy_count = len_b - i_b;
9490 Copy(array_b + i_b, array_r + i_r, copy_count, UV);
9492 len_r = i_r + copy_count;
9495 /* Set the result to the final length, which can change the pointer to
9496 * array_r, so re-find it. (Note that it is unlikely that this will
9497 * change, as we are shrinking the space, not enlarging it) */
9498 if (len_r != _invlist_len(r)) {
9499 invlist_set_len(r, len_r, *get_invlist_offset_addr(r));
9501 array_r = invlist_array(r);
9504 if (*i == NULL) { /* Simply return the calculated intersection */
9507 else { /* Otherwise, replace the existing inversion list in '*i'. We could
9508 instead free '*i', and then set it to 'r', but experience has
9509 shown [perl #127392] that if the input is a mortal, we can get a
9510 huge build-up of these during regex compilation before they get
9513 invlist_replace_list_destroys_src(*i, r);
9525 Perl__add_range_to_invlist(pTHX_ SV* invlist, UV start, UV end)
9527 /* Add the range from 'start' to 'end' inclusive to the inversion list's
9528 * set. A pointer to the inversion list is returned. This may actually be
9529 * a new list, in which case the passed in one has been destroyed. The
9530 * passed-in inversion list can be NULL, in which case a new one is created
9531 * with just the one range in it. The new list is not necessarily
9532 * NUL-terminated. Space is not freed if the inversion list shrinks as a
9533 * result of this function. The gain would not be large, and in many
9534 * cases, this is called multiple times on a single inversion list, so
9535 * anything freed may almost immediately be needed again.
9537 * This used to mostly call the 'union' routine, but that is much more
9538 * heavyweight than really needed for a single range addition */
9540 UV* array; /* The array implementing the inversion list */
9541 UV len; /* How many elements in 'array' */
9542 SSize_t i_s; /* index into the invlist array where 'start'
9544 SSize_t i_e = 0; /* And the index where 'end' should go */
9545 UV cur_highest; /* The highest code point in the inversion list
9546 upon entry to this function */
9548 /* This range becomes the whole inversion list if none already existed */
9549 if (invlist == NULL) {
9550 invlist = _new_invlist(2);
9551 _append_range_to_invlist(invlist, start, end);
9555 /* Likewise, if the inversion list is currently empty */
9556 len = _invlist_len(invlist);
9558 _append_range_to_invlist(invlist, start, end);
9562 /* Starting here, we have to know the internals of the list */
9563 array = invlist_array(invlist);
9565 /* If the new range ends higher than the current highest ... */
9566 cur_highest = invlist_highest(invlist);
9567 if (end > cur_highest) {
9569 /* If the whole range is higher, we can just append it */
9570 if (start > cur_highest) {
9571 _append_range_to_invlist(invlist, start, end);
9575 /* Otherwise, add the portion that is higher ... */
9576 _append_range_to_invlist(invlist, cur_highest + 1, end);
9578 /* ... and continue on below to handle the rest. As a result of the
9579 * above append, we know that the index of the end of the range is the
9580 * final even numbered one of the array. Recall that the final element
9581 * always starts a range that extends to infinity. If that range is in
9582 * the set (meaning the set goes from here to infinity), it will be an
9583 * even index, but if it isn't in the set, it's odd, and the final
9584 * range in the set is one less, which is even. */
9585 if (end == UV_MAX) {
9593 /* We have dealt with appending, now see about prepending. If the new
9594 * range starts lower than the current lowest ... */
9595 if (start < array[0]) {
9597 /* Adding something which has 0 in it is somewhat tricky, and uncommon.
9598 * Let the union code handle it, rather than having to know the
9599 * trickiness in two code places. */
9600 if (UNLIKELY(start == 0)) {
9603 range_invlist = _new_invlist(2);
9604 _append_range_to_invlist(range_invlist, start, end);
9606 _invlist_union(invlist, range_invlist, &invlist);
9608 SvREFCNT_dec_NN(range_invlist);
9613 /* If the whole new range comes before the first entry, and doesn't
9614 * extend it, we have to insert it as an additional range */
9615 if (end < array[0] - 1) {
9617 goto splice_in_new_range;
9620 /* Here the new range adjoins the existing first range, extending it
9624 /* And continue on below to handle the rest. We know that the index of
9625 * the beginning of the range is the first one of the array */
9628 else { /* Not prepending any part of the new range to the existing list.
9629 * Find where in the list it should go. This finds i_s, such that:
9630 * invlist[i_s] <= start < array[i_s+1]
9632 i_s = _invlist_search(invlist, start);
9635 /* At this point, any extending before the beginning of the inversion list
9636 * and/or after the end has been done. This has made it so that, in the
9637 * code below, each endpoint of the new range is either in a range that is
9638 * in the set, or is in a gap between two ranges that are. This means we
9639 * don't have to worry about exceeding the array bounds.
9641 * Find where in the list the new range ends (but we can skip this if we
9642 * have already determined what it is, or if it will be the same as i_s,
9643 * which we already have computed) */
9645 i_e = (start == end)
9647 : _invlist_search(invlist, end);
9650 /* Here generally invlist[i_e] <= end < array[i_e+1]. But if invlist[i_e]
9651 * is a range that goes to infinity there is no element at invlist[i_e+1],
9652 * so only the first relation holds. */
9654 if ( ! ELEMENT_RANGE_MATCHES_INVLIST(i_s)) {
9656 /* Here, the ranges on either side of the beginning of the new range
9657 * are in the set, and this range starts in the gap between them.
9659 * The new range extends the range above it downwards if the new range
9660 * ends at or above that range's start */
9661 const bool extends_the_range_above = ( end == UV_MAX
9662 || end + 1 >= array[i_s+1]);
9664 /* The new range extends the range below it upwards if it begins just
9665 * after where that range ends */
9666 if (start == array[i_s]) {
9668 /* If the new range fills the entire gap between the other ranges,
9669 * they will get merged together. Other ranges may also get
9670 * merged, depending on how many of them the new range spans. In
9671 * the general case, we do the merge later, just once, after we
9672 * figure out how many to merge. But in the case where the new
9673 * range exactly spans just this one gap (possibly extending into
9674 * the one above), we do the merge here, and an early exit. This
9675 * is done here to avoid having to special case later. */
9676 if (i_e - i_s <= 1) {
9678 /* If i_e - i_s == 1, it means that the new range terminates
9679 * within the range above, and hence 'extends_the_range_above'
9680 * must be true. (If the range above it extends to infinity,
9681 * 'i_s+2' will be above the array's limit, but 'len-i_s-2'
9682 * will be 0, so no harm done.) */
9683 if (extends_the_range_above) {
9684 Move(array + i_s + 2, array + i_s, len - i_s - 2, UV);
9685 invlist_set_len(invlist,
9687 *(get_invlist_offset_addr(invlist)));
9691 /* Here, i_e must == i_s. We keep them in sync, as they apply
9692 * to the same range, and below we are about to decrement i_s
9697 /* Here, the new range is adjacent to the one below. (It may also
9698 * span beyond the range above, but that will get resolved later.)
9699 * Extend the range below to include this one. */
9700 array[i_s] = (end == UV_MAX) ? UV_MAX : end + 1;
9704 else if (extends_the_range_above) {
9706 /* Here the new range only extends the range above it, but not the
9707 * one below. It merges with the one above. Again, we keep i_e
9708 * and i_s in sync if they point to the same range */
9717 /* Here, we've dealt with the new range start extending any adjoining
9720 * If the new range extends to infinity, it is now the final one,
9721 * regardless of what was there before */
9722 if (UNLIKELY(end == UV_MAX)) {
9723 invlist_set_len(invlist, i_s + 1, *(get_invlist_offset_addr(invlist)));
9727 /* If i_e started as == i_s, it has also been dealt with,
9728 * and been updated to the new i_s, which will fail the following if */
9729 if (! ELEMENT_RANGE_MATCHES_INVLIST(i_e)) {
9731 /* Here, the ranges on either side of the end of the new range are in
9732 * the set, and this range ends in the gap between them.
9734 * If this range is adjacent to (hence extends) the range above it, it
9735 * becomes part of that range; likewise if it extends the range below,
9736 * it becomes part of that range */
9737 if (end + 1 == array[i_e+1]) {
9741 else if (start <= array[i_e]) {
9742 array[i_e] = end + 1;
9749 /* If the range fits entirely in an existing range (as possibly already
9750 * extended above), it doesn't add anything new */
9751 if (ELEMENT_RANGE_MATCHES_INVLIST(i_s)) {
9755 /* Here, no part of the range is in the list. Must add it. It will
9756 * occupy 2 more slots */
9757 splice_in_new_range:
9759 invlist_extend(invlist, len + 2);
9760 array = invlist_array(invlist);
9761 /* Move the rest of the array down two slots. Don't include any
9763 Move(array + i_e + 1, array + i_e + 3, len - i_e - 1, UV);
9765 /* Do the actual splice */
9766 array[i_e+1] = start;
9767 array[i_e+2] = end + 1;
9768 invlist_set_len(invlist, len + 2, *(get_invlist_offset_addr(invlist)));
9772 /* Here the new range crossed the boundaries of a pre-existing range. The
9773 * code above has adjusted things so that both ends are in ranges that are
9774 * in the set. This means everything in between must also be in the set.
9775 * Just squash things together */
9776 Move(array + i_e + 1, array + i_s + 1, len - i_e - 1, UV);
9777 invlist_set_len(invlist,
9779 *(get_invlist_offset_addr(invlist)));
9785 Perl__setup_canned_invlist(pTHX_ const STRLEN size, const UV element0,
9786 UV** other_elements_ptr)
9788 /* Create and return an inversion list whose contents are to be populated
9789 * by the caller. The caller gives the number of elements (in 'size') and
9790 * the very first element ('element0'). This function will set
9791 * '*other_elements_ptr' to an array of UVs, where the remaining elements
9794 * Obviously there is some trust involved that the caller will properly
9795 * fill in the other elements of the array.
9797 * (The first element needs to be passed in, as the underlying code does
9798 * things differently depending on whether it is zero or non-zero) */
9800 SV* invlist = _new_invlist(size);
9803 PERL_ARGS_ASSERT__SETUP_CANNED_INVLIST;
9805 invlist = add_cp_to_invlist(invlist, element0);
9806 offset = *get_invlist_offset_addr(invlist);
9808 invlist_set_len(invlist, size, offset);
9809 *other_elements_ptr = invlist_array(invlist) + 1;
9815 PERL_STATIC_INLINE SV*
9816 S_add_cp_to_invlist(pTHX_ SV* invlist, const UV cp) {
9817 return _add_range_to_invlist(invlist, cp, cp);
9820 #ifndef PERL_IN_XSUB_RE
9822 Perl__invlist_invert(pTHX_ SV* const invlist)
9824 /* Complement the input inversion list. This adds a 0 if the list didn't
9825 * have a zero; removes it otherwise. As described above, the data
9826 * structure is set up so that this is very efficient */
9828 PERL_ARGS_ASSERT__INVLIST_INVERT;
9830 assert(! invlist_is_iterating(invlist));
9832 /* The inverse of matching nothing is matching everything */
9833 if (_invlist_len(invlist) == 0) {
9834 _append_range_to_invlist(invlist, 0, UV_MAX);
9838 *get_invlist_offset_addr(invlist) = ! *get_invlist_offset_addr(invlist);
9843 PERL_STATIC_INLINE SV*
9844 S_invlist_clone(pTHX_ SV* const invlist)
9847 /* Return a new inversion list that is a copy of the input one, which is
9848 * unchanged. The new list will not be mortal even if the old one was. */
9850 /* Need to allocate extra space to accommodate Perl's addition of a
9851 * trailing NUL to SvPV's, since it thinks they are always strings */
9852 SV* new_invlist = _new_invlist(_invlist_len(invlist) + 1);
9853 STRLEN physical_length = SvCUR(invlist);
9854 bool offset = *(get_invlist_offset_addr(invlist));
9856 PERL_ARGS_ASSERT_INVLIST_CLONE;
9858 *(get_invlist_offset_addr(new_invlist)) = offset;
9859 invlist_set_len(new_invlist, _invlist_len(invlist), offset);
9860 Copy(SvPVX(invlist), SvPVX(new_invlist), physical_length, char);
9865 PERL_STATIC_INLINE STRLEN*
9866 S_get_invlist_iter_addr(SV* invlist)
9868 /* Return the address of the UV that contains the current iteration
9871 PERL_ARGS_ASSERT_GET_INVLIST_ITER_ADDR;
9873 assert(SvTYPE(invlist) == SVt_INVLIST);
9875 return &(((XINVLIST*) SvANY(invlist))->iterator);
9878 PERL_STATIC_INLINE void
9879 S_invlist_iterinit(SV* invlist) /* Initialize iterator for invlist */
9881 PERL_ARGS_ASSERT_INVLIST_ITERINIT;
9883 *get_invlist_iter_addr(invlist) = 0;
9886 PERL_STATIC_INLINE void
9887 S_invlist_iterfinish(SV* invlist)
9889 /* Terminate iterator for invlist. This is to catch development errors.
9890 * Any iteration that is interrupted before completed should call this
9891 * function. Functions that add code points anywhere else but to the end
9892 * of an inversion list assert that they are not in the middle of an
9893 * iteration. If they were, the addition would make the iteration
9894 * problematical: if the iteration hadn't reached the place where things
9895 * were being added, it would be ok */
9897 PERL_ARGS_ASSERT_INVLIST_ITERFINISH;
9899 *get_invlist_iter_addr(invlist) = (STRLEN) UV_MAX;
9903 S_invlist_iternext(SV* invlist, UV* start, UV* end)
9905 /* An C<invlist_iterinit> call on <invlist> must be used to set this up.
9906 * This call sets in <*start> and <*end>, the next range in <invlist>.
9907 * Returns <TRUE> if successful and the next call will return the next
9908 * range; <FALSE> if was already at the end of the list. If the latter,
9909 * <*start> and <*end> are unchanged, and the next call to this function
9910 * will start over at the beginning of the list */
9912 STRLEN* pos = get_invlist_iter_addr(invlist);
9913 UV len = _invlist_len(invlist);
9916 PERL_ARGS_ASSERT_INVLIST_ITERNEXT;
9919 *pos = (STRLEN) UV_MAX; /* Force iterinit() to be required next time */
9923 array = invlist_array(invlist);
9925 *start = array[(*pos)++];
9931 *end = array[(*pos)++] - 1;
9937 PERL_STATIC_INLINE UV
9938 S_invlist_highest(SV* const invlist)
9940 /* Returns the highest code point that matches an inversion list. This API
9941 * has an ambiguity, as it returns 0 under either the highest is actually
9942 * 0, or if the list is empty. If this distinction matters to you, check
9943 * for emptiness before calling this function */
9945 UV len = _invlist_len(invlist);
9948 PERL_ARGS_ASSERT_INVLIST_HIGHEST;
9954 array = invlist_array(invlist);
9956 /* The last element in the array in the inversion list always starts a
9957 * range that goes to infinity. That range may be for code points that are
9958 * matched in the inversion list, or it may be for ones that aren't
9959 * matched. In the latter case, the highest code point in the set is one
9960 * less than the beginning of this range; otherwise it is the final element
9961 * of this range: infinity */
9962 return (ELEMENT_RANGE_MATCHES_INVLIST(len - 1))
9964 : array[len - 1] - 1;
9968 S_invlist_contents(pTHX_ SV* const invlist, const bool traditional_style)
9970 /* Get the contents of an inversion list into a string SV so that they can
9971 * be printed out. If 'traditional_style' is TRUE, it uses the format
9972 * traditionally done for debug tracing; otherwise it uses a format
9973 * suitable for just copying to the output, with blanks between ranges and
9974 * a dash between range components */
9978 const char intra_range_delimiter = (traditional_style ? '\t' : '-');
9979 const char inter_range_delimiter = (traditional_style ? '\n' : ' ');
9981 if (traditional_style) {
9982 output = newSVpvs("\n");
9985 output = newSVpvs("");
9988 PERL_ARGS_ASSERT_INVLIST_CONTENTS;
9990 assert(! invlist_is_iterating(invlist));
9992 invlist_iterinit(invlist);
9993 while (invlist_iternext(invlist, &start, &end)) {
9994 if (end == UV_MAX) {
9995 Perl_sv_catpvf(aTHX_ output, "%04" UVXf "%cINFINITY%c",
9996 start, intra_range_delimiter,
9997 inter_range_delimiter);
9999 else if (end != start) {
10000 Perl_sv_catpvf(aTHX_ output, "%04" UVXf "%c%04" UVXf "%c",
10002 intra_range_delimiter,
10003 end, inter_range_delimiter);
10006 Perl_sv_catpvf(aTHX_ output, "%04" UVXf "%c",
10007 start, inter_range_delimiter);
10011 if (SvCUR(output) && ! traditional_style) {/* Get rid of trailing blank */
10012 SvCUR_set(output, SvCUR(output) - 1);
10018 #ifndef PERL_IN_XSUB_RE
10020 Perl__invlist_dump(pTHX_ PerlIO *file, I32 level,
10021 const char * const indent, SV* const invlist)
10023 /* Designed to be called only by do_sv_dump(). Dumps out the ranges of the
10024 * inversion list 'invlist' to 'file' at 'level' Each line is prefixed by
10025 * the string 'indent'. The output looks like this:
10026 [0] 0x000A .. 0x000D
10028 [4] 0x2028 .. 0x2029
10029 [6] 0x3104 .. INFINITY
10030 * This means that the first range of code points matched by the list are
10031 * 0xA through 0xD; the second range contains only the single code point
10032 * 0x85, etc. An inversion list is an array of UVs. Two array elements
10033 * are used to define each range (except if the final range extends to
10034 * infinity, only a single element is needed). The array index of the
10035 * first element for the corresponding range is given in brackets. */
10040 PERL_ARGS_ASSERT__INVLIST_DUMP;
10042 if (invlist_is_iterating(invlist)) {
10043 Perl_dump_indent(aTHX_ level, file,
10044 "%sCan't dump inversion list because is in middle of iterating\n",
10049 invlist_iterinit(invlist);
10050 while (invlist_iternext(invlist, &start, &end)) {
10051 if (end == UV_MAX) {
10052 Perl_dump_indent(aTHX_ level, file,
10053 "%s[%" UVuf "] 0x%04" UVXf " .. INFINITY\n",
10054 indent, (UV)count, start);
10056 else if (end != start) {
10057 Perl_dump_indent(aTHX_ level, file,
10058 "%s[%" UVuf "] 0x%04" UVXf " .. 0x%04" UVXf "\n",
10059 indent, (UV)count, start, end);
10062 Perl_dump_indent(aTHX_ level, file, "%s[%" UVuf "] 0x%04" UVXf "\n",
10063 indent, (UV)count, start);
10070 Perl__load_PL_utf8_foldclosures (pTHX)
10072 assert(! PL_utf8_foldclosures);
10074 /* If the folds haven't been read in, call a fold function
10076 if (! PL_utf8_tofold) {
10077 U8 dummy[UTF8_MAXBYTES_CASE+1];
10078 const U8 hyphen[] = HYPHEN_UTF8;
10080 /* This string is just a short named one above \xff */
10081 toFOLD_utf8_safe(hyphen, hyphen + sizeof(hyphen) - 1, dummy, NULL);
10082 assert(PL_utf8_tofold); /* Verify that worked */
10084 PL_utf8_foldclosures = _swash_inversion_hash(PL_utf8_tofold);
10088 #if defined(PERL_ARGS_ASSERT__INVLISTEQ) && !defined(PERL_IN_XSUB_RE)
10090 Perl__invlistEQ(pTHX_ SV* const a, SV* const b, const bool complement_b)
10092 /* Return a boolean as to if the two passed in inversion lists are
10093 * identical. The final argument, if TRUE, says to take the complement of
10094 * the second inversion list before doing the comparison */
10096 const UV* array_a = invlist_array(a);
10097 const UV* array_b = invlist_array(b);
10098 UV len_a = _invlist_len(a);
10099 UV len_b = _invlist_len(b);
10101 PERL_ARGS_ASSERT__INVLISTEQ;
10103 /* If are to compare 'a' with the complement of b, set it
10104 * up so are looking at b's complement. */
10105 if (complement_b) {
10107 /* The complement of nothing is everything, so <a> would have to have
10108 * just one element, starting at zero (ending at infinity) */
10110 return (len_a == 1 && array_a[0] == 0);
10112 else if (array_b[0] == 0) {
10114 /* Otherwise, to complement, we invert. Here, the first element is
10115 * 0, just remove it. To do this, we just pretend the array starts
10123 /* But if the first element is not zero, we pretend the list starts
10124 * at the 0 that is always stored immediately before the array. */
10130 return len_a == len_b
10131 && memEQ(array_a, array_b, len_a * sizeof(array_a[0]));
10137 * As best we can, determine the characters that can match the start of
10138 * the given EXACTF-ish node.
10140 * Returns the invlist as a new SV*; it is the caller's responsibility to
10141 * call SvREFCNT_dec() when done with it.
10144 S__make_exactf_invlist(pTHX_ RExC_state_t *pRExC_state, regnode *node)
10146 const U8 * s = (U8*)STRING(node);
10147 SSize_t bytelen = STR_LEN(node);
10149 /* Start out big enough for 2 separate code points */
10150 SV* invlist = _new_invlist(4);
10152 PERL_ARGS_ASSERT__MAKE_EXACTF_INVLIST;
10157 /* We punt and assume can match anything if the node begins
10158 * with a multi-character fold. Things are complicated. For
10159 * example, /ffi/i could match any of:
10160 * "\N{LATIN SMALL LIGATURE FFI}"
10161 * "\N{LATIN SMALL LIGATURE FF}I"
10162 * "F\N{LATIN SMALL LIGATURE FI}"
10163 * plus several other things; and making sure we have all the
10164 * possibilities is hard. */
10165 if (is_MULTI_CHAR_FOLD_latin1_safe(s, s + bytelen)) {
10166 invlist = _add_range_to_invlist(invlist, 0, UV_MAX);
10169 /* Any Latin1 range character can potentially match any
10170 * other depending on the locale */
10171 if (OP(node) == EXACTFL) {
10172 _invlist_union(invlist, PL_Latin1, &invlist);
10175 /* But otherwise, it matches at least itself. We can
10176 * quickly tell if it has a distinct fold, and if so,
10177 * it matches that as well */
10178 invlist = add_cp_to_invlist(invlist, uc);
10179 if (IS_IN_SOME_FOLD_L1(uc))
10180 invlist = add_cp_to_invlist(invlist, PL_fold_latin1[uc]);
10183 /* Some characters match above-Latin1 ones under /i. This
10184 * is true of EXACTFL ones when the locale is UTF-8 */
10185 if (HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE(uc)
10186 && (! isASCII(uc) || (OP(node) != EXACTFA
10187 && OP(node) != EXACTFA_NO_TRIE)))
10189 add_above_Latin1_folds(pRExC_state, (U8) uc, &invlist);
10193 else { /* Pattern is UTF-8 */
10194 U8 folded[UTF8_MAX_FOLD_CHAR_EXPAND * UTF8_MAXBYTES_CASE + 1] = { '\0' };
10195 STRLEN foldlen = UTF8SKIP(s);
10196 const U8* e = s + bytelen;
10199 uc = utf8_to_uvchr_buf(s, s + bytelen, NULL);
10201 /* The only code points that aren't folded in a UTF EXACTFish
10202 * node are are the problematic ones in EXACTFL nodes */
10203 if (OP(node) == EXACTFL && is_PROBLEMATIC_LOCALE_FOLDEDS_START_cp(uc)) {
10204 /* We need to check for the possibility that this EXACTFL
10205 * node begins with a multi-char fold. Therefore we fold
10206 * the first few characters of it so that we can make that
10211 for (i = 0; i < UTF8_MAX_FOLD_CHAR_EXPAND && s < e; i++) {
10213 *(d++) = (U8) toFOLD(*s);
10218 toFOLD_utf8_safe(s, e, d, &len);
10224 /* And set up so the code below that looks in this folded
10225 * buffer instead of the node's string */
10227 foldlen = UTF8SKIP(folded);
10231 /* When we reach here 's' points to the fold of the first
10232 * character(s) of the node; and 'e' points to far enough along
10233 * the folded string to be just past any possible multi-char
10234 * fold. 'foldlen' is the length in bytes of the first
10237 * Unlike the non-UTF-8 case, the macro for determining if a
10238 * string is a multi-char fold requires all the characters to
10239 * already be folded. This is because of all the complications
10240 * if not. Note that they are folded anyway, except in EXACTFL
10241 * nodes. Like the non-UTF case above, we punt if the node
10242 * begins with a multi-char fold */
10244 if (is_MULTI_CHAR_FOLD_utf8_safe(s, e)) {
10245 invlist = _add_range_to_invlist(invlist, 0, UV_MAX);
10247 else { /* Single char fold */
10249 /* It matches all the things that fold to it, which are
10250 * found in PL_utf8_foldclosures (including itself) */
10251 invlist = add_cp_to_invlist(invlist, uc);
10252 if (! PL_utf8_foldclosures)
10253 _load_PL_utf8_foldclosures();
10254 if ((listp = hv_fetch(PL_utf8_foldclosures,
10255 (char *) s, foldlen, FALSE)))
10257 AV* list = (AV*) *listp;
10259 for (k = 0; k <= av_tindex_skip_len_mg(list); k++) {
10260 SV** c_p = av_fetch(list, k, FALSE);
10266 /* /aa doesn't allow folds between ASCII and non- */
10267 if ((OP(node) == EXACTFA || OP(node) == EXACTFA_NO_TRIE)
10268 && isASCII(c) != isASCII(uc))
10273 invlist = add_cp_to_invlist(invlist, c);
10282 #undef HEADER_LENGTH
10283 #undef TO_INTERNAL_SIZE
10284 #undef FROM_INTERNAL_SIZE
10285 #undef INVLIST_VERSION_ID
10287 /* End of inversion list object */
10290 S_parse_lparen_question_flags(pTHX_ RExC_state_t *pRExC_state)
10292 /* This parses the flags that are in either the '(?foo)' or '(?foo:bar)'
10293 * constructs, and updates RExC_flags with them. On input, RExC_parse
10294 * should point to the first flag; it is updated on output to point to the
10295 * final ')' or ':'. There needs to be at least one flag, or this will
10298 /* for (?g), (?gc), and (?o) warnings; warning
10299 about (?c) will warn about (?g) -- japhy */
10301 #define WASTED_O 0x01
10302 #define WASTED_G 0x02
10303 #define WASTED_C 0x04
10304 #define WASTED_GC (WASTED_G|WASTED_C)
10305 I32 wastedflags = 0x00;
10306 U32 posflags = 0, negflags = 0;
10307 U32 *flagsp = &posflags;
10308 char has_charset_modifier = '\0';
10310 bool has_use_defaults = FALSE;
10311 const char* const seqstart = RExC_parse - 1; /* Point to the '?' */
10312 int x_mod_count = 0;
10314 PERL_ARGS_ASSERT_PARSE_LPAREN_QUESTION_FLAGS;
10316 /* '^' as an initial flag sets certain defaults */
10317 if (UCHARAT(RExC_parse) == '^') {
10319 has_use_defaults = TRUE;
10320 STD_PMMOD_FLAGS_CLEAR(&RExC_flags);
10321 set_regex_charset(&RExC_flags, (RExC_utf8 || RExC_uni_semantics)
10322 ? REGEX_UNICODE_CHARSET
10323 : REGEX_DEPENDS_CHARSET);
10326 cs = get_regex_charset(RExC_flags);
10327 if (cs == REGEX_DEPENDS_CHARSET
10328 && (RExC_utf8 || RExC_uni_semantics))
10330 cs = REGEX_UNICODE_CHARSET;
10333 while (RExC_parse < RExC_end) {
10334 /* && strchr("iogcmsx", *RExC_parse) */
10335 /* (?g), (?gc) and (?o) are useless here
10336 and must be globally applied -- japhy */
10337 switch (*RExC_parse) {
10339 /* Code for the imsxn flags */
10340 CASE_STD_PMMOD_FLAGS_PARSE_SET(flagsp, x_mod_count);
10342 case LOCALE_PAT_MOD:
10343 if (has_charset_modifier) {
10344 goto excess_modifier;
10346 else if (flagsp == &negflags) {
10349 cs = REGEX_LOCALE_CHARSET;
10350 has_charset_modifier = LOCALE_PAT_MOD;
10352 case UNICODE_PAT_MOD:
10353 if (has_charset_modifier) {
10354 goto excess_modifier;
10356 else if (flagsp == &negflags) {
10359 cs = REGEX_UNICODE_CHARSET;
10360 has_charset_modifier = UNICODE_PAT_MOD;
10362 case ASCII_RESTRICT_PAT_MOD:
10363 if (flagsp == &negflags) {
10366 if (has_charset_modifier) {
10367 if (cs != REGEX_ASCII_RESTRICTED_CHARSET) {
10368 goto excess_modifier;
10370 /* Doubled modifier implies more restricted */
10371 cs = REGEX_ASCII_MORE_RESTRICTED_CHARSET;
10374 cs = REGEX_ASCII_RESTRICTED_CHARSET;
10376 has_charset_modifier = ASCII_RESTRICT_PAT_MOD;
10378 case DEPENDS_PAT_MOD:
10379 if (has_use_defaults) {
10380 goto fail_modifiers;
10382 else if (flagsp == &negflags) {
10385 else if (has_charset_modifier) {
10386 goto excess_modifier;
10389 /* The dual charset means unicode semantics if the
10390 * pattern (or target, not known until runtime) are
10391 * utf8, or something in the pattern indicates unicode
10393 cs = (RExC_utf8 || RExC_uni_semantics)
10394 ? REGEX_UNICODE_CHARSET
10395 : REGEX_DEPENDS_CHARSET;
10396 has_charset_modifier = DEPENDS_PAT_MOD;
10400 if (has_charset_modifier == ASCII_RESTRICT_PAT_MOD) {
10401 vFAIL2("Regexp modifier \"%c\" may appear a maximum of twice", ASCII_RESTRICT_PAT_MOD);
10403 else if (has_charset_modifier == *(RExC_parse - 1)) {
10404 vFAIL2("Regexp modifier \"%c\" may not appear twice",
10405 *(RExC_parse - 1));
10408 vFAIL3("Regexp modifiers \"%c\" and \"%c\" are mutually exclusive", has_charset_modifier, *(RExC_parse - 1));
10410 NOT_REACHED; /*NOTREACHED*/
10413 vFAIL2("Regexp modifier \"%c\" may not appear after the \"-\"",
10414 *(RExC_parse - 1));
10415 NOT_REACHED; /*NOTREACHED*/
10416 case ONCE_PAT_MOD: /* 'o' */
10417 case GLOBAL_PAT_MOD: /* 'g' */
10418 if (PASS2 && ckWARN(WARN_REGEXP)) {
10419 const I32 wflagbit = *RExC_parse == 'o'
10422 if (! (wastedflags & wflagbit) ) {
10423 wastedflags |= wflagbit;
10424 /* diag_listed_as: Useless (?-%s) - don't use /%s modifier in regex; marked by <-- HERE in m/%s/ */
10427 "Useless (%s%c) - %suse /%c modifier",
10428 flagsp == &negflags ? "?-" : "?",
10430 flagsp == &negflags ? "don't " : "",
10437 case CONTINUE_PAT_MOD: /* 'c' */
10438 if (PASS2 && ckWARN(WARN_REGEXP)) {
10439 if (! (wastedflags & WASTED_C) ) {
10440 wastedflags |= WASTED_GC;
10441 /* diag_listed_as: Useless (?-%s) - don't use /%s modifier in regex; marked by <-- HERE in m/%s/ */
10444 "Useless (%sc) - %suse /gc modifier",
10445 flagsp == &negflags ? "?-" : "?",
10446 flagsp == &negflags ? "don't " : ""
10451 case KEEPCOPY_PAT_MOD: /* 'p' */
10452 if (flagsp == &negflags) {
10454 ckWARNreg(RExC_parse + 1,"Useless use of (?-p)");
10456 *flagsp |= RXf_PMf_KEEPCOPY;
10460 /* A flag is a default iff it is following a minus, so
10461 * if there is a minus, it means will be trying to
10462 * re-specify a default which is an error */
10463 if (has_use_defaults || flagsp == &negflags) {
10464 goto fail_modifiers;
10466 flagsp = &negflags;
10467 wastedflags = 0; /* reset so (?g-c) warns twice */
10473 if ((posflags & (RXf_PMf_EXTENDED|RXf_PMf_EXTENDED_MORE)) == RXf_PMf_EXTENDED) {
10474 negflags |= RXf_PMf_EXTENDED_MORE;
10476 RExC_flags |= posflags;
10478 if (negflags & RXf_PMf_EXTENDED) {
10479 negflags |= RXf_PMf_EXTENDED_MORE;
10481 RExC_flags &= ~negflags;
10482 set_regex_charset(&RExC_flags, cs);
10487 RExC_parse += SKIP_IF_CHAR(RExC_parse);
10488 /* diag_listed_as: Sequence (?%s...) not recognized in regex; marked by <-- HERE in m/%s/ */
10489 vFAIL2utf8f("Sequence (%" UTF8f "...) not recognized",
10490 UTF8fARG(UTF, RExC_parse-seqstart, seqstart));
10491 NOT_REACHED; /*NOTREACHED*/
10494 RExC_parse += UTF ? UTF8SKIP(RExC_parse) : 1;
10497 vFAIL("Sequence (?... not terminated");
10501 - reg - regular expression, i.e. main body or parenthesized thing
10503 * Caller must absorb opening parenthesis.
10505 * Combining parenthesis handling with the base level of regular expression
10506 * is a trifle forced, but the need to tie the tails of the branches to what
10507 * follows makes it hard to avoid.
10509 #define REGTAIL(x,y,z) regtail((x),(y),(z),depth+1)
10511 #define REGTAIL_STUDY(x,y,z) regtail_study((x),(y),(z),depth+1)
10513 #define REGTAIL_STUDY(x,y,z) regtail((x),(y),(z),depth+1)
10516 PERL_STATIC_INLINE regnode *
10517 S_handle_named_backref(pTHX_ RExC_state_t *pRExC_state,
10519 char * parse_start,
10524 char* name_start = RExC_parse;
10526 SV *sv_dat = reg_scan_name(pRExC_state, SIZE_ONLY
10527 ? REG_RSN_RETURN_NULL
10528 : REG_RSN_RETURN_DATA);
10529 GET_RE_DEBUG_FLAGS_DECL;
10531 PERL_ARGS_ASSERT_HANDLE_NAMED_BACKREF;
10533 if (RExC_parse == name_start || *RExC_parse != ch) {
10534 /* diag_listed_as: Sequence \%s... not terminated in regex; marked by <-- HERE in m/%s/ */
10535 vFAIL2("Sequence %.3s... not terminated",parse_start);
10539 num = add_data( pRExC_state, STR_WITH_LEN("S"));
10540 RExC_rxi->data->data[num]=(void*)sv_dat;
10541 SvREFCNT_inc_simple_void(sv_dat);
10544 ret = reganode(pRExC_state,
10547 : (ASCII_FOLD_RESTRICTED)
10549 : (AT_LEAST_UNI_SEMANTICS)
10555 *flagp |= HASWIDTH;
10557 Set_Node_Offset(ret, parse_start+1);
10558 Set_Node_Cur_Length(ret, parse_start);
10560 nextchar(pRExC_state);
10564 /* Returns NULL, setting *flagp to TRYAGAIN at the end of (?) that only sets
10565 flags. Returns NULL, setting *flagp to RESTART_PASS1 if the sizing scan
10566 needs to be restarted, or'd with NEED_UTF8 if the pattern needs to be
10567 upgraded to UTF-8. Otherwise would only return NULL if regbranch() returns
10568 NULL, which cannot happen. */
10570 S_reg(pTHX_ RExC_state_t *pRExC_state, I32 paren, I32 *flagp,U32 depth)
10571 /* paren: Parenthesized? 0=top; 1,2=inside '(': changed to letter.
10572 * 2 is like 1, but indicates that nextchar() has been called to advance
10573 * RExC_parse beyond the '('. Things like '(?' are indivisible tokens, and
10574 * this flag alerts us to the need to check for that */
10576 regnode *ret; /* Will be the head of the group. */
10579 regnode *ender = NULL;
10582 U32 oregflags = RExC_flags;
10583 bool have_branch = 0;
10585 I32 freeze_paren = 0;
10586 I32 after_freeze = 0;
10587 I32 num; /* numeric backreferences */
10589 char * parse_start = RExC_parse; /* MJD */
10590 char * const oregcomp_parse = RExC_parse;
10592 GET_RE_DEBUG_FLAGS_DECL;
10594 PERL_ARGS_ASSERT_REG;
10595 DEBUG_PARSE("reg ");
10597 *flagp = 0; /* Tentatively. */
10599 /* Having this true makes it feasible to have a lot fewer tests for the
10600 * parse pointer being in scope. For example, we can write
10601 * while(isFOO(*RExC_parse)) RExC_parse++;
10603 * while(RExC_parse < RExC_end && isFOO(*RExC_parse)) RExC_parse++;
10605 assert(*RExC_end == '\0');
10607 /* Make an OPEN node, if parenthesized. */
10610 /* Under /x, space and comments can be gobbled up between the '(' and
10611 * here (if paren ==2). The forms '(*VERB' and '(?...' disallow such
10612 * intervening space, as the sequence is a token, and a token should be
10614 bool has_intervening_patws = paren == 2 && *(RExC_parse - 1) != '(';
10616 if (RExC_parse >= RExC_end) {
10617 vFAIL("Unmatched (");
10620 if ( *RExC_parse == '*') { /* (*VERB:ARG) */
10621 char *start_verb = RExC_parse + 1;
10623 char *start_arg = NULL;
10624 unsigned char op = 0;
10625 int arg_required = 0;
10626 int internal_argval = -1; /* if >-1 we are not allowed an argument*/
10628 if (has_intervening_patws) {
10629 RExC_parse++; /* past the '*' */
10630 vFAIL("In '(*VERB...)', the '(' and '*' must be adjacent");
10632 while (RExC_parse < RExC_end && *RExC_parse != ')' ) {
10633 if ( *RExC_parse == ':' ) {
10634 start_arg = RExC_parse + 1;
10637 RExC_parse += UTF ? UTF8SKIP(RExC_parse) : 1;
10639 verb_len = RExC_parse - start_verb;
10641 if (RExC_parse >= RExC_end) {
10642 goto unterminated_verb_pattern;
10644 RExC_parse += UTF ? UTF8SKIP(RExC_parse) : 1;
10645 while ( RExC_parse < RExC_end && *RExC_parse != ')' )
10646 RExC_parse += UTF ? UTF8SKIP(RExC_parse) : 1;
10647 if ( RExC_parse >= RExC_end || *RExC_parse != ')' )
10648 unterminated_verb_pattern:
10649 vFAIL("Unterminated verb pattern argument");
10650 if ( RExC_parse == start_arg )
10653 if ( RExC_parse >= RExC_end || *RExC_parse != ')' )
10654 vFAIL("Unterminated verb pattern");
10657 /* Here, we know that RExC_parse < RExC_end */
10659 switch ( *start_verb ) {
10660 case 'A': /* (*ACCEPT) */
10661 if ( memEQs(start_verb,verb_len,"ACCEPT") ) {
10663 internal_argval = RExC_nestroot;
10666 case 'C': /* (*COMMIT) */
10667 if ( memEQs(start_verb,verb_len,"COMMIT") )
10670 case 'F': /* (*FAIL) */
10671 if ( verb_len==1 || memEQs(start_verb,verb_len,"FAIL") ) {
10675 case ':': /* (*:NAME) */
10676 case 'M': /* (*MARK:NAME) */
10677 if ( verb_len==0 || memEQs(start_verb,verb_len,"MARK") ) {
10682 case 'P': /* (*PRUNE) */
10683 if ( memEQs(start_verb,verb_len,"PRUNE") )
10686 case 'S': /* (*SKIP) */
10687 if ( memEQs(start_verb,verb_len,"SKIP") )
10690 case 'T': /* (*THEN) */
10691 /* [19:06] <TimToady> :: is then */
10692 if ( memEQs(start_verb,verb_len,"THEN") ) {
10694 RExC_seen |= REG_CUTGROUP_SEEN;
10699 RExC_parse += UTF ? UTF8SKIP(RExC_parse) : 1;
10701 "Unknown verb pattern '%" UTF8f "'",
10702 UTF8fARG(UTF, verb_len, start_verb));
10704 if ( arg_required && !start_arg ) {
10705 vFAIL3("Verb pattern '%.*s' has a mandatory argument",
10706 verb_len, start_verb);
10708 if (internal_argval == -1) {
10709 ret = reganode(pRExC_state, op, 0);
10711 ret = reg2Lanode(pRExC_state, op, 0, internal_argval);
10713 RExC_seen |= REG_VERBARG_SEEN;
10714 if ( ! SIZE_ONLY ) {
10716 SV *sv = newSVpvn( start_arg,
10717 RExC_parse - start_arg);
10718 ARG(ret) = add_data( pRExC_state,
10719 STR_WITH_LEN("S"));
10720 RExC_rxi->data->data[ARG(ret)]=(void*)sv;
10725 if ( internal_argval != -1 )
10726 ARG2L_SET(ret, internal_argval);
10728 nextchar(pRExC_state);
10731 else if (*RExC_parse == '?') { /* (?...) */
10732 bool is_logical = 0;
10733 const char * const seqstart = RExC_parse;
10734 const char * endptr;
10735 if (has_intervening_patws) {
10737 vFAIL("In '(?...)', the '(' and '?' must be adjacent");
10740 RExC_parse++; /* past the '?' */
10741 paren = *RExC_parse; /* might be a trailing NUL, if not
10743 RExC_parse += UTF ? UTF8SKIP(RExC_parse) : 1;
10744 if (RExC_parse > RExC_end) {
10747 ret = NULL; /* For look-ahead/behind. */
10750 case 'P': /* (?P...) variants for those used to PCRE/Python */
10751 paren = *RExC_parse;
10752 if ( paren == '<') { /* (?P<...>) named capture */
10754 if (RExC_parse >= RExC_end) {
10755 vFAIL("Sequence (?P<... not terminated");
10757 goto named_capture;
10759 else if (paren == '>') { /* (?P>name) named recursion */
10761 if (RExC_parse >= RExC_end) {
10762 vFAIL("Sequence (?P>... not terminated");
10764 goto named_recursion;
10766 else if (paren == '=') { /* (?P=...) named backref */
10768 return handle_named_backref(pRExC_state, flagp,
10771 RExC_parse += SKIP_IF_CHAR(RExC_parse);
10772 /* diag_listed_as: Sequence (?%s...) not recognized in regex; marked by <-- HERE in m/%s/ */
10773 vFAIL3("Sequence (%.*s...) not recognized",
10774 RExC_parse-seqstart, seqstart);
10775 NOT_REACHED; /*NOTREACHED*/
10776 case '<': /* (?<...) */
10777 if (*RExC_parse == '!')
10779 else if (*RExC_parse != '=')
10786 case '\'': /* (?'...') */
10787 name_start = RExC_parse;
10788 svname = reg_scan_name(pRExC_state,
10789 SIZE_ONLY /* reverse test from the others */
10790 ? REG_RSN_RETURN_NAME
10791 : REG_RSN_RETURN_NULL);
10792 if ( RExC_parse == name_start
10793 || RExC_parse >= RExC_end
10794 || *RExC_parse != paren)
10796 vFAIL2("Sequence (?%c... not terminated",
10797 paren=='>' ? '<' : paren);
10802 if (!svname) /* shouldn't happen */
10804 "panic: reg_scan_name returned NULL");
10805 if (!RExC_paren_names) {
10806 RExC_paren_names= newHV();
10807 sv_2mortal(MUTABLE_SV(RExC_paren_names));
10809 RExC_paren_name_list= newAV();
10810 sv_2mortal(MUTABLE_SV(RExC_paren_name_list));
10813 he_str = hv_fetch_ent( RExC_paren_names, svname, 1, 0 );
10815 sv_dat = HeVAL(he_str);
10817 /* croak baby croak */
10819 "panic: paren_name hash element allocation failed");
10820 } else if ( SvPOK(sv_dat) ) {
10821 /* (?|...) can mean we have dupes so scan to check
10822 its already been stored. Maybe a flag indicating
10823 we are inside such a construct would be useful,
10824 but the arrays are likely to be quite small, so
10825 for now we punt -- dmq */
10826 IV count = SvIV(sv_dat);
10827 I32 *pv = (I32*)SvPVX(sv_dat);
10829 for ( i = 0 ; i < count ; i++ ) {
10830 if ( pv[i] == RExC_npar ) {
10836 pv = (I32*)SvGROW(sv_dat,
10837 SvCUR(sv_dat) + sizeof(I32)+1);
10838 SvCUR_set(sv_dat, SvCUR(sv_dat) + sizeof(I32));
10839 pv[count] = RExC_npar;
10840 SvIV_set(sv_dat, SvIVX(sv_dat) + 1);
10843 (void)SvUPGRADE(sv_dat,SVt_PVNV);
10844 sv_setpvn(sv_dat, (char *)&(RExC_npar),
10847 SvIV_set(sv_dat, 1);
10850 /* Yes this does cause a memory leak in debugging Perls
10852 if (!av_store(RExC_paren_name_list,
10853 RExC_npar, SvREFCNT_inc(svname)))
10854 SvREFCNT_dec_NN(svname);
10857 /*sv_dump(sv_dat);*/
10859 nextchar(pRExC_state);
10861 goto capturing_parens;
10863 RExC_seen |= REG_LOOKBEHIND_SEEN;
10864 RExC_in_lookbehind++;
10866 if (RExC_parse >= RExC_end) {
10867 vFAIL("Sequence (?... not terminated");
10871 case '=': /* (?=...) */
10872 RExC_seen_zerolen++;
10874 case '!': /* (?!...) */
10875 RExC_seen_zerolen++;
10876 /* check if we're really just a "FAIL" assertion */
10877 skip_to_be_ignored_text(pRExC_state, &RExC_parse,
10878 FALSE /* Don't force to /x */ );
10879 if (*RExC_parse == ')') {
10880 ret=reganode(pRExC_state, OPFAIL, 0);
10881 nextchar(pRExC_state);
10885 case '|': /* (?|...) */
10886 /* branch reset, behave like a (?:...) except that
10887 buffers in alternations share the same numbers */
10889 after_freeze = freeze_paren = RExC_npar;
10891 case ':': /* (?:...) */
10892 case '>': /* (?>...) */
10894 case '$': /* (?$...) */
10895 case '@': /* (?@...) */
10896 vFAIL2("Sequence (?%c...) not implemented", (int)paren);
10898 case '0' : /* (?0) */
10899 case 'R' : /* (?R) */
10900 if (RExC_parse == RExC_end || *RExC_parse != ')')
10901 FAIL("Sequence (?R) not terminated");
10903 RExC_seen |= REG_RECURSE_SEEN;
10904 *flagp |= POSTPONED;
10905 goto gen_recurse_regop;
10907 /* named and numeric backreferences */
10908 case '&': /* (?&NAME) */
10909 parse_start = RExC_parse - 1;
10912 SV *sv_dat = reg_scan_name(pRExC_state,
10913 SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
10914 num = sv_dat ? *((I32 *)SvPVX(sv_dat)) : 0;
10916 if (RExC_parse >= RExC_end || *RExC_parse != ')')
10917 vFAIL("Sequence (?&... not terminated");
10918 goto gen_recurse_regop;
10921 if (!(RExC_parse[0] >= '1' && RExC_parse[0] <= '9')) {
10923 vFAIL("Illegal pattern");
10925 goto parse_recursion;
10927 case '-': /* (?-1) */
10928 if (!(RExC_parse[0] >= '1' && RExC_parse[0] <= '9')) {
10929 RExC_parse--; /* rewind to let it be handled later */
10933 case '1': case '2': case '3': case '4': /* (?1) */
10934 case '5': case '6': case '7': case '8': case '9':
10935 RExC_parse = (char *) seqstart + 1; /* Point to the digit */
10938 bool is_neg = FALSE;
10940 parse_start = RExC_parse - 1; /* MJD */
10941 if (*RExC_parse == '-') {
10945 if (grok_atoUV(RExC_parse, &unum, &endptr)
10949 RExC_parse = (char*)endptr;
10953 /* Some limit for num? */
10957 if (*RExC_parse!=')')
10958 vFAIL("Expecting close bracket");
10961 if ( paren == '-' ) {
10963 Diagram of capture buffer numbering.
10964 Top line is the normal capture buffer numbers
10965 Bottom line is the negative indexing as from
10969 /(a(x)y)(a(b(c(?-2)d)e)f)(g(h))/
10973 num = RExC_npar + num;
10976 vFAIL("Reference to nonexistent group");
10978 } else if ( paren == '+' ) {
10979 num = RExC_npar + num - 1;
10981 /* We keep track how many GOSUB items we have produced.
10982 To start off the ARG2L() of the GOSUB holds its "id",
10983 which is used later in conjunction with RExC_recurse
10984 to calculate the offset we need to jump for the GOSUB,
10985 which it will store in the final representation.
10986 We have to defer the actual calculation until much later
10987 as the regop may move.
10990 ret = reg2Lanode(pRExC_state, GOSUB, num, RExC_recurse_count);
10992 if (num > (I32)RExC_rx->nparens) {
10994 vFAIL("Reference to nonexistent group");
10996 RExC_recurse_count++;
10997 DEBUG_OPTIMISE_MORE_r(Perl_re_printf( aTHX_
10998 "%*s%*s Recurse #%" UVuf " to %" IVdf "\n",
10999 22, "| |", (int)(depth * 2 + 1), "",
11000 (UV)ARG(ret), (IV)ARG2L(ret)));
11002 RExC_seen |= REG_RECURSE_SEEN;
11004 Set_Node_Length(ret, 1 + regarglen[OP(ret)]); /* MJD */
11005 Set_Node_Offset(ret, parse_start); /* MJD */
11007 *flagp |= POSTPONED;
11008 assert(*RExC_parse == ')');
11009 nextchar(pRExC_state);
11014 case '?': /* (??...) */
11016 if (*RExC_parse != '{') {
11017 RExC_parse += SKIP_IF_CHAR(RExC_parse);
11018 /* diag_listed_as: Sequence (?%s...) not recognized in regex; marked by <-- HERE in m/%s/ */
11020 "Sequence (%" UTF8f "...) not recognized",
11021 UTF8fARG(UTF, RExC_parse-seqstart, seqstart));
11022 NOT_REACHED; /*NOTREACHED*/
11024 *flagp |= POSTPONED;
11028 case '{': /* (?{...}) */
11031 struct reg_code_block *cb;
11033 RExC_seen_zerolen++;
11035 if ( !pRExC_state->code_blocks
11036 || pRExC_state->code_index
11037 >= pRExC_state->code_blocks->count
11038 || pRExC_state->code_blocks->cb[pRExC_state->code_index].start
11039 != (STRLEN)((RExC_parse -3 - (is_logical ? 1 : 0))
11042 if (RExC_pm_flags & PMf_USE_RE_EVAL)
11043 FAIL("panic: Sequence (?{...}): no code block found\n");
11044 FAIL("Eval-group not allowed at runtime, use re 'eval'");
11046 /* this is a pre-compiled code block (?{...}) */
11047 cb = &pRExC_state->code_blocks->cb[pRExC_state->code_index];
11048 RExC_parse = RExC_start + cb->end;
11051 if (cb->src_regex) {
11052 n = add_data(pRExC_state, STR_WITH_LEN("rl"));
11053 RExC_rxi->data->data[n] =
11054 (void*)SvREFCNT_inc((SV*)cb->src_regex);
11055 RExC_rxi->data->data[n+1] = (void*)o;
11058 n = add_data(pRExC_state,
11059 (RExC_pm_flags & PMf_HAS_CV) ? "L" : "l", 1);
11060 RExC_rxi->data->data[n] = (void*)o;
11063 pRExC_state->code_index++;
11064 nextchar(pRExC_state);
11068 ret = reg_node(pRExC_state, LOGICAL);
11070 eval = reg2Lanode(pRExC_state, EVAL,
11073 /* for later propagation into (??{})
11075 RExC_flags & RXf_PMf_COMPILETIME
11080 REGTAIL(pRExC_state, ret, eval);
11081 /* deal with the length of this later - MJD */
11084 ret = reg2Lanode(pRExC_state, EVAL, n, 0);
11085 Set_Node_Length(ret, RExC_parse - parse_start + 1);
11086 Set_Node_Offset(ret, parse_start);
11089 case '(': /* (?(?{...})...) and (?(?=...)...) */
11092 const int DEFINE_len = sizeof("DEFINE") - 1;
11093 if (RExC_parse[0] == '?') { /* (?(?...)) */
11094 if ( RExC_parse < RExC_end - 1
11095 && ( RExC_parse[1] == '='
11096 || RExC_parse[1] == '!'
11097 || RExC_parse[1] == '<'
11098 || RExC_parse[1] == '{')
11099 ) { /* Lookahead or eval. */
11103 ret = reg_node(pRExC_state, LOGICAL);
11107 tail = reg(pRExC_state, 1, &flag, depth+1);
11108 if (flag & (RESTART_PASS1|NEED_UTF8)) {
11109 *flagp = flag & (RESTART_PASS1|NEED_UTF8);
11112 REGTAIL(pRExC_state, ret, tail);
11115 /* Fall through to ‘Unknown switch condition’ at the
11116 end of the if/else chain. */
11118 else if ( RExC_parse[0] == '<' /* (?(<NAME>)...) */
11119 || RExC_parse[0] == '\'' ) /* (?('NAME')...) */
11121 char ch = RExC_parse[0] == '<' ? '>' : '\'';
11122 char *name_start= RExC_parse++;
11124 SV *sv_dat=reg_scan_name(pRExC_state,
11125 SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
11126 if ( RExC_parse == name_start
11127 || RExC_parse >= RExC_end
11128 || *RExC_parse != ch)
11130 vFAIL2("Sequence (?(%c... not terminated",
11131 (ch == '>' ? '<' : ch));
11135 num = add_data( pRExC_state, STR_WITH_LEN("S"));
11136 RExC_rxi->data->data[num]=(void*)sv_dat;
11137 SvREFCNT_inc_simple_void(sv_dat);
11139 ret = reganode(pRExC_state,NGROUPP,num);
11140 goto insert_if_check_paren;
11142 else if (RExC_end - RExC_parse >= DEFINE_len
11143 && strnEQ(RExC_parse, "DEFINE", DEFINE_len))
11145 ret = reganode(pRExC_state,DEFINEP,0);
11146 RExC_parse += DEFINE_len;
11148 goto insert_if_check_paren;
11150 else if (RExC_parse[0] == 'R') {
11152 /* parno == 0 => /(?(R)YES|NO)/ "in any form of recursion OR eval"
11153 * parno == 1 => /(?(R0)YES|NO)/ "in GOSUB (?0) / (?R)"
11154 * parno == 2 => /(?(R1)YES|NO)/ "in GOSUB (?1) (parno-1)"
11157 if (RExC_parse[0] == '0') {
11161 else if (RExC_parse[0] >= '1' && RExC_parse[0] <= '9' ) {
11163 if (grok_atoUV(RExC_parse, &uv, &endptr)
11166 parno = (I32)uv + 1;
11167 RExC_parse = (char*)endptr;
11169 /* else "Switch condition not recognized" below */
11170 } else if (RExC_parse[0] == '&') {
11173 sv_dat = reg_scan_name(pRExC_state,
11175 ? REG_RSN_RETURN_NULL
11176 : REG_RSN_RETURN_DATA);
11178 /* we should only have a false sv_dat when
11179 * SIZE_ONLY is true, and we always have false
11180 * sv_dat when SIZE_ONLY is true.
11181 * reg_scan_name() will VFAIL() if the name is
11182 * unknown when SIZE_ONLY is false, and otherwise
11183 * will return something, and when SIZE_ONLY is
11184 * true, reg_scan_name() just parses the string,
11185 * and doesnt return anything. (in theory) */
11186 assert(SIZE_ONLY ? !sv_dat : !!sv_dat);
11189 parno = 1 + *((I32 *)SvPVX(sv_dat));
11191 ret = reganode(pRExC_state,INSUBP,parno);
11192 goto insert_if_check_paren;
11194 else if (RExC_parse[0] >= '1' && RExC_parse[0] <= '9' ) {
11198 if (grok_atoUV(RExC_parse, &uv, &endptr)
11202 RExC_parse = (char*)endptr;
11205 vFAIL("panic: grok_atoUV returned FALSE");
11207 ret = reganode(pRExC_state, GROUPP, parno);
11209 insert_if_check_paren:
11210 if (UCHARAT(RExC_parse) != ')') {
11211 RExC_parse += UTF ? UTF8SKIP(RExC_parse) : 1;
11212 vFAIL("Switch condition not recognized");
11214 nextchar(pRExC_state);
11216 REGTAIL(pRExC_state, ret, reganode(pRExC_state, IFTHEN, 0));
11217 br = regbranch(pRExC_state, &flags, 1,depth+1);
11219 if (flags & (RESTART_PASS1|NEED_UTF8)) {
11220 *flagp = flags & (RESTART_PASS1|NEED_UTF8);
11223 FAIL2("panic: regbranch returned NULL, flags=%#" UVxf,
11226 REGTAIL(pRExC_state, br, reganode(pRExC_state,
11228 c = UCHARAT(RExC_parse);
11229 nextchar(pRExC_state);
11230 if (flags&HASWIDTH)
11231 *flagp |= HASWIDTH;
11234 vFAIL("(?(DEFINE)....) does not allow branches");
11236 /* Fake one for optimizer. */
11237 lastbr = reganode(pRExC_state, IFTHEN, 0);
11239 if (!regbranch(pRExC_state, &flags, 1,depth+1)) {
11240 if (flags & (RESTART_PASS1|NEED_UTF8)) {
11241 *flagp = flags & (RESTART_PASS1|NEED_UTF8);
11244 FAIL2("panic: regbranch returned NULL, flags=%#" UVxf,
11247 REGTAIL(pRExC_state, ret, lastbr);
11248 if (flags&HASWIDTH)
11249 *flagp |= HASWIDTH;
11250 c = UCHARAT(RExC_parse);
11251 nextchar(pRExC_state);
11256 if (RExC_parse >= RExC_end)
11257 vFAIL("Switch (?(condition)... not terminated");
11259 vFAIL("Switch (?(condition)... contains too many branches");
11261 ender = reg_node(pRExC_state, TAIL);
11262 REGTAIL(pRExC_state, br, ender);
11264 REGTAIL(pRExC_state, lastbr, ender);
11265 REGTAIL(pRExC_state, NEXTOPER(NEXTOPER(lastbr)), ender);
11268 REGTAIL(pRExC_state, ret, ender);
11269 RExC_size++; /* XXX WHY do we need this?!!
11270 For large programs it seems to be required
11271 but I can't figure out why. -- dmq*/
11274 RExC_parse += UTF ? UTF8SKIP(RExC_parse) : 1;
11275 vFAIL("Unknown switch condition (?(...))");
11277 case '[': /* (?[ ... ]) */
11278 return handle_regex_sets(pRExC_state, NULL, flagp, depth,
11280 case 0: /* A NUL */
11281 RExC_parse--; /* for vFAIL to print correctly */
11282 vFAIL("Sequence (? incomplete");
11284 default: /* e.g., (?i) */
11285 RExC_parse = (char *) seqstart + 1;
11287 parse_lparen_question_flags(pRExC_state);
11288 if (UCHARAT(RExC_parse) != ':') {
11289 if (RExC_parse < RExC_end)
11290 nextchar(pRExC_state);
11295 nextchar(pRExC_state);
11300 else if (!(RExC_flags & RXf_PMf_NOCAPTURE)) { /* (...) */
11305 ret = reganode(pRExC_state, OPEN, parno);
11307 if (!RExC_nestroot)
11308 RExC_nestroot = parno;
11309 if (RExC_open_parens && !RExC_open_parens[parno])
11311 DEBUG_OPTIMISE_MORE_r(Perl_re_printf( aTHX_
11312 "%*s%*s Setting open paren #%" IVdf " to %d\n",
11313 22, "| |", (int)(depth * 2 + 1), "",
11314 (IV)parno, REG_NODE_NUM(ret)));
11315 RExC_open_parens[parno]= ret;
11318 Set_Node_Length(ret, 1); /* MJD */
11319 Set_Node_Offset(ret, RExC_parse); /* MJD */
11322 /* with RXf_PMf_NOCAPTURE treat (...) as (?:...) */
11331 /* Pick up the branches, linking them together. */
11332 parse_start = RExC_parse; /* MJD */
11333 br = regbranch(pRExC_state, &flags, 1,depth+1);
11335 /* branch_len = (paren != 0); */
11338 if (flags & (RESTART_PASS1|NEED_UTF8)) {
11339 *flagp = flags & (RESTART_PASS1|NEED_UTF8);
11342 FAIL2("panic: regbranch returned NULL, flags=%#" UVxf, (UV) flags);
11344 if (*RExC_parse == '|') {
11345 if (!SIZE_ONLY && RExC_extralen) {
11346 reginsert(pRExC_state, BRANCHJ, br, depth+1);
11349 reginsert(pRExC_state, BRANCH, br, depth+1);
11350 Set_Node_Length(br, paren != 0);
11351 Set_Node_Offset_To_R(br-RExC_emit_start, parse_start-RExC_start);
11355 RExC_extralen += 1; /* For BRANCHJ-BRANCH. */
11357 else if (paren == ':') {
11358 *flagp |= flags&SIMPLE;
11360 if (is_open) { /* Starts with OPEN. */
11361 REGTAIL(pRExC_state, ret, br); /* OPEN -> first. */
11363 else if (paren != '?') /* Not Conditional */
11365 *flagp |= flags & (SPSTART | HASWIDTH | POSTPONED);
11367 while (*RExC_parse == '|') {
11368 if (!SIZE_ONLY && RExC_extralen) {
11369 ender = reganode(pRExC_state, LONGJMP,0);
11371 /* Append to the previous. */
11372 REGTAIL(pRExC_state, NEXTOPER(NEXTOPER(lastbr)), ender);
11375 RExC_extralen += 2; /* Account for LONGJMP. */
11376 nextchar(pRExC_state);
11377 if (freeze_paren) {
11378 if (RExC_npar > after_freeze)
11379 after_freeze = RExC_npar;
11380 RExC_npar = freeze_paren;
11382 br = regbranch(pRExC_state, &flags, 0, depth+1);
11385 if (flags & (RESTART_PASS1|NEED_UTF8)) {
11386 *flagp = flags & (RESTART_PASS1|NEED_UTF8);
11389 FAIL2("panic: regbranch returned NULL, flags=%#" UVxf, (UV) flags);
11391 REGTAIL(pRExC_state, lastbr, br); /* BRANCH -> BRANCH. */
11393 *flagp |= flags & (SPSTART | HASWIDTH | POSTPONED);
11396 if (have_branch || paren != ':') {
11397 /* Make a closing node, and hook it on the end. */
11400 ender = reg_node(pRExC_state, TAIL);
11403 ender = reganode(pRExC_state, CLOSE, parno);
11404 if ( RExC_close_parens ) {
11405 DEBUG_OPTIMISE_MORE_r(Perl_re_printf( aTHX_
11406 "%*s%*s Setting close paren #%" IVdf " to %d\n",
11407 22, "| |", (int)(depth * 2 + 1), "", (IV)parno, REG_NODE_NUM(ender)));
11408 RExC_close_parens[parno]= ender;
11409 if (RExC_nestroot == parno)
11412 Set_Node_Offset(ender,RExC_parse+1); /* MJD */
11413 Set_Node_Length(ender,1); /* MJD */
11419 *flagp &= ~HASWIDTH;
11422 ender = reg_node(pRExC_state, SUCCEED);
11425 ender = reg_node(pRExC_state, END);
11427 assert(!RExC_end_op); /* there can only be one! */
11428 RExC_end_op = ender;
11429 if (RExC_close_parens) {
11430 DEBUG_OPTIMISE_MORE_r(Perl_re_printf( aTHX_
11431 "%*s%*s Setting close paren #0 (END) to %d\n",
11432 22, "| |", (int)(depth * 2 + 1), "", REG_NODE_NUM(ender)));
11434 RExC_close_parens[0]= ender;
11439 DEBUG_PARSE_r(if (!SIZE_ONLY) {
11440 DEBUG_PARSE_MSG("lsbr");
11441 regprop(RExC_rx, RExC_mysv1, lastbr, NULL, pRExC_state);
11442 regprop(RExC_rx, RExC_mysv2, ender, NULL, pRExC_state);
11443 Perl_re_printf( aTHX_ "~ tying lastbr %s (%" IVdf ") to ender %s (%" IVdf ") offset %" IVdf "\n",
11444 SvPV_nolen_const(RExC_mysv1),
11445 (IV)REG_NODE_NUM(lastbr),
11446 SvPV_nolen_const(RExC_mysv2),
11447 (IV)REG_NODE_NUM(ender),
11448 (IV)(ender - lastbr)
11451 REGTAIL(pRExC_state, lastbr, ender);
11453 if (have_branch && !SIZE_ONLY) {
11454 char is_nothing= 1;
11456 RExC_seen |= REG_TOP_LEVEL_BRANCHES_SEEN;
11458 /* Hook the tails of the branches to the closing node. */
11459 for (br = ret; br; br = regnext(br)) {
11460 const U8 op = PL_regkind[OP(br)];
11461 if (op == BRANCH) {
11462 REGTAIL_STUDY(pRExC_state, NEXTOPER(br), ender);
11463 if ( OP(NEXTOPER(br)) != NOTHING
11464 || regnext(NEXTOPER(br)) != ender)
11467 else if (op == BRANCHJ) {
11468 REGTAIL_STUDY(pRExC_state, NEXTOPER(NEXTOPER(br)), ender);
11469 /* for now we always disable this optimisation * /
11470 if ( OP(NEXTOPER(NEXTOPER(br))) != NOTHING
11471 || regnext(NEXTOPER(NEXTOPER(br))) != ender)
11477 br= PL_regkind[OP(ret)] != BRANCH ? regnext(ret) : ret;
11478 DEBUG_PARSE_r(if (!SIZE_ONLY) {
11479 DEBUG_PARSE_MSG("NADA");
11480 regprop(RExC_rx, RExC_mysv1, ret, NULL, pRExC_state);
11481 regprop(RExC_rx, RExC_mysv2, ender, NULL, pRExC_state);
11482 Perl_re_printf( aTHX_ "~ converting ret %s (%" IVdf ") to ender %s (%" IVdf ") offset %" IVdf "\n",
11483 SvPV_nolen_const(RExC_mysv1),
11484 (IV)REG_NODE_NUM(ret),
11485 SvPV_nolen_const(RExC_mysv2),
11486 (IV)REG_NODE_NUM(ender),
11491 if (OP(ender) == TAIL) {
11496 for ( opt= br + 1; opt < ender ; opt++ )
11497 OP(opt)= OPTIMIZED;
11498 NEXT_OFF(br)= ender - br;
11506 static const char parens[] = "=!<,>";
11508 if (paren && (p = strchr(parens, paren))) {
11509 U8 node = ((p - parens) % 2) ? UNLESSM : IFMATCH;
11510 int flag = (p - parens) > 1;
11513 node = SUSPEND, flag = 0;
11514 reginsert(pRExC_state, node,ret, depth+1);
11515 Set_Node_Cur_Length(ret, parse_start);
11516 Set_Node_Offset(ret, parse_start + 1);
11518 REGTAIL_STUDY(pRExC_state, ret, reg_node(pRExC_state, TAIL));
11522 /* Check for proper termination. */
11524 /* restore original flags, but keep (?p) and, if we've changed from /d
11525 * rules to /u, keep the /u */
11526 RExC_flags = oregflags | (RExC_flags & RXf_PMf_KEEPCOPY);
11527 if (DEPENDS_SEMANTICS && RExC_uni_semantics) {
11528 set_regex_charset(&RExC_flags, REGEX_UNICODE_CHARSET);
11530 if (RExC_parse >= RExC_end || UCHARAT(RExC_parse) != ')') {
11531 RExC_parse = oregcomp_parse;
11532 vFAIL("Unmatched (");
11534 nextchar(pRExC_state);
11536 else if (!paren && RExC_parse < RExC_end) {
11537 if (*RExC_parse == ')') {
11539 vFAIL("Unmatched )");
11542 FAIL("Junk on end of regexp"); /* "Can't happen". */
11543 NOT_REACHED; /* NOTREACHED */
11546 if (RExC_in_lookbehind) {
11547 RExC_in_lookbehind--;
11549 if (after_freeze > RExC_npar)
11550 RExC_npar = after_freeze;
11555 - regbranch - one alternative of an | operator
11557 * Implements the concatenation operator.
11559 * Returns NULL, setting *flagp to RESTART_PASS1 if the sizing scan needs to be
11560 * restarted, or'd with NEED_UTF8 if the pattern needs to be upgraded to UTF-8
11563 S_regbranch(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, I32 first, U32 depth)
11566 regnode *chain = NULL;
11568 I32 flags = 0, c = 0;
11569 GET_RE_DEBUG_FLAGS_DECL;
11571 PERL_ARGS_ASSERT_REGBRANCH;
11573 DEBUG_PARSE("brnc");
11578 if (!SIZE_ONLY && RExC_extralen)
11579 ret = reganode(pRExC_state, BRANCHJ,0);
11581 ret = reg_node(pRExC_state, BRANCH);
11582 Set_Node_Length(ret, 1);
11586 if (!first && SIZE_ONLY)
11587 RExC_extralen += 1; /* BRANCHJ */
11589 *flagp = WORST; /* Tentatively. */
11591 skip_to_be_ignored_text(pRExC_state, &RExC_parse,
11592 FALSE /* Don't force to /x */ );
11593 while (RExC_parse < RExC_end && *RExC_parse != '|' && *RExC_parse != ')') {
11594 flags &= ~TRYAGAIN;
11595 latest = regpiece(pRExC_state, &flags,depth+1);
11596 if (latest == NULL) {
11597 if (flags & TRYAGAIN)
11599 if (flags & (RESTART_PASS1|NEED_UTF8)) {
11600 *flagp = flags & (RESTART_PASS1|NEED_UTF8);
11603 FAIL2("panic: regpiece returned NULL, flags=%#" UVxf, (UV) flags);
11605 else if (ret == NULL)
11607 *flagp |= flags&(HASWIDTH|POSTPONED);
11608 if (chain == NULL) /* First piece. */
11609 *flagp |= flags&SPSTART;
11611 /* FIXME adding one for every branch after the first is probably
11612 * excessive now we have TRIE support. (hv) */
11614 REGTAIL(pRExC_state, chain, latest);
11619 if (chain == NULL) { /* Loop ran zero times. */
11620 chain = reg_node(pRExC_state, NOTHING);
11625 *flagp |= flags&SIMPLE;
11632 - regpiece - something followed by possible quantifier * + ? {n,m}
11634 * Note that the branching code sequences used for ? and the general cases
11635 * of * and + are somewhat optimized: they use the same NOTHING node as
11636 * both the endmarker for their branch list and the body of the last branch.
11637 * It might seem that this node could be dispensed with entirely, but the
11638 * endmarker role is not redundant.
11640 * Returns NULL, setting *flagp to TRYAGAIN if regatom() returns NULL with
11642 * Returns NULL, setting *flagp to RESTART_PASS1 if the sizing scan needs to be
11643 * restarted, or'd with NEED_UTF8 if the pattern needs to be upgraded to UTF-8
11646 S_regpiece(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, U32 depth)
11652 const char * const origparse = RExC_parse;
11654 I32 max = REG_INFTY;
11655 #ifdef RE_TRACK_PATTERN_OFFSETS
11658 const char *maxpos = NULL;
11661 /* Save the original in case we change the emitted regop to a FAIL. */
11662 regnode * const orig_emit = RExC_emit;
11664 GET_RE_DEBUG_FLAGS_DECL;
11666 PERL_ARGS_ASSERT_REGPIECE;
11668 DEBUG_PARSE("piec");
11670 ret = regatom(pRExC_state, &flags,depth+1);
11672 if (flags & (TRYAGAIN|RESTART_PASS1|NEED_UTF8))
11673 *flagp |= flags & (TRYAGAIN|RESTART_PASS1|NEED_UTF8);
11675 FAIL2("panic: regatom returned NULL, flags=%#" UVxf, (UV) flags);
11681 if (op == '{' && regcurly(RExC_parse)) {
11683 #ifdef RE_TRACK_PATTERN_OFFSETS
11684 parse_start = RExC_parse; /* MJD */
11686 next = RExC_parse + 1;
11687 while (isDIGIT(*next) || *next == ',') {
11688 if (*next == ',') {
11696 if (*next == '}') { /* got one */
11697 const char* endptr;
11701 if (isDIGIT(*RExC_parse)) {
11702 if (!grok_atoUV(RExC_parse, &uv, &endptr))
11703 vFAIL("Invalid quantifier in {,}");
11704 if (uv >= REG_INFTY)
11705 vFAIL2("Quantifier in {,} bigger than %d", REG_INFTY - 1);
11710 if (*maxpos == ',')
11713 maxpos = RExC_parse;
11714 if (isDIGIT(*maxpos)) {
11715 if (!grok_atoUV(maxpos, &uv, &endptr))
11716 vFAIL("Invalid quantifier in {,}");
11717 if (uv >= REG_INFTY)
11718 vFAIL2("Quantifier in {,} bigger than %d", REG_INFTY - 1);
11721 max = REG_INFTY; /* meaning "infinity" */
11724 nextchar(pRExC_state);
11725 if (max < min) { /* If can't match, warn and optimize to fail
11727 reginsert(pRExC_state, OPFAIL, orig_emit, depth+1);
11729 ckWARNreg(RExC_parse, "Quantifier {n,m} with n > m can't match");
11730 NEXT_OFF(orig_emit)= regarglen[OPFAIL] + NODE_STEP_REGNODE;
11734 else if (min == max && *RExC_parse == '?')
11737 ckWARN2reg(RExC_parse + 1,
11738 "Useless use of greediness modifier '%c'",
11744 if ((flags&SIMPLE)) {
11745 if (min == 0 && max == REG_INFTY) {
11746 reginsert(pRExC_state, STAR, ret, depth+1);
11749 RExC_seen |= REG_UNBOUNDED_QUANTIFIER_SEEN;
11752 if (min == 1 && max == REG_INFTY) {
11753 reginsert(pRExC_state, PLUS, ret, depth+1);
11756 RExC_seen |= REG_UNBOUNDED_QUANTIFIER_SEEN;
11759 MARK_NAUGHTY_EXP(2, 2);
11760 reginsert(pRExC_state, CURLY, ret, depth+1);
11761 Set_Node_Offset(ret, parse_start+1); /* MJD */
11762 Set_Node_Cur_Length(ret, parse_start);
11765 regnode * const w = reg_node(pRExC_state, WHILEM);
11768 REGTAIL(pRExC_state, ret, w);
11769 if (!SIZE_ONLY && RExC_extralen) {
11770 reginsert(pRExC_state, LONGJMP,ret, depth+1);
11771 reginsert(pRExC_state, NOTHING,ret, depth+1);
11772 NEXT_OFF(ret) = 3; /* Go over LONGJMP. */
11774 reginsert(pRExC_state, CURLYX,ret, depth+1);
11776 Set_Node_Offset(ret, parse_start+1);
11777 Set_Node_Length(ret,
11778 op == '{' ? (RExC_parse - parse_start) : 1);
11780 if (!SIZE_ONLY && RExC_extralen)
11781 NEXT_OFF(ret) = 3; /* Go over NOTHING to LONGJMP. */
11782 REGTAIL(pRExC_state, ret, reg_node(pRExC_state, NOTHING));
11784 RExC_whilem_seen++, RExC_extralen += 3;
11785 MARK_NAUGHTY_EXP(1, 4); /* compound interest */
11792 *flagp |= HASWIDTH;
11794 ARG1_SET(ret, (U16)min);
11795 ARG2_SET(ret, (U16)max);
11797 if (max == REG_INFTY)
11798 RExC_seen |= REG_UNBOUNDED_QUANTIFIER_SEEN;
11804 if (!ISMULT1(op)) {
11809 #if 0 /* Now runtime fix should be reliable. */
11811 /* if this is reinstated, don't forget to put this back into perldiag:
11813 =item Regexp *+ operand could be empty at {#} in regex m/%s/
11815 (F) The part of the regexp subject to either the * or + quantifier
11816 could match an empty string. The {#} shows in the regular
11817 expression about where the problem was discovered.
11821 if (!(flags&HASWIDTH) && op != '?')
11822 vFAIL("Regexp *+ operand could be empty");
11825 #ifdef RE_TRACK_PATTERN_OFFSETS
11826 parse_start = RExC_parse;
11828 nextchar(pRExC_state);
11830 *flagp = (op != '+') ? (WORST|SPSTART|HASWIDTH) : (WORST|HASWIDTH);
11836 else if (op == '+') {
11840 else if (op == '?') {
11845 if (!SIZE_ONLY && !(flags&(HASWIDTH|POSTPONED)) && max > REG_INFTY/3) {
11846 SAVEFREESV(RExC_rx_sv); /* in case of fatal warnings */
11847 ckWARN2reg(RExC_parse,
11848 "%" UTF8f " matches null string many times",
11849 UTF8fARG(UTF, (RExC_parse >= origparse
11850 ? RExC_parse - origparse
11853 (void)ReREFCNT_inc(RExC_rx_sv);
11856 if (*RExC_parse == '?') {
11857 nextchar(pRExC_state);
11858 reginsert(pRExC_state, MINMOD, ret, depth+1);
11859 REGTAIL(pRExC_state, ret, ret + NODE_STEP_REGNODE);
11861 else if (*RExC_parse == '+') {
11863 nextchar(pRExC_state);
11864 ender = reg_node(pRExC_state, SUCCEED);
11865 REGTAIL(pRExC_state, ret, ender);
11866 reginsert(pRExC_state, SUSPEND, ret, depth+1);
11868 ender = reg_node(pRExC_state, TAIL);
11869 REGTAIL(pRExC_state, ret, ender);
11872 if (ISMULT2(RExC_parse)) {
11874 vFAIL("Nested quantifiers");
11881 S_grok_bslash_N(pTHX_ RExC_state_t *pRExC_state,
11890 /* This routine teases apart the various meanings of \N and returns
11891 * accordingly. The input parameters constrain which meaning(s) is/are valid
11892 * in the current context.
11894 * Exactly one of <node_p> and <code_point_p> must be non-NULL.
11896 * If <code_point_p> is not NULL, the context is expecting the result to be a
11897 * single code point. If this \N instance turns out to a single code point,
11898 * the function returns TRUE and sets *code_point_p to that code point.
11900 * If <node_p> is not NULL, the context is expecting the result to be one of
11901 * the things representable by a regnode. If this \N instance turns out to be
11902 * one such, the function generates the regnode, returns TRUE and sets *node_p
11903 * to point to that regnode.
11905 * If this instance of \N isn't legal in any context, this function will
11906 * generate a fatal error and not return.
11908 * On input, RExC_parse should point to the first char following the \N at the
11909 * time of the call. On successful return, RExC_parse will have been updated
11910 * to point to just after the sequence identified by this routine. Also
11911 * *flagp has been updated as needed.
11913 * When there is some problem with the current context and this \N instance,
11914 * the function returns FALSE, without advancing RExC_parse, nor setting
11915 * *node_p, nor *code_point_p, nor *flagp.
11917 * If <cp_count> is not NULL, the caller wants to know the length (in code
11918 * points) that this \N sequence matches. This is set even if the function
11919 * returns FALSE, as detailed below.
11921 * There are 5 possibilities here, as detailed in the next 5 paragraphs.
11923 * Probably the most common case is for the \N to specify a single code point.
11924 * *cp_count will be set to 1, and *code_point_p will be set to that code
11927 * Another possibility is for the input to be an empty \N{}, which for
11928 * backwards compatibility we accept. *cp_count will be set to 0. *node_p
11929 * will be set to a generated NOTHING node.
11931 * Still another possibility is for the \N to mean [^\n]. *cp_count will be
11932 * set to 0. *node_p will be set to a generated REG_ANY node.
11934 * The fourth possibility is that \N resolves to a sequence of more than one
11935 * code points. *cp_count will be set to the number of code points in the
11936 * sequence. *node_p * will be set to a generated node returned by this
11937 * function calling S_reg().
11939 * The final possibility is that it is premature to be calling this function;
11940 * that pass1 needs to be restarted. This can happen when this changes from
11941 * /d to /u rules, or when the pattern needs to be upgraded to UTF-8. The
11942 * latter occurs only when the fourth possibility would otherwise be in
11943 * effect, and is because one of those code points requires the pattern to be
11944 * recompiled as UTF-8. The function returns FALSE, and sets the
11945 * RESTART_PASS1 and NEED_UTF8 flags in *flagp, as appropriate. When this
11946 * happens, the caller needs to desist from continuing parsing, and return
11947 * this information to its caller. This is not set for when there is only one
11948 * code point, as this can be called as part of an ANYOF node, and they can
11949 * store above-Latin1 code points without the pattern having to be in UTF-8.
11951 * For non-single-quoted regexes, the tokenizer has resolved character and
11952 * sequence names inside \N{...} into their Unicode values, normalizing the
11953 * result into what we should see here: '\N{U+c1.c2...}', where c1... are the
11954 * hex-represented code points in the sequence. This is done there because
11955 * the names can vary based on what charnames pragma is in scope at the time,
11956 * so we need a way to take a snapshot of what they resolve to at the time of
11957 * the original parse. [perl #56444].
11959 * That parsing is skipped for single-quoted regexes, so we may here get
11960 * '\N{NAME}'. This is a fatal error. These names have to be resolved by the
11961 * parser. But if the single-quoted regex is something like '\N{U+41}', that
11962 * is legal and handled here. The code point is Unicode, and has to be
11963 * translated into the native character set for non-ASCII platforms.
11966 char * endbrace; /* points to '}' following the name */
11967 char *endchar; /* Points to '.' or '}' ending cur char in the input
11969 char* p = RExC_parse; /* Temporary */
11971 GET_RE_DEBUG_FLAGS_DECL;
11973 PERL_ARGS_ASSERT_GROK_BSLASH_N;
11975 GET_RE_DEBUG_FLAGS;
11977 assert(cBOOL(node_p) ^ cBOOL(code_point_p)); /* Exactly one should be set */
11978 assert(! (node_p && cp_count)); /* At most 1 should be set */
11980 if (cp_count) { /* Initialize return for the most common case */
11984 /* The [^\n] meaning of \N ignores spaces and comments under the /x
11985 * modifier. The other meanings do not, so use a temporary until we find
11986 * out which we are being called with */
11987 skip_to_be_ignored_text(pRExC_state, &p,
11988 FALSE /* Don't force to /x */ );
11990 /* Disambiguate between \N meaning a named character versus \N meaning
11991 * [^\n]. The latter is assumed when the {...} following the \N is a legal
11992 * quantifier, or there is no '{' at all */
11993 if (*p != '{' || regcurly(p)) {
12003 *node_p = reg_node(pRExC_state, REG_ANY);
12004 *flagp |= HASWIDTH|SIMPLE;
12006 Set_Node_Length(*node_p, 1); /* MJD */
12010 /* Here, we have decided it should be a named character or sequence */
12012 /* The test above made sure that the next real character is a '{', but
12013 * under the /x modifier, it could be separated by space (or a comment and
12014 * \n) and this is not allowed (for consistency with \x{...} and the
12015 * tokenizer handling of \N{NAME}). */
12016 if (*RExC_parse != '{') {
12017 vFAIL("Missing braces on \\N{}");
12020 RExC_parse++; /* Skip past the '{' */
12022 if (! (endbrace = strchr(RExC_parse, '}'))) { /* no trailing brace */
12023 vFAIL2("Missing right brace on \\%c{}", 'N');
12025 else if(!(endbrace == RExC_parse /* nothing between the {} */
12026 || (endbrace - RExC_parse >= 2 /* U+ (bad hex is checked... */
12027 && strnEQ(RExC_parse, "U+", 2)))) /* ... below for a better
12030 RExC_parse = endbrace; /* position msg's '<--HERE' */
12031 vFAIL("\\N{NAME} must be resolved by the lexer");
12034 REQUIRE_UNI_RULES(flagp, FALSE); /* Unicode named chars imply Unicode
12037 if (endbrace == RExC_parse) { /* empty: \N{} */
12039 RExC_parse++; /* Position after the "}" */
12040 vFAIL("Zero length \\N{}");
12045 nextchar(pRExC_state);
12050 *node_p = reg_node(pRExC_state,NOTHING);
12054 RExC_parse += 2; /* Skip past the 'U+' */
12056 /* Because toke.c has generated a special construct for us guaranteed not
12057 * to have NULs, we can use a str function */
12058 endchar = RExC_parse + strcspn(RExC_parse, ".}");
12060 /* Code points are separated by dots. If none, there is only one code
12061 * point, and is terminated by the brace */
12063 if (endchar >= endbrace) {
12064 STRLEN length_of_hex;
12065 I32 grok_hex_flags;
12067 /* Here, exactly one code point. If that isn't what is wanted, fail */
12068 if (! code_point_p) {
12073 /* Convert code point from hex */
12074 length_of_hex = (STRLEN)(endchar - RExC_parse);
12075 grok_hex_flags = PERL_SCAN_ALLOW_UNDERSCORES
12076 | PERL_SCAN_DISALLOW_PREFIX
12078 /* No errors in the first pass (See [perl
12079 * #122671].) We let the code below find the
12080 * errors when there are multiple chars. */
12082 ? PERL_SCAN_SILENT_ILLDIGIT
12085 /* This routine is the one place where both single- and double-quotish
12086 * \N{U+xxxx} are evaluated. The value is a Unicode code point which
12087 * must be converted to native. */
12088 *code_point_p = UNI_TO_NATIVE(grok_hex(RExC_parse,
12093 /* The tokenizer should have guaranteed validity, but it's possible to
12094 * bypass it by using single quoting, so check. Don't do the check
12095 * here when there are multiple chars; we do it below anyway. */
12096 if (length_of_hex == 0
12097 || length_of_hex != (STRLEN)(endchar - RExC_parse) )
12099 RExC_parse += length_of_hex; /* Includes all the valid */
12100 RExC_parse += (RExC_orig_utf8) /* point to after 1st invalid */
12101 ? UTF8SKIP(RExC_parse)
12103 /* Guard against malformed utf8 */
12104 if (RExC_parse >= endchar) {
12105 RExC_parse = endchar;
12107 vFAIL("Invalid hexadecimal number in \\N{U+...}");
12110 RExC_parse = endbrace + 1;
12113 else { /* Is a multiple character sequence */
12114 SV * substitute_parse;
12116 char *orig_end = RExC_end;
12117 char *save_start = RExC_start;
12120 /* Count the code points, if desired, in the sequence */
12123 while (RExC_parse < endbrace) {
12124 /* Point to the beginning of the next character in the sequence. */
12125 RExC_parse = endchar + 1;
12126 endchar = RExC_parse + strcspn(RExC_parse, ".}");
12131 /* Fail if caller doesn't want to handle a multi-code-point sequence.
12132 * But don't backup up the pointer if the caller want to know how many
12133 * code points there are (they can then handle things) */
12141 /* What is done here is to convert this to a sub-pattern of the form
12142 * \x{char1}\x{char2}... and then call reg recursively to parse it
12143 * (enclosing in "(?: ... )" ). That way, it retains its atomicness,
12144 * while not having to worry about special handling that some code
12145 * points may have. */
12147 substitute_parse = newSVpvs("?:");
12149 while (RExC_parse < endbrace) {
12151 /* Convert to notation the rest of the code understands */
12152 sv_catpv(substitute_parse, "\\x{");
12153 sv_catpvn(substitute_parse, RExC_parse, endchar - RExC_parse);
12154 sv_catpv(substitute_parse, "}");
12156 /* Point to the beginning of the next character in the sequence. */
12157 RExC_parse = endchar + 1;
12158 endchar = RExC_parse + strcspn(RExC_parse, ".}");
12161 sv_catpv(substitute_parse, ")");
12163 RExC_parse = RExC_start = RExC_adjusted_start = SvPV(substitute_parse,
12166 /* Don't allow empty number */
12167 if (len < (STRLEN) 8) {
12168 RExC_parse = endbrace;
12169 vFAIL("Invalid hexadecimal number in \\N{U+...}");
12171 RExC_end = RExC_parse + len;
12173 /* The values are Unicode, and therefore not subject to recoding, but
12174 * have to be converted to native on a non-Unicode (meaning non-ASCII)
12177 RExC_recode_x_to_native = 1;
12181 if (!(*node_p = reg(pRExC_state, 1, &flags, depth+1))) {
12182 if (flags & (RESTART_PASS1|NEED_UTF8)) {
12183 *flagp = flags & (RESTART_PASS1|NEED_UTF8);
12186 FAIL2("panic: reg returned NULL to grok_bslash_N, flags=%#" UVxf,
12189 *flagp |= flags&(HASWIDTH|SPSTART|SIMPLE|POSTPONED);
12192 /* Restore the saved values */
12193 RExC_start = RExC_adjusted_start = save_start;
12194 RExC_parse = endbrace;
12195 RExC_end = orig_end;
12197 RExC_recode_x_to_native = 0;
12200 SvREFCNT_dec_NN(substitute_parse);
12201 nextchar(pRExC_state);
12208 PERL_STATIC_INLINE U8
12209 S_compute_EXACTish(RExC_state_t *pRExC_state)
12213 PERL_ARGS_ASSERT_COMPUTE_EXACTISH;
12221 op = get_regex_charset(RExC_flags);
12222 if (op >= REGEX_ASCII_RESTRICTED_CHARSET) {
12223 op--; /* /a is same as /u, and map /aa's offset to what /a's would have
12224 been, so there is no hole */
12227 return op + EXACTF;
12230 PERL_STATIC_INLINE void
12231 S_alloc_maybe_populate_EXACT(pTHX_ RExC_state_t *pRExC_state,
12232 regnode *node, I32* flagp, STRLEN len, UV code_point,
12235 /* This knows the details about sizing an EXACTish node, setting flags for
12236 * it (by setting <*flagp>, and potentially populating it with a single
12239 * If <len> (the length in bytes) is non-zero, this function assumes that
12240 * the node has already been populated, and just does the sizing. In this
12241 * case <code_point> should be the final code point that has already been
12242 * placed into the node. This value will be ignored except that under some
12243 * circumstances <*flagp> is set based on it.
12245 * If <len> is zero, the function assumes that the node is to contain only
12246 * the single character given by <code_point> and calculates what <len>
12247 * should be. In pass 1, it sizes the node appropriately. In pass 2, it
12248 * additionally will populate the node's STRING with <code_point> or its
12251 * In both cases <*flagp> is appropriately set
12253 * It knows that under FOLD, the Latin Sharp S and UTF characters above
12254 * 255, must be folded (the former only when the rules indicate it can
12257 * When it does the populating, it looks at the flag 'downgradable'. If
12258 * true with a node that folds, it checks if the single code point
12259 * participates in a fold, and if not downgrades the node to an EXACT.
12260 * This helps the optimizer */
12262 bool len_passed_in = cBOOL(len != 0);
12263 U8 character[UTF8_MAXBYTES_CASE+1];
12265 PERL_ARGS_ASSERT_ALLOC_MAYBE_POPULATE_EXACT;
12267 /* Don't bother to check for downgrading in PASS1, as it doesn't make any
12268 * sizing difference, and is extra work that is thrown away */
12269 if (downgradable && ! PASS2) {
12270 downgradable = FALSE;
12273 if (! len_passed_in) {
12275 if (UVCHR_IS_INVARIANT(code_point)) {
12276 if (LOC || ! FOLD) { /* /l defers folding until runtime */
12277 *character = (U8) code_point;
12279 else { /* Here is /i and not /l. (toFOLD() is defined on just
12280 ASCII, which isn't the same thing as INVARIANT on
12281 EBCDIC, but it works there, as the extra invariants
12282 fold to themselves) */
12283 *character = toFOLD((U8) code_point);
12285 /* We can downgrade to an EXACT node if this character
12286 * isn't a folding one. Note that this assumes that
12287 * nothing above Latin1 folds to some other invariant than
12288 * one of these alphabetics; otherwise we would also have
12290 * && (! HAS_NONLATIN1_FOLD_CLOSURE(code_point)
12291 * || ASCII_FOLD_RESTRICTED))
12293 if (downgradable && PL_fold[code_point] == code_point) {
12299 else if (FOLD && (! LOC
12300 || ! is_PROBLEMATIC_LOCALE_FOLD_cp(code_point)))
12301 { /* Folding, and ok to do so now */
12302 UV folded = _to_uni_fold_flags(
12306 FOLD_FLAGS_FULL | ((ASCII_FOLD_RESTRICTED)
12307 ? FOLD_FLAGS_NOMIX_ASCII
12310 && folded == code_point /* This quickly rules out many
12311 cases, avoiding the
12312 _invlist_contains_cp() overhead
12314 && ! _invlist_contains_cp(PL_utf8_foldable, code_point))
12321 else if (code_point <= MAX_UTF8_TWO_BYTE) {
12323 /* Not folding this cp, and can output it directly */
12324 *character = UTF8_TWO_BYTE_HI(code_point);
12325 *(character + 1) = UTF8_TWO_BYTE_LO(code_point);
12329 uvchr_to_utf8( character, code_point);
12330 len = UTF8SKIP(character);
12332 } /* Else pattern isn't UTF8. */
12334 *character = (U8) code_point;
12336 } /* Else is folded non-UTF8 */
12337 #if UNICODE_MAJOR_VERSION > 3 /* no multifolds in early Unicode */ \
12338 || (UNICODE_MAJOR_VERSION == 3 && ( UNICODE_DOT_VERSION > 0) \
12339 || UNICODE_DOT_DOT_VERSION > 0)
12340 else if (LIKELY(code_point != LATIN_SMALL_LETTER_SHARP_S)) {
12344 /* We don't fold any non-UTF8 except possibly the Sharp s (see
12345 * comments at join_exact()); */
12346 *character = (U8) code_point;
12349 /* Can turn into an EXACT node if we know the fold at compile time,
12350 * and it folds to itself and doesn't particpate in other folds */
12353 && PL_fold_latin1[code_point] == code_point
12354 && (! HAS_NONLATIN1_FOLD_CLOSURE(code_point)
12355 || (isASCII(code_point) && ASCII_FOLD_RESTRICTED)))
12359 } /* else is Sharp s. May need to fold it */
12360 else if (AT_LEAST_UNI_SEMANTICS && ! ASCII_FOLD_RESTRICTED) {
12362 *(character + 1) = 's';
12366 *character = LATIN_SMALL_LETTER_SHARP_S;
12372 RExC_size += STR_SZ(len);
12375 RExC_emit += STR_SZ(len);
12376 STR_LEN(node) = len;
12377 if (! len_passed_in) {
12378 Copy((char *) character, STRING(node), len, char);
12382 *flagp |= HASWIDTH;
12384 /* A single character node is SIMPLE, except for the special-cased SHARP S
12386 if ((len == 1 || (UTF && len == UVCHR_SKIP(code_point)))
12387 #if UNICODE_MAJOR_VERSION > 3 /* no multifolds in early Unicode */ \
12388 || (UNICODE_MAJOR_VERSION == 3 && ( UNICODE_DOT_VERSION > 0) \
12389 || UNICODE_DOT_DOT_VERSION > 0)
12390 && ( code_point != LATIN_SMALL_LETTER_SHARP_S
12391 || ! FOLD || ! DEPENDS_SEMANTICS)
12397 /* The OP may not be well defined in PASS1 */
12398 if (PASS2 && OP(node) == EXACTFL) {
12399 RExC_contains_locale = 1;
12404 /* Parse backref decimal value, unless it's too big to sensibly be a backref,
12405 * in which case return I32_MAX (rather than possibly 32-bit wrapping) */
12408 S_backref_value(char *p)
12410 const char* endptr;
12412 if (grok_atoUV(p, &val, &endptr) && val <= I32_MAX)
12419 - regatom - the lowest level
12421 Try to identify anything special at the start of the current parse position.
12422 If there is, then handle it as required. This may involve generating a
12423 single regop, such as for an assertion; or it may involve recursing, such as
12424 to handle a () structure.
12426 If the string doesn't start with something special then we gobble up
12427 as much literal text as we can. If we encounter a quantifier, we have to
12428 back off the final literal character, as that quantifier applies to just it
12429 and not to the whole string of literals.
12431 Once we have been able to handle whatever type of thing started the
12432 sequence, we return.
12434 Note: we have to be careful with escapes, as they can be both literal
12435 and special, and in the case of \10 and friends, context determines which.
12437 A summary of the code structure is:
12439 switch (first_byte) {
12440 cases for each special:
12441 handle this special;
12444 switch (2nd byte) {
12445 cases for each unambiguous special:
12446 handle this special;
12448 cases for each ambigous special/literal:
12450 if (special) handle here
12452 default: // unambiguously literal:
12455 default: // is a literal char
12458 create EXACTish node for literal;
12459 while (more input and node isn't full) {
12460 switch (input_byte) {
12461 cases for each special;
12462 make sure parse pointer is set so that the next call to
12463 regatom will see this special first
12464 goto loopdone; // EXACTish node terminated by prev. char
12466 append char to EXACTISH node;
12468 get next input byte;
12472 return the generated node;
12474 Specifically there are two separate switches for handling
12475 escape sequences, with the one for handling literal escapes requiring
12476 a dummy entry for all of the special escapes that are actually handled
12479 Returns NULL, setting *flagp to TRYAGAIN if reg() returns NULL with
12481 Returns NULL, setting *flagp to RESTART_PASS1 if the sizing scan needs to be
12482 restarted, or'd with NEED_UTF8 if the pattern needs to be upgraded to UTF-8
12483 Otherwise does not return NULL.
12487 S_regatom(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, U32 depth)
12489 regnode *ret = NULL;
12496 GET_RE_DEBUG_FLAGS_DECL;
12498 *flagp = WORST; /* Tentatively. */
12500 DEBUG_PARSE("atom");
12502 PERL_ARGS_ASSERT_REGATOM;
12505 parse_start = RExC_parse;
12506 assert(RExC_parse < RExC_end);
12507 switch ((U8)*RExC_parse) {
12509 RExC_seen_zerolen++;
12510 nextchar(pRExC_state);
12511 if (RExC_flags & RXf_PMf_MULTILINE)
12512 ret = reg_node(pRExC_state, MBOL);
12514 ret = reg_node(pRExC_state, SBOL);
12515 Set_Node_Length(ret, 1); /* MJD */
12518 nextchar(pRExC_state);
12520 RExC_seen_zerolen++;
12521 if (RExC_flags & RXf_PMf_MULTILINE)
12522 ret = reg_node(pRExC_state, MEOL);
12524 ret = reg_node(pRExC_state, SEOL);
12525 Set_Node_Length(ret, 1); /* MJD */
12528 nextchar(pRExC_state);
12529 if (RExC_flags & RXf_PMf_SINGLELINE)
12530 ret = reg_node(pRExC_state, SANY);
12532 ret = reg_node(pRExC_state, REG_ANY);
12533 *flagp |= HASWIDTH|SIMPLE;
12535 Set_Node_Length(ret, 1); /* MJD */
12539 char * const oregcomp_parse = ++RExC_parse;
12540 ret = regclass(pRExC_state, flagp,depth+1,
12541 FALSE, /* means parse the whole char class */
12542 TRUE, /* allow multi-char folds */
12543 FALSE, /* don't silence non-portable warnings. */
12544 (bool) RExC_strict,
12545 TRUE, /* Allow an optimized regnode result */
12549 if (*flagp & (RESTART_PASS1|NEED_UTF8))
12551 FAIL2("panic: regclass returned NULL to regatom, flags=%#" UVxf,
12554 if (*RExC_parse != ']') {
12555 RExC_parse = oregcomp_parse;
12556 vFAIL("Unmatched [");
12558 nextchar(pRExC_state);
12559 Set_Node_Length(ret, RExC_parse - oregcomp_parse + 1); /* MJD */
12563 nextchar(pRExC_state);
12564 ret = reg(pRExC_state, 2, &flags,depth+1);
12566 if (flags & TRYAGAIN) {
12567 if (RExC_parse >= RExC_end) {
12568 /* Make parent create an empty node if needed. */
12569 *flagp |= TRYAGAIN;
12574 if (flags & (RESTART_PASS1|NEED_UTF8)) {
12575 *flagp = flags & (RESTART_PASS1|NEED_UTF8);
12578 FAIL2("panic: reg returned NULL to regatom, flags=%#" UVxf,
12581 *flagp |= flags&(HASWIDTH|SPSTART|SIMPLE|POSTPONED);
12585 if (flags & TRYAGAIN) {
12586 *flagp |= TRYAGAIN;
12589 vFAIL("Internal urp");
12590 /* Supposed to be caught earlier. */
12596 vFAIL("Quantifier follows nothing");
12601 This switch handles escape sequences that resolve to some kind
12602 of special regop and not to literal text. Escape sequnces that
12603 resolve to literal text are handled below in the switch marked
12606 Every entry in this switch *must* have a corresponding entry
12607 in the literal escape switch. However, the opposite is not
12608 required, as the default for this switch is to jump to the
12609 literal text handling code.
12612 switch ((U8)*RExC_parse) {
12613 /* Special Escapes */
12615 RExC_seen_zerolen++;
12616 ret = reg_node(pRExC_state, SBOL);
12617 /* SBOL is shared with /^/ so we set the flags so we can tell
12618 * /\A/ from /^/ in split. We check ret because first pass we
12619 * have no regop struct to set the flags on. */
12623 goto finish_meta_pat;
12625 ret = reg_node(pRExC_state, GPOS);
12626 RExC_seen |= REG_GPOS_SEEN;
12628 goto finish_meta_pat;
12630 RExC_seen_zerolen++;
12631 ret = reg_node(pRExC_state, KEEPS);
12633 /* XXX:dmq : disabling in-place substitution seems to
12634 * be necessary here to avoid cases of memory corruption, as
12635 * with: C<$_="x" x 80; s/x\K/y/> -- rgs
12637 RExC_seen |= REG_LOOKBEHIND_SEEN;
12638 goto finish_meta_pat;
12640 ret = reg_node(pRExC_state, SEOL);
12642 RExC_seen_zerolen++; /* Do not optimize RE away */
12643 goto finish_meta_pat;
12645 ret = reg_node(pRExC_state, EOS);
12647 RExC_seen_zerolen++; /* Do not optimize RE away */
12648 goto finish_meta_pat;
12650 vFAIL("\\C no longer supported");
12652 ret = reg_node(pRExC_state, CLUMP);
12653 *flagp |= HASWIDTH;
12654 goto finish_meta_pat;
12660 arg = ANYOF_WORDCHAR;
12668 regex_charset charset = get_regex_charset(RExC_flags);
12670 RExC_seen_zerolen++;
12671 RExC_seen |= REG_LOOKBEHIND_SEEN;
12672 op = BOUND + charset;
12674 if (op == BOUNDL) {
12675 RExC_contains_locale = 1;
12678 ret = reg_node(pRExC_state, op);
12680 if (RExC_parse >= RExC_end || *(RExC_parse + 1) != '{') {
12681 FLAGS(ret) = TRADITIONAL_BOUND;
12682 if (PASS2 && op > BOUNDA) { /* /aa is same as /a */
12688 char name = *RExC_parse;
12691 endbrace = strchr(RExC_parse, '}');
12694 vFAIL2("Missing right brace on \\%c{}", name);
12696 /* XXX Need to decide whether to take spaces or not. Should be
12697 * consistent with \p{}, but that currently is SPACE, which
12698 * means vertical too, which seems wrong
12699 * while (isBLANK(*RExC_parse)) {
12702 if (endbrace == RExC_parse) {
12703 RExC_parse++; /* After the '}' */
12704 vFAIL2("Empty \\%c{}", name);
12706 length = endbrace - RExC_parse;
12707 /*while (isBLANK(*(RExC_parse + length - 1))) {
12710 switch (*RExC_parse) {
12713 && (length != 3 || strnNE(RExC_parse + 1, "cb", 2)))
12715 goto bad_bound_type;
12717 FLAGS(ret) = GCB_BOUND;
12720 if (length != 2 || *(RExC_parse + 1) != 'b') {
12721 goto bad_bound_type;
12723 FLAGS(ret) = LB_BOUND;
12726 if (length != 2 || *(RExC_parse + 1) != 'b') {
12727 goto bad_bound_type;
12729 FLAGS(ret) = SB_BOUND;
12732 if (length != 2 || *(RExC_parse + 1) != 'b') {
12733 goto bad_bound_type;
12735 FLAGS(ret) = WB_BOUND;
12739 RExC_parse = endbrace;
12741 "'%" UTF8f "' is an unknown bound type",
12742 UTF8fARG(UTF, length, endbrace - length));
12743 NOT_REACHED; /*NOTREACHED*/
12745 RExC_parse = endbrace;
12746 REQUIRE_UNI_RULES(flagp, NULL);
12748 if (PASS2 && op >= BOUNDA) { /* /aa is same as /a */
12752 /* Don't have to worry about UTF-8, in this message because
12753 * to get here the contents of the \b must be ASCII */
12754 ckWARN4reg(RExC_parse + 1, /* Include the '}' in msg */
12755 "Using /u for '%.*s' instead of /%s",
12757 endbrace - length + 1,
12758 (charset == REGEX_ASCII_RESTRICTED_CHARSET)
12759 ? ASCII_RESTRICT_PAT_MODS
12760 : ASCII_MORE_RESTRICT_PAT_MODS);
12764 if (PASS2 && invert) {
12765 OP(ret) += NBOUND - BOUND;
12767 goto finish_meta_pat;
12775 if (! DEPENDS_SEMANTICS) {
12779 /* \d doesn't have any matches in the upper Latin1 range, hence /d
12780 * is equivalent to /u. Changing to /u saves some branches at
12783 goto join_posix_op_known;
12786 ret = reg_node(pRExC_state, LNBREAK);
12787 *flagp |= HASWIDTH|SIMPLE;
12788 goto finish_meta_pat;
12796 goto join_posix_op_known;
12802 arg = ANYOF_VERTWS;
12804 goto join_posix_op_known;
12814 op = POSIXD + get_regex_charset(RExC_flags);
12815 if (op > POSIXA) { /* /aa is same as /a */
12818 else if (op == POSIXL) {
12819 RExC_contains_locale = 1;
12822 join_posix_op_known:
12825 op += NPOSIXD - POSIXD;
12828 ret = reg_node(pRExC_state, op);
12830 FLAGS(ret) = namedclass_to_classnum(arg);
12833 *flagp |= HASWIDTH|SIMPLE;
12837 nextchar(pRExC_state);
12838 Set_Node_Length(ret, 2); /* MJD */
12844 ret = regclass(pRExC_state, flagp,depth+1,
12845 TRUE, /* means just parse this element */
12846 FALSE, /* don't allow multi-char folds */
12847 FALSE, /* don't silence non-portable warnings. It
12848 would be a bug if these returned
12850 (bool) RExC_strict,
12851 TRUE, /* Allow an optimized regnode result */
12854 if (*flagp & RESTART_PASS1)
12856 /* regclass() can only return RESTART_PASS1 and NEED_UTF8 if
12857 * multi-char folds are allowed. */
12859 FAIL2("panic: regclass returned NULL to regatom, flags=%#" UVxf,
12864 Set_Node_Offset(ret, parse_start);
12865 Set_Node_Cur_Length(ret, parse_start - 2);
12866 nextchar(pRExC_state);
12869 /* Handle \N, \N{} and \N{NAMED SEQUENCE} (the latter meaning the
12870 * \N{...} evaluates to a sequence of more than one code points).
12871 * The function call below returns a regnode, which is our result.
12872 * The parameters cause it to fail if the \N{} evaluates to a
12873 * single code point; we handle those like any other literal. The
12874 * reason that the multicharacter case is handled here and not as
12875 * part of the EXACtish code is because of quantifiers. In
12876 * /\N{BLAH}+/, the '+' applies to the whole thing, and doing it
12877 * this way makes that Just Happen. dmq.
12878 * join_exact() will join this up with adjacent EXACTish nodes
12879 * later on, if appropriate. */
12881 if (grok_bslash_N(pRExC_state,
12882 &ret, /* Want a regnode returned */
12883 NULL, /* Fail if evaluates to a single code
12885 NULL, /* Don't need a count of how many code
12894 if (*flagp & RESTART_PASS1)
12897 /* Here, evaluates to a single code point. Go get that */
12898 RExC_parse = parse_start;
12901 case 'k': /* Handle \k<NAME> and \k'NAME' */
12905 if ( RExC_parse >= RExC_end - 1
12906 || (( ch = RExC_parse[1]) != '<'
12911 /* diag_listed_as: Sequence \%s... not terminated in regex; marked by <-- HERE in m/%s/ */
12912 vFAIL2("Sequence %.2s... not terminated",parse_start);
12915 ret = handle_named_backref(pRExC_state,
12927 case '1': case '2': case '3': case '4':
12928 case '5': case '6': case '7': case '8': case '9':
12933 if (*RExC_parse == 'g') {
12937 if (*RExC_parse == '{') {
12941 if (*RExC_parse == '-') {
12945 if (hasbrace && !isDIGIT(*RExC_parse)) {
12946 if (isrel) RExC_parse--;
12948 goto parse_named_seq;
12951 if (RExC_parse >= RExC_end) {
12952 goto unterminated_g;
12954 num = S_backref_value(RExC_parse);
12956 vFAIL("Reference to invalid group 0");
12957 else if (num == I32_MAX) {
12958 if (isDIGIT(*RExC_parse))
12959 vFAIL("Reference to nonexistent group");
12962 vFAIL("Unterminated \\g... pattern");
12966 num = RExC_npar - num;
12968 vFAIL("Reference to nonexistent or unclosed group");
12972 num = S_backref_value(RExC_parse);
12973 /* bare \NNN might be backref or octal - if it is larger
12974 * than or equal RExC_npar then it is assumed to be an
12975 * octal escape. Note RExC_npar is +1 from the actual
12976 * number of parens. */
12977 /* Note we do NOT check if num == I32_MAX here, as that is
12978 * handled by the RExC_npar check */
12981 /* any numeric escape < 10 is always a backref */
12983 /* any numeric escape < RExC_npar is a backref */
12984 && num >= RExC_npar
12985 /* cannot be an octal escape if it starts with 8 */
12986 && *RExC_parse != '8'
12987 /* cannot be an octal escape it it starts with 9 */
12988 && *RExC_parse != '9'
12991 /* Probably not a backref, instead likely to be an
12992 * octal character escape, e.g. \35 or \777.
12993 * The above logic should make it obvious why using
12994 * octal escapes in patterns is problematic. - Yves */
12995 RExC_parse = parse_start;
13000 /* At this point RExC_parse points at a numeric escape like
13001 * \12 or \88 or something similar, which we should NOT treat
13002 * as an octal escape. It may or may not be a valid backref
13003 * escape. For instance \88888888 is unlikely to be a valid
13005 while (isDIGIT(*RExC_parse))
13008 if (*RExC_parse != '}')
13009 vFAIL("Unterminated \\g{...} pattern");
13013 if (num > (I32)RExC_rx->nparens)
13014 vFAIL("Reference to nonexistent group");
13017 ret = reganode(pRExC_state,
13020 : (ASCII_FOLD_RESTRICTED)
13022 : (AT_LEAST_UNI_SEMANTICS)
13028 *flagp |= HASWIDTH;
13030 /* override incorrect value set in reganode MJD */
13031 Set_Node_Offset(ret, parse_start);
13032 Set_Node_Cur_Length(ret, parse_start-1);
13033 skip_to_be_ignored_text(pRExC_state, &RExC_parse,
13034 FALSE /* Don't force to /x */ );
13038 if (RExC_parse >= RExC_end)
13039 FAIL("Trailing \\");
13042 /* Do not generate "unrecognized" warnings here, we fall
13043 back into the quick-grab loop below */
13044 RExC_parse = parse_start;
13046 } /* end of switch on a \foo sequence */
13051 /* '#' comments should have been spaced over before this function was
13053 assert((RExC_flags & RXf_PMf_EXTENDED) == 0);
13055 if (RExC_flags & RXf_PMf_EXTENDED) {
13056 RExC_parse = reg_skipcomment( pRExC_state, RExC_parse );
13057 if (RExC_parse < RExC_end)
13067 /* Here, we have determined that the next thing is probably a
13068 * literal character. RExC_parse points to the first byte of its
13069 * definition. (It still may be an escape sequence that evaluates
13070 * to a single character) */
13076 #define MAX_NODE_STRING_SIZE 127
13077 char foldbuf[MAX_NODE_STRING_SIZE+UTF8_MAXBYTES_CASE];
13079 U8 upper_parse = MAX_NODE_STRING_SIZE;
13080 U8 node_type = compute_EXACTish(pRExC_state);
13081 bool next_is_quantifier;
13082 char * oldp = NULL;
13084 /* We can convert EXACTF nodes to EXACTFU if they contain only
13085 * characters that match identically regardless of the target
13086 * string's UTF8ness. The reason to do this is that EXACTF is not
13087 * trie-able, EXACTFU is.
13089 * Similarly, we can convert EXACTFL nodes to EXACTFLU8 if they
13090 * contain only above-Latin1 characters (hence must be in UTF8),
13091 * which don't participate in folds with Latin1-range characters,
13092 * as the latter's folds aren't known until runtime. (We don't
13093 * need to figure this out until pass 2) */
13094 bool maybe_exactfu = PASS2
13095 && (node_type == EXACTF || node_type == EXACTFL);
13097 /* If a folding node contains only code points that don't
13098 * participate in folds, it can be changed into an EXACT node,
13099 * which allows the optimizer more things to look for */
13102 ret = reg_node(pRExC_state, node_type);
13104 /* In pass1, folded, we use a temporary buffer instead of the
13105 * actual node, as the node doesn't exist yet */
13106 s = (SIZE_ONLY && FOLD) ? foldbuf : STRING(ret);
13112 /* We look for the EXACTFish to EXACT node optimizaton only if
13113 * folding. (And we don't need to figure this out until pass 2).
13114 * XXX It might actually make sense to split the node into portions
13115 * that are exact and ones that aren't, so that we could later use
13116 * the exact ones to find the longest fixed and floating strings.
13117 * One would want to join them back into a larger node. One could
13118 * use a pseudo regnode like 'EXACT_ORIG_FOLD' */
13119 maybe_exact = FOLD && PASS2;
13121 /* XXX The node can hold up to 255 bytes, yet this only goes to
13122 * 127. I (khw) do not know why. Keeping it somewhat less than
13123 * 255 allows us to not have to worry about overflow due to
13124 * converting to utf8 and fold expansion, but that value is
13125 * 255-UTF8_MAXBYTES_CASE. join_exact() may join adjacent nodes
13126 * split up by this limit into a single one using the real max of
13127 * 255. Even at 127, this breaks under rare circumstances. If
13128 * folding, we do not want to split a node at a character that is a
13129 * non-final in a multi-char fold, as an input string could just
13130 * happen to want to match across the node boundary. The join
13131 * would solve that problem if the join actually happens. But a
13132 * series of more than two nodes in a row each of 127 would cause
13133 * the first join to succeed to get to 254, but then there wouldn't
13134 * be room for the next one, which could at be one of those split
13135 * multi-char folds. I don't know of any fool-proof solution. One
13136 * could back off to end with only a code point that isn't such a
13137 * non-final, but it is possible for there not to be any in the
13140 assert( ! UTF /* Is at the beginning of a character */
13141 || UTF8_IS_INVARIANT(UCHARAT(RExC_parse))
13142 || UTF8_IS_START(UCHARAT(RExC_parse)));
13144 /* Here, we have a literal character. Find the maximal string of
13145 * them in the input that we can fit into a single EXACTish node.
13146 * We quit at the first non-literal or when the node gets full */
13147 for (p = RExC_parse;
13148 len < upper_parse && p < RExC_end;
13153 /* White space has already been ignored */
13154 assert( (RExC_flags & RXf_PMf_EXTENDED) == 0
13155 || ! is_PATWS_safe((p), RExC_end, UTF));
13167 /* Literal Escapes Switch
13169 This switch is meant to handle escape sequences that
13170 resolve to a literal character.
13172 Every escape sequence that represents something
13173 else, like an assertion or a char class, is handled
13174 in the switch marked 'Special Escapes' above in this
13175 routine, but also has an entry here as anything that
13176 isn't explicitly mentioned here will be treated as
13177 an unescaped equivalent literal.
13180 switch ((U8)*++p) {
13181 /* These are all the special escapes. */
13182 case 'A': /* Start assertion */
13183 case 'b': case 'B': /* Word-boundary assertion*/
13184 case 'C': /* Single char !DANGEROUS! */
13185 case 'd': case 'D': /* digit class */
13186 case 'g': case 'G': /* generic-backref, pos assertion */
13187 case 'h': case 'H': /* HORIZWS */
13188 case 'k': case 'K': /* named backref, keep marker */
13189 case 'p': case 'P': /* Unicode property */
13190 case 'R': /* LNBREAK */
13191 case 's': case 'S': /* space class */
13192 case 'v': case 'V': /* VERTWS */
13193 case 'w': case 'W': /* word class */
13194 case 'X': /* eXtended Unicode "combining
13195 character sequence" */
13196 case 'z': case 'Z': /* End of line/string assertion */
13200 /* Anything after here is an escape that resolves to a
13201 literal. (Except digits, which may or may not)
13207 case 'N': /* Handle a single-code point named character. */
13208 RExC_parse = p + 1;
13209 if (! grok_bslash_N(pRExC_state,
13210 NULL, /* Fail if evaluates to
13211 anything other than a
13212 single code point */
13213 &ender, /* The returned single code
13215 NULL, /* Don't need a count of
13216 how many code points */
13221 if (*flagp & NEED_UTF8)
13222 FAIL("panic: grok_bslash_N set NEED_UTF8");
13223 if (*flagp & RESTART_PASS1)
13226 /* Here, it wasn't a single code point. Go close
13227 * up this EXACTish node. The switch() prior to
13228 * this switch handles the other cases */
13229 RExC_parse = p = oldp;
13233 if (ender > 0xff) {
13234 REQUIRE_UTF8(flagp);
13250 ender = ESC_NATIVE;
13260 const char* error_msg;
13262 bool valid = grok_bslash_o(&p,
13265 PASS2, /* out warnings */
13266 (bool) RExC_strict,
13267 TRUE, /* Output warnings
13272 RExC_parse = p; /* going to die anyway; point
13273 to exact spot of failure */
13277 if (ender > 0xff) {
13278 REQUIRE_UTF8(flagp);
13284 UV result = UV_MAX; /* initialize to erroneous
13286 const char* error_msg;
13288 bool valid = grok_bslash_x(&p,
13291 PASS2, /* out warnings */
13292 (bool) RExC_strict,
13293 TRUE, /* Silence warnings
13298 RExC_parse = p; /* going to die anyway; point
13299 to exact spot of failure */
13304 if (ender < 0x100) {
13306 if (RExC_recode_x_to_native) {
13307 ender = LATIN1_TO_NATIVE(ender);
13312 REQUIRE_UTF8(flagp);
13318 ender = grok_bslash_c(*p++, PASS2);
13320 case '8': case '9': /* must be a backreference */
13322 /* we have an escape like \8 which cannot be an octal escape
13323 * so we exit the loop, and let the outer loop handle this
13324 * escape which may or may not be a legitimate backref. */
13326 case '1': case '2': case '3':case '4':
13327 case '5': case '6': case '7':
13328 /* When we parse backslash escapes there is ambiguity
13329 * between backreferences and octal escapes. Any escape
13330 * from \1 - \9 is a backreference, any multi-digit
13331 * escape which does not start with 0 and which when
13332 * evaluated as decimal could refer to an already
13333 * parsed capture buffer is a back reference. Anything
13336 * Note this implies that \118 could be interpreted as
13337 * 118 OR as "\11" . "8" depending on whether there
13338 * were 118 capture buffers defined already in the
13341 /* NOTE, RExC_npar is 1 more than the actual number of
13342 * parens we have seen so far, hence the < RExC_npar below. */
13344 if ( !isDIGIT(p[1]) || S_backref_value(p) < RExC_npar)
13345 { /* Not to be treated as an octal constant, go
13353 I32 flags = PERL_SCAN_SILENT_ILLDIGIT;
13355 ender = grok_oct(p, &numlen, &flags, NULL);
13356 if (ender > 0xff) {
13357 REQUIRE_UTF8(flagp);
13360 if (PASS2 /* like \08, \178 */
13362 && isDIGIT(*p) && ckWARN(WARN_REGEXP))
13364 reg_warn_non_literal_string(
13366 form_short_octal_warning(p, numlen));
13372 FAIL("Trailing \\");
13375 if (!SIZE_ONLY&& isALPHANUMERIC(*p)) {
13376 /* Include any left brace following the alpha to emphasize
13377 * that it could be part of an escape at some point
13379 int len = (isALPHA(*p) && *(p + 1) == '{') ? 2 : 1;
13380 ckWARN3reg(p + len, "Unrecognized escape \\%.*s passed through", len, p);
13382 goto normal_default;
13383 } /* End of switch on '\' */
13386 /* Currently we don't care if the lbrace is at the start
13387 * of a construct. This catches it in the middle of a
13388 * literal string, or when it's the first thing after
13389 * something like "\b" */
13390 if (len || (p > RExC_start && isALPHA_A(*(p -1)))) {
13391 RExC_parse = p + 1;
13392 vFAIL("Unescaped left brace in regex is illegal here");
13394 goto normal_default;
13397 if (PASS2 && p > RExC_parse && RExC_strict) {
13398 ckWARN2reg(p + 1, "Unescaped literal '%c'", *p);
13401 default: /* A literal character */
13403 if (! UTF8_IS_INVARIANT(*p) && UTF) {
13405 ender = utf8n_to_uvchr((U8*)p, RExC_end - p,
13406 &numlen, UTF8_ALLOW_DEFAULT);
13412 } /* End of switch on the literal */
13414 /* Here, have looked at the literal character and <ender>
13415 * contains its ordinal, <p> points to the character after it.
13416 * We need to check if the next non-ignored thing is a
13417 * quantifier. Move <p> to after anything that should be
13418 * ignored, which, as a side effect, positions <p> for the next
13419 * loop iteration */
13420 skip_to_be_ignored_text(pRExC_state, &p,
13421 FALSE /* Don't force to /x */ );
13423 /* If the next thing is a quantifier, it applies to this
13424 * character only, which means that this character has to be in
13425 * its own node and can't just be appended to the string in an
13426 * existing node, so if there are already other characters in
13427 * the node, close the node with just them, and set up to do
13428 * this character again next time through, when it will be the
13429 * only thing in its new node */
13431 if ((next_is_quantifier = ( LIKELY(p < RExC_end)
13432 && UNLIKELY(ISMULT2(p))))
13439 /* Ready to add 'ender' to the node */
13441 if (! FOLD) { /* The simple case, just append the literal */
13443 /* In the sizing pass, we need only the size of the
13444 * character we are appending, hence we can delay getting
13445 * its representation until PASS2. */
13448 const STRLEN unilen = UVCHR_SKIP(ender);
13451 /* We have to subtract 1 just below (and again in
13452 * the corresponding PASS2 code) because the loop
13453 * increments <len> each time, as all but this path
13454 * (and one other) through it add a single byte to
13455 * the EXACTish node. But these paths would change
13456 * len to be the correct final value, so cancel out
13457 * the increment that follows */
13463 } else { /* PASS2 */
13466 U8 * new_s = uvchr_to_utf8((U8*)s, ender);
13467 len += (char *) new_s - s - 1;
13468 s = (char *) new_s;
13471 *(s++) = (char) ender;
13475 else if (LOC && is_PROBLEMATIC_LOCALE_FOLD_cp(ender)) {
13477 /* Here are folding under /l, and the code point is
13478 * problematic. First, we know we can't simplify things */
13479 maybe_exact = FALSE;
13480 maybe_exactfu = FALSE;
13482 /* A problematic code point in this context means that its
13483 * fold isn't known until runtime, so we can't fold it now.
13484 * (The non-problematic code points are the above-Latin1
13485 * ones that fold to also all above-Latin1. Their folds
13486 * don't vary no matter what the locale is.) But here we
13487 * have characters whose fold depends on the locale.
13488 * Unlike the non-folding case above, we have to keep track
13489 * of these in the sizing pass, so that we can make sure we
13490 * don't split too-long nodes in the middle of a potential
13491 * multi-char fold. And unlike the regular fold case
13492 * handled in the else clauses below, we don't actually
13493 * fold and don't have special cases to consider. What we
13494 * do for both passes is the PASS2 code for non-folding */
13495 goto not_fold_common;
13497 else /* A regular FOLD code point */
13499 #if UNICODE_MAJOR_VERSION > 3 /* no multifolds in early Unicode */ \
13500 || (UNICODE_MAJOR_VERSION == 3 && ( UNICODE_DOT_VERSION > 0) \
13501 || UNICODE_DOT_DOT_VERSION > 0)
13502 /* See comments for join_exact() as to why we fold
13503 * this non-UTF at compile time */
13504 || ( node_type == EXACTFU
13505 && ender == LATIN_SMALL_LETTER_SHARP_S)
13508 /* Here, are folding and are not UTF-8 encoded; therefore
13509 * the character must be in the range 0-255, and is not /l
13510 * (Not /l because we already handled these under /l in
13511 * is_PROBLEMATIC_LOCALE_FOLD_cp) */
13512 if (IS_IN_SOME_FOLD_L1(ender)) {
13513 maybe_exact = FALSE;
13515 /* See if the character's fold differs between /d and
13516 * /u. This includes the multi-char fold SHARP S to
13518 if (UNLIKELY(ender == LATIN_SMALL_LETTER_SHARP_S)) {
13519 RExC_seen_unfolded_sharp_s = 1;
13520 maybe_exactfu = FALSE;
13522 else if (maybe_exactfu
13523 && (PL_fold[ender] != PL_fold_latin1[ender]
13524 #if UNICODE_MAJOR_VERSION > 3 /* no multifolds in early Unicode */ \
13525 || (UNICODE_MAJOR_VERSION == 3 && ( UNICODE_DOT_VERSION > 0) \
13526 || UNICODE_DOT_DOT_VERSION > 0)
13528 && isALPHA_FOLD_EQ(ender, 's')
13529 && isALPHA_FOLD_EQ(*(s-1), 's'))
13532 maybe_exactfu = FALSE;
13536 /* Even when folding, we store just the input character, as
13537 * we have an array that finds its fold quickly */
13538 *(s++) = (char) ender;
13540 else { /* FOLD, and UTF (or sharp s) */
13541 /* Unlike the non-fold case, we do actually have to
13542 * calculate the results here in pass 1. This is for two
13543 * reasons, the folded length may be longer than the
13544 * unfolded, and we have to calculate how many EXACTish
13545 * nodes it will take; and we may run out of room in a node
13546 * in the middle of a potential multi-char fold, and have
13547 * to back off accordingly. */
13550 if (isASCII_uni(ender)) {
13551 folded = toFOLD(ender);
13552 *(s)++ = (U8) folded;
13557 folded = _to_uni_fold_flags(
13561 FOLD_FLAGS_FULL | ((ASCII_FOLD_RESTRICTED)
13562 ? FOLD_FLAGS_NOMIX_ASCII
13566 /* The loop increments <len> each time, as all but this
13567 * path (and one other) through it add a single byte to
13568 * the EXACTish node. But this one has changed len to
13569 * be the correct final value, so subtract one to
13570 * cancel out the increment that follows */
13571 len += foldlen - 1;
13573 /* If this node only contains non-folding code points so
13574 * far, see if this new one is also non-folding */
13576 if (folded != ender) {
13577 maybe_exact = FALSE;
13580 /* Here the fold is the original; we have to check
13581 * further to see if anything folds to it */
13582 if (_invlist_contains_cp(PL_utf8_foldable,
13585 maybe_exact = FALSE;
13592 if (next_is_quantifier) {
13594 /* Here, the next input is a quantifier, and to get here,
13595 * the current character is the only one in the node.
13596 * Also, here <len> doesn't include the final byte for this
13602 } /* End of loop through literal characters */
13604 /* Here we have either exhausted the input or ran out of room in
13605 * the node. (If we encountered a character that can't be in the
13606 * node, transfer is made directly to <loopdone>, and so we
13607 * wouldn't have fallen off the end of the loop.) In the latter
13608 * case, we artificially have to split the node into two, because
13609 * we just don't have enough space to hold everything. This
13610 * creates a problem if the final character participates in a
13611 * multi-character fold in the non-final position, as a match that
13612 * should have occurred won't, due to the way nodes are matched,
13613 * and our artificial boundary. So back off until we find a non-
13614 * problematic character -- one that isn't at the beginning or
13615 * middle of such a fold. (Either it doesn't participate in any
13616 * folds, or appears only in the final position of all the folds it
13617 * does participate in.) A better solution with far fewer false
13618 * positives, and that would fill the nodes more completely, would
13619 * be to actually have available all the multi-character folds to
13620 * test against, and to back-off only far enough to be sure that
13621 * this node isn't ending with a partial one. <upper_parse> is set
13622 * further below (if we need to reparse the node) to include just
13623 * up through that final non-problematic character that this code
13624 * identifies, so when it is set to less than the full node, we can
13625 * skip the rest of this */
13626 if (FOLD && p < RExC_end && upper_parse == MAX_NODE_STRING_SIZE) {
13628 const STRLEN full_len = len;
13630 assert(len >= MAX_NODE_STRING_SIZE);
13632 /* Here, <s> points to the final byte of the final character.
13633 * Look backwards through the string until find a non-
13634 * problematic character */
13638 /* This has no multi-char folds to non-UTF characters */
13639 if (ASCII_FOLD_RESTRICTED) {
13643 while (--s >= s0 && IS_NON_FINAL_FOLD(*s)) { }
13647 if (! PL_NonL1NonFinalFold) {
13648 PL_NonL1NonFinalFold = _new_invlist_C_array(
13649 NonL1_Perl_Non_Final_Folds_invlist);
13652 /* Point to the first byte of the final character */
13653 s = (char *) utf8_hop((U8 *) s, -1);
13655 while (s >= s0) { /* Search backwards until find
13656 non-problematic char */
13657 if (UTF8_IS_INVARIANT(*s)) {
13659 /* There are no ascii characters that participate
13660 * in multi-char folds under /aa. In EBCDIC, the
13661 * non-ascii invariants are all control characters,
13662 * so don't ever participate in any folds. */
13663 if (ASCII_FOLD_RESTRICTED
13664 || ! IS_NON_FINAL_FOLD(*s))
13669 else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
13670 if (! IS_NON_FINAL_FOLD(EIGHT_BIT_UTF8_TO_NATIVE(
13676 else if (! _invlist_contains_cp(
13677 PL_NonL1NonFinalFold,
13678 valid_utf8_to_uvchr((U8 *) s, NULL)))
13683 /* Here, the current character is problematic in that
13684 * it does occur in the non-final position of some
13685 * fold, so try the character before it, but have to
13686 * special case the very first byte in the string, so
13687 * we don't read outside the string */
13688 s = (s == s0) ? s -1 : (char *) utf8_hop((U8 *) s, -1);
13689 } /* End of loop backwards through the string */
13691 /* If there were only problematic characters in the string,
13692 * <s> will point to before s0, in which case the length
13693 * should be 0, otherwise include the length of the
13694 * non-problematic character just found */
13695 len = (s < s0) ? 0 : s - s0 + UTF8SKIP(s);
13698 /* Here, have found the final character, if any, that is
13699 * non-problematic as far as ending the node without splitting
13700 * it across a potential multi-char fold. <len> contains the
13701 * number of bytes in the node up-to and including that
13702 * character, or is 0 if there is no such character, meaning
13703 * the whole node contains only problematic characters. In
13704 * this case, give up and just take the node as-is. We can't
13709 /* If the node ends in an 's' we make sure it stays EXACTF,
13710 * as if it turns into an EXACTFU, it could later get
13711 * joined with another 's' that would then wrongly match
13713 if (maybe_exactfu && isALPHA_FOLD_EQ(ender, 's'))
13715 maybe_exactfu = FALSE;
13719 /* Here, the node does contain some characters that aren't
13720 * problematic. If one such is the final character in the
13721 * node, we are done */
13722 if (len == full_len) {
13725 else if (len + ((UTF) ? UTF8SKIP(s) : 1) == full_len) {
13727 /* If the final character is problematic, but the
13728 * penultimate is not, back-off that last character to
13729 * later start a new node with it */
13734 /* Here, the final non-problematic character is earlier
13735 * in the input than the penultimate character. What we do
13736 * is reparse from the beginning, going up only as far as
13737 * this final ok one, thus guaranteeing that the node ends
13738 * in an acceptable character. The reason we reparse is
13739 * that we know how far in the character is, but we don't
13740 * know how to correlate its position with the input parse.
13741 * An alternate implementation would be to build that
13742 * correlation as we go along during the original parse,
13743 * but that would entail extra work for every node, whereas
13744 * this code gets executed only when the string is too
13745 * large for the node, and the final two characters are
13746 * problematic, an infrequent occurrence. Yet another
13747 * possible strategy would be to save the tail of the
13748 * string, and the next time regatom is called, initialize
13749 * with that. The problem with this is that unless you
13750 * back off one more character, you won't be guaranteed
13751 * regatom will get called again, unless regbranch,
13752 * regpiece ... are also changed. If you do back off that
13753 * extra character, so that there is input guaranteed to
13754 * force calling regatom, you can't handle the case where
13755 * just the first character in the node is acceptable. I
13756 * (khw) decided to try this method which doesn't have that
13757 * pitfall; if performance issues are found, we can do a
13758 * combination of the current approach plus that one */
13764 } /* End of verifying node ends with an appropriate char */
13766 loopdone: /* Jumped to when encounters something that shouldn't be
13769 /* I (khw) don't know if you can get here with zero length, but the
13770 * old code handled this situation by creating a zero-length EXACT
13771 * node. Might as well be NOTHING instead */
13777 /* If 'maybe_exact' is still set here, means there are no
13778 * code points in the node that participate in folds;
13779 * similarly for 'maybe_exactfu' and code points that match
13780 * differently depending on UTF8ness of the target string
13781 * (for /u), or depending on locale for /l */
13787 else if (maybe_exactfu) {
13793 alloc_maybe_populate_EXACT(pRExC_state, ret, flagp, len, ender,
13794 FALSE /* Don't look to see if could
13795 be turned into an EXACT
13796 node, as we have already
13801 RExC_parse = p - 1;
13802 Set_Node_Cur_Length(ret, parse_start);
13805 /* len is STRLEN which is unsigned, need to copy to signed */
13808 vFAIL("Internal disaster");
13811 } /* End of label 'defchar:' */
13813 } /* End of giant switch on input character */
13815 /* Position parse to next real character */
13816 skip_to_be_ignored_text(pRExC_state, &RExC_parse,
13817 FALSE /* Don't force to /x */ );
13818 if (PASS2 && *RExC_parse == '{' && OP(ret) != SBOL && ! regcurly(RExC_parse)) {
13819 ckWARNregdep(RExC_parse + 1, "Unescaped left brace in regex is deprecated here (and will be fatal in Perl 5.30), passed through");
13827 S_populate_ANYOF_from_invlist(pTHX_ regnode *node, SV** invlist_ptr)
13829 /* Uses the inversion list '*invlist_ptr' to populate the ANYOF 'node'. It
13830 * sets up the bitmap and any flags, removing those code points from the
13831 * inversion list, setting it to NULL should it become completely empty */
13833 PERL_ARGS_ASSERT_POPULATE_ANYOF_FROM_INVLIST;
13834 assert(PL_regkind[OP(node)] == ANYOF);
13836 ANYOF_BITMAP_ZERO(node);
13837 if (*invlist_ptr) {
13839 /* This gets set if we actually need to modify things */
13840 bool change_invlist = FALSE;
13844 /* Start looking through *invlist_ptr */
13845 invlist_iterinit(*invlist_ptr);
13846 while (invlist_iternext(*invlist_ptr, &start, &end)) {
13850 if (end == UV_MAX && start <= NUM_ANYOF_CODE_POINTS) {
13851 ANYOF_FLAGS(node) |= ANYOF_MATCHES_ALL_ABOVE_BITMAP;
13854 /* Quit if are above what we should change */
13855 if (start >= NUM_ANYOF_CODE_POINTS) {
13859 change_invlist = TRUE;
13861 /* Set all the bits in the range, up to the max that we are doing */
13862 high = (end < NUM_ANYOF_CODE_POINTS - 1)
13864 : NUM_ANYOF_CODE_POINTS - 1;
13865 for (i = start; i <= (int) high; i++) {
13866 if (! ANYOF_BITMAP_TEST(node, i)) {
13867 ANYOF_BITMAP_SET(node, i);
13871 invlist_iterfinish(*invlist_ptr);
13873 /* Done with loop; remove any code points that are in the bitmap from
13874 * *invlist_ptr; similarly for code points above the bitmap if we have
13875 * a flag to match all of them anyways */
13876 if (change_invlist) {
13877 _invlist_subtract(*invlist_ptr, PL_InBitmap, invlist_ptr);
13879 if (ANYOF_FLAGS(node) & ANYOF_MATCHES_ALL_ABOVE_BITMAP) {
13880 _invlist_intersection(*invlist_ptr, PL_InBitmap, invlist_ptr);
13883 /* If have completely emptied it, remove it completely */
13884 if (_invlist_len(*invlist_ptr) == 0) {
13885 SvREFCNT_dec_NN(*invlist_ptr);
13886 *invlist_ptr = NULL;
13891 /* Parse POSIX character classes: [[:foo:]], [[=foo=]], [[.foo.]].
13892 Character classes ([:foo:]) can also be negated ([:^foo:]).
13893 Returns a named class id (ANYOF_XXX) if successful, -1 otherwise.
13894 Equivalence classes ([=foo=]) and composites ([.foo.]) are parsed,
13895 but trigger failures because they are currently unimplemented. */
13897 #define POSIXCC_DONE(c) ((c) == ':')
13898 #define POSIXCC_NOTYET(c) ((c) == '=' || (c) == '.')
13899 #define POSIXCC(c) (POSIXCC_DONE(c) || POSIXCC_NOTYET(c))
13900 #define MAYBE_POSIXCC(c) (POSIXCC(c) || (c) == '^' || (c) == ';')
13902 #define WARNING_PREFIX "Assuming NOT a POSIX class since "
13903 #define NO_BLANKS_POSIX_WARNING "no blanks are allowed in one"
13904 #define SEMI_COLON_POSIX_WARNING "a semi-colon was found instead of a colon"
13906 #define NOT_MEANT_TO_BE_A_POSIX_CLASS (OOB_NAMEDCLASS - 1)
13908 /* 'posix_warnings' and 'warn_text' are names of variables in the following
13910 #define ADD_POSIX_WARNING(p, text) STMT_START { \
13911 if (posix_warnings) { \
13912 if (! RExC_warn_text ) RExC_warn_text = (AV *) sv_2mortal((SV *) newAV()); \
13913 av_push(RExC_warn_text, Perl_newSVpvf(aTHX_ \
13917 REPORT_LOCATION_ARGS(p))); \
13922 S_handle_possible_posix(pTHX_ RExC_state_t *pRExC_state,
13924 const char * const s, /* Where the putative posix class begins.
13925 Normally, this is one past the '['. This
13926 parameter exists so it can be somewhere
13927 besides RExC_parse. */
13928 char ** updated_parse_ptr, /* Where to set the updated parse pointer, or
13930 AV ** posix_warnings, /* Where to place any generated warnings, or
13932 const bool check_only /* Don't die if error */
13935 /* This parses what the caller thinks may be one of the three POSIX
13937 * 1) a character class, like [:blank:]
13938 * 2) a collating symbol, like [. .]
13939 * 3) an equivalence class, like [= =]
13940 * In the latter two cases, it croaks if it finds a syntactically legal
13941 * one, as these are not handled by Perl.
13943 * The main purpose is to look for a POSIX character class. It returns:
13944 * a) the class number
13945 * if it is a completely syntactically and semantically legal class.
13946 * 'updated_parse_ptr', if not NULL, is set to point to just after the
13947 * closing ']' of the class
13948 * b) OOB_NAMEDCLASS
13949 * if it appears that one of the three POSIX constructs was meant, but
13950 * its specification was somehow defective. 'updated_parse_ptr', if
13951 * not NULL, is set to point to the character just after the end
13952 * character of the class. See below for handling of warnings.
13953 * c) NOT_MEANT_TO_BE_A_POSIX_CLASS
13954 * if it doesn't appear that a POSIX construct was intended.
13955 * 'updated_parse_ptr' is not changed. No warnings nor errors are
13958 * In b) there may be errors or warnings generated. If 'check_only' is
13959 * TRUE, then any errors are discarded. Warnings are returned to the
13960 * caller via an AV* created into '*posix_warnings' if it is not NULL. If
13961 * instead it is NULL, warnings are suppressed. This is done in all
13962 * passes. The reason for this is that the rest of the parsing is heavily
13963 * dependent on whether this routine found a valid posix class or not. If
13964 * it did, the closing ']' is absorbed as part of the class. If no class,
13965 * or an invalid one is found, any ']' will be considered the terminator of
13966 * the outer bracketed character class, leading to very different results.
13967 * In particular, a '(?[ ])' construct will likely have a syntax error if
13968 * the class is parsed other than intended, and this will happen in pass1,
13969 * before the warnings would normally be output. This mechanism allows the
13970 * caller to output those warnings in pass1 just before dieing, giving a
13971 * much better clue as to what is wrong.
13973 * The reason for this function, and its complexity is that a bracketed
13974 * character class can contain just about anything. But it's easy to
13975 * mistype the very specific posix class syntax but yielding a valid
13976 * regular bracketed class, so it silently gets compiled into something
13977 * quite unintended.
13979 * The solution adopted here maintains backward compatibility except that
13980 * it adds a warning if it looks like a posix class was intended but
13981 * improperly specified. The warning is not raised unless what is input
13982 * very closely resembles one of the 14 legal posix classes. To do this,
13983 * it uses fuzzy parsing. It calculates how many single-character edits it
13984 * would take to transform what was input into a legal posix class. Only
13985 * if that number is quite small does it think that the intention was a
13986 * posix class. Obviously these are heuristics, and there will be cases
13987 * where it errs on one side or another, and they can be tweaked as
13988 * experience informs.
13990 * The syntax for a legal posix class is:
13992 * qr/(?xa: \[ : \^? [:lower:]{4,6} : \] )/
13994 * What this routine considers syntactically to be an intended posix class
13995 * is this (the comments indicate some restrictions that the pattern
13998 * qr/(?x: \[? # The left bracket, possibly
14000 * \h* # possibly followed by blanks
14001 * (?: \^ \h* )? # possibly a misplaced caret
14002 * [:;]? # The opening class character,
14003 * # possibly omitted. A typo
14004 * # semi-colon can also be used.
14006 * \^? # possibly a correctly placed
14007 * # caret, but not if there was also
14008 * # a misplaced one
14010 * .{3,15} # The class name. If there are
14011 * # deviations from the legal syntax,
14012 * # its edit distance must be close
14013 * # to a real class name in order
14014 * # for it to be considered to be
14015 * # an intended posix class.
14017 * [:punct:]? # The closing class character,
14018 * # possibly omitted. If not a colon
14019 * # nor semi colon, the class name
14020 * # must be even closer to a valid
14023 * \]? # The right bracket, possibly
14027 * In the above, \h must be ASCII-only.
14029 * These are heuristics, and can be tweaked as field experience dictates.
14030 * There will be cases when someone didn't intend to specify a posix class
14031 * that this warns as being so. The goal is to minimize these, while
14032 * maximizing the catching of things intended to be a posix class that
14033 * aren't parsed as such.
14037 const char * const e = RExC_end;
14038 unsigned complement = 0; /* If to complement the class */
14039 bool found_problem = FALSE; /* Assume OK until proven otherwise */
14040 bool has_opening_bracket = FALSE;
14041 bool has_opening_colon = FALSE;
14042 int class_number = OOB_NAMEDCLASS; /* Out-of-bounds until find
14044 const char * possible_end = NULL; /* used for a 2nd parse pass */
14045 const char* name_start; /* ptr to class name first char */
14047 /* If the number of single-character typos the input name is away from a
14048 * legal name is no more than this number, it is considered to have meant
14049 * the legal name */
14050 int max_distance = 2;
14052 /* to store the name. The size determines the maximum length before we
14053 * decide that no posix class was intended. Should be at least
14054 * sizeof("alphanumeric") */
14057 PERL_ARGS_ASSERT_HANDLE_POSSIBLE_POSIX;
14059 if (posix_warnings && RExC_warn_text)
14060 av_clear(RExC_warn_text);
14063 return NOT_MEANT_TO_BE_A_POSIX_CLASS;
14066 if (*(p - 1) != '[') {
14067 ADD_POSIX_WARNING(p, "it doesn't start with a '['");
14068 found_problem = TRUE;
14071 has_opening_bracket = TRUE;
14074 /* They could be confused and think you can put spaces between the
14077 found_problem = TRUE;
14081 } while (p < e && isBLANK(*p));
14083 ADD_POSIX_WARNING(p, NO_BLANKS_POSIX_WARNING);
14086 /* For [. .] and [= =]. These are quite different internally from [: :],
14087 * so they are handled separately. */
14088 if (POSIXCC_NOTYET(*p) && p < e - 3) /* 1 for the close, and 1 for the ']'
14089 and 1 for at least one char in it
14092 const char open_char = *p;
14093 const char * temp_ptr = p + 1;
14095 /* These two constructs are not handled by perl, and if we find a
14096 * syntactically valid one, we croak. khw, who wrote this code, finds
14097 * this explanation of them very unclear:
14098 * http://pubs.opengroup.org/onlinepubs/009696899/basedefs/xbd_chap09.html
14099 * And searching the rest of the internet wasn't very helpful either.
14100 * It looks like just about any byte can be in these constructs,
14101 * depending on the locale. But unless the pattern is being compiled
14102 * under /l, which is very rare, Perl runs under the C or POSIX locale.
14103 * In that case, it looks like [= =] isn't allowed at all, and that
14104 * [. .] could be any single code point, but for longer strings the
14105 * constituent characters would have to be the ASCII alphabetics plus
14106 * the minus-hyphen. Any sensible locale definition would limit itself
14107 * to these. And any portable one definitely should. Trying to parse
14108 * the general case is a nightmare (see [perl #127604]). So, this code
14109 * looks only for interiors of these constructs that match:
14111 * Using \w relaxes the apparent rules a little, without adding much
14112 * danger of mistaking something else for one of these constructs.
14114 * [. .] in some implementations described on the internet is usable to
14115 * escape a character that otherwise is special in bracketed character
14116 * classes. For example [.].] means a literal right bracket instead of
14117 * the ending of the class
14119 * [= =] can legitimately contain a [. .] construct, but we don't
14120 * handle this case, as that [. .] construct will later get parsed
14121 * itself and croak then. And [= =] is checked for even when not under
14122 * /l, as Perl has long done so.
14124 * The code below relies on there being a trailing NUL, so it doesn't
14125 * have to keep checking if the parse ptr < e.
14127 if (temp_ptr[1] == open_char) {
14130 else while ( temp_ptr < e
14131 && (isWORDCHAR(*temp_ptr) || *temp_ptr == '-'))
14136 if (*temp_ptr == open_char) {
14138 if (*temp_ptr == ']') {
14140 if (! found_problem && ! check_only) {
14141 RExC_parse = (char *) temp_ptr;
14142 vFAIL3("POSIX syntax [%c %c] is reserved for future "
14143 "extensions", open_char, open_char);
14146 /* Here, the syntax wasn't completely valid, or else the call
14147 * is to check-only */
14148 if (updated_parse_ptr) {
14149 *updated_parse_ptr = (char *) temp_ptr;
14152 return OOB_NAMEDCLASS;
14156 /* If we find something that started out to look like one of these
14157 * constructs, but isn't, we continue below so that it can be checked
14158 * for being a class name with a typo of '.' or '=' instead of a colon.
14162 /* Here, we think there is a possibility that a [: :] class was meant, and
14163 * we have the first real character. It could be they think the '^' comes
14166 found_problem = TRUE;
14167 ADD_POSIX_WARNING(p + 1, "the '^' must come after the colon");
14172 found_problem = TRUE;
14176 } while (p < e && isBLANK(*p));
14178 ADD_POSIX_WARNING(p, NO_BLANKS_POSIX_WARNING);
14182 /* But the first character should be a colon, which they could have easily
14183 * mistyped on a qwerty keyboard as a semi-colon (and which may be hard to
14184 * distinguish from a colon, so treat that as a colon). */
14187 has_opening_colon = TRUE;
14189 else if (*p == ';') {
14190 found_problem = TRUE;
14192 ADD_POSIX_WARNING(p, SEMI_COLON_POSIX_WARNING);
14193 has_opening_colon = TRUE;
14196 found_problem = TRUE;
14197 ADD_POSIX_WARNING(p, "there must be a starting ':'");
14199 /* Consider an initial punctuation (not one of the recognized ones) to
14200 * be a left terminator */
14201 if (*p != '^' && *p != ']' && isPUNCT(*p)) {
14206 /* They may think that you can put spaces between the components */
14208 found_problem = TRUE;
14212 } while (p < e && isBLANK(*p));
14214 ADD_POSIX_WARNING(p, NO_BLANKS_POSIX_WARNING);
14219 /* We consider something like [^:^alnum:]] to not have been intended to
14220 * be a posix class, but XXX maybe we should */
14222 return NOT_MEANT_TO_BE_A_POSIX_CLASS;
14229 /* Again, they may think that you can put spaces between the components */
14231 found_problem = TRUE;
14235 } while (p < e && isBLANK(*p));
14237 ADD_POSIX_WARNING(p, NO_BLANKS_POSIX_WARNING);
14242 /* XXX This ']' may be a typo, and something else was meant. But
14243 * treating it as such creates enough complications, that that
14244 * possibility isn't currently considered here. So we assume that the
14245 * ']' is what is intended, and if we've already found an initial '[',
14246 * this leaves this construct looking like [:] or [:^], which almost
14247 * certainly weren't intended to be posix classes */
14248 if (has_opening_bracket) {
14249 return NOT_MEANT_TO_BE_A_POSIX_CLASS;
14252 /* But this function can be called when we parse the colon for
14253 * something like qr/[alpha:]]/, so we back up to look for the
14258 found_problem = TRUE;
14259 ADD_POSIX_WARNING(p, SEMI_COLON_POSIX_WARNING);
14261 else if (*p != ':') {
14263 /* XXX We are currently very restrictive here, so this code doesn't
14264 * consider the possibility that, say, /[alpha.]]/ was intended to
14265 * be a posix class. */
14266 return NOT_MEANT_TO_BE_A_POSIX_CLASS;
14269 /* Here we have something like 'foo:]'. There was no initial colon,
14270 * and we back up over 'foo. XXX Unlike the going forward case, we
14271 * don't handle typos of non-word chars in the middle */
14272 has_opening_colon = FALSE;
14275 while (p > RExC_start && isWORDCHAR(*p)) {
14280 /* Here, we have positioned ourselves to where we think the first
14281 * character in the potential class is */
14284 /* Now the interior really starts. There are certain key characters that
14285 * can end the interior, or these could just be typos. To catch both
14286 * cases, we may have to do two passes. In the first pass, we keep on
14287 * going unless we come to a sequence that matches
14288 * qr/ [[:punct:]] [[:blank:]]* \] /xa
14289 * This means it takes a sequence to end the pass, so two typos in a row if
14290 * that wasn't what was intended. If the class is perfectly formed, just
14291 * this one pass is needed. We also stop if there are too many characters
14292 * being accumulated, but this number is deliberately set higher than any
14293 * real class. It is set high enough so that someone who thinks that
14294 * 'alphanumeric' is a correct name would get warned that it wasn't.
14295 * While doing the pass, we keep track of where the key characters were in
14296 * it. If we don't find an end to the class, and one of the key characters
14297 * was found, we redo the pass, but stop when we get to that character.
14298 * Thus the key character was considered a typo in the first pass, but a
14299 * terminator in the second. If two key characters are found, we stop at
14300 * the second one in the first pass. Again this can miss two typos, but
14301 * catches a single one
14303 * In the first pass, 'possible_end' starts as NULL, and then gets set to
14304 * point to the first key character. For the second pass, it starts as -1.
14310 bool has_blank = FALSE;
14311 bool has_upper = FALSE;
14312 bool has_terminating_colon = FALSE;
14313 bool has_terminating_bracket = FALSE;
14314 bool has_semi_colon = FALSE;
14315 unsigned int name_len = 0;
14316 int punct_count = 0;
14320 /* Squeeze out blanks when looking up the class name below */
14321 if (isBLANK(*p) ) {
14323 found_problem = TRUE;
14328 /* The name will end with a punctuation */
14330 const char * peek = p + 1;
14332 /* Treat any non-']' punctuation followed by a ']' (possibly
14333 * with intervening blanks) as trying to terminate the class.
14334 * ']]' is very likely to mean a class was intended (but
14335 * missing the colon), but the warning message that gets
14336 * generated shows the error position better if we exit the
14337 * loop at the bottom (eventually), so skip it here. */
14339 if (peek < e && isBLANK(*peek)) {
14341 found_problem = TRUE;
14344 } while (peek < e && isBLANK(*peek));
14347 if (peek < e && *peek == ']') {
14348 has_terminating_bracket = TRUE;
14350 has_terminating_colon = TRUE;
14352 else if (*p == ';') {
14353 has_semi_colon = TRUE;
14354 has_terminating_colon = TRUE;
14357 found_problem = TRUE;
14364 /* Here we have punctuation we thought didn't end the class.
14365 * Keep track of the position of the key characters that are
14366 * more likely to have been class-enders */
14367 if (*p == ']' || *p == '[' || *p == ':' || *p == ';') {
14369 /* Allow just one such possible class-ender not actually
14370 * ending the class. */
14371 if (possible_end) {
14377 /* If we have too many punctuation characters, no use in
14379 if (++punct_count > max_distance) {
14383 /* Treat the punctuation as a typo. */
14384 input_text[name_len++] = *p;
14387 else if (isUPPER(*p)) { /* Use lowercase for lookup */
14388 input_text[name_len++] = toLOWER(*p);
14390 found_problem = TRUE;
14392 } else if (! UTF || UTF8_IS_INVARIANT(*p)) {
14393 input_text[name_len++] = *p;
14397 input_text[name_len++] = utf8_to_uvchr_buf((U8 *) p, e, NULL);
14401 /* The declaration of 'input_text' is how long we allow a potential
14402 * class name to be, before saying they didn't mean a class name at
14404 if (name_len >= C_ARRAY_LENGTH(input_text)) {
14409 /* We get to here when the possible class name hasn't been properly
14410 * terminated before:
14411 * 1) we ran off the end of the pattern; or
14412 * 2) found two characters, each of which might have been intended to
14413 * be the name's terminator
14414 * 3) found so many punctuation characters in the purported name,
14415 * that the edit distance to a valid one is exceeded
14416 * 4) we decided it was more characters than anyone could have
14417 * intended to be one. */
14419 found_problem = TRUE;
14421 /* In the final two cases, we know that looking up what we've
14422 * accumulated won't lead to a match, even a fuzzy one. */
14423 if ( name_len >= C_ARRAY_LENGTH(input_text)
14424 || punct_count > max_distance)
14426 /* If there was an intermediate key character that could have been
14427 * an intended end, redo the parse, but stop there */
14428 if (possible_end && possible_end != (char *) -1) {
14429 possible_end = (char *) -1; /* Special signal value to say
14430 we've done a first pass */
14435 /* Otherwise, it can't have meant to have been a class */
14436 return NOT_MEANT_TO_BE_A_POSIX_CLASS;
14439 /* If we ran off the end, and the final character was a punctuation
14440 * one, back up one, to look at that final one just below. Later, we
14441 * will restore the parse pointer if appropriate */
14442 if (name_len && p == e && isPUNCT(*(p-1))) {
14447 if (p < e && isPUNCT(*p)) {
14449 has_terminating_bracket = TRUE;
14451 /* If this is a 2nd ']', and the first one is just below this
14452 * one, consider that to be the real terminator. This gives a
14453 * uniform and better positioning for the warning message */
14455 && possible_end != (char *) -1
14456 && *possible_end == ']'
14457 && name_len && input_text[name_len - 1] == ']')
14462 /* And this is actually equivalent to having done the 2nd
14463 * pass now, so set it to not try again */
14464 possible_end = (char *) -1;
14469 has_terminating_colon = TRUE;
14471 else if (*p == ';') {
14472 has_semi_colon = TRUE;
14473 has_terminating_colon = TRUE;
14481 /* Here, we have a class name to look up. We can short circuit the
14482 * stuff below for short names that can't possibly be meant to be a
14483 * class name. (We can do this on the first pass, as any second pass
14484 * will yield an even shorter name) */
14485 if (name_len < 3) {
14486 return NOT_MEANT_TO_BE_A_POSIX_CLASS;
14489 /* Find which class it is. Initially switch on the length of the name.
14491 switch (name_len) {
14493 if (memEQ(name_start, "word", 4)) {
14494 /* this is not POSIX, this is the Perl \w */
14495 class_number = ANYOF_WORDCHAR;
14499 /* Names all of length 5: alnum alpha ascii blank cntrl digit
14500 * graph lower print punct space upper
14501 * Offset 4 gives the best switch position. */
14502 switch (name_start[4]) {
14504 if (memEQ(name_start, "alph", 4)) /* alpha */
14505 class_number = ANYOF_ALPHA;
14508 if (memEQ(name_start, "spac", 4)) /* space */
14509 class_number = ANYOF_SPACE;
14512 if (memEQ(name_start, "grap", 4)) /* graph */
14513 class_number = ANYOF_GRAPH;
14516 if (memEQ(name_start, "asci", 4)) /* ascii */
14517 class_number = ANYOF_ASCII;
14520 if (memEQ(name_start, "blan", 4)) /* blank */
14521 class_number = ANYOF_BLANK;
14524 if (memEQ(name_start, "cntr", 4)) /* cntrl */
14525 class_number = ANYOF_CNTRL;
14528 if (memEQ(name_start, "alnu", 4)) /* alnum */
14529 class_number = ANYOF_ALPHANUMERIC;
14532 if (memEQ(name_start, "lowe", 4)) /* lower */
14533 class_number = (FOLD) ? ANYOF_CASED : ANYOF_LOWER;
14534 else if (memEQ(name_start, "uppe", 4)) /* upper */
14535 class_number = (FOLD) ? ANYOF_CASED : ANYOF_UPPER;
14538 if (memEQ(name_start, "digi", 4)) /* digit */
14539 class_number = ANYOF_DIGIT;
14540 else if (memEQ(name_start, "prin", 4)) /* print */
14541 class_number = ANYOF_PRINT;
14542 else if (memEQ(name_start, "punc", 4)) /* punct */
14543 class_number = ANYOF_PUNCT;
14548 if (memEQ(name_start, "xdigit", 6))
14549 class_number = ANYOF_XDIGIT;
14553 /* If the name exactly matches a posix class name the class number will
14554 * here be set to it, and the input almost certainly was meant to be a
14555 * posix class, so we can skip further checking. If instead the syntax
14556 * is exactly correct, but the name isn't one of the legal ones, we
14557 * will return that as an error below. But if neither of these apply,
14558 * it could be that no posix class was intended at all, or that one
14559 * was, but there was a typo. We tease these apart by doing fuzzy
14560 * matching on the name */
14561 if (class_number == OOB_NAMEDCLASS && found_problem) {
14562 const UV posix_names[][6] = {
14563 { 'a', 'l', 'n', 'u', 'm' },
14564 { 'a', 'l', 'p', 'h', 'a' },
14565 { 'a', 's', 'c', 'i', 'i' },
14566 { 'b', 'l', 'a', 'n', 'k' },
14567 { 'c', 'n', 't', 'r', 'l' },
14568 { 'd', 'i', 'g', 'i', 't' },
14569 { 'g', 'r', 'a', 'p', 'h' },
14570 { 'l', 'o', 'w', 'e', 'r' },
14571 { 'p', 'r', 'i', 'n', 't' },
14572 { 'p', 'u', 'n', 'c', 't' },
14573 { 's', 'p', 'a', 'c', 'e' },
14574 { 'u', 'p', 'p', 'e', 'r' },
14575 { 'w', 'o', 'r', 'd' },
14576 { 'x', 'd', 'i', 'g', 'i', 't' }
14578 /* The names of the above all have added NULs to make them the same
14579 * size, so we need to also have the real lengths */
14580 const UV posix_name_lengths[] = {
14581 sizeof("alnum") - 1,
14582 sizeof("alpha") - 1,
14583 sizeof("ascii") - 1,
14584 sizeof("blank") - 1,
14585 sizeof("cntrl") - 1,
14586 sizeof("digit") - 1,
14587 sizeof("graph") - 1,
14588 sizeof("lower") - 1,
14589 sizeof("print") - 1,
14590 sizeof("punct") - 1,
14591 sizeof("space") - 1,
14592 sizeof("upper") - 1,
14593 sizeof("word") - 1,
14594 sizeof("xdigit")- 1
14597 int temp_max = max_distance; /* Use a temporary, so if we
14598 reparse, we haven't changed the
14601 /* Use a smaller max edit distance if we are missing one of the
14603 if ( has_opening_bracket + has_opening_colon < 2
14604 || has_terminating_bracket + has_terminating_colon < 2)
14609 /* See if the input name is close to a legal one */
14610 for (i = 0; i < C_ARRAY_LENGTH(posix_names); i++) {
14612 /* Short circuit call if the lengths are too far apart to be
14614 if (abs( (int) (name_len - posix_name_lengths[i]))
14620 if (edit_distance(input_text,
14623 posix_name_lengths[i],
14627 { /* If it is close, it probably was intended to be a class */
14628 goto probably_meant_to_be;
14632 /* Here the input name is not close enough to a valid class name
14633 * for us to consider it to be intended to be a posix class. If
14634 * we haven't already done so, and the parse found a character that
14635 * could have been terminators for the name, but which we absorbed
14636 * as typos during the first pass, repeat the parse, signalling it
14637 * to stop at that character */
14638 if (possible_end && possible_end != (char *) -1) {
14639 possible_end = (char *) -1;
14644 /* Here neither pass found a close-enough class name */
14645 return NOT_MEANT_TO_BE_A_POSIX_CLASS;
14648 probably_meant_to_be:
14650 /* Here we think that a posix specification was intended. Update any
14652 if (updated_parse_ptr) {
14653 *updated_parse_ptr = (char *) p;
14656 /* If a posix class name was intended but incorrectly specified, we
14657 * output or return the warnings */
14658 if (found_problem) {
14660 /* We set flags for these issues in the parse loop above instead of
14661 * adding them to the list of warnings, because we can parse it
14662 * twice, and we only want one warning instance */
14664 ADD_POSIX_WARNING(p, "the name must be all lowercase letters");
14667 ADD_POSIX_WARNING(p, NO_BLANKS_POSIX_WARNING);
14669 if (has_semi_colon) {
14670 ADD_POSIX_WARNING(p, SEMI_COLON_POSIX_WARNING);
14672 else if (! has_terminating_colon) {
14673 ADD_POSIX_WARNING(p, "there is no terminating ':'");
14675 if (! has_terminating_bracket) {
14676 ADD_POSIX_WARNING(p, "there is no terminating ']'");
14679 if (posix_warnings && RExC_warn_text && av_top_index(RExC_warn_text) > -1) {
14680 *posix_warnings = RExC_warn_text;
14683 else if (class_number != OOB_NAMEDCLASS) {
14684 /* If it is a known class, return the class. The class number
14685 * #defines are structured so each complement is +1 to the normal
14687 return class_number + complement;
14689 else if (! check_only) {
14691 /* Here, it is an unrecognized class. This is an error (unless the
14692 * call is to check only, which we've already handled above) */
14693 const char * const complement_string = (complement)
14696 RExC_parse = (char *) p;
14697 vFAIL3utf8f("POSIX class [:%s%" UTF8f ":] unknown",
14699 UTF8fARG(UTF, RExC_parse - name_start - 2, name_start));
14703 return OOB_NAMEDCLASS;
14705 #undef ADD_POSIX_WARNING
14707 STATIC unsigned int
14708 S_regex_set_precedence(const U8 my_operator) {
14710 /* Returns the precedence in the (?[...]) construct of the input operator,
14711 * specified by its character representation. The precedence follows
14712 * general Perl rules, but it extends this so that ')' and ']' have (low)
14713 * precedence even though they aren't really operators */
14715 switch (my_operator) {
14731 NOT_REACHED; /* NOTREACHED */
14732 return 0; /* Silence compiler warning */
14736 S_handle_regex_sets(pTHX_ RExC_state_t *pRExC_state, SV** return_invlist,
14737 I32 *flagp, U32 depth,
14738 char * const oregcomp_parse)
14740 /* Handle the (?[...]) construct to do set operations */
14742 U8 curchar; /* Current character being parsed */
14743 UV start, end; /* End points of code point ranges */
14744 SV* final = NULL; /* The end result inversion list */
14745 SV* result_string; /* 'final' stringified */
14746 AV* stack; /* stack of operators and operands not yet
14748 AV* fence_stack = NULL; /* A stack containing the positions in
14749 'stack' of where the undealt-with left
14750 parens would be if they were actually
14752 /* The 'VOL' (expanding to 'volatile') is a workaround for an optimiser bug
14753 * in Solaris Studio 12.3. See RT #127455 */
14754 VOL IV fence = 0; /* Position of where most recent undealt-
14755 with left paren in stack is; -1 if none.
14757 STRLEN len; /* Temporary */
14758 regnode* node; /* Temporary, and final regnode returned by
14760 const bool save_fold = FOLD; /* Temporary */
14761 char *save_end, *save_parse; /* Temporaries */
14762 const bool in_locale = LOC; /* we turn off /l during processing */
14763 AV* posix_warnings = NULL;
14765 GET_RE_DEBUG_FLAGS_DECL;
14767 PERL_ARGS_ASSERT_HANDLE_REGEX_SETS;
14770 set_regex_charset(&RExC_flags, REGEX_UNICODE_CHARSET);
14773 REQUIRE_UNI_RULES(flagp, NULL); /* The use of this operator implies /u.
14774 This is required so that the compile
14775 time values are valid in all runtime
14778 /* This will return only an ANYOF regnode, or (unlikely) something smaller
14779 * (such as EXACT). Thus we can skip most everything if just sizing. We
14780 * call regclass to handle '[]' so as to not have to reinvent its parsing
14781 * rules here (throwing away the size it computes each time). And, we exit
14782 * upon an unescaped ']' that isn't one ending a regclass. To do both
14783 * these things, we need to realize that something preceded by a backslash
14784 * is escaped, so we have to keep track of backslashes */
14786 UV depth = 0; /* how many nested (?[...]) constructs */
14788 while (RExC_parse < RExC_end) {
14789 SV* current = NULL;
14791 skip_to_be_ignored_text(pRExC_state, &RExC_parse,
14792 TRUE /* Force /x */ );
14794 switch (*RExC_parse) {
14796 if (RExC_parse[1] == '[') depth++, RExC_parse++;
14801 /* Skip past this, so the next character gets skipped, after
14804 if (*RExC_parse == 'c') {
14805 /* Skip the \cX notation for control characters */
14806 RExC_parse += UTF ? UTF8SKIP(RExC_parse) : 1;
14812 /* See if this is a [:posix:] class. */
14813 bool is_posix_class = (OOB_NAMEDCLASS
14814 < handle_possible_posix(pRExC_state,
14818 TRUE /* checking only */));
14819 /* If it is a posix class, leave the parse pointer at the
14820 * '[' to fool regclass() into thinking it is part of a
14821 * '[[:posix:]]'. */
14822 if (! is_posix_class) {
14826 /* regclass() can only return RESTART_PASS1 and NEED_UTF8
14827 * if multi-char folds are allowed. */
14828 if (!regclass(pRExC_state, flagp,depth+1,
14829 is_posix_class, /* parse the whole char
14830 class only if not a
14832 FALSE, /* don't allow multi-char folds */
14833 TRUE, /* silence non-portable warnings. */
14835 FALSE, /* Require return to be an ANYOF */
14839 FAIL2("panic: regclass returned NULL to handle_sets, "
14840 "flags=%#" UVxf, (UV) *flagp);
14842 /* function call leaves parse pointing to the ']', except
14843 * if we faked it */
14844 if (is_posix_class) {
14848 SvREFCNT_dec(current); /* In case it returned something */
14853 if (depth--) break;
14855 if (*RExC_parse == ')') {
14856 node = reganode(pRExC_state, ANYOF, 0);
14857 RExC_size += ANYOF_SKIP;
14858 nextchar(pRExC_state);
14859 Set_Node_Length(node,
14860 RExC_parse - oregcomp_parse + 1); /* MJD */
14862 set_regex_charset(&RExC_flags, REGEX_LOCALE_CHARSET);
14870 RExC_parse += UTF ? UTF8SKIP(RExC_parse) : 1;
14874 /* We output the messages even if warnings are off, because we'll fail
14875 * the very next thing, and these give a likely diagnosis for that */
14876 if (posix_warnings && av_tindex_skip_len_mg(posix_warnings) >= 0) {
14877 output_or_return_posix_warnings(pRExC_state, posix_warnings, NULL);
14880 FAIL("Syntax error in (?[...])");
14883 /* Pass 2 only after this. */
14884 Perl_ck_warner_d(aTHX_
14885 packWARN(WARN_EXPERIMENTAL__REGEX_SETS),
14886 "The regex_sets feature is experimental" REPORT_LOCATION,
14887 REPORT_LOCATION_ARGS(RExC_parse));
14889 /* Everything in this construct is a metacharacter. Operands begin with
14890 * either a '\' (for an escape sequence), or a '[' for a bracketed
14891 * character class. Any other character should be an operator, or
14892 * parenthesis for grouping. Both types of operands are handled by calling
14893 * regclass() to parse them. It is called with a parameter to indicate to
14894 * return the computed inversion list. The parsing here is implemented via
14895 * a stack. Each entry on the stack is a single character representing one
14896 * of the operators; or else a pointer to an operand inversion list. */
14898 #define IS_OPERATOR(a) SvIOK(a)
14899 #define IS_OPERAND(a) (! IS_OPERATOR(a))
14901 /* The stack is kept in Łukasiewicz order. (That's pronounced similar
14902 * to luke-a-shave-itch (or -itz), but people who didn't want to bother
14903 * with pronouncing it called it Reverse Polish instead, but now that YOU
14904 * know how to pronounce it you can use the correct term, thus giving due
14905 * credit to the person who invented it, and impressing your geek friends.
14906 * Wikipedia says that the pronounciation of "Ł" has been changing so that
14907 * it is now more like an English initial W (as in wonk) than an L.)
14909 * This means that, for example, 'a | b & c' is stored on the stack as
14917 * where the numbers in brackets give the stack [array] element number.
14918 * In this implementation, parentheses are not stored on the stack.
14919 * Instead a '(' creates a "fence" so that the part of the stack below the
14920 * fence is invisible except to the corresponding ')' (this allows us to
14921 * replace testing for parens, by using instead subtraction of the fence
14922 * position). As new operands are processed they are pushed onto the stack
14923 * (except as noted in the next paragraph). New operators of higher
14924 * precedence than the current final one are inserted on the stack before
14925 * the lhs operand (so that when the rhs is pushed next, everything will be
14926 * in the correct positions shown above. When an operator of equal or
14927 * lower precedence is encountered in parsing, all the stacked operations
14928 * of equal or higher precedence are evaluated, leaving the result as the
14929 * top entry on the stack. This makes higher precedence operations
14930 * evaluate before lower precedence ones, and causes operations of equal
14931 * precedence to left associate.
14933 * The only unary operator '!' is immediately pushed onto the stack when
14934 * encountered. When an operand is encountered, if the top of the stack is
14935 * a '!", the complement is immediately performed, and the '!' popped. The
14936 * resulting value is treated as a new operand, and the logic in the
14937 * previous paragraph is executed. Thus in the expression
14939 * the stack looks like
14945 * as 'b' gets parsed, the latter gets evaluated to '!b', and the stack
14952 * A ')' is treated as an operator with lower precedence than all the
14953 * aforementioned ones, which causes all operations on the stack above the
14954 * corresponding '(' to be evaluated down to a single resultant operand.
14955 * Then the fence for the '(' is removed, and the operand goes through the
14956 * algorithm above, without the fence.
14958 * A separate stack is kept of the fence positions, so that the position of
14959 * the latest so-far unbalanced '(' is at the top of it.
14961 * The ']' ending the construct is treated as the lowest operator of all,
14962 * so that everything gets evaluated down to a single operand, which is the
14965 sv_2mortal((SV *)(stack = newAV()));
14966 sv_2mortal((SV *)(fence_stack = newAV()));
14968 while (RExC_parse < RExC_end) {
14969 I32 top_index; /* Index of top-most element in 'stack' */
14970 SV** top_ptr; /* Pointer to top 'stack' element */
14971 SV* current = NULL; /* To contain the current inversion list
14973 SV* only_to_avoid_leaks;
14975 skip_to_be_ignored_text(pRExC_state, &RExC_parse,
14976 TRUE /* Force /x */ );
14977 if (RExC_parse >= RExC_end) {
14978 Perl_croak(aTHX_ "panic: Read past end of '(?[ ])'");
14981 curchar = UCHARAT(RExC_parse);
14985 #ifdef ENABLE_REGEX_SETS_DEBUGGING
14986 /* Enable with -Accflags=-DENABLE_REGEX_SETS_DEBUGGING */
14987 DEBUG_U(dump_regex_sets_structures(pRExC_state,
14988 stack, fence, fence_stack));
14991 top_index = av_tindex_skip_len_mg(stack);
14994 SV** stacked_ptr; /* Ptr to something already on 'stack' */
14995 char stacked_operator; /* The topmost operator on the 'stack'. */
14996 SV* lhs; /* Operand to the left of the operator */
14997 SV* rhs; /* Operand to the right of the operator */
14998 SV* fence_ptr; /* Pointer to top element of the fence
15003 if ( RExC_parse < RExC_end - 1
15004 && (UCHARAT(RExC_parse + 1) == '?'))
15006 /* If is a '(?', could be an embedded '(?flags:(?[...])'.
15007 * This happens when we have some thing like
15009 * my $thai_or_lao = qr/(?[ \p{Thai} + \p{Lao} ])/;
15011 * qr/(?[ \p{Digit} & $thai_or_lao ])/;
15013 * Here we would be handling the interpolated
15014 * '$thai_or_lao'. We handle this by a recursive call to
15015 * ourselves which returns the inversion list the
15016 * interpolated expression evaluates to. We use the flags
15017 * from the interpolated pattern. */
15018 U32 save_flags = RExC_flags;
15019 const char * save_parse;
15021 RExC_parse += 2; /* Skip past the '(?' */
15022 save_parse = RExC_parse;
15024 /* Parse any flags for the '(?' */
15025 parse_lparen_question_flags(pRExC_state);
15027 if (RExC_parse == save_parse /* Makes sure there was at
15028 least one flag (or else
15029 this embedding wasn't
15031 || RExC_parse >= RExC_end - 4
15032 || UCHARAT(RExC_parse) != ':'
15033 || UCHARAT(++RExC_parse) != '('
15034 || UCHARAT(++RExC_parse) != '?'
15035 || UCHARAT(++RExC_parse) != '[')
15038 /* In combination with the above, this moves the
15039 * pointer to the point just after the first erroneous
15040 * character (or if there are no flags, to where they
15041 * should have been) */
15042 if (RExC_parse >= RExC_end - 4) {
15043 RExC_parse = RExC_end;
15045 else if (RExC_parse != save_parse) {
15046 RExC_parse += (UTF) ? UTF8SKIP(RExC_parse) : 1;
15048 vFAIL("Expecting '(?flags:(?[...'");
15051 /* Recurse, with the meat of the embedded expression */
15053 (void) handle_regex_sets(pRExC_state, ¤t, flagp,
15054 depth+1, oregcomp_parse);
15056 /* Here, 'current' contains the embedded expression's
15057 * inversion list, and RExC_parse points to the trailing
15058 * ']'; the next character should be the ')' */
15060 assert(UCHARAT(RExC_parse) == ')');
15062 /* Then the ')' matching the original '(' handled by this
15063 * case: statement */
15065 assert(UCHARAT(RExC_parse) == ')');
15068 RExC_flags = save_flags;
15069 goto handle_operand;
15072 /* A regular '('. Look behind for illegal syntax */
15073 if (top_index - fence >= 0) {
15074 /* If the top entry on the stack is an operator, it had
15075 * better be a '!', otherwise the entry below the top
15076 * operand should be an operator */
15077 if ( ! (top_ptr = av_fetch(stack, top_index, FALSE))
15078 || (IS_OPERATOR(*top_ptr) && SvUV(*top_ptr) != '!')
15079 || ( IS_OPERAND(*top_ptr)
15080 && ( top_index - fence < 1
15081 || ! (stacked_ptr = av_fetch(stack,
15084 || ! IS_OPERATOR(*stacked_ptr))))
15087 vFAIL("Unexpected '(' with no preceding operator");
15091 /* Stack the position of this undealt-with left paren */
15092 av_push(fence_stack, newSViv(fence));
15093 fence = top_index + 1;
15097 /* regclass() can only return RESTART_PASS1 and NEED_UTF8 if
15098 * multi-char folds are allowed. */
15099 if (!regclass(pRExC_state, flagp,depth+1,
15100 TRUE, /* means parse just the next thing */
15101 FALSE, /* don't allow multi-char folds */
15102 FALSE, /* don't silence non-portable warnings. */
15104 FALSE, /* Require return to be an ANYOF */
15108 FAIL2("panic: regclass returned NULL to handle_sets, "
15109 "flags=%#" UVxf, (UV) *flagp);
15112 /* regclass() will return with parsing just the \ sequence,
15113 * leaving the parse pointer at the next thing to parse */
15115 goto handle_operand;
15117 case '[': /* Is a bracketed character class */
15119 /* See if this is a [:posix:] class. */
15120 bool is_posix_class = (OOB_NAMEDCLASS
15121 < handle_possible_posix(pRExC_state,
15125 TRUE /* checking only */));
15126 /* If it is a posix class, leave the parse pointer at the '['
15127 * to fool regclass() into thinking it is part of a
15128 * '[[:posix:]]'. */
15129 if (! is_posix_class) {
15133 /* regclass() can only return RESTART_PASS1 and NEED_UTF8 if
15134 * multi-char folds are allowed. */
15135 if (!regclass(pRExC_state, flagp,depth+1,
15136 is_posix_class, /* parse the whole char
15137 class only if not a
15139 FALSE, /* don't allow multi-char folds */
15140 TRUE, /* silence non-portable warnings. */
15142 FALSE, /* Require return to be an ANYOF */
15147 FAIL2("panic: regclass returned NULL to handle_sets, "
15148 "flags=%#" UVxf, (UV) *flagp);
15151 /* function call leaves parse pointing to the ']', except if we
15153 if (is_posix_class) {
15157 goto handle_operand;
15161 if (top_index >= 1) {
15162 goto join_operators;
15165 /* Only a single operand on the stack: are done */
15169 if (av_tindex_skip_len_mg(fence_stack) < 0) {
15171 vFAIL("Unexpected ')'");
15174 /* If nothing after the fence, is missing an operand */
15175 if (top_index - fence < 0) {
15179 /* If at least two things on the stack, treat this as an
15181 if (top_index - fence >= 1) {
15182 goto join_operators;
15185 /* Here only a single thing on the fenced stack, and there is a
15186 * fence. Get rid of it */
15187 fence_ptr = av_pop(fence_stack);
15189 fence = SvIV(fence_ptr) - 1;
15190 SvREFCNT_dec_NN(fence_ptr);
15197 /* Having gotten rid of the fence, we pop the operand at the
15198 * stack top and process it as a newly encountered operand */
15199 current = av_pop(stack);
15200 if (IS_OPERAND(current)) {
15201 goto handle_operand;
15213 /* These binary operators should have a left operand already
15215 if ( top_index - fence < 0
15216 || top_index - fence == 1
15217 || ( ! (top_ptr = av_fetch(stack, top_index, FALSE)))
15218 || ! IS_OPERAND(*top_ptr))
15220 goto unexpected_binary;
15223 /* If only the one operand is on the part of the stack visible
15224 * to us, we just place this operator in the proper position */
15225 if (top_index - fence < 2) {
15227 /* Place the operator before the operand */
15229 SV* lhs = av_pop(stack);
15230 av_push(stack, newSVuv(curchar));
15231 av_push(stack, lhs);
15235 /* But if there is something else on the stack, we need to
15236 * process it before this new operator if and only if the
15237 * stacked operation has equal or higher precedence than the
15242 /* The operator on the stack is supposed to be below both its
15244 if ( ! (stacked_ptr = av_fetch(stack, top_index - 2, FALSE))
15245 || IS_OPERAND(*stacked_ptr))
15247 /* But if not, it's legal and indicates we are completely
15248 * done if and only if we're currently processing a ']',
15249 * which should be the final thing in the expression */
15250 if (curchar == ']') {
15256 vFAIL2("Unexpected binary operator '%c' with no "
15257 "preceding operand", curchar);
15259 stacked_operator = (char) SvUV(*stacked_ptr);
15261 if (regex_set_precedence(curchar)
15262 > regex_set_precedence(stacked_operator))
15264 /* Here, the new operator has higher precedence than the
15265 * stacked one. This means we need to add the new one to
15266 * the stack to await its rhs operand (and maybe more
15267 * stuff). We put it before the lhs operand, leaving
15268 * untouched the stacked operator and everything below it
15270 lhs = av_pop(stack);
15271 assert(IS_OPERAND(lhs));
15273 av_push(stack, newSVuv(curchar));
15274 av_push(stack, lhs);
15278 /* Here, the new operator has equal or lower precedence than
15279 * what's already there. This means the operation already
15280 * there should be performed now, before the new one. */
15282 rhs = av_pop(stack);
15283 if (! IS_OPERAND(rhs)) {
15285 /* This can happen when a ! is not followed by an operand,
15286 * like in /(?[\t &!])/ */
15290 lhs = av_pop(stack);
15292 if (! IS_OPERAND(lhs)) {
15294 /* This can happen when there is an empty (), like in
15295 * /(?[[0]+()+])/ */
15299 switch (stacked_operator) {
15301 _invlist_intersection(lhs, rhs, &rhs);
15306 _invlist_union(lhs, rhs, &rhs);
15310 _invlist_subtract(lhs, rhs, &rhs);
15313 case '^': /* The union minus the intersection */
15318 _invlist_union(lhs, rhs, &u);
15319 _invlist_intersection(lhs, rhs, &i);
15320 _invlist_subtract(u, i, &rhs);
15321 SvREFCNT_dec_NN(i);
15322 SvREFCNT_dec_NN(u);
15328 /* Here, the higher precedence operation has been done, and the
15329 * result is in 'rhs'. We overwrite the stacked operator with
15330 * the result. Then we redo this code to either push the new
15331 * operator onto the stack or perform any higher precedence
15332 * stacked operation */
15333 only_to_avoid_leaks = av_pop(stack);
15334 SvREFCNT_dec(only_to_avoid_leaks);
15335 av_push(stack, rhs);
15338 case '!': /* Highest priority, right associative */
15340 /* If what's already at the top of the stack is another '!",
15341 * they just cancel each other out */
15342 if ( (top_ptr = av_fetch(stack, top_index, FALSE))
15343 && (IS_OPERATOR(*top_ptr) && SvUV(*top_ptr) == '!'))
15345 only_to_avoid_leaks = av_pop(stack);
15346 SvREFCNT_dec(only_to_avoid_leaks);
15348 else { /* Otherwise, since it's right associative, just push
15350 av_push(stack, newSVuv(curchar));
15355 RExC_parse += (UTF) ? UTF8SKIP(RExC_parse) : 1;
15356 vFAIL("Unexpected character");
15360 /* Here 'current' is the operand. If something is already on the
15361 * stack, we have to check if it is a !. But first, the code above
15362 * may have altered the stack in the time since we earlier set
15365 top_index = av_tindex_skip_len_mg(stack);
15366 if (top_index - fence >= 0) {
15367 /* If the top entry on the stack is an operator, it had better
15368 * be a '!', otherwise the entry below the top operand should
15369 * be an operator */
15370 top_ptr = av_fetch(stack, top_index, FALSE);
15372 if (IS_OPERATOR(*top_ptr)) {
15374 /* The only permissible operator at the top of the stack is
15375 * '!', which is applied immediately to this operand. */
15376 curchar = (char) SvUV(*top_ptr);
15377 if (curchar != '!') {
15378 SvREFCNT_dec(current);
15379 vFAIL2("Unexpected binary operator '%c' with no "
15380 "preceding operand", curchar);
15383 _invlist_invert(current);
15385 only_to_avoid_leaks = av_pop(stack);
15386 SvREFCNT_dec(only_to_avoid_leaks);
15388 /* And we redo with the inverted operand. This allows
15389 * handling multiple ! in a row */
15390 goto handle_operand;
15392 /* Single operand is ok only for the non-binary ')'
15394 else if ((top_index - fence == 0 && curchar != ')')
15395 || (top_index - fence > 0
15396 && (! (stacked_ptr = av_fetch(stack,
15399 || IS_OPERAND(*stacked_ptr))))
15401 SvREFCNT_dec(current);
15402 vFAIL("Operand with no preceding operator");
15406 /* Here there was nothing on the stack or the top element was
15407 * another operand. Just add this new one */
15408 av_push(stack, current);
15410 } /* End of switch on next parse token */
15412 RExC_parse += (UTF) ? UTF8SKIP(RExC_parse) : 1;
15413 } /* End of loop parsing through the construct */
15416 if (av_tindex_skip_len_mg(fence_stack) >= 0) {
15417 vFAIL("Unmatched (");
15420 if (av_tindex_skip_len_mg(stack) < 0 /* Was empty */
15421 || ((final = av_pop(stack)) == NULL)
15422 || ! IS_OPERAND(final)
15423 || SvTYPE(final) != SVt_INVLIST
15424 || av_tindex_skip_len_mg(stack) >= 0) /* More left on stack */
15427 SvREFCNT_dec(final);
15428 vFAIL("Incomplete expression within '(?[ ])'");
15431 /* Here, 'final' is the resultant inversion list from evaluating the
15432 * expression. Return it if so requested */
15433 if (return_invlist) {
15434 *return_invlist = final;
15438 /* Otherwise generate a resultant node, based on 'final'. regclass() is
15439 * expecting a string of ranges and individual code points */
15440 invlist_iterinit(final);
15441 result_string = newSVpvs("");
15442 while (invlist_iternext(final, &start, &end)) {
15443 if (start == end) {
15444 Perl_sv_catpvf(aTHX_ result_string, "\\x{%" UVXf "}", start);
15447 Perl_sv_catpvf(aTHX_ result_string, "\\x{%" UVXf "}-\\x{%" UVXf "}",
15452 /* About to generate an ANYOF (or similar) node from the inversion list we
15453 * have calculated */
15454 save_parse = RExC_parse;
15455 RExC_parse = SvPV(result_string, len);
15456 save_end = RExC_end;
15457 RExC_end = RExC_parse + len;
15459 /* We turn off folding around the call, as the class we have constructed
15460 * already has all folding taken into consideration, and we don't want
15461 * regclass() to add to that */
15462 RExC_flags &= ~RXf_PMf_FOLD;
15463 /* regclass() can only return RESTART_PASS1 and NEED_UTF8 if multi-char
15464 * folds are allowed. */
15465 node = regclass(pRExC_state, flagp,depth+1,
15466 FALSE, /* means parse the whole char class */
15467 FALSE, /* don't allow multi-char folds */
15468 TRUE, /* silence non-portable warnings. The above may very
15469 well have generated non-portable code points, but
15470 they're valid on this machine */
15471 FALSE, /* similarly, no need for strict */
15472 FALSE, /* Require return to be an ANYOF */
15477 FAIL2("panic: regclass returned NULL to handle_sets, flags=%#" UVxf,
15480 /* Fix up the node type if we are in locale. (We have pretended we are
15481 * under /u for the purposes of regclass(), as this construct will only
15482 * work under UTF-8 locales. But now we change the opcode to be ANYOFL (so
15483 * as to cause any warnings about bad locales to be output in regexec.c),
15484 * and add the flag that indicates to check if not in a UTF-8 locale. The
15485 * reason we above forbid optimization into something other than an ANYOF
15486 * node is simply to minimize the number of code changes in regexec.c.
15487 * Otherwise we would have to create new EXACTish node types and deal with
15488 * them. This decision could be revisited should this construct become
15491 * (One might think we could look at the resulting ANYOF node and suppress
15492 * the flag if everything is above 255, as those would be UTF-8 only,
15493 * but this isn't true, as the components that led to that result could
15494 * have been locale-affected, and just happen to cancel each other out
15495 * under UTF-8 locales.) */
15497 set_regex_charset(&RExC_flags, REGEX_LOCALE_CHARSET);
15499 assert(OP(node) == ANYOF);
15503 |= ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD;
15507 RExC_flags |= RXf_PMf_FOLD;
15510 RExC_parse = save_parse + 1;
15511 RExC_end = save_end;
15512 SvREFCNT_dec_NN(final);
15513 SvREFCNT_dec_NN(result_string);
15515 nextchar(pRExC_state);
15516 Set_Node_Length(node, RExC_parse - oregcomp_parse + 1); /* MJD */
15520 #ifdef ENABLE_REGEX_SETS_DEBUGGING
15523 S_dump_regex_sets_structures(pTHX_ RExC_state_t *pRExC_state,
15524 AV * stack, const IV fence, AV * fence_stack)
15525 { /* Dumps the stacks in handle_regex_sets() */
15527 const SSize_t stack_top = av_tindex_skip_len_mg(stack);
15528 const SSize_t fence_stack_top = av_tindex_skip_len_mg(fence_stack);
15531 PERL_ARGS_ASSERT_DUMP_REGEX_SETS_STRUCTURES;
15533 PerlIO_printf(Perl_debug_log, "\nParse position is:%s\n", RExC_parse);
15535 if (stack_top < 0) {
15536 PerlIO_printf(Perl_debug_log, "Nothing on stack\n");
15539 PerlIO_printf(Perl_debug_log, "Stack: (fence=%d)\n", (int) fence);
15540 for (i = stack_top; i >= 0; i--) {
15541 SV ** element_ptr = av_fetch(stack, i, FALSE);
15542 if (! element_ptr) {
15545 if (IS_OPERATOR(*element_ptr)) {
15546 PerlIO_printf(Perl_debug_log, "[%d]: %c\n",
15547 (int) i, (int) SvIV(*element_ptr));
15550 PerlIO_printf(Perl_debug_log, "[%d] ", (int) i);
15551 sv_dump(*element_ptr);
15556 if (fence_stack_top < 0) {
15557 PerlIO_printf(Perl_debug_log, "Nothing on fence_stack\n");
15560 PerlIO_printf(Perl_debug_log, "Fence_stack: \n");
15561 for (i = fence_stack_top; i >= 0; i--) {
15562 SV ** element_ptr = av_fetch(fence_stack, i, FALSE);
15563 if (! element_ptr) {
15566 PerlIO_printf(Perl_debug_log, "[%d]: %d\n",
15567 (int) i, (int) SvIV(*element_ptr));
15578 S_add_above_Latin1_folds(pTHX_ RExC_state_t *pRExC_state, const U8 cp, SV** invlist)
15580 /* This hard-codes the Latin1/above-Latin1 folding rules, so that an
15581 * innocent-looking character class, like /[ks]/i won't have to go out to
15582 * disk to find the possible matches.
15584 * This should be called only for a Latin1-range code points, cp, which is
15585 * known to be involved in a simple fold with other code points above
15586 * Latin1. It would give false results if /aa has been specified.
15587 * Multi-char folds are outside the scope of this, and must be handled
15590 * XXX It would be better to generate these via regen, in case a new
15591 * version of the Unicode standard adds new mappings, though that is not
15592 * really likely, and may be caught by the default: case of the switch
15595 PERL_ARGS_ASSERT_ADD_ABOVE_LATIN1_FOLDS;
15597 assert(HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE(cp));
15603 add_cp_to_invlist(*invlist, KELVIN_SIGN);
15607 *invlist = add_cp_to_invlist(*invlist, LATIN_SMALL_LETTER_LONG_S);
15610 *invlist = add_cp_to_invlist(*invlist, GREEK_CAPITAL_LETTER_MU);
15611 *invlist = add_cp_to_invlist(*invlist, GREEK_SMALL_LETTER_MU);
15613 case LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE:
15614 case LATIN_SMALL_LETTER_A_WITH_RING_ABOVE:
15615 *invlist = add_cp_to_invlist(*invlist, ANGSTROM_SIGN);
15617 case LATIN_SMALL_LETTER_Y_WITH_DIAERESIS:
15618 *invlist = add_cp_to_invlist(*invlist,
15619 LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS);
15622 #ifdef LATIN_CAPITAL_LETTER_SHARP_S /* not defined in early Unicode releases */
15624 case LATIN_SMALL_LETTER_SHARP_S:
15625 *invlist = add_cp_to_invlist(*invlist, LATIN_CAPITAL_LETTER_SHARP_S);
15630 #if UNICODE_MAJOR_VERSION < 3 \
15631 || (UNICODE_MAJOR_VERSION == 3 && UNICODE_DOT_VERSION == 0)
15633 /* In 3.0 and earlier, U+0130 folded simply to 'i'; and in 3.0.1 so did
15638 add_cp_to_invlist(*invlist, LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE);
15639 # if UNICODE_DOT_DOT_VERSION == 1
15640 *invlist = add_cp_to_invlist(*invlist, LATIN_SMALL_LETTER_DOTLESS_I);
15646 /* Use deprecated warning to increase the chances of this being
15649 ckWARN2reg_d(RExC_parse, "Perl folding rules are not up-to-date for 0x%02X; please use the perlbug utility to report;", cp);
15656 S_output_or_return_posix_warnings(pTHX_ RExC_state_t *pRExC_state, AV* posix_warnings, AV** return_posix_warnings)
15658 /* If the final parameter is NULL, output the elements of the array given
15659 * by '*posix_warnings' as REGEXP warnings. Otherwise, the elements are
15660 * pushed onto it, (creating if necessary) */
15663 const bool first_is_fatal = ! return_posix_warnings
15664 && ckDEAD(packWARN(WARN_REGEXP));
15666 PERL_ARGS_ASSERT_OUTPUT_OR_RETURN_POSIX_WARNINGS;
15668 while ((msg = av_shift(posix_warnings)) != &PL_sv_undef) {
15669 if (return_posix_warnings) {
15670 if (! *return_posix_warnings) { /* mortalize to not leak if
15671 warnings are fatal */
15672 *return_posix_warnings = (AV *) sv_2mortal((SV *) newAV());
15674 av_push(*return_posix_warnings, msg);
15677 if (first_is_fatal) { /* Avoid leaking this */
15678 av_undef(posix_warnings); /* This isn't necessary if the
15679 array is mortal, but is a
15681 (void) sv_2mortal(msg);
15683 SAVEFREESV(RExC_rx_sv);
15686 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s", SvPVX(msg));
15687 SvREFCNT_dec_NN(msg);
15693 S_add_multi_match(pTHX_ AV* multi_char_matches, SV* multi_string, const STRLEN cp_count)
15695 /* This adds the string scalar <multi_string> to the array
15696 * <multi_char_matches>. <multi_string> is known to have exactly
15697 * <cp_count> code points in it. This is used when constructing a
15698 * bracketed character class and we find something that needs to match more
15699 * than a single character.
15701 * <multi_char_matches> is actually an array of arrays. Each top-level
15702 * element is an array that contains all the strings known so far that are
15703 * the same length. And that length (in number of code points) is the same
15704 * as the index of the top-level array. Hence, the [2] element is an
15705 * array, each element thereof is a string containing TWO code points;
15706 * while element [3] is for strings of THREE characters, and so on. Since
15707 * this is for multi-char strings there can never be a [0] nor [1] element.
15709 * When we rewrite the character class below, we will do so such that the
15710 * longest strings are written first, so that it prefers the longest
15711 * matching strings first. This is done even if it turns out that any
15712 * quantifier is non-greedy, out of this programmer's (khw) laziness. Tom
15713 * Christiansen has agreed that this is ok. This makes the test for the
15714 * ligature 'ffi' come before the test for 'ff', for example */
15717 AV** this_array_ptr;
15719 PERL_ARGS_ASSERT_ADD_MULTI_MATCH;
15721 if (! multi_char_matches) {
15722 multi_char_matches = newAV();
15725 if (av_exists(multi_char_matches, cp_count)) {
15726 this_array_ptr = (AV**) av_fetch(multi_char_matches, cp_count, FALSE);
15727 this_array = *this_array_ptr;
15730 this_array = newAV();
15731 av_store(multi_char_matches, cp_count,
15734 av_push(this_array, multi_string);
15736 return multi_char_matches;
15739 /* The names of properties whose definitions are not known at compile time are
15740 * stored in this SV, after a constant heading. So if the length has been
15741 * changed since initialization, then there is a run-time definition. */
15742 #define HAS_NONLOCALE_RUNTIME_PROPERTY_DEFINITION \
15743 (SvCUR(listsv) != initial_listsv_len)
15745 /* There is a restricted set of white space characters that are legal when
15746 * ignoring white space in a bracketed character class. This generates the
15747 * code to skip them.
15749 * There is a line below that uses the same white space criteria but is outside
15750 * this macro. Both here and there must use the same definition */
15751 #define SKIP_BRACKETED_WHITE_SPACE(do_skip, p) \
15754 while (isBLANK_A(UCHARAT(p))) \
15762 S_regclass(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, U32 depth,
15763 const bool stop_at_1, /* Just parse the next thing, don't
15764 look for a full character class */
15765 bool allow_multi_folds,
15766 const bool silence_non_portable, /* Don't output warnings
15770 bool optimizable, /* ? Allow a non-ANYOF return
15772 SV** ret_invlist, /* Return an inversion list, not a node */
15773 AV** return_posix_warnings
15776 /* parse a bracketed class specification. Most of these will produce an
15777 * ANYOF node; but something like [a] will produce an EXACT node; [aA], an
15778 * EXACTFish node; [[:ascii:]], a POSIXA node; etc. It is more complex
15779 * under /i with multi-character folds: it will be rewritten following the
15780 * paradigm of this example, where the <multi-fold>s are characters which
15781 * fold to multiple character sequences:
15782 * /[abc\x{multi-fold1}def\x{multi-fold2}ghi]/i
15783 * gets effectively rewritten as:
15784 * /(?:\x{multi-fold1}|\x{multi-fold2}|[abcdefghi]/i
15785 * reg() gets called (recursively) on the rewritten version, and this
15786 * function will return what it constructs. (Actually the <multi-fold>s
15787 * aren't physically removed from the [abcdefghi], it's just that they are
15788 * ignored in the recursion by means of a flag:
15789 * <RExC_in_multi_char_class>.)
15791 * ANYOF nodes contain a bit map for the first NUM_ANYOF_CODE_POINTS
15792 * characters, with the corresponding bit set if that character is in the
15793 * list. For characters above this, a range list or swash is used. There
15794 * are extra bits for \w, etc. in locale ANYOFs, as what these match is not
15795 * determinable at compile time
15797 * Returns NULL, setting *flagp to RESTART_PASS1 if the sizing scan needs
15798 * to be restarted, or'd with NEED_UTF8 if the pattern needs to be upgraded
15799 * to UTF-8. This can only happen if ret_invlist is non-NULL.
15802 UV prevvalue = OOB_UNICODE, save_prevvalue = OOB_UNICODE;
15804 UV value = OOB_UNICODE, save_value = OOB_UNICODE;
15807 int namedclass = OOB_NAMEDCLASS;
15808 char *rangebegin = NULL;
15809 bool need_class = 0;
15811 STRLEN initial_listsv_len = 0; /* Kind of a kludge to see if it is more
15812 than just initialized. */
15813 SV* properties = NULL; /* Code points that match \p{} \P{} */
15814 SV* posixes = NULL; /* Code points that match classes like [:word:],
15815 extended beyond the Latin1 range. These have to
15816 be kept separate from other code points for much
15817 of this function because their handling is
15818 different under /i, and for most classes under
15820 SV* nposixes = NULL; /* Similarly for [:^word:]. These are kept
15821 separate for a while from the non-complemented
15822 versions because of complications with /d
15824 SV* simple_posixes = NULL; /* But under some conditions, the classes can be
15825 treated more simply than the general case,
15826 leading to less compilation and execution
15828 UV element_count = 0; /* Number of distinct elements in the class.
15829 Optimizations may be possible if this is tiny */
15830 AV * multi_char_matches = NULL; /* Code points that fold to more than one
15831 character; used under /i */
15833 char * stop_ptr = RExC_end; /* where to stop parsing */
15835 /* ignore unescaped whitespace? */
15836 const bool skip_white = cBOOL( ret_invlist
15837 || (RExC_flags & RXf_PMf_EXTENDED_MORE));
15839 /* Unicode properties are stored in a swash; this holds the current one
15840 * being parsed. If this swash is the only above-latin1 component of the
15841 * character class, an optimization is to pass it directly on to the
15842 * execution engine. Otherwise, it is set to NULL to indicate that there
15843 * are other things in the class that have to be dealt with at execution
15845 SV* swash = NULL; /* Code points that match \p{} \P{} */
15847 /* Set if a component of this character class is user-defined; just passed
15848 * on to the engine */
15849 bool has_user_defined_property = FALSE;
15851 /* inversion list of code points this node matches only when the target
15852 * string is in UTF-8. These are all non-ASCII, < 256. (Because is under
15854 SV* has_upper_latin1_only_utf8_matches = NULL;
15856 /* Inversion list of code points this node matches regardless of things
15857 * like locale, folding, utf8ness of the target string */
15858 SV* cp_list = NULL;
15860 /* Like cp_list, but code points on this list need to be checked for things
15861 * that fold to/from them under /i */
15862 SV* cp_foldable_list = NULL;
15864 /* Like cp_list, but code points on this list are valid only when the
15865 * runtime locale is UTF-8 */
15866 SV* only_utf8_locale_list = NULL;
15868 /* In a range, if one of the endpoints is non-character-set portable,
15869 * meaning that it hard-codes a code point that may mean a different
15870 * charactger in ASCII vs. EBCDIC, as opposed to, say, a literal 'A' or a
15871 * mnemonic '\t' which each mean the same character no matter which
15872 * character set the platform is on. */
15873 unsigned int non_portable_endpoint = 0;
15875 /* Is the range unicode? which means on a platform that isn't 1-1 native
15876 * to Unicode (i.e. non-ASCII), each code point in it should be considered
15877 * to be a Unicode value. */
15878 bool unicode_range = FALSE;
15879 bool invert = FALSE; /* Is this class to be complemented */
15881 bool warn_super = ALWAYS_WARN_SUPER;
15883 regnode * const orig_emit = RExC_emit; /* Save the original RExC_emit in
15884 case we need to change the emitted regop to an EXACT. */
15885 const char * orig_parse = RExC_parse;
15886 const SSize_t orig_size = RExC_size;
15887 bool posixl_matches_all = FALSE; /* Does /l class have both e.g. \W,\w ? */
15889 /* This variable is used to mark where the end in the input is of something
15890 * that looks like a POSIX construct but isn't. During the parse, when
15891 * something looks like it could be such a construct is encountered, it is
15892 * checked for being one, but not if we've already checked this area of the
15893 * input. Only after this position is reached do we check again */
15894 char *not_posix_region_end = RExC_parse - 1;
15896 AV* posix_warnings = NULL;
15897 const bool do_posix_warnings = return_posix_warnings
15898 || (PASS2 && ckWARN(WARN_REGEXP));
15900 GET_RE_DEBUG_FLAGS_DECL;
15902 PERL_ARGS_ASSERT_REGCLASS;
15904 PERL_UNUSED_ARG(depth);
15907 DEBUG_PARSE("clas");
15909 #if UNICODE_MAJOR_VERSION < 3 /* no multifolds in early Unicode */ \
15910 || (UNICODE_MAJOR_VERSION == 3 && UNICODE_DOT_VERSION == 0 \
15911 && UNICODE_DOT_DOT_VERSION == 0)
15912 allow_multi_folds = FALSE;
15915 /* Assume we are going to generate an ANYOF node. */
15916 ret = reganode(pRExC_state,
15923 RExC_size += ANYOF_SKIP;
15924 listsv = &PL_sv_undef; /* For code scanners: listsv always non-NULL. */
15927 ANYOF_FLAGS(ret) = 0;
15929 RExC_emit += ANYOF_SKIP;
15930 listsv = newSVpvs_flags("# comment\n", SVs_TEMP);
15931 initial_listsv_len = SvCUR(listsv);
15932 SvTEMP_off(listsv); /* Grr, TEMPs and mortals are conflated. */
15935 SKIP_BRACKETED_WHITE_SPACE(skip_white, RExC_parse);
15937 assert(RExC_parse <= RExC_end);
15939 if (UCHARAT(RExC_parse) == '^') { /* Complement the class */
15942 allow_multi_folds = FALSE;
15944 SKIP_BRACKETED_WHITE_SPACE(skip_white, RExC_parse);
15947 /* Check that they didn't say [:posix:] instead of [[:posix:]] */
15948 if (! ret_invlist && MAYBE_POSIXCC(UCHARAT(RExC_parse))) {
15949 int maybe_class = handle_possible_posix(pRExC_state,
15951 ¬_posix_region_end,
15953 TRUE /* checking only */);
15954 if (PASS2 && maybe_class >= OOB_NAMEDCLASS && do_posix_warnings) {
15955 SAVEFREESV(RExC_rx_sv);
15956 ckWARN4reg(not_posix_region_end,
15957 "POSIX syntax [%c %c] belongs inside character classes%s",
15958 *RExC_parse, *RExC_parse,
15959 (maybe_class == OOB_NAMEDCLASS)
15960 ? ((POSIXCC_NOTYET(*RExC_parse))
15961 ? " (but this one isn't implemented)"
15962 : " (but this one isn't fully valid)")
15965 (void)ReREFCNT_inc(RExC_rx_sv);
15969 /* If the caller wants us to just parse a single element, accomplish this
15970 * by faking the loop ending condition */
15971 if (stop_at_1 && RExC_end > RExC_parse) {
15972 stop_ptr = RExC_parse + 1;
15975 /* allow 1st char to be ']' (allowing it to be '-' is dealt with later) */
15976 if (UCHARAT(RExC_parse) == ']')
15977 goto charclassloop;
15981 if ( posix_warnings
15982 && av_tindex_skip_len_mg(posix_warnings) >= 0
15983 && RExC_parse > not_posix_region_end)
15985 /* Warnings about posix class issues are considered tentative until
15986 * we are far enough along in the parse that we can no longer
15987 * change our mind, at which point we either output them or add
15988 * them, if it has so specified, to what gets returned to the
15989 * caller. This is done each time through the loop so that a later
15990 * class won't zap them before they have been dealt with. */
15991 output_or_return_posix_warnings(pRExC_state, posix_warnings,
15992 return_posix_warnings);
15995 if (RExC_parse >= stop_ptr) {
15999 SKIP_BRACKETED_WHITE_SPACE(skip_white, RExC_parse);
16001 if (UCHARAT(RExC_parse) == ']') {
16007 namedclass = OOB_NAMEDCLASS; /* initialize as illegal */
16008 save_value = value;
16009 save_prevvalue = prevvalue;
16012 rangebegin = RExC_parse;
16014 non_portable_endpoint = 0;
16016 if (UTF && ! UTF8_IS_INVARIANT(* RExC_parse)) {
16017 value = utf8n_to_uvchr((U8*)RExC_parse,
16018 RExC_end - RExC_parse,
16019 &numlen, UTF8_ALLOW_DEFAULT);
16020 RExC_parse += numlen;
16023 value = UCHARAT(RExC_parse++);
16025 if (value == '[') {
16026 char * posix_class_end;
16027 namedclass = handle_possible_posix(pRExC_state,
16030 do_posix_warnings ? &posix_warnings : NULL,
16031 FALSE /* die if error */);
16032 if (namedclass > OOB_NAMEDCLASS) {
16034 /* If there was an earlier attempt to parse this particular
16035 * posix class, and it failed, it was a false alarm, as this
16036 * successful one proves */
16037 if ( posix_warnings
16038 && av_tindex_skip_len_mg(posix_warnings) >= 0
16039 && not_posix_region_end >= RExC_parse
16040 && not_posix_region_end <= posix_class_end)
16042 av_undef(posix_warnings);
16045 RExC_parse = posix_class_end;
16047 else if (namedclass == OOB_NAMEDCLASS) {
16048 not_posix_region_end = posix_class_end;
16051 namedclass = OOB_NAMEDCLASS;
16054 else if ( RExC_parse - 1 > not_posix_region_end
16055 && MAYBE_POSIXCC(value))
16057 (void) handle_possible_posix(
16059 RExC_parse - 1, /* -1 because parse has already been
16061 ¬_posix_region_end,
16062 do_posix_warnings ? &posix_warnings : NULL,
16063 TRUE /* checking only */);
16065 else if (value == '\\') {
16066 /* Is a backslash; get the code point of the char after it */
16068 if (RExC_parse >= RExC_end) {
16069 vFAIL("Unmatched [");
16072 if (UTF && ! UTF8_IS_INVARIANT(UCHARAT(RExC_parse))) {
16073 value = utf8n_to_uvchr((U8*)RExC_parse,
16074 RExC_end - RExC_parse,
16075 &numlen, UTF8_ALLOW_DEFAULT);
16076 RExC_parse += numlen;
16079 value = UCHARAT(RExC_parse++);
16081 /* Some compilers cannot handle switching on 64-bit integer
16082 * values, therefore value cannot be an UV. Yes, this will
16083 * be a problem later if we want switch on Unicode.
16084 * A similar issue a little bit later when switching on
16085 * namedclass. --jhi */
16087 /* If the \ is escaping white space when white space is being
16088 * skipped, it means that that white space is wanted literally, and
16089 * is already in 'value'. Otherwise, need to translate the escape
16090 * into what it signifies. */
16091 if (! skip_white || ! isBLANK_A(value)) switch ((I32)value) {
16093 case 'w': namedclass = ANYOF_WORDCHAR; break;
16094 case 'W': namedclass = ANYOF_NWORDCHAR; break;
16095 case 's': namedclass = ANYOF_SPACE; break;
16096 case 'S': namedclass = ANYOF_NSPACE; break;
16097 case 'd': namedclass = ANYOF_DIGIT; break;
16098 case 'D': namedclass = ANYOF_NDIGIT; break;
16099 case 'v': namedclass = ANYOF_VERTWS; break;
16100 case 'V': namedclass = ANYOF_NVERTWS; break;
16101 case 'h': namedclass = ANYOF_HORIZWS; break;
16102 case 'H': namedclass = ANYOF_NHORIZWS; break;
16103 case 'N': /* Handle \N{NAME} in class */
16105 const char * const backslash_N_beg = RExC_parse - 2;
16108 if (! grok_bslash_N(pRExC_state,
16109 NULL, /* No regnode */
16110 &value, /* Yes single value */
16111 &cp_count, /* Multiple code pt count */
16117 if (*flagp & NEED_UTF8)
16118 FAIL("panic: grok_bslash_N set NEED_UTF8");
16119 if (*flagp & RESTART_PASS1)
16122 if (cp_count < 0) {
16123 vFAIL("\\N in a character class must be a named character: \\N{...}");
16125 else if (cp_count == 0) {
16127 ckWARNreg(RExC_parse,
16128 "Ignoring zero length \\N{} in character class");
16131 else { /* cp_count > 1 */
16132 if (! RExC_in_multi_char_class) {
16133 if (invert || range || *RExC_parse == '-') {
16136 vFAIL("\\N{} in inverted character class or as a range end-point is restricted to one character");
16139 ckWARNreg(RExC_parse, "Using just the first character returned by \\N{} in character class");
16141 break; /* <value> contains the first code
16142 point. Drop out of the switch to
16146 SV * multi_char_N = newSVpvn(backslash_N_beg,
16147 RExC_parse - backslash_N_beg);
16149 = add_multi_match(multi_char_matches,
16154 } /* End of cp_count != 1 */
16156 /* This element should not be processed further in this
16159 value = save_value;
16160 prevvalue = save_prevvalue;
16161 continue; /* Back to top of loop to get next char */
16164 /* Here, is a single code point, and <value> contains it */
16165 unicode_range = TRUE; /* \N{} are Unicode */
16173 /* We will handle any undefined properties ourselves */
16174 U8 swash_init_flags = _CORE_SWASH_INIT_RETURN_IF_UNDEF
16175 /* And we actually would prefer to get
16176 * the straight inversion list of the
16177 * swash, since we will be accessing it
16178 * anyway, to save a little time */
16179 |_CORE_SWASH_INIT_ACCEPT_INVLIST;
16181 if (RExC_parse >= RExC_end)
16182 vFAIL2("Empty \\%c", (U8)value);
16183 if (*RExC_parse == '{') {
16184 const U8 c = (U8)value;
16185 e = strchr(RExC_parse, '}');
16188 vFAIL2("Missing right brace on \\%c{}", c);
16192 while (isSPACE(*RExC_parse)) {
16196 if (UCHARAT(RExC_parse) == '^') {
16198 /* toggle. (The rhs xor gets the single bit that
16199 * differs between P and p; the other xor inverts just
16201 value ^= 'P' ^ 'p';
16204 while (isSPACE(*RExC_parse)) {
16209 if (e == RExC_parse)
16210 vFAIL2("Empty \\%c{}", c);
16212 n = e - RExC_parse;
16213 while (isSPACE(*(RExC_parse + n - 1)))
16215 } /* The \p isn't immediately followed by a '{' */
16216 else if (! isALPHA(*RExC_parse)) {
16217 RExC_parse += (UTF) ? UTF8SKIP(RExC_parse) : 1;
16218 vFAIL2("Character following \\%c must be '{' or a "
16219 "single-character Unicode property name",
16229 char* base_name; /* name after any packages are stripped */
16230 char* lookup_name = NULL;
16231 const char * const colon_colon = "::";
16233 /* Try to get the definition of the property into
16234 * <invlist>. If /i is in effect, the effective property
16235 * will have its name be <__NAME_i>. The design is
16236 * discussed in commit
16237 * 2f833f5208e26b208886e51e09e2c072b5eabb46 */
16238 name = savepv(Perl_form(aTHX_ "%.*s", (int)n, RExC_parse));
16241 lookup_name = savepv(Perl_form(aTHX_ "__%s_i", name));
16243 /* The function call just below that uses this can fail
16244 * to return, leaking memory if we don't do this */
16245 SAVEFREEPV(lookup_name);
16248 /* Look up the property name, and get its swash and
16249 * inversion list, if the property is found */
16250 SvREFCNT_dec(swash); /* Free any left-overs */
16251 swash = _core_swash_init("utf8",
16258 NULL, /* No inversion list */
16261 if (! swash || ! (invlist = _get_swash_invlist(swash))) {
16262 HV* curpkg = (IN_PERL_COMPILETIME)
16264 : CopSTASH(PL_curcop);
16268 if (swash) { /* Got a swash but no inversion list.
16269 Something is likely wrong that will
16270 be sorted-out later */
16271 SvREFCNT_dec_NN(swash);
16275 /* Here didn't find it. It could be a an error (like a
16276 * typo) in specifying a Unicode property, or it could
16277 * be a user-defined property that will be available at
16278 * run-time. The names of these must begin with 'In'
16279 * or 'Is' (after any packages are stripped off). So
16280 * if not one of those, or if we accept only
16281 * compile-time properties, is an error; otherwise add
16282 * it to the list for run-time look up. */
16283 if ((base_name = rninstr(name, name + n,
16284 colon_colon, colon_colon + 2)))
16285 { /* Has ::. We know this must be a user-defined
16288 final_n -= base_name - name;
16297 || base_name[0] != 'I'
16298 || (base_name[1] != 's' && base_name[1] != 'n')
16301 const char * const msg
16303 ? "Illegal user-defined property name"
16304 : "Can't find Unicode property definition";
16305 RExC_parse = e + 1;
16307 /* diag_listed_as: Can't find Unicode property definition "%s" */
16308 vFAIL3utf8f("%s \"%" UTF8f "\"",
16309 msg, UTF8fARG(UTF, n, name));
16312 /* If the property name doesn't already have a package
16313 * name, add the current one to it so that it can be
16314 * referred to outside it. [perl #121777] */
16315 if (! has_pkg && curpkg) {
16316 char* pkgname = HvNAME(curpkg);
16317 if (strNE(pkgname, "main")) {
16318 char* full_name = Perl_form(aTHX_
16322 n = strlen(full_name);
16323 name = savepvn(full_name, n);
16327 Perl_sv_catpvf(aTHX_ listsv, "%cutf8::%s%" UTF8f "%s\n",
16328 (value == 'p' ? '+' : '!'),
16329 (FOLD) ? "__" : "",
16330 UTF8fARG(UTF, n, name),
16331 (FOLD) ? "_i" : "");
16332 has_user_defined_property = TRUE;
16333 optimizable = FALSE; /* Will have to leave this an
16336 /* We don't know yet what this matches, so have to flag
16338 ANYOF_FLAGS(ret) |= ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP;
16342 /* Here, did get the swash and its inversion list. If
16343 * the swash is from a user-defined property, then this
16344 * whole character class should be regarded as such */
16345 if (swash_init_flags
16346 & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY)
16348 has_user_defined_property = TRUE;
16351 /* We warn on matching an above-Unicode code point
16352 * if the match would return true, except don't
16353 * warn for \p{All}, which has exactly one element
16355 (_invlist_contains_cp(invlist, 0x110000)
16356 && (! (_invlist_len(invlist) == 1
16357 && *invlist_array(invlist) == 0)))
16363 /* Invert if asking for the complement */
16364 if (value == 'P') {
16365 _invlist_union_complement_2nd(properties,
16369 /* The swash can't be used as-is, because we've
16370 * inverted things; delay removing it to here after
16371 * have copied its invlist above */
16372 SvREFCNT_dec_NN(swash);
16376 _invlist_union(properties, invlist, &properties);
16380 RExC_parse = e + 1;
16381 namedclass = ANYOF_UNIPROP; /* no official name, but it's
16384 /* \p means they want Unicode semantics */
16385 REQUIRE_UNI_RULES(flagp, NULL);
16388 case 'n': value = '\n'; break;
16389 case 'r': value = '\r'; break;
16390 case 't': value = '\t'; break;
16391 case 'f': value = '\f'; break;
16392 case 'b': value = '\b'; break;
16393 case 'e': value = ESC_NATIVE; break;
16394 case 'a': value = '\a'; break;
16396 RExC_parse--; /* function expects to be pointed at the 'o' */
16398 const char* error_msg;
16399 bool valid = grok_bslash_o(&RExC_parse,
16402 PASS2, /* warnings only in
16405 silence_non_portable,
16411 non_portable_endpoint++;
16414 RExC_parse--; /* function expects to be pointed at the 'x' */
16416 const char* error_msg;
16417 bool valid = grok_bslash_x(&RExC_parse,
16420 PASS2, /* Output warnings */
16422 silence_non_portable,
16428 non_portable_endpoint++;
16431 value = grok_bslash_c(*RExC_parse++, PASS2);
16432 non_portable_endpoint++;
16434 case '0': case '1': case '2': case '3': case '4':
16435 case '5': case '6': case '7':
16437 /* Take 1-3 octal digits */
16438 I32 flags = PERL_SCAN_SILENT_ILLDIGIT;
16439 numlen = (strict) ? 4 : 3;
16440 value = grok_oct(--RExC_parse, &numlen, &flags, NULL);
16441 RExC_parse += numlen;
16444 RExC_parse += (UTF) ? UTF8SKIP(RExC_parse) : 1;
16445 vFAIL("Need exactly 3 octal digits");
16447 else if (! SIZE_ONLY /* like \08, \178 */
16449 && RExC_parse < RExC_end
16450 && isDIGIT(*RExC_parse)
16451 && ckWARN(WARN_REGEXP))
16453 SAVEFREESV(RExC_rx_sv);
16454 reg_warn_non_literal_string(
16456 form_short_octal_warning(RExC_parse, numlen));
16457 (void)ReREFCNT_inc(RExC_rx_sv);
16460 non_portable_endpoint++;
16464 /* Allow \_ to not give an error */
16465 if (!SIZE_ONLY && isWORDCHAR(value) && value != '_') {
16467 vFAIL2("Unrecognized escape \\%c in character class",
16471 SAVEFREESV(RExC_rx_sv);
16472 ckWARN2reg(RExC_parse,
16473 "Unrecognized escape \\%c in character class passed through",
16475 (void)ReREFCNT_inc(RExC_rx_sv);
16479 } /* End of switch on char following backslash */
16480 } /* end of handling backslash escape sequences */
16482 /* Here, we have the current token in 'value' */
16484 if (namedclass > OOB_NAMEDCLASS) { /* this is a named class \blah */
16487 /* a bad range like a-\d, a-[:digit:]. The '-' is taken as a
16488 * literal, as is the character that began the false range, i.e.
16489 * the 'a' in the examples */
16492 const int w = (RExC_parse >= rangebegin)
16493 ? RExC_parse - rangebegin
16497 "False [] range \"%" UTF8f "\"",
16498 UTF8fARG(UTF, w, rangebegin));
16501 SAVEFREESV(RExC_rx_sv); /* in case of fatal warnings */
16502 ckWARN2reg(RExC_parse,
16503 "False [] range \"%" UTF8f "\"",
16504 UTF8fARG(UTF, w, rangebegin));
16505 (void)ReREFCNT_inc(RExC_rx_sv);
16506 cp_list = add_cp_to_invlist(cp_list, '-');
16507 cp_foldable_list = add_cp_to_invlist(cp_foldable_list,
16512 range = 0; /* this was not a true range */
16513 element_count += 2; /* So counts for three values */
16516 classnum = namedclass_to_classnum(namedclass);
16518 if (LOC && namedclass < ANYOF_POSIXL_MAX
16519 #ifndef HAS_ISASCII
16520 && classnum != _CC_ASCII
16523 /* What the Posix classes (like \w, [:space:]) match in locale
16524 * isn't knowable under locale until actual match time. Room
16525 * must be reserved (one time per outer bracketed class) to
16526 * store such classes. The space will contain a bit for each
16527 * named class that is to be matched against. This isn't
16528 * needed for \p{} and pseudo-classes, as they are not affected
16529 * by locale, and hence are dealt with separately */
16530 if (! need_class) {
16533 RExC_size += ANYOF_POSIXL_SKIP - ANYOF_SKIP;
16536 RExC_emit += ANYOF_POSIXL_SKIP - ANYOF_SKIP;
16538 ANYOF_FLAGS(ret) |= ANYOF_MATCHES_POSIXL;
16539 ANYOF_POSIXL_ZERO(ret);
16541 /* We can't change this into some other type of node
16542 * (unless this is the only element, in which case there
16543 * are nodes that mean exactly this) as has runtime
16545 optimizable = FALSE;
16548 /* Coverity thinks it is possible for this to be negative; both
16549 * jhi and khw think it's not, but be safer */
16550 assert(! (ANYOF_FLAGS(ret) & ANYOF_MATCHES_POSIXL)
16551 || (namedclass + ((namedclass % 2) ? -1 : 1)) >= 0);
16553 /* See if it already matches the complement of this POSIX
16555 if ((ANYOF_FLAGS(ret) & ANYOF_MATCHES_POSIXL)
16556 && ANYOF_POSIXL_TEST(ret, namedclass + ((namedclass % 2)
16560 posixl_matches_all = TRUE;
16561 break; /* No need to continue. Since it matches both
16562 e.g., \w and \W, it matches everything, and the
16563 bracketed class can be optimized into qr/./s */
16566 /* Add this class to those that should be checked at runtime */
16567 ANYOF_POSIXL_SET(ret, namedclass);
16569 /* The above-Latin1 characters are not subject to locale rules.
16570 * Just add them, in the second pass, to the
16571 * unconditionally-matched list */
16573 SV* scratch_list = NULL;
16575 /* Get the list of the above-Latin1 code points this
16577 _invlist_intersection_maybe_complement_2nd(PL_AboveLatin1,
16578 PL_XPosix_ptrs[classnum],
16580 /* Odd numbers are complements, like
16581 * NDIGIT, NASCII, ... */
16582 namedclass % 2 != 0,
16584 /* Checking if 'cp_list' is NULL first saves an extra
16585 * clone. Its reference count will be decremented at the
16586 * next union, etc, or if this is the only instance, at the
16587 * end of the routine */
16589 cp_list = scratch_list;
16592 _invlist_union(cp_list, scratch_list, &cp_list);
16593 SvREFCNT_dec_NN(scratch_list);
16595 continue; /* Go get next character */
16598 else if (! SIZE_ONLY) {
16600 /* Here, not in pass1 (in that pass we skip calculating the
16601 * contents of this class), and is not /l, or is a POSIX class
16602 * for which /l doesn't matter (or is a Unicode property, which
16603 * is skipped here). */
16604 if (namedclass >= ANYOF_POSIXL_MAX) { /* If a special class */
16605 if (namedclass != ANYOF_UNIPROP) { /* UNIPROP = \p and \P */
16607 /* Here, should be \h, \H, \v, or \V. None of /d, /i
16608 * nor /l make a difference in what these match,
16609 * therefore we just add what they match to cp_list. */
16610 if (classnum != _CC_VERTSPACE) {
16611 assert( namedclass == ANYOF_HORIZWS
16612 || namedclass == ANYOF_NHORIZWS);
16614 /* It turns out that \h is just a synonym for
16616 classnum = _CC_BLANK;
16619 _invlist_union_maybe_complement_2nd(
16621 PL_XPosix_ptrs[classnum],
16622 namedclass % 2 != 0, /* Complement if odd
16623 (NHORIZWS, NVERTWS)
16628 else if ( UNI_SEMANTICS
16629 || classnum == _CC_ASCII
16630 || (DEPENDS_SEMANTICS && ( classnum == _CC_DIGIT
16631 || classnum == _CC_XDIGIT)))
16633 /* We usually have to worry about /d and /a affecting what
16634 * POSIX classes match, with special code needed for /d
16635 * because we won't know until runtime what all matches.
16636 * But there is no extra work needed under /u, and
16637 * [:ascii:] is unaffected by /a and /d; and :digit: and
16638 * :xdigit: don't have runtime differences under /d. So we
16639 * can special case these, and avoid some extra work below,
16640 * and at runtime. */
16641 _invlist_union_maybe_complement_2nd(
16643 PL_XPosix_ptrs[classnum],
16644 namedclass % 2 != 0,
16647 else { /* Garden variety class. If is NUPPER, NALPHA, ...
16648 complement and use nposixes */
16649 SV** posixes_ptr = namedclass % 2 == 0
16652 _invlist_union_maybe_complement_2nd(
16654 PL_XPosix_ptrs[classnum],
16655 namedclass % 2 != 0,
16659 } /* end of namedclass \blah */
16661 SKIP_BRACKETED_WHITE_SPACE(skip_white, RExC_parse);
16663 /* If 'range' is set, 'value' is the ending of a range--check its
16664 * validity. (If value isn't a single code point in the case of a
16665 * range, we should have figured that out above in the code that
16666 * catches false ranges). Later, we will handle each individual code
16667 * point in the range. If 'range' isn't set, this could be the
16668 * beginning of a range, so check for that by looking ahead to see if
16669 * the next real character to be processed is the range indicator--the
16674 /* For unicode ranges, we have to test that the Unicode as opposed
16675 * to the native values are not decreasing. (Above 255, there is
16676 * no difference between native and Unicode) */
16677 if (unicode_range && prevvalue < 255 && value < 255) {
16678 if (NATIVE_TO_LATIN1(prevvalue) > NATIVE_TO_LATIN1(value)) {
16679 goto backwards_range;
16684 if (prevvalue > value) /* b-a */ {
16689 w = RExC_parse - rangebegin;
16691 "Invalid [] range \"%" UTF8f "\"",
16692 UTF8fARG(UTF, w, rangebegin));
16693 NOT_REACHED; /* NOTREACHED */
16697 prevvalue = value; /* save the beginning of the potential range */
16698 if (! stop_at_1 /* Can't be a range if parsing just one thing */
16699 && *RExC_parse == '-')
16701 char* next_char_ptr = RExC_parse + 1;
16703 /* Get the next real char after the '-' */
16704 SKIP_BRACKETED_WHITE_SPACE(skip_white, next_char_ptr);
16706 /* If the '-' is at the end of the class (just before the ']',
16707 * it is a literal minus; otherwise it is a range */
16708 if (next_char_ptr < RExC_end && *next_char_ptr != ']') {
16709 RExC_parse = next_char_ptr;
16711 /* a bad range like \w-, [:word:]- ? */
16712 if (namedclass > OOB_NAMEDCLASS) {
16713 if (strict || (PASS2 && ckWARN(WARN_REGEXP))) {
16714 const int w = RExC_parse >= rangebegin
16715 ? RExC_parse - rangebegin
16718 vFAIL4("False [] range \"%*.*s\"",
16723 "False [] range \"%*.*s\"",
16728 cp_list = add_cp_to_invlist(cp_list, '-');
16732 range = 1; /* yeah, it's a range! */
16733 continue; /* but do it the next time */
16738 if (namedclass > OOB_NAMEDCLASS) {
16742 /* Here, we have a single value this time through the loop, and
16743 * <prevvalue> is the beginning of the range, if any; or <value> if
16746 /* non-Latin1 code point implies unicode semantics. Must be set in
16747 * pass1 so is there for the whole of pass 2 */
16749 REQUIRE_UNI_RULES(flagp, NULL);
16752 /* Ready to process either the single value, or the completed range.
16753 * For single-valued non-inverted ranges, we consider the possibility
16754 * of multi-char folds. (We made a conscious decision to not do this
16755 * for the other cases because it can often lead to non-intuitive
16756 * results. For example, you have the peculiar case that:
16757 * "s s" =~ /^[^\xDF]+$/i => Y
16758 * "ss" =~ /^[^\xDF]+$/i => N
16760 * See [perl #89750] */
16761 if (FOLD && allow_multi_folds && value == prevvalue) {
16762 if (value == LATIN_SMALL_LETTER_SHARP_S
16763 || (value > 255 && _invlist_contains_cp(PL_HasMultiCharFold,
16766 /* Here <value> is indeed a multi-char fold. Get what it is */
16768 U8 foldbuf[UTF8_MAXBYTES_CASE];
16771 UV folded = _to_uni_fold_flags(
16775 FOLD_FLAGS_FULL | (ASCII_FOLD_RESTRICTED
16776 ? FOLD_FLAGS_NOMIX_ASCII
16780 /* Here, <folded> should be the first character of the
16781 * multi-char fold of <value>, with <foldbuf> containing the
16782 * whole thing. But, if this fold is not allowed (because of
16783 * the flags), <fold> will be the same as <value>, and should
16784 * be processed like any other character, so skip the special
16786 if (folded != value) {
16788 /* Skip if we are recursed, currently parsing the class
16789 * again. Otherwise add this character to the list of
16790 * multi-char folds. */
16791 if (! RExC_in_multi_char_class) {
16792 STRLEN cp_count = utf8_length(foldbuf,
16793 foldbuf + foldlen);
16794 SV* multi_fold = sv_2mortal(newSVpvs(""));
16796 Perl_sv_catpvf(aTHX_ multi_fold, "\\x{%" UVXf "}", value);
16799 = add_multi_match(multi_char_matches,
16805 /* This element should not be processed further in this
16808 value = save_value;
16809 prevvalue = save_prevvalue;
16815 if (strict && PASS2 && ckWARN(WARN_REGEXP)) {
16818 /* If the range starts above 255, everything is portable and
16819 * likely to be so for any forseeable character set, so don't
16821 if (unicode_range && non_portable_endpoint && prevvalue < 256) {
16822 vWARN(RExC_parse, "Both or neither range ends should be Unicode");
16824 else if (prevvalue != value) {
16826 /* Under strict, ranges that stop and/or end in an ASCII
16827 * printable should have each end point be a portable value
16828 * for it (preferably like 'A', but we don't warn if it is
16829 * a (portable) Unicode name or code point), and the range
16830 * must be be all digits or all letters of the same case.
16831 * Otherwise, the range is non-portable and unclear as to
16832 * what it contains */
16833 if ( (isPRINT_A(prevvalue) || isPRINT_A(value))
16834 && ( non_portable_endpoint
16835 || ! ( (isDIGIT_A(prevvalue) && isDIGIT_A(value))
16836 || (isLOWER_A(prevvalue) && isLOWER_A(value))
16837 || (isUPPER_A(prevvalue) && isUPPER_A(value))
16839 vWARN(RExC_parse, "Ranges of ASCII printables should"
16840 " be some subset of \"0-9\","
16841 " \"A-Z\", or \"a-z\"");
16843 else if (prevvalue >= 0x660) { /* ARABIC_INDIC_DIGIT_ZERO */
16844 SSize_t index_start;
16845 SSize_t index_final;
16847 /* But the nature of Unicode and languages mean we
16848 * can't do the same checks for above-ASCII ranges,
16849 * except in the case of digit ones. These should
16850 * contain only digits from the same group of 10. The
16851 * ASCII case is handled just above. 0x660 is the
16852 * first digit character beyond ASCII. Hence here, the
16853 * range could be a range of digits. First some
16854 * unlikely special cases. Grandfather in that a range
16855 * ending in 19DA (NEW TAI LUE THAM DIGIT ONE) is bad
16856 * if its starting value is one of the 10 digits prior
16857 * to it. This is because it is an alternate way of
16858 * writing 19D1, and some people may expect it to be in
16859 * that group. But it is bad, because it won't give
16860 * the expected results. In Unicode 5.2 it was
16861 * considered to be in that group (of 11, hence), but
16862 * this was fixed in the next version */
16864 if (UNLIKELY(value == 0x19DA && prevvalue >= 0x19D0)) {
16865 goto warn_bad_digit_range;
16867 else if (UNLIKELY( prevvalue >= 0x1D7CE
16868 && value <= 0x1D7FF))
16870 /* This is the only other case currently in Unicode
16871 * where the algorithm below fails. The code
16872 * points just above are the end points of a single
16873 * range containing only decimal digits. It is 5
16874 * different series of 0-9. All other ranges of
16875 * digits currently in Unicode are just a single
16876 * series. (And mktables will notify us if a later
16877 * Unicode version breaks this.)
16879 * If the range being checked is at most 9 long,
16880 * and the digit values represented are in
16881 * numerical order, they are from the same series.
16883 if ( value - prevvalue > 9
16884 || ((( value - 0x1D7CE) % 10)
16885 <= (prevvalue - 0x1D7CE) % 10))
16887 goto warn_bad_digit_range;
16892 /* For all other ranges of digits in Unicode, the
16893 * algorithm is just to check if both end points
16894 * are in the same series, which is the same range.
16896 index_start = _invlist_search(
16897 PL_XPosix_ptrs[_CC_DIGIT],
16900 /* Warn if the range starts and ends with a digit,
16901 * and they are not in the same group of 10. */
16902 if ( index_start >= 0
16903 && ELEMENT_RANGE_MATCHES_INVLIST(index_start)
16905 _invlist_search(PL_XPosix_ptrs[_CC_DIGIT],
16906 value)) != index_start
16907 && index_final >= 0
16908 && ELEMENT_RANGE_MATCHES_INVLIST(index_final))
16910 warn_bad_digit_range:
16911 vWARN(RExC_parse, "Ranges of digits should be"
16912 " from the same group of"
16919 if ((! range || prevvalue == value) && non_portable_endpoint) {
16920 if (isPRINT_A(value)) {
16923 if (isBACKSLASHED_PUNCT(value)) {
16924 literal[d++] = '\\';
16926 literal[d++] = (char) value;
16927 literal[d++] = '\0';
16930 "\"%.*s\" is more clearly written simply as \"%s\"",
16931 (int) (RExC_parse - rangebegin),
16936 else if isMNEMONIC_CNTRL(value) {
16938 "\"%.*s\" is more clearly written simply as \"%s\"",
16939 (int) (RExC_parse - rangebegin),
16941 cntrl_to_mnemonic((U8) value)
16947 /* Deal with this element of the class */
16951 cp_foldable_list = _add_range_to_invlist(cp_foldable_list,
16954 /* On non-ASCII platforms, for ranges that span all of 0..255, and
16955 * ones that don't require special handling, we can just add the
16956 * range like we do for ASCII platforms */
16957 if ((UNLIKELY(prevvalue == 0) && value >= 255)
16958 || ! (prevvalue < 256
16960 || (! non_portable_endpoint
16961 && ((isLOWER_A(prevvalue) && isLOWER_A(value))
16962 || (isUPPER_A(prevvalue)
16963 && isUPPER_A(value)))))))
16965 cp_foldable_list = _add_range_to_invlist(cp_foldable_list,
16969 /* Here, requires special handling. This can be because it is
16970 * a range whose code points are considered to be Unicode, and
16971 * so must be individually translated into native, or because
16972 * its a subrange of 'A-Z' or 'a-z' which each aren't
16973 * contiguous in EBCDIC, but we have defined them to include
16974 * only the "expected" upper or lower case ASCII alphabetics.
16975 * Subranges above 255 are the same in native and Unicode, so
16976 * can be added as a range */
16977 U8 start = NATIVE_TO_LATIN1(prevvalue);
16979 U8 end = (value < 256) ? NATIVE_TO_LATIN1(value) : 255;
16980 for (j = start; j <= end; j++) {
16981 cp_foldable_list = add_cp_to_invlist(cp_foldable_list, LATIN1_TO_NATIVE(j));
16984 cp_foldable_list = _add_range_to_invlist(cp_foldable_list,
16991 range = 0; /* this range (if it was one) is done now */
16992 } /* End of loop through all the text within the brackets */
16995 if ( posix_warnings && av_tindex_skip_len_mg(posix_warnings) >= 0) {
16996 output_or_return_posix_warnings(pRExC_state, posix_warnings,
16997 return_posix_warnings);
17000 /* If anything in the class expands to more than one character, we have to
17001 * deal with them by building up a substitute parse string, and recursively
17002 * calling reg() on it, instead of proceeding */
17003 if (multi_char_matches) {
17004 SV * substitute_parse = newSVpvn_flags("?:", 2, SVs_TEMP);
17007 char *save_end = RExC_end;
17008 char *save_parse = RExC_parse;
17009 char *save_start = RExC_start;
17010 STRLEN prefix_end = 0; /* We copy the character class after a
17011 prefix supplied here. This is the size
17012 + 1 of that prefix */
17013 bool first_time = TRUE; /* First multi-char occurrence doesn't get
17018 assert(RExC_precomp_adj == 0); /* Only one level of recursion allowed */
17020 #if 0 /* Have decided not to deal with multi-char folds in inverted classes,
17021 because too confusing */
17023 sv_catpv(substitute_parse, "(?:");
17027 /* Look at the longest folds first */
17028 for (cp_count = av_tindex_skip_len_mg(multi_char_matches);
17033 if (av_exists(multi_char_matches, cp_count)) {
17034 AV** this_array_ptr;
17037 this_array_ptr = (AV**) av_fetch(multi_char_matches,
17039 while ((this_sequence = av_pop(*this_array_ptr)) !=
17042 if (! first_time) {
17043 sv_catpv(substitute_parse, "|");
17045 first_time = FALSE;
17047 sv_catpv(substitute_parse, SvPVX(this_sequence));
17052 /* If the character class contains anything else besides these
17053 * multi-character folds, have to include it in recursive parsing */
17054 if (element_count) {
17055 sv_catpv(substitute_parse, "|[");
17056 prefix_end = SvCUR(substitute_parse);
17057 sv_catpvn(substitute_parse, orig_parse, RExC_parse - orig_parse);
17059 /* Put in a closing ']' only if not going off the end, as otherwise
17060 * we are adding something that really isn't there */
17061 if (RExC_parse < RExC_end) {
17062 sv_catpv(substitute_parse, "]");
17066 sv_catpv(substitute_parse, ")");
17069 /* This is a way to get the parse to skip forward a whole named
17070 * sequence instead of matching the 2nd character when it fails the
17072 sv_catpv(substitute_parse, "(*THEN)(*SKIP)(*FAIL)|.)");
17076 /* Set up the data structure so that any errors will be properly
17077 * reported. See the comments at the definition of
17078 * REPORT_LOCATION_ARGS for details */
17079 RExC_precomp_adj = orig_parse - RExC_precomp;
17080 RExC_start = RExC_parse = SvPV(substitute_parse, len);
17081 RExC_adjusted_start = RExC_start + prefix_end;
17082 RExC_end = RExC_parse + len;
17083 RExC_in_multi_char_class = 1;
17084 RExC_emit = (regnode *)orig_emit;
17086 ret = reg(pRExC_state, 1, ®_flags, depth+1);
17088 *flagp |= reg_flags&(HASWIDTH|SIMPLE|SPSTART|POSTPONED|RESTART_PASS1|NEED_UTF8);
17090 /* And restore so can parse the rest of the pattern */
17091 RExC_parse = save_parse;
17092 RExC_start = RExC_adjusted_start = save_start;
17093 RExC_precomp_adj = 0;
17094 RExC_end = save_end;
17095 RExC_in_multi_char_class = 0;
17096 SvREFCNT_dec_NN(multi_char_matches);
17100 /* Here, we've gone through the entire class and dealt with multi-char
17101 * folds. We are now in a position that we can do some checks to see if we
17102 * can optimize this ANYOF node into a simpler one, even in Pass 1.
17103 * Currently we only do two checks:
17104 * 1) is in the unlikely event that the user has specified both, eg. \w and
17105 * \W under /l, then the class matches everything. (This optimization
17106 * is done only to make the optimizer code run later work.)
17107 * 2) if the character class contains only a single element (including a
17108 * single range), we see if there is an equivalent node for it.
17109 * Other checks are possible */
17111 && ! ret_invlist /* Can't optimize if returning the constructed
17113 && (UNLIKELY(posixl_matches_all) || element_count == 1))
17118 if (UNLIKELY(posixl_matches_all)) {
17121 else if (namedclass > OOB_NAMEDCLASS) { /* this is a single named
17122 class, like \w or [:digit:]
17125 /* All named classes are mapped into POSIXish nodes, with its FLAG
17126 * argument giving which class it is */
17127 switch ((I32)namedclass) {
17128 case ANYOF_UNIPROP:
17131 /* These don't depend on the charset modifiers. They always
17132 * match under /u rules */
17133 case ANYOF_NHORIZWS:
17134 case ANYOF_HORIZWS:
17135 namedclass = ANYOF_BLANK + namedclass - ANYOF_HORIZWS;
17138 case ANYOF_NVERTWS:
17143 /* The actual POSIXish node for all the rest depends on the
17144 * charset modifier. The ones in the first set depend only on
17145 * ASCII or, if available on this platform, also locale */
17149 op = (LOC) ? POSIXL : POSIXA;
17155 /* The following don't have any matches in the upper Latin1
17156 * range, hence /d is equivalent to /u for them. Making it /u
17157 * saves some branches at runtime */
17161 case ANYOF_NXDIGIT:
17162 if (! DEPENDS_SEMANTICS) {
17163 goto treat_as_default;
17169 /* The following change to CASED under /i */
17175 namedclass = ANYOF_CASED + (namedclass % 2);
17179 /* The rest have more possibilities depending on the charset.
17180 * We take advantage of the enum ordering of the charset
17181 * modifiers to get the exact node type, */
17184 op = POSIXD + get_regex_charset(RExC_flags);
17185 if (op > POSIXA) { /* /aa is same as /a */
17190 /* The odd numbered ones are the complements of the
17191 * next-lower even number one */
17192 if (namedclass % 2 == 1) {
17196 arg = namedclass_to_classnum(namedclass);
17200 else if (value == prevvalue) {
17202 /* Here, the class consists of just a single code point */
17205 if (! LOC && value == '\n') {
17206 op = REG_ANY; /* Optimize [^\n] */
17207 *flagp |= HASWIDTH|SIMPLE;
17211 else if (value < 256 || UTF) {
17213 /* Optimize a single value into an EXACTish node, but not if it
17214 * would require converting the pattern to UTF-8. */
17215 op = compute_EXACTish(pRExC_state);
17217 } /* Otherwise is a range */
17218 else if (! LOC) { /* locale could vary these */
17219 if (prevvalue == '0') {
17220 if (value == '9') {
17225 else if (! FOLD || ASCII_FOLD_RESTRICTED) {
17226 /* We can optimize A-Z or a-z, but not if they could match
17227 * something like the KELVIN SIGN under /i. */
17228 if (prevvalue == 'A') {
17231 && ! non_portable_endpoint
17234 arg = (FOLD) ? _CC_ALPHA : _CC_UPPER;
17238 else if (prevvalue == 'a') {
17241 && ! non_portable_endpoint
17244 arg = (FOLD) ? _CC_ALPHA : _CC_LOWER;
17251 /* Here, we have changed <op> away from its initial value iff we found
17252 * an optimization */
17255 /* Throw away this ANYOF regnode, and emit the calculated one,
17256 * which should correspond to the beginning, not current, state of
17258 const char * cur_parse = RExC_parse;
17259 RExC_parse = (char *)orig_parse;
17263 /* To get locale nodes to not use the full ANYOF size would
17264 * require moving the code above that writes the portions
17265 * of it that aren't in other nodes to after this point.
17266 * e.g. ANYOF_POSIXL_SET */
17267 RExC_size = orig_size;
17271 RExC_emit = (regnode *)orig_emit;
17272 if (PL_regkind[op] == POSIXD) {
17273 if (op == POSIXL) {
17274 RExC_contains_locale = 1;
17277 op += NPOSIXD - POSIXD;
17282 ret = reg_node(pRExC_state, op);
17284 if (PL_regkind[op] == POSIXD || PL_regkind[op] == NPOSIXD) {
17288 *flagp |= HASWIDTH|SIMPLE;
17290 else if (PL_regkind[op] == EXACT) {
17291 alloc_maybe_populate_EXACT(pRExC_state, ret, flagp, 0, value,
17292 TRUE /* downgradable to EXACT */
17296 RExC_parse = (char *) cur_parse;
17298 SvREFCNT_dec(posixes);
17299 SvREFCNT_dec(nposixes);
17300 SvREFCNT_dec(simple_posixes);
17301 SvREFCNT_dec(cp_list);
17302 SvREFCNT_dec(cp_foldable_list);
17309 /****** !SIZE_ONLY (Pass 2) AFTER HERE *********/
17311 /* If folding, we calculate all characters that could fold to or from the
17312 * ones already on the list */
17313 if (cp_foldable_list) {
17315 UV start, end; /* End points of code point ranges */
17317 SV* fold_intersection = NULL;
17320 /* Our calculated list will be for Unicode rules. For locale
17321 * matching, we have to keep a separate list that is consulted at
17322 * runtime only when the locale indicates Unicode rules. For
17323 * non-locale, we just use the general list */
17325 use_list = &only_utf8_locale_list;
17328 use_list = &cp_list;
17331 /* Only the characters in this class that participate in folds need
17332 * be checked. Get the intersection of this class and all the
17333 * possible characters that are foldable. This can quickly narrow
17334 * down a large class */
17335 _invlist_intersection(PL_utf8_foldable, cp_foldable_list,
17336 &fold_intersection);
17338 /* The folds for all the Latin1 characters are hard-coded into this
17339 * program, but we have to go out to disk to get the others. */
17340 if (invlist_highest(cp_foldable_list) >= 256) {
17342 /* This is a hash that for a particular fold gives all
17343 * characters that are involved in it */
17344 if (! PL_utf8_foldclosures) {
17345 _load_PL_utf8_foldclosures();
17349 /* Now look at the foldable characters in this class individually */
17350 invlist_iterinit(fold_intersection);
17351 while (invlist_iternext(fold_intersection, &start, &end)) {
17354 /* Look at every character in the range */
17355 for (j = start; j <= end; j++) {
17356 U8 foldbuf[UTF8_MAXBYTES_CASE+1];
17362 if (IS_IN_SOME_FOLD_L1(j)) {
17364 /* ASCII is always matched; non-ASCII is matched
17365 * only under Unicode rules (which could happen
17366 * under /l if the locale is a UTF-8 one */
17367 if (isASCII(j) || ! DEPENDS_SEMANTICS) {
17368 *use_list = add_cp_to_invlist(*use_list,
17369 PL_fold_latin1[j]);
17372 has_upper_latin1_only_utf8_matches
17373 = add_cp_to_invlist(
17374 has_upper_latin1_only_utf8_matches,
17375 PL_fold_latin1[j]);
17379 if (HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE(j)
17380 && (! isASCII(j) || ! ASCII_FOLD_RESTRICTED))
17382 add_above_Latin1_folds(pRExC_state,
17389 /* Here is an above Latin1 character. We don't have the
17390 * rules hard-coded for it. First, get its fold. This is
17391 * the simple fold, as the multi-character folds have been
17392 * handled earlier and separated out */
17393 _to_uni_fold_flags(j, foldbuf, &foldlen,
17394 (ASCII_FOLD_RESTRICTED)
17395 ? FOLD_FLAGS_NOMIX_ASCII
17398 /* Single character fold of above Latin1. Add everything in
17399 * its fold closure to the list that this node should match.
17400 * The fold closures data structure is a hash with the keys
17401 * being the UTF-8 of every character that is folded to, like
17402 * 'k', and the values each an array of all code points that
17403 * fold to its key. e.g. [ 'k', 'K', KELVIN_SIGN ].
17404 * Multi-character folds are not included */
17405 if ((listp = hv_fetch(PL_utf8_foldclosures,
17406 (char *) foldbuf, foldlen, FALSE)))
17408 AV* list = (AV*) *listp;
17410 for (k = 0; k <= av_tindex_skip_len_mg(list); k++) {
17411 SV** c_p = av_fetch(list, k, FALSE);
17417 /* /aa doesn't allow folds between ASCII and non- */
17418 if ((ASCII_FOLD_RESTRICTED
17419 && (isASCII(c) != isASCII(j))))
17424 /* Folds under /l which cross the 255/256 boundary
17425 * are added to a separate list. (These are valid
17426 * only when the locale is UTF-8.) */
17427 if (c < 256 && LOC) {
17428 *use_list = add_cp_to_invlist(*use_list, c);
17432 if (isASCII(c) || c > 255 || AT_LEAST_UNI_SEMANTICS)
17434 cp_list = add_cp_to_invlist(cp_list, c);
17437 /* Similarly folds involving non-ascii Latin1
17438 * characters under /d are added to their list */
17439 has_upper_latin1_only_utf8_matches
17440 = add_cp_to_invlist(
17441 has_upper_latin1_only_utf8_matches,
17448 SvREFCNT_dec_NN(fold_intersection);
17451 /* Now that we have finished adding all the folds, there is no reason
17452 * to keep the foldable list separate */
17453 _invlist_union(cp_list, cp_foldable_list, &cp_list);
17454 SvREFCNT_dec_NN(cp_foldable_list);
17457 /* And combine the result (if any) with any inversion lists from posix
17458 * classes. The lists are kept separate up to now because we don't want to
17459 * fold the classes (folding of those is automatically handled by the swash
17460 * fetching code) */
17461 if (simple_posixes) { /* These are the classes known to be unaffected by
17464 _invlist_union(cp_list, simple_posixes, &cp_list);
17465 SvREFCNT_dec_NN(simple_posixes);
17468 cp_list = simple_posixes;
17471 if (posixes || nposixes) {
17473 /* We have to adjust /a and /aa */
17474 if (AT_LEAST_ASCII_RESTRICTED) {
17476 /* Under /a and /aa, nothing above ASCII matches these */
17478 _invlist_intersection(posixes,
17479 PL_XPosix_ptrs[_CC_ASCII],
17483 /* Under /a and /aa, everything above ASCII matches these
17486 _invlist_union_complement_2nd(nposixes,
17487 PL_XPosix_ptrs[_CC_ASCII],
17492 if (! DEPENDS_SEMANTICS) {
17494 /* For everything but /d, we can just add the current 'posixes' and
17495 * 'nposixes' to the main list */
17498 _invlist_union(cp_list, posixes, &cp_list);
17499 SvREFCNT_dec_NN(posixes);
17507 _invlist_union(cp_list, nposixes, &cp_list);
17508 SvREFCNT_dec_NN(nposixes);
17511 cp_list = nposixes;
17516 /* Under /d, things like \w match upper Latin1 characters only if
17517 * the target string is in UTF-8. But things like \W match all the
17518 * upper Latin1 characters if the target string is not in UTF-8.
17520 * Handle the case where there something like \W separately */
17522 SV* only_non_utf8_list = invlist_clone(PL_UpperLatin1);
17524 /* A complemented posix class matches all upper Latin1
17525 * characters if not in UTF-8. And it matches just certain
17526 * ones when in UTF-8. That means those certain ones are
17527 * matched regardless, so can just be added to the
17528 * unconditional list */
17530 _invlist_union(cp_list, nposixes, &cp_list);
17531 SvREFCNT_dec_NN(nposixes);
17535 cp_list = nposixes;
17538 /* Likewise for 'posixes' */
17539 _invlist_union(posixes, cp_list, &cp_list);
17541 /* Likewise for anything else in the range that matched only
17543 if (has_upper_latin1_only_utf8_matches) {
17544 _invlist_union(cp_list,
17545 has_upper_latin1_only_utf8_matches,
17547 SvREFCNT_dec_NN(has_upper_latin1_only_utf8_matches);
17548 has_upper_latin1_only_utf8_matches = NULL;
17551 /* If we don't match all the upper Latin1 characters regardless
17552 * of UTF-8ness, we have to set a flag to match the rest when
17554 _invlist_subtract(only_non_utf8_list, cp_list,
17555 &only_non_utf8_list);
17556 if (_invlist_len(only_non_utf8_list) != 0) {
17557 ANYOF_FLAGS(ret) |= ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER;
17561 /* Here there were no complemented posix classes. That means
17562 * the upper Latin1 characters in 'posixes' match only when the
17563 * target string is in UTF-8. So we have to add them to the
17564 * list of those types of code points, while adding the
17565 * remainder to the unconditional list.
17567 * First calculate what they are */
17568 SV* nonascii_but_latin1_properties = NULL;
17569 _invlist_intersection(posixes, PL_UpperLatin1,
17570 &nonascii_but_latin1_properties);
17572 /* And add them to the final list of such characters. */
17573 _invlist_union(has_upper_latin1_only_utf8_matches,
17574 nonascii_but_latin1_properties,
17575 &has_upper_latin1_only_utf8_matches);
17577 /* Remove them from what now becomes the unconditional list */
17578 _invlist_subtract(posixes, nonascii_but_latin1_properties,
17581 /* And add those unconditional ones to the final list */
17583 _invlist_union(cp_list, posixes, &cp_list);
17584 SvREFCNT_dec_NN(posixes);
17591 SvREFCNT_dec(nonascii_but_latin1_properties);
17593 /* Get rid of any characters that we now know are matched
17594 * unconditionally from the conditional list, which may make
17595 * that list empty */
17596 _invlist_subtract(has_upper_latin1_only_utf8_matches,
17598 &has_upper_latin1_only_utf8_matches);
17599 if (_invlist_len(has_upper_latin1_only_utf8_matches) == 0) {
17600 SvREFCNT_dec_NN(has_upper_latin1_only_utf8_matches);
17601 has_upper_latin1_only_utf8_matches = NULL;
17607 /* And combine the result (if any) with any inversion list from properties.
17608 * The lists are kept separate up to now so that we can distinguish the two
17609 * in regards to matching above-Unicode. A run-time warning is generated
17610 * if a Unicode property is matched against a non-Unicode code point. But,
17611 * we allow user-defined properties to match anything, without any warning,
17612 * and we also suppress the warning if there is a portion of the character
17613 * class that isn't a Unicode property, and which matches above Unicode, \W
17614 * or [\x{110000}] for example.
17615 * (Note that in this case, unlike the Posix one above, there is no
17616 * <has_upper_latin1_only_utf8_matches>, because having a Unicode property
17617 * forces Unicode semantics */
17621 /* If it matters to the final outcome, see if a non-property
17622 * component of the class matches above Unicode. If so, the
17623 * warning gets suppressed. This is true even if just a single
17624 * such code point is specified, as, though not strictly correct if
17625 * another such code point is matched against, the fact that they
17626 * are using above-Unicode code points indicates they should know
17627 * the issues involved */
17629 warn_super = ! (invert
17630 ^ (invlist_highest(cp_list) > PERL_UNICODE_MAX));
17633 _invlist_union(properties, cp_list, &cp_list);
17634 SvREFCNT_dec_NN(properties);
17637 cp_list = properties;
17642 |= ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER;
17644 /* Because an ANYOF node is the only one that warns, this node
17645 * can't be optimized into something else */
17646 optimizable = FALSE;
17650 /* Here, we have calculated what code points should be in the character
17653 * Now we can see about various optimizations. Fold calculation (which we
17654 * did above) needs to take place before inversion. Otherwise /[^k]/i
17655 * would invert to include K, which under /i would match k, which it
17656 * shouldn't. Therefore we can't invert folded locale now, as it won't be
17657 * folded until runtime */
17659 /* If we didn't do folding, it's because some information isn't available
17660 * until runtime; set the run-time fold flag for these. (We don't have to
17661 * worry about properties folding, as that is taken care of by the swash
17662 * fetching). We know to set the flag if we have a non-NULL list for UTF-8
17663 * locales, or the class matches at least one 0-255 range code point */
17666 /* Some things on the list might be unconditionally included because of
17667 * other components. Remove them, and clean up the list if it goes to
17669 if (only_utf8_locale_list && cp_list) {
17670 _invlist_subtract(only_utf8_locale_list, cp_list,
17671 &only_utf8_locale_list);
17673 if (_invlist_len(only_utf8_locale_list) == 0) {
17674 SvREFCNT_dec_NN(only_utf8_locale_list);
17675 only_utf8_locale_list = NULL;
17678 if (only_utf8_locale_list) {
17681 |ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD;
17683 else if (cp_list) { /* Look to see if a 0-255 code point is in list */
17685 invlist_iterinit(cp_list);
17686 if (invlist_iternext(cp_list, &start, &end) && start < 256) {
17687 ANYOF_FLAGS(ret) |= ANYOFL_FOLD;
17689 invlist_iterfinish(cp_list);
17692 else if ( DEPENDS_SEMANTICS
17693 && ( has_upper_latin1_only_utf8_matches
17694 || (ANYOF_FLAGS(ret) & ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER)))
17697 optimizable = FALSE;
17701 /* Optimize inverted simple patterns (e.g. [^a-z]) when everything is known
17702 * at compile time. Besides not inverting folded locale now, we can't
17703 * invert if there are things such as \w, which aren't known until runtime
17707 && OP(ret) != ANYOFD
17708 && ! (ANYOF_FLAGS(ret) & (ANYOF_LOCALE_FLAGS))
17709 && ! HAS_NONLOCALE_RUNTIME_PROPERTY_DEFINITION)
17711 _invlist_invert(cp_list);
17713 /* Any swash can't be used as-is, because we've inverted things */
17715 SvREFCNT_dec_NN(swash);
17719 /* Clear the invert flag since have just done it here */
17726 *ret_invlist = cp_list;
17727 SvREFCNT_dec(swash);
17729 /* Discard the generated node */
17731 RExC_size = orig_size;
17734 RExC_emit = orig_emit;
17739 /* Some character classes are equivalent to other nodes. Such nodes take
17740 * up less room and generally fewer operations to execute than ANYOF nodes.
17741 * Above, we checked for and optimized into some such equivalents for
17742 * certain common classes that are easy to test. Getting to this point in
17743 * the code means that the class didn't get optimized there. Since this
17744 * code is only executed in Pass 2, it is too late to save space--it has
17745 * been allocated in Pass 1, and currently isn't given back. But turning
17746 * things into an EXACTish node can allow the optimizer to join it to any
17747 * adjacent such nodes. And if the class is equivalent to things like /./,
17748 * expensive run-time swashes can be avoided. Now that we have more
17749 * complete information, we can find things necessarily missed by the
17750 * earlier code. Another possible "optimization" that isn't done is that
17751 * something like [Ee] could be changed into an EXACTFU. khw tried this
17752 * and found that the ANYOF is faster, including for code points not in the
17753 * bitmap. This still might make sense to do, provided it got joined with
17754 * an adjacent node(s) to create a longer EXACTFU one. This could be
17755 * accomplished by creating a pseudo ANYOF_EXACTFU node type that the join
17756 * routine would know is joinable. If that didn't happen, the node type
17757 * could then be made a straight ANYOF */
17759 if (optimizable && cp_list && ! invert) {
17761 U8 op = END; /* The optimzation node-type */
17762 int posix_class = -1; /* Illegal value */
17763 const char * cur_parse= RExC_parse;
17765 invlist_iterinit(cp_list);
17766 if (! invlist_iternext(cp_list, &start, &end)) {
17768 /* Here, the list is empty. This happens, for example, when a
17769 * Unicode property that doesn't match anything is the only element
17770 * in the character class (perluniprops.pod notes such properties).
17773 *flagp |= HASWIDTH|SIMPLE;
17775 else if (start == end) { /* The range is a single code point */
17776 if (! invlist_iternext(cp_list, &start, &end)
17778 /* Don't do this optimization if it would require changing
17779 * the pattern to UTF-8 */
17780 && (start < 256 || UTF))
17782 /* Here, the list contains a single code point. Can optimize
17783 * into an EXACTish node */
17794 /* A locale node under folding with one code point can be
17795 * an EXACTFL, as its fold won't be calculated until
17801 /* Here, we are generally folding, but there is only one
17802 * code point to match. If we have to, we use an EXACT
17803 * node, but it would be better for joining with adjacent
17804 * nodes in the optimization pass if we used the same
17805 * EXACTFish node that any such are likely to be. We can
17806 * do this iff the code point doesn't participate in any
17807 * folds. For example, an EXACTF of a colon is the same as
17808 * an EXACT one, since nothing folds to or from a colon. */
17810 if (IS_IN_SOME_FOLD_L1(value)) {
17815 if (_invlist_contains_cp(PL_utf8_foldable, value)) {
17820 /* If we haven't found the node type, above, it means we
17821 * can use the prevailing one */
17823 op = compute_EXACTish(pRExC_state);
17827 } /* End of first range contains just a single code point */
17828 else if (start == 0) {
17829 if (end == UV_MAX) {
17831 *flagp |= HASWIDTH|SIMPLE;
17834 else if (end == '\n' - 1
17835 && invlist_iternext(cp_list, &start, &end)
17836 && start == '\n' + 1 && end == UV_MAX)
17839 *flagp |= HASWIDTH|SIMPLE;
17843 invlist_iterfinish(cp_list);
17846 const UV cp_list_len = _invlist_len(cp_list);
17847 const UV* cp_list_array = invlist_array(cp_list);
17849 /* Here, didn't find an optimization. See if this matches any of
17850 * the POSIX classes. These run slightly faster for above-Unicode
17851 * code points, so don't bother with POSIXA ones nor the 2 that
17852 * have no above-Unicode matches. We can avoid these checks unless
17853 * the ANYOF matches at least as high as the lowest POSIX one
17854 * (which was manually found to be \v. The actual code point may
17855 * increase in later Unicode releases, if a higher code point is
17856 * assigned to be \v, but this code will never break. It would
17857 * just mean we could execute the checks for posix optimizations
17858 * unnecessarily) */
17860 if (cp_list_array[cp_list_len-1] > 0x2029) {
17861 for (posix_class = 0;
17862 posix_class <= _HIGHEST_REGCOMP_DOT_H_SYNC;
17866 if (posix_class == _CC_ASCII || posix_class == _CC_CNTRL) {
17869 for (try_inverted = 0; try_inverted < 2; try_inverted++) {
17871 /* Check if matches normal or inverted */
17872 if (_invlistEQ(cp_list,
17873 PL_XPosix_ptrs[posix_class],
17876 op = (try_inverted)
17879 *flagp |= HASWIDTH|SIMPLE;
17889 RExC_parse = (char *)orig_parse;
17890 RExC_emit = (regnode *)orig_emit;
17892 if (regarglen[op]) {
17893 ret = reganode(pRExC_state, op, 0);
17895 ret = reg_node(pRExC_state, op);
17898 RExC_parse = (char *)cur_parse;
17900 if (PL_regkind[op] == EXACT) {
17901 alloc_maybe_populate_EXACT(pRExC_state, ret, flagp, 0, value,
17902 TRUE /* downgradable to EXACT */
17905 else if (PL_regkind[op] == POSIXD || PL_regkind[op] == NPOSIXD) {
17906 FLAGS(ret) = posix_class;
17909 SvREFCNT_dec_NN(cp_list);
17914 /* Here, <cp_list> contains all the code points we can determine at
17915 * compile time that match under all conditions. Go through it, and
17916 * for things that belong in the bitmap, put them there, and delete from
17917 * <cp_list>. While we are at it, see if everything above 255 is in the
17918 * list, and if so, set a flag to speed up execution */
17920 populate_ANYOF_from_invlist(ret, &cp_list);
17923 ANYOF_FLAGS(ret) |= ANYOF_INVERT;
17926 /* Here, the bitmap has been populated with all the Latin1 code points that
17927 * always match. Can now add to the overall list those that match only
17928 * when the target string is UTF-8 (<has_upper_latin1_only_utf8_matches>).
17930 if (has_upper_latin1_only_utf8_matches) {
17932 _invlist_union(cp_list,
17933 has_upper_latin1_only_utf8_matches,
17935 SvREFCNT_dec_NN(has_upper_latin1_only_utf8_matches);
17938 cp_list = has_upper_latin1_only_utf8_matches;
17940 ANYOF_FLAGS(ret) |= ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP;
17943 /* If there is a swash and more than one element, we can't use the swash in
17944 * the optimization below. */
17945 if (swash && element_count > 1) {
17946 SvREFCNT_dec_NN(swash);
17950 /* Note that the optimization of using 'swash' if it is the only thing in
17951 * the class doesn't have us change swash at all, so it can include things
17952 * that are also in the bitmap; otherwise we have purposely deleted that
17953 * duplicate information */
17954 set_ANYOF_arg(pRExC_state, ret, cp_list,
17955 (HAS_NONLOCALE_RUNTIME_PROPERTY_DEFINITION)
17957 only_utf8_locale_list,
17958 swash, has_user_defined_property);
17960 *flagp |= HASWIDTH|SIMPLE;
17962 if (ANYOF_FLAGS(ret) & ANYOF_LOCALE_FLAGS) {
17963 RExC_contains_locale = 1;
17969 #undef HAS_NONLOCALE_RUNTIME_PROPERTY_DEFINITION
17972 S_set_ANYOF_arg(pTHX_ RExC_state_t* const pRExC_state,
17973 regnode* const node,
17975 SV* const runtime_defns,
17976 SV* const only_utf8_locale_list,
17978 const bool has_user_defined_property)
17980 /* Sets the arg field of an ANYOF-type node 'node', using information about
17981 * the node passed-in. If there is nothing outside the node's bitmap, the
17982 * arg is set to ANYOF_ONLY_HAS_BITMAP. Otherwise, it sets the argument to
17983 * the count returned by add_data(), having allocated and stored an array,
17984 * av, that that count references, as follows:
17985 * av[0] stores the character class description in its textual form.
17986 * This is used later (regexec.c:Perl_regclass_swash()) to
17987 * initialize the appropriate swash, and is also useful for dumping
17988 * the regnode. This is set to &PL_sv_undef if the textual
17989 * description is not needed at run-time (as happens if the other
17990 * elements completely define the class)
17991 * av[1] if &PL_sv_undef, is a placeholder to later contain the swash
17992 * computed from av[0]. But if no further computation need be done,
17993 * the swash is stored here now (and av[0] is &PL_sv_undef).
17994 * av[2] stores the inversion list of code points that match only if the
17995 * current locale is UTF-8
17996 * av[3] stores the cp_list inversion list for use in addition or instead
17997 * of av[0]; used only if cp_list exists and av[1] is &PL_sv_undef.
17998 * (Otherwise everything needed is already in av[0] and av[1])
17999 * av[4] is set if any component of the class is from a user-defined
18000 * property; used only if av[3] exists */
18004 PERL_ARGS_ASSERT_SET_ANYOF_ARG;
18006 if (! cp_list && ! runtime_defns && ! only_utf8_locale_list) {
18007 assert(! (ANYOF_FLAGS(node)
18008 & ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP));
18009 ARG_SET(node, ANYOF_ONLY_HAS_BITMAP);
18012 AV * const av = newAV();
18015 av_store(av, 0, (runtime_defns)
18016 ? SvREFCNT_inc(runtime_defns) : &PL_sv_undef);
18019 av_store(av, 1, swash);
18020 SvREFCNT_dec_NN(cp_list);
18023 av_store(av, 1, &PL_sv_undef);
18025 av_store(av, 3, cp_list);
18026 av_store(av, 4, newSVuv(has_user_defined_property));
18030 if (only_utf8_locale_list) {
18031 av_store(av, 2, only_utf8_locale_list);
18034 av_store(av, 2, &PL_sv_undef);
18037 rv = newRV_noinc(MUTABLE_SV(av));
18038 n = add_data(pRExC_state, STR_WITH_LEN("s"));
18039 RExC_rxi->data->data[n] = (void*)rv;
18044 #if !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION)
18046 Perl__get_regclass_nonbitmap_data(pTHX_ const regexp *prog,
18047 const regnode* node,
18050 SV** only_utf8_locale_ptr,
18051 SV** output_invlist)
18054 /* For internal core use only.
18055 * Returns the swash for the input 'node' in the regex 'prog'.
18056 * If <doinit> is 'true', will attempt to create the swash if not already
18058 * If <listsvp> is non-null, will return the printable contents of the
18059 * swash. This can be used to get debugging information even before the
18060 * swash exists, by calling this function with 'doinit' set to false, in
18061 * which case the components that will be used to eventually create the
18062 * swash are returned (in a printable form).
18063 * If <only_utf8_locale_ptr> is not NULL, it is where this routine is to
18064 * store an inversion list of code points that should match only if the
18065 * execution-time locale is a UTF-8 one.
18066 * If <output_invlist> is not NULL, it is where this routine is to store an
18067 * inversion list of the code points that would be instead returned in
18068 * <listsvp> if this were NULL. Thus, what gets output in <listsvp>
18069 * when this parameter is used, is just the non-code point data that
18070 * will go into creating the swash. This currently should be just
18071 * user-defined properties whose definitions were not known at compile
18072 * time. Using this parameter allows for easier manipulation of the
18073 * swash's data by the caller. It is illegal to call this function with
18074 * this parameter set, but not <listsvp>
18076 * Tied intimately to how S_set_ANYOF_arg sets up the data structure. Note
18077 * that, in spite of this function's name, the swash it returns may include
18078 * the bitmap data as well */
18081 SV *si = NULL; /* Input swash initialization string */
18082 SV* invlist = NULL;
18084 RXi_GET_DECL(prog,progi);
18085 const struct reg_data * const data = prog ? progi->data : NULL;
18087 PERL_ARGS_ASSERT__GET_REGCLASS_NONBITMAP_DATA;
18088 assert(! output_invlist || listsvp);
18090 if (data && data->count) {
18091 const U32 n = ARG(node);
18093 if (data->what[n] == 's') {
18094 SV * const rv = MUTABLE_SV(data->data[n]);
18095 AV * const av = MUTABLE_AV(SvRV(rv));
18096 SV **const ary = AvARRAY(av);
18097 U8 swash_init_flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
18099 si = *ary; /* ary[0] = the string to initialize the swash with */
18101 if (av_tindex_skip_len_mg(av) >= 2) {
18102 if (only_utf8_locale_ptr
18104 && ary[2] != &PL_sv_undef)
18106 *only_utf8_locale_ptr = ary[2];
18109 assert(only_utf8_locale_ptr);
18110 *only_utf8_locale_ptr = NULL;
18113 /* Elements 3 and 4 are either both present or both absent. [3]
18114 * is any inversion list generated at compile time; [4]
18115 * indicates if that inversion list has any user-defined
18116 * properties in it. */
18117 if (av_tindex_skip_len_mg(av) >= 3) {
18119 if (SvUV(ary[4])) {
18120 swash_init_flags |= _CORE_SWASH_INIT_USER_DEFINED_PROPERTY;
18128 /* Element [1] is reserved for the set-up swash. If already there,
18129 * return it; if not, create it and store it there */
18130 if (ary[1] && SvROK(ary[1])) {
18133 else if (doinit && ((si && si != &PL_sv_undef)
18134 || (invlist && invlist != &PL_sv_undef))) {
18136 sw = _core_swash_init("utf8", /* the utf8 package */
18140 0, /* not from tr/// */
18142 &swash_init_flags);
18143 (void)av_store(av, 1, sw);
18148 /* If requested, return a printable version of what this swash matches */
18150 SV* matches_string = NULL;
18152 /* The swash should be used, if possible, to get the data, as it
18153 * contains the resolved data. But this function can be called at
18154 * compile-time, before everything gets resolved, in which case we
18155 * return the currently best available information, which is the string
18156 * that will eventually be used to do that resolving, 'si' */
18157 if ((! sw || (invlist = _get_swash_invlist(sw)) == NULL)
18158 && (si && si != &PL_sv_undef))
18160 /* Here, we only have 'si' (and possibly some passed-in data in
18161 * 'invlist', which is handled below) If the caller only wants
18162 * 'si', use that. */
18163 if (! output_invlist) {
18164 matches_string = newSVsv(si);
18167 /* But if the caller wants an inversion list of the node, we
18168 * need to parse 'si' and place as much as possible in the
18169 * desired output inversion list, making 'matches_string' only
18170 * contain the currently unresolvable things */
18171 const char *si_string = SvPVX(si);
18172 STRLEN remaining = SvCUR(si);
18176 /* Ignore everything before the first new-line */
18177 while (*si_string != '\n' && remaining > 0) {
18181 assert(remaining > 0);
18186 while (remaining > 0) {
18188 /* The data consists of just strings defining user-defined
18189 * property names, but in prior incarnations, and perhaps
18190 * somehow from pluggable regex engines, it could still
18191 * hold hex code point definitions. Each component of a
18192 * range would be separated by a tab, and each range by a
18193 * new-line. If these are found, instead add them to the
18194 * inversion list */
18195 I32 grok_flags = PERL_SCAN_SILENT_ILLDIGIT
18196 |PERL_SCAN_SILENT_NON_PORTABLE;
18197 STRLEN len = remaining;
18198 UV cp = grok_hex(si_string, &len, &grok_flags, NULL);
18200 /* If the hex decode routine found something, it should go
18201 * up to the next \n */
18202 if ( *(si_string + len) == '\n') {
18203 if (count) { /* 2nd code point on line */
18204 *output_invlist = _add_range_to_invlist(*output_invlist, prev_cp, cp);
18207 *output_invlist = add_cp_to_invlist(*output_invlist, cp);
18210 goto prepare_for_next_iteration;
18213 /* If the hex decode was instead for the lower range limit,
18214 * save it, and go parse the upper range limit */
18215 if (*(si_string + len) == '\t') {
18216 assert(count == 0);
18220 prepare_for_next_iteration:
18221 si_string += len + 1;
18222 remaining -= len + 1;
18226 /* Here, didn't find a legal hex number. Just add it from
18227 * here to the next \n */
18230 while (*(si_string + len) != '\n' && remaining > 0) {
18234 if (*(si_string + len) == '\n') {
18238 if (matches_string) {
18239 sv_catpvn(matches_string, si_string, len - 1);
18242 matches_string = newSVpvn(si_string, len - 1);
18245 sv_catpvs(matches_string, " ");
18246 } /* end of loop through the text */
18248 assert(matches_string);
18249 if (SvCUR(matches_string)) { /* Get rid of trailing blank */
18250 SvCUR_set(matches_string, SvCUR(matches_string) - 1);
18252 } /* end of has an 'si' but no swash */
18255 /* If we have a swash in place, its equivalent inversion list was above
18256 * placed into 'invlist'. If not, this variable may contain a stored
18257 * inversion list which is information beyond what is in 'si' */
18260 /* Again, if the caller doesn't want the output inversion list, put
18261 * everything in 'matches-string' */
18262 if (! output_invlist) {
18263 if ( ! matches_string) {
18264 matches_string = newSVpvs("\n");
18266 sv_catsv(matches_string, invlist_contents(invlist,
18267 TRUE /* traditional style */
18270 else if (! *output_invlist) {
18271 *output_invlist = invlist_clone(invlist);
18274 _invlist_union(*output_invlist, invlist, output_invlist);
18278 *listsvp = matches_string;
18283 #endif /* !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION) */
18285 /* reg_skipcomment()
18287 Absorbs an /x style # comment from the input stream,
18288 returning a pointer to the first character beyond the comment, or if the
18289 comment terminates the pattern without anything following it, this returns
18290 one past the final character of the pattern (in other words, RExC_end) and
18291 sets the REG_RUN_ON_COMMENT_SEEN flag.
18293 Note it's the callers responsibility to ensure that we are
18294 actually in /x mode
18298 PERL_STATIC_INLINE char*
18299 S_reg_skipcomment(RExC_state_t *pRExC_state, char* p)
18301 PERL_ARGS_ASSERT_REG_SKIPCOMMENT;
18305 while (p < RExC_end) {
18306 if (*(++p) == '\n') {
18311 /* we ran off the end of the pattern without ending the comment, so we have
18312 * to add an \n when wrapping */
18313 RExC_seen |= REG_RUN_ON_COMMENT_SEEN;
18318 S_skip_to_be_ignored_text(pTHX_ RExC_state_t *pRExC_state,
18320 const bool force_to_xmod
18323 /* If the text at the current parse position '*p' is a '(?#...)' comment,
18324 * or if we are under /x or 'force_to_xmod' is TRUE, and the text at '*p'
18325 * is /x whitespace, advance '*p' so that on exit it points to the first
18326 * byte past all such white space and comments */
18328 const bool use_xmod = force_to_xmod || (RExC_flags & RXf_PMf_EXTENDED);
18330 PERL_ARGS_ASSERT_SKIP_TO_BE_IGNORED_TEXT;
18332 assert( ! UTF || UTF8_IS_INVARIANT(**p) || UTF8_IS_START(**p));
18335 if (RExC_end - (*p) >= 3
18337 && *(*p + 1) == '?'
18338 && *(*p + 2) == '#')
18340 while (*(*p) != ')') {
18341 if ((*p) == RExC_end)
18342 FAIL("Sequence (?#... not terminated");
18350 const char * save_p = *p;
18351 while ((*p) < RExC_end) {
18353 if ((len = is_PATWS_safe((*p), RExC_end, UTF))) {
18356 else if (*(*p) == '#') {
18357 (*p) = reg_skipcomment(pRExC_state, (*p));
18363 if (*p != save_p) {
18376 Advances the parse position by one byte, unless that byte is the beginning
18377 of a '(?#...)' style comment, or is /x whitespace and /x is in effect. In
18378 those two cases, the parse position is advanced beyond all such comments and
18381 This is the UTF, (?#...), and /x friendly way of saying RExC_parse++.
18385 S_nextchar(pTHX_ RExC_state_t *pRExC_state)
18387 PERL_ARGS_ASSERT_NEXTCHAR;
18389 if (RExC_parse < RExC_end) {
18391 || UTF8_IS_INVARIANT(*RExC_parse)
18392 || UTF8_IS_START(*RExC_parse));
18394 RExC_parse += (UTF) ? UTF8SKIP(RExC_parse) : 1;
18396 skip_to_be_ignored_text(pRExC_state, &RExC_parse,
18397 FALSE /* Don't force /x */ );
18402 S_regnode_guts(pTHX_ RExC_state_t *pRExC_state, const U8 op, const STRLEN extra_size, const char* const name)
18404 /* Allocate a regnode for 'op' and returns it, with 'extra_size' extra
18405 * space. In pass1, it aligns and increments RExC_size; in pass2,
18408 regnode * const ret = RExC_emit;
18409 GET_RE_DEBUG_FLAGS_DECL;
18411 PERL_ARGS_ASSERT_REGNODE_GUTS;
18413 assert(extra_size >= regarglen[op]);
18416 SIZE_ALIGN(RExC_size);
18417 RExC_size += 1 + extra_size;
18420 if (RExC_emit >= RExC_emit_bound)
18421 Perl_croak(aTHX_ "panic: reg_node overrun trying to emit %d, %p>=%p",
18422 op, (void*)RExC_emit, (void*)RExC_emit_bound);
18424 NODE_ALIGN_FILL(ret);
18425 #ifndef RE_TRACK_PATTERN_OFFSETS
18426 PERL_UNUSED_ARG(name);
18428 if (RExC_offsets) { /* MJD */
18430 ("%s:%d: (op %s) %s %" UVuf " (len %" UVuf ") (max %" UVuf ").\n",
18433 (UV)(RExC_emit - RExC_emit_start) > RExC_offsets[0]
18434 ? "Overwriting end of array!\n" : "OK",
18435 (UV)(RExC_emit - RExC_emit_start),
18436 (UV)(RExC_parse - RExC_start),
18437 (UV)RExC_offsets[0]));
18438 Set_Node_Offset(RExC_emit, RExC_parse + (op == END));
18445 - reg_node - emit a node
18447 STATIC regnode * /* Location. */
18448 S_reg_node(pTHX_ RExC_state_t *pRExC_state, U8 op)
18450 regnode * const ret = regnode_guts(pRExC_state, op, regarglen[op], "reg_node");
18452 PERL_ARGS_ASSERT_REG_NODE;
18454 assert(regarglen[op] == 0);
18457 regnode *ptr = ret;
18458 FILL_ADVANCE_NODE(ptr, op);
18465 - reganode - emit a node with an argument
18467 STATIC regnode * /* Location. */
18468 S_reganode(pTHX_ RExC_state_t *pRExC_state, U8 op, U32 arg)
18470 regnode * const ret = regnode_guts(pRExC_state, op, regarglen[op], "reganode");
18472 PERL_ARGS_ASSERT_REGANODE;
18474 assert(regarglen[op] == 1);
18477 regnode *ptr = ret;
18478 FILL_ADVANCE_NODE_ARG(ptr, op, arg);
18485 S_reg2Lanode(pTHX_ RExC_state_t *pRExC_state, const U8 op, const U32 arg1, const I32 arg2)
18487 /* emit a node with U32 and I32 arguments */
18489 regnode * const ret = regnode_guts(pRExC_state, op, regarglen[op], "reg2Lanode");
18491 PERL_ARGS_ASSERT_REG2LANODE;
18493 assert(regarglen[op] == 2);
18496 regnode *ptr = ret;
18497 FILL_ADVANCE_NODE_2L_ARG(ptr, op, arg1, arg2);
18504 - reginsert - insert an operator in front of already-emitted operand
18506 * Means relocating the operand.
18508 * IMPORTANT NOTE - it is the *callers* responsibility to correctly
18509 * set up NEXT_OFF() of the inserted node if needed. Something like this:
18511 * reginsert(pRExC, OPFAIL, orig_emit, depth+1);
18513 * NEXT_OFF(orig_emit) = regarglen[OPFAIL] + NODE_STEP_REGNODE;
18517 S_reginsert(pTHX_ RExC_state_t *pRExC_state, U8 op, regnode *operand, U32 depth)
18522 const int offset = regarglen[(U8)op];
18523 const int size = NODE_STEP_REGNODE + offset;
18524 GET_RE_DEBUG_FLAGS_DECL;
18526 PERL_ARGS_ASSERT_REGINSERT;
18527 PERL_UNUSED_CONTEXT;
18528 PERL_UNUSED_ARG(depth);
18529 /* (PL_regkind[(U8)op] == CURLY ? EXTRA_STEP_2ARGS : 0); */
18530 DEBUG_PARSE_FMT("inst"," - %s",PL_reg_name[op]);
18535 assert(!RExC_study_started); /* I believe we should never use reginsert once we have started
18536 studying. If this is wrong then we need to adjust RExC_recurse
18537 below like we do with RExC_open_parens/RExC_close_parens. */
18541 if (RExC_open_parens) {
18543 /*DEBUG_PARSE_FMT("inst"," - %" IVdf, (IV)RExC_npar);*/
18544 /* remember that RExC_npar is rex->nparens + 1,
18545 * iow it is 1 more than the number of parens seen in
18546 * the pattern so far. */
18547 for ( paren=0 ; paren < RExC_npar ; paren++ ) {
18548 /* note, RExC_open_parens[0] is the start of the
18549 * regex, it can't move. RExC_close_parens[0] is the end
18550 * of the regex, it *can* move. */
18551 if ( paren && RExC_open_parens[paren] >= operand ) {
18552 /*DEBUG_PARSE_FMT("open"," - %d",size);*/
18553 RExC_open_parens[paren] += size;
18555 /*DEBUG_PARSE_FMT("open"," - %s","ok");*/
18557 if ( RExC_close_parens[paren] >= operand ) {
18558 /*DEBUG_PARSE_FMT("close"," - %d",size);*/
18559 RExC_close_parens[paren] += size;
18561 /*DEBUG_PARSE_FMT("close"," - %s","ok");*/
18566 RExC_end_op += size;
18568 while (src > operand) {
18569 StructCopy(--src, --dst, regnode);
18570 #ifdef RE_TRACK_PATTERN_OFFSETS
18571 if (RExC_offsets) { /* MJD 20010112 */
18573 ("%s(%d): (op %s) %s copy %" UVuf " -> %" UVuf " (max %" UVuf ").\n",
18577 (UV)(dst - RExC_emit_start) > RExC_offsets[0]
18578 ? "Overwriting end of array!\n" : "OK",
18579 (UV)(src - RExC_emit_start),
18580 (UV)(dst - RExC_emit_start),
18581 (UV)RExC_offsets[0]));
18582 Set_Node_Offset_To_R(dst-RExC_emit_start, Node_Offset(src));
18583 Set_Node_Length_To_R(dst-RExC_emit_start, Node_Length(src));
18589 place = operand; /* Op node, where operand used to be. */
18590 #ifdef RE_TRACK_PATTERN_OFFSETS
18591 if (RExC_offsets) { /* MJD */
18593 ("%s(%d): (op %s) %s %" UVuf " <- %" UVuf " (max %" UVuf ").\n",
18597 (UV)(place - RExC_emit_start) > RExC_offsets[0]
18598 ? "Overwriting end of array!\n" : "OK",
18599 (UV)(place - RExC_emit_start),
18600 (UV)(RExC_parse - RExC_start),
18601 (UV)RExC_offsets[0]));
18602 Set_Node_Offset(place, RExC_parse);
18603 Set_Node_Length(place, 1);
18606 src = NEXTOPER(place);
18607 FILL_ADVANCE_NODE(place, op);
18608 Zero(src, offset, regnode);
18612 - regtail - set the next-pointer at the end of a node chain of p to val.
18613 - SEE ALSO: regtail_study
18616 S_regtail(pTHX_ RExC_state_t * pRExC_state,
18617 const regnode * const p,
18618 const regnode * const val,
18622 GET_RE_DEBUG_FLAGS_DECL;
18624 PERL_ARGS_ASSERT_REGTAIL;
18626 PERL_UNUSED_ARG(depth);
18632 /* Find last node. */
18633 scan = (regnode *) p;
18635 regnode * const temp = regnext(scan);
18637 DEBUG_PARSE_MSG((scan==p ? "tail" : ""));
18638 regprop(RExC_rx, RExC_mysv, scan, NULL, pRExC_state);
18639 Perl_re_printf( aTHX_ "~ %s (%d) %s %s\n",
18640 SvPV_nolen_const(RExC_mysv), REG_NODE_NUM(scan),
18641 (temp == NULL ? "->" : ""),
18642 (temp == NULL ? PL_reg_name[OP(val)] : "")
18650 if (reg_off_by_arg[OP(scan)]) {
18651 ARG_SET(scan, val - scan);
18654 NEXT_OFF(scan) = val - scan;
18660 - regtail_study - set the next-pointer at the end of a node chain of p to val.
18661 - Look for optimizable sequences at the same time.
18662 - currently only looks for EXACT chains.
18664 This is experimental code. The idea is to use this routine to perform
18665 in place optimizations on branches and groups as they are constructed,
18666 with the long term intention of removing optimization from study_chunk so
18667 that it is purely analytical.
18669 Currently only used when in DEBUG mode. The macro REGTAIL_STUDY() is used
18670 to control which is which.
18673 /* TODO: All four parms should be const */
18676 S_regtail_study(pTHX_ RExC_state_t *pRExC_state, regnode *p,
18677 const regnode *val,U32 depth)
18681 #ifdef EXPERIMENTAL_INPLACESCAN
18684 GET_RE_DEBUG_FLAGS_DECL;
18686 PERL_ARGS_ASSERT_REGTAIL_STUDY;
18692 /* Find last node. */
18696 regnode * const temp = regnext(scan);
18697 #ifdef EXPERIMENTAL_INPLACESCAN
18698 if (PL_regkind[OP(scan)] == EXACT) {
18699 bool unfolded_multi_char; /* Unexamined in this routine */
18700 if (join_exact(pRExC_state, scan, &min,
18701 &unfolded_multi_char, 1, val, depth+1))
18706 switch (OP(scan)) {
18710 case EXACTFA_NO_TRIE:
18716 if( exact == PSEUDO )
18718 else if ( exact != OP(scan) )
18727 DEBUG_PARSE_MSG((scan==p ? "tsdy" : ""));
18728 regprop(RExC_rx, RExC_mysv, scan, NULL, pRExC_state);
18729 Perl_re_printf( aTHX_ "~ %s (%d) -> %s\n",
18730 SvPV_nolen_const(RExC_mysv),
18731 REG_NODE_NUM(scan),
18732 PL_reg_name[exact]);
18739 DEBUG_PARSE_MSG("");
18740 regprop(RExC_rx, RExC_mysv, val, NULL, pRExC_state);
18741 Perl_re_printf( aTHX_
18742 "~ attach to %s (%" IVdf ") offset to %" IVdf "\n",
18743 SvPV_nolen_const(RExC_mysv),
18744 (IV)REG_NODE_NUM(val),
18748 if (reg_off_by_arg[OP(scan)]) {
18749 ARG_SET(scan, val - scan);
18752 NEXT_OFF(scan) = val - scan;
18760 - regdump - dump a regexp onto Perl_debug_log in vaguely comprehensible form
18765 S_regdump_intflags(pTHX_ const char *lead, const U32 flags)
18770 ASSUME(REG_INTFLAGS_NAME_SIZE <= sizeof(flags)*8);
18772 for (bit=0; bit<REG_INTFLAGS_NAME_SIZE; bit++) {
18773 if (flags & (1<<bit)) {
18774 if (!set++ && lead)
18775 Perl_re_printf( aTHX_ "%s",lead);
18776 Perl_re_printf( aTHX_ "%s ",PL_reg_intflags_name[bit]);
18781 Perl_re_printf( aTHX_ "\n");
18783 Perl_re_printf( aTHX_ "%s[none-set]\n",lead);
18788 S_regdump_extflags(pTHX_ const char *lead, const U32 flags)
18794 ASSUME(REG_EXTFLAGS_NAME_SIZE <= sizeof(flags)*8);
18796 for (bit=0; bit<REG_EXTFLAGS_NAME_SIZE; bit++) {
18797 if (flags & (1<<bit)) {
18798 if ((1<<bit) & RXf_PMf_CHARSET) { /* Output separately, below */
18801 if (!set++ && lead)
18802 Perl_re_printf( aTHX_ "%s",lead);
18803 Perl_re_printf( aTHX_ "%s ",PL_reg_extflags_name[bit]);
18806 if ((cs = get_regex_charset(flags)) != REGEX_DEPENDS_CHARSET) {
18807 if (!set++ && lead) {
18808 Perl_re_printf( aTHX_ "%s",lead);
18811 case REGEX_UNICODE_CHARSET:
18812 Perl_re_printf( aTHX_ "UNICODE");
18814 case REGEX_LOCALE_CHARSET:
18815 Perl_re_printf( aTHX_ "LOCALE");
18817 case REGEX_ASCII_RESTRICTED_CHARSET:
18818 Perl_re_printf( aTHX_ "ASCII-RESTRICTED");
18820 case REGEX_ASCII_MORE_RESTRICTED_CHARSET:
18821 Perl_re_printf( aTHX_ "ASCII-MORE_RESTRICTED");
18824 Perl_re_printf( aTHX_ "UNKNOWN CHARACTER SET");
18830 Perl_re_printf( aTHX_ "\n");
18832 Perl_re_printf( aTHX_ "%s[none-set]\n",lead);
18838 Perl_regdump(pTHX_ const regexp *r)
18841 SV * const sv = sv_newmortal();
18842 SV *dsv= sv_newmortal();
18843 RXi_GET_DECL(r,ri);
18844 GET_RE_DEBUG_FLAGS_DECL;
18846 PERL_ARGS_ASSERT_REGDUMP;
18848 (void)dumpuntil(r, ri->program, ri->program + 1, NULL, NULL, sv, 0, 0);
18850 /* Header fields of interest. */
18851 if (r->anchored_substr) {
18852 RE_PV_QUOTED_DECL(s, 0, dsv, SvPVX_const(r->anchored_substr),
18853 RE_SV_DUMPLEN(r->anchored_substr), 30);
18854 Perl_re_printf( aTHX_
18855 "anchored %s%s at %" IVdf " ",
18856 s, RE_SV_TAIL(r->anchored_substr),
18857 (IV)r->anchored_offset);
18858 } else if (r->anchored_utf8) {
18859 RE_PV_QUOTED_DECL(s, 1, dsv, SvPVX_const(r->anchored_utf8),
18860 RE_SV_DUMPLEN(r->anchored_utf8), 30);
18861 Perl_re_printf( aTHX_
18862 "anchored utf8 %s%s at %" IVdf " ",
18863 s, RE_SV_TAIL(r->anchored_utf8),
18864 (IV)r->anchored_offset);
18866 if (r->float_substr) {
18867 RE_PV_QUOTED_DECL(s, 0, dsv, SvPVX_const(r->float_substr),
18868 RE_SV_DUMPLEN(r->float_substr), 30);
18869 Perl_re_printf( aTHX_
18870 "floating %s%s at %" IVdf "..%" UVuf " ",
18871 s, RE_SV_TAIL(r->float_substr),
18872 (IV)r->float_min_offset, (UV)r->float_max_offset);
18873 } else if (r->float_utf8) {
18874 RE_PV_QUOTED_DECL(s, 1, dsv, SvPVX_const(r->float_utf8),
18875 RE_SV_DUMPLEN(r->float_utf8), 30);
18876 Perl_re_printf( aTHX_
18877 "floating utf8 %s%s at %" IVdf "..%" UVuf " ",
18878 s, RE_SV_TAIL(r->float_utf8),
18879 (IV)r->float_min_offset, (UV)r->float_max_offset);
18881 if (r->check_substr || r->check_utf8)
18882 Perl_re_printf( aTHX_
18884 (r->check_substr == r->float_substr
18885 && r->check_utf8 == r->float_utf8
18886 ? "(checking floating" : "(checking anchored"));
18887 if (r->intflags & PREGf_NOSCAN)
18888 Perl_re_printf( aTHX_ " noscan");
18889 if (r->extflags & RXf_CHECK_ALL)
18890 Perl_re_printf( aTHX_ " isall");
18891 if (r->check_substr || r->check_utf8)
18892 Perl_re_printf( aTHX_ ") ");
18894 if (ri->regstclass) {
18895 regprop(r, sv, ri->regstclass, NULL, NULL);
18896 Perl_re_printf( aTHX_ "stclass %s ", SvPVX_const(sv));
18898 if (r->intflags & PREGf_ANCH) {
18899 Perl_re_printf( aTHX_ "anchored");
18900 if (r->intflags & PREGf_ANCH_MBOL)
18901 Perl_re_printf( aTHX_ "(MBOL)");
18902 if (r->intflags & PREGf_ANCH_SBOL)
18903 Perl_re_printf( aTHX_ "(SBOL)");
18904 if (r->intflags & PREGf_ANCH_GPOS)
18905 Perl_re_printf( aTHX_ "(GPOS)");
18906 Perl_re_printf( aTHX_ " ");
18908 if (r->intflags & PREGf_GPOS_SEEN)
18909 Perl_re_printf( aTHX_ "GPOS:%" UVuf " ", (UV)r->gofs);
18910 if (r->intflags & PREGf_SKIP)
18911 Perl_re_printf( aTHX_ "plus ");
18912 if (r->intflags & PREGf_IMPLICIT)
18913 Perl_re_printf( aTHX_ "implicit ");
18914 Perl_re_printf( aTHX_ "minlen %" IVdf " ", (IV)r->minlen);
18915 if (r->extflags & RXf_EVAL_SEEN)
18916 Perl_re_printf( aTHX_ "with eval ");
18917 Perl_re_printf( aTHX_ "\n");
18919 regdump_extflags("r->extflags: ",r->extflags);
18920 regdump_intflags("r->intflags: ",r->intflags);
18923 PERL_ARGS_ASSERT_REGDUMP;
18924 PERL_UNUSED_CONTEXT;
18925 PERL_UNUSED_ARG(r);
18926 #endif /* DEBUGGING */
18929 /* Should be synchronized with ANYOF_ #defines in regcomp.h */
18932 # if _CC_WORDCHAR != 0 || _CC_DIGIT != 1 || _CC_ALPHA != 2 \
18933 || _CC_LOWER != 3 || _CC_UPPER != 4 || _CC_PUNCT != 5 \
18934 || _CC_PRINT != 6 || _CC_ALPHANUMERIC != 7 || _CC_GRAPH != 8 \
18935 || _CC_CASED != 9 || _CC_SPACE != 10 || _CC_BLANK != 11 \
18936 || _CC_XDIGIT != 12 || _CC_CNTRL != 13 || _CC_ASCII != 14 \
18937 || _CC_VERTSPACE != 15
18938 # error Need to adjust order of anyofs[]
18940 static const char * const anyofs[] = {
18977 - regprop - printable representation of opcode, with run time support
18981 Perl_regprop(pTHX_ const regexp *prog, SV *sv, const regnode *o, const regmatch_info *reginfo, const RExC_state_t *pRExC_state)
18985 RXi_GET_DECL(prog,progi);
18986 GET_RE_DEBUG_FLAGS_DECL;
18988 PERL_ARGS_ASSERT_REGPROP;
18992 if (OP(o) > REGNODE_MAX) /* regnode.type is unsigned */
18993 /* It would be nice to FAIL() here, but this may be called from
18994 regexec.c, and it would be hard to supply pRExC_state. */
18995 Perl_croak(aTHX_ "Corrupted regexp opcode %d > %d",
18996 (int)OP(o), (int)REGNODE_MAX);
18997 sv_catpv(sv, PL_reg_name[OP(o)]); /* Take off const! */
18999 k = PL_regkind[OP(o)];
19002 sv_catpvs(sv, " ");
19003 /* Using is_utf8_string() (via PERL_PV_UNI_DETECT)
19004 * is a crude hack but it may be the best for now since
19005 * we have no flag "this EXACTish node was UTF-8"
19007 pv_pretty(sv, STRING(o), STR_LEN(o), 60, PL_colors[0], PL_colors[1],
19008 PERL_PV_ESCAPE_UNI_DETECT |
19009 PERL_PV_ESCAPE_NONASCII |
19010 PERL_PV_PRETTY_ELLIPSES |
19011 PERL_PV_PRETTY_LTGT |
19012 PERL_PV_PRETTY_NOCLEAR
19014 } else if (k == TRIE) {
19015 /* print the details of the trie in dumpuntil instead, as
19016 * progi->data isn't available here */
19017 const char op = OP(o);
19018 const U32 n = ARG(o);
19019 const reg_ac_data * const ac = IS_TRIE_AC(op) ?
19020 (reg_ac_data *)progi->data->data[n] :
19022 const reg_trie_data * const trie
19023 = (reg_trie_data*)progi->data->data[!IS_TRIE_AC(op) ? n : ac->trie];
19025 Perl_sv_catpvf(aTHX_ sv, "-%s",PL_reg_name[o->flags]);
19026 DEBUG_TRIE_COMPILE_r({
19028 sv_catpvs(sv, "(JUMP)");
19029 Perl_sv_catpvf(aTHX_ sv,
19030 "<S:%" UVuf "/%" IVdf " W:%" UVuf " L:%" UVuf "/%" UVuf " C:%" UVuf "/%" UVuf ">",
19031 (UV)trie->startstate,
19032 (IV)trie->statecount-1, /* -1 because of the unused 0 element */
19033 (UV)trie->wordcount,
19036 (UV)TRIE_CHARCOUNT(trie),
19037 (UV)trie->uniquecharcount
19040 if ( IS_ANYOF_TRIE(op) || trie->bitmap ) {
19041 sv_catpvs(sv, "[");
19042 (void) put_charclass_bitmap_innards(sv,
19043 ((IS_ANYOF_TRIE(op))
19045 : TRIE_BITMAP(trie)),
19051 sv_catpvs(sv, "]");
19053 } else if (k == CURLY) {
19054 U32 lo = ARG1(o), hi = ARG2(o);
19055 if (OP(o) == CURLYM || OP(o) == CURLYN || OP(o) == CURLYX)
19056 Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* Parenth number */
19057 Perl_sv_catpvf(aTHX_ sv, "{%u,", (unsigned) lo);
19058 if (hi == REG_INFTY)
19059 sv_catpvs(sv, "INFTY");
19061 Perl_sv_catpvf(aTHX_ sv, "%u", (unsigned) hi);
19062 sv_catpvs(sv, "}");
19064 else if (k == WHILEM && o->flags) /* Ordinal/of */
19065 Perl_sv_catpvf(aTHX_ sv, "[%d/%d]", o->flags & 0xf, o->flags>>4);
19066 else if (k == REF || k == OPEN || k == CLOSE
19067 || k == GROUPP || OP(o)==ACCEPT)
19069 AV *name_list= NULL;
19070 U32 parno= OP(o) == ACCEPT ? (U32)ARG2L(o) : ARG(o);
19071 Perl_sv_catpvf(aTHX_ sv, "%" UVuf, (UV)parno); /* Parenth number */
19072 if ( RXp_PAREN_NAMES(prog) ) {
19073 name_list= MUTABLE_AV(progi->data->data[progi->name_list_idx]);
19074 } else if ( pRExC_state ) {
19075 name_list= RExC_paren_name_list;
19078 if ( k != REF || (OP(o) < NREF)) {
19079 SV **name= av_fetch(name_list, parno, 0 );
19081 Perl_sv_catpvf(aTHX_ sv, " '%" SVf "'", SVfARG(*name));
19084 SV *sv_dat= MUTABLE_SV(progi->data->data[ parno ]);
19085 I32 *nums=(I32*)SvPVX(sv_dat);
19086 SV **name= av_fetch(name_list, nums[0], 0 );
19089 for ( n=0; n<SvIVX(sv_dat); n++ ) {
19090 Perl_sv_catpvf(aTHX_ sv, "%s%" IVdf,
19091 (n ? "," : ""), (IV)nums[n]);
19093 Perl_sv_catpvf(aTHX_ sv, " '%" SVf "'", SVfARG(*name));
19097 if ( k == REF && reginfo) {
19098 U32 n = ARG(o); /* which paren pair */
19099 I32 ln = prog->offs[n].start;
19100 if (prog->lastparen < n || ln == -1)
19101 Perl_sv_catpvf(aTHX_ sv, ": FAIL");
19102 else if (ln == prog->offs[n].end)
19103 Perl_sv_catpvf(aTHX_ sv, ": ACCEPT - EMPTY STRING");
19105 const char *s = reginfo->strbeg + ln;
19106 Perl_sv_catpvf(aTHX_ sv, ": ");
19107 Perl_pv_pretty( aTHX_ sv, s, prog->offs[n].end - prog->offs[n].start, 32, 0, 0,
19108 PERL_PV_ESCAPE_UNI_DETECT|PERL_PV_PRETTY_NOCLEAR|PERL_PV_PRETTY_ELLIPSES|PERL_PV_PRETTY_QUOTE );
19111 } else if (k == GOSUB) {
19112 AV *name_list= NULL;
19113 if ( RXp_PAREN_NAMES(prog) ) {
19114 name_list= MUTABLE_AV(progi->data->data[progi->name_list_idx]);
19115 } else if ( pRExC_state ) {
19116 name_list= RExC_paren_name_list;
19119 /* Paren and offset */
19120 Perl_sv_catpvf(aTHX_ sv, "%d[%+d:%d]", (int)ARG(o),(int)ARG2L(o),
19121 (int)((o + (int)ARG2L(o)) - progi->program) );
19123 SV **name= av_fetch(name_list, ARG(o), 0 );
19125 Perl_sv_catpvf(aTHX_ sv, " '%" SVf "'", SVfARG(*name));
19128 else if (k == LOGICAL)
19129 /* 2: embedded, otherwise 1 */
19130 Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags);
19131 else if (k == ANYOF) {
19132 const U8 flags = ANYOF_FLAGS(o);
19133 bool do_sep = FALSE; /* Do we need to separate various components of
19135 /* Set if there is still an unresolved user-defined property */
19136 SV *unresolved = NULL;
19138 /* Things that are ignored except when the runtime locale is UTF-8 */
19139 SV *only_utf8_locale_invlist = NULL;
19141 /* Code points that don't fit in the bitmap */
19142 SV *nonbitmap_invlist = NULL;
19144 /* And things that aren't in the bitmap, but are small enough to be */
19145 SV* bitmap_range_not_in_bitmap = NULL;
19147 const bool inverted = flags & ANYOF_INVERT;
19149 if (OP(o) == ANYOFL) {
19150 if (ANYOFL_UTF8_LOCALE_REQD(flags)) {
19151 sv_catpvs(sv, "{utf8-locale-reqd}");
19153 if (flags & ANYOFL_FOLD) {
19154 sv_catpvs(sv, "{i}");
19158 /* If there is stuff outside the bitmap, get it */
19159 if (ARG(o) != ANYOF_ONLY_HAS_BITMAP) {
19160 (void) _get_regclass_nonbitmap_data(prog, o, FALSE,
19162 &only_utf8_locale_invlist,
19163 &nonbitmap_invlist);
19164 /* The non-bitmap data may contain stuff that could fit in the
19165 * bitmap. This could come from a user-defined property being
19166 * finally resolved when this call was done; or much more likely
19167 * because there are matches that require UTF-8 to be valid, and so
19168 * aren't in the bitmap. This is teased apart later */
19169 _invlist_intersection(nonbitmap_invlist,
19171 &bitmap_range_not_in_bitmap);
19172 /* Leave just the things that don't fit into the bitmap */
19173 _invlist_subtract(nonbitmap_invlist,
19175 &nonbitmap_invlist);
19178 /* Obey this flag to add all above-the-bitmap code points */
19179 if (flags & ANYOF_MATCHES_ALL_ABOVE_BITMAP) {
19180 nonbitmap_invlist = _add_range_to_invlist(nonbitmap_invlist,
19181 NUM_ANYOF_CODE_POINTS,
19185 /* Ready to start outputting. First, the initial left bracket */
19186 Perl_sv_catpvf(aTHX_ sv, "[%s", PL_colors[0]);
19188 /* Then all the things that could fit in the bitmap */
19189 do_sep = put_charclass_bitmap_innards(sv,
19191 bitmap_range_not_in_bitmap,
19192 only_utf8_locale_invlist,
19195 /* Can't try inverting for a
19196 * better display if there are
19197 * things that haven't been
19199 unresolved != NULL);
19200 SvREFCNT_dec(bitmap_range_not_in_bitmap);
19202 /* If there are user-defined properties which haven't been defined yet,
19203 * output them. If the result is not to be inverted, it is clearest to
19204 * output them in a separate [] from the bitmap range stuff. If the
19205 * result is to be complemented, we have to show everything in one [],
19206 * as the inversion applies to the whole thing. Use {braces} to
19207 * separate them from anything in the bitmap and anything above the
19211 if (! do_sep) { /* If didn't output anything in the bitmap */
19212 sv_catpvs(sv, "^");
19214 sv_catpvs(sv, "{");
19217 Perl_sv_catpvf(aTHX_ sv,"%s][%s",PL_colors[1],PL_colors[0]);
19219 sv_catsv(sv, unresolved);
19221 sv_catpvs(sv, "}");
19223 do_sep = ! inverted;
19226 /* And, finally, add the above-the-bitmap stuff */
19227 if (nonbitmap_invlist && _invlist_len(nonbitmap_invlist)) {
19230 /* See if truncation size is overridden */
19231 const STRLEN dump_len = (PL_dump_re_max_len)
19232 ? PL_dump_re_max_len
19235 /* This is output in a separate [] */
19237 Perl_sv_catpvf(aTHX_ sv,"%s][%s",PL_colors[1],PL_colors[0]);
19240 /* And, for easy of understanding, it is shown in the
19241 * uncomplemented form if possible. The one exception being if
19242 * there are unresolved items, where the inversion has to be
19243 * delayed until runtime */
19244 if (inverted && ! unresolved) {
19245 _invlist_invert(nonbitmap_invlist);
19246 _invlist_subtract(nonbitmap_invlist, PL_InBitmap, &nonbitmap_invlist);
19249 contents = invlist_contents(nonbitmap_invlist,
19250 FALSE /* output suitable for catsv */
19253 /* If the output is shorter than the permissible maximum, just do it. */
19254 if (SvCUR(contents) <= dump_len) {
19255 sv_catsv(sv, contents);
19258 const char * contents_string = SvPVX(contents);
19259 STRLEN i = dump_len;
19261 /* Otherwise, start at the permissible max and work back to the
19262 * first break possibility */
19263 while (i > 0 && contents_string[i] != ' ') {
19266 if (i == 0) { /* Fail-safe. Use the max if we couldn't
19267 find a legal break */
19271 sv_catpvn(sv, contents_string, i);
19272 sv_catpvs(sv, "...");
19275 SvREFCNT_dec_NN(contents);
19276 SvREFCNT_dec_NN(nonbitmap_invlist);
19279 /* And finally the matching, closing ']' */
19280 Perl_sv_catpvf(aTHX_ sv, "%s]", PL_colors[1]);
19282 SvREFCNT_dec(unresolved);
19284 else if (k == POSIXD || k == NPOSIXD) {
19285 U8 index = FLAGS(o) * 2;
19286 if (index < C_ARRAY_LENGTH(anyofs)) {
19287 if (*anyofs[index] != '[') {
19290 sv_catpv(sv, anyofs[index]);
19291 if (*anyofs[index] != '[') {
19296 Perl_sv_catpvf(aTHX_ sv, "[illegal type=%d])", index);
19299 else if (k == BOUND || k == NBOUND) {
19300 /* Must be synced with order of 'bound_type' in regcomp.h */
19301 const char * const bounds[] = {
19302 "", /* Traditional */
19308 assert(FLAGS(o) < C_ARRAY_LENGTH(bounds));
19309 sv_catpv(sv, bounds[FLAGS(o)]);
19311 else if (k == BRANCHJ && (OP(o) == UNLESSM || OP(o) == IFMATCH))
19312 Perl_sv_catpvf(aTHX_ sv, "[%d]", -(o->flags));
19313 else if (OP(o) == SBOL)
19314 Perl_sv_catpvf(aTHX_ sv, " /%s/", o->flags ? "\\A" : "^");
19316 /* add on the verb argument if there is one */
19317 if ( ( k == VERB || OP(o) == ACCEPT || OP(o) == OPFAIL ) && o->flags) {
19318 Perl_sv_catpvf(aTHX_ sv, ":%" SVf,
19319 SVfARG((MUTABLE_SV(progi->data->data[ ARG( o ) ]))));
19322 PERL_UNUSED_CONTEXT;
19323 PERL_UNUSED_ARG(sv);
19324 PERL_UNUSED_ARG(o);
19325 PERL_UNUSED_ARG(prog);
19326 PERL_UNUSED_ARG(reginfo);
19327 PERL_UNUSED_ARG(pRExC_state);
19328 #endif /* DEBUGGING */
19334 Perl_re_intuit_string(pTHX_ REGEXP * const r)
19335 { /* Assume that RE_INTUIT is set */
19336 struct regexp *const prog = ReANY(r);
19337 GET_RE_DEBUG_FLAGS_DECL;
19339 PERL_ARGS_ASSERT_RE_INTUIT_STRING;
19340 PERL_UNUSED_CONTEXT;
19344 const char * const s = SvPV_nolen_const(RX_UTF8(r)
19345 ? prog->check_utf8 : prog->check_substr);
19347 if (!PL_colorset) reginitcolors();
19348 Perl_re_printf( aTHX_
19349 "%sUsing REx %ssubstr:%s \"%s%.60s%s%s\"\n",
19351 RX_UTF8(r) ? "utf8 " : "",
19352 PL_colors[5],PL_colors[0],
19355 (strlen(s) > 60 ? "..." : ""));
19358 /* use UTF8 check substring if regexp pattern itself is in UTF8 */
19359 return RX_UTF8(r) ? prog->check_utf8 : prog->check_substr;
19365 handles refcounting and freeing the perl core regexp structure. When
19366 it is necessary to actually free the structure the first thing it
19367 does is call the 'free' method of the regexp_engine associated to
19368 the regexp, allowing the handling of the void *pprivate; member
19369 first. (This routine is not overridable by extensions, which is why
19370 the extensions free is called first.)
19372 See regdupe and regdupe_internal if you change anything here.
19374 #ifndef PERL_IN_XSUB_RE
19376 Perl_pregfree(pTHX_ REGEXP *r)
19382 Perl_pregfree2(pTHX_ REGEXP *rx)
19384 struct regexp *const r = ReANY(rx);
19385 GET_RE_DEBUG_FLAGS_DECL;
19387 PERL_ARGS_ASSERT_PREGFREE2;
19389 if (r->mother_re) {
19390 ReREFCNT_dec(r->mother_re);
19392 CALLREGFREE_PVT(rx); /* free the private data */
19393 SvREFCNT_dec(RXp_PAREN_NAMES(r));
19394 Safefree(r->xpv_len_u.xpvlenu_pv);
19397 SvREFCNT_dec(r->anchored_substr);
19398 SvREFCNT_dec(r->anchored_utf8);
19399 SvREFCNT_dec(r->float_substr);
19400 SvREFCNT_dec(r->float_utf8);
19401 Safefree(r->substrs);
19403 RX_MATCH_COPY_FREE(rx);
19404 #ifdef PERL_ANY_COW
19405 SvREFCNT_dec(r->saved_copy);
19408 SvREFCNT_dec(r->qr_anoncv);
19409 if (r->recurse_locinput)
19410 Safefree(r->recurse_locinput);
19411 rx->sv_u.svu_rx = 0;
19416 This is a hacky workaround to the structural issue of match results
19417 being stored in the regexp structure which is in turn stored in
19418 PL_curpm/PL_reg_curpm. The problem is that due to qr// the pattern
19419 could be PL_curpm in multiple contexts, and could require multiple
19420 result sets being associated with the pattern simultaneously, such
19421 as when doing a recursive match with (??{$qr})
19423 The solution is to make a lightweight copy of the regexp structure
19424 when a qr// is returned from the code executed by (??{$qr}) this
19425 lightweight copy doesn't actually own any of its data except for
19426 the starp/end and the actual regexp structure itself.
19432 Perl_reg_temp_copy (pTHX_ REGEXP *ret_x, REGEXP *rx)
19434 struct regexp *ret;
19435 struct regexp *const r = ReANY(rx);
19436 const bool islv = ret_x && SvTYPE(ret_x) == SVt_PVLV;
19438 PERL_ARGS_ASSERT_REG_TEMP_COPY;
19441 ret_x = (REGEXP*) newSV_type(SVt_REGEXP);
19443 SvOK_off((SV *)ret_x);
19445 /* For PVLVs, SvANY points to the xpvlv body while sv_u points
19446 to the regexp. (For SVt_REGEXPs, sv_upgrade has already
19447 made both spots point to the same regexp body.) */
19448 REGEXP *temp = (REGEXP *)newSV_type(SVt_REGEXP);
19449 assert(!SvPVX(ret_x));
19450 ret_x->sv_u.svu_rx = temp->sv_any;
19451 temp->sv_any = NULL;
19452 SvFLAGS(temp) = (SvFLAGS(temp) & ~SVTYPEMASK) | SVt_NULL;
19453 SvREFCNT_dec_NN(temp);
19454 /* SvCUR still resides in the xpvlv struct, so the regexp copy-
19455 ing below will not set it. */
19456 SvCUR_set(ret_x, SvCUR(rx));
19459 /* This ensures that SvTHINKFIRST(sv) is true, and hence that
19460 sv_force_normal(sv) is called. */
19462 ret = ReANY(ret_x);
19464 SvFLAGS(ret_x) |= SvUTF8(rx);
19465 /* We share the same string buffer as the original regexp, on which we
19466 hold a reference count, incremented when mother_re is set below.
19467 The string pointer is copied here, being part of the regexp struct.
19469 memcpy(&(ret->xpv_cur), &(r->xpv_cur),
19470 sizeof(regexp) - STRUCT_OFFSET(regexp, xpv_cur));
19472 const I32 npar = r->nparens+1;
19473 Newx(ret->offs, npar, regexp_paren_pair);
19474 Copy(r->offs, ret->offs, npar, regexp_paren_pair);
19477 Newx(ret->substrs, 1, struct reg_substr_data);
19478 StructCopy(r->substrs, ret->substrs, struct reg_substr_data);
19480 SvREFCNT_inc_void(ret->anchored_substr);
19481 SvREFCNT_inc_void(ret->anchored_utf8);
19482 SvREFCNT_inc_void(ret->float_substr);
19483 SvREFCNT_inc_void(ret->float_utf8);
19485 /* check_substr and check_utf8, if non-NULL, point to either their
19486 anchored or float namesakes, and don't hold a second reference. */
19488 RX_MATCH_COPIED_off(ret_x);
19489 #ifdef PERL_ANY_COW
19490 ret->saved_copy = NULL;
19492 ret->mother_re = ReREFCNT_inc(r->mother_re ? r->mother_re : rx);
19493 SvREFCNT_inc_void(ret->qr_anoncv);
19494 if (r->recurse_locinput)
19495 Newxz(ret->recurse_locinput,r->nparens + 1,char *);
19501 /* regfree_internal()
19503 Free the private data in a regexp. This is overloadable by
19504 extensions. Perl takes care of the regexp structure in pregfree(),
19505 this covers the *pprivate pointer which technically perl doesn't
19506 know about, however of course we have to handle the
19507 regexp_internal structure when no extension is in use.
19509 Note this is called before freeing anything in the regexp
19514 Perl_regfree_internal(pTHX_ REGEXP * const rx)
19516 struct regexp *const r = ReANY(rx);
19517 RXi_GET_DECL(r,ri);
19518 GET_RE_DEBUG_FLAGS_DECL;
19520 PERL_ARGS_ASSERT_REGFREE_INTERNAL;
19526 SV *dsv= sv_newmortal();
19527 RE_PV_QUOTED_DECL(s, RX_UTF8(rx),
19528 dsv, RX_PRECOMP(rx), RX_PRELEN(rx), 60);
19529 Perl_re_printf( aTHX_ "%sFreeing REx:%s %s\n",
19530 PL_colors[4],PL_colors[5],s);
19533 #ifdef RE_TRACK_PATTERN_OFFSETS
19535 Safefree(ri->u.offsets); /* 20010421 MJD */
19537 if (ri->code_blocks)
19538 S_free_codeblocks(aTHX_ ri->code_blocks);
19541 int n = ri->data->count;
19544 /* If you add a ->what type here, update the comment in regcomp.h */
19545 switch (ri->data->what[n]) {
19551 SvREFCNT_dec(MUTABLE_SV(ri->data->data[n]));
19554 Safefree(ri->data->data[n]);
19560 { /* Aho Corasick add-on structure for a trie node.
19561 Used in stclass optimization only */
19563 reg_ac_data *aho=(reg_ac_data*)ri->data->data[n];
19564 #ifdef USE_ITHREADS
19568 refcount = --aho->refcount;
19571 PerlMemShared_free(aho->states);
19572 PerlMemShared_free(aho->fail);
19573 /* do this last!!!! */
19574 PerlMemShared_free(ri->data->data[n]);
19575 /* we should only ever get called once, so
19576 * assert as much, and also guard the free
19577 * which /might/ happen twice. At the least
19578 * it will make code anlyzers happy and it
19579 * doesn't cost much. - Yves */
19580 assert(ri->regstclass);
19581 if (ri->regstclass) {
19582 PerlMemShared_free(ri->regstclass);
19583 ri->regstclass = 0;
19590 /* trie structure. */
19592 reg_trie_data *trie=(reg_trie_data*)ri->data->data[n];
19593 #ifdef USE_ITHREADS
19597 refcount = --trie->refcount;
19600 PerlMemShared_free(trie->charmap);
19601 PerlMemShared_free(trie->states);
19602 PerlMemShared_free(trie->trans);
19604 PerlMemShared_free(trie->bitmap);
19606 PerlMemShared_free(trie->jump);
19607 PerlMemShared_free(trie->wordinfo);
19608 /* do this last!!!! */
19609 PerlMemShared_free(ri->data->data[n]);
19614 Perl_croak(aTHX_ "panic: regfree data code '%c'",
19615 ri->data->what[n]);
19618 Safefree(ri->data->what);
19619 Safefree(ri->data);
19625 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
19626 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
19627 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
19630 re_dup_guts - duplicate a regexp.
19632 This routine is expected to clone a given regexp structure. It is only
19633 compiled under USE_ITHREADS.
19635 After all of the core data stored in struct regexp is duplicated
19636 the regexp_engine.dupe method is used to copy any private data
19637 stored in the *pprivate pointer. This allows extensions to handle
19638 any duplication it needs to do.
19640 See pregfree() and regfree_internal() if you change anything here.
19642 #if defined(USE_ITHREADS)
19643 #ifndef PERL_IN_XSUB_RE
19645 Perl_re_dup_guts(pTHX_ const REGEXP *sstr, REGEXP *dstr, CLONE_PARAMS *param)
19649 const struct regexp *r = ReANY(sstr);
19650 struct regexp *ret = ReANY(dstr);
19652 PERL_ARGS_ASSERT_RE_DUP_GUTS;
19654 npar = r->nparens+1;
19655 Newx(ret->offs, npar, regexp_paren_pair);
19656 Copy(r->offs, ret->offs, npar, regexp_paren_pair);
19658 if (ret->substrs) {
19659 /* Do it this way to avoid reading from *r after the StructCopy().
19660 That way, if any of the sv_dup_inc()s dislodge *r from the L1
19661 cache, it doesn't matter. */
19662 const bool anchored = r->check_substr
19663 ? r->check_substr == r->anchored_substr
19664 : r->check_utf8 == r->anchored_utf8;
19665 Newx(ret->substrs, 1, struct reg_substr_data);
19666 StructCopy(r->substrs, ret->substrs, struct reg_substr_data);
19668 ret->anchored_substr = sv_dup_inc(ret->anchored_substr, param);
19669 ret->anchored_utf8 = sv_dup_inc(ret->anchored_utf8, param);
19670 ret->float_substr = sv_dup_inc(ret->float_substr, param);
19671 ret->float_utf8 = sv_dup_inc(ret->float_utf8, param);
19673 /* check_substr and check_utf8, if non-NULL, point to either their
19674 anchored or float namesakes, and don't hold a second reference. */
19676 if (ret->check_substr) {
19678 assert(r->check_utf8 == r->anchored_utf8);
19679 ret->check_substr = ret->anchored_substr;
19680 ret->check_utf8 = ret->anchored_utf8;
19682 assert(r->check_substr == r->float_substr);
19683 assert(r->check_utf8 == r->float_utf8);
19684 ret->check_substr = ret->float_substr;
19685 ret->check_utf8 = ret->float_utf8;
19687 } else if (ret->check_utf8) {
19689 ret->check_utf8 = ret->anchored_utf8;
19691 ret->check_utf8 = ret->float_utf8;
19696 RXp_PAREN_NAMES(ret) = hv_dup_inc(RXp_PAREN_NAMES(ret), param);
19697 ret->qr_anoncv = MUTABLE_CV(sv_dup_inc((const SV *)ret->qr_anoncv, param));
19698 if (r->recurse_locinput)
19699 Newxz(ret->recurse_locinput,r->nparens + 1,char *);
19702 RXi_SET(ret,CALLREGDUPE_PVT(dstr,param));
19704 if (RX_MATCH_COPIED(dstr))
19705 ret->subbeg = SAVEPVN(ret->subbeg, ret->sublen);
19707 ret->subbeg = NULL;
19708 #ifdef PERL_ANY_COW
19709 ret->saved_copy = NULL;
19712 /* Whether mother_re be set or no, we need to copy the string. We
19713 cannot refrain from copying it when the storage points directly to
19714 our mother regexp, because that's
19715 1: a buffer in a different thread
19716 2: something we no longer hold a reference on
19717 so we need to copy it locally. */
19718 RX_WRAPPED(dstr) = SAVEPVN(RX_WRAPPED(sstr), SvCUR(sstr)+1);
19719 ret->mother_re = NULL;
19721 #endif /* PERL_IN_XSUB_RE */
19726 This is the internal complement to regdupe() which is used to copy
19727 the structure pointed to by the *pprivate pointer in the regexp.
19728 This is the core version of the extension overridable cloning hook.
19729 The regexp structure being duplicated will be copied by perl prior
19730 to this and will be provided as the regexp *r argument, however
19731 with the /old/ structures pprivate pointer value. Thus this routine
19732 may override any copying normally done by perl.
19734 It returns a pointer to the new regexp_internal structure.
19738 Perl_regdupe_internal(pTHX_ REGEXP * const rx, CLONE_PARAMS *param)
19741 struct regexp *const r = ReANY(rx);
19742 regexp_internal *reti;
19744 RXi_GET_DECL(r,ri);
19746 PERL_ARGS_ASSERT_REGDUPE_INTERNAL;
19750 Newxc(reti, sizeof(regexp_internal) + len*sizeof(regnode),
19751 char, regexp_internal);
19752 Copy(ri->program, reti->program, len+1, regnode);
19755 if (ri->code_blocks) {
19757 Newx(reti->code_blocks, 1, struct reg_code_blocks);
19758 Newx(reti->code_blocks->cb, ri->code_blocks->count,
19759 struct reg_code_block);
19760 Copy(ri->code_blocks->cb, reti->code_blocks->cb,
19761 ri->code_blocks->count, struct reg_code_block);
19762 for (n = 0; n < ri->code_blocks->count; n++)
19763 reti->code_blocks->cb[n].src_regex = (REGEXP*)
19764 sv_dup_inc((SV*)(ri->code_blocks->cb[n].src_regex), param);
19765 reti->code_blocks->count = ri->code_blocks->count;
19766 reti->code_blocks->refcnt = 1;
19769 reti->code_blocks = NULL;
19771 reti->regstclass = NULL;
19774 struct reg_data *d;
19775 const int count = ri->data->count;
19778 Newxc(d, sizeof(struct reg_data) + count*sizeof(void *),
19779 char, struct reg_data);
19780 Newx(d->what, count, U8);
19783 for (i = 0; i < count; i++) {
19784 d->what[i] = ri->data->what[i];
19785 switch (d->what[i]) {
19786 /* see also regcomp.h and regfree_internal() */
19787 case 'a': /* actually an AV, but the dup function is identical. */
19791 case 'u': /* actually an HV, but the dup function is identical. */
19792 d->data[i] = sv_dup_inc((const SV *)ri->data->data[i], param);
19795 /* This is cheating. */
19796 Newx(d->data[i], 1, regnode_ssc);
19797 StructCopy(ri->data->data[i], d->data[i], regnode_ssc);
19798 reti->regstclass = (regnode*)d->data[i];
19801 /* Trie stclasses are readonly and can thus be shared
19802 * without duplication. We free the stclass in pregfree
19803 * when the corresponding reg_ac_data struct is freed.
19805 reti->regstclass= ri->regstclass;
19809 ((reg_trie_data*)ri->data->data[i])->refcount++;
19814 d->data[i] = ri->data->data[i];
19817 Perl_croak(aTHX_ "panic: re_dup_guts unknown data code '%c'",
19818 ri->data->what[i]);
19827 reti->name_list_idx = ri->name_list_idx;
19829 #ifdef RE_TRACK_PATTERN_OFFSETS
19830 if (ri->u.offsets) {
19831 Newx(reti->u.offsets, 2*len+1, U32);
19832 Copy(ri->u.offsets, reti->u.offsets, 2*len+1, U32);
19835 SetProgLen(reti,len);
19838 return (void*)reti;
19841 #endif /* USE_ITHREADS */
19843 #ifndef PERL_IN_XSUB_RE
19846 - regnext - dig the "next" pointer out of a node
19849 Perl_regnext(pTHX_ regnode *p)
19856 if (OP(p) > REGNODE_MAX) { /* regnode.type is unsigned */
19857 Perl_croak(aTHX_ "Corrupted regexp opcode %d > %d",
19858 (int)OP(p), (int)REGNODE_MAX);
19861 offset = (reg_off_by_arg[OP(p)] ? ARG(p) : NEXT_OFF(p));
19870 S_re_croak2(pTHX_ bool utf8, const char* pat1,const char* pat2,...)
19873 STRLEN l1 = strlen(pat1);
19874 STRLEN l2 = strlen(pat2);
19877 const char *message;
19879 PERL_ARGS_ASSERT_RE_CROAK2;
19885 Copy(pat1, buf, l1 , char);
19886 Copy(pat2, buf + l1, l2 , char);
19887 buf[l1 + l2] = '\n';
19888 buf[l1 + l2 + 1] = '\0';
19889 va_start(args, pat2);
19890 msv = vmess(buf, &args);
19892 message = SvPV_const(msv,l1);
19895 Copy(message, buf, l1 , char);
19896 /* l1-1 to avoid \n */
19897 Perl_croak(aTHX_ "%" UTF8f, UTF8fARG(utf8, l1-1, buf));
19900 /* XXX Here's a total kludge. But we need to re-enter for swash routines. */
19902 #ifndef PERL_IN_XSUB_RE
19904 Perl_save_re_context(pTHX)
19909 /* Save $1..$n (#18107: UTF-8 s/(\w+)/uc($1)/e); AMS 20021106. */
19912 const REGEXP * const rx = PM_GETRE(PL_curpm);
19914 nparens = RX_NPARENS(rx);
19917 /* RT #124109. This is a complete hack; in the SWASHNEW case we know
19918 * that PL_curpm will be null, but that utf8.pm and the modules it
19919 * loads will only use $1..$3.
19920 * The t/porting/re_context.t test file checks this assumption.
19925 for (i = 1; i <= nparens; i++) {
19926 char digits[TYPE_CHARS(long)];
19927 const STRLEN len = my_snprintf(digits, sizeof(digits),
19929 GV *const *const gvp
19930 = (GV**)hv_fetch(PL_defstash, digits, len, 0);
19933 GV * const gv = *gvp;
19934 if (SvTYPE(gv) == SVt_PVGV && GvSV(gv))
19944 S_put_code_point(pTHX_ SV *sv, UV c)
19946 PERL_ARGS_ASSERT_PUT_CODE_POINT;
19949 Perl_sv_catpvf(aTHX_ sv, "\\x{%04" UVXf "}", c);
19951 else if (isPRINT(c)) {
19952 const char string = (char) c;
19954 /* We use {phrase} as metanotation in the class, so also escape literal
19956 if (isBACKSLASHED_PUNCT(c) || c == '{' || c == '}')
19957 sv_catpvs(sv, "\\");
19958 sv_catpvn(sv, &string, 1);
19960 else if (isMNEMONIC_CNTRL(c)) {
19961 Perl_sv_catpvf(aTHX_ sv, "%s", cntrl_to_mnemonic((U8) c));
19964 Perl_sv_catpvf(aTHX_ sv, "\\x%02X", (U8) c);
19968 #define MAX_PRINT_A MAX_PRINT_A_FOR_USE_ONLY_BY_REGCOMP_DOT_C
19971 S_put_range(pTHX_ SV *sv, UV start, const UV end, const bool allow_literals)
19973 /* Appends to 'sv' a displayable version of the range of code points from
19974 * 'start' to 'end'. Mnemonics (like '\r') are used for the few controls
19975 * that have them, when they occur at the beginning or end of the range.
19976 * It uses hex to output the remaining code points, unless 'allow_literals'
19977 * is true, in which case the printable ASCII ones are output as-is (though
19978 * some of these will be escaped by put_code_point()).
19980 * NOTE: This is designed only for printing ranges of code points that fit
19981 * inside an ANYOF bitmap. Higher code points are simply suppressed
19984 const unsigned int min_range_count = 3;
19986 assert(start <= end);
19988 PERL_ARGS_ASSERT_PUT_RANGE;
19990 while (start <= end) {
19992 const char * format;
19994 if (end - start < min_range_count) {
19996 /* Output chars individually when they occur in short ranges */
19997 for (; start <= end; start++) {
19998 put_code_point(sv, start);
20003 /* If permitted by the input options, and there is a possibility that
20004 * this range contains a printable literal, look to see if there is
20006 if (allow_literals && start <= MAX_PRINT_A) {
20008 /* If the character at the beginning of the range isn't an ASCII
20009 * printable, effectively split the range into two parts:
20010 * 1) the portion before the first such printable,
20012 * and output them separately. */
20013 if (! isPRINT_A(start)) {
20014 UV temp_end = start + 1;
20016 /* There is no point looking beyond the final possible
20017 * printable, in MAX_PRINT_A */
20018 UV max = MIN(end, MAX_PRINT_A);
20020 while (temp_end <= max && ! isPRINT_A(temp_end)) {
20024 /* Here, temp_end points to one beyond the first printable if
20025 * found, or to one beyond 'max' if not. If none found, make
20026 * sure that we use the entire range */
20027 if (temp_end > MAX_PRINT_A) {
20028 temp_end = end + 1;
20031 /* Output the first part of the split range: the part that
20032 * doesn't have printables, with the parameter set to not look
20033 * for literals (otherwise we would infinitely recurse) */
20034 put_range(sv, start, temp_end - 1, FALSE);
20036 /* The 2nd part of the range (if any) starts here. */
20039 /* We do a continue, instead of dropping down, because even if
20040 * the 2nd part is non-empty, it could be so short that we want
20041 * to output it as individual characters, as tested for at the
20042 * top of this loop. */
20046 /* Here, 'start' is a printable ASCII. If it is an alphanumeric,
20047 * output a sub-range of just the digits or letters, then process
20048 * the remaining portion as usual. */
20049 if (isALPHANUMERIC_A(start)) {
20050 UV mask = (isDIGIT_A(start))
20055 UV temp_end = start + 1;
20057 /* Find the end of the sub-range that includes just the
20058 * characters in the same class as the first character in it */
20059 while (temp_end <= end && _generic_isCC_A(temp_end, mask)) {
20064 /* For short ranges, don't duplicate the code above to output
20065 * them; just call recursively */
20066 if (temp_end - start < min_range_count) {
20067 put_range(sv, start, temp_end, FALSE);
20069 else { /* Output as a range */
20070 put_code_point(sv, start);
20071 sv_catpvs(sv, "-");
20072 put_code_point(sv, temp_end);
20074 start = temp_end + 1;
20078 /* We output any other printables as individual characters */
20079 if (isPUNCT_A(start) || isSPACE_A(start)) {
20080 while (start <= end && (isPUNCT_A(start)
20081 || isSPACE_A(start)))
20083 put_code_point(sv, start);
20088 } /* End of looking for literals */
20090 /* Here is not to output as a literal. Some control characters have
20091 * mnemonic names. Split off any of those at the beginning and end of
20092 * the range to print mnemonically. It isn't possible for many of
20093 * these to be in a row, so this won't overwhelm with output */
20095 && (isMNEMONIC_CNTRL(start) || isMNEMONIC_CNTRL(end)))
20097 while (isMNEMONIC_CNTRL(start) && start <= end) {
20098 put_code_point(sv, start);
20102 /* If this didn't take care of the whole range ... */
20103 if (start <= end) {
20105 /* Look backwards from the end to find the final non-mnemonic
20108 while (isMNEMONIC_CNTRL(temp_end)) {
20112 /* And separately output the interior range that doesn't start
20113 * or end with mnemonics */
20114 put_range(sv, start, temp_end, FALSE);
20116 /* Then output the mnemonic trailing controls */
20117 start = temp_end + 1;
20118 while (start <= end) {
20119 put_code_point(sv, start);
20126 /* As a final resort, output the range or subrange as hex. */
20128 this_end = (end < NUM_ANYOF_CODE_POINTS)
20130 : NUM_ANYOF_CODE_POINTS - 1;
20131 #if NUM_ANYOF_CODE_POINTS > 256
20132 format = (this_end < 256)
20133 ? "\\x%02" UVXf "-\\x%02" UVXf
20134 : "\\x{%04" UVXf "}-\\x{%04" UVXf "}";
20136 format = "\\x%02" UVXf "-\\x%02" UVXf;
20138 GCC_DIAG_IGNORE(-Wformat-nonliteral);
20139 Perl_sv_catpvf(aTHX_ sv, format, start, this_end);
20146 S_put_charclass_bitmap_innards_invlist(pTHX_ SV *sv, SV* invlist)
20148 /* Concatenate onto the PV in 'sv' a displayable form of the inversion list
20152 bool allow_literals = TRUE;
20154 PERL_ARGS_ASSERT_PUT_CHARCLASS_BITMAP_INNARDS_INVLIST;
20156 /* Generally, it is more readable if printable characters are output as
20157 * literals, but if a range (nearly) spans all of them, it's best to output
20158 * it as a single range. This code will use a single range if all but 2
20159 * ASCII printables are in it */
20160 invlist_iterinit(invlist);
20161 while (invlist_iternext(invlist, &start, &end)) {
20163 /* If the range starts beyond the final printable, it doesn't have any
20165 if (start > MAX_PRINT_A) {
20169 /* In both ASCII and EBCDIC, a SPACE is the lowest printable. To span
20170 * all but two, the range must start and end no later than 2 from
20172 if (start < ' ' + 2 && end > MAX_PRINT_A - 2) {
20173 if (end > MAX_PRINT_A) {
20179 if (end - start >= MAX_PRINT_A - ' ' - 2) {
20180 allow_literals = FALSE;
20185 invlist_iterfinish(invlist);
20187 /* Here we have figured things out. Output each range */
20188 invlist_iterinit(invlist);
20189 while (invlist_iternext(invlist, &start, &end)) {
20190 if (start >= NUM_ANYOF_CODE_POINTS) {
20193 put_range(sv, start, end, allow_literals);
20195 invlist_iterfinish(invlist);
20201 S_put_charclass_bitmap_innards_common(pTHX_
20202 SV* invlist, /* The bitmap */
20203 SV* posixes, /* Under /l, things like [:word:], \S */
20204 SV* only_utf8, /* Under /d, matches iff the target is UTF-8 */
20205 SV* not_utf8, /* /d, matches iff the target isn't UTF-8 */
20206 SV* only_utf8_locale, /* Under /l, matches if the locale is UTF-8 */
20207 const bool invert /* Is the result to be inverted? */
20210 /* Create and return an SV containing a displayable version of the bitmap
20211 * and associated information determined by the input parameters. If the
20212 * output would have been only the inversion indicator '^', NULL is instead
20217 PERL_ARGS_ASSERT_PUT_CHARCLASS_BITMAP_INNARDS_COMMON;
20220 output = newSVpvs("^");
20223 output = newSVpvs("");
20226 /* First, the code points in the bitmap that are unconditionally there */
20227 put_charclass_bitmap_innards_invlist(output, invlist);
20229 /* Traditionally, these have been placed after the main code points */
20231 sv_catsv(output, posixes);
20234 if (only_utf8 && _invlist_len(only_utf8)) {
20235 Perl_sv_catpvf(aTHX_ output, "%s{utf8}%s", PL_colors[1], PL_colors[0]);
20236 put_charclass_bitmap_innards_invlist(output, only_utf8);
20239 if (not_utf8 && _invlist_len(not_utf8)) {
20240 Perl_sv_catpvf(aTHX_ output, "%s{not utf8}%s", PL_colors[1], PL_colors[0]);
20241 put_charclass_bitmap_innards_invlist(output, not_utf8);
20244 if (only_utf8_locale && _invlist_len(only_utf8_locale)) {
20245 Perl_sv_catpvf(aTHX_ output, "%s{utf8 locale}%s", PL_colors[1], PL_colors[0]);
20246 put_charclass_bitmap_innards_invlist(output, only_utf8_locale);
20248 /* This is the only list in this routine that can legally contain code
20249 * points outside the bitmap range. The call just above to
20250 * 'put_charclass_bitmap_innards_invlist' will simply suppress them, so
20251 * output them here. There's about a half-dozen possible, and none in
20252 * contiguous ranges longer than 2 */
20253 if (invlist_highest(only_utf8_locale) >= NUM_ANYOF_CODE_POINTS) {
20255 SV* above_bitmap = NULL;
20257 _invlist_subtract(only_utf8_locale, PL_InBitmap, &above_bitmap);
20259 invlist_iterinit(above_bitmap);
20260 while (invlist_iternext(above_bitmap, &start, &end)) {
20263 for (i = start; i <= end; i++) {
20264 put_code_point(output, i);
20267 invlist_iterfinish(above_bitmap);
20268 SvREFCNT_dec_NN(above_bitmap);
20272 if (invert && SvCUR(output) == 1) {
20280 S_put_charclass_bitmap_innards(pTHX_ SV *sv,
20282 SV *nonbitmap_invlist,
20283 SV *only_utf8_locale_invlist,
20284 const regnode * const node,
20285 const bool force_as_is_display)
20287 /* Appends to 'sv' a displayable version of the innards of the bracketed
20288 * character class defined by the other arguments:
20289 * 'bitmap' points to the bitmap.
20290 * 'nonbitmap_invlist' is an inversion list of the code points that are in
20291 * the bitmap range, but for some reason aren't in the bitmap; NULL if
20292 * none. The reasons for this could be that they require some
20293 * condition such as the target string being or not being in UTF-8
20294 * (under /d), or because they came from a user-defined property that
20295 * was not resolved at the time of the regex compilation (under /u)
20296 * 'only_utf8_locale_invlist' is an inversion list of the code points that
20297 * are valid only if the runtime locale is a UTF-8 one; NULL if none
20298 * 'node' is the regex pattern node. It is needed only when the above two
20299 * parameters are not null, and is passed so that this routine can
20300 * tease apart the various reasons for them.
20301 * 'force_as_is_display' is TRUE if this routine should definitely NOT try
20302 * to invert things to see if that leads to a cleaner display. If
20303 * FALSE, this routine is free to use its judgment about doing this.
20305 * It returns TRUE if there was actually something output. (It may be that
20306 * the bitmap, etc is empty.)
20308 * When called for outputting the bitmap of a non-ANYOF node, just pass the
20309 * bitmap, with the succeeding parameters set to NULL, and the final one to
20313 /* In general, it tries to display the 'cleanest' representation of the
20314 * innards, choosing whether to display them inverted or not, regardless of
20315 * whether the class itself is to be inverted. However, there are some
20316 * cases where it can't try inverting, as what actually matches isn't known
20317 * until runtime, and hence the inversion isn't either. */
20318 bool inverting_allowed = ! force_as_is_display;
20321 STRLEN orig_sv_cur = SvCUR(sv);
20323 SV* invlist; /* Inversion list we accumulate of code points that
20324 are unconditionally matched */
20325 SV* only_utf8 = NULL; /* Under /d, list of matches iff the target is
20327 SV* not_utf8 = NULL; /* /d, list of matches iff the target isn't UTF-8
20329 SV* posixes = NULL; /* Under /l, string of things like [:word:], \D */
20330 SV* only_utf8_locale = NULL; /* Under /l, list of matches if the locale
20333 SV* as_is_display; /* The output string when we take the inputs
20335 SV* inverted_display; /* The output string when we invert the inputs */
20337 U8 flags = (node) ? ANYOF_FLAGS(node) : 0;
20339 bool invert = cBOOL(flags & ANYOF_INVERT); /* Is the input to be inverted
20341 /* We are biased in favor of displaying things without them being inverted,
20342 * as that is generally easier to understand */
20343 const int bias = 5;
20345 PERL_ARGS_ASSERT_PUT_CHARCLASS_BITMAP_INNARDS;
20347 /* Start off with whatever code points are passed in. (We clone, so we
20348 * don't change the caller's list) */
20349 if (nonbitmap_invlist) {
20350 assert(invlist_highest(nonbitmap_invlist) < NUM_ANYOF_CODE_POINTS);
20351 invlist = invlist_clone(nonbitmap_invlist);
20353 else { /* Worst case size is every other code point is matched */
20354 invlist = _new_invlist(NUM_ANYOF_CODE_POINTS / 2);
20358 if (OP(node) == ANYOFD) {
20360 /* This flag indicates that the code points below 0x100 in the
20361 * nonbitmap list are precisely the ones that match only when the
20362 * target is UTF-8 (they should all be non-ASCII). */
20363 if (flags & ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP)
20365 _invlist_intersection(invlist, PL_UpperLatin1, &only_utf8);
20366 _invlist_subtract(invlist, only_utf8, &invlist);
20369 /* And this flag for matching all non-ASCII 0xFF and below */
20370 if (flags & ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER)
20372 not_utf8 = invlist_clone(PL_UpperLatin1);
20375 else if (OP(node) == ANYOFL) {
20377 /* If either of these flags are set, what matches isn't
20378 * determinable except during execution, so don't know enough here
20380 if (flags & (ANYOFL_FOLD|ANYOF_MATCHES_POSIXL)) {
20381 inverting_allowed = FALSE;
20384 /* What the posix classes match also varies at runtime, so these
20385 * will be output symbolically. */
20386 if (ANYOF_POSIXL_TEST_ANY_SET(node)) {
20389 posixes = newSVpvs("");
20390 for (i = 0; i < ANYOF_POSIXL_MAX; i++) {
20391 if (ANYOF_POSIXL_TEST(node,i)) {
20392 sv_catpv(posixes, anyofs[i]);
20399 /* Accumulate the bit map into the unconditional match list */
20400 for (i = 0; i < NUM_ANYOF_CODE_POINTS; i++) {
20401 if (BITMAP_TEST(bitmap, i)) {
20403 for (; i < NUM_ANYOF_CODE_POINTS && BITMAP_TEST(bitmap, i); i++) {
20406 invlist = _add_range_to_invlist(invlist, start, i-1);
20410 /* Make sure that the conditional match lists don't have anything in them
20411 * that match unconditionally; otherwise the output is quite confusing.
20412 * This could happen if the code that populates these misses some
20415 _invlist_subtract(only_utf8, invlist, &only_utf8);
20418 _invlist_subtract(not_utf8, invlist, ¬_utf8);
20421 if (only_utf8_locale_invlist) {
20423 /* Since this list is passed in, we have to make a copy before
20425 only_utf8_locale = invlist_clone(only_utf8_locale_invlist);
20427 _invlist_subtract(only_utf8_locale, invlist, &only_utf8_locale);
20429 /* And, it can get really weird for us to try outputting an inverted
20430 * form of this list when it has things above the bitmap, so don't even
20432 if (invlist_highest(only_utf8_locale) >= NUM_ANYOF_CODE_POINTS) {
20433 inverting_allowed = FALSE;
20437 /* Calculate what the output would be if we take the input as-is */
20438 as_is_display = put_charclass_bitmap_innards_common(invlist,
20445 /* If have to take the output as-is, just do that */
20446 if (! inverting_allowed) {
20447 if (as_is_display) {
20448 sv_catsv(sv, as_is_display);
20449 SvREFCNT_dec_NN(as_is_display);
20452 else { /* But otherwise, create the output again on the inverted input, and
20453 use whichever version is shorter */
20455 int inverted_bias, as_is_bias;
20457 /* We will apply our bias to whichever of the the results doesn't have
20467 inverted_bias = bias;
20470 /* Now invert each of the lists that contribute to the output,
20471 * excluding from the result things outside the possible range */
20473 /* For the unconditional inversion list, we have to add in all the
20474 * conditional code points, so that when inverted, they will be gone
20476 _invlist_union(only_utf8, invlist, &invlist);
20477 _invlist_union(not_utf8, invlist, &invlist);
20478 _invlist_union(only_utf8_locale, invlist, &invlist);
20479 _invlist_invert(invlist);
20480 _invlist_intersection(invlist, PL_InBitmap, &invlist);
20483 _invlist_invert(only_utf8);
20484 _invlist_intersection(only_utf8, PL_UpperLatin1, &only_utf8);
20486 else if (not_utf8) {
20488 /* If a code point matches iff the target string is not in UTF-8,
20489 * then complementing the result has it not match iff not in UTF-8,
20490 * which is the same thing as matching iff it is UTF-8. */
20491 only_utf8 = not_utf8;
20495 if (only_utf8_locale) {
20496 _invlist_invert(only_utf8_locale);
20497 _invlist_intersection(only_utf8_locale,
20499 &only_utf8_locale);
20502 inverted_display = put_charclass_bitmap_innards_common(
20507 only_utf8_locale, invert);
20509 /* Use the shortest representation, taking into account our bias
20510 * against showing it inverted */
20511 if ( inverted_display
20512 && ( ! as_is_display
20513 || ( SvCUR(inverted_display) + inverted_bias
20514 < SvCUR(as_is_display) + as_is_bias)))
20516 sv_catsv(sv, inverted_display);
20518 else if (as_is_display) {
20519 sv_catsv(sv, as_is_display);
20522 SvREFCNT_dec(as_is_display);
20523 SvREFCNT_dec(inverted_display);
20526 SvREFCNT_dec_NN(invlist);
20527 SvREFCNT_dec(only_utf8);
20528 SvREFCNT_dec(not_utf8);
20529 SvREFCNT_dec(posixes);
20530 SvREFCNT_dec(only_utf8_locale);
20532 return SvCUR(sv) > orig_sv_cur;
20535 #define CLEAR_OPTSTART \
20536 if (optstart) STMT_START { \
20537 DEBUG_OPTIMISE_r(Perl_re_printf( aTHX_ \
20538 " (%" IVdf " nodes)\n", (IV)(node - optstart))); \
20542 #define DUMPUNTIL(b,e) \
20544 node=dumpuntil(r,start,(b),(e),last,sv,indent+1,depth+1);
20546 STATIC const regnode *
20547 S_dumpuntil(pTHX_ const regexp *r, const regnode *start, const regnode *node,
20548 const regnode *last, const regnode *plast,
20549 SV* sv, I32 indent, U32 depth)
20551 U8 op = PSEUDO; /* Arbitrary non-END op. */
20552 const regnode *next;
20553 const regnode *optstart= NULL;
20555 RXi_GET_DECL(r,ri);
20556 GET_RE_DEBUG_FLAGS_DECL;
20558 PERL_ARGS_ASSERT_DUMPUNTIL;
20560 #ifdef DEBUG_DUMPUNTIL
20561 Perl_re_printf( aTHX_ "--- %d : %d - %d - %d\n",indent,node-start,
20562 last ? last-start : 0,plast ? plast-start : 0);
20565 if (plast && plast < last)
20568 while (PL_regkind[op] != END && (!last || node < last)) {
20570 /* While that wasn't END last time... */
20573 if (op == CLOSE || op == WHILEM)
20575 next = regnext((regnode *)node);
20578 if (OP(node) == OPTIMIZED) {
20579 if (!optstart && RE_DEBUG_FLAG(RE_DEBUG_COMPILE_OPTIMISE))
20586 regprop(r, sv, node, NULL, NULL);
20587 Perl_re_printf( aTHX_ "%4" IVdf ":%*s%s", (IV)(node - start),
20588 (int)(2*indent + 1), "", SvPVX_const(sv));
20590 if (OP(node) != OPTIMIZED) {
20591 if (next == NULL) /* Next ptr. */
20592 Perl_re_printf( aTHX_ " (0)");
20593 else if (PL_regkind[(U8)op] == BRANCH
20594 && PL_regkind[OP(next)] != BRANCH )
20595 Perl_re_printf( aTHX_ " (FAIL)");
20597 Perl_re_printf( aTHX_ " (%" IVdf ")", (IV)(next - start));
20598 Perl_re_printf( aTHX_ "\n");
20602 if (PL_regkind[(U8)op] == BRANCHJ) {
20605 const regnode *nnode = (OP(next) == LONGJMP
20606 ? regnext((regnode *)next)
20608 if (last && nnode > last)
20610 DUMPUNTIL(NEXTOPER(NEXTOPER(node)), nnode);
20613 else if (PL_regkind[(U8)op] == BRANCH) {
20615 DUMPUNTIL(NEXTOPER(node), next);
20617 else if ( PL_regkind[(U8)op] == TRIE ) {
20618 const regnode *this_trie = node;
20619 const char op = OP(node);
20620 const U32 n = ARG(node);
20621 const reg_ac_data * const ac = op>=AHOCORASICK ?
20622 (reg_ac_data *)ri->data->data[n] :
20624 const reg_trie_data * const trie =
20625 (reg_trie_data*)ri->data->data[op<AHOCORASICK ? n : ac->trie];
20627 AV *const trie_words
20628 = MUTABLE_AV(ri->data->data[n + TRIE_WORDS_OFFSET]);
20630 const regnode *nextbranch= NULL;
20633 for (word_idx= 0; word_idx < (I32)trie->wordcount; word_idx++) {
20634 SV ** const elem_ptr = av_fetch(trie_words,word_idx,0);
20636 Perl_re_indentf( aTHX_ "%s ",
20639 ? pv_pretty(sv, SvPV_nolen_const(*elem_ptr),
20640 SvCUR(*elem_ptr), 60,
20641 PL_colors[0], PL_colors[1],
20643 ? PERL_PV_ESCAPE_UNI
20645 | PERL_PV_PRETTY_ELLIPSES
20646 | PERL_PV_PRETTY_LTGT
20651 U16 dist= trie->jump[word_idx+1];
20652 Perl_re_printf( aTHX_ "(%" UVuf ")\n",
20653 (UV)((dist ? this_trie + dist : next) - start));
20656 nextbranch= this_trie + trie->jump[0];
20657 DUMPUNTIL(this_trie + dist, nextbranch);
20659 if (nextbranch && PL_regkind[OP(nextbranch)]==BRANCH)
20660 nextbranch= regnext((regnode *)nextbranch);
20662 Perl_re_printf( aTHX_ "\n");
20665 if (last && next > last)
20670 else if ( op == CURLY ) { /* "next" might be very big: optimizer */
20671 DUMPUNTIL(NEXTOPER(node) + EXTRA_STEP_2ARGS,
20672 NEXTOPER(node) + EXTRA_STEP_2ARGS + 1);
20674 else if (PL_regkind[(U8)op] == CURLY && op != CURLYX) {
20676 DUMPUNTIL(NEXTOPER(node) + EXTRA_STEP_2ARGS, next);
20678 else if ( op == PLUS || op == STAR) {
20679 DUMPUNTIL(NEXTOPER(node), NEXTOPER(node) + 1);
20681 else if (PL_regkind[(U8)op] == ANYOF) {
20682 /* arglen 1 + class block */
20683 node += 1 + ((ANYOF_FLAGS(node) & ANYOF_MATCHES_POSIXL)
20684 ? ANYOF_POSIXL_SKIP
20686 node = NEXTOPER(node);
20688 else if (PL_regkind[(U8)op] == EXACT) {
20689 /* Literal string, where present. */
20690 node += NODE_SZ_STR(node) - 1;
20691 node = NEXTOPER(node);
20694 node = NEXTOPER(node);
20695 node += regarglen[(U8)op];
20697 if (op == CURLYX || op == OPEN)
20701 #ifdef DEBUG_DUMPUNTIL
20702 Perl_re_printf( aTHX_ "--- %d\n", (int)indent);
20707 #endif /* DEBUGGING */
20710 * ex: set ts=8 sts=4 sw=4 et: