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
78 #ifdef PERL_IN_XSUB_RE
80 EXTERN_C const struct regexp_engine my_reg_engine;
85 #include "dquote_inline.h"
86 #include "invlist_inline.h"
87 #include "unicode_constants.h"
89 #define HAS_NONLATIN1_FOLD_CLOSURE(i) \
90 _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(i)
91 #define HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE(i) \
92 _HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(i)
93 #define IS_NON_FINAL_FOLD(c) _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c)
94 #define IS_IN_SOME_FOLD_L1(c) _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c)
100 /* this is a chain of data about sub patterns we are processing that
101 need to be handled separately/specially in study_chunk. Its so
102 we can simulate recursion without losing state. */
104 typedef struct scan_frame {
105 regnode *last_regnode; /* last node to process in this frame */
106 regnode *next_regnode; /* next node to process when last is reached */
107 U32 prev_recursed_depth;
108 I32 stopparen; /* what stopparen do we use */
110 struct scan_frame *this_prev_frame; /* this previous frame */
111 struct scan_frame *prev_frame; /* previous frame */
112 struct scan_frame *next_frame; /* next frame */
115 /* Certain characters are output as a sequence with the first being a
117 #define isBACKSLASHED_PUNCT(c) strchr("-[]\\^", c)
120 struct RExC_state_t {
121 U32 flags; /* RXf_* are we folding, multilining? */
122 U32 pm_flags; /* PMf_* stuff from the calling PMOP */
123 char *precomp; /* uncompiled string. */
124 char *precomp_end; /* pointer to end of uncompiled string. */
125 REGEXP *rx_sv; /* The SV that is the regexp. */
126 regexp *rx; /* perl core regexp structure */
127 regexp_internal *rxi; /* internal data for regexp object
129 char *start; /* Start of input for compile */
130 char *end; /* End of input for compile */
131 char *parse; /* Input-scan pointer. */
132 char *copy_start; /* start of copy of input within
133 constructed parse string */
134 char *save_copy_start; /* Provides one level of saving
135 and restoring 'copy_start' */
136 char *copy_start_in_input; /* Position in input string
137 corresponding to copy_start */
138 SSize_t whilem_seen; /* number of WHILEM in this expr */
139 regnode *emit_start; /* Start of emitted-code area */
140 regnode_offset emit; /* Code-emit pointer */
141 I32 naughty; /* How bad is this pattern? */
142 I32 sawback; /* Did we see \1, ...? */
144 SSize_t size; /* Number of regnode equivalents in
147 /* position beyond 'precomp' of the warning message furthest away from
148 * 'precomp'. During the parse, no warnings are raised for any problems
149 * earlier in the parse than this position. This works if warnings are
150 * raised the first time a given spot is parsed, and if only one
151 * independent warning is raised for any given spot */
152 Size_t latest_warn_offset;
154 I32 npar; /* Capture buffer count so far in the
155 parse, (OPEN) plus one. ("par" 0 is
157 I32 total_par; /* During initial parse, is either 0,
158 or -1; the latter indicating a
159 reparse is needed. After that pass,
160 it is what 'npar' became after the
161 pass. Hence, it being > 0 indicates
162 we are in a reparse situation */
163 I32 nestroot; /* root parens we are in - used by
166 regnode_offset *open_parens; /* offsets to open parens */
167 regnode_offset *close_parens; /* offsets to close parens */
168 I32 parens_buf_size; /* #slots malloced open/close_parens */
169 regnode *end_op; /* END node in program */
170 I32 utf8; /* whether the pattern is utf8 or not */
171 I32 orig_utf8; /* whether the pattern was originally in utf8 */
172 /* XXX use this for future optimisation of case
173 * where pattern must be upgraded to utf8. */
174 I32 uni_semantics; /* If a d charset modifier should use unicode
175 rules, even if the pattern is not in
177 HV *paren_names; /* Paren names */
179 regnode **recurse; /* Recurse regops */
180 I32 recurse_count; /* Number of recurse regops we have generated */
181 U8 *study_chunk_recursed; /* bitmap of which subs we have moved
183 U32 study_chunk_recursed_bytes; /* bytes in bitmap */
187 I32 override_recoding;
188 I32 recode_x_to_native;
189 I32 in_multi_char_class;
190 struct reg_code_blocks *code_blocks;/* positions of literal (?{})
192 int code_index; /* next code_blocks[] slot */
193 SSize_t maxlen; /* mininum possible number of chars in string to match */
194 scan_frame *frame_head;
195 scan_frame *frame_last;
199 #ifdef ADD_TO_REGEXEC
200 char *starttry; /* -Dr: where regtry was called. */
201 #define RExC_starttry (pRExC_state->starttry)
203 SV *runtime_code_qr; /* qr with the runtime code blocks */
205 const char *lastparse;
207 AV *paren_name_list; /* idx -> name */
208 U32 study_chunk_recursed_count;
212 #define RExC_lastparse (pRExC_state->lastparse)
213 #define RExC_lastnum (pRExC_state->lastnum)
214 #define RExC_paren_name_list (pRExC_state->paren_name_list)
215 #define RExC_study_chunk_recursed_count (pRExC_state->study_chunk_recursed_count)
216 #define RExC_mysv (pRExC_state->mysv1)
217 #define RExC_mysv1 (pRExC_state->mysv1)
218 #define RExC_mysv2 (pRExC_state->mysv2)
228 #define RExC_flags (pRExC_state->flags)
229 #define RExC_pm_flags (pRExC_state->pm_flags)
230 #define RExC_precomp (pRExC_state->precomp)
231 #define RExC_copy_start_in_input (pRExC_state->copy_start_in_input)
232 #define RExC_copy_start_in_constructed (pRExC_state->copy_start)
233 #define RExC_save_copy_start_in_constructed (pRExC_state->save_copy_start)
234 #define RExC_precomp_end (pRExC_state->precomp_end)
235 #define RExC_rx_sv (pRExC_state->rx_sv)
236 #define RExC_rx (pRExC_state->rx)
237 #define RExC_rxi (pRExC_state->rxi)
238 #define RExC_start (pRExC_state->start)
239 #define RExC_end (pRExC_state->end)
240 #define RExC_parse (pRExC_state->parse)
241 #define RExC_latest_warn_offset (pRExC_state->latest_warn_offset )
242 #define RExC_whilem_seen (pRExC_state->whilem_seen)
243 #define RExC_seen_d_op (pRExC_state->seen_d_op) /* Seen something that differs
244 under /d from /u ? */
246 #ifdef RE_TRACK_PATTERN_OFFSETS
247 # define RExC_offsets (RExC_rxi->u.offsets) /* I am not like the
250 #define RExC_emit (pRExC_state->emit)
251 #define RExC_emit_start (pRExC_state->emit_start)
252 #define RExC_sawback (pRExC_state->sawback)
253 #define RExC_seen (pRExC_state->seen)
254 #define RExC_size (pRExC_state->size)
255 #define RExC_maxlen (pRExC_state->maxlen)
256 #define RExC_npar (pRExC_state->npar)
257 #define RExC_total_parens (pRExC_state->total_par)
258 #define RExC_parens_buf_size (pRExC_state->parens_buf_size)
259 #define RExC_nestroot (pRExC_state->nestroot)
260 #define RExC_seen_zerolen (pRExC_state->seen_zerolen)
261 #define RExC_utf8 (pRExC_state->utf8)
262 #define RExC_uni_semantics (pRExC_state->uni_semantics)
263 #define RExC_orig_utf8 (pRExC_state->orig_utf8)
264 #define RExC_open_parens (pRExC_state->open_parens)
265 #define RExC_close_parens (pRExC_state->close_parens)
266 #define RExC_end_op (pRExC_state->end_op)
267 #define RExC_paren_names (pRExC_state->paren_names)
268 #define RExC_recurse (pRExC_state->recurse)
269 #define RExC_recurse_count (pRExC_state->recurse_count)
270 #define RExC_study_chunk_recursed (pRExC_state->study_chunk_recursed)
271 #define RExC_study_chunk_recursed_bytes \
272 (pRExC_state->study_chunk_recursed_bytes)
273 #define RExC_in_lookbehind (pRExC_state->in_lookbehind)
274 #define RExC_in_lookahead (pRExC_state->in_lookahead)
275 #define RExC_contains_locale (pRExC_state->contains_locale)
276 #define RExC_recode_x_to_native (pRExC_state->recode_x_to_native)
279 # define SET_recode_x_to_native(x) \
280 STMT_START { RExC_recode_x_to_native = (x); } STMT_END
282 # define SET_recode_x_to_native(x) NOOP
285 #define RExC_in_multi_char_class (pRExC_state->in_multi_char_class)
286 #define RExC_frame_head (pRExC_state->frame_head)
287 #define RExC_frame_last (pRExC_state->frame_last)
288 #define RExC_frame_count (pRExC_state->frame_count)
289 #define RExC_strict (pRExC_state->strict)
290 #define RExC_study_started (pRExC_state->study_started)
291 #define RExC_warn_text (pRExC_state->warn_text)
292 #define RExC_in_script_run (pRExC_state->in_script_run)
293 #define RExC_use_BRANCHJ (pRExC_state->use_BRANCHJ)
294 #define RExC_unlexed_names (pRExC_state->unlexed_names)
296 /* Heuristic check on the complexity of the pattern: if TOO_NAUGHTY, we set
297 * a flag to disable back-off on the fixed/floating substrings - if it's
298 * a high complexity pattern we assume the benefit of avoiding a full match
299 * is worth the cost of checking for the substrings even if they rarely help.
301 #define RExC_naughty (pRExC_state->naughty)
302 #define TOO_NAUGHTY (10)
303 #define MARK_NAUGHTY(add) \
304 if (RExC_naughty < TOO_NAUGHTY) \
305 RExC_naughty += (add)
306 #define MARK_NAUGHTY_EXP(exp, add) \
307 if (RExC_naughty < TOO_NAUGHTY) \
308 RExC_naughty += RExC_naughty / (exp) + (add)
310 #define ISMULT1(c) ((c) == '*' || (c) == '+' || (c) == '?')
311 #define ISMULT2(s) ((*s) == '*' || (*s) == '+' || (*s) == '?' || \
312 ((*s) == '{' && regcurly(s)))
315 * Flags to be passed up and down.
317 #define WORST 0 /* Worst case. */
318 #define HASWIDTH 0x01 /* Known to not match null strings, could match
321 /* Simple enough to be STAR/PLUS operand; in an EXACTish node must be a single
322 * character. (There needs to be a case: in the switch statement in regexec.c
323 * for any node marked SIMPLE.) Note that this is not the same thing as
326 #define SPSTART 0x04 /* Starts with * or + */
327 #define POSTPONED 0x08 /* (?1),(?&name), (??{...}) or similar */
328 #define TRYAGAIN 0x10 /* Weeded out a declaration. */
329 #define RESTART_PARSE 0x20 /* Need to redo the parse */
330 #define NEED_UTF8 0x40 /* In conjunction with RESTART_PARSE, need to
331 calcuate sizes as UTF-8 */
333 #define REG_NODE_NUM(x) ((x) ? (int)((x)-RExC_emit_start) : -1)
335 /* whether trie related optimizations are enabled */
336 #if PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION
337 #define TRIE_STUDY_OPT
338 #define FULL_TRIE_STUDY
344 #define PBYTE(u8str,paren) ((U8*)(u8str))[(paren) >> 3]
345 #define PBITVAL(paren) (1 << ((paren) & 7))
346 #define PAREN_TEST(u8str,paren) ( PBYTE(u8str,paren) & PBITVAL(paren))
347 #define PAREN_SET(u8str,paren) PBYTE(u8str,paren) |= PBITVAL(paren)
348 #define PAREN_UNSET(u8str,paren) PBYTE(u8str,paren) &= (~PBITVAL(paren))
350 #define REQUIRE_UTF8(flagp) STMT_START { \
352 *flagp = RESTART_PARSE|NEED_UTF8; \
357 /* Change from /d into /u rules, and restart the parse. RExC_uni_semantics is
358 * a flag that indicates we need to override /d with /u as a result of
359 * something in the pattern. It should only be used in regards to calling
360 * set_regex_charset() or get_regex_charse() */
361 #define REQUIRE_UNI_RULES(flagp, restart_retval) \
363 if (DEPENDS_SEMANTICS) { \
364 set_regex_charset(&RExC_flags, REGEX_UNICODE_CHARSET); \
365 RExC_uni_semantics = 1; \
366 if (RExC_seen_d_op && LIKELY(! IN_PARENS_PASS)) { \
367 /* No need to restart the parse if we haven't seen \
368 * anything that differs between /u and /d, and no need \
369 * to restart immediately if we're going to reparse \
370 * anyway to count parens */ \
371 *flagp |= RESTART_PARSE; \
372 return restart_retval; \
377 #define REQUIRE_BRANCHJ(flagp, restart_retval) \
379 RExC_use_BRANCHJ = 1; \
380 *flagp |= RESTART_PARSE; \
381 return restart_retval; \
384 /* Until we have completed the parse, we leave RExC_total_parens at 0 or
385 * less. After that, it must always be positive, because the whole re is
386 * considered to be surrounded by virtual parens. Setting it to negative
387 * indicates there is some construct that needs to know the actual number of
388 * parens to be properly handled. And that means an extra pass will be
389 * required after we've counted them all */
390 #define ALL_PARENS_COUNTED (RExC_total_parens > 0)
391 #define REQUIRE_PARENS_PASS \
392 STMT_START { /* No-op if have completed a pass */ \
393 if (! ALL_PARENS_COUNTED) RExC_total_parens = -1; \
395 #define IN_PARENS_PASS (RExC_total_parens < 0)
398 /* This is used to return failure (zero) early from the calling function if
399 * various flags in 'flags' are set. Two flags always cause a return:
400 * 'RESTART_PARSE' and 'NEED_UTF8'. 'extra' can be used to specify any
401 * additional flags that should cause a return; 0 if none. If the return will
402 * be done, '*flagp' is first set to be all of the flags that caused the
404 #define RETURN_FAIL_ON_RESTART_OR_FLAGS(flags,flagp,extra) \
406 if ((flags) & (RESTART_PARSE|NEED_UTF8|(extra))) { \
407 *(flagp) = (flags) & (RESTART_PARSE|NEED_UTF8|(extra)); \
412 #define MUST_RESTART(flags) ((flags) & (RESTART_PARSE))
414 #define RETURN_FAIL_ON_RESTART(flags,flagp) \
415 RETURN_FAIL_ON_RESTART_OR_FLAGS( flags, flagp, 0)
416 #define RETURN_FAIL_ON_RESTART_FLAGP(flagp) \
417 if (MUST_RESTART(*(flagp))) return 0
419 /* This converts the named class defined in regcomp.h to its equivalent class
420 * number defined in handy.h. */
421 #define namedclass_to_classnum(class) ((int) ((class) / 2))
422 #define classnum_to_namedclass(classnum) ((classnum) * 2)
424 #define _invlist_union_complement_2nd(a, b, output) \
425 _invlist_union_maybe_complement_2nd(a, b, TRUE, output)
426 #define _invlist_intersection_complement_2nd(a, b, output) \
427 _invlist_intersection_maybe_complement_2nd(a, b, TRUE, output)
429 /* About scan_data_t.
431 During optimisation we recurse through the regexp program performing
432 various inplace (keyhole style) optimisations. In addition study_chunk
433 and scan_commit populate this data structure with information about
434 what strings MUST appear in the pattern. We look for the longest
435 string that must appear at a fixed location, and we look for the
436 longest string that may appear at a floating location. So for instance
441 Both 'FOO' and 'A' are fixed strings. Both 'B' and 'BAR' are floating
442 strings (because they follow a .* construct). study_chunk will identify
443 both FOO and BAR as being the longest fixed and floating strings respectively.
445 The strings can be composites, for instance
449 will result in a composite fixed substring 'foo'.
451 For each string some basic information is maintained:
454 This is the position the string must appear at, or not before.
455 It also implicitly (when combined with minlenp) tells us how many
456 characters must match before the string we are searching for.
457 Likewise when combined with minlenp and the length of the string it
458 tells us how many characters must appear after the string we have
462 Only used for floating strings. This is the rightmost point that
463 the string can appear at. If set to SSize_t_MAX it indicates that the
464 string can occur infinitely far to the right.
465 For fixed strings, it is equal to min_offset.
468 A pointer to the minimum number of characters of the pattern that the
469 string was found inside. This is important as in the case of positive
470 lookahead or positive lookbehind we can have multiple patterns
475 The minimum length of the pattern overall is 3, the minimum length
476 of the lookahead part is 3, but the minimum length of the part that
477 will actually match is 1. So 'FOO's minimum length is 3, but the
478 minimum length for the F is 1. This is important as the minimum length
479 is used to determine offsets in front of and behind the string being
480 looked for. Since strings can be composites this is the length of the
481 pattern at the time it was committed with a scan_commit. Note that
482 the length is calculated by study_chunk, so that the minimum lengths
483 are not known until the full pattern has been compiled, thus the
484 pointer to the value.
488 In the case of lookbehind the string being searched for can be
489 offset past the start point of the final matching string.
490 If this value was just blithely removed from the min_offset it would
491 invalidate some of the calculations for how many chars must match
492 before or after (as they are derived from min_offset and minlen and
493 the length of the string being searched for).
494 When the final pattern is compiled and the data is moved from the
495 scan_data_t structure into the regexp structure the information
496 about lookbehind is factored in, with the information that would
497 have been lost precalculated in the end_shift field for the
500 The fields pos_min and pos_delta are used to store the minimum offset
501 and the delta to the maximum offset at the current point in the pattern.
505 struct scan_data_substrs {
506 SV *str; /* longest substring found in pattern */
507 SSize_t min_offset; /* earliest point in string it can appear */
508 SSize_t max_offset; /* latest point in string it can appear */
509 SSize_t *minlenp; /* pointer to the minlen relevant to the string */
510 SSize_t lookbehind; /* is the pos of the string modified by LB */
511 I32 flags; /* per substring SF_* and SCF_* flags */
514 typedef struct scan_data_t {
515 /*I32 len_min; unused */
516 /*I32 len_delta; unused */
520 SSize_t last_end; /* min value, <0 unless valid. */
521 SSize_t last_start_min;
522 SSize_t last_start_max;
523 U8 cur_is_floating; /* whether the last_* values should be set as
524 * the next fixed (0) or floating (1)
527 /* [0] is longest fixed substring so far, [1] is longest float so far */
528 struct scan_data_substrs substrs[2];
530 I32 flags; /* common SF_* and SCF_* flags */
532 SSize_t *last_closep;
533 regnode_ssc *start_class;
537 * Forward declarations for pregcomp()'s friends.
540 static const scan_data_t zero_scan_data = {
541 0, 0, NULL, 0, 0, 0, 0,
543 { NULL, 0, 0, 0, 0, 0 },
544 { NULL, 0, 0, 0, 0, 0 },
551 #define SF_BEFORE_SEOL 0x0001
552 #define SF_BEFORE_MEOL 0x0002
553 #define SF_BEFORE_EOL (SF_BEFORE_SEOL|SF_BEFORE_MEOL)
555 #define SF_IS_INF 0x0040
556 #define SF_HAS_PAR 0x0080
557 #define SF_IN_PAR 0x0100
558 #define SF_HAS_EVAL 0x0200
561 /* SCF_DO_SUBSTR is the flag that tells the regexp analyzer to track the
562 * longest substring in the pattern. When it is not set the optimiser keeps
563 * track of position, but does not keep track of the actual strings seen,
565 * So for instance /foo/ will be parsed with SCF_DO_SUBSTR being true, but
568 * Similarly, /foo.*(blah|erm|huh).*fnorble/ will have "foo" and "fnorble"
569 * parsed with SCF_DO_SUBSTR on, but while processing the (...) it will be
570 * turned off because of the alternation (BRANCH). */
571 #define SCF_DO_SUBSTR 0x0400
573 #define SCF_DO_STCLASS_AND 0x0800
574 #define SCF_DO_STCLASS_OR 0x1000
575 #define SCF_DO_STCLASS (SCF_DO_STCLASS_AND|SCF_DO_STCLASS_OR)
576 #define SCF_WHILEM_VISITED_POS 0x2000
578 #define SCF_TRIE_RESTUDY 0x4000 /* Do restudy? */
579 #define SCF_SEEN_ACCEPT 0x8000
580 #define SCF_TRIE_DOING_RESTUDY 0x10000
581 #define SCF_IN_DEFINE 0x20000
586 #define UTF cBOOL(RExC_utf8)
588 /* The enums for all these are ordered so things work out correctly */
589 #define LOC (get_regex_charset(RExC_flags) == REGEX_LOCALE_CHARSET)
590 #define DEPENDS_SEMANTICS (get_regex_charset(RExC_flags) \
591 == REGEX_DEPENDS_CHARSET)
592 #define UNI_SEMANTICS (get_regex_charset(RExC_flags) == REGEX_UNICODE_CHARSET)
593 #define AT_LEAST_UNI_SEMANTICS (get_regex_charset(RExC_flags) \
594 >= REGEX_UNICODE_CHARSET)
595 #define ASCII_RESTRICTED (get_regex_charset(RExC_flags) \
596 == REGEX_ASCII_RESTRICTED_CHARSET)
597 #define AT_LEAST_ASCII_RESTRICTED (get_regex_charset(RExC_flags) \
598 >= REGEX_ASCII_RESTRICTED_CHARSET)
599 #define ASCII_FOLD_RESTRICTED (get_regex_charset(RExC_flags) \
600 == REGEX_ASCII_MORE_RESTRICTED_CHARSET)
602 #define FOLD cBOOL(RExC_flags & RXf_PMf_FOLD)
604 /* For programs that want to be strictly Unicode compatible by dying if any
605 * attempt is made to match a non-Unicode code point against a Unicode
607 #define ALWAYS_WARN_SUPER ckDEAD(packWARN(WARN_NON_UNICODE))
609 #define OOB_NAMEDCLASS -1
611 /* There is no code point that is out-of-bounds, so this is problematic. But
612 * its only current use is to initialize a variable that is always set before
614 #define OOB_UNICODE 0xDEADBEEF
616 #define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv))
619 /* length of regex to show in messages that don't mark a position within */
620 #define RegexLengthToShowInErrorMessages 127
623 * If MARKER[12] are adjusted, be sure to adjust the constants at the top
624 * of t/op/regmesg.t, the tests in t/op/re_tests, and those in
625 * op/pragma/warn/regcomp.
627 #define MARKER1 "<-- HERE" /* marker as it appears in the description */
628 #define MARKER2 " <-- HERE " /* marker as it appears within the regex */
630 #define REPORT_LOCATION " in regex; marked by " MARKER1 \
631 " in m/%" UTF8f MARKER2 "%" UTF8f "/"
633 /* The code in this file in places uses one level of recursion with parsing
634 * rebased to an alternate string constructed by us in memory. This can take
635 * the form of something that is completely different from the input, or
636 * something that uses the input as part of the alternate. In the first case,
637 * there should be no possibility of an error, as we are in complete control of
638 * the alternate string. But in the second case we don't completely control
639 * the input portion, so there may be errors in that. Here's an example:
641 * is handled specially because \x{df} folds to a sequence of more than one
642 * character: 'ss'. What is done is to create and parse an alternate string,
643 * which looks like this:
644 * /(?:\x{DF}|[abc\x{DF}def])/ui
645 * where it uses the input unchanged in the middle of something it constructs,
646 * which is a branch for the DF outside the character class, and clustering
647 * parens around the whole thing. (It knows enough to skip the DF inside the
648 * class while in this substitute parse.) 'abc' and 'def' may have errors that
649 * need to be reported. The general situation looks like this:
651 * |<------- identical ------>|
653 * Input: ---------------------------------------------------------------
654 * Constructed: ---------------------------------------------------
656 * |<------- identical ------>|
658 * sI..eI is the portion of the input pattern we are concerned with here.
659 * sC..EC is the constructed substitute parse string.
660 * sC..tC is constructed by us
661 * tC..eC is an exact duplicate of the portion of the input pattern tI..eI.
662 * In the diagram, these are vertically aligned.
663 * eC..EC is also constructed by us.
664 * xC is the position in the substitute parse string where we found a
666 * xI is the position in the original pattern corresponding to xC.
668 * We want to display a message showing the real input string. Thus we need to
669 * translate from xC to xI. We know that xC >= tC, since the portion of the
670 * string sC..tC has been constructed by us, and so shouldn't have errors. We
672 * xI = tI + (xC - tC)
674 * When the substitute parse is constructed, the code needs to set:
677 * RExC_copy_start_in_input (tI)
678 * RExC_copy_start_in_constructed (tC)
679 * and restore them when done.
681 * During normal processing of the input pattern, both
682 * 'RExC_copy_start_in_input' and 'RExC_copy_start_in_constructed' are set to
683 * sI, so that xC equals xI.
686 #define sI RExC_precomp
687 #define eI RExC_precomp_end
688 #define sC RExC_start
690 #define tI RExC_copy_start_in_input
691 #define tC RExC_copy_start_in_constructed
692 #define xI(xC) (tI + (xC - tC))
693 #define xI_offset(xC) (xI(xC) - sI)
695 #define REPORT_LOCATION_ARGS(xC) \
697 (xI(xC) > eI) /* Don't run off end */ \
698 ? eI - sI /* Length before the <--HERE */ \
699 : ((xI_offset(xC) >= 0) \
701 : (Perl_croak(aTHX_ "panic: %s: %d: negative offset: %" \
702 IVdf " trying to output message for " \
704 __FILE__, __LINE__, (IV) xI_offset(xC), \
705 ((int) (eC - sC)), sC), 0)), \
706 sI), /* The input pattern printed up to the <--HERE */ \
708 (xI(xC) > eI) ? 0 : eI - xI(xC), /* Length after <--HERE */ \
709 (xI(xC) > eI) ? eI : xI(xC)) /* pattern after <--HERE */
711 /* Used to point after bad bytes for an error message, but avoid skipping
712 * past a nul byte. */
713 #define SKIP_IF_CHAR(s, e) (!*(s) ? 0 : UTF ? UTF8_SAFE_SKIP(s, e) : 1)
715 /* Set up to clean up after our imminent demise */
716 #define PREPARE_TO_DIE \
719 SAVEFREESV(RExC_rx_sv); \
720 if (RExC_open_parens) \
721 SAVEFREEPV(RExC_open_parens); \
722 if (RExC_close_parens) \
723 SAVEFREEPV(RExC_close_parens); \
727 * Calls SAVEDESTRUCTOR_X if needed, then calls Perl_croak with the given
728 * arg. Show regex, up to a maximum length. If it's too long, chop and add
731 #define _FAIL(code) STMT_START { \
732 const char *ellipses = ""; \
733 IV len = RExC_precomp_end - RExC_precomp; \
736 if (len > RegexLengthToShowInErrorMessages) { \
737 /* chop 10 shorter than the max, to ensure meaning of "..." */ \
738 len = RegexLengthToShowInErrorMessages - 10; \
744 #define FAIL(msg) _FAIL( \
745 Perl_croak(aTHX_ "%s in regex m/%" UTF8f "%s/", \
746 msg, UTF8fARG(UTF, len, RExC_precomp), ellipses))
748 #define FAIL2(msg,arg) _FAIL( \
749 Perl_croak(aTHX_ msg " in regex m/%" UTF8f "%s/", \
750 arg, UTF8fARG(UTF, len, RExC_precomp), ellipses))
752 #define FAIL3(msg,arg1,arg2) _FAIL( \
753 Perl_croak(aTHX_ msg " in regex m/%" UTF8f "%s/", \
754 arg1, arg2, UTF8fARG(UTF, len, RExC_precomp), ellipses))
757 * Simple_vFAIL -- like FAIL, but marks the current location in the scan
759 #define Simple_vFAIL(m) STMT_START { \
760 Perl_croak(aTHX_ "%s" REPORT_LOCATION, \
761 m, REPORT_LOCATION_ARGS(RExC_parse)); \
765 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL()
767 #define vFAIL(m) STMT_START { \
773 * Like Simple_vFAIL(), but accepts two arguments.
775 #define Simple_vFAIL2(m,a1) STMT_START { \
776 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, \
777 REPORT_LOCATION_ARGS(RExC_parse)); \
781 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL2().
783 #define vFAIL2(m,a1) STMT_START { \
785 Simple_vFAIL2(m, a1); \
790 * Like Simple_vFAIL(), but accepts three arguments.
792 #define Simple_vFAIL3(m, a1, a2) STMT_START { \
793 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, a2, \
794 REPORT_LOCATION_ARGS(RExC_parse)); \
798 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL3().
800 #define vFAIL3(m,a1,a2) STMT_START { \
802 Simple_vFAIL3(m, a1, a2); \
806 * Like Simple_vFAIL(), but accepts four arguments.
808 #define Simple_vFAIL4(m, a1, a2, a3) STMT_START { \
809 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, a2, a3, \
810 REPORT_LOCATION_ARGS(RExC_parse)); \
813 #define vFAIL4(m,a1,a2,a3) STMT_START { \
815 Simple_vFAIL4(m, a1, a2, a3); \
818 /* A specialized version of vFAIL2 that works with UTF8f */
819 #define vFAIL2utf8f(m, a1) STMT_START { \
821 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, \
822 REPORT_LOCATION_ARGS(RExC_parse)); \
825 #define vFAIL3utf8f(m, a1, a2) STMT_START { \
827 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, a2, \
828 REPORT_LOCATION_ARGS(RExC_parse)); \
831 /* Setting this to NULL is a signal to not output warnings */
832 #define TURN_OFF_WARNINGS_IN_SUBSTITUTE_PARSE \
834 RExC_save_copy_start_in_constructed = RExC_copy_start_in_constructed;\
835 RExC_copy_start_in_constructed = NULL; \
837 #define RESTORE_WARNINGS \
838 RExC_copy_start_in_constructed = RExC_save_copy_start_in_constructed
840 /* Since a warning can be generated multiple times as the input is reparsed, we
841 * output it the first time we come to that point in the parse, but suppress it
842 * otherwise. 'RExC_copy_start_in_constructed' being NULL is a flag to not
843 * generate any warnings */
844 #define TO_OUTPUT_WARNINGS(loc) \
845 ( RExC_copy_start_in_constructed \
846 && ((xI(loc)) - RExC_precomp) > (Ptrdiff_t) RExC_latest_warn_offset)
848 /* After we've emitted a warning, we save the position in the input so we don't
850 #define UPDATE_WARNINGS_LOC(loc) \
852 if (TO_OUTPUT_WARNINGS(loc)) { \
853 RExC_latest_warn_offset = (xI(loc)) - RExC_precomp; \
857 /* 'warns' is the output of the packWARNx macro used in 'code' */
858 #define _WARN_HELPER(loc, warns, code) \
860 if (! RExC_copy_start_in_constructed) { \
861 Perl_croak( aTHX_ "panic! %s: %d: Tried to warn when none" \
862 " expected at '%s'", \
863 __FILE__, __LINE__, loc); \
865 if (TO_OUTPUT_WARNINGS(loc)) { \
869 UPDATE_WARNINGS_LOC(loc); \
873 /* m is not necessarily a "literal string", in this macro */
874 #define reg_warn_non_literal_string(loc, m) \
875 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
876 Perl_warner(aTHX_ packWARN(WARN_REGEXP), \
877 "%s" REPORT_LOCATION, \
878 m, REPORT_LOCATION_ARGS(loc)))
880 #define ckWARNreg(loc,m) \
881 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
882 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), \
884 REPORT_LOCATION_ARGS(loc)))
886 #define vWARN(loc, m) \
887 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
888 Perl_warner(aTHX_ packWARN(WARN_REGEXP), \
890 REPORT_LOCATION_ARGS(loc))) \
892 #define vWARN_dep(loc, m) \
893 _WARN_HELPER(loc, packWARN(WARN_DEPRECATED), \
894 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), \
896 REPORT_LOCATION_ARGS(loc)))
898 #define ckWARNdep(loc,m) \
899 _WARN_HELPER(loc, packWARN(WARN_DEPRECATED), \
900 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
902 REPORT_LOCATION_ARGS(loc)))
904 #define ckWARNregdep(loc,m) \
905 _WARN_HELPER(loc, packWARN2(WARN_DEPRECATED, WARN_REGEXP), \
906 Perl_ck_warner_d(aTHX_ packWARN2(WARN_DEPRECATED, \
909 REPORT_LOCATION_ARGS(loc)))
911 #define ckWARN2reg_d(loc,m, a1) \
912 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
913 Perl_ck_warner_d(aTHX_ packWARN(WARN_REGEXP), \
915 a1, REPORT_LOCATION_ARGS(loc)))
917 #define ckWARN2reg(loc, m, a1) \
918 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
919 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), \
921 a1, REPORT_LOCATION_ARGS(loc)))
923 #define vWARN3(loc, m, a1, a2) \
924 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
925 Perl_warner(aTHX_ packWARN(WARN_REGEXP), \
927 a1, a2, REPORT_LOCATION_ARGS(loc)))
929 #define ckWARN3reg(loc, m, a1, a2) \
930 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
931 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), \
934 REPORT_LOCATION_ARGS(loc)))
936 #define vWARN4(loc, m, a1, a2, a3) \
937 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
938 Perl_warner(aTHX_ packWARN(WARN_REGEXP), \
941 REPORT_LOCATION_ARGS(loc)))
943 #define ckWARN4reg(loc, m, a1, a2, a3) \
944 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
945 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), \
948 REPORT_LOCATION_ARGS(loc)))
950 #define vWARN5(loc, m, a1, a2, a3, a4) \
951 _WARN_HELPER(loc, packWARN(WARN_REGEXP), \
952 Perl_warner(aTHX_ packWARN(WARN_REGEXP), \
955 REPORT_LOCATION_ARGS(loc)))
957 #define ckWARNexperimental(loc, class, m) \
958 _WARN_HELPER(loc, packWARN(class), \
959 Perl_ck_warner_d(aTHX_ packWARN(class), \
961 REPORT_LOCATION_ARGS(loc)))
963 /* Convert between a pointer to a node and its offset from the beginning of the
965 #define REGNODE_p(offset) (RExC_emit_start + (offset))
966 #define REGNODE_OFFSET(node) ((node) - RExC_emit_start)
968 /* Macros for recording node offsets. 20001227 mjd@plover.com
969 * Nodes are numbered 1, 2, 3, 4. Node #n's position is recorded in
970 * element 2*n-1 of the array. Element #2n holds the byte length node #n.
971 * Element 0 holds the number n.
972 * Position is 1 indexed.
974 #ifndef RE_TRACK_PATTERN_OFFSETS
975 #define Set_Node_Offset_To_R(offset,byte)
976 #define Set_Node_Offset(node,byte)
977 #define Set_Cur_Node_Offset
978 #define Set_Node_Length_To_R(node,len)
979 #define Set_Node_Length(node,len)
980 #define Set_Node_Cur_Length(node,start)
981 #define Node_Offset(n)
982 #define Node_Length(n)
983 #define Set_Node_Offset_Length(node,offset,len)
984 #define ProgLen(ri) ri->u.proglen
985 #define SetProgLen(ri,x) ri->u.proglen = x
986 #define Track_Code(code)
988 #define ProgLen(ri) ri->u.offsets[0]
989 #define SetProgLen(ri,x) ri->u.offsets[0] = x
990 #define Set_Node_Offset_To_R(offset,byte) STMT_START { \
991 MJD_OFFSET_DEBUG(("** (%d) offset of node %d is %d.\n", \
992 __LINE__, (int)(offset), (int)(byte))); \
994 Perl_croak(aTHX_ "value of node is %d in Offset macro", \
997 RExC_offsets[2*(offset)-1] = (byte); \
1001 #define Set_Node_Offset(node,byte) \
1002 Set_Node_Offset_To_R(REGNODE_OFFSET(node), (byte)-RExC_start)
1003 #define Set_Cur_Node_Offset Set_Node_Offset(RExC_emit, RExC_parse)
1005 #define Set_Node_Length_To_R(node,len) STMT_START { \
1006 MJD_OFFSET_DEBUG(("** (%d) size of node %d is %d.\n", \
1007 __LINE__, (int)(node), (int)(len))); \
1009 Perl_croak(aTHX_ "value of node is %d in Length macro", \
1012 RExC_offsets[2*(node)] = (len); \
1016 #define Set_Node_Length(node,len) \
1017 Set_Node_Length_To_R(REGNODE_OFFSET(node), len)
1018 #define Set_Node_Cur_Length(node, start) \
1019 Set_Node_Length(node, RExC_parse - start)
1021 /* Get offsets and lengths */
1022 #define Node_Offset(n) (RExC_offsets[2*(REGNODE_OFFSET(n))-1])
1023 #define Node_Length(n) (RExC_offsets[2*(REGNODE_OFFSET(n))])
1025 #define Set_Node_Offset_Length(node,offset,len) STMT_START { \
1026 Set_Node_Offset_To_R(REGNODE_OFFSET(node), (offset)); \
1027 Set_Node_Length_To_R(REGNODE_OFFSET(node), (len)); \
1030 #define Track_Code(code) STMT_START { code } STMT_END
1033 #if PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS
1034 #define EXPERIMENTAL_INPLACESCAN
1035 #endif /*PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS*/
1039 Perl_re_printf(pTHX_ const char *fmt, ...)
1043 PerlIO *f= Perl_debug_log;
1044 PERL_ARGS_ASSERT_RE_PRINTF;
1046 result = PerlIO_vprintf(f, fmt, ap);
1052 Perl_re_indentf(pTHX_ const char *fmt, U32 depth, ...)
1056 PerlIO *f= Perl_debug_log;
1057 PERL_ARGS_ASSERT_RE_INDENTF;
1058 va_start(ap, depth);
1059 PerlIO_printf(f, "%*s", ( (int)depth % 20 ) * 2, "");
1060 result = PerlIO_vprintf(f, fmt, ap);
1064 #endif /* DEBUGGING */
1066 #define DEBUG_RExC_seen() \
1067 DEBUG_OPTIMISE_MORE_r({ \
1068 Perl_re_printf( aTHX_ "RExC_seen: "); \
1070 if (RExC_seen & REG_ZERO_LEN_SEEN) \
1071 Perl_re_printf( aTHX_ "REG_ZERO_LEN_SEEN "); \
1073 if (RExC_seen & REG_LOOKBEHIND_SEEN) \
1074 Perl_re_printf( aTHX_ "REG_LOOKBEHIND_SEEN "); \
1076 if (RExC_seen & REG_GPOS_SEEN) \
1077 Perl_re_printf( aTHX_ "REG_GPOS_SEEN "); \
1079 if (RExC_seen & REG_RECURSE_SEEN) \
1080 Perl_re_printf( aTHX_ "REG_RECURSE_SEEN "); \
1082 if (RExC_seen & REG_TOP_LEVEL_BRANCHES_SEEN) \
1083 Perl_re_printf( aTHX_ "REG_TOP_LEVEL_BRANCHES_SEEN "); \
1085 if (RExC_seen & REG_VERBARG_SEEN) \
1086 Perl_re_printf( aTHX_ "REG_VERBARG_SEEN "); \
1088 if (RExC_seen & REG_CUTGROUP_SEEN) \
1089 Perl_re_printf( aTHX_ "REG_CUTGROUP_SEEN "); \
1091 if (RExC_seen & REG_RUN_ON_COMMENT_SEEN) \
1092 Perl_re_printf( aTHX_ "REG_RUN_ON_COMMENT_SEEN "); \
1094 if (RExC_seen & REG_UNFOLDED_MULTI_SEEN) \
1095 Perl_re_printf( aTHX_ "REG_UNFOLDED_MULTI_SEEN "); \
1097 if (RExC_seen & REG_UNBOUNDED_QUANTIFIER_SEEN) \
1098 Perl_re_printf( aTHX_ "REG_UNBOUNDED_QUANTIFIER_SEEN "); \
1100 Perl_re_printf( aTHX_ "\n"); \
1103 #define DEBUG_SHOW_STUDY_FLAG(flags,flag) \
1104 if ((flags) & flag) Perl_re_printf( aTHX_ "%s ", #flag)
1109 S_debug_show_study_flags(pTHX_ U32 flags, const char *open_str,
1110 const char *close_str)
1115 Perl_re_printf( aTHX_ "%s", open_str);
1116 DEBUG_SHOW_STUDY_FLAG(flags, SF_BEFORE_SEOL);
1117 DEBUG_SHOW_STUDY_FLAG(flags, SF_BEFORE_MEOL);
1118 DEBUG_SHOW_STUDY_FLAG(flags, SF_IS_INF);
1119 DEBUG_SHOW_STUDY_FLAG(flags, SF_HAS_PAR);
1120 DEBUG_SHOW_STUDY_FLAG(flags, SF_IN_PAR);
1121 DEBUG_SHOW_STUDY_FLAG(flags, SF_HAS_EVAL);
1122 DEBUG_SHOW_STUDY_FLAG(flags, SCF_DO_SUBSTR);
1123 DEBUG_SHOW_STUDY_FLAG(flags, SCF_DO_STCLASS_AND);
1124 DEBUG_SHOW_STUDY_FLAG(flags, SCF_DO_STCLASS_OR);
1125 DEBUG_SHOW_STUDY_FLAG(flags, SCF_DO_STCLASS);
1126 DEBUG_SHOW_STUDY_FLAG(flags, SCF_WHILEM_VISITED_POS);
1127 DEBUG_SHOW_STUDY_FLAG(flags, SCF_TRIE_RESTUDY);
1128 DEBUG_SHOW_STUDY_FLAG(flags, SCF_SEEN_ACCEPT);
1129 DEBUG_SHOW_STUDY_FLAG(flags, SCF_TRIE_DOING_RESTUDY);
1130 DEBUG_SHOW_STUDY_FLAG(flags, SCF_IN_DEFINE);
1131 Perl_re_printf( aTHX_ "%s", close_str);
1136 S_debug_studydata(pTHX_ const char *where, scan_data_t *data,
1137 U32 depth, int is_inf)
1139 GET_RE_DEBUG_FLAGS_DECL;
1141 DEBUG_OPTIMISE_MORE_r({
1144 Perl_re_indentf(aTHX_ "%s: Pos:%" IVdf "/%" IVdf " Flags: 0x%" UVXf,
1148 (IV)data->pos_delta,
1152 S_debug_show_study_flags(aTHX_ data->flags," [","]");
1154 Perl_re_printf( aTHX_
1155 " Whilem_c: %" IVdf " Lcp: %" IVdf " %s",
1157 (IV)(data->last_closep ? *((data)->last_closep) : -1),
1158 is_inf ? "INF " : ""
1161 if (data->last_found) {
1163 Perl_re_printf(aTHX_
1164 "Last:'%s' %" IVdf ":%" IVdf "/%" IVdf,
1165 SvPVX_const(data->last_found),
1167 (IV)data->last_start_min,
1168 (IV)data->last_start_max
1171 for (i = 0; i < 2; i++) {
1172 Perl_re_printf(aTHX_
1173 " %s%s: '%s' @ %" IVdf "/%" IVdf,
1174 data->cur_is_floating == i ? "*" : "",
1175 i ? "Float" : "Fixed",
1176 SvPVX_const(data->substrs[i].str),
1177 (IV)data->substrs[i].min_offset,
1178 (IV)data->substrs[i].max_offset
1180 S_debug_show_study_flags(aTHX_ data->substrs[i].flags," [","]");
1184 Perl_re_printf( aTHX_ "\n");
1190 S_debug_peep(pTHX_ const char *str, const RExC_state_t *pRExC_state,
1191 regnode *scan, U32 depth, U32 flags)
1193 GET_RE_DEBUG_FLAGS_DECL;
1200 Next = regnext(scan);
1201 regprop(RExC_rx, RExC_mysv, scan, NULL, pRExC_state);
1202 Perl_re_indentf( aTHX_ "%s>%3d: %s (%d)",
1205 REG_NODE_NUM(scan), SvPV_nolen_const(RExC_mysv),
1206 Next ? (REG_NODE_NUM(Next)) : 0 );
1207 S_debug_show_study_flags(aTHX_ flags," [ ","]");
1208 Perl_re_printf( aTHX_ "\n");
1213 # define DEBUG_STUDYDATA(where, data, depth, is_inf) \
1214 S_debug_studydata(aTHX_ where, data, depth, is_inf)
1216 # define DEBUG_PEEP(str, scan, depth, flags) \
1217 S_debug_peep(aTHX_ str, pRExC_state, scan, depth, flags)
1220 # define DEBUG_STUDYDATA(where, data, depth, is_inf) NOOP
1221 # define DEBUG_PEEP(str, scan, depth, flags) NOOP
1225 /* =========================================================
1226 * BEGIN edit_distance stuff.
1228 * This calculates how many single character changes of any type are needed to
1229 * transform a string into another one. It is taken from version 3.1 of
1231 * https://metacpan.org/pod/Text::Levenshtein::Damerau::XS
1234 /* Our unsorted dictionary linked list. */
1235 /* Note we use UVs, not chars. */
1240 struct dictionary* next;
1242 typedef struct dictionary item;
1245 PERL_STATIC_INLINE item*
1246 push(UV key, item* curr)
1249 Newx(head, 1, item);
1257 PERL_STATIC_INLINE item*
1258 find(item* head, UV key)
1260 item* iterator = head;
1262 if (iterator->key == key){
1265 iterator = iterator->next;
1271 PERL_STATIC_INLINE item*
1272 uniquePush(item* head, UV key)
1274 item* iterator = head;
1277 if (iterator->key == key) {
1280 iterator = iterator->next;
1283 return push(key, head);
1286 PERL_STATIC_INLINE void
1287 dict_free(item* head)
1289 item* iterator = head;
1292 item* temp = iterator;
1293 iterator = iterator->next;
1300 /* End of Dictionary Stuff */
1302 /* All calculations/work are done here */
1304 S_edit_distance(const UV* src,
1306 const STRLEN x, /* length of src[] */
1307 const STRLEN y, /* length of tgt[] */
1308 const SSize_t maxDistance
1312 UV swapCount, swapScore, targetCharCount, i, j;
1314 UV score_ceil = x + y;
1316 PERL_ARGS_ASSERT_EDIT_DISTANCE;
1318 /* intialize matrix start values */
1319 Newx(scores, ( (x + 2) * (y + 2)), UV);
1320 scores[0] = score_ceil;
1321 scores[1 * (y + 2) + 0] = score_ceil;
1322 scores[0 * (y + 2) + 1] = score_ceil;
1323 scores[1 * (y + 2) + 1] = 0;
1324 head = uniquePush(uniquePush(head, src[0]), tgt[0]);
1329 for (i=1;i<=x;i++) {
1331 head = uniquePush(head, src[i]);
1332 scores[(i+1) * (y + 2) + 1] = i;
1333 scores[(i+1) * (y + 2) + 0] = score_ceil;
1336 for (j=1;j<=y;j++) {
1339 head = uniquePush(head, tgt[j]);
1340 scores[1 * (y + 2) + (j + 1)] = j;
1341 scores[0 * (y + 2) + (j + 1)] = score_ceil;
1344 targetCharCount = find(head, tgt[j-1])->value;
1345 swapScore = scores[targetCharCount * (y + 2) + swapCount] + i - targetCharCount - 1 + j - swapCount;
1347 if (src[i-1] != tgt[j-1]){
1348 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));
1352 scores[(i+1) * (y + 2) + (j + 1)] = MIN(scores[i * (y + 2) + j], swapScore);
1356 find(head, src[i-1])->value = i;
1360 IV score = scores[(x+1) * (y + 2) + (y + 1)];
1363 return (maxDistance != 0 && maxDistance < score)?(-1):score;
1367 /* END of edit_distance() stuff
1368 * ========================================================= */
1370 /* is c a control character for which we have a mnemonic? */
1371 #define isMNEMONIC_CNTRL(c) _IS_MNEMONIC_CNTRL_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c)
1374 S_cntrl_to_mnemonic(const U8 c)
1376 /* Returns the mnemonic string that represents character 'c', if one
1377 * exists; NULL otherwise. The only ones that exist for the purposes of
1378 * this routine are a few control characters */
1381 case '\a': return "\\a";
1382 case '\b': return "\\b";
1383 case ESC_NATIVE: return "\\e";
1384 case '\f': return "\\f";
1385 case '\n': return "\\n";
1386 case '\r': return "\\r";
1387 case '\t': return "\\t";
1393 /* Mark that we cannot extend a found fixed substring at this point.
1394 Update the longest found anchored substring or the longest found
1395 floating substrings if needed. */
1398 S_scan_commit(pTHX_ const RExC_state_t *pRExC_state, scan_data_t *data,
1399 SSize_t *minlenp, int is_inf)
1401 const STRLEN l = CHR_SVLEN(data->last_found);
1402 SV * const longest_sv = data->substrs[data->cur_is_floating].str;
1403 const STRLEN old_l = CHR_SVLEN(longest_sv);
1404 GET_RE_DEBUG_FLAGS_DECL;
1406 PERL_ARGS_ASSERT_SCAN_COMMIT;
1408 if ((l >= old_l) && ((l > old_l) || (data->flags & SF_BEFORE_EOL))) {
1409 const U8 i = data->cur_is_floating;
1410 SvSetMagicSV(longest_sv, data->last_found);
1411 data->substrs[i].min_offset = l ? data->last_start_min : data->pos_min;
1414 data->substrs[0].max_offset = data->substrs[0].min_offset;
1416 data->substrs[1].max_offset = (l
1417 ? data->last_start_max
1418 : (data->pos_delta > SSize_t_MAX - data->pos_min
1420 : data->pos_min + data->pos_delta));
1422 || (STRLEN)data->substrs[1].max_offset > (STRLEN)SSize_t_MAX)
1423 data->substrs[1].max_offset = SSize_t_MAX;
1426 if (data->flags & SF_BEFORE_EOL)
1427 data->substrs[i].flags |= (data->flags & SF_BEFORE_EOL);
1429 data->substrs[i].flags &= ~SF_BEFORE_EOL;
1430 data->substrs[i].minlenp = minlenp;
1431 data->substrs[i].lookbehind = 0;
1434 SvCUR_set(data->last_found, 0);
1436 SV * const sv = data->last_found;
1437 if (SvUTF8(sv) && SvMAGICAL(sv)) {
1438 MAGIC * const mg = mg_find(sv, PERL_MAGIC_utf8);
1443 data->last_end = -1;
1444 data->flags &= ~SF_BEFORE_EOL;
1445 DEBUG_STUDYDATA("commit", data, 0, is_inf);
1448 /* An SSC is just a regnode_charclass_posix with an extra field: the inversion
1449 * list that describes which code points it matches */
1452 S_ssc_anything(pTHX_ regnode_ssc *ssc)
1454 /* Set the SSC 'ssc' to match an empty string or any code point */
1456 PERL_ARGS_ASSERT_SSC_ANYTHING;
1458 assert(is_ANYOF_SYNTHETIC(ssc));
1460 /* mortalize so won't leak */
1461 ssc->invlist = sv_2mortal(_add_range_to_invlist(NULL, 0, UV_MAX));
1462 ANYOF_FLAGS(ssc) |= SSC_MATCHES_EMPTY_STRING; /* Plus matches empty */
1466 S_ssc_is_anything(const regnode_ssc *ssc)
1468 /* Returns TRUE if the SSC 'ssc' can match the empty string and any code
1469 * point; FALSE otherwise. Thus, this is used to see if using 'ssc' buys
1470 * us anything: if the function returns TRUE, 'ssc' hasn't been restricted
1471 * in any way, so there's no point in using it */
1476 PERL_ARGS_ASSERT_SSC_IS_ANYTHING;
1478 assert(is_ANYOF_SYNTHETIC(ssc));
1480 if (! (ANYOF_FLAGS(ssc) & SSC_MATCHES_EMPTY_STRING)) {
1484 /* See if the list consists solely of the range 0 - Infinity */
1485 invlist_iterinit(ssc->invlist);
1486 ret = invlist_iternext(ssc->invlist, &start, &end)
1490 invlist_iterfinish(ssc->invlist);
1496 /* If e.g., both \w and \W are set, matches everything */
1497 if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)) {
1499 for (i = 0; i < ANYOF_POSIXL_MAX; i += 2) {
1500 if (ANYOF_POSIXL_TEST(ssc, i) && ANYOF_POSIXL_TEST(ssc, i+1)) {
1510 S_ssc_init(pTHX_ const RExC_state_t *pRExC_state, regnode_ssc *ssc)
1512 /* Initializes the SSC 'ssc'. This includes setting it to match an empty
1513 * string, any code point, or any posix class under locale */
1515 PERL_ARGS_ASSERT_SSC_INIT;
1517 Zero(ssc, 1, regnode_ssc);
1518 set_ANYOF_SYNTHETIC(ssc);
1519 ARG_SET(ssc, ANYOF_ONLY_HAS_BITMAP);
1522 /* If any portion of the regex is to operate under locale rules that aren't
1523 * fully known at compile time, initialization includes it. The reason
1524 * this isn't done for all regexes is that the optimizer was written under
1525 * the assumption that locale was all-or-nothing. Given the complexity and
1526 * lack of documentation in the optimizer, and that there are inadequate
1527 * test cases for locale, many parts of it may not work properly, it is
1528 * safest to avoid locale unless necessary. */
1529 if (RExC_contains_locale) {
1530 ANYOF_POSIXL_SETALL(ssc);
1533 ANYOF_POSIXL_ZERO(ssc);
1538 S_ssc_is_cp_posixl_init(const RExC_state_t *pRExC_state,
1539 const regnode_ssc *ssc)
1541 /* Returns TRUE if the SSC 'ssc' is in its initial state with regard only
1542 * to the list of code points matched, and locale posix classes; hence does
1543 * not check its flags) */
1548 PERL_ARGS_ASSERT_SSC_IS_CP_POSIXL_INIT;
1550 assert(is_ANYOF_SYNTHETIC(ssc));
1552 invlist_iterinit(ssc->invlist);
1553 ret = invlist_iternext(ssc->invlist, &start, &end)
1557 invlist_iterfinish(ssc->invlist);
1563 if (RExC_contains_locale && ! ANYOF_POSIXL_SSC_TEST_ALL_SET(ssc)) {
1570 #define INVLIST_INDEX 0
1571 #define ONLY_LOCALE_MATCHES_INDEX 1
1572 #define DEFERRED_USER_DEFINED_INDEX 2
1575 S_get_ANYOF_cp_list_for_ssc(pTHX_ const RExC_state_t *pRExC_state,
1576 const regnode_charclass* const node)
1578 /* Returns a mortal inversion list defining which code points are matched
1579 * by 'node', which is of type ANYOF. Handles complementing the result if
1580 * appropriate. If some code points aren't knowable at this time, the
1581 * returned list must, and will, contain every code point that is a
1586 SV* only_utf8_locale_invlist = NULL;
1588 const U32 n = ARG(node);
1589 bool new_node_has_latin1 = FALSE;
1590 const U8 flags = (inRANGE(OP(node), ANYOFH, ANYOFHr))
1592 : ANYOF_FLAGS(node);
1594 PERL_ARGS_ASSERT_GET_ANYOF_CP_LIST_FOR_SSC;
1596 /* Look at the data structure created by S_set_ANYOF_arg() */
1597 if (n != ANYOF_ONLY_HAS_BITMAP) {
1598 SV * const rv = MUTABLE_SV(RExC_rxi->data->data[n]);
1599 AV * const av = MUTABLE_AV(SvRV(rv));
1600 SV **const ary = AvARRAY(av);
1601 assert(RExC_rxi->data->what[n] == 's');
1603 if (av_tindex_skip_len_mg(av) >= DEFERRED_USER_DEFINED_INDEX) {
1605 /* Here there are things that won't be known until runtime -- we
1606 * have to assume it could be anything */
1607 invlist = sv_2mortal(_new_invlist(1));
1608 return _add_range_to_invlist(invlist, 0, UV_MAX);
1610 else if (ary[INVLIST_INDEX]) {
1612 /* Use the node's inversion list */
1613 invlist = sv_2mortal(invlist_clone(ary[INVLIST_INDEX], NULL));
1616 /* Get the code points valid only under UTF-8 locales */
1617 if ( (flags & ANYOFL_FOLD)
1618 && av_tindex_skip_len_mg(av) >= ONLY_LOCALE_MATCHES_INDEX)
1620 only_utf8_locale_invlist = ary[ONLY_LOCALE_MATCHES_INDEX];
1625 invlist = sv_2mortal(_new_invlist(0));
1628 /* An ANYOF node contains a bitmap for the first NUM_ANYOF_CODE_POINTS
1629 * code points, and an inversion list for the others, but if there are code
1630 * points that should match only conditionally on the target string being
1631 * UTF-8, those are placed in the inversion list, and not the bitmap.
1632 * Since there are circumstances under which they could match, they are
1633 * included in the SSC. But if the ANYOF node is to be inverted, we have
1634 * to exclude them here, so that when we invert below, the end result
1635 * actually does include them. (Think about "\xe0" =~ /[^\xc0]/di;). We
1636 * have to do this here before we add the unconditionally matched code
1638 if (flags & ANYOF_INVERT) {
1639 _invlist_intersection_complement_2nd(invlist,
1644 /* Add in the points from the bit map */
1645 if (! inRANGE(OP(node), ANYOFH, ANYOFHr)) {
1646 for (i = 0; i < NUM_ANYOF_CODE_POINTS; i++) {
1647 if (ANYOF_BITMAP_TEST(node, i)) {
1648 unsigned int start = i++;
1650 for (; i < NUM_ANYOF_CODE_POINTS
1651 && ANYOF_BITMAP_TEST(node, i); ++i)
1655 invlist = _add_range_to_invlist(invlist, start, i-1);
1656 new_node_has_latin1 = TRUE;
1661 /* If this can match all upper Latin1 code points, have to add them
1662 * as well. But don't add them if inverting, as when that gets done below,
1663 * it would exclude all these characters, including the ones it shouldn't
1664 * that were added just above */
1665 if (! (flags & ANYOF_INVERT) && OP(node) == ANYOFD
1666 && (flags & ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER))
1668 _invlist_union(invlist, PL_UpperLatin1, &invlist);
1671 /* Similarly for these */
1672 if (flags & ANYOF_MATCHES_ALL_ABOVE_BITMAP) {
1673 _invlist_union_complement_2nd(invlist, PL_InBitmap, &invlist);
1676 if (flags & ANYOF_INVERT) {
1677 _invlist_invert(invlist);
1679 else if (flags & ANYOFL_FOLD) {
1680 if (new_node_has_latin1) {
1682 /* Under /li, any 0-255 could fold to any other 0-255, depending on
1683 * the locale. We can skip this if there are no 0-255 at all. */
1684 _invlist_union(invlist, PL_Latin1, &invlist);
1686 invlist = add_cp_to_invlist(invlist, LATIN_SMALL_LETTER_DOTLESS_I);
1687 invlist = add_cp_to_invlist(invlist, LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE);
1690 if (_invlist_contains_cp(invlist, LATIN_SMALL_LETTER_DOTLESS_I)) {
1691 invlist = add_cp_to_invlist(invlist, 'I');
1693 if (_invlist_contains_cp(invlist,
1694 LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE))
1696 invlist = add_cp_to_invlist(invlist, 'i');
1701 /* Similarly add the UTF-8 locale possible matches. These have to be
1702 * deferred until after the non-UTF-8 locale ones are taken care of just
1703 * above, or it leads to wrong results under ANYOF_INVERT */
1704 if (only_utf8_locale_invlist) {
1705 _invlist_union_maybe_complement_2nd(invlist,
1706 only_utf8_locale_invlist,
1707 flags & ANYOF_INVERT,
1714 /* These two functions currently do the exact same thing */
1715 #define ssc_init_zero ssc_init
1717 #define ssc_add_cp(ssc, cp) ssc_add_range((ssc), (cp), (cp))
1718 #define ssc_match_all_cp(ssc) ssc_add_range(ssc, 0, UV_MAX)
1720 /* 'AND' a given class with another one. Can create false positives. 'ssc'
1721 * should not be inverted. 'and_with->flags & ANYOF_MATCHES_POSIXL' should be
1722 * 0 if 'and_with' is a regnode_charclass instead of a regnode_ssc. */
1725 S_ssc_and(pTHX_ const RExC_state_t *pRExC_state, regnode_ssc *ssc,
1726 const regnode_charclass *and_with)
1728 /* Accumulate into SSC 'ssc' its 'AND' with 'and_with', which is either
1729 * another SSC or a regular ANYOF class. Can create false positives. */
1732 U8 and_with_flags = inRANGE(OP(and_with), ANYOFH, ANYOFHr)
1734 : ANYOF_FLAGS(and_with);
1737 PERL_ARGS_ASSERT_SSC_AND;
1739 assert(is_ANYOF_SYNTHETIC(ssc));
1741 /* 'and_with' is used as-is if it too is an SSC; otherwise have to extract
1742 * the code point inversion list and just the relevant flags */
1743 if (is_ANYOF_SYNTHETIC(and_with)) {
1744 anded_cp_list = ((regnode_ssc *)and_with)->invlist;
1745 anded_flags = and_with_flags;
1747 /* XXX This is a kludge around what appears to be deficiencies in the
1748 * optimizer. If we make S_ssc_anything() add in the WARN_SUPER flag,
1749 * there are paths through the optimizer where it doesn't get weeded
1750 * out when it should. And if we don't make some extra provision for
1751 * it like the code just below, it doesn't get added when it should.
1752 * This solution is to add it only when AND'ing, which is here, and
1753 * only when what is being AND'ed is the pristine, original node
1754 * matching anything. Thus it is like adding it to ssc_anything() but
1755 * only when the result is to be AND'ed. Probably the same solution
1756 * could be adopted for the same problem we have with /l matching,
1757 * which is solved differently in S_ssc_init(), and that would lead to
1758 * fewer false positives than that solution has. But if this solution
1759 * creates bugs, the consequences are only that a warning isn't raised
1760 * that should be; while the consequences for having /l bugs is
1761 * incorrect matches */
1762 if (ssc_is_anything((regnode_ssc *)and_with)) {
1763 anded_flags |= ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER;
1767 anded_cp_list = get_ANYOF_cp_list_for_ssc(pRExC_state, and_with);
1768 if (OP(and_with) == ANYOFD) {
1769 anded_flags = and_with_flags & ANYOF_COMMON_FLAGS;
1772 anded_flags = and_with_flags
1773 &( ANYOF_COMMON_FLAGS
1774 |ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER
1775 |ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP);
1776 if (ANYOFL_UTF8_LOCALE_REQD(and_with_flags)) {
1778 ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD;
1783 ANYOF_FLAGS(ssc) &= anded_flags;
1785 /* Below, C1 is the list of code points in 'ssc'; P1, its posix classes.
1786 * C2 is the list of code points in 'and-with'; P2, its posix classes.
1787 * 'and_with' may be inverted. When not inverted, we have the situation of
1789 * (C1 | P1) & (C2 | P2)
1790 * = (C1 & (C2 | P2)) | (P1 & (C2 | P2))
1791 * = ((C1 & C2) | (C1 & P2)) | ((P1 & C2) | (P1 & P2))
1792 * <= ((C1 & C2) | P2)) | ( P1 | (P1 & P2))
1793 * <= ((C1 & C2) | P1 | P2)
1794 * Alternatively, the last few steps could be:
1795 * = ((C1 & C2) | (C1 & P2)) | ((P1 & C2) | (P1 & P2))
1796 * <= ((C1 & C2) | C1 ) | ( C2 | (P1 & P2))
1797 * <= (C1 | C2 | (P1 & P2))
1798 * We favor the second approach if either P1 or P2 is non-empty. This is
1799 * because these components are a barrier to doing optimizations, as what
1800 * they match cannot be known until the moment of matching as they are
1801 * dependent on the current locale, 'AND"ing them likely will reduce or
1803 * But we can do better if we know that C1,P1 are in their initial state (a
1804 * frequent occurrence), each matching everything:
1805 * (<everything>) & (C2 | P2) = C2 | P2
1806 * Similarly, if C2,P2 are in their initial state (again a frequent
1807 * occurrence), the result is a no-op
1808 * (C1 | P1) & (<everything>) = C1 | P1
1811 * (C1 | P1) & ~(C2 | P2) = (C1 | P1) & (~C2 & ~P2)
1812 * = (C1 & (~C2 & ~P2)) | (P1 & (~C2 & ~P2))
1813 * <= (C1 & ~C2) | (P1 & ~P2)
1816 if ((and_with_flags & ANYOF_INVERT)
1817 && ! is_ANYOF_SYNTHETIC(and_with))
1821 ssc_intersection(ssc,
1823 FALSE /* Has already been inverted */
1826 /* If either P1 or P2 is empty, the intersection will be also; can skip
1828 if (! (and_with_flags & ANYOF_MATCHES_POSIXL)) {
1829 ANYOF_POSIXL_ZERO(ssc);
1831 else if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)) {
1833 /* Note that the Posix class component P from 'and_with' actually
1835 * P = Pa | Pb | ... | Pn
1836 * where each component is one posix class, such as in [\w\s].
1838 * ~P = ~(Pa | Pb | ... | Pn)
1839 * = ~Pa & ~Pb & ... & ~Pn
1840 * <= ~Pa | ~Pb | ... | ~Pn
1841 * The last is something we can easily calculate, but unfortunately
1842 * is likely to have many false positives. We could do better
1843 * in some (but certainly not all) instances if two classes in
1844 * P have known relationships. For example
1845 * :lower: <= :alpha: <= :alnum: <= \w <= :graph: <= :print:
1847 * :lower: & :print: = :lower:
1848 * And similarly for classes that must be disjoint. For example,
1849 * since \s and \w can have no elements in common based on rules in
1850 * the POSIX standard,
1851 * \w & ^\S = nothing
1852 * Unfortunately, some vendor locales do not meet the Posix
1853 * standard, in particular almost everything by Microsoft.
1854 * The loop below just changes e.g., \w into \W and vice versa */
1856 regnode_charclass_posixl temp;
1857 int add = 1; /* To calculate the index of the complement */
1859 Zero(&temp, 1, regnode_charclass_posixl);
1860 ANYOF_POSIXL_ZERO(&temp);
1861 for (i = 0; i < ANYOF_MAX; i++) {
1863 || ! ANYOF_POSIXL_TEST((regnode_charclass_posixl*) and_with, i)
1864 || ! ANYOF_POSIXL_TEST((regnode_charclass_posixl*) and_with, i + 1));
1866 if (ANYOF_POSIXL_TEST((regnode_charclass_posixl*) and_with, i)) {
1867 ANYOF_POSIXL_SET(&temp, i + add);
1869 add = 0 - add; /* 1 goes to -1; -1 goes to 1 */
1871 ANYOF_POSIXL_AND(&temp, ssc);
1873 } /* else ssc already has no posixes */
1874 } /* else: Not inverted. This routine is a no-op if 'and_with' is an SSC
1875 in its initial state */
1876 else if (! is_ANYOF_SYNTHETIC(and_with)
1877 || ! ssc_is_cp_posixl_init(pRExC_state, (regnode_ssc *)and_with))
1879 /* But if 'ssc' is in its initial state, the result is just 'and_with';
1880 * copy it over 'ssc' */
1881 if (ssc_is_cp_posixl_init(pRExC_state, ssc)) {
1882 if (is_ANYOF_SYNTHETIC(and_with)) {
1883 StructCopy(and_with, ssc, regnode_ssc);
1886 ssc->invlist = anded_cp_list;
1887 ANYOF_POSIXL_ZERO(ssc);
1888 if (and_with_flags & ANYOF_MATCHES_POSIXL) {
1889 ANYOF_POSIXL_OR((regnode_charclass_posixl*) and_with, ssc);
1893 else if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)
1894 || (and_with_flags & ANYOF_MATCHES_POSIXL))
1896 /* One or the other of P1, P2 is non-empty. */
1897 if (and_with_flags & ANYOF_MATCHES_POSIXL) {
1898 ANYOF_POSIXL_AND((regnode_charclass_posixl*) and_with, ssc);
1900 ssc_union(ssc, anded_cp_list, FALSE);
1902 else { /* P1 = P2 = empty */
1903 ssc_intersection(ssc, anded_cp_list, FALSE);
1909 S_ssc_or(pTHX_ const RExC_state_t *pRExC_state, regnode_ssc *ssc,
1910 const regnode_charclass *or_with)
1912 /* Accumulate into SSC 'ssc' its 'OR' with 'or_with', which is either
1913 * another SSC or a regular ANYOF class. Can create false positives if
1914 * 'or_with' is to be inverted. */
1918 U8 or_with_flags = inRANGE(OP(or_with), ANYOFH, ANYOFHr)
1920 : ANYOF_FLAGS(or_with);
1922 PERL_ARGS_ASSERT_SSC_OR;
1924 assert(is_ANYOF_SYNTHETIC(ssc));
1926 /* 'or_with' is used as-is if it too is an SSC; otherwise have to extract
1927 * the code point inversion list and just the relevant flags */
1928 if (is_ANYOF_SYNTHETIC(or_with)) {
1929 ored_cp_list = ((regnode_ssc*) or_with)->invlist;
1930 ored_flags = or_with_flags;
1933 ored_cp_list = get_ANYOF_cp_list_for_ssc(pRExC_state, or_with);
1934 ored_flags = or_with_flags & ANYOF_COMMON_FLAGS;
1935 if (OP(or_with) != ANYOFD) {
1938 & ( ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER
1939 |ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP);
1940 if (ANYOFL_UTF8_LOCALE_REQD(or_with_flags)) {
1942 ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD;
1947 ANYOF_FLAGS(ssc) |= ored_flags;
1949 /* Below, C1 is the list of code points in 'ssc'; P1, its posix classes.
1950 * C2 is the list of code points in 'or-with'; P2, its posix classes.
1951 * 'or_with' may be inverted. When not inverted, we have the simple
1952 * situation of computing:
1953 * (C1 | P1) | (C2 | P2) = (C1 | C2) | (P1 | P2)
1954 * If P1|P2 yields a situation with both a class and its complement are
1955 * set, like having both \w and \W, this matches all code points, and we
1956 * can delete these from the P component of the ssc going forward. XXX We
1957 * might be able to delete all the P components, but I (khw) am not certain
1958 * about this, and it is better to be safe.
1961 * (C1 | P1) | ~(C2 | P2) = (C1 | P1) | (~C2 & ~P2)
1962 * <= (C1 | P1) | ~C2
1963 * <= (C1 | ~C2) | P1
1964 * (which results in actually simpler code than the non-inverted case)
1967 if ((or_with_flags & ANYOF_INVERT)
1968 && ! is_ANYOF_SYNTHETIC(or_with))
1970 /* We ignore P2, leaving P1 going forward */
1971 } /* else Not inverted */
1972 else if (or_with_flags & ANYOF_MATCHES_POSIXL) {
1973 ANYOF_POSIXL_OR((regnode_charclass_posixl*)or_with, ssc);
1974 if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)) {
1976 for (i = 0; i < ANYOF_MAX; i += 2) {
1977 if (ANYOF_POSIXL_TEST(ssc, i) && ANYOF_POSIXL_TEST(ssc, i + 1))
1979 ssc_match_all_cp(ssc);
1980 ANYOF_POSIXL_CLEAR(ssc, i);
1981 ANYOF_POSIXL_CLEAR(ssc, i+1);
1989 FALSE /* Already has been inverted */
1993 PERL_STATIC_INLINE void
1994 S_ssc_union(pTHX_ regnode_ssc *ssc, SV* const invlist, const bool invert2nd)
1996 PERL_ARGS_ASSERT_SSC_UNION;
1998 assert(is_ANYOF_SYNTHETIC(ssc));
2000 _invlist_union_maybe_complement_2nd(ssc->invlist,
2006 PERL_STATIC_INLINE void
2007 S_ssc_intersection(pTHX_ regnode_ssc *ssc,
2009 const bool invert2nd)
2011 PERL_ARGS_ASSERT_SSC_INTERSECTION;
2013 assert(is_ANYOF_SYNTHETIC(ssc));
2015 _invlist_intersection_maybe_complement_2nd(ssc->invlist,
2021 PERL_STATIC_INLINE void
2022 S_ssc_add_range(pTHX_ regnode_ssc *ssc, const UV start, const UV end)
2024 PERL_ARGS_ASSERT_SSC_ADD_RANGE;
2026 assert(is_ANYOF_SYNTHETIC(ssc));
2028 ssc->invlist = _add_range_to_invlist(ssc->invlist, start, end);
2031 PERL_STATIC_INLINE void
2032 S_ssc_cp_and(pTHX_ regnode_ssc *ssc, const UV cp)
2034 /* AND just the single code point 'cp' into the SSC 'ssc' */
2036 SV* cp_list = _new_invlist(2);
2038 PERL_ARGS_ASSERT_SSC_CP_AND;
2040 assert(is_ANYOF_SYNTHETIC(ssc));
2042 cp_list = add_cp_to_invlist(cp_list, cp);
2043 ssc_intersection(ssc, cp_list,
2044 FALSE /* Not inverted */
2046 SvREFCNT_dec_NN(cp_list);
2049 PERL_STATIC_INLINE void
2050 S_ssc_clear_locale(regnode_ssc *ssc)
2052 /* Set the SSC 'ssc' to not match any locale things */
2053 PERL_ARGS_ASSERT_SSC_CLEAR_LOCALE;
2055 assert(is_ANYOF_SYNTHETIC(ssc));
2057 ANYOF_POSIXL_ZERO(ssc);
2058 ANYOF_FLAGS(ssc) &= ~ANYOF_LOCALE_FLAGS;
2061 #define NON_OTHER_COUNT NON_OTHER_COUNT_FOR_USE_ONLY_BY_REGCOMP_DOT_C
2064 S_is_ssc_worth_it(const RExC_state_t * pRExC_state, const regnode_ssc * ssc)
2066 /* The synthetic start class is used to hopefully quickly winnow down
2067 * places where a pattern could start a match in the target string. If it
2068 * doesn't really narrow things down that much, there isn't much point to
2069 * having the overhead of using it. This function uses some very crude
2070 * heuristics to decide if to use the ssc or not.
2072 * It returns TRUE if 'ssc' rules out more than half what it considers to
2073 * be the "likely" possible matches, but of course it doesn't know what the
2074 * actual things being matched are going to be; these are only guesses
2076 * For /l matches, it assumes that the only likely matches are going to be
2077 * in the 0-255 range, uniformly distributed, so half of that is 127
2078 * For /a and /d matches, it assumes that the likely matches will be just
2079 * the ASCII range, so half of that is 63
2080 * For /u and there isn't anything matching above the Latin1 range, it
2081 * assumes that that is the only range likely to be matched, and uses
2082 * half that as the cut-off: 127. If anything matches above Latin1,
2083 * it assumes that all of Unicode could match (uniformly), except for
2084 * non-Unicode code points and things in the General Category "Other"
2085 * (unassigned, private use, surrogates, controls and formats). This
2086 * is a much large number. */
2088 U32 count = 0; /* Running total of number of code points matched by
2090 UV start, end; /* Start and end points of current range in inversion
2091 XXX outdated. UTF-8 locales are common, what about invert? list */
2092 const U32 max_code_points = (LOC)
2094 : (( ! UNI_SEMANTICS
2095 || invlist_highest(ssc->invlist) < 256)
2098 const U32 max_match = max_code_points / 2;
2100 PERL_ARGS_ASSERT_IS_SSC_WORTH_IT;
2102 invlist_iterinit(ssc->invlist);
2103 while (invlist_iternext(ssc->invlist, &start, &end)) {
2104 if (start >= max_code_points) {
2107 end = MIN(end, max_code_points - 1);
2108 count += end - start + 1;
2109 if (count >= max_match) {
2110 invlist_iterfinish(ssc->invlist);
2120 S_ssc_finalize(pTHX_ RExC_state_t *pRExC_state, regnode_ssc *ssc)
2122 /* The inversion list in the SSC is marked mortal; now we need a more
2123 * permanent copy, which is stored the same way that is done in a regular
2124 * ANYOF node, with the first NUM_ANYOF_CODE_POINTS code points in a bit
2127 SV* invlist = invlist_clone(ssc->invlist, NULL);
2129 PERL_ARGS_ASSERT_SSC_FINALIZE;
2131 assert(is_ANYOF_SYNTHETIC(ssc));
2133 /* The code in this file assumes that all but these flags aren't relevant
2134 * to the SSC, except SSC_MATCHES_EMPTY_STRING, which should be cleared
2135 * by the time we reach here */
2136 assert(! (ANYOF_FLAGS(ssc)
2137 & ~( ANYOF_COMMON_FLAGS
2138 |ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER
2139 |ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP)));
2141 populate_ANYOF_from_invlist( (regnode *) ssc, &invlist);
2143 set_ANYOF_arg(pRExC_state, (regnode *) ssc, invlist, NULL, NULL);
2145 /* Make sure is clone-safe */
2146 ssc->invlist = NULL;
2148 if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)) {
2149 ANYOF_FLAGS(ssc) |= ANYOF_MATCHES_POSIXL;
2150 OP(ssc) = ANYOFPOSIXL;
2152 else if (RExC_contains_locale) {
2156 assert(! (ANYOF_FLAGS(ssc) & ANYOF_LOCALE_FLAGS) || RExC_contains_locale);
2159 #define TRIE_LIST_ITEM(state,idx) (trie->states[state].trans.list)[ idx ]
2160 #define TRIE_LIST_CUR(state) ( TRIE_LIST_ITEM( state, 0 ).forid )
2161 #define TRIE_LIST_LEN(state) ( TRIE_LIST_ITEM( state, 0 ).newstate )
2162 #define TRIE_LIST_USED(idx) ( trie->states[state].trans.list \
2163 ? (TRIE_LIST_CUR( idx ) - 1) \
2169 dump_trie(trie,widecharmap,revcharmap)
2170 dump_trie_interim_list(trie,widecharmap,revcharmap,next_alloc)
2171 dump_trie_interim_table(trie,widecharmap,revcharmap,next_alloc)
2173 These routines dump out a trie in a somewhat readable format.
2174 The _interim_ variants are used for debugging the interim
2175 tables that are used to generate the final compressed
2176 representation which is what dump_trie expects.
2178 Part of the reason for their existence is to provide a form
2179 of documentation as to how the different representations function.
2184 Dumps the final compressed table form of the trie to Perl_debug_log.
2185 Used for debugging make_trie().
2189 S_dump_trie(pTHX_ const struct _reg_trie_data *trie, HV *widecharmap,
2190 AV *revcharmap, U32 depth)
2193 SV *sv=sv_newmortal();
2194 int colwidth= widecharmap ? 6 : 4;
2196 GET_RE_DEBUG_FLAGS_DECL;
2198 PERL_ARGS_ASSERT_DUMP_TRIE;
2200 Perl_re_indentf( aTHX_ "Char : %-6s%-6s%-4s ",
2201 depth+1, "Match","Base","Ofs" );
2203 for( state = 0 ; state < trie->uniquecharcount ; state++ ) {
2204 SV ** const tmp = av_fetch( revcharmap, state, 0);
2206 Perl_re_printf( aTHX_ "%*s",
2208 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
2209 PL_colors[0], PL_colors[1],
2210 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
2211 PERL_PV_ESCAPE_FIRSTCHAR
2216 Perl_re_printf( aTHX_ "\n");
2217 Perl_re_indentf( aTHX_ "State|-----------------------", depth+1);
2219 for( state = 0 ; state < trie->uniquecharcount ; state++ )
2220 Perl_re_printf( aTHX_ "%.*s", colwidth, "--------");
2221 Perl_re_printf( aTHX_ "\n");
2223 for( state = 1 ; state < trie->statecount ; state++ ) {
2224 const U32 base = trie->states[ state ].trans.base;
2226 Perl_re_indentf( aTHX_ "#%4" UVXf "|", depth+1, (UV)state);
2228 if ( trie->states[ state ].wordnum ) {
2229 Perl_re_printf( aTHX_ " W%4X", trie->states[ state ].wordnum );
2231 Perl_re_printf( aTHX_ "%6s", "" );
2234 Perl_re_printf( aTHX_ " @%4" UVXf " ", (UV)base );
2239 while( ( base + ofs < trie->uniquecharcount ) ||
2240 ( base + ofs - trie->uniquecharcount < trie->lasttrans
2241 && trie->trans[ base + ofs - trie->uniquecharcount ].check
2245 Perl_re_printf( aTHX_ "+%2" UVXf "[ ", (UV)ofs);
2247 for ( ofs = 0 ; ofs < trie->uniquecharcount ; ofs++ ) {
2248 if ( ( base + ofs >= trie->uniquecharcount )
2249 && ( base + ofs - trie->uniquecharcount
2251 && trie->trans[ base + ofs
2252 - trie->uniquecharcount ].check == state )
2254 Perl_re_printf( aTHX_ "%*" UVXf, colwidth,
2255 (UV)trie->trans[ base + ofs - trie->uniquecharcount ].next
2258 Perl_re_printf( aTHX_ "%*s", colwidth," ." );
2262 Perl_re_printf( aTHX_ "]");
2265 Perl_re_printf( aTHX_ "\n" );
2267 Perl_re_indentf( aTHX_ "word_info N:(prev,len)=",
2269 for (word=1; word <= trie->wordcount; word++) {
2270 Perl_re_printf( aTHX_ " %d:(%d,%d)",
2271 (int)word, (int)(trie->wordinfo[word].prev),
2272 (int)(trie->wordinfo[word].len));
2274 Perl_re_printf( aTHX_ "\n" );
2277 Dumps a fully constructed but uncompressed trie in list form.
2278 List tries normally only are used for construction when the number of
2279 possible chars (trie->uniquecharcount) is very high.
2280 Used for debugging make_trie().
2283 S_dump_trie_interim_list(pTHX_ const struct _reg_trie_data *trie,
2284 HV *widecharmap, AV *revcharmap, U32 next_alloc,
2288 SV *sv=sv_newmortal();
2289 int colwidth= widecharmap ? 6 : 4;
2290 GET_RE_DEBUG_FLAGS_DECL;
2292 PERL_ARGS_ASSERT_DUMP_TRIE_INTERIM_LIST;
2294 /* print out the table precompression. */
2295 Perl_re_indentf( aTHX_ "State :Word | Transition Data\n",
2297 Perl_re_indentf( aTHX_ "%s",
2298 depth+1, "------:-----+-----------------\n" );
2300 for( state=1 ; state < next_alloc ; state ++ ) {
2303 Perl_re_indentf( aTHX_ " %4" UVXf " :",
2304 depth+1, (UV)state );
2305 if ( ! trie->states[ state ].wordnum ) {
2306 Perl_re_printf( aTHX_ "%5s| ","");
2308 Perl_re_printf( aTHX_ "W%4x| ",
2309 trie->states[ state ].wordnum
2312 for( charid = 1 ; charid <= TRIE_LIST_USED( state ) ; charid++ ) {
2313 SV ** const tmp = av_fetch( revcharmap,
2314 TRIE_LIST_ITEM(state, charid).forid, 0);
2316 Perl_re_printf( aTHX_ "%*s:%3X=%4" UVXf " | ",
2318 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp),
2320 PL_colors[0], PL_colors[1],
2321 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)
2322 | PERL_PV_ESCAPE_FIRSTCHAR
2324 TRIE_LIST_ITEM(state, charid).forid,
2325 (UV)TRIE_LIST_ITEM(state, charid).newstate
2328 Perl_re_printf( aTHX_ "\n%*s| ",
2329 (int)((depth * 2) + 14), "");
2332 Perl_re_printf( aTHX_ "\n");
2337 Dumps a fully constructed but uncompressed trie in table form.
2338 This is the normal DFA style state transition table, with a few
2339 twists to facilitate compression later.
2340 Used for debugging make_trie().
2343 S_dump_trie_interim_table(pTHX_ const struct _reg_trie_data *trie,
2344 HV *widecharmap, AV *revcharmap, U32 next_alloc,
2349 SV *sv=sv_newmortal();
2350 int colwidth= widecharmap ? 6 : 4;
2351 GET_RE_DEBUG_FLAGS_DECL;
2353 PERL_ARGS_ASSERT_DUMP_TRIE_INTERIM_TABLE;
2356 print out the table precompression so that we can do a visual check
2357 that they are identical.
2360 Perl_re_indentf( aTHX_ "Char : ", depth+1 );
2362 for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) {
2363 SV ** const tmp = av_fetch( revcharmap, charid, 0);
2365 Perl_re_printf( aTHX_ "%*s",
2367 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
2368 PL_colors[0], PL_colors[1],
2369 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
2370 PERL_PV_ESCAPE_FIRSTCHAR
2376 Perl_re_printf( aTHX_ "\n");
2377 Perl_re_indentf( aTHX_ "State+-", depth+1 );
2379 for( charid=0 ; charid < trie->uniquecharcount ; charid++ ) {
2380 Perl_re_printf( aTHX_ "%.*s", colwidth,"--------");
2383 Perl_re_printf( aTHX_ "\n" );
2385 for( state=1 ; state < next_alloc ; state += trie->uniquecharcount ) {
2387 Perl_re_indentf( aTHX_ "%4" UVXf " : ",
2389 (UV)TRIE_NODENUM( state ) );
2391 for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) {
2392 UV v=(UV)SAFE_TRIE_NODENUM( trie->trans[ state + charid ].next );
2394 Perl_re_printf( aTHX_ "%*" UVXf, colwidth, v );
2396 Perl_re_printf( aTHX_ "%*s", colwidth, "." );
2398 if ( ! trie->states[ TRIE_NODENUM( state ) ].wordnum ) {
2399 Perl_re_printf( aTHX_ " (%4" UVXf ")\n",
2400 (UV)trie->trans[ state ].check );
2402 Perl_re_printf( aTHX_ " (%4" UVXf ") W%4X\n",
2403 (UV)trie->trans[ state ].check,
2404 trie->states[ TRIE_NODENUM( state ) ].wordnum );
2412 /* make_trie(startbranch,first,last,tail,word_count,flags,depth)
2413 startbranch: the first branch in the whole branch sequence
2414 first : start branch of sequence of branch-exact nodes.
2415 May be the same as startbranch
2416 last : Thing following the last branch.
2417 May be the same as tail.
2418 tail : item following the branch sequence
2419 count : words in the sequence
2420 flags : currently the OP() type we will be building one of /EXACT(|F|FA|FU|FU_SS|L|FLU8)/
2421 depth : indent depth
2423 Inplace optimizes a sequence of 2 or more Branch-Exact nodes into a TRIE node.
2425 A trie is an N'ary tree where the branches are determined by digital
2426 decomposition of the key. IE, at the root node you look up the 1st character and
2427 follow that branch repeat until you find the end of the branches. Nodes can be
2428 marked as "accepting" meaning they represent a complete word. Eg:
2432 would convert into the following structure. Numbers represent states, letters
2433 following numbers represent valid transitions on the letter from that state, if
2434 the number is in square brackets it represents an accepting state, otherwise it
2435 will be in parenthesis.
2437 +-h->+-e->[3]-+-r->(8)-+-s->[9]
2441 (1) +-i->(6)-+-s->[7]
2443 +-s->(3)-+-h->(4)-+-e->[5]
2445 Accept Word Mapping: 3=>1 (he),5=>2 (she), 7=>3 (his), 9=>4 (hers)
2447 This shows that when matching against the string 'hers' we will begin at state 1
2448 read 'h' and move to state 2, read 'e' and move to state 3 which is accepting,
2449 then read 'r' and go to state 8 followed by 's' which takes us to state 9 which
2450 is also accepting. Thus we know that we can match both 'he' and 'hers' with a
2451 single traverse. We store a mapping from accepting to state to which word was
2452 matched, and then when we have multiple possibilities we try to complete the
2453 rest of the regex in the order in which they occurred in the alternation.
2455 The only prior NFA like behaviour that would be changed by the TRIE support is
2456 the silent ignoring of duplicate alternations which are of the form:
2458 / (DUPE|DUPE) X? (?{ ... }) Y /x
2460 Thus EVAL blocks following a trie may be called a different number of times with
2461 and without the optimisation. With the optimisations dupes will be silently
2462 ignored. This inconsistent behaviour of EVAL type nodes is well established as
2463 the following demonstrates:
2465 'words'=~/(word|word|word)(?{ print $1 })[xyz]/
2467 which prints out 'word' three times, but
2469 'words'=~/(word|word|word)(?{ print $1 })S/
2471 which doesnt print it out at all. This is due to other optimisations kicking in.
2473 Example of what happens on a structural level:
2475 The regexp /(ac|ad|ab)+/ will produce the following debug output:
2477 1: CURLYM[1] {1,32767}(18)
2488 This would be optimizable with startbranch=5, first=5, last=16, tail=16
2489 and should turn into:
2491 1: CURLYM[1] {1,32767}(18)
2493 [Words:3 Chars Stored:6 Unique Chars:4 States:5 NCP:1]
2501 Cases where tail != last would be like /(?foo|bar)baz/:
2511 which would be optimizable with startbranch=1, first=1, last=7, tail=8
2512 and would end up looking like:
2515 [Words:2 Chars Stored:6 Unique Chars:5 States:7 NCP:1]
2522 d = uvchr_to_utf8_flags(d, uv, 0);
2524 is the recommended Unicode-aware way of saying
2529 #define TRIE_STORE_REVCHAR(val) \
2532 SV *zlopp = newSV(UTF8_MAXBYTES); \
2533 unsigned char *flrbbbbb = (unsigned char *) SvPVX(zlopp); \
2534 unsigned char *const kapow = uvchr_to_utf8(flrbbbbb, val); \
2536 SvCUR_set(zlopp, kapow - flrbbbbb); \
2539 av_push(revcharmap, zlopp); \
2541 char ooooff = (char)val; \
2542 av_push(revcharmap, newSVpvn(&ooooff, 1)); \
2546 /* This gets the next character from the input, folding it if not already
2548 #define TRIE_READ_CHAR STMT_START { \
2551 /* if it is UTF then it is either already folded, or does not need \
2553 uvc = valid_utf8_to_uvchr( (const U8*) uc, &len); \
2555 else if (folder == PL_fold_latin1) { \
2556 /* This folder implies Unicode rules, which in the range expressible \
2557 * by not UTF is the lower case, with the two exceptions, one of \
2558 * which should have been taken care of before calling this */ \
2559 assert(*uc != LATIN_SMALL_LETTER_SHARP_S); \
2560 uvc = toLOWER_L1(*uc); \
2561 if (UNLIKELY(uvc == MICRO_SIGN)) uvc = GREEK_SMALL_LETTER_MU; \
2564 /* raw data, will be folded later if needed */ \
2572 #define TRIE_LIST_PUSH(state,fid,ns) STMT_START { \
2573 if ( TRIE_LIST_CUR( state ) >=TRIE_LIST_LEN( state ) ) { \
2574 U32 ging = TRIE_LIST_LEN( state ) * 2; \
2575 Renew( trie->states[ state ].trans.list, ging, reg_trie_trans_le ); \
2576 TRIE_LIST_LEN( state ) = ging; \
2578 TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).forid = fid; \
2579 TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).newstate = ns; \
2580 TRIE_LIST_CUR( state )++; \
2583 #define TRIE_LIST_NEW(state) STMT_START { \
2584 Newx( trie->states[ state ].trans.list, \
2585 4, reg_trie_trans_le ); \
2586 TRIE_LIST_CUR( state ) = 1; \
2587 TRIE_LIST_LEN( state ) = 4; \
2590 #define TRIE_HANDLE_WORD(state) STMT_START { \
2591 U16 dupe= trie->states[ state ].wordnum; \
2592 regnode * const noper_next = regnext( noper ); \
2595 /* store the word for dumping */ \
2597 if (OP(noper) != NOTHING) \
2598 tmp = newSVpvn_utf8(STRING(noper), STR_LEN(noper), UTF); \
2600 tmp = newSVpvn_utf8( "", 0, UTF ); \
2601 av_push( trie_words, tmp ); \
2605 trie->wordinfo[curword].prev = 0; \
2606 trie->wordinfo[curword].len = wordlen; \
2607 trie->wordinfo[curword].accept = state; \
2609 if ( noper_next < tail ) { \
2611 trie->jump = (U16 *) PerlMemShared_calloc( word_count + 1, \
2613 trie->jump[curword] = (U16)(noper_next - convert); \
2615 jumper = noper_next; \
2617 nextbranch= regnext(cur); \
2621 /* It's a dupe. Pre-insert into the wordinfo[].prev */\
2622 /* chain, so that when the bits of chain are later */\
2623 /* linked together, the dups appear in the chain */\
2624 trie->wordinfo[curword].prev = trie->wordinfo[dupe].prev; \
2625 trie->wordinfo[dupe].prev = curword; \
2627 /* we haven't inserted this word yet. */ \
2628 trie->states[ state ].wordnum = curword; \
2633 #define TRIE_TRANS_STATE(state,base,ucharcount,charid,special) \
2634 ( ( base + charid >= ucharcount \
2635 && base + charid < ubound \
2636 && state == trie->trans[ base - ucharcount + charid ].check \
2637 && trie->trans[ base - ucharcount + charid ].next ) \
2638 ? trie->trans[ base - ucharcount + charid ].next \
2639 : ( state==1 ? special : 0 ) \
2642 #define TRIE_BITMAP_SET_FOLDED(trie, uvc, folder) \
2644 TRIE_BITMAP_SET(trie, uvc); \
2645 /* store the folded codepoint */ \
2647 TRIE_BITMAP_SET(trie, folder[(U8) uvc ]); \
2650 /* store first byte of utf8 representation of */ \
2651 /* variant codepoints */ \
2652 if (! UVCHR_IS_INVARIANT(uvc)) { \
2653 TRIE_BITMAP_SET(trie, UTF8_TWO_BYTE_HI(uvc)); \
2658 #define MADE_JUMP_TRIE 2
2659 #define MADE_EXACT_TRIE 4
2662 S_make_trie(pTHX_ RExC_state_t *pRExC_state, regnode *startbranch,
2663 regnode *first, regnode *last, regnode *tail,
2664 U32 word_count, U32 flags, U32 depth)
2666 /* first pass, loop through and scan words */
2667 reg_trie_data *trie;
2668 HV *widecharmap = NULL;
2669 AV *revcharmap = newAV();
2675 regnode *jumper = NULL;
2676 regnode *nextbranch = NULL;
2677 regnode *convert = NULL;
2678 U32 *prev_states; /* temp array mapping each state to previous one */
2679 /* we just use folder as a flag in utf8 */
2680 const U8 * folder = NULL;
2682 /* in the below add_data call we are storing either 'tu' or 'tuaa'
2683 * which stands for one trie structure, one hash, optionally followed
2686 const U32 data_slot = add_data( pRExC_state, STR_WITH_LEN("tuaa"));
2687 AV *trie_words = NULL;
2688 /* along with revcharmap, this only used during construction but both are
2689 * useful during debugging so we store them in the struct when debugging.
2692 const U32 data_slot = add_data( pRExC_state, STR_WITH_LEN("tu"));
2693 STRLEN trie_charcount=0;
2695 SV *re_trie_maxbuff;
2696 GET_RE_DEBUG_FLAGS_DECL;
2698 PERL_ARGS_ASSERT_MAKE_TRIE;
2700 PERL_UNUSED_ARG(depth);
2704 case EXACT: case EXACT_ONLY8: case EXACTL: break;
2708 case EXACTFLU8: folder = PL_fold_latin1; break;
2709 case EXACTF: folder = PL_fold; break;
2710 default: Perl_croak( aTHX_ "panic! In trie construction, unknown node type %u %s", (unsigned) flags, PL_reg_name[flags] );
2713 trie = (reg_trie_data *) PerlMemShared_calloc( 1, sizeof(reg_trie_data) );
2715 trie->startstate = 1;
2716 trie->wordcount = word_count;
2717 RExC_rxi->data->data[ data_slot ] = (void*)trie;
2718 trie->charmap = (U16 *) PerlMemShared_calloc( 256, sizeof(U16) );
2719 if (flags == EXACT || flags == EXACT_ONLY8 || flags == EXACTL)
2720 trie->bitmap = (char *) PerlMemShared_calloc( ANYOF_BITMAP_SIZE, 1 );
2721 trie->wordinfo = (reg_trie_wordinfo *) PerlMemShared_calloc(
2722 trie->wordcount+1, sizeof(reg_trie_wordinfo));
2725 trie_words = newAV();
2728 re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, GV_ADD);
2729 assert(re_trie_maxbuff);
2730 if (!SvIOK(re_trie_maxbuff)) {
2731 sv_setiv(re_trie_maxbuff, RE_TRIE_MAXBUF_INIT);
2733 DEBUG_TRIE_COMPILE_r({
2734 Perl_re_indentf( aTHX_
2735 "make_trie start==%d, first==%d, last==%d, tail==%d depth=%d\n",
2737 REG_NODE_NUM(startbranch), REG_NODE_NUM(first),
2738 REG_NODE_NUM(last), REG_NODE_NUM(tail), (int)depth);
2741 /* Find the node we are going to overwrite */
2742 if ( first == startbranch && OP( last ) != BRANCH ) {
2743 /* whole branch chain */
2746 /* branch sub-chain */
2747 convert = NEXTOPER( first );
2750 /* -- First loop and Setup --
2752 We first traverse the branches and scan each word to determine if it
2753 contains widechars, and how many unique chars there are, this is
2754 important as we have to build a table with at least as many columns as we
2757 We use an array of integers to represent the character codes 0..255
2758 (trie->charmap) and we use a an HV* to store Unicode characters. We use
2759 the native representation of the character value as the key and IV's for
2762 *TODO* If we keep track of how many times each character is used we can
2763 remap the columns so that the table compression later on is more
2764 efficient in terms of memory by ensuring the most common value is in the
2765 middle and the least common are on the outside. IMO this would be better
2766 than a most to least common mapping as theres a decent chance the most
2767 common letter will share a node with the least common, meaning the node
2768 will not be compressible. With a middle is most common approach the worst
2769 case is when we have the least common nodes twice.
2773 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
2774 regnode *noper = NEXTOPER( cur );
2778 U32 wordlen = 0; /* required init */
2779 STRLEN minchars = 0;
2780 STRLEN maxchars = 0;
2781 bool set_bit = trie->bitmap ? 1 : 0; /*store the first char in the
2784 if (OP(noper) == NOTHING) {
2785 /* skip past a NOTHING at the start of an alternation
2786 * eg, /(?:)a|(?:b)/ should be the same as /a|b/
2788 regnode *noper_next= regnext(noper);
2789 if (noper_next < tail)
2794 && ( OP(noper) == flags
2795 || (flags == EXACT && OP(noper) == EXACT_ONLY8)
2796 || (flags == EXACTFU && ( OP(noper) == EXACTFU_ONLY8
2797 || OP(noper) == EXACTFUP))))
2799 uc= (U8*)STRING(noper);
2800 e= uc + STR_LEN(noper);
2807 if ( set_bit ) { /* bitmap only alloced when !(UTF&&Folding) */
2808 TRIE_BITMAP_SET(trie,*uc); /* store the raw first byte
2809 regardless of encoding */
2810 if (OP( noper ) == EXACTFUP) {
2811 /* false positives are ok, so just set this */
2812 TRIE_BITMAP_SET(trie, LATIN_SMALL_LETTER_SHARP_S);
2816 for ( ; uc < e ; uc += len ) { /* Look at each char in the current
2818 TRIE_CHARCOUNT(trie)++;
2821 /* TRIE_READ_CHAR returns the current character, or its fold if /i
2822 * is in effect. Under /i, this character can match itself, or
2823 * anything that folds to it. If not under /i, it can match just
2824 * itself. Most folds are 1-1, for example k, K, and KELVIN SIGN
2825 * all fold to k, and all are single characters. But some folds
2826 * expand to more than one character, so for example LATIN SMALL
2827 * LIGATURE FFI folds to the three character sequence 'ffi'. If
2828 * the string beginning at 'uc' is 'ffi', it could be matched by
2829 * three characters, or just by the one ligature character. (It
2830 * could also be matched by two characters: LATIN SMALL LIGATURE FF
2831 * followed by 'i', or by 'f' followed by LATIN SMALL LIGATURE FI).
2832 * (Of course 'I' and/or 'F' instead of 'i' and 'f' can also
2833 * match.) The trie needs to know the minimum and maximum number
2834 * of characters that could match so that it can use size alone to
2835 * quickly reject many match attempts. The max is simple: it is
2836 * the number of folded characters in this branch (since a fold is
2837 * never shorter than what folds to it. */
2841 /* And the min is equal to the max if not under /i (indicated by
2842 * 'folder' being NULL), or there are no multi-character folds. If
2843 * there is a multi-character fold, the min is incremented just
2844 * once, for the character that folds to the sequence. Each
2845 * character in the sequence needs to be added to the list below of
2846 * characters in the trie, but we count only the first towards the
2847 * min number of characters needed. This is done through the
2848 * variable 'foldlen', which is returned by the macros that look
2849 * for these sequences as the number of bytes the sequence
2850 * occupies. Each time through the loop, we decrement 'foldlen' by
2851 * how many bytes the current char occupies. Only when it reaches
2852 * 0 do we increment 'minchars' or look for another multi-character
2854 if (folder == NULL) {
2857 else if (foldlen > 0) {
2858 foldlen -= (UTF) ? UTF8SKIP(uc) : 1;
2863 /* See if *uc is the beginning of a multi-character fold. If
2864 * so, we decrement the length remaining to look at, to account
2865 * for the current character this iteration. (We can use 'uc'
2866 * instead of the fold returned by TRIE_READ_CHAR because for
2867 * non-UTF, the latin1_safe macro is smart enough to account
2868 * for all the unfolded characters, and because for UTF, the
2869 * string will already have been folded earlier in the
2870 * compilation process */
2872 if ((foldlen = is_MULTI_CHAR_FOLD_utf8_safe(uc, e))) {
2873 foldlen -= UTF8SKIP(uc);
2876 else if ((foldlen = is_MULTI_CHAR_FOLD_latin1_safe(uc, e))) {
2881 /* The current character (and any potential folds) should be added
2882 * to the possible matching characters for this position in this
2886 U8 folded= folder[ (U8) uvc ];
2887 if ( !trie->charmap[ folded ] ) {
2888 trie->charmap[ folded ]=( ++trie->uniquecharcount );
2889 TRIE_STORE_REVCHAR( folded );
2892 if ( !trie->charmap[ uvc ] ) {
2893 trie->charmap[ uvc ]=( ++trie->uniquecharcount );
2894 TRIE_STORE_REVCHAR( uvc );
2897 /* store the codepoint in the bitmap, and its folded
2899 TRIE_BITMAP_SET_FOLDED(trie, uvc, folder);
2900 set_bit = 0; /* We've done our bit :-) */
2904 /* XXX We could come up with the list of code points that fold
2905 * to this using PL_utf8_foldclosures, except not for
2906 * multi-char folds, as there may be multiple combinations
2907 * there that could work, which needs to wait until runtime to
2908 * resolve (The comment about LIGATURE FFI above is such an
2913 widecharmap = newHV();
2915 svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 1 );
2918 Perl_croak( aTHX_ "error creating/fetching widecharmap entry for 0x%" UVXf, uvc );
2920 if ( !SvTRUE( *svpp ) ) {
2921 sv_setiv( *svpp, ++trie->uniquecharcount );
2922 TRIE_STORE_REVCHAR(uvc);
2925 } /* end loop through characters in this branch of the trie */
2927 /* We take the min and max for this branch and combine to find the min
2928 * and max for all branches processed so far */
2929 if( cur == first ) {
2930 trie->minlen = minchars;
2931 trie->maxlen = maxchars;
2932 } else if (minchars < trie->minlen) {
2933 trie->minlen = minchars;
2934 } else if (maxchars > trie->maxlen) {
2935 trie->maxlen = maxchars;
2937 } /* end first pass */
2938 DEBUG_TRIE_COMPILE_r(
2939 Perl_re_indentf( aTHX_
2940 "TRIE(%s): W:%d C:%d Uq:%d Min:%d Max:%d\n",
2942 ( widecharmap ? "UTF8" : "NATIVE" ), (int)word_count,
2943 (int)TRIE_CHARCOUNT(trie), trie->uniquecharcount,
2944 (int)trie->minlen, (int)trie->maxlen )
2948 We now know what we are dealing with in terms of unique chars and
2949 string sizes so we can calculate how much memory a naive
2950 representation using a flat table will take. If it's over a reasonable
2951 limit (as specified by ${^RE_TRIE_MAXBUF}) we use a more memory
2952 conservative but potentially much slower representation using an array
2955 At the end we convert both representations into the same compressed
2956 form that will be used in regexec.c for matching with. The latter
2957 is a form that cannot be used to construct with but has memory
2958 properties similar to the list form and access properties similar
2959 to the table form making it both suitable for fast searches and
2960 small enough that its feasable to store for the duration of a program.
2962 See the comment in the code where the compressed table is produced
2963 inplace from the flat tabe representation for an explanation of how
2964 the compression works.
2969 Newx(prev_states, TRIE_CHARCOUNT(trie) + 2, U32);
2972 if ( (IV)( ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount + 1)
2973 > SvIV(re_trie_maxbuff) )
2976 Second Pass -- Array Of Lists Representation
2978 Each state will be represented by a list of charid:state records
2979 (reg_trie_trans_le) the first such element holds the CUR and LEN
2980 points of the allocated array. (See defines above).
2982 We build the initial structure using the lists, and then convert
2983 it into the compressed table form which allows faster lookups
2984 (but cant be modified once converted).
2987 STRLEN transcount = 1;
2989 DEBUG_TRIE_COMPILE_MORE_r( Perl_re_indentf( aTHX_ "Compiling trie using list compiler\n",
2992 trie->states = (reg_trie_state *)
2993 PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2,
2994 sizeof(reg_trie_state) );
2998 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
3000 regnode *noper = NEXTOPER( cur );
3001 U32 state = 1; /* required init */
3002 U16 charid = 0; /* sanity init */
3003 U32 wordlen = 0; /* required init */
3005 if (OP(noper) == NOTHING) {
3006 regnode *noper_next= regnext(noper);
3007 if (noper_next < tail)
3012 && ( OP(noper) == flags
3013 || (flags == EXACT && OP(noper) == EXACT_ONLY8)
3014 || (flags == EXACTFU && ( OP(noper) == EXACTFU_ONLY8
3015 || OP(noper) == EXACTFUP))))
3017 const U8 *uc= (U8*)STRING(noper);
3018 const U8 *e= uc + STR_LEN(noper);
3020 for ( ; uc < e ; uc += len ) {
3025 charid = trie->charmap[ uvc ];
3027 SV** const svpp = hv_fetch( widecharmap,
3034 charid=(U16)SvIV( *svpp );
3037 /* charid is now 0 if we dont know the char read, or
3038 * nonzero if we do */
3045 if ( !trie->states[ state ].trans.list ) {
3046 TRIE_LIST_NEW( state );
3049 check <= TRIE_LIST_USED( state );
3052 if ( TRIE_LIST_ITEM( state, check ).forid
3055 newstate = TRIE_LIST_ITEM( state, check ).newstate;
3060 newstate = next_alloc++;
3061 prev_states[newstate] = state;
3062 TRIE_LIST_PUSH( state, charid, newstate );
3067 Perl_croak( aTHX_ "panic! In trie construction, no char mapping for %" IVdf, uvc );
3071 TRIE_HANDLE_WORD(state);
3073 } /* end second pass */
3075 /* next alloc is the NEXT state to be allocated */
3076 trie->statecount = next_alloc;
3077 trie->states = (reg_trie_state *)
3078 PerlMemShared_realloc( trie->states,
3080 * sizeof(reg_trie_state) );
3082 /* and now dump it out before we compress it */
3083 DEBUG_TRIE_COMPILE_MORE_r(dump_trie_interim_list(trie, widecharmap,
3084 revcharmap, next_alloc,
3088 trie->trans = (reg_trie_trans *)
3089 PerlMemShared_calloc( transcount, sizeof(reg_trie_trans) );
3096 for( state=1 ; state < next_alloc ; state ++ ) {
3100 DEBUG_TRIE_COMPILE_MORE_r(
3101 Perl_re_printf( aTHX_ "tp: %d zp: %d ",tp,zp)
3105 if (trie->states[state].trans.list) {
3106 U16 minid=TRIE_LIST_ITEM( state, 1).forid;
3110 for( idx = 2 ; idx <= TRIE_LIST_USED( state ) ; idx++ ) {
3111 const U16 forid = TRIE_LIST_ITEM( state, idx).forid;
3112 if ( forid < minid ) {
3114 } else if ( forid > maxid ) {
3118 if ( transcount < tp + maxid - minid + 1) {
3120 trie->trans = (reg_trie_trans *)
3121 PerlMemShared_realloc( trie->trans,
3123 * sizeof(reg_trie_trans) );
3124 Zero( trie->trans + (transcount / 2),
3128 base = trie->uniquecharcount + tp - minid;
3129 if ( maxid == minid ) {
3131 for ( ; zp < tp ; zp++ ) {
3132 if ( ! trie->trans[ zp ].next ) {
3133 base = trie->uniquecharcount + zp - minid;
3134 trie->trans[ zp ].next = TRIE_LIST_ITEM( state,
3136 trie->trans[ zp ].check = state;
3142 trie->trans[ tp ].next = TRIE_LIST_ITEM( state,
3144 trie->trans[ tp ].check = state;
3149 for ( idx=1; idx <= TRIE_LIST_USED( state ) ; idx++ ) {
3150 const U32 tid = base
3151 - trie->uniquecharcount
3152 + TRIE_LIST_ITEM( state, idx ).forid;
3153 trie->trans[ tid ].next = TRIE_LIST_ITEM( state,
3155 trie->trans[ tid ].check = state;
3157 tp += ( maxid - minid + 1 );
3159 Safefree(trie->states[ state ].trans.list);
3162 DEBUG_TRIE_COMPILE_MORE_r(
3163 Perl_re_printf( aTHX_ " base: %d\n",base);
3166 trie->states[ state ].trans.base=base;
3168 trie->lasttrans = tp + 1;
3172 Second Pass -- Flat Table Representation.
3174 we dont use the 0 slot of either trans[] or states[] so we add 1 to
3175 each. We know that we will need Charcount+1 trans at most to store
3176 the data (one row per char at worst case) So we preallocate both
3177 structures assuming worst case.
3179 We then construct the trie using only the .next slots of the entry
3182 We use the .check field of the first entry of the node temporarily
3183 to make compression both faster and easier by keeping track of how
3184 many non zero fields are in the node.
3186 Since trans are numbered from 1 any 0 pointer in the table is a FAIL
3189 There are two terms at use here: state as a TRIE_NODEIDX() which is
3190 a number representing the first entry of the node, and state as a
3191 TRIE_NODENUM() which is the trans number. state 1 is TRIE_NODEIDX(1)
3192 and TRIE_NODENUM(1), state 2 is TRIE_NODEIDX(2) and TRIE_NODENUM(3)
3193 if there are 2 entrys per node. eg:
3201 The table is internally in the right hand, idx form. However as we
3202 also have to deal with the states array which is indexed by nodenum
3203 we have to use TRIE_NODENUM() to convert.
3206 DEBUG_TRIE_COMPILE_MORE_r( Perl_re_indentf( aTHX_ "Compiling trie using table compiler\n",
3209 trie->trans = (reg_trie_trans *)
3210 PerlMemShared_calloc( ( TRIE_CHARCOUNT(trie) + 1 )
3211 * trie->uniquecharcount + 1,
3212 sizeof(reg_trie_trans) );
3213 trie->states = (reg_trie_state *)
3214 PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2,
3215 sizeof(reg_trie_state) );
3216 next_alloc = trie->uniquecharcount + 1;
3219 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
3221 regnode *noper = NEXTOPER( cur );
3223 U32 state = 1; /* required init */
3225 U16 charid = 0; /* sanity init */
3226 U32 accept_state = 0; /* sanity init */
3228 U32 wordlen = 0; /* required init */
3230 if (OP(noper) == NOTHING) {
3231 regnode *noper_next= regnext(noper);
3232 if (noper_next < tail)
3237 && ( OP(noper) == flags
3238 || (flags == EXACT && OP(noper) == EXACT_ONLY8)
3239 || (flags == EXACTFU && ( OP(noper) == EXACTFU_ONLY8
3240 || OP(noper) == EXACTFUP))))
3242 const U8 *uc= (U8*)STRING(noper);
3243 const U8 *e= uc + STR_LEN(noper);
3245 for ( ; uc < e ; uc += len ) {
3250 charid = trie->charmap[ uvc ];
3252 SV* const * const svpp = hv_fetch( widecharmap,
3256 charid = svpp ? (U16)SvIV(*svpp) : 0;
3260 if ( !trie->trans[ state + charid ].next ) {
3261 trie->trans[ state + charid ].next = next_alloc;
3262 trie->trans[ state ].check++;
3263 prev_states[TRIE_NODENUM(next_alloc)]
3264 = TRIE_NODENUM(state);
3265 next_alloc += trie->uniquecharcount;
3267 state = trie->trans[ state + charid ].next;
3269 Perl_croak( aTHX_ "panic! In trie construction, no char mapping for %" IVdf, uvc );
3271 /* charid is now 0 if we dont know the char read, or
3272 * nonzero if we do */
3275 accept_state = TRIE_NODENUM( state );
3276 TRIE_HANDLE_WORD(accept_state);
3278 } /* end second pass */
3280 /* and now dump it out before we compress it */
3281 DEBUG_TRIE_COMPILE_MORE_r(dump_trie_interim_table(trie, widecharmap,
3283 next_alloc, depth+1));
3287 * Inplace compress the table.*
3289 For sparse data sets the table constructed by the trie algorithm will
3290 be mostly 0/FAIL transitions or to put it another way mostly empty.
3291 (Note that leaf nodes will not contain any transitions.)
3293 This algorithm compresses the tables by eliminating most such
3294 transitions, at the cost of a modest bit of extra work during lookup:
3296 - Each states[] entry contains a .base field which indicates the
3297 index in the state[] array wheres its transition data is stored.
3299 - If .base is 0 there are no valid transitions from that node.
3301 - If .base is nonzero then charid is added to it to find an entry in
3304 -If trans[states[state].base+charid].check!=state then the
3305 transition is taken to be a 0/Fail transition. Thus if there are fail
3306 transitions at the front of the node then the .base offset will point
3307 somewhere inside the previous nodes data (or maybe even into a node
3308 even earlier), but the .check field determines if the transition is
3312 The following process inplace converts the table to the compressed
3313 table: We first do not compress the root node 1,and mark all its
3314 .check pointers as 1 and set its .base pointer as 1 as well. This
3315 allows us to do a DFA construction from the compressed table later,
3316 and ensures that any .base pointers we calculate later are greater
3319 - We set 'pos' to indicate the first entry of the second node.
3321 - We then iterate over the columns of the node, finding the first and
3322 last used entry at l and m. We then copy l..m into pos..(pos+m-l),
3323 and set the .check pointers accordingly, and advance pos
3324 appropriately and repreat for the next node. Note that when we copy
3325 the next pointers we have to convert them from the original
3326 NODEIDX form to NODENUM form as the former is not valid post
3329 - If a node has no transitions used we mark its base as 0 and do not
3330 advance the pos pointer.
3332 - If a node only has one transition we use a second pointer into the
3333 structure to fill in allocated fail transitions from other states.
3334 This pointer is independent of the main pointer and scans forward
3335 looking for null transitions that are allocated to a state. When it
3336 finds one it writes the single transition into the "hole". If the
3337 pointer doesnt find one the single transition is appended as normal.
3339 - Once compressed we can Renew/realloc the structures to release the
3342 See "Table-Compression Methods" in sec 3.9 of the Red Dragon,
3343 specifically Fig 3.47 and the associated pseudocode.
3347 const U32 laststate = TRIE_NODENUM( next_alloc );
3350 trie->statecount = laststate;
3352 for ( state = 1 ; state < laststate ; state++ ) {
3354 const U32 stateidx = TRIE_NODEIDX( state );
3355 const U32 o_used = trie->trans[ stateidx ].check;
3356 U32 used = trie->trans[ stateidx ].check;
3357 trie->trans[ stateidx ].check = 0;
3360 used && charid < trie->uniquecharcount;
3363 if ( flag || trie->trans[ stateidx + charid ].next ) {
3364 if ( trie->trans[ stateidx + charid ].next ) {
3366 for ( ; zp < pos ; zp++ ) {
3367 if ( ! trie->trans[ zp ].next ) {
3371 trie->states[ state ].trans.base
3373 + trie->uniquecharcount
3375 trie->trans[ zp ].next
3376 = SAFE_TRIE_NODENUM( trie->trans[ stateidx
3378 trie->trans[ zp ].check = state;
3379 if ( ++zp > pos ) pos = zp;
3386 trie->states[ state ].trans.base
3387 = pos + trie->uniquecharcount - charid ;
3389 trie->trans[ pos ].next
3390 = SAFE_TRIE_NODENUM(
3391 trie->trans[ stateidx + charid ].next );
3392 trie->trans[ pos ].check = state;
3397 trie->lasttrans = pos + 1;
3398 trie->states = (reg_trie_state *)
3399 PerlMemShared_realloc( trie->states, laststate
3400 * sizeof(reg_trie_state) );
3401 DEBUG_TRIE_COMPILE_MORE_r(
3402 Perl_re_indentf( aTHX_ "Alloc: %d Orig: %" IVdf " elements, Final:%" IVdf ". Savings of %%%5.2f\n",
3404 (int)( ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount
3408 ( ( next_alloc - pos ) * 100 ) / (double)next_alloc );
3411 } /* end table compress */
3413 DEBUG_TRIE_COMPILE_MORE_r(
3414 Perl_re_indentf( aTHX_ "Statecount:%" UVxf " Lasttrans:%" UVxf "\n",
3416 (UV)trie->statecount,
3417 (UV)trie->lasttrans)
3419 /* resize the trans array to remove unused space */
3420 trie->trans = (reg_trie_trans *)
3421 PerlMemShared_realloc( trie->trans, trie->lasttrans
3422 * sizeof(reg_trie_trans) );
3424 { /* Modify the program and insert the new TRIE node */
3425 U8 nodetype =(U8)(flags & 0xFF);
3429 regnode *optimize = NULL;
3430 #ifdef RE_TRACK_PATTERN_OFFSETS
3433 U32 mjd_nodelen = 0;
3434 #endif /* RE_TRACK_PATTERN_OFFSETS */
3435 #endif /* DEBUGGING */
3437 This means we convert either the first branch or the first Exact,
3438 depending on whether the thing following (in 'last') is a branch
3439 or not and whther first is the startbranch (ie is it a sub part of
3440 the alternation or is it the whole thing.)
3441 Assuming its a sub part we convert the EXACT otherwise we convert
3442 the whole branch sequence, including the first.
3444 /* Find the node we are going to overwrite */
3445 if ( first != startbranch || OP( last ) == BRANCH ) {
3446 /* branch sub-chain */
3447 NEXT_OFF( first ) = (U16)(last - first);
3448 #ifdef RE_TRACK_PATTERN_OFFSETS
3450 mjd_offset= Node_Offset((convert));
3451 mjd_nodelen= Node_Length((convert));
3454 /* whole branch chain */
3456 #ifdef RE_TRACK_PATTERN_OFFSETS
3459 const regnode *nop = NEXTOPER( convert );
3460 mjd_offset= Node_Offset((nop));
3461 mjd_nodelen= Node_Length((nop));
3465 Perl_re_indentf( aTHX_ "MJD offset:%" UVuf " MJD length:%" UVuf "\n",
3467 (UV)mjd_offset, (UV)mjd_nodelen)
3470 /* But first we check to see if there is a common prefix we can
3471 split out as an EXACT and put in front of the TRIE node. */
3472 trie->startstate= 1;
3473 if ( trie->bitmap && !widecharmap && !trie->jump ) {
3474 /* we want to find the first state that has more than
3475 * one transition, if that state is not the first state
3476 * then we have a common prefix which we can remove.
3479 for ( state = 1 ; state < trie->statecount-1 ; state++ ) {
3481 I32 first_ofs = -1; /* keeps track of the ofs of the first
3482 transition, -1 means none */
3484 const U32 base = trie->states[ state ].trans.base;
3486 /* does this state terminate an alternation? */
3487 if ( trie->states[state].wordnum )
3490 for ( ofs = 0 ; ofs < trie->uniquecharcount ; ofs++ ) {
3491 if ( ( base + ofs >= trie->uniquecharcount ) &&
3492 ( base + ofs - trie->uniquecharcount < trie->lasttrans ) &&
3493 trie->trans[ base + ofs - trie->uniquecharcount ].check == state )
3495 if ( ++count > 1 ) {
3496 /* we have more than one transition */
3499 /* if this is the first state there is no common prefix
3500 * to extract, so we can exit */
3501 if ( state == 1 ) break;
3502 tmp = av_fetch( revcharmap, ofs, 0);
3503 ch = (U8*)SvPV_nolen_const( *tmp );
3505 /* if we are on count 2 then we need to initialize the
3506 * bitmap, and store the previous char if there was one
3509 /* clear the bitmap */
3510 Zero(trie->bitmap, ANYOF_BITMAP_SIZE, char);
3512 Perl_re_indentf( aTHX_ "New Start State=%" UVuf " Class: [",
3515 if (first_ofs >= 0) {
3516 SV ** const tmp = av_fetch( revcharmap, first_ofs, 0);
3517 const U8 * const ch = (U8*)SvPV_nolen_const( *tmp );
3519 TRIE_BITMAP_SET_FOLDED(trie,*ch, folder);
3521 Perl_re_printf( aTHX_ "%s", (char*)ch)
3525 /* store the current firstchar in the bitmap */
3526 TRIE_BITMAP_SET_FOLDED(trie,*ch, folder);
3527 DEBUG_OPTIMISE_r(Perl_re_printf( aTHX_ "%s", ch));
3533 /* This state has only one transition, its transition is part
3534 * of a common prefix - we need to concatenate the char it
3535 * represents to what we have so far. */
3536 SV **tmp = av_fetch( revcharmap, first_ofs, 0);
3538 char *ch = SvPV( *tmp, len );
3540 SV *sv=sv_newmortal();
3541 Perl_re_indentf( aTHX_ "Prefix State: %" UVuf " Ofs:%" UVuf " Char='%s'\n",
3543 (UV)state, (UV)first_ofs,
3544 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 6,
3545 PL_colors[0], PL_colors[1],
3546 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
3547 PERL_PV_ESCAPE_FIRSTCHAR
3552 OP( convert ) = nodetype;
3553 str=STRING(convert);
3556 STR_LEN(convert) += len;
3562 DEBUG_OPTIMISE_r(Perl_re_printf( aTHX_ "]\n"));
3567 trie->prefixlen = (state-1);
3569 regnode *n = convert+NODE_SZ_STR(convert);
3570 NEXT_OFF(convert) = NODE_SZ_STR(convert);
3571 trie->startstate = state;
3572 trie->minlen -= (state - 1);
3573 trie->maxlen -= (state - 1);
3575 /* At least the UNICOS C compiler choked on this
3576 * being argument to DEBUG_r(), so let's just have
3579 #ifdef PERL_EXT_RE_BUILD
3585 regnode *fix = convert;
3586 U32 word = trie->wordcount;
3587 #ifdef RE_TRACK_PATTERN_OFFSETS
3590 Set_Node_Offset_Length(convert, mjd_offset, state - 1);
3591 while( ++fix < n ) {
3592 Set_Node_Offset_Length(fix, 0, 0);
3595 SV ** const tmp = av_fetch( trie_words, word, 0 );
3597 if ( STR_LEN(convert) <= SvCUR(*tmp) )
3598 sv_chop(*tmp, SvPV_nolen(*tmp) + STR_LEN(convert));
3600 sv_chop(*tmp, SvPV_nolen(*tmp) + SvCUR(*tmp));
3608 NEXT_OFF(convert) = (U16)(tail - convert);
3609 DEBUG_r(optimize= n);
3615 if ( trie->maxlen ) {
3616 NEXT_OFF( convert ) = (U16)(tail - convert);
3617 ARG_SET( convert, data_slot );
3618 /* Store the offset to the first unabsorbed branch in
3619 jump[0], which is otherwise unused by the jump logic.
3620 We use this when dumping a trie and during optimisation. */
3622 trie->jump[0] = (U16)(nextbranch - convert);
3624 /* If the start state is not accepting (meaning there is no empty string/NOTHING)
3625 * and there is a bitmap
3626 * and the first "jump target" node we found leaves enough room
3627 * then convert the TRIE node into a TRIEC node, with the bitmap
3628 * embedded inline in the opcode - this is hypothetically faster.
3630 if ( !trie->states[trie->startstate].wordnum
3632 && ( (char *)jumper - (char *)convert) >= (int)sizeof(struct regnode_charclass) )
3634 OP( convert ) = TRIEC;
3635 Copy(trie->bitmap, ((struct regnode_charclass *)convert)->bitmap, ANYOF_BITMAP_SIZE, char);
3636 PerlMemShared_free(trie->bitmap);
3639 OP( convert ) = TRIE;
3641 /* store the type in the flags */
3642 convert->flags = nodetype;
3646 + regarglen[ OP( convert ) ];
3648 /* XXX We really should free up the resource in trie now,
3649 as we won't use them - (which resources?) dmq */
3651 /* needed for dumping*/
3652 DEBUG_r(if (optimize) {
3653 regnode *opt = convert;
3655 while ( ++opt < optimize) {
3656 Set_Node_Offset_Length(opt, 0, 0);
3659 Try to clean up some of the debris left after the
3662 while( optimize < jumper ) {
3663 Track_Code( mjd_nodelen += Node_Length((optimize)); );
3664 OP( optimize ) = OPTIMIZED;
3665 Set_Node_Offset_Length(optimize, 0, 0);
3668 Set_Node_Offset_Length(convert, mjd_offset, mjd_nodelen);
3670 } /* end node insert */
3672 /* Finish populating the prev field of the wordinfo array. Walk back
3673 * from each accept state until we find another accept state, and if
3674 * so, point the first word's .prev field at the second word. If the
3675 * second already has a .prev field set, stop now. This will be the
3676 * case either if we've already processed that word's accept state,
3677 * or that state had multiple words, and the overspill words were
3678 * already linked up earlier.
3685 for (word=1; word <= trie->wordcount; word++) {
3687 if (trie->wordinfo[word].prev)
3689 state = trie->wordinfo[word].accept;
3691 state = prev_states[state];
3694 prev = trie->states[state].wordnum;
3698 trie->wordinfo[word].prev = prev;
3700 Safefree(prev_states);
3704 /* and now dump out the compressed format */
3705 DEBUG_TRIE_COMPILE_r(dump_trie(trie, widecharmap, revcharmap, depth+1));
3707 RExC_rxi->data->data[ data_slot + 1 ] = (void*)widecharmap;
3709 RExC_rxi->data->data[ data_slot + TRIE_WORDS_OFFSET ] = (void*)trie_words;
3710 RExC_rxi->data->data[ data_slot + 3 ] = (void*)revcharmap;
3712 SvREFCNT_dec_NN(revcharmap);
3716 : trie->startstate>1
3722 S_construct_ahocorasick_from_trie(pTHX_ RExC_state_t *pRExC_state, regnode *source, U32 depth)
3724 /* The Trie is constructed and compressed now so we can build a fail array if
3727 This is basically the Aho-Corasick algorithm. Its from exercise 3.31 and
3729 "Red Dragon" -- Compilers, principles, techniques, and tools. Aho, Sethi,
3733 We find the fail state for each state in the trie, this state is the longest
3734 proper suffix of the current state's 'word' that is also a proper prefix of
3735 another word in our trie. State 1 represents the word '' and is thus the
3736 default fail state. This allows the DFA not to have to restart after its
3737 tried and failed a word at a given point, it simply continues as though it
3738 had been matching the other word in the first place.
3740 'abcdgu'=~/abcdefg|cdgu/
3741 When we get to 'd' we are still matching the first word, we would encounter
3742 'g' which would fail, which would bring us to the state representing 'd' in
3743 the second word where we would try 'g' and succeed, proceeding to match
3746 /* add a fail transition */
3747 const U32 trie_offset = ARG(source);
3748 reg_trie_data *trie=(reg_trie_data *)RExC_rxi->data->data[trie_offset];
3750 const U32 ucharcount = trie->uniquecharcount;
3751 const U32 numstates = trie->statecount;
3752 const U32 ubound = trie->lasttrans + ucharcount;
3756 U32 base = trie->states[ 1 ].trans.base;
3759 const U32 data_slot = add_data( pRExC_state, STR_WITH_LEN("T"));
3761 GET_RE_DEBUG_FLAGS_DECL;
3763 PERL_ARGS_ASSERT_CONSTRUCT_AHOCORASICK_FROM_TRIE;
3764 PERL_UNUSED_CONTEXT;
3766 PERL_UNUSED_ARG(depth);
3769 if ( OP(source) == TRIE ) {
3770 struct regnode_1 *op = (struct regnode_1 *)
3771 PerlMemShared_calloc(1, sizeof(struct regnode_1));
3772 StructCopy(source, op, struct regnode_1);
3773 stclass = (regnode *)op;
3775 struct regnode_charclass *op = (struct regnode_charclass *)
3776 PerlMemShared_calloc(1, sizeof(struct regnode_charclass));
3777 StructCopy(source, op, struct regnode_charclass);
3778 stclass = (regnode *)op;
3780 OP(stclass)+=2; /* convert the TRIE type to its AHO-CORASICK equivalent */
3782 ARG_SET( stclass, data_slot );
3783 aho = (reg_ac_data *) PerlMemShared_calloc( 1, sizeof(reg_ac_data) );
3784 RExC_rxi->data->data[ data_slot ] = (void*)aho;
3785 aho->trie=trie_offset;
3786 aho->states=(reg_trie_state *)PerlMemShared_malloc( numstates * sizeof(reg_trie_state) );
3787 Copy( trie->states, aho->states, numstates, reg_trie_state );
3788 Newx( q, numstates, U32);
3789 aho->fail = (U32 *) PerlMemShared_calloc( numstates, sizeof(U32) );
3792 /* initialize fail[0..1] to be 1 so that we always have
3793 a valid final fail state */
3794 fail[ 0 ] = fail[ 1 ] = 1;
3796 for ( charid = 0; charid < ucharcount ; charid++ ) {
3797 const U32 newstate = TRIE_TRANS_STATE( 1, base, ucharcount, charid, 0 );
3799 q[ q_write ] = newstate;
3800 /* set to point at the root */
3801 fail[ q[ q_write++ ] ]=1;
3804 while ( q_read < q_write) {
3805 const U32 cur = q[ q_read++ % numstates ];
3806 base = trie->states[ cur ].trans.base;
3808 for ( charid = 0 ; charid < ucharcount ; charid++ ) {
3809 const U32 ch_state = TRIE_TRANS_STATE( cur, base, ucharcount, charid, 1 );
3811 U32 fail_state = cur;
3814 fail_state = fail[ fail_state ];
3815 fail_base = aho->states[ fail_state ].trans.base;
3816 } while ( !TRIE_TRANS_STATE( fail_state, fail_base, ucharcount, charid, 1 ) );
3818 fail_state = TRIE_TRANS_STATE( fail_state, fail_base, ucharcount, charid, 1 );
3819 fail[ ch_state ] = fail_state;
3820 if ( !aho->states[ ch_state ].wordnum && aho->states[ fail_state ].wordnum )
3822 aho->states[ ch_state ].wordnum = aho->states[ fail_state ].wordnum;
3824 q[ q_write++ % numstates] = ch_state;
3828 /* restore fail[0..1] to 0 so that we "fall out" of the AC loop
3829 when we fail in state 1, this allows us to use the
3830 charclass scan to find a valid start char. This is based on the principle
3831 that theres a good chance the string being searched contains lots of stuff
3832 that cant be a start char.
3834 fail[ 0 ] = fail[ 1 ] = 0;
3835 DEBUG_TRIE_COMPILE_r({
3836 Perl_re_indentf( aTHX_ "Stclass Failtable (%" UVuf " states): 0",
3837 depth, (UV)numstates
3839 for( q_read=1; q_read<numstates; q_read++ ) {
3840 Perl_re_printf( aTHX_ ", %" UVuf, (UV)fail[q_read]);
3842 Perl_re_printf( aTHX_ "\n");
3845 /*RExC_seen |= REG_TRIEDFA_SEEN;*/
3850 /* The below joins as many adjacent EXACTish nodes as possible into a single
3851 * one. The regop may be changed if the node(s) contain certain sequences that
3852 * require special handling. The joining is only done if:
3853 * 1) there is room in the current conglomerated node to entirely contain the
3855 * 2) they are compatible node types
3857 * The adjacent nodes actually may be separated by NOTHING-kind nodes, and
3858 * these get optimized out
3860 * XXX khw thinks this should be enhanced to fill EXACT (at least) nodes as full
3861 * as possible, even if that means splitting an existing node so that its first
3862 * part is moved to the preceeding node. This would maximise the efficiency of
3863 * memEQ during matching.
3865 * If a node is to match under /i (folded), the number of characters it matches
3866 * can be different than its character length if it contains a multi-character
3867 * fold. *min_subtract is set to the total delta number of characters of the
3870 * And *unfolded_multi_char is set to indicate whether or not the node contains
3871 * an unfolded multi-char fold. This happens when it won't be known until
3872 * runtime whether the fold is valid or not; namely
3873 * 1) for EXACTF nodes that contain LATIN SMALL LETTER SHARP S, as only if the
3874 * target string being matched against turns out to be UTF-8 is that fold
3876 * 2) for EXACTFL nodes whose folding rules depend on the locale in force at
3878 * (Multi-char folds whose components are all above the Latin1 range are not
3879 * run-time locale dependent, and have already been folded by the time this
3880 * function is called.)
3882 * This is as good a place as any to discuss the design of handling these
3883 * multi-character fold sequences. It's been wrong in Perl for a very long
3884 * time. There are three code points in Unicode whose multi-character folds
3885 * were long ago discovered to mess things up. The previous designs for
3886 * dealing with these involved assigning a special node for them. This
3887 * approach doesn't always work, as evidenced by this example:
3888 * "\xDFs" =~ /s\xDF/ui # Used to fail before these patches
3889 * Both sides fold to "sss", but if the pattern is parsed to create a node that
3890 * would match just the \xDF, it won't be able to handle the case where a
3891 * successful match would have to cross the node's boundary. The new approach
3892 * that hopefully generally solves the problem generates an EXACTFUP node
3893 * that is "sss" in this case.
3895 * It turns out that there are problems with all multi-character folds, and not
3896 * just these three. Now the code is general, for all such cases. The
3897 * approach taken is:
3898 * 1) This routine examines each EXACTFish node that could contain multi-
3899 * character folded sequences. Since a single character can fold into
3900 * such a sequence, the minimum match length for this node is less than
3901 * the number of characters in the node. This routine returns in
3902 * *min_subtract how many characters to subtract from the the actual
3903 * length of the string to get a real minimum match length; it is 0 if
3904 * there are no multi-char foldeds. This delta is used by the caller to
3905 * adjust the min length of the match, and the delta between min and max,
3906 * so that the optimizer doesn't reject these possibilities based on size
3909 * 2) For the sequence involving the LATIN SMALL LETTER SHARP S (U+00DF)
3910 * under /u, we fold it to 'ss' in regatom(), and in this routine, after
3911 * joining, we scan for occurrences of the sequence 'ss' in non-UTF-8
3912 * EXACTFU nodes. The node type of such nodes is then changed to
3913 * EXACTFUP, indicating it is problematic, and needs careful handling.
3914 * (The procedures in step 1) above are sufficient to handle this case in
3915 * UTF-8 encoded nodes.) The reason this is problematic is that this is
3916 * the only case where there is a possible fold length change in non-UTF-8
3917 * patterns. By reserving a special node type for problematic cases, the
3918 * far more common regular EXACTFU nodes can be processed faster.
3919 * regexec.c takes advantage of this.
3921 * EXACTFUP has been created as a grab-bag for (hopefully uncommon)
3922 * problematic cases. These all only occur when the pattern is not
3923 * UTF-8. In addition to the 'ss' sequence where there is a possible fold
3924 * length change, it handles the situation where the string cannot be
3925 * entirely folded. The strings in an EXACTFish node are folded as much
3926 * as possible during compilation in regcomp.c. This saves effort in
3927 * regex matching. By using an EXACTFUP node when it is not possible to
3928 * fully fold at compile time, regexec.c can know that everything in an
3929 * EXACTFU node is folded, so folding can be skipped at runtime. The only
3930 * case where folding in EXACTFU nodes can't be done at compile time is
3931 * the presumably uncommon MICRO SIGN, when the pattern isn't UTF-8. This
3932 * is because its fold requires UTF-8 to represent. Thus EXACTFUP nodes
3933 * handle two very different cases. Alternatively, there could have been
3934 * a node type where there are length changes, one for unfolded, and one
3935 * for both. If yet another special case needed to be created, the number
3936 * of required node types would have to go to 7. khw figures that even
3937 * though there are plenty of node types to spare, that the maintenance
3938 * cost wasn't worth the small speedup of doing it that way, especially
3939 * since he thinks the MICRO SIGN is rarely encountered in practice.
3941 * There are other cases where folding isn't done at compile time, but
3942 * none of them are under /u, and hence not for EXACTFU nodes. The folds
3943 * in EXACTFL nodes aren't known until runtime, and vary as the locale
3944 * changes. Some folds in EXACTF depend on if the runtime target string
3945 * is UTF-8 or not. (regatom() will create an EXACTFU node even under /di
3946 * when no fold in it depends on the UTF-8ness of the target string.)
3948 * 3) A problem remains for unfolded multi-char folds. (These occur when the
3949 * validity of the fold won't be known until runtime, and so must remain
3950 * unfolded for now. This happens for the sharp s in EXACTF and EXACTFAA
3951 * nodes when the pattern isn't in UTF-8. (Note, BTW, that there cannot
3952 * be an EXACTF node with a UTF-8 pattern.) They also occur for various
3953 * folds in EXACTFL nodes, regardless of the UTF-ness of the pattern.)
3954 * The reason this is a problem is that the optimizer part of regexec.c
3955 * (probably unwittingly, in Perl_regexec_flags()) makes an assumption
3956 * that a character in the pattern corresponds to at most a single
3957 * character in the target string. (And I do mean character, and not byte
3958 * here, unlike other parts of the documentation that have never been
3959 * updated to account for multibyte Unicode.) Sharp s in EXACTF and
3960 * EXACTFL nodes can match the two character string 'ss'; in EXACTFAA
3961 * nodes it can match "\x{17F}\x{17F}". These, along with other ones in
3962 * EXACTFL nodes, violate the assumption, and they are the only instances
3963 * where it is violated. I'm reluctant to try to change the assumption,
3964 * as the code involved is impenetrable to me (khw), so instead the code
3965 * here punts. This routine examines EXACTFL nodes, and (when the pattern
3966 * isn't UTF-8) EXACTF and EXACTFAA for such unfolded folds, and returns a
3967 * boolean indicating whether or not the node contains such a fold. When
3968 * it is true, the caller sets a flag that later causes the optimizer in
3969 * this file to not set values for the floating and fixed string lengths,
3970 * and thus avoids the optimizer code in regexec.c that makes the invalid
3971 * assumption. Thus, there is no optimization based on string lengths for
3972 * EXACTFL nodes that contain these few folds, nor for non-UTF8-pattern
3973 * EXACTF and EXACTFAA nodes that contain the sharp s. (The reason the
3974 * assumption is wrong only in these cases is that all other non-UTF-8
3975 * folds are 1-1; and, for UTF-8 patterns, we pre-fold all other folds to
3976 * their expanded versions. (Again, we can't prefold sharp s to 'ss' in
3977 * EXACTF nodes because we don't know at compile time if it actually
3978 * matches 'ss' or not. For EXACTF nodes it will match iff the target
3979 * string is in UTF-8. This is in contrast to EXACTFU nodes, where it
3980 * always matches; and EXACTFAA where it never does. In an EXACTFAA node
3981 * in a UTF-8 pattern, sharp s is folded to "\x{17F}\x{17F}, avoiding the
3982 * problem; but in a non-UTF8 pattern, folding it to that above-Latin1
3983 * string would require the pattern to be forced into UTF-8, the overhead
3984 * of which we want to avoid. Similarly the unfolded multi-char folds in
3985 * EXACTFL nodes will match iff the locale at the time of match is a UTF-8
3988 * Similarly, the code that generates tries doesn't currently handle
3989 * not-already-folded multi-char folds, and it looks like a pain to change
3990 * that. Therefore, trie generation of EXACTFAA nodes with the sharp s
3991 * doesn't work. Instead, such an EXACTFAA is turned into a new regnode,
3992 * EXACTFAA_NO_TRIE, which the trie code knows not to handle. Most people
3993 * using /iaa matching will be doing so almost entirely with ASCII
3994 * strings, so this should rarely be encountered in practice */
3996 #define JOIN_EXACT(scan,min_subtract,unfolded_multi_char, flags) \
3997 if (PL_regkind[OP(scan)] == EXACT) \
3998 join_exact(pRExC_state,(scan),(min_subtract),unfolded_multi_char, (flags), NULL, depth+1)
4001 S_join_exact(pTHX_ RExC_state_t *pRExC_state, regnode *scan,
4002 UV *min_subtract, bool *unfolded_multi_char,
4003 U32 flags, regnode *val, U32 depth)
4005 /* Merge several consecutive EXACTish nodes into one. */
4007 regnode *n = regnext(scan);
4009 regnode *next = scan + NODE_SZ_STR(scan);
4013 regnode *stop = scan;
4014 GET_RE_DEBUG_FLAGS_DECL;
4016 PERL_UNUSED_ARG(depth);
4019 PERL_ARGS_ASSERT_JOIN_EXACT;
4020 #ifndef EXPERIMENTAL_INPLACESCAN
4021 PERL_UNUSED_ARG(flags);
4022 PERL_UNUSED_ARG(val);
4024 DEBUG_PEEP("join", scan, depth, 0);
4026 assert(PL_regkind[OP(scan)] == EXACT);
4028 /* Look through the subsequent nodes in the chain. Skip NOTHING, merge
4029 * EXACT ones that are mergeable to the current one. */
4031 && ( PL_regkind[OP(n)] == NOTHING
4032 || (stringok && PL_regkind[OP(n)] == EXACT))
4034 && NEXT_OFF(scan) + NEXT_OFF(n) < I16_MAX)
4037 if (OP(n) == TAIL || n > next)
4039 if (PL_regkind[OP(n)] == NOTHING) {
4040 DEBUG_PEEP("skip:", n, depth, 0);
4041 NEXT_OFF(scan) += NEXT_OFF(n);
4042 next = n + NODE_STEP_REGNODE;
4049 else if (stringok) {
4050 const unsigned int oldl = STR_LEN(scan);
4051 regnode * const nnext = regnext(n);
4053 /* XXX I (khw) kind of doubt that this works on platforms (should
4054 * Perl ever run on one) where U8_MAX is above 255 because of lots
4055 * of other assumptions */
4056 /* Don't join if the sum can't fit into a single node */
4057 if (oldl + STR_LEN(n) > U8_MAX)
4060 /* Joining something that requires UTF-8 with something that
4061 * doesn't, means the result requires UTF-8. */
4062 if (OP(scan) == EXACT && (OP(n) == EXACT_ONLY8)) {
4063 OP(scan) = EXACT_ONLY8;
4065 else if (OP(scan) == EXACT_ONLY8 && (OP(n) == EXACT)) {
4066 ; /* join is compatible, no need to change OP */
4068 else if ((OP(scan) == EXACTFU) && (OP(n) == EXACTFU_ONLY8)) {
4069 OP(scan) = EXACTFU_ONLY8;
4071 else if ((OP(scan) == EXACTFU_ONLY8) && (OP(n) == EXACTFU)) {
4072 ; /* join is compatible, no need to change OP */
4074 else if (OP(scan) == EXACTFU && OP(n) == EXACTFU) {
4075 ; /* join is compatible, no need to change OP */
4077 else if (OP(scan) == EXACTFU && OP(n) == EXACTFU_S_EDGE) {
4079 /* Under /di, temporary EXACTFU_S_EDGE nodes are generated,
4080 * which can join with EXACTFU ones. We check for this case
4081 * here. These need to be resolved to either EXACTFU or
4082 * EXACTF at joining time. They have nothing in them that
4083 * would forbid them from being the more desirable EXACTFU
4084 * nodes except that they begin and/or end with a single [Ss].
4085 * The reason this is problematic is because they could be
4086 * joined in this loop with an adjacent node that ends and/or
4087 * begins with [Ss] which would then form the sequence 'ss',
4088 * which matches differently under /di than /ui, in which case
4089 * EXACTFU can't be used. If the 'ss' sequence doesn't get
4090 * formed, the nodes get absorbed into any adjacent EXACTFU
4091 * node. And if the only adjacent node is EXACTF, they get
4092 * absorbed into that, under the theory that a longer node is
4093 * better than two shorter ones, even if one is EXACTFU. Note
4094 * that EXACTFU_ONLY8 is generated only for UTF-8 patterns,
4095 * and the EXACTFU_S_EDGE ones only for non-UTF-8. */
4097 if (STRING(n)[STR_LEN(n)-1] == 's') {
4099 /* Here the joined node would end with 's'. If the node
4100 * following the combination is an EXACTF one, it's better to
4101 * join this trailing edge 's' node with that one, leaving the
4102 * current one in 'scan' be the more desirable EXACTFU */
4103 if (OP(nnext) == EXACTF) {
4107 OP(scan) = EXACTFU_S_EDGE;
4109 } /* Otherwise, the beginning 's' of the 2nd node just
4110 becomes an interior 's' in 'scan' */
4112 else if (OP(scan) == EXACTF && OP(n) == EXACTF) {
4113 ; /* join is compatible, no need to change OP */
4115 else if (OP(scan) == EXACTF && OP(n) == EXACTFU_S_EDGE) {
4117 /* EXACTF nodes are compatible for joining with EXACTFU_S_EDGE
4118 * nodes. But the latter nodes can be also joined with EXACTFU
4119 * ones, and that is a better outcome, so if the node following
4120 * 'n' is EXACTFU, quit now so that those two can be joined
4122 if (OP(nnext) == EXACTFU) {
4126 /* The join is compatible, and the combined node will be
4127 * EXACTF. (These don't care if they begin or end with 's' */
4129 else if (OP(scan) == EXACTFU_S_EDGE && OP(n) == EXACTFU_S_EDGE) {
4130 if ( STRING(scan)[STR_LEN(scan)-1] == 's'
4131 && STRING(n)[0] == 's')
4133 /* When combined, we have the sequence 'ss', which means we
4134 * have to remain /di */
4138 else if (OP(scan) == EXACTFU_S_EDGE && OP(n) == EXACTFU) {
4139 if (STRING(n)[0] == 's') {
4140 ; /* Here the join is compatible and the combined node
4141 starts with 's', no need to change OP */
4143 else { /* Now the trailing 's' is in the interior */
4147 else if (OP(scan) == EXACTFU_S_EDGE && OP(n) == EXACTF) {
4149 /* The join is compatible, and the combined node will be
4150 * EXACTF. (These don't care if they begin or end with 's' */
4153 else if (OP(scan) != OP(n)) {
4155 /* The only other compatible joinings are the same node type */
4159 DEBUG_PEEP("merg", n, depth, 0);
4162 NEXT_OFF(scan) += NEXT_OFF(n);
4163 STR_LEN(scan) += STR_LEN(n);
4164 next = n + NODE_SZ_STR(n);
4165 /* Now we can overwrite *n : */
4166 Move(STRING(n), STRING(scan) + oldl, STR_LEN(n), char);
4174 #ifdef EXPERIMENTAL_INPLACESCAN
4175 if (flags && !NEXT_OFF(n)) {
4176 DEBUG_PEEP("atch", val, depth, 0);
4177 if (reg_off_by_arg[OP(n)]) {
4178 ARG_SET(n, val - n);
4181 NEXT_OFF(n) = val - n;
4188 /* This temporary node can now be turned into EXACTFU, and must, as
4189 * regexec.c doesn't handle it */
4190 if (OP(scan) == EXACTFU_S_EDGE) {
4195 *unfolded_multi_char = FALSE;
4197 /* Here, all the adjacent mergeable EXACTish nodes have been merged. We
4198 * can now analyze for sequences of problematic code points. (Prior to
4199 * this final joining, sequences could have been split over boundaries, and
4200 * hence missed). The sequences only happen in folding, hence for any
4201 * non-EXACT EXACTish node */
4202 if (OP(scan) != EXACT && OP(scan) != EXACT_ONLY8 && OP(scan) != EXACTL) {
4203 U8* s0 = (U8*) STRING(scan);
4205 U8* s_end = s0 + STR_LEN(scan);
4207 int total_count_delta = 0; /* Total delta number of characters that
4208 multi-char folds expand to */
4210 /* One pass is made over the node's string looking for all the
4211 * possibilities. To avoid some tests in the loop, there are two main
4212 * cases, for UTF-8 patterns (which can't have EXACTF nodes) and
4217 if (OP(scan) == EXACTFL) {
4220 /* An EXACTFL node would already have been changed to another
4221 * node type unless there is at least one character in it that
4222 * is problematic; likely a character whose fold definition
4223 * won't be known until runtime, and so has yet to be folded.
4224 * For all but the UTF-8 locale, folds are 1-1 in length, but
4225 * to handle the UTF-8 case, we need to create a temporary
4226 * folded copy using UTF-8 locale rules in order to analyze it.
4227 * This is because our macros that look to see if a sequence is
4228 * a multi-char fold assume everything is folded (otherwise the
4229 * tests in those macros would be too complicated and slow).
4230 * Note that here, the non-problematic folds will have already
4231 * been done, so we can just copy such characters. We actually
4232 * don't completely fold the EXACTFL string. We skip the
4233 * unfolded multi-char folds, as that would just create work
4234 * below to figure out the size they already are */
4236 Newx(folded, UTF8_MAX_FOLD_CHAR_EXPAND * STR_LEN(scan) + 1, U8);
4239 STRLEN s_len = UTF8SKIP(s);
4240 if (! is_PROBLEMATIC_LOCALE_FOLD_utf8(s)) {
4241 Copy(s, d, s_len, U8);
4244 else if (is_FOLDS_TO_MULTI_utf8(s)) {
4245 *unfolded_multi_char = TRUE;
4246 Copy(s, d, s_len, U8);
4249 else if (isASCII(*s)) {
4250 *(d++) = toFOLD(*s);
4254 _toFOLD_utf8_flags(s, s_end, d, &len, FOLD_FLAGS_FULL);
4260 /* Point the remainder of the routine to look at our temporary
4264 } /* End of creating folded copy of EXACTFL string */
4266 /* Examine the string for a multi-character fold sequence. UTF-8
4267 * patterns have all characters pre-folded by the time this code is
4269 while (s < s_end - 1) /* Can stop 1 before the end, as minimum
4270 length sequence we are looking for is 2 */
4272 int count = 0; /* How many characters in a multi-char fold */
4273 int len = is_MULTI_CHAR_FOLD_utf8_safe(s, s_end);
4274 if (! len) { /* Not a multi-char fold: get next char */
4279 { /* Here is a generic multi-char fold. */
4280 U8* multi_end = s + len;
4282 /* Count how many characters are in it. In the case of
4283 * /aa, no folds which contain ASCII code points are
4284 * allowed, so check for those, and skip if found. */
4285 if (OP(scan) != EXACTFAA && OP(scan) != EXACTFAA_NO_TRIE) {
4286 count = utf8_length(s, multi_end);
4290 while (s < multi_end) {
4293 goto next_iteration;
4303 /* The delta is how long the sequence is minus 1 (1 is how long
4304 * the character that folds to the sequence is) */
4305 total_count_delta += count - 1;
4309 /* We created a temporary folded copy of the string in EXACTFL
4310 * nodes. Therefore we need to be sure it doesn't go below zero,
4311 * as the real string could be shorter */
4312 if (OP(scan) == EXACTFL) {
4313 int total_chars = utf8_length((U8*) STRING(scan),
4314 (U8*) STRING(scan) + STR_LEN(scan));
4315 if (total_count_delta > total_chars) {
4316 total_count_delta = total_chars;
4320 *min_subtract += total_count_delta;
4323 else if (OP(scan) == EXACTFAA) {
4325 /* Non-UTF-8 pattern, EXACTFAA node. There can't be a multi-char
4326 * fold to the ASCII range (and there are no existing ones in the
4327 * upper latin1 range). But, as outlined in the comments preceding
4328 * this function, we need to flag any occurrences of the sharp s.
4329 * This character forbids trie formation (because of added
4331 #if UNICODE_MAJOR_VERSION > 3 /* no multifolds in early Unicode */ \
4332 || (UNICODE_MAJOR_VERSION == 3 && ( UNICODE_DOT_VERSION > 0) \
4333 || UNICODE_DOT_DOT_VERSION > 0)
4335 if (*s == LATIN_SMALL_LETTER_SHARP_S) {
4336 OP(scan) = EXACTFAA_NO_TRIE;
4337 *unfolded_multi_char = TRUE;
4345 /* Non-UTF-8 pattern, not EXACTFAA node. Look for the multi-char
4346 * folds that are all Latin1. As explained in the comments
4347 * preceding this function, we look also for the sharp s in EXACTF
4348 * and EXACTFL nodes; it can be in the final position. Otherwise
4349 * we can stop looking 1 byte earlier because have to find at least
4350 * two characters for a multi-fold */
4351 const U8* upper = (OP(scan) == EXACTF || OP(scan) == EXACTFL)
4356 int len = is_MULTI_CHAR_FOLD_latin1_safe(s, s_end);
4357 if (! len) { /* Not a multi-char fold. */
4358 if (*s == LATIN_SMALL_LETTER_SHARP_S
4359 && (OP(scan) == EXACTF || OP(scan) == EXACTFL))
4361 *unfolded_multi_char = TRUE;
4368 && isALPHA_FOLD_EQ(*s, 's')
4369 && isALPHA_FOLD_EQ(*(s+1), 's'))
4372 /* EXACTF nodes need to know that the minimum length
4373 * changed so that a sharp s in the string can match this
4374 * ss in the pattern, but they remain EXACTF nodes, as they
4375 * won't match this unless the target string is is UTF-8,
4376 * which we don't know until runtime. EXACTFL nodes can't
4377 * transform into EXACTFU nodes */
4378 if (OP(scan) != EXACTF && OP(scan) != EXACTFL) {
4379 OP(scan) = EXACTFUP;
4383 *min_subtract += len - 1;
4389 if ( STR_LEN(scan) == 1
4390 && isALPHA_A(* STRING(scan))
4391 && ( OP(scan) == EXACTFAA
4392 || ( OP(scan) == EXACTFU
4393 && ! HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE(* STRING(scan)))))
4395 U8 mask = ~ ('A' ^ 'a'); /* These differ in just one bit */
4397 /* Replace a length 1 ASCII fold pair node with an ANYOFM node,
4398 * with the mask set to the complement of the bit that differs
4399 * between upper and lower case, and the lowest code point of the
4400 * pair (which the '&' forces) */
4402 ARG_SET(scan, *STRING(scan) & mask);
4408 /* Allow dumping but overwriting the collection of skipped
4409 * ops and/or strings with fake optimized ops */
4410 n = scan + NODE_SZ_STR(scan);
4418 DEBUG_OPTIMISE_r(if (merged){DEBUG_PEEP("finl", scan, depth, 0);});
4422 /* REx optimizer. Converts nodes into quicker variants "in place".
4423 Finds fixed substrings. */
4425 /* Stops at toplevel WHILEM as well as at "last". At end *scanp is set
4426 to the position after last scanned or to NULL. */
4428 #define INIT_AND_WITHP \
4429 assert(!and_withp); \
4430 Newx(and_withp, 1, regnode_ssc); \
4431 SAVEFREEPV(and_withp)
4435 S_unwind_scan_frames(pTHX_ const void *p)
4437 scan_frame *f= (scan_frame *)p;
4439 scan_frame *n= f->next_frame;
4445 /* the return from this sub is the minimum length that could possibly match */
4447 S_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp,
4448 SSize_t *minlenp, SSize_t *deltap,
4453 regnode_ssc *and_withp,
4454 U32 flags, U32 depth)
4455 /* scanp: Start here (read-write). */
4456 /* deltap: Write maxlen-minlen here. */
4457 /* last: Stop before this one. */
4458 /* data: string data about the pattern */
4459 /* stopparen: treat close N as END */
4460 /* recursed: which subroutines have we recursed into */
4461 /* and_withp: Valid if flags & SCF_DO_STCLASS_OR */
4464 /* There must be at least this number of characters to match */
4467 regnode *scan = *scanp, *next;
4469 int is_inf = (flags & SCF_DO_SUBSTR) && (data->flags & SF_IS_INF);
4470 int is_inf_internal = 0; /* The studied chunk is infinite */
4471 I32 is_par = OP(scan) == OPEN ? ARG(scan) : 0;
4472 scan_data_t data_fake;
4473 SV *re_trie_maxbuff = NULL;
4474 regnode *first_non_open = scan;
4475 SSize_t stopmin = SSize_t_MAX;
4476 scan_frame *frame = NULL;
4477 GET_RE_DEBUG_FLAGS_DECL;
4479 PERL_ARGS_ASSERT_STUDY_CHUNK;
4480 RExC_study_started= 1;
4482 Zero(&data_fake, 1, scan_data_t);
4485 while (first_non_open && OP(first_non_open) == OPEN)
4486 first_non_open=regnext(first_non_open);
4492 RExC_study_chunk_recursed_count++;
4494 DEBUG_OPTIMISE_MORE_r(
4496 Perl_re_indentf( aTHX_ "study_chunk stopparen=%ld recursed_count=%lu depth=%lu recursed_depth=%lu scan=%p last=%p",
4497 depth, (long)stopparen,
4498 (unsigned long)RExC_study_chunk_recursed_count,
4499 (unsigned long)depth, (unsigned long)recursed_depth,
4502 if (recursed_depth) {
4505 for ( j = 0 ; j < recursed_depth ; j++ ) {
4506 for ( i = 0 ; i < (U32)RExC_total_parens ; i++ ) {
4508 PAREN_TEST(RExC_study_chunk_recursed +
4509 ( j * RExC_study_chunk_recursed_bytes), i )
4512 !PAREN_TEST(RExC_study_chunk_recursed +
4513 (( j - 1 ) * RExC_study_chunk_recursed_bytes), i)
4516 Perl_re_printf( aTHX_ " %d",(int)i);
4520 if ( j + 1 < recursed_depth ) {
4521 Perl_re_printf( aTHX_ ",");
4525 Perl_re_printf( aTHX_ "\n");
4528 while ( scan && OP(scan) != END && scan < last ){
4529 UV min_subtract = 0; /* How mmany chars to subtract from the minimum
4530 node length to get a real minimum (because
4531 the folded version may be shorter) */
4532 bool unfolded_multi_char = FALSE;
4533 /* Peephole optimizer: */
4534 DEBUG_STUDYDATA("Peep", data, depth, is_inf);
4535 DEBUG_PEEP("Peep", scan, depth, flags);
4538 /* The reason we do this here is that we need to deal with things like
4539 * /(?:f)(?:o)(?:o)/ which cant be dealt with by the normal EXACT
4540 * parsing code, as each (?:..) is handled by a different invocation of
4543 JOIN_EXACT(scan,&min_subtract, &unfolded_multi_char, 0);
4545 /* Follow the next-chain of the current node and optimize
4546 away all the NOTHINGs from it. */
4547 if (OP(scan) != CURLYX) {
4548 const int max = (reg_off_by_arg[OP(scan)]
4550 /* I32 may be smaller than U16 on CRAYs! */
4551 : (I32_MAX < U16_MAX ? I32_MAX : U16_MAX));
4552 int off = (reg_off_by_arg[OP(scan)] ? ARG(scan) : NEXT_OFF(scan));
4556 /* Skip NOTHING and LONGJMP. */
4557 while ((n = regnext(n))
4558 && ((PL_regkind[OP(n)] == NOTHING && (noff = NEXT_OFF(n)))
4559 || ((OP(n) == LONGJMP) && (noff = ARG(n))))
4560 && off + noff < max)
4562 if (reg_off_by_arg[OP(scan)])
4565 NEXT_OFF(scan) = off;
4568 /* The principal pseudo-switch. Cannot be a switch, since we
4569 look into several different things. */
4570 if ( OP(scan) == DEFINEP ) {
4572 SSize_t deltanext = 0;
4573 SSize_t fake_last_close = 0;
4574 I32 f = SCF_IN_DEFINE;
4576 StructCopy(&zero_scan_data, &data_fake, scan_data_t);
4577 scan = regnext(scan);
4578 assert( OP(scan) == IFTHEN );
4579 DEBUG_PEEP("expect IFTHEN", scan, depth, flags);
4581 data_fake.last_closep= &fake_last_close;
4583 next = regnext(scan);
4584 scan = NEXTOPER(NEXTOPER(scan));
4585 DEBUG_PEEP("scan", scan, depth, flags);
4586 DEBUG_PEEP("next", next, depth, flags);
4588 /* we suppose the run is continuous, last=next...
4589 * NOTE we dont use the return here! */
4590 /* DEFINEP study_chunk() recursion */
4591 (void)study_chunk(pRExC_state, &scan, &minlen,
4592 &deltanext, next, &data_fake, stopparen,
4593 recursed_depth, NULL, f, depth+1);
4598 OP(scan) == BRANCH ||
4599 OP(scan) == BRANCHJ ||
4602 next = regnext(scan);
4605 /* The op(next)==code check below is to see if we
4606 * have "BRANCH-BRANCH", "BRANCHJ-BRANCHJ", "IFTHEN-IFTHEN"
4607 * IFTHEN is special as it might not appear in pairs.
4608 * Not sure whether BRANCH-BRANCHJ is possible, regardless
4609 * we dont handle it cleanly. */
4610 if (OP(next) == code || code == IFTHEN) {
4611 /* NOTE - There is similar code to this block below for
4612 * handling TRIE nodes on a re-study. If you change stuff here
4613 * check there too. */
4614 SSize_t max1 = 0, min1 = SSize_t_MAX, num = 0;
4616 regnode * const startbranch=scan;
4618 if (flags & SCF_DO_SUBSTR) {
4619 /* Cannot merge strings after this. */
4620 scan_commit(pRExC_state, data, minlenp, is_inf);
4623 if (flags & SCF_DO_STCLASS)
4624 ssc_init_zero(pRExC_state, &accum);
4626 while (OP(scan) == code) {
4627 SSize_t deltanext, minnext, fake;
4629 regnode_ssc this_class;
4631 DEBUG_PEEP("Branch", scan, depth, flags);
4634 StructCopy(&zero_scan_data, &data_fake, scan_data_t);
4636 data_fake.whilem_c = data->whilem_c;
4637 data_fake.last_closep = data->last_closep;
4640 data_fake.last_closep = &fake;
4642 data_fake.pos_delta = delta;
4643 next = regnext(scan);
4645 scan = NEXTOPER(scan); /* everything */
4646 if (code != BRANCH) /* everything but BRANCH */
4647 scan = NEXTOPER(scan);
4649 if (flags & SCF_DO_STCLASS) {
4650 ssc_init(pRExC_state, &this_class);
4651 data_fake.start_class = &this_class;
4652 f = SCF_DO_STCLASS_AND;
4654 if (flags & SCF_WHILEM_VISITED_POS)
4655 f |= SCF_WHILEM_VISITED_POS;
4657 /* we suppose the run is continuous, last=next...*/
4658 /* recurse study_chunk() for each BRANCH in an alternation */
4659 minnext = study_chunk(pRExC_state, &scan, minlenp,
4660 &deltanext, next, &data_fake, stopparen,
4661 recursed_depth, NULL, f, depth+1);
4665 if (deltanext == SSize_t_MAX) {
4666 is_inf = is_inf_internal = 1;
4668 } else if (max1 < minnext + deltanext)
4669 max1 = minnext + deltanext;
4671 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
4673 if (data_fake.flags & SCF_SEEN_ACCEPT) {
4674 if ( stopmin > minnext)
4675 stopmin = min + min1;
4676 flags &= ~SCF_DO_SUBSTR;
4678 data->flags |= SCF_SEEN_ACCEPT;
4681 if (data_fake.flags & SF_HAS_EVAL)
4682 data->flags |= SF_HAS_EVAL;
4683 data->whilem_c = data_fake.whilem_c;
4685 if (flags & SCF_DO_STCLASS)
4686 ssc_or(pRExC_state, &accum, (regnode_charclass*)&this_class);
4688 if (code == IFTHEN && num < 2) /* Empty ELSE branch */
4690 if (flags & SCF_DO_SUBSTR) {
4691 data->pos_min += min1;
4692 if (data->pos_delta >= SSize_t_MAX - (max1 - min1))
4693 data->pos_delta = SSize_t_MAX;
4695 data->pos_delta += max1 - min1;
4696 if (max1 != min1 || is_inf)
4697 data->cur_is_floating = 1;
4700 if (delta == SSize_t_MAX
4701 || SSize_t_MAX - delta - (max1 - min1) < 0)
4702 delta = SSize_t_MAX;
4704 delta += max1 - min1;
4705 if (flags & SCF_DO_STCLASS_OR) {
4706 ssc_or(pRExC_state, data->start_class, (regnode_charclass*) &accum);
4708 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
4709 flags &= ~SCF_DO_STCLASS;
4712 else if (flags & SCF_DO_STCLASS_AND) {
4714 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) &accum);
4715 flags &= ~SCF_DO_STCLASS;
4718 /* Switch to OR mode: cache the old value of
4719 * data->start_class */
4721 StructCopy(data->start_class, and_withp, regnode_ssc);
4722 flags &= ~SCF_DO_STCLASS_AND;
4723 StructCopy(&accum, data->start_class, regnode_ssc);
4724 flags |= SCF_DO_STCLASS_OR;
4728 if (PERL_ENABLE_TRIE_OPTIMISATION &&
4729 OP( startbranch ) == BRANCH )
4733 Assuming this was/is a branch we are dealing with: 'scan'
4734 now points at the item that follows the branch sequence,
4735 whatever it is. We now start at the beginning of the
4736 sequence and look for subsequences of
4742 which would be constructed from a pattern like
4745 If we can find such a subsequence we need to turn the first
4746 element into a trie and then add the subsequent branch exact
4747 strings to the trie.
4751 1. patterns where the whole set of branches can be
4754 2. patterns where only a subset can be converted.
4756 In case 1 we can replace the whole set with a single regop
4757 for the trie. In case 2 we need to keep the start and end
4760 'BRANCH EXACT; BRANCH EXACT; BRANCH X'
4761 becomes BRANCH TRIE; BRANCH X;
4763 There is an additional case, that being where there is a
4764 common prefix, which gets split out into an EXACT like node
4765 preceding the TRIE node.
4767 If x(1..n)==tail then we can do a simple trie, if not we make
4768 a "jump" trie, such that when we match the appropriate word
4769 we "jump" to the appropriate tail node. Essentially we turn
4770 a nested if into a case structure of sorts.
4775 if (!re_trie_maxbuff) {
4776 re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, 1);
4777 if (!SvIOK(re_trie_maxbuff))
4778 sv_setiv(re_trie_maxbuff, RE_TRIE_MAXBUF_INIT);
4780 if ( SvIV(re_trie_maxbuff)>=0 ) {
4782 regnode *first = (regnode *)NULL;
4783 regnode *last = (regnode *)NULL;
4784 regnode *tail = scan;
4788 /* var tail is used because there may be a TAIL
4789 regop in the way. Ie, the exacts will point to the
4790 thing following the TAIL, but the last branch will
4791 point at the TAIL. So we advance tail. If we
4792 have nested (?:) we may have to move through several
4796 while ( OP( tail ) == TAIL ) {
4797 /* this is the TAIL generated by (?:) */
4798 tail = regnext( tail );
4802 DEBUG_TRIE_COMPILE_r({
4803 regprop(RExC_rx, RExC_mysv, tail, NULL, pRExC_state);
4804 Perl_re_indentf( aTHX_ "%s %" UVuf ":%s\n",
4806 "Looking for TRIE'able sequences. Tail node is ",
4807 (UV) REGNODE_OFFSET(tail),
4808 SvPV_nolen_const( RExC_mysv )
4814 Step through the branches
4815 cur represents each branch,
4816 noper is the first thing to be matched as part
4818 noper_next is the regnext() of that node.
4820 We normally handle a case like this
4821 /FOO[xyz]|BAR[pqr]/ via a "jump trie" but we also
4822 support building with NOJUMPTRIE, which restricts
4823 the trie logic to structures like /FOO|BAR/.
4825 If noper is a trieable nodetype then the branch is
4826 a possible optimization target. If we are building
4827 under NOJUMPTRIE then we require that noper_next is
4828 the same as scan (our current position in the regex
4831 Once we have two or more consecutive such branches
4832 we can create a trie of the EXACT's contents and
4833 stitch it in place into the program.
4835 If the sequence represents all of the branches in
4836 the alternation we replace the entire thing with a
4839 Otherwise when it is a subsequence we need to
4840 stitch it in place and replace only the relevant
4841 branches. This means the first branch has to remain
4842 as it is used by the alternation logic, and its
4843 next pointer, and needs to be repointed at the item
4844 on the branch chain following the last branch we
4845 have optimized away.
4847 This could be either a BRANCH, in which case the
4848 subsequence is internal, or it could be the item
4849 following the branch sequence in which case the
4850 subsequence is at the end (which does not
4851 necessarily mean the first node is the start of the
4854 TRIE_TYPE(X) is a define which maps the optype to a
4858 ----------------+-----------
4863 EXACTFU_ONLY8 | EXACTFU
4867 EXACTFLU8 | EXACTFLU8
4871 #define TRIE_TYPE(X) ( ( NOTHING == (X) ) \
4873 : ( EXACT == (X) || EXACT_ONLY8 == (X) ) \
4875 : ( EXACTFU == (X) \
4876 || EXACTFU_ONLY8 == (X) \
4877 || EXACTFUP == (X) ) \
4879 : ( EXACTFAA == (X) ) \
4881 : ( EXACTL == (X) ) \
4883 : ( EXACTFLU8 == (X) ) \
4887 /* dont use tail as the end marker for this traverse */
4888 for ( cur = startbranch ; cur != scan ; cur = regnext( cur ) ) {
4889 regnode * const noper = NEXTOPER( cur );
4890 U8 noper_type = OP( noper );
4891 U8 noper_trietype = TRIE_TYPE( noper_type );
4892 #if defined(DEBUGGING) || defined(NOJUMPTRIE)
4893 regnode * const noper_next = regnext( noper );
4894 U8 noper_next_type = (noper_next && noper_next < tail) ? OP(noper_next) : 0;
4895 U8 noper_next_trietype = (noper_next && noper_next < tail) ? TRIE_TYPE( noper_next_type ) :0;
4898 DEBUG_TRIE_COMPILE_r({
4899 regprop(RExC_rx, RExC_mysv, cur, NULL, pRExC_state);
4900 Perl_re_indentf( aTHX_ "- %d:%s (%d)",
4902 REG_NODE_NUM(cur), SvPV_nolen_const( RExC_mysv ), REG_NODE_NUM(cur) );
4904 regprop(RExC_rx, RExC_mysv, noper, NULL, pRExC_state);
4905 Perl_re_printf( aTHX_ " -> %d:%s",
4906 REG_NODE_NUM(noper), SvPV_nolen_const(RExC_mysv));
4909 regprop(RExC_rx, RExC_mysv, noper_next, NULL, pRExC_state);
4910 Perl_re_printf( aTHX_ "\t=> %d:%s\t",
4911 REG_NODE_NUM(noper_next), SvPV_nolen_const(RExC_mysv));
4913 Perl_re_printf( aTHX_ "(First==%d,Last==%d,Cur==%d,tt==%s,ntt==%s,nntt==%s)\n",
4914 REG_NODE_NUM(first), REG_NODE_NUM(last), REG_NODE_NUM(cur),
4915 PL_reg_name[trietype], PL_reg_name[noper_trietype], PL_reg_name[noper_next_trietype]
4919 /* Is noper a trieable nodetype that can be merged
4920 * with the current trie (if there is one)? */
4924 ( noper_trietype == NOTHING )
4925 || ( trietype == NOTHING )
4926 || ( trietype == noper_trietype )
4929 && noper_next >= tail
4933 /* Handle mergable triable node Either we are
4934 * the first node in a new trieable sequence,
4935 * in which case we do some bookkeeping,
4936 * otherwise we update the end pointer. */
4939 if ( noper_trietype == NOTHING ) {
4940 #if !defined(DEBUGGING) && !defined(NOJUMPTRIE)
4941 regnode * const noper_next = regnext( noper );
4942 U8 noper_next_type = (noper_next && noper_next < tail) ? OP(noper_next) : 0;
4943 U8 noper_next_trietype = noper_next_type ? TRIE_TYPE( noper_next_type ) :0;
4946 if ( noper_next_trietype ) {
4947 trietype = noper_next_trietype;
4948 } else if (noper_next_type) {
4949 /* a NOTHING regop is 1 regop wide.
4950 * We need at least two for a trie
4951 * so we can't merge this in */
4955 trietype = noper_trietype;
4958 if ( trietype == NOTHING )
4959 trietype = noper_trietype;
4964 } /* end handle mergable triable node */
4966 /* handle unmergable node -
4967 * noper may either be a triable node which can
4968 * not be tried together with the current trie,
4969 * or a non triable node */
4971 /* If last is set and trietype is not
4972 * NOTHING then we have found at least two
4973 * triable branch sequences in a row of a
4974 * similar trietype so we can turn them
4975 * into a trie. If/when we allow NOTHING to
4976 * start a trie sequence this condition
4977 * will be required, and it isn't expensive
4978 * so we leave it in for now. */
4979 if ( trietype && trietype != NOTHING )
4980 make_trie( pRExC_state,
4981 startbranch, first, cur, tail,
4982 count, trietype, depth+1 );
4983 last = NULL; /* note: we clear/update
4984 first, trietype etc below,
4985 so we dont do it here */
4989 && noper_next >= tail
4992 /* noper is triable, so we can start a new
4996 trietype = noper_trietype;
4998 /* if we already saw a first but the
4999 * current node is not triable then we have
5000 * to reset the first information. */
5005 } /* end handle unmergable node */
5006 } /* loop over branches */
5007 DEBUG_TRIE_COMPILE_r({
5008 regprop(RExC_rx, RExC_mysv, cur, NULL, pRExC_state);
5009 Perl_re_indentf( aTHX_ "- %s (%d) <SCAN FINISHED> ",
5010 depth+1, SvPV_nolen_const( RExC_mysv ), REG_NODE_NUM(cur));
5011 Perl_re_printf( aTHX_ "(First==%d, Last==%d, Cur==%d, tt==%s)\n",
5012 REG_NODE_NUM(first), REG_NODE_NUM(last), REG_NODE_NUM(cur),
5013 PL_reg_name[trietype]
5017 if ( last && trietype ) {
5018 if ( trietype != NOTHING ) {
5019 /* the last branch of the sequence was part of
5020 * a trie, so we have to construct it here
5021 * outside of the loop */
5022 made= make_trie( pRExC_state, startbranch,
5023 first, scan, tail, count,
5024 trietype, depth+1 );
5025 #ifdef TRIE_STUDY_OPT
5026 if ( ((made == MADE_EXACT_TRIE &&
5027 startbranch == first)
5028 || ( first_non_open == first )) &&
5030 flags |= SCF_TRIE_RESTUDY;
5031 if ( startbranch == first
5034 RExC_seen &=~REG_TOP_LEVEL_BRANCHES_SEEN;
5039 /* at this point we know whatever we have is a
5040 * NOTHING sequence/branch AND if 'startbranch'
5041 * is 'first' then we can turn the whole thing
5044 if ( startbranch == first ) {
5046 /* the entire thing is a NOTHING sequence,
5047 * something like this: (?:|) So we can
5048 * turn it into a plain NOTHING op. */
5049 DEBUG_TRIE_COMPILE_r({
5050 regprop(RExC_rx, RExC_mysv, cur, NULL, pRExC_state);
5051 Perl_re_indentf( aTHX_ "- %s (%d) <NOTHING BRANCH SEQUENCE>\n",
5053 SvPV_nolen_const( RExC_mysv ), REG_NODE_NUM(cur));
5056 OP(startbranch)= NOTHING;
5057 NEXT_OFF(startbranch)= tail - startbranch;
5058 for ( opt= startbranch + 1; opt < tail ; opt++ )
5062 } /* end if ( last) */
5063 } /* TRIE_MAXBUF is non zero */
5068 else if ( code == BRANCHJ ) { /* single branch is optimized. */
5069 scan = NEXTOPER(NEXTOPER(scan));
5070 } else /* single branch is optimized. */
5071 scan = NEXTOPER(scan);
5073 } else if (OP(scan) == SUSPEND || OP(scan) == GOSUB) {
5075 regnode *start = NULL;
5076 regnode *end = NULL;
5077 U32 my_recursed_depth= recursed_depth;
5079 if (OP(scan) != SUSPEND) { /* GOSUB */
5080 /* Do setup, note this code has side effects beyond
5081 * the rest of this block. Specifically setting
5082 * RExC_recurse[] must happen at least once during
5085 RExC_recurse[ARG2L(scan)] = scan;
5086 start = REGNODE_p(RExC_open_parens[paren]);
5087 end = REGNODE_p(RExC_close_parens[paren]);
5089 /* NOTE we MUST always execute the above code, even
5090 * if we do nothing with a GOSUB */
5092 ( flags & SCF_IN_DEFINE )
5095 (is_inf_internal || is_inf || (data && data->flags & SF_IS_INF))
5097 ( (flags & (SCF_DO_STCLASS | SCF_DO_SUBSTR)) == 0 )
5100 /* no need to do anything here if we are in a define. */
5101 /* or we are after some kind of infinite construct
5102 * so we can skip recursing into this item.
5103 * Since it is infinite we will not change the maxlen
5104 * or delta, and if we miss something that might raise
5105 * the minlen it will merely pessimise a little.
5107 * Iow /(?(DEFINE)(?<foo>foo|food))a+(?&foo)/
5108 * might result in a minlen of 1 and not of 4,
5109 * but this doesn't make us mismatch, just try a bit
5110 * harder than we should.
5112 scan= regnext(scan);
5119 !PAREN_TEST(RExC_study_chunk_recursed + ((recursed_depth-1) * RExC_study_chunk_recursed_bytes), paren)
5121 /* it is quite possible that there are more efficient ways
5122 * to do this. We maintain a bitmap per level of recursion
5123 * of which patterns we have entered so we can detect if a
5124 * pattern creates a possible infinite loop. When we
5125 * recurse down a level we copy the previous levels bitmap
5126 * down. When we are at recursion level 0 we zero the top
5127 * level bitmap. It would be nice to implement a different
5128 * more efficient way of doing this. In particular the top
5129 * level bitmap may be unnecessary.
5131 if (!recursed_depth) {
5132 Zero(RExC_study_chunk_recursed, RExC_study_chunk_recursed_bytes, U8);
5134 Copy(RExC_study_chunk_recursed + ((recursed_depth-1) * RExC_study_chunk_recursed_bytes),
5135 RExC_study_chunk_recursed + (recursed_depth * RExC_study_chunk_recursed_bytes),
5136 RExC_study_chunk_recursed_bytes, U8);
5138 /* we havent recursed into this paren yet, so recurse into it */
5139 DEBUG_STUDYDATA("gosub-set", data, depth, is_inf);
5140 PAREN_SET(RExC_study_chunk_recursed + (recursed_depth * RExC_study_chunk_recursed_bytes), paren);
5141 my_recursed_depth= recursed_depth + 1;
5143 DEBUG_STUDYDATA("gosub-inf", data, depth, is_inf);
5144 /* some form of infinite recursion, assume infinite length
5146 if (flags & SCF_DO_SUBSTR) {
5147 scan_commit(pRExC_state, data, minlenp, is_inf);
5148 data->cur_is_floating = 1;
5150 is_inf = is_inf_internal = 1;
5151 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
5152 ssc_anything(data->start_class);
5153 flags &= ~SCF_DO_STCLASS;
5155 start= NULL; /* reset start so we dont recurse later on. */
5160 end = regnext(scan);
5163 scan_frame *newframe;
5165 if (!RExC_frame_last) {
5166 Newxz(newframe, 1, scan_frame);
5167 SAVEDESTRUCTOR_X(S_unwind_scan_frames, newframe);
5168 RExC_frame_head= newframe;
5170 } else if (!RExC_frame_last->next_frame) {
5171 Newxz(newframe, 1, scan_frame);
5172 RExC_frame_last->next_frame= newframe;
5173 newframe->prev_frame= RExC_frame_last;
5176 newframe= RExC_frame_last->next_frame;
5178 RExC_frame_last= newframe;
5180 newframe->next_regnode = regnext(scan);
5181 newframe->last_regnode = last;
5182 newframe->stopparen = stopparen;
5183 newframe->prev_recursed_depth = recursed_depth;
5184 newframe->this_prev_frame= frame;
5186 DEBUG_STUDYDATA("frame-new", data, depth, is_inf);
5187 DEBUG_PEEP("fnew", scan, depth, flags);
5194 recursed_depth= my_recursed_depth;
5199 else if ( OP(scan) == EXACT
5200 || OP(scan) == EXACT_ONLY8
5201 || OP(scan) == EXACTL)
5203 SSize_t l = STR_LEN(scan);
5207 const U8 * const s = (U8*)STRING(scan);
5208 uc = utf8_to_uvchr_buf(s, s + l, NULL);
5209 l = utf8_length(s, s + l);
5211 uc = *((U8*)STRING(scan));
5214 if (flags & SCF_DO_SUBSTR) { /* Update longest substr. */
5215 /* The code below prefers earlier match for fixed
5216 offset, later match for variable offset. */
5217 if (data->last_end == -1) { /* Update the start info. */
5218 data->last_start_min = data->pos_min;
5219 data->last_start_max = is_inf
5220 ? SSize_t_MAX : data->pos_min + data->pos_delta;
5222 sv_catpvn(data->last_found, STRING(scan), STR_LEN(scan));
5224 SvUTF8_on(data->last_found);
5226 SV * const sv = data->last_found;
5227 MAGIC * const mg = SvUTF8(sv) && SvMAGICAL(sv) ?
5228 mg_find(sv, PERL_MAGIC_utf8) : NULL;
5229 if (mg && mg->mg_len >= 0)
5230 mg->mg_len += utf8_length((U8*)STRING(scan),
5231 (U8*)STRING(scan)+STR_LEN(scan));
5233 data->last_end = data->pos_min + l;
5234 data->pos_min += l; /* As in the first entry. */
5235 data->flags &= ~SF_BEFORE_EOL;
5238 /* ANDing the code point leaves at most it, and not in locale, and
5239 * can't match null string */
5240 if (flags & SCF_DO_STCLASS_AND) {
5241 ssc_cp_and(data->start_class, uc);
5242 ANYOF_FLAGS(data->start_class) &= ~SSC_MATCHES_EMPTY_STRING;
5243 ssc_clear_locale(data->start_class);
5245 else if (flags & SCF_DO_STCLASS_OR) {
5246 ssc_add_cp(data->start_class, uc);
5247 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
5249 /* See commit msg 749e076fceedeb708a624933726e7989f2302f6a */
5250 ANYOF_FLAGS(data->start_class) &= ~SSC_MATCHES_EMPTY_STRING;
5252 flags &= ~SCF_DO_STCLASS;
5254 else if (PL_regkind[OP(scan)] == EXACT) {
5255 /* But OP != EXACT!, so is EXACTFish */
5256 SSize_t l = STR_LEN(scan);
5257 const U8 * s = (U8*)STRING(scan);
5259 /* Search for fixed substrings supports EXACT only. */
5260 if (flags & SCF_DO_SUBSTR) {
5262 scan_commit(pRExC_state, data, minlenp, is_inf);
5265 l = utf8_length(s, s + l);
5267 if (unfolded_multi_char) {
5268 RExC_seen |= REG_UNFOLDED_MULTI_SEEN;
5270 min += l - min_subtract;
5272 delta += min_subtract;
5273 if (flags & SCF_DO_SUBSTR) {
5274 data->pos_min += l - min_subtract;
5275 if (data->pos_min < 0) {
5278 data->pos_delta += min_subtract;
5280 data->cur_is_floating = 1; /* float */
5284 if (flags & SCF_DO_STCLASS) {
5285 SV* EXACTF_invlist = _make_exactf_invlist(pRExC_state, scan);
5287 assert(EXACTF_invlist);
5288 if (flags & SCF_DO_STCLASS_AND) {
5289 if (OP(scan) != EXACTFL)
5290 ssc_clear_locale(data->start_class);
5291 ANYOF_FLAGS(data->start_class) &= ~SSC_MATCHES_EMPTY_STRING;
5292 ANYOF_POSIXL_ZERO(data->start_class);
5293 ssc_intersection(data->start_class, EXACTF_invlist, FALSE);
5295 else { /* SCF_DO_STCLASS_OR */
5296 ssc_union(data->start_class, EXACTF_invlist, FALSE);
5297 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
5299 /* See commit msg 749e076fceedeb708a624933726e7989f2302f6a */
5300 ANYOF_FLAGS(data->start_class) &= ~SSC_MATCHES_EMPTY_STRING;
5302 flags &= ~SCF_DO_STCLASS;
5303 SvREFCNT_dec(EXACTF_invlist);
5306 else if (REGNODE_VARIES(OP(scan))) {
5307 SSize_t mincount, maxcount, minnext, deltanext, pos_before = 0;
5308 I32 fl = 0, f = flags;
5309 regnode * const oscan = scan;
5310 regnode_ssc this_class;
5311 regnode_ssc *oclass = NULL;
5312 I32 next_is_eval = 0;
5314 switch (PL_regkind[OP(scan)]) {
5315 case WHILEM: /* End of (?:...)* . */
5316 scan = NEXTOPER(scan);
5319 if (flags & (SCF_DO_SUBSTR | SCF_DO_STCLASS)) {
5320 next = NEXTOPER(scan);
5321 if ( OP(next) == EXACT
5322 || OP(next) == EXACT_ONLY8
5323 || OP(next) == EXACTL
5324 || (flags & SCF_DO_STCLASS))
5327 maxcount = REG_INFTY;
5328 next = regnext(scan);
5329 scan = NEXTOPER(scan);
5333 if (flags & SCF_DO_SUBSTR)
5338 next = NEXTOPER(scan);
5340 /* This temporary node can now be turned into EXACTFU, and
5341 * must, as regexec.c doesn't handle it */
5342 if (OP(next) == EXACTFU_S_EDGE) {
5346 if ( STR_LEN(next) == 1
5347 && isALPHA_A(* STRING(next))
5348 && ( OP(next) == EXACTFAA
5349 || ( OP(next) == EXACTFU
5350 && ! HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE(* STRING(next)))))
5352 /* These differ in just one bit */
5353 U8 mask = ~ ('A' ^ 'a');
5355 assert(isALPHA_A(* STRING(next)));
5357 /* Then replace it by an ANYOFM node, with
5358 * the mask set to the complement of the
5359 * bit that differs between upper and lower
5360 * case, and the lowest code point of the
5361 * pair (which the '&' forces) */
5363 ARG_SET(next, *STRING(next) & mask);
5367 if (flags & SCF_DO_STCLASS) {
5369 maxcount = REG_INFTY;
5370 next = regnext(scan);
5371 scan = NEXTOPER(scan);
5374 if (flags & SCF_DO_SUBSTR) {
5375 scan_commit(pRExC_state, data, minlenp, is_inf);
5376 /* Cannot extend fixed substrings */
5377 data->cur_is_floating = 1; /* float */
5379 is_inf = is_inf_internal = 1;
5380 scan = regnext(scan);
5381 goto optimize_curly_tail;
5383 if (stopparen>0 && (OP(scan)==CURLYN || OP(scan)==CURLYM)
5384 && (scan->flags == stopparen))
5389 mincount = ARG1(scan);
5390 maxcount = ARG2(scan);
5392 next = regnext(scan);
5393 if (OP(scan) == CURLYX) {
5394 I32 lp = (data ? *(data->last_closep) : 0);
5395 scan->flags = ((lp <= (I32)U8_MAX) ? (U8)lp : U8_MAX);
5397 scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
5398 next_is_eval = (OP(scan) == EVAL);
5400 if (flags & SCF_DO_SUBSTR) {
5402 scan_commit(pRExC_state, data, minlenp, is_inf);
5403 /* Cannot extend fixed substrings */
5404 pos_before = data->pos_min;
5408 data->flags &= ~(SF_HAS_PAR|SF_IN_PAR|SF_HAS_EVAL);
5410 data->flags |= SF_IS_INF;
5412 if (flags & SCF_DO_STCLASS) {
5413 ssc_init(pRExC_state, &this_class);
5414 oclass = data->start_class;
5415 data->start_class = &this_class;
5416 f |= SCF_DO_STCLASS_AND;
5417 f &= ~SCF_DO_STCLASS_OR;
5419 /* Exclude from super-linear cache processing any {n,m}
5420 regops for which the combination of input pos and regex
5421 pos is not enough information to determine if a match
5424 For example, in the regex /foo(bar\s*){4,8}baz/ with the
5425 regex pos at the \s*, the prospects for a match depend not
5426 only on the input position but also on how many (bar\s*)
5427 repeats into the {4,8} we are. */
5428 if ((mincount > 1) || (maxcount > 1 && maxcount != REG_INFTY))
5429 f &= ~SCF_WHILEM_VISITED_POS;
5431 /* This will finish on WHILEM, setting scan, or on NULL: */
5432 /* recurse study_chunk() on loop bodies */
5433 minnext = study_chunk(pRExC_state, &scan, minlenp, &deltanext,
5434 last, data, stopparen, recursed_depth, NULL,
5436 ? (f & ~SCF_DO_SUBSTR)
5440 if (flags & SCF_DO_STCLASS)
5441 data->start_class = oclass;
5442 if (mincount == 0 || minnext == 0) {
5443 if (flags & SCF_DO_STCLASS_OR) {
5444 ssc_or(pRExC_state, data->start_class, (regnode_charclass *) &this_class);
5446 else if (flags & SCF_DO_STCLASS_AND) {
5447 /* Switch to OR mode: cache the old value of
5448 * data->start_class */
5450 StructCopy(data->start_class, and_withp, regnode_ssc);
5451 flags &= ~SCF_DO_STCLASS_AND;
5452 StructCopy(&this_class, data->start_class, regnode_ssc);
5453 flags |= SCF_DO_STCLASS_OR;
5454 ANYOF_FLAGS(data->start_class)
5455 |= SSC_MATCHES_EMPTY_STRING;
5457 } else { /* Non-zero len */
5458 if (flags & SCF_DO_STCLASS_OR) {
5459 ssc_or(pRExC_state, data->start_class, (regnode_charclass *) &this_class);
5460 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
5462 else if (flags & SCF_DO_STCLASS_AND)
5463 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) &this_class);
5464 flags &= ~SCF_DO_STCLASS;
5466 if (!scan) /* It was not CURLYX, but CURLY. */
5468 if (((flags & (SCF_TRIE_DOING_RESTUDY|SCF_DO_SUBSTR))==SCF_DO_SUBSTR)
5469 /* ? quantifier ok, except for (?{ ... }) */
5470 && (next_is_eval || !(mincount == 0 && maxcount == 1))
5471 && (minnext == 0) && (deltanext == 0)
5472 && data && !(data->flags & (SF_HAS_PAR|SF_IN_PAR))
5473 && maxcount <= REG_INFTY/3) /* Complement check for big
5476 _WARN_HELPER(RExC_precomp_end, packWARN(WARN_REGEXP),
5477 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
5478 "Quantifier unexpected on zero-length expression "
5479 "in regex m/%" UTF8f "/",
5480 UTF8fARG(UTF, RExC_precomp_end - RExC_precomp,
5484 min += minnext * mincount;
5485 is_inf_internal |= deltanext == SSize_t_MAX
5486 || (maxcount == REG_INFTY && minnext + deltanext > 0);
5487 is_inf |= is_inf_internal;
5489 delta = SSize_t_MAX;
5491 delta += (minnext + deltanext) * maxcount
5492 - minnext * mincount;
5494 /* Try powerful optimization CURLYX => CURLYN. */
5495 if ( OP(oscan) == CURLYX && data
5496 && data->flags & SF_IN_PAR
5497 && !(data->flags & SF_HAS_EVAL)
5498 && !deltanext && minnext == 1 ) {
5499 /* Try to optimize to CURLYN. */
5500 regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS;
5501 regnode * const nxt1 = nxt;
5508 if (!REGNODE_SIMPLE(OP(nxt))
5509 && !(PL_regkind[OP(nxt)] == EXACT
5510 && STR_LEN(nxt) == 1))
5516 if (OP(nxt) != CLOSE)
5518 if (RExC_open_parens) {
5521 RExC_open_parens[ARG(nxt1)] = REGNODE_OFFSET(oscan);
5524 RExC_close_parens[ARG(nxt1)] = REGNODE_OFFSET(nxt) + 2;
5526 /* Now we know that nxt2 is the only contents: */
5527 oscan->flags = (U8)ARG(nxt);
5529 OP(nxt1) = NOTHING; /* was OPEN. */
5532 OP(nxt1 + 1) = OPTIMIZED; /* was count. */
5533 NEXT_OFF(nxt1+ 1) = 0; /* just for consistency. */
5534 NEXT_OFF(nxt2) = 0; /* just for consistency with CURLY. */
5535 OP(nxt) = OPTIMIZED; /* was CLOSE. */
5536 OP(nxt + 1) = OPTIMIZED; /* was count. */
5537 NEXT_OFF(nxt+ 1) = 0; /* just for consistency. */
5542 /* Try optimization CURLYX => CURLYM. */
5543 if ( OP(oscan) == CURLYX && data
5544 && !(data->flags & SF_HAS_PAR)
5545 && !(data->flags & SF_HAS_EVAL)
5546 && !deltanext /* atom is fixed width */
5547 && minnext != 0 /* CURLYM can't handle zero width */
5549 /* Nor characters whose fold at run-time may be
5550 * multi-character */
5551 && ! (RExC_seen & REG_UNFOLDED_MULTI_SEEN)
5553 /* XXXX How to optimize if data == 0? */
5554 /* Optimize to a simpler form. */
5555 regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN */
5559 while ( (nxt2 = regnext(nxt)) /* skip over embedded stuff*/
5560 && (OP(nxt2) != WHILEM))
5562 OP(nxt2) = SUCCEED; /* Whas WHILEM */
5563 /* Need to optimize away parenths. */
5564 if ((data->flags & SF_IN_PAR) && OP(nxt) == CLOSE) {
5565 /* Set the parenth number. */
5566 regnode *nxt1 = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN*/
5568 oscan->flags = (U8)ARG(nxt);
5569 if (RExC_open_parens) {
5571 RExC_open_parens[ARG(nxt1)] = REGNODE_OFFSET(oscan);
5574 RExC_close_parens[ARG(nxt1)] = REGNODE_OFFSET(nxt2)
5577 OP(nxt1) = OPTIMIZED; /* was OPEN. */
5578 OP(nxt) = OPTIMIZED; /* was CLOSE. */
5581 OP(nxt1 + 1) = OPTIMIZED; /* was count. */
5582 OP(nxt + 1) = OPTIMIZED; /* was count. */
5583 NEXT_OFF(nxt1 + 1) = 0; /* just for consistency. */
5584 NEXT_OFF(nxt + 1) = 0; /* just for consistency. */
5587 while ( nxt1 && (OP(nxt1) != WHILEM)) {
5588 regnode *nnxt = regnext(nxt1);
5590 if (reg_off_by_arg[OP(nxt1)])
5591 ARG_SET(nxt1, nxt2 - nxt1);
5592 else if (nxt2 - nxt1 < U16_MAX)
5593 NEXT_OFF(nxt1) = nxt2 - nxt1;
5595 OP(nxt) = NOTHING; /* Cannot beautify */
5600 /* Optimize again: */
5601 /* recurse study_chunk() on optimised CURLYX => CURLYM */
5602 study_chunk(pRExC_state, &nxt1, minlenp, &deltanext, nxt,
5603 NULL, stopparen, recursed_depth, NULL, 0,
5609 else if ((OP(oscan) == CURLYX)
5610 && (flags & SCF_WHILEM_VISITED_POS)
5611 /* See the comment on a similar expression above.
5612 However, this time it's not a subexpression
5613 we care about, but the expression itself. */
5614 && (maxcount == REG_INFTY)
5616 /* This stays as CURLYX, we can put the count/of pair. */
5617 /* Find WHILEM (as in regexec.c) */
5618 regnode *nxt = oscan + NEXT_OFF(oscan);
5620 if (OP(PREVOPER(nxt)) == NOTHING) /* LONGJMP */
5622 nxt = PREVOPER(nxt);
5623 if (nxt->flags & 0xf) {
5624 /* we've already set whilem count on this node */
5625 } else if (++data->whilem_c < 16) {
5626 assert(data->whilem_c <= RExC_whilem_seen);
5627 nxt->flags = (U8)(data->whilem_c
5628 | (RExC_whilem_seen << 4)); /* On WHILEM */
5631 if (data && fl & (SF_HAS_PAR|SF_IN_PAR))
5633 if (flags & SCF_DO_SUBSTR) {
5634 SV *last_str = NULL;
5635 STRLEN last_chrs = 0;
5636 int counted = mincount != 0;
5638 if (data->last_end > 0 && mincount != 0) { /* Ends with a
5640 SSize_t b = pos_before >= data->last_start_min
5641 ? pos_before : data->last_start_min;
5643 const char * const s = SvPV_const(data->last_found, l);
5644 SSize_t old = b - data->last_start_min;
5648 old = utf8_hop_forward((U8*)s, old,
5649 (U8 *) SvEND(data->last_found))
5652 /* Get the added string: */
5653 last_str = newSVpvn_utf8(s + old, l, UTF);
5654 last_chrs = UTF ? utf8_length((U8*)(s + old),
5655 (U8*)(s + old + l)) : l;
5656 if (deltanext == 0 && pos_before == b) {
5657 /* What was added is a constant string */
5660 SvGROW(last_str, (mincount * l) + 1);
5661 repeatcpy(SvPVX(last_str) + l,
5662 SvPVX_const(last_str), l,
5664 SvCUR_set(last_str, SvCUR(last_str) * mincount);
5665 /* Add additional parts. */
5666 SvCUR_set(data->last_found,
5667 SvCUR(data->last_found) - l);
5668 sv_catsv(data->last_found, last_str);
5670 SV * sv = data->last_found;
5672 SvUTF8(sv) && SvMAGICAL(sv) ?
5673 mg_find(sv, PERL_MAGIC_utf8) : NULL;
5674 if (mg && mg->mg_len >= 0)
5675 mg->mg_len += last_chrs * (mincount-1);
5677 last_chrs *= mincount;
5678 data->last_end += l * (mincount - 1);
5681 /* start offset must point into the last copy */
5682 data->last_start_min += minnext * (mincount - 1);
5683 data->last_start_max =
5686 : data->last_start_max +
5687 (maxcount - 1) * (minnext + data->pos_delta);
5690 /* It is counted once already... */
5691 data->pos_min += minnext * (mincount - counted);
5693 Perl_re_printf( aTHX_ "counted=%" UVuf " deltanext=%" UVuf
5694 " SSize_t_MAX=%" UVuf " minnext=%" UVuf
5695 " maxcount=%" UVuf " mincount=%" UVuf "\n",
5696 (UV)counted, (UV)deltanext, (UV)SSize_t_MAX, (UV)minnext, (UV)maxcount,
5698 if (deltanext != SSize_t_MAX)
5699 Perl_re_printf( aTHX_ "LHS=%" UVuf " RHS=%" UVuf "\n",
5700 (UV)(-counted * deltanext + (minnext + deltanext) * maxcount
5701 - minnext * mincount), (UV)(SSize_t_MAX - data->pos_delta));
5703 if (deltanext == SSize_t_MAX
5704 || -counted * deltanext + (minnext + deltanext) * maxcount - minnext * mincount >= SSize_t_MAX - data->pos_delta)
5705 data->pos_delta = SSize_t_MAX;
5707 data->pos_delta += - counted * deltanext +
5708 (minnext + deltanext) * maxcount - minnext * mincount;
5709 if (mincount != maxcount) {
5710 /* Cannot extend fixed substrings found inside
5712 scan_commit(pRExC_state, data, minlenp, is_inf);
5713 if (mincount && last_str) {
5714 SV * const sv = data->last_found;
5715 MAGIC * const mg = SvUTF8(sv) && SvMAGICAL(sv) ?
5716 mg_find(sv, PERL_MAGIC_utf8) : NULL;
5720 sv_setsv(sv, last_str);
5721 data->last_end = data->pos_min;
5722 data->last_start_min = data->pos_min - last_chrs;
5723 data->last_start_max = is_inf
5725 : data->pos_min + data->pos_delta - last_chrs;
5727 data->cur_is_floating = 1; /* float */
5729 SvREFCNT_dec(last_str);
5731 if (data && (fl & SF_HAS_EVAL))
5732 data->flags |= SF_HAS_EVAL;
5733 optimize_curly_tail:
5734 if (OP(oscan) != CURLYX) {
5735 while (PL_regkind[OP(next = regnext(oscan))] == NOTHING
5737 NEXT_OFF(oscan) += NEXT_OFF(next);
5743 Perl_croak(aTHX_ "panic: unexpected varying REx opcode %d",
5748 if (flags & SCF_DO_SUBSTR) {
5749 /* Cannot expect anything... */
5750 scan_commit(pRExC_state, data, minlenp, is_inf);
5751 data->cur_is_floating = 1; /* float */
5753 is_inf = is_inf_internal = 1;
5754 if (flags & SCF_DO_STCLASS_OR) {
5755 if (OP(scan) == CLUMP) {
5756 /* Actually is any start char, but very few code points
5757 * aren't start characters */
5758 ssc_match_all_cp(data->start_class);
5761 ssc_anything(data->start_class);
5764 flags &= ~SCF_DO_STCLASS;
5768 else if (OP(scan) == LNBREAK) {
5769 if (flags & SCF_DO_STCLASS) {
5770 if (flags & SCF_DO_STCLASS_AND) {
5771 ssc_intersection(data->start_class,
5772 PL_XPosix_ptrs[_CC_VERTSPACE], FALSE);
5773 ssc_clear_locale(data->start_class);
5774 ANYOF_FLAGS(data->start_class)
5775 &= ~SSC_MATCHES_EMPTY_STRING;
5777 else if (flags & SCF_DO_STCLASS_OR) {
5778 ssc_union(data->start_class,
5779 PL_XPosix_ptrs[_CC_VERTSPACE],
5781 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
5783 /* See commit msg for
5784 * 749e076fceedeb708a624933726e7989f2302f6a */
5785 ANYOF_FLAGS(data->start_class)
5786 &= ~SSC_MATCHES_EMPTY_STRING;
5788 flags &= ~SCF_DO_STCLASS;
5791 if (delta != SSize_t_MAX)
5792 delta++; /* Because of the 2 char string cr-lf */
5793 if (flags & SCF_DO_SUBSTR) {
5794 /* Cannot expect anything... */
5795 scan_commit(pRExC_state, data, minlenp, is_inf);
5797 if (data->pos_delta != SSize_t_MAX) {
5798 data->pos_delta += 1;
5800 data->cur_is_floating = 1; /* float */
5803 else if (REGNODE_SIMPLE(OP(scan))) {
5805 if (flags & SCF_DO_SUBSTR) {
5806 scan_commit(pRExC_state, data, minlenp, is_inf);
5810 if (flags & SCF_DO_STCLASS) {
5812 SV* my_invlist = NULL;
5815 /* See commit msg 749e076fceedeb708a624933726e7989f2302f6a */
5816 ANYOF_FLAGS(data->start_class) &= ~SSC_MATCHES_EMPTY_STRING;
5818 /* Some of the logic below assumes that switching
5819 locale on will only add false positives. */
5824 Perl_croak(aTHX_ "panic: unexpected simple REx opcode %d",
5828 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
5829 ssc_match_all_cp(data->start_class);
5834 SV* REG_ANY_invlist = _new_invlist(2);
5835 REG_ANY_invlist = add_cp_to_invlist(REG_ANY_invlist,
5837 if (flags & SCF_DO_STCLASS_OR) {
5838 ssc_union(data->start_class,
5840 TRUE /* TRUE => invert, hence all but \n
5844 else if (flags & SCF_DO_STCLASS_AND) {
5845 ssc_intersection(data->start_class,
5847 TRUE /* TRUE => invert */
5849 ssc_clear_locale(data->start_class);
5851 SvREFCNT_dec_NN(REG_ANY_invlist);
5862 if (flags & SCF_DO_STCLASS_AND)
5863 ssc_and(pRExC_state, data->start_class,
5864 (regnode_charclass *) scan);
5866 ssc_or(pRExC_state, data->start_class,
5867 (regnode_charclass *) scan);
5873 SV* cp_list = get_ANYOFM_contents(scan);
5875 if (flags & SCF_DO_STCLASS_OR) {
5876 ssc_union(data->start_class, cp_list, invert);
5878 else if (flags & SCF_DO_STCLASS_AND) {
5879 ssc_intersection(data->start_class, cp_list, invert);
5882 SvREFCNT_dec_NN(cp_list);
5891 namedclass = classnum_to_namedclass(FLAGS(scan)) + invert;
5892 if (flags & SCF_DO_STCLASS_AND) {
5893 bool was_there = cBOOL(
5894 ANYOF_POSIXL_TEST(data->start_class,
5896 ANYOF_POSIXL_ZERO(data->start_class);
5897 if (was_there) { /* Do an AND */
5898 ANYOF_POSIXL_SET(data->start_class, namedclass);
5900 /* No individual code points can now match */
5901 data->start_class->invlist
5902 = sv_2mortal(_new_invlist(0));
5905 int complement = namedclass + ((invert) ? -1 : 1);
5907 assert(flags & SCF_DO_STCLASS_OR);
5909 /* If the complement of this class was already there,
5910 * the result is that they match all code points,
5911 * (\d + \D == everything). Remove the classes from
5912 * future consideration. Locale is not relevant in
5914 if (ANYOF_POSIXL_TEST(data->start_class, complement)) {
5915 ssc_match_all_cp(data->start_class);
5916 ANYOF_POSIXL_CLEAR(data->start_class, namedclass);
5917 ANYOF_POSIXL_CLEAR(data->start_class, complement);
5919 else { /* The usual case; just add this class to the
5921 ANYOF_POSIXL_SET(data->start_class, namedclass);
5926 case NPOSIXA: /* For these, we always know the exact set of
5931 my_invlist = invlist_clone(PL_Posix_ptrs[FLAGS(scan)], NULL);
5932 goto join_posix_and_ascii;
5940 my_invlist = invlist_clone(PL_XPosix_ptrs[FLAGS(scan)], NULL);
5942 /* NPOSIXD matches all upper Latin1 code points unless the
5943 * target string being matched is UTF-8, which is
5944 * unknowable until match time. Since we are going to
5945 * invert, we want to get rid of all of them so that the
5946 * inversion will match all */
5947 if (OP(scan) == NPOSIXD) {
5948 _invlist_subtract(my_invlist, PL_UpperLatin1,
5952 join_posix_and_ascii:
5954 if (flags & SCF_DO_STCLASS_AND) {
5955 ssc_intersection(data->start_class, my_invlist, invert);
5956 ssc_clear_locale(data->start_class);
5959 assert(flags & SCF_DO_STCLASS_OR);
5960 ssc_union(data->start_class, my_invlist, invert);
5962 SvREFCNT_dec(my_invlist);
5964 if (flags & SCF_DO_STCLASS_OR)
5965 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
5966 flags &= ~SCF_DO_STCLASS;
5969 else if (PL_regkind[OP(scan)] == EOL && flags & SCF_DO_SUBSTR) {
5970 data->flags |= (OP(scan) == MEOL
5973 scan_commit(pRExC_state, data, minlenp, is_inf);
5976 else if ( PL_regkind[OP(scan)] == BRANCHJ
5977 /* Lookbehind, or need to calculate parens/evals/stclass: */
5978 && (scan->flags || data || (flags & SCF_DO_STCLASS))
5979 && (OP(scan) == IFMATCH || OP(scan) == UNLESSM))
5981 if ( !PERL_ENABLE_POSITIVE_ASSERTION_STUDY
5982 || OP(scan) == UNLESSM )
5984 /* Negative Lookahead/lookbehind
5985 In this case we can't do fixed string optimisation.
5988 SSize_t deltanext, minnext, fake = 0;
5993 StructCopy(&zero_scan_data, &data_fake, scan_data_t);
5995 data_fake.whilem_c = data->whilem_c;
5996 data_fake.last_closep = data->last_closep;
5999 data_fake.last_closep = &fake;
6000 data_fake.pos_delta = delta;
6001 if ( flags & SCF_DO_STCLASS && !scan->flags
6002 && OP(scan) == IFMATCH ) { /* Lookahead */
6003 ssc_init(pRExC_state, &intrnl);
6004 data_fake.start_class = &intrnl;
6005 f |= SCF_DO_STCLASS_AND;
6007 if (flags & SCF_WHILEM_VISITED_POS)
6008 f |= SCF_WHILEM_VISITED_POS;
6009 next = regnext(scan);
6010 nscan = NEXTOPER(NEXTOPER(scan));
6012 /* recurse study_chunk() for lookahead body */
6013 minnext = study_chunk(pRExC_state, &nscan, minlenp, &deltanext,
6014 last, &data_fake, stopparen,
6015 recursed_depth, NULL, f, depth+1);
6018 || deltanext > (I32) U8_MAX
6019 || minnext > (I32)U8_MAX
6020 || minnext + deltanext > (I32)U8_MAX)
6022 FAIL2("Lookbehind longer than %" UVuf " not implemented",
6026 /* The 'next_off' field has been repurposed to count the
6027 * additional starting positions to try beyond the initial
6028 * one. (This leaves it at 0 for non-variable length
6029 * matches to avoid breakage for those not using this
6032 scan->next_off = deltanext;
6033 ckWARNexperimental(RExC_parse,
6034 WARN_EXPERIMENTAL__VLB,
6035 "Variable length lookbehind is experimental");
6037 scan->flags = (U8)minnext + deltanext;
6040 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
6042 if (data_fake.flags & SF_HAS_EVAL)
6043 data->flags |= SF_HAS_EVAL;
6044 data->whilem_c = data_fake.whilem_c;
6046 if (f & SCF_DO_STCLASS_AND) {
6047 if (flags & SCF_DO_STCLASS_OR) {
6048 /* OR before, AND after: ideally we would recurse with
6049 * data_fake to get the AND applied by study of the
6050 * remainder of the pattern, and then derecurse;
6051 * *** HACK *** for now just treat as "no information".
6052 * See [perl #56690].
6054 ssc_init(pRExC_state, data->start_class);
6056 /* AND before and after: combine and continue. These
6057 * assertions are zero-length, so can match an EMPTY
6059 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) &intrnl);
6060 ANYOF_FLAGS(data->start_class)
6061 |= SSC_MATCHES_EMPTY_STRING;
6065 #if PERL_ENABLE_POSITIVE_ASSERTION_STUDY
6067 /* Positive Lookahead/lookbehind
6068 In this case we can do fixed string optimisation,
6069 but we must be careful about it. Note in the case of
6070 lookbehind the positions will be offset by the minimum
6071 length of the pattern, something we won't know about
6072 until after the recurse.
6074 SSize_t deltanext, fake = 0;
6078 /* We use SAVEFREEPV so that when the full compile
6079 is finished perl will clean up the allocated
6080 minlens when it's all done. This way we don't
6081 have to worry about freeing them when we know
6082 they wont be used, which would be a pain.
6085 Newx( minnextp, 1, SSize_t );
6086 SAVEFREEPV(minnextp);
6089 StructCopy(data, &data_fake, scan_data_t);
6090 if ((flags & SCF_DO_SUBSTR) && data->last_found) {
6093 scan_commit(pRExC_state, &data_fake, minlenp, is_inf);
6094 data_fake.last_found=newSVsv(data->last_found);
6098 data_fake.last_closep = &fake;
6099 data_fake.flags = 0;
6100 data_fake.substrs[0].flags = 0;
6101 data_fake.substrs[1].flags = 0;
6102 data_fake.pos_delta = delta;
6104 data_fake.flags |= SF_IS_INF;
6105 if ( flags & SCF_DO_STCLASS && !scan->flags
6106 && OP(scan) == IFMATCH ) { /* Lookahead */
6107 ssc_init(pRExC_state, &intrnl);
6108 data_fake.start_class = &intrnl;
6109 f |= SCF_DO_STCLASS_AND;
6111 if (flags & SCF_WHILEM_VISITED_POS)
6112 f |= SCF_WHILEM_VISITED_POS;
6113 next = regnext(scan);
6114 nscan = NEXTOPER(NEXTOPER(scan));
6116 /* positive lookahead study_chunk() recursion */
6117 *minnextp = study_chunk(pRExC_state, &nscan, minnextp,
6118 &deltanext, last, &data_fake,
6119 stopparen, recursed_depth, NULL,
6122 assert(0); /* This code has never been tested since this
6123 is normally not compiled */
6125 || deltanext > (I32) U8_MAX
6126 || *minnextp > (I32)U8_MAX
6127 || *minnextp + deltanext > (I32)U8_MAX)
6129 FAIL2("Lookbehind longer than %" UVuf " not implemented",
6134 scan->next_off = deltanext;
6136 scan->flags = (U8)*minnextp + deltanext;
6141 if (f & SCF_DO_STCLASS_AND) {
6142 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) &intrnl);
6143 ANYOF_FLAGS(data->start_class) |= SSC_MATCHES_EMPTY_STRING;
6146 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
6148 if (data_fake.flags & SF_HAS_EVAL)
6149 data->flags |= SF_HAS_EVAL;
6150 data->whilem_c = data_fake.whilem_c;
6151 if ((flags & SCF_DO_SUBSTR) && data_fake.last_found) {
6153 if (RExC_rx->minlen<*minnextp)
6154 RExC_rx->minlen=*minnextp;
6155 scan_commit(pRExC_state, &data_fake, minnextp, is_inf);
6156 SvREFCNT_dec_NN(data_fake.last_found);
6158 for (i = 0; i < 2; i++) {
6159 if (data_fake.substrs[i].minlenp != minlenp) {
6160 data->substrs[i].min_offset =
6161 data_fake.substrs[i].min_offset;
6162 data->substrs[i].max_offset =
6163 data_fake.substrs[i].max_offset;
6164 data->substrs[i].minlenp =
6165 data_fake.substrs[i].minlenp;
6166 data->substrs[i].lookbehind += scan->flags;
6175 else if (OP(scan) == OPEN) {
6176 if (stopparen != (I32)ARG(scan))
6179 else if (OP(scan) == CLOSE) {
6180 if (stopparen == (I32)ARG(scan)) {
6183 if ((I32)ARG(scan) == is_par) {
6184 next = regnext(scan);
6186 if ( next && (OP(next) != WHILEM) && next < last)
6187 is_par = 0; /* Disable optimization */
6190 *(data->last_closep) = ARG(scan);
6192 else if (OP(scan) == EVAL) {
6194 data->flags |= SF_HAS_EVAL;
6196 else if ( PL_regkind[OP(scan)] == ENDLIKE ) {
6197 if (flags & SCF_DO_SUBSTR) {
6198 scan_commit(pRExC_state, data, minlenp, is_inf);
6199 flags &= ~SCF_DO_SUBSTR;
6201 if (data && OP(scan)==ACCEPT) {
6202 data->flags |= SCF_SEEN_ACCEPT;
6207 else if (OP(scan) == LOGICAL && scan->flags == 2) /* Embedded follows */
6209 if (flags & SCF_DO_SUBSTR) {
6210 scan_commit(pRExC_state, data, minlenp, is_inf);
6211 data->cur_is_floating = 1; /* float */
6213 is_inf = is_inf_internal = 1;
6214 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
6215 ssc_anything(data->start_class);
6216 flags &= ~SCF_DO_STCLASS;
6218 else if (OP(scan) == GPOS) {
6219 if (!(RExC_rx->intflags & PREGf_GPOS_FLOAT) &&
6220 !(delta || is_inf || (data && data->pos_delta)))
6222 if (!(RExC_rx->intflags & PREGf_ANCH) && (flags & SCF_DO_SUBSTR))
6223 RExC_rx->intflags |= PREGf_ANCH_GPOS;
6224 if (RExC_rx->gofs < (STRLEN)min)
6225 RExC_rx->gofs = min;
6227 RExC_rx->intflags |= PREGf_GPOS_FLOAT;
6231 #ifdef TRIE_STUDY_OPT
6232 #ifdef FULL_TRIE_STUDY
6233 else if (PL_regkind[OP(scan)] == TRIE) {
6234 /* NOTE - There is similar code to this block above for handling
6235 BRANCH nodes on the initial study. If you change stuff here
6237 regnode *trie_node= scan;
6238 regnode *tail= regnext(scan);
6239 reg_trie_data *trie = (reg_trie_data*)RExC_rxi->data->data[ ARG(scan) ];
6240 SSize_t max1 = 0, min1 = SSize_t_MAX;
6243 if (flags & SCF_DO_SUBSTR) { /* XXXX Add !SUSPEND? */
6244 /* Cannot merge strings after this. */
6245 scan_commit(pRExC_state, data, minlenp, is_inf);
6247 if (flags & SCF_DO_STCLASS)
6248 ssc_init_zero(pRExC_state, &accum);
6254 const regnode *nextbranch= NULL;
6257 for ( word=1 ; word <= trie->wordcount ; word++)
6259 SSize_t deltanext=0, minnext=0, f = 0, fake;
6260 regnode_ssc this_class;
6262 StructCopy(&zero_scan_data, &data_fake, scan_data_t);
6264 data_fake.whilem_c = data->whilem_c;
6265 data_fake.last_closep = data->last_closep;
6268 data_fake.last_closep = &fake;
6269 data_fake.pos_delta = delta;
6270 if (flags & SCF_DO_STCLASS) {
6271 ssc_init(pRExC_state, &this_class);
6272 data_fake.start_class = &this_class;
6273 f = SCF_DO_STCLASS_AND;
6275 if (flags & SCF_WHILEM_VISITED_POS)
6276 f |= SCF_WHILEM_VISITED_POS;
6278 if (trie->jump[word]) {
6280 nextbranch = trie_node + trie->jump[0];
6281 scan= trie_node + trie->jump[word];
6282 /* We go from the jump point to the branch that follows
6283 it. Note this means we need the vestigal unused
6284 branches even though they arent otherwise used. */
6285 /* optimise study_chunk() for TRIE */
6286 minnext = study_chunk(pRExC_state, &scan, minlenp,
6287 &deltanext, (regnode *)nextbranch, &data_fake,
6288 stopparen, recursed_depth, NULL, f, depth+1);
6290 if (nextbranch && PL_regkind[OP(nextbranch)]==BRANCH)
6291 nextbranch= regnext((regnode*)nextbranch);
6293 if (min1 > (SSize_t)(minnext + trie->minlen))
6294 min1 = minnext + trie->minlen;
6295 if (deltanext == SSize_t_MAX) {
6296 is_inf = is_inf_internal = 1;
6298 } else if (max1 < (SSize_t)(minnext + deltanext + trie->maxlen))
6299 max1 = minnext + deltanext + trie->maxlen;
6301 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
6303 if (data_fake.flags & SCF_SEEN_ACCEPT) {
6304 if ( stopmin > min + min1)
6305 stopmin = min + min1;
6306 flags &= ~SCF_DO_SUBSTR;
6308 data->flags |= SCF_SEEN_ACCEPT;
6311 if (data_fake.flags & SF_HAS_EVAL)
6312 data->flags |= SF_HAS_EVAL;
6313 data->whilem_c = data_fake.whilem_c;
6315 if (flags & SCF_DO_STCLASS)
6316 ssc_or(pRExC_state, &accum, (regnode_charclass *) &this_class);
6319 if (flags & SCF_DO_SUBSTR) {
6320 data->pos_min += min1;
6321 data->pos_delta += max1 - min1;
6322 if (max1 != min1 || is_inf)
6323 data->cur_is_floating = 1; /* float */
6326 if (delta != SSize_t_MAX) {
6327 if (SSize_t_MAX - (max1 - min1) >= delta)
6328 delta += max1 - min1;
6330 delta = SSize_t_MAX;
6332 if (flags & SCF_DO_STCLASS_OR) {
6333 ssc_or(pRExC_state, data->start_class, (regnode_charclass *) &accum);
6335 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
6336 flags &= ~SCF_DO_STCLASS;
6339 else if (flags & SCF_DO_STCLASS_AND) {
6341 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) &accum);
6342 flags &= ~SCF_DO_STCLASS;
6345 /* Switch to OR mode: cache the old value of
6346 * data->start_class */
6348 StructCopy(data->start_class, and_withp, regnode_ssc);
6349 flags &= ~SCF_DO_STCLASS_AND;
6350 StructCopy(&accum, data->start_class, regnode_ssc);
6351 flags |= SCF_DO_STCLASS_OR;
6358 else if (PL_regkind[OP(scan)] == TRIE) {
6359 reg_trie_data *trie = (reg_trie_data*)RExC_rxi->data->data[ ARG(scan) ];
6362 min += trie->minlen;
6363 delta += (trie->maxlen - trie->minlen);
6364 flags &= ~SCF_DO_STCLASS; /* xxx */
6365 if (flags & SCF_DO_SUBSTR) {
6366 /* Cannot expect anything... */
6367 scan_commit(pRExC_state, data, minlenp, is_inf);
6368 data->pos_min += trie->minlen;
6369 data->pos_delta += (trie->maxlen - trie->minlen);
6370 if (trie->maxlen != trie->minlen)
6371 data->cur_is_floating = 1; /* float */
6373 if (trie->jump) /* no more substrings -- for now /grr*/
6374 flags &= ~SCF_DO_SUBSTR;
6376 #endif /* old or new */
6377 #endif /* TRIE_STUDY_OPT */
6379 /* Else: zero-length, ignore. */
6380 scan = regnext(scan);
6385 /* we need to unwind recursion. */
6388 DEBUG_STUDYDATA("frame-end", data, depth, is_inf);
6389 DEBUG_PEEP("fend", scan, depth, flags);
6391 /* restore previous context */
6392 last = frame->last_regnode;
6393 scan = frame->next_regnode;
6394 stopparen = frame->stopparen;
6395 recursed_depth = frame->prev_recursed_depth;
6397 RExC_frame_last = frame->prev_frame;
6398 frame = frame->this_prev_frame;
6399 goto fake_study_recurse;
6403 DEBUG_STUDYDATA("pre-fin", data, depth, is_inf);
6406 *deltap = is_inf_internal ? SSize_t_MAX : delta;
6408 if (flags & SCF_DO_SUBSTR && is_inf)
6409 data->pos_delta = SSize_t_MAX - data->pos_min;
6410 if (is_par > (I32)U8_MAX)
6412 if (is_par && pars==1 && data) {
6413 data->flags |= SF_IN_PAR;
6414 data->flags &= ~SF_HAS_PAR;
6416 else if (pars && data) {
6417 data->flags |= SF_HAS_PAR;
6418 data->flags &= ~SF_IN_PAR;
6420 if (flags & SCF_DO_STCLASS_OR)
6421 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
6422 if (flags & SCF_TRIE_RESTUDY)
6423 data->flags |= SCF_TRIE_RESTUDY;
6425 DEBUG_STUDYDATA("post-fin", data, depth, is_inf);
6428 SSize_t final_minlen= min < stopmin ? min : stopmin;
6430 if (!(RExC_seen & REG_UNBOUNDED_QUANTIFIER_SEEN)) {
6431 if (final_minlen > SSize_t_MAX - delta)
6432 RExC_maxlen = SSize_t_MAX;
6433 else if (RExC_maxlen < final_minlen + delta)
6434 RExC_maxlen = final_minlen + delta;
6436 return final_minlen;
6438 NOT_REACHED; /* NOTREACHED */
6442 S_add_data(RExC_state_t* const pRExC_state, const char* const s, const U32 n)
6444 U32 count = RExC_rxi->data ? RExC_rxi->data->count : 0;
6446 PERL_ARGS_ASSERT_ADD_DATA;
6448 Renewc(RExC_rxi->data,
6449 sizeof(*RExC_rxi->data) + sizeof(void*) * (count + n - 1),
6450 char, struct reg_data);
6452 Renew(RExC_rxi->data->what, count + n, U8);
6454 Newx(RExC_rxi->data->what, n, U8);
6455 RExC_rxi->data->count = count + n;
6456 Copy(s, RExC_rxi->data->what + count, n, U8);
6460 /*XXX: todo make this not included in a non debugging perl, but appears to be
6461 * used anyway there, in 'use re' */
6462 #ifndef PERL_IN_XSUB_RE
6464 Perl_reginitcolors(pTHX)
6466 const char * const s = PerlEnv_getenv("PERL_RE_COLORS");
6468 char *t = savepv(s);
6472 t = strchr(t, '\t');
6478 PL_colors[i] = t = (char *)"";
6483 PL_colors[i++] = (char *)"";
6490 #ifdef TRIE_STUDY_OPT
6491 #define CHECK_RESTUDY_GOTO_butfirst(dOsomething) \
6494 (data.flags & SCF_TRIE_RESTUDY) \
6502 #define CHECK_RESTUDY_GOTO_butfirst
6506 * pregcomp - compile a regular expression into internal code
6508 * Decides which engine's compiler to call based on the hint currently in
6512 #ifndef PERL_IN_XSUB_RE
6514 /* return the currently in-scope regex engine (or the default if none) */
6516 regexp_engine const *
6517 Perl_current_re_engine(pTHX)
6519 if (IN_PERL_COMPILETIME) {
6520 HV * const table = GvHV(PL_hintgv);
6523 if (!table || !(PL_hints & HINT_LOCALIZE_HH))
6524 return &PL_core_reg_engine;
6525 ptr = hv_fetchs(table, "regcomp", FALSE);
6526 if ( !(ptr && SvIOK(*ptr) && SvIV(*ptr)))
6527 return &PL_core_reg_engine;
6528 return INT2PTR(regexp_engine*, SvIV(*ptr));
6532 if (!PL_curcop->cop_hints_hash)
6533 return &PL_core_reg_engine;
6534 ptr = cop_hints_fetch_pvs(PL_curcop, "regcomp", 0);
6535 if ( !(ptr && SvIOK(ptr) && SvIV(ptr)))
6536 return &PL_core_reg_engine;
6537 return INT2PTR(regexp_engine*, SvIV(ptr));
6543 Perl_pregcomp(pTHX_ SV * const pattern, const U32 flags)
6545 regexp_engine const *eng = current_re_engine();
6546 GET_RE_DEBUG_FLAGS_DECL;
6548 PERL_ARGS_ASSERT_PREGCOMP;
6550 /* Dispatch a request to compile a regexp to correct regexp engine. */
6552 Perl_re_printf( aTHX_ "Using engine %" UVxf "\n",
6555 return CALLREGCOMP_ENG(eng, pattern, flags);
6559 /* public(ish) entry point for the perl core's own regex compiling code.
6560 * It's actually a wrapper for Perl_re_op_compile that only takes an SV
6561 * pattern rather than a list of OPs, and uses the internal engine rather
6562 * than the current one */
6565 Perl_re_compile(pTHX_ SV * const pattern, U32 rx_flags)
6567 SV *pat = pattern; /* defeat constness! */
6568 PERL_ARGS_ASSERT_RE_COMPILE;
6569 return Perl_re_op_compile(aTHX_ &pat, 1, NULL,
6570 #ifdef PERL_IN_XSUB_RE
6573 &PL_core_reg_engine,
6575 NULL, NULL, rx_flags, 0);
6580 S_free_codeblocks(pTHX_ struct reg_code_blocks *cbs)
6584 if (--cbs->refcnt > 0)
6586 for (n = 0; n < cbs->count; n++) {
6587 REGEXP *rx = cbs->cb[n].src_regex;
6589 cbs->cb[n].src_regex = NULL;
6590 SvREFCNT_dec_NN(rx);
6598 static struct reg_code_blocks *
6599 S_alloc_code_blocks(pTHX_ int ncode)
6601 struct reg_code_blocks *cbs;
6602 Newx(cbs, 1, struct reg_code_blocks);
6605 SAVEDESTRUCTOR_X(S_free_codeblocks, cbs);
6607 Newx(cbs->cb, ncode, struct reg_code_block);
6614 /* upgrade pattern pat_p of length plen_p to UTF8, and if there are code
6615 * blocks, recalculate the indices. Update pat_p and plen_p in-place to
6616 * point to the realloced string and length.
6618 * This is essentially a copy of Perl_bytes_to_utf8() with the code index
6622 S_pat_upgrade_to_utf8(pTHX_ RExC_state_t * const pRExC_state,
6623 char **pat_p, STRLEN *plen_p, int num_code_blocks)
6625 U8 *const src = (U8*)*pat_p;
6630 GET_RE_DEBUG_FLAGS_DECL;
6632 DEBUG_PARSE_r(Perl_re_printf( aTHX_
6633 "UTF8 mismatch! Converting to utf8 for resizing and compile\n"));
6635 /* 1 for each byte + 1 for each byte that expands to two, + trailing NUL */
6636 Newx(dst, *plen_p + variant_under_utf8_count(src, src + *plen_p) + 1, U8);
6639 while (s < *plen_p) {
6640 append_utf8_from_native_byte(src[s], &d);
6642 if (n < num_code_blocks) {
6643 assert(pRExC_state->code_blocks);
6644 if (!do_end && pRExC_state->code_blocks->cb[n].start == s) {
6645 pRExC_state->code_blocks->cb[n].start = d - dst - 1;
6646 assert(*(d - 1) == '(');
6649 else if (do_end && pRExC_state->code_blocks->cb[n].end == s) {
6650 pRExC_state->code_blocks->cb[n].end = d - dst - 1;
6651 assert(*(d - 1) == ')');
6660 *pat_p = (char*) dst;
6662 RExC_orig_utf8 = RExC_utf8 = 1;
6667 /* S_concat_pat(): concatenate a list of args to the pattern string pat,
6668 * while recording any code block indices, and handling overloading,
6669 * nested qr// objects etc. If pat is null, it will allocate a new
6670 * string, or just return the first arg, if there's only one.
6672 * Returns the malloced/updated pat.
6673 * patternp and pat_count is the array of SVs to be concatted;
6674 * oplist is the optional list of ops that generated the SVs;
6675 * recompile_p is a pointer to a boolean that will be set if
6676 * the regex will need to be recompiled.
6677 * delim, if non-null is an SV that will be inserted between each element
6681 S_concat_pat(pTHX_ RExC_state_t * const pRExC_state,
6682 SV *pat, SV ** const patternp, int pat_count,
6683 OP *oplist, bool *recompile_p, SV *delim)
6687 bool use_delim = FALSE;
6688 bool alloced = FALSE;
6690 /* if we know we have at least two args, create an empty string,
6691 * then concatenate args to that. For no args, return an empty string */
6692 if (!pat && pat_count != 1) {
6698 for (svp = patternp; svp < patternp + pat_count; svp++) {
6701 STRLEN orig_patlen = 0;
6703 SV *msv = use_delim ? delim : *svp;
6704 if (!msv) msv = &PL_sv_undef;
6706 /* if we've got a delimiter, we go round the loop twice for each
6707 * svp slot (except the last), using the delimiter the second
6716 if (SvTYPE(msv) == SVt_PVAV) {
6717 /* we've encountered an interpolated array within
6718 * the pattern, e.g. /...@a..../. Expand the list of elements,
6719 * then recursively append elements.
6720 * The code in this block is based on S_pushav() */
6722 AV *const av = (AV*)msv;
6723 const SSize_t maxarg = AvFILL(av) + 1;
6727 assert(oplist->op_type == OP_PADAV
6728 || oplist->op_type == OP_RV2AV);
6729 oplist = OpSIBLING(oplist);
6732 if (SvRMAGICAL(av)) {
6735 Newx(array, maxarg, SV*);
6737 for (i=0; i < maxarg; i++) {
6738 SV ** const svp = av_fetch(av, i, FALSE);
6739 array[i] = svp ? *svp : &PL_sv_undef;
6743 array = AvARRAY(av);
6745 pat = S_concat_pat(aTHX_ pRExC_state, pat,
6746 array, maxarg, NULL, recompile_p,
6748 GvSV((gv_fetchpvs("\"", GV_ADDMULTI, SVt_PV))));
6754 /* we make the assumption here that each op in the list of
6755 * op_siblings maps to one SV pushed onto the stack,
6756 * except for code blocks, with have both an OP_NULL and
6758 * This allows us to match up the list of SVs against the
6759 * list of OPs to find the next code block.
6761 * Note that PUSHMARK PADSV PADSV ..
6763 * PADRANGE PADSV PADSV ..
6764 * so the alignment still works. */
6767 if (oplist->op_type == OP_NULL
6768 && (oplist->op_flags & OPf_SPECIAL))
6770 assert(n < pRExC_state->code_blocks->count);
6771 pRExC_state->code_blocks->cb[n].start = pat ? SvCUR(pat) : 0;
6772 pRExC_state->code_blocks->cb[n].block = oplist;
6773 pRExC_state->code_blocks->cb[n].src_regex = NULL;
6776 oplist = OpSIBLING(oplist); /* skip CONST */
6779 oplist = OpSIBLING(oplist);;
6782 /* apply magic and QR overloading to arg */
6785 if (SvROK(msv) && SvAMAGIC(msv)) {
6786 SV *sv = AMG_CALLunary(msv, regexp_amg);
6790 if (SvTYPE(sv) != SVt_REGEXP)
6791 Perl_croak(aTHX_ "Overloaded qr did not return a REGEXP");
6796 /* try concatenation overload ... */
6797 if (pat && (SvAMAGIC(pat) || SvAMAGIC(msv)) &&
6798 (sv = amagic_call(pat, msv, concat_amg, AMGf_assign)))
6801 /* overloading involved: all bets are off over literal
6802 * code. Pretend we haven't seen it */
6804 pRExC_state->code_blocks->count -= n;
6808 /* ... or failing that, try "" overload */
6809 while (SvAMAGIC(msv)
6810 && (sv = AMG_CALLunary(msv, string_amg))
6814 && SvRV(msv) == SvRV(sv))
6819 if (SvROK(msv) && SvTYPE(SvRV(msv)) == SVt_REGEXP)
6823 /* this is a partially unrolled
6824 * sv_catsv_nomg(pat, msv);
6825 * that allows us to adjust code block indices if
6828 char *dst = SvPV_force_nomg(pat, dlen);
6830 if (SvUTF8(msv) && !SvUTF8(pat)) {
6831 S_pat_upgrade_to_utf8(aTHX_ pRExC_state, &dst, &dlen, n);
6832 sv_setpvn(pat, dst, dlen);
6835 sv_catsv_nomg(pat, msv);
6839 /* We have only one SV to process, but we need to verify
6840 * it is properly null terminated or we will fail asserts
6841 * later. In theory we probably shouldn't get such SV's,
6842 * but if we do we should handle it gracefully. */
6843 if ( SvTYPE(msv) != SVt_PV || (SvLEN(msv) > SvCUR(msv) && *(SvEND(msv)) == 0) || SvIsCOW_shared_hash(msv) ) {
6844 /* not a string, or a string with a trailing null */
6847 /* a string with no trailing null, we need to copy it
6848 * so it has a trailing null */
6849 pat = sv_2mortal(newSVsv(msv));
6854 pRExC_state->code_blocks->cb[n-1].end = SvCUR(pat)-1;
6857 /* extract any code blocks within any embedded qr//'s */
6858 if (rx && SvTYPE(rx) == SVt_REGEXP
6859 && RX_ENGINE((REGEXP*)rx)->op_comp)
6862 RXi_GET_DECL(ReANY((REGEXP *)rx), ri);
6863 if (ri->code_blocks && ri->code_blocks->count) {
6865 /* the presence of an embedded qr// with code means
6866 * we should always recompile: the text of the
6867 * qr// may not have changed, but it may be a
6868 * different closure than last time */
6870 if (pRExC_state->code_blocks) {
6871 int new_count = pRExC_state->code_blocks->count
6872 + ri->code_blocks->count;
6873 Renew(pRExC_state->code_blocks->cb,
6874 new_count, struct reg_code_block);
6875 pRExC_state->code_blocks->count = new_count;
6878 pRExC_state->code_blocks = S_alloc_code_blocks(aTHX_
6879 ri->code_blocks->count);
6881 for (i=0; i < ri->code_blocks->count; i++) {
6882 struct reg_code_block *src, *dst;
6883 STRLEN offset = orig_patlen
6884 + ReANY((REGEXP *)rx)->pre_prefix;
6885 assert(n < pRExC_state->code_blocks->count);
6886 src = &ri->code_blocks->cb[i];
6887 dst = &pRExC_state->code_blocks->cb[n];
6888 dst->start = src->start + offset;
6889 dst->end = src->end + offset;
6890 dst->block = src->block;
6891 dst->src_regex = (REGEXP*) SvREFCNT_inc( (SV*)
6900 /* avoid calling magic multiple times on a single element e.g. =~ $qr */
6909 /* see if there are any run-time code blocks in the pattern.
6910 * False positives are allowed */
6913 S_has_runtime_code(pTHX_ RExC_state_t * const pRExC_state,
6914 char *pat, STRLEN plen)
6919 PERL_UNUSED_CONTEXT;
6921 for (s = 0; s < plen; s++) {
6922 if ( pRExC_state->code_blocks
6923 && n < pRExC_state->code_blocks->count
6924 && s == pRExC_state->code_blocks->cb[n].start)
6926 s = pRExC_state->code_blocks->cb[n].end;
6930 /* TODO ideally should handle [..], (#..), /#.../x to reduce false
6932 if (pat[s] == '(' && s+2 <= plen && pat[s+1] == '?' &&
6934 || (s + 2 <= plen && pat[s+2] == '?' && pat[s+3] == '{'))
6941 /* Handle run-time code blocks. We will already have compiled any direct
6942 * or indirect literal code blocks. Now, take the pattern 'pat' and make a
6943 * copy of it, but with any literal code blocks blanked out and
6944 * appropriate chars escaped; then feed it into
6946 * eval "qr'modified_pattern'"
6950 * a\bc(?{"this was literal"})def'ghi\\jkl(?{"this is runtime"})mno
6954 * qr'a\\bc_______________________def\'ghi\\\\jkl(?{"this is runtime"})mno'
6956 * After eval_sv()-ing that, grab any new code blocks from the returned qr
6957 * and merge them with any code blocks of the original regexp.
6959 * If the pat is non-UTF8, while the evalled qr is UTF8, don't merge;
6960 * instead, just save the qr and return FALSE; this tells our caller that
6961 * the original pattern needs upgrading to utf8.
6965 S_compile_runtime_code(pTHX_ RExC_state_t * const pRExC_state,
6966 char *pat, STRLEN plen)
6970 GET_RE_DEBUG_FLAGS_DECL;
6972 if (pRExC_state->runtime_code_qr) {
6973 /* this is the second time we've been called; this should
6974 * only happen if the main pattern got upgraded to utf8
6975 * during compilation; re-use the qr we compiled first time
6976 * round (which should be utf8 too)
6978 qr = pRExC_state->runtime_code_qr;
6979 pRExC_state->runtime_code_qr = NULL;
6980 assert(RExC_utf8 && SvUTF8(qr));
6986 int newlen = plen + 7; /* allow for "qr''xx\0" extra chars */
6990 /* determine how many extra chars we need for ' and \ escaping */
6991 for (s = 0; s < plen; s++) {
6992 if (pat[s] == '\'' || pat[s] == '\\')
6996 Newx(newpat, newlen, char);
6998 *p++ = 'q'; *p++ = 'r'; *p++ = '\'';
7000 for (s = 0; s < plen; s++) {
7001 if ( pRExC_state->code_blocks
7002 && n < pRExC_state->code_blocks->count
7003 && s == pRExC_state->code_blocks->cb[n].start)
7005 /* blank out literal code block so that they aren't
7006 * recompiled: eg change from/to:
7016 assert(pat[s] == '(');
7017 assert(pat[s+1] == '?');
7021 while (s < pRExC_state->code_blocks->cb[n].end) {
7029 if (pat[s] == '\'' || pat[s] == '\\')
7034 if (pRExC_state->pm_flags & RXf_PMf_EXTENDED) {
7036 if (pRExC_state->pm_flags & RXf_PMf_EXTENDED_MORE) {
7042 Perl_re_printf( aTHX_
7043 "%sre-parsing pattern for runtime code:%s %s\n",
7044 PL_colors[4], PL_colors[5], newpat);
7047 sv = newSVpvn_flags(newpat, p-newpat-1, RExC_utf8 ? SVf_UTF8 : 0);
7053 PUSHSTACKi(PERLSI_REQUIRE);
7054 /* G_RE_REPARSING causes the toker to collapse \\ into \ when
7055 * parsing qr''; normally only q'' does this. It also alters
7057 eval_sv(sv, G_SCALAR|G_RE_REPARSING);
7058 SvREFCNT_dec_NN(sv);
7063 SV * const errsv = ERRSV;
7064 if (SvTRUE_NN(errsv))
7065 /* use croak_sv ? */
7066 Perl_croak_nocontext("%" SVf, SVfARG(errsv));
7068 assert(SvROK(qr_ref));
7070 assert(SvTYPE(qr) == SVt_REGEXP && RX_ENGINE((REGEXP*)qr)->op_comp);
7071 /* the leaving below frees the tmp qr_ref.
7072 * Give qr a life of its own */
7080 if (!RExC_utf8 && SvUTF8(qr)) {
7081 /* first time through; the pattern got upgraded; save the
7082 * qr for the next time through */
7083 assert(!pRExC_state->runtime_code_qr);
7084 pRExC_state->runtime_code_qr = qr;
7089 /* extract any code blocks within the returned qr// */
7092 /* merge the main (r1) and run-time (r2) code blocks into one */
7094 RXi_GET_DECL(ReANY((REGEXP *)qr), r2);
7095 struct reg_code_block *new_block, *dst;
7096 RExC_state_t * const r1 = pRExC_state; /* convenient alias */
7100 if (!r2->code_blocks || !r2->code_blocks->count) /* we guessed wrong */
7102 SvREFCNT_dec_NN(qr);
7106 if (!r1->code_blocks)
7107 r1->code_blocks = S_alloc_code_blocks(aTHX_ 0);
7109 r1c = r1->code_blocks->count;
7110 r2c = r2->code_blocks->count;
7112 Newx(new_block, r1c + r2c, struct reg_code_block);
7116 while (i1 < r1c || i2 < r2c) {
7117 struct reg_code_block *src;
7121 src = &r2->code_blocks->cb[i2++];
7125 src = &r1->code_blocks->cb[i1++];
7126 else if ( r1->code_blocks->cb[i1].start
7127 < r2->code_blocks->cb[i2].start)
7129 src = &r1->code_blocks->cb[i1++];
7130 assert(src->end < r2->code_blocks->cb[i2].start);
7133 assert( r1->code_blocks->cb[i1].start
7134 > r2->code_blocks->cb[i2].start);
7135 src = &r2->code_blocks->cb[i2++];
7137 assert(src->end < r1->code_blocks->cb[i1].start);
7140 assert(pat[src->start] == '(');
7141 assert(pat[src->end] == ')');
7142 dst->start = src->start;
7143 dst->end = src->end;
7144 dst->block = src->block;
7145 dst->src_regex = is_qr ? (REGEXP*) SvREFCNT_inc( (SV*) qr)
7149 r1->code_blocks->count += r2c;
7150 Safefree(r1->code_blocks->cb);
7151 r1->code_blocks->cb = new_block;
7154 SvREFCNT_dec_NN(qr);
7160 S_setup_longest(pTHX_ RExC_state_t *pRExC_state,
7161 struct reg_substr_datum *rsd,
7162 struct scan_data_substrs *sub,
7163 STRLEN longest_length)
7165 /* This is the common code for setting up the floating and fixed length
7166 * string data extracted from Perl_re_op_compile() below. Returns a boolean
7167 * as to whether succeeded or not */
7171 bool eol = cBOOL(sub->flags & SF_BEFORE_EOL);
7172 bool meol = cBOOL(sub->flags & SF_BEFORE_MEOL);
7174 if (! (longest_length
7175 || (eol /* Can't have SEOL and MULTI */
7176 && (! meol || (RExC_flags & RXf_PMf_MULTILINE)))
7178 /* See comments for join_exact for why REG_UNFOLDED_MULTI_SEEN */
7179 || (RExC_seen & REG_UNFOLDED_MULTI_SEEN))
7184 /* copy the information about the longest from the reg_scan_data
7185 over to the program. */
7186 if (SvUTF8(sub->str)) {
7188 rsd->utf8_substr = sub->str;
7190 rsd->substr = sub->str;
7191 rsd->utf8_substr = NULL;
7193 /* end_shift is how many chars that must be matched that
7194 follow this item. We calculate it ahead of time as once the
7195 lookbehind offset is added in we lose the ability to correctly
7197 ml = sub->minlenp ? *(sub->minlenp) : (SSize_t)longest_length;
7198 rsd->end_shift = ml - sub->min_offset
7200 /* XXX SvTAIL is always false here - did you mean FBMcf_TAIL
7202 + (SvTAIL(sub->str) != 0)
7206 t = (eol/* Can't have SEOL and MULTI */
7207 && (! meol || (RExC_flags & RXf_PMf_MULTILINE)));
7208 fbm_compile(sub->str, t ? FBMcf_TAIL : 0);
7214 S_set_regex_pv(pTHX_ RExC_state_t *pRExC_state, REGEXP *Rx)
7216 /* Calculates and sets in the compiled pattern 'Rx' the string to compile,
7217 * properly wrapped with the right modifiers */
7219 bool has_p = ((RExC_rx->extflags & RXf_PMf_KEEPCOPY) == RXf_PMf_KEEPCOPY);
7220 bool has_charset = RExC_utf8 || (get_regex_charset(RExC_rx->extflags)
7221 != REGEX_DEPENDS_CHARSET);
7223 /* The caret is output if there are any defaults: if not all the STD
7224 * flags are set, or if no character set specifier is needed */
7226 (((RExC_rx->extflags & RXf_PMf_STD_PMMOD) != RXf_PMf_STD_PMMOD)
7228 bool has_runon = ((RExC_seen & REG_RUN_ON_COMMENT_SEEN)
7229 == REG_RUN_ON_COMMENT_SEEN);
7230 U8 reganch = (U8)((RExC_rx->extflags & RXf_PMf_STD_PMMOD)
7231 >> RXf_PMf_STD_PMMOD_SHIFT);
7232 const char *fptr = STD_PAT_MODS; /*"msixxn"*/
7234 STRLEN pat_len = RExC_precomp_end - RExC_precomp;
7236 /* We output all the necessary flags; we never output a minus, as all
7237 * those are defaults, so are
7238 * covered by the caret */
7239 const STRLEN wraplen = pat_len + has_p + has_runon
7240 + has_default /* If needs a caret */
7241 + PL_bitcount[reganch] /* 1 char for each set standard flag */
7243 /* If needs a character set specifier */
7244 + ((has_charset) ? MAX_CHARSET_NAME_LENGTH : 0)
7245 + (sizeof("(?:)") - 1);
7247 PERL_ARGS_ASSERT_SET_REGEX_PV;
7249 /* make sure PL_bitcount bounds not exceeded */
7250 assert(sizeof(STD_PAT_MODS) <= 8);
7252 p = sv_grow(MUTABLE_SV(Rx), wraplen + 1); /* +1 for the ending NUL */
7255 SvFLAGS(Rx) |= SVf_UTF8;
7258 /* If a default, cover it using the caret */
7260 *p++= DEFAULT_PAT_MOD;
7266 name = get_regex_charset_name(RExC_rx->extflags, &len);
7267 if (strEQ(name, DEPENDS_PAT_MODS)) { /* /d under UTF-8 => /u */
7269 name = UNICODE_PAT_MODS;
7270 len = sizeof(UNICODE_PAT_MODS) - 1;
7272 Copy(name, p, len, char);
7276 *p++ = KEEPCOPY_PAT_MOD; /*'p'*/
7279 while((ch = *fptr++)) {
7287 Copy(RExC_precomp, p, pat_len, char);
7288 assert ((RX_WRAPPED(Rx) - p) < 16);
7289 RExC_rx->pre_prefix = p - RX_WRAPPED(Rx);
7292 /* Adding a trailing \n causes this to compile properly:
7293 my $R = qr / A B C # D E/x; /($R)/
7294 Otherwise the parens are considered part of the comment */
7299 SvCUR_set(Rx, p - RX_WRAPPED(Rx));
7303 * Perl_re_op_compile - the perl internal RE engine's function to compile a
7304 * regular expression into internal code.
7305 * The pattern may be passed either as:
7306 * a list of SVs (patternp plus pat_count)
7307 * a list of OPs (expr)
7308 * If both are passed, the SV list is used, but the OP list indicates
7309 * which SVs are actually pre-compiled code blocks
7311 * The SVs in the list have magic and qr overloading applied to them (and
7312 * the list may be modified in-place with replacement SVs in the latter
7315 * If the pattern hasn't changed from old_re, then old_re will be
7318 * eng is the current engine. If that engine has an op_comp method, then
7319 * handle directly (i.e. we assume that op_comp was us); otherwise, just
7320 * do the initial concatenation of arguments and pass on to the external
7323 * If is_bare_re is not null, set it to a boolean indicating whether the
7324 * arg list reduced (after overloading) to a single bare regex which has
7325 * been returned (i.e. /$qr/).
7327 * orig_rx_flags contains RXf_* flags. See perlreapi.pod for more details.
7329 * pm_flags contains the PMf_* flags, typically based on those from the
7330 * pm_flags field of the related PMOP. Currently we're only interested in
7331 * PMf_HAS_CV, PMf_IS_QR, PMf_USE_RE_EVAL.
7333 * For many years this code had an initial sizing pass that calculated
7334 * (sometimes incorrectly, leading to security holes) the size needed for the
7335 * compiled pattern. That was changed by commit
7336 * 7c932d07cab18751bfc7515b4320436273a459e2 in 5.29, which reallocs the size, a
7337 * node at a time, as parsing goes along. Patches welcome to fix any obsolete
7338 * references to this sizing pass.
7340 * Now, an initial crude guess as to the size needed is made, based on the
7341 * length of the pattern. Patches welcome to improve that guess. That amount
7342 * of space is malloc'd and then immediately freed, and then clawed back node
7343 * by node. This design is to minimze, to the extent possible, memory churn
7344 * when doing the the reallocs.
7346 * A separate parentheses counting pass may be needed in some cases.
7347 * (Previously the sizing pass did this.) Patches welcome to reduce the number
7350 * The existence of a sizing pass necessitated design decisions that are no
7351 * longer needed. There are potential areas of simplification.
7353 * Beware that the optimization-preparation code in here knows about some
7354 * of the structure of the compiled regexp. [I'll say.]
7358 Perl_re_op_compile(pTHX_ SV ** const patternp, int pat_count,
7359 OP *expr, const regexp_engine* eng, REGEXP *old_re,
7360 bool *is_bare_re, const U32 orig_rx_flags, const U32 pm_flags)
7363 REGEXP *Rx; /* Capital 'R' means points to a REGEXP */
7371 SV** new_patternp = patternp;
7373 /* these are all flags - maybe they should be turned
7374 * into a single int with different bit masks */
7375 I32 sawlookahead = 0;
7380 regex_charset initial_charset = get_regex_charset(orig_rx_flags);
7382 bool runtime_code = 0;
7384 RExC_state_t RExC_state;
7385 RExC_state_t * const pRExC_state = &RExC_state;
7386 #ifdef TRIE_STUDY_OPT
7388 RExC_state_t copyRExC_state;
7390 GET_RE_DEBUG_FLAGS_DECL;
7392 PERL_ARGS_ASSERT_RE_OP_COMPILE;
7394 DEBUG_r(if (!PL_colorset) reginitcolors());
7396 /* Initialize these here instead of as-needed, as is quick and avoids
7397 * having to test them each time otherwise */
7398 if (! PL_InBitmap) {
7400 char * dump_len_string;
7403 /* This is calculated here, because the Perl program that generates the
7404 * static global ones doesn't currently have access to
7405 * NUM_ANYOF_CODE_POINTS */
7406 PL_InBitmap = _new_invlist(2);
7407 PL_InBitmap = _add_range_to_invlist(PL_InBitmap, 0,
7408 NUM_ANYOF_CODE_POINTS - 1);
7410 dump_len_string = PerlEnv_getenv("PERL_DUMP_RE_MAX_LEN");
7411 if ( ! dump_len_string
7412 || ! grok_atoUV(dump_len_string, (UV *)&PL_dump_re_max_len, NULL))
7414 PL_dump_re_max_len = 60; /* A reasonable default */
7419 pRExC_state->warn_text = NULL;
7420 pRExC_state->unlexed_names = NULL;
7421 pRExC_state->code_blocks = NULL;
7424 *is_bare_re = FALSE;
7426 if (expr && (expr->op_type == OP_LIST ||
7427 (expr->op_type == OP_NULL && expr->op_targ == OP_LIST))) {
7428 /* allocate code_blocks if needed */
7432 for (o = cLISTOPx(expr)->op_first; o; o = OpSIBLING(o))
7433 if (o->op_type == OP_NULL && (o->op_flags & OPf_SPECIAL))
7434 ncode++; /* count of DO blocks */
7437 pRExC_state->code_blocks = S_alloc_code_blocks(aTHX_ ncode);
7441 /* compile-time pattern with just OP_CONSTs and DO blocks */
7446 /* find how many CONSTs there are */
7449 if (expr->op_type == OP_CONST)
7452 for (o = cLISTOPx(expr)->op_first; o; o = OpSIBLING(o)) {
7453 if (o->op_type == OP_CONST)
7457 /* fake up an SV array */
7459 assert(!new_patternp);
7460 Newx(new_patternp, n, SV*);
7461 SAVEFREEPV(new_patternp);
7465 if (expr->op_type == OP_CONST)
7466 new_patternp[n] = cSVOPx_sv(expr);
7468 for (o = cLISTOPx(expr)->op_first; o; o = OpSIBLING(o)) {
7469 if (o->op_type == OP_CONST)
7470 new_patternp[n++] = cSVOPo_sv;
7475 DEBUG_PARSE_r(Perl_re_printf( aTHX_
7476 "Assembling pattern from %d elements%s\n", pat_count,
7477 orig_rx_flags & RXf_SPLIT ? " for split" : ""));
7479 /* set expr to the first arg op */
7481 if (pRExC_state->code_blocks && pRExC_state->code_blocks->count
7482 && expr->op_type != OP_CONST)
7484 expr = cLISTOPx(expr)->op_first;
7485 assert( expr->op_type == OP_PUSHMARK
7486 || (expr->op_type == OP_NULL && expr->op_targ == OP_PUSHMARK)
7487 || expr->op_type == OP_PADRANGE);
7488 expr = OpSIBLING(expr);
7491 pat = S_concat_pat(aTHX_ pRExC_state, NULL, new_patternp, pat_count,
7492 expr, &recompile, NULL);
7494 /* handle bare (possibly after overloading) regex: foo =~ $re */
7499 if (SvTYPE(re) == SVt_REGEXP) {
7503 DEBUG_PARSE_r(Perl_re_printf( aTHX_
7504 "Precompiled pattern%s\n",
7505 orig_rx_flags & RXf_SPLIT ? " for split" : ""));
7511 exp = SvPV_nomg(pat, plen);
7513 if (!eng->op_comp) {
7514 if ((SvUTF8(pat) && IN_BYTES)
7515 || SvGMAGICAL(pat) || SvAMAGIC(pat))
7517 /* make a temporary copy; either to convert to bytes,
7518 * or to avoid repeating get-magic / overloaded stringify */
7519 pat = newSVpvn_flags(exp, plen, SVs_TEMP |
7520 (IN_BYTES ? 0 : SvUTF8(pat)));
7522 return CALLREGCOMP_ENG(eng, pat, orig_rx_flags);
7525 /* ignore the utf8ness if the pattern is 0 length */
7526 RExC_utf8 = RExC_orig_utf8 = (plen == 0 || IN_BYTES) ? 0 : SvUTF8(pat);
7527 RExC_uni_semantics = 0;
7528 RExC_contains_locale = 0;
7529 RExC_strict = cBOOL(pm_flags & RXf_PMf_STRICT);
7530 RExC_in_script_run = 0;
7531 RExC_study_started = 0;
7532 pRExC_state->runtime_code_qr = NULL;
7533 RExC_frame_head= NULL;
7534 RExC_frame_last= NULL;
7535 RExC_frame_count= 0;
7536 RExC_latest_warn_offset = 0;
7537 RExC_use_BRANCHJ = 0;
7538 RExC_total_parens = 0;
7539 RExC_open_parens = NULL;
7540 RExC_close_parens = NULL;
7541 RExC_paren_names = NULL;
7543 RExC_seen_d_op = FALSE;
7545 RExC_paren_name_list = NULL;
7549 RExC_mysv1= sv_newmortal();
7550 RExC_mysv2= sv_newmortal();
7554 SV *dsv= sv_newmortal();
7555 RE_PV_QUOTED_DECL(s, RExC_utf8, dsv, exp, plen, PL_dump_re_max_len);
7556 Perl_re_printf( aTHX_ "%sCompiling REx%s %s\n",
7557 PL_colors[4], PL_colors[5], s);
7560 /* we jump here if we have to recompile, e.g., from upgrading the pattern
7563 if ((pm_flags & PMf_USE_RE_EVAL)
7564 /* this second condition covers the non-regex literal case,
7565 * i.e. $foo =~ '(?{})'. */
7566 || (IN_PERL_COMPILETIME && (PL_hints & HINT_RE_EVAL))
7568 runtime_code = S_has_runtime_code(aTHX_ pRExC_state, exp, plen);
7571 /* return old regex if pattern hasn't changed */
7572 /* XXX: note in the below we have to check the flags as well as the
7575 * Things get a touch tricky as we have to compare the utf8 flag
7576 * independently from the compile flags. */
7580 && !!RX_UTF8(old_re) == !!RExC_utf8
7581 && ( RX_COMPFLAGS(old_re) == ( orig_rx_flags & RXf_PMf_FLAGCOPYMASK ) )
7582 && RX_PRECOMP(old_re)
7583 && RX_PRELEN(old_re) == plen
7584 && memEQ(RX_PRECOMP(old_re), exp, plen)
7585 && !runtime_code /* with runtime code, always recompile */ )
7590 /* Allocate the pattern's SV */
7591 RExC_rx_sv = Rx = (REGEXP*) newSV_type(SVt_REGEXP);
7592 RExC_rx = ReANY(Rx);
7593 if ( RExC_rx == NULL )
7594 FAIL("Regexp out of space");
7596 rx_flags = orig_rx_flags;
7598 if ( (UTF || RExC_uni_semantics)
7599 && initial_charset == REGEX_DEPENDS_CHARSET)
7602 /* Set to use unicode semantics if the pattern is in utf8 and has the
7603 * 'depends' charset specified, as it means unicode when utf8 */
7604 set_regex_charset(&rx_flags, REGEX_UNICODE_CHARSET);
7605 RExC_uni_semantics = 1;
7608 RExC_pm_flags = pm_flags;
7611 assert(TAINTING_get || !TAINT_get);
7613 Perl_croak(aTHX_ "Eval-group in insecure regular expression");
7615 if (!S_compile_runtime_code(aTHX_ pRExC_state, exp, plen)) {
7616 /* whoops, we have a non-utf8 pattern, whilst run-time code
7617 * got compiled as utf8. Try again with a utf8 pattern */
7618 S_pat_upgrade_to_utf8(aTHX_ pRExC_state, &exp, &plen,
7619 pRExC_state->code_blocks ? pRExC_state->code_blocks->count : 0);
7623 assert(!pRExC_state->runtime_code_qr);
7629 RExC_in_lookbehind = 0;
7630 RExC_in_lookahead = 0;
7631 RExC_seen_zerolen = *exp == '^' ? -1 : 0;
7632 RExC_recode_x_to_native = 0;
7633 RExC_in_multi_char_class = 0;
7635 RExC_start = RExC_copy_start_in_constructed = RExC_copy_start_in_input = RExC_precomp = exp;
7636 RExC_precomp_end = RExC_end = exp + plen;
7638 RExC_whilem_seen = 0;
7640 RExC_recurse = NULL;
7641 RExC_study_chunk_recursed = NULL;
7642 RExC_study_chunk_recursed_bytes= 0;
7643 RExC_recurse_count = 0;
7644 pRExC_state->code_index = 0;
7646 /* Initialize the string in the compiled pattern. This is so that there is
7647 * something to output if necessary */
7648 set_regex_pv(pRExC_state, Rx);
7651 Perl_re_printf( aTHX_
7652 "Starting parse and generation\n");
7654 RExC_lastparse=NULL;
7657 /* Allocate space and zero-initialize. Note, the two step process
7658 of zeroing when in debug mode, thus anything assigned has to
7659 happen after that */
7662 /* On the first pass of the parse, we guess how big this will be. Then
7663 * we grow in one operation to that amount and then give it back. As
7664 * we go along, we re-allocate what we need.
7666 * XXX Currently the guess is essentially that the pattern will be an
7667 * EXACT node with one byte input, one byte output. This is crude, and
7668 * better heuristics are welcome.
7670 * On any subsequent passes, we guess what we actually computed in the
7671 * latest earlier pass. Such a pass probably didn't complete so is
7672 * missing stuff. We could improve those guesses by knowing where the
7673 * parse stopped, and use the length so far plus apply the above
7674 * assumption to what's left. */
7675 RExC_size = STR_SZ(RExC_end - RExC_start);
7678 Newxc(RExC_rxi, sizeof(regexp_internal) + RExC_size, char, regexp_internal);
7679 if ( RExC_rxi == NULL )
7680 FAIL("Regexp out of space");
7682 Zero(RExC_rxi, sizeof(regexp_internal) + RExC_size, char);
7683 RXi_SET( RExC_rx, RExC_rxi );
7685 /* We start from 0 (over from 0 in the case this is a reparse. The first
7686 * node parsed will give back any excess memory we have allocated so far).
7690 /* non-zero initialization begins here */
7691 RExC_rx->engine= eng;
7692 RExC_rx->extflags = rx_flags;
7693 RXp_COMPFLAGS(RExC_rx) = orig_rx_flags & RXf_PMf_FLAGCOPYMASK;
7695 if (pm_flags & PMf_IS_QR) {
7696 RExC_rxi->code_blocks = pRExC_state->code_blocks;
7697 if (RExC_rxi->code_blocks) {
7698 RExC_rxi->code_blocks->refcnt++;
7702 RExC_rx->intflags = 0;
7704 RExC_flags = rx_flags; /* don't let top level (?i) bleed */
7707 /* This NUL is guaranteed because the pattern comes from an SV*, and the sv
7708 * code makes sure the final byte is an uncounted NUL. But should this
7709 * ever not be the case, lots of things could read beyond the end of the
7710 * buffer: loops like
7711 * while(isFOO(*RExC_parse)) RExC_parse++;
7712 * strchr(RExC_parse, "foo");
7713 * etc. So it is worth noting. */
7714 assert(*RExC_end == '\0');
7718 RExC_parens_buf_size = 0;
7719 RExC_emit_start = RExC_rxi->program;
7720 pRExC_state->code_index = 0;
7722 *((char*) RExC_emit_start) = (char) REG_MAGIC;
7726 if (reg(pRExC_state, 0, &flags, 1)) {
7728 /* Success!, But we may need to redo the parse knowing how many parens
7729 * there actually are */
7730 if (IN_PARENS_PASS) {
7731 flags |= RESTART_PARSE;
7734 /* We have that number in RExC_npar */
7735 RExC_total_parens = RExC_npar;
7737 else if (! MUST_RESTART(flags)) {
7739 Perl_croak(aTHX_ "panic: reg returned failure to re_op_compile, flags=%#" UVxf, (UV) flags);
7742 /* Here, we either have success, or we have to redo the parse for some reason */
7743 if (MUST_RESTART(flags)) {
7745 /* It's possible to write a regexp in ascii that represents Unicode
7746 codepoints outside of the byte range, such as via \x{100}. If we
7747 detect such a sequence we have to convert the entire pattern to utf8
7748 and then recompile, as our sizing calculation will have been based
7749 on 1 byte == 1 character, but we will need to use utf8 to encode
7750 at least some part of the pattern, and therefore must convert the whole
7753 if (flags & NEED_UTF8) {
7755 /* We have stored the offset of the final warning output so far.
7756 * That must be adjusted. Any variant characters between the start
7757 * of the pattern and this warning count for 2 bytes in the final,
7758 * so just add them again */
7759 if (UNLIKELY(RExC_latest_warn_offset > 0)) {
7760 RExC_latest_warn_offset +=
7761 variant_under_utf8_count((U8 *) exp, (U8 *) exp
7762 + RExC_latest_warn_offset);
7764 S_pat_upgrade_to_utf8(aTHX_ pRExC_state, &exp, &plen,
7765 pRExC_state->code_blocks ? pRExC_state->code_blocks->count : 0);
7766 DEBUG_PARSE_r(Perl_re_printf( aTHX_ "Need to redo parse after upgrade\n"));
7769 DEBUG_PARSE_r(Perl_re_printf( aTHX_ "Need to redo parse\n"));
7772 if (ALL_PARENS_COUNTED) {
7773 /* Make enough room for all the known parens, and zero it */
7774 Renew(RExC_open_parens, RExC_total_parens, regnode_offset);
7775 Zero(RExC_open_parens, RExC_total_parens, regnode_offset);
7776 RExC_open_parens[0] = 1; /* +1 for REG_MAGIC */
7778 Renew(RExC_close_parens, RExC_total_parens, regnode_offset);
7779 Zero(RExC_close_parens, RExC_total_parens, regnode_offset);
7781 else { /* Parse did not complete. Reinitialize the parentheses
7783 RExC_total_parens = 0;
7784 if (RExC_open_parens) {
7785 Safefree(RExC_open_parens);
7786 RExC_open_parens = NULL;
7788 if (RExC_close_parens) {
7789 Safefree(RExC_close_parens);
7790 RExC_close_parens = NULL;
7794 /* Clean up what we did in this parse */
7795 SvREFCNT_dec_NN(RExC_rx_sv);
7800 /* Here, we have successfully parsed and generated the pattern's program
7801 * for the regex engine. We are ready to finish things up and look for
7804 /* Update the string to compile, with correct modifiers, etc */
7805 set_regex_pv(pRExC_state, Rx);
7807 RExC_rx->nparens = RExC_total_parens - 1;
7809 /* Uses the upper 4 bits of the FLAGS field, so keep within that size */
7810 if (RExC_whilem_seen > 15)
7811 RExC_whilem_seen = 15;
7814 Perl_re_printf( aTHX_
7815 "Required size %" IVdf " nodes\n", (IV)RExC_size);
7817 RExC_lastparse=NULL;
7820 #ifdef RE_TRACK_PATTERN_OFFSETS
7821 DEBUG_OFFSETS_r(Perl_re_printf( aTHX_
7822 "%s %" UVuf " bytes for offset annotations.\n",
7823 RExC_offsets ? "Got" : "Couldn't get",
7824 (UV)((RExC_offsets[0] * 2 + 1))));
7825 DEBUG_OFFSETS_r(if (RExC_offsets) {
7826 const STRLEN len = RExC_offsets[0];
7828 GET_RE_DEBUG_FLAGS_DECL;
7829 Perl_re_printf( aTHX_
7830 "Offsets: [%" UVuf "]\n\t", (UV)RExC_offsets[0]);
7831 for (i = 1; i <= len; i++) {
7832 if (RExC_offsets[i*2-1] || RExC_offsets[i*2])
7833 Perl_re_printf( aTHX_ "%" UVuf ":%" UVuf "[%" UVuf "] ",
7834 (UV)i, (UV)RExC_offsets[i*2-1], (UV)RExC_offsets[i*2]);
7836 Perl_re_printf( aTHX_ "\n");
7840 SetProgLen(RExC_rxi,RExC_size);
7843 DEBUG_DUMP_PRE_OPTIMIZE_r({
7844 SV * const sv = sv_newmortal();
7845 RXi_GET_DECL(RExC_rx, ri);
7847 Perl_re_printf( aTHX_ "Program before optimization:\n");
7849 (void)dumpuntil(RExC_rx, ri->program, ri->program + 1, NULL, NULL,
7854 Perl_re_printf( aTHX_ "Starting post parse optimization\n");
7857 /* XXXX To minimize changes to RE engine we always allocate
7858 3-units-long substrs field. */
7859 Newx(RExC_rx->substrs, 1, struct reg_substr_data);
7860 if (RExC_recurse_count) {
7861 Newx(RExC_recurse, RExC_recurse_count, regnode *);
7862 SAVEFREEPV(RExC_recurse);
7865 if (RExC_seen & REG_RECURSE_SEEN) {
7866 /* Note, RExC_total_parens is 1 + the number of parens in a pattern.
7867 * So its 1 if there are no parens. */
7868 RExC_study_chunk_recursed_bytes= (RExC_total_parens >> 3) +
7869 ((RExC_total_parens & 0x07) != 0);
7870 Newx(RExC_study_chunk_recursed,
7871 RExC_study_chunk_recursed_bytes * RExC_total_parens, U8);
7872 SAVEFREEPV(RExC_study_chunk_recursed);
7876 RExC_rx->minlen = minlen = sawlookahead = sawplus = sawopen = sawminmod = 0;
7878 RExC_study_chunk_recursed_count= 0;
7880 Zero(RExC_rx->substrs, 1, struct reg_substr_data);
7881 if (RExC_study_chunk_recursed) {
7882 Zero(RExC_study_chunk_recursed,
7883 RExC_study_chunk_recursed_bytes * RExC_total_parens, U8);
7887 #ifdef TRIE_STUDY_OPT
7889 StructCopy(&zero_scan_data, &data, scan_data_t);
7890 copyRExC_state = RExC_state;
7893 DEBUG_OPTIMISE_r(Perl_re_printf( aTHX_ "Restudying\n"));
7895 RExC_state = copyRExC_state;
7896 if (seen & REG_TOP_LEVEL_BRANCHES_SEEN)
7897 RExC_seen |= REG_TOP_LEVEL_BRANCHES_SEEN;
7899 RExC_seen &= ~REG_TOP_LEVEL_BRANCHES_SEEN;
7900 StructCopy(&zero_scan_data, &data, scan_data_t);
7903 StructCopy(&zero_scan_data, &data, scan_data_t);
7906 /* Dig out information for optimizations. */
7907 RExC_rx->extflags = RExC_flags; /* was pm_op */
7908 /*dmq: removed as part of de-PMOP: pm->op_pmflags = RExC_flags; */
7911 SvUTF8_on(Rx); /* Unicode in it? */
7912 RExC_rxi->regstclass = NULL;
7913 if (RExC_naughty >= TOO_NAUGHTY) /* Probably an expensive pattern. */
7914 RExC_rx->intflags |= PREGf_NAUGHTY;
7915 scan = RExC_rxi->program + 1; /* First BRANCH. */
7917 /* testing for BRANCH here tells us whether there is "must appear"
7918 data in the pattern. If there is then we can use it for optimisations */
7919 if (!(RExC_seen & REG_TOP_LEVEL_BRANCHES_SEEN)) { /* Only one top-level choice.
7922 STRLEN longest_length[2];
7923 regnode_ssc ch_class; /* pointed to by data */
7925 SSize_t last_close = 0; /* pointed to by data */
7926 regnode *first= scan;
7927 regnode *first_next= regnext(first);
7931 * Skip introductions and multiplicators >= 1
7932 * so that we can extract the 'meat' of the pattern that must
7933 * match in the large if() sequence following.
7934 * NOTE that EXACT is NOT covered here, as it is normally
7935 * picked up by the optimiser separately.
7937 * This is unfortunate as the optimiser isnt handling lookahead
7938 * properly currently.
7941 while ((OP(first) == OPEN && (sawopen = 1)) ||
7942 /* An OR of *one* alternative - should not happen now. */
7943 (OP(first) == BRANCH && OP(first_next) != BRANCH) ||
7944 /* for now we can't handle lookbehind IFMATCH*/
7945 (OP(first) == IFMATCH && !first->flags && (sawlookahead = 1)) ||
7946 (OP(first) == PLUS) ||
7947 (OP(first) == MINMOD) ||
7948 /* An {n,m} with n>0 */
7949 (PL_regkind[OP(first)] == CURLY && ARG1(first) > 0) ||
7950 (OP(first) == NOTHING && PL_regkind[OP(first_next)] != END ))
7953 * the only op that could be a regnode is PLUS, all the rest
7954 * will be regnode_1 or regnode_2.
7956 * (yves doesn't think this is true)
7958 if (OP(first) == PLUS)
7961 if (OP(first) == MINMOD)
7963 first += regarglen[OP(first)];
7965 first = NEXTOPER(first);
7966 first_next= regnext(first);
7969 /* Starting-point info. */
7971 DEBUG_PEEP("first:", first, 0, 0);
7972 /* Ignore EXACT as we deal with it later. */
7973 if (PL_regkind[OP(first)] == EXACT) {
7974 if ( OP(first) == EXACT
7975 || OP(first) == EXACT_ONLY8
7976 || OP(first) == EXACTL)
7978 NOOP; /* Empty, get anchored substr later. */
7981 RExC_rxi->regstclass = first;
7984 else if (PL_regkind[OP(first)] == TRIE &&
7985 ((reg_trie_data *)RExC_rxi->data->data[ ARG(first) ])->minlen>0)
7987 /* this can happen only on restudy */
7988 RExC_rxi->regstclass = construct_ahocorasick_from_trie(pRExC_state, (regnode *)first, 0);
7991 else if (REGNODE_SIMPLE(OP(first)))
7992 RExC_rxi->regstclass = first;
7993 else if (PL_regkind[OP(first)] == BOUND ||
7994 PL_regkind[OP(first)] == NBOUND)
7995 RExC_rxi->regstclass = first;
7996 else if (PL_regkind[OP(first)] == BOL) {
7997 RExC_rx->intflags |= (OP(first) == MBOL
8000 first = NEXTOPER(first);
8003 else if (OP(first) == GPOS) {
8004 RExC_rx->intflags |= PREGf_ANCH_GPOS;
8005 first = NEXTOPER(first);
8008 else if ((!sawopen || !RExC_sawback) &&
8010 (OP(first) == STAR &&
8011 PL_regkind[OP(NEXTOPER(first))] == REG_ANY) &&
8012 !(RExC_rx->intflags & PREGf_ANCH) && !pRExC_state->code_blocks)
8014 /* turn .* into ^.* with an implied $*=1 */
8016 (OP(NEXTOPER(first)) == REG_ANY)
8019 RExC_rx->intflags |= (type | PREGf_IMPLICIT);
8020 first = NEXTOPER(first);
8023 if (sawplus && !sawminmod && !sawlookahead
8024 && (!sawopen || !RExC_sawback)
8025 && !pRExC_state->code_blocks) /* May examine pos and $& */
8026 /* x+ must match at the 1st pos of run of x's */
8027 RExC_rx->intflags |= PREGf_SKIP;
8029 /* Scan is after the zeroth branch, first is atomic matcher. */
8030 #ifdef TRIE_STUDY_OPT
8033 Perl_re_printf( aTHX_ "first at %" IVdf "\n",
8034 (IV)(first - scan + 1))
8038 Perl_re_printf( aTHX_ "first at %" IVdf "\n",
8039 (IV)(first - scan + 1))
8045 * If there's something expensive in the r.e., find the
8046 * longest literal string that must appear and make it the
8047 * regmust. Resolve ties in favor of later strings, since
8048 * the regstart check works with the beginning of the r.e.
8049 * and avoiding duplication strengthens checking. Not a
8050 * strong reason, but sufficient in the absence of others.
8051 * [Now we resolve ties in favor of the earlier string if
8052 * it happens that c_offset_min has been invalidated, since the
8053 * earlier string may buy us something the later one won't.]
8056 data.substrs[0].str = newSVpvs("");
8057 data.substrs[1].str = newSVpvs("");
8058 data.last_found = newSVpvs("");
8059 data.cur_is_floating = 0; /* initially any found substring is fixed */
8060 ENTER_with_name("study_chunk");
8061 SAVEFREESV(data.substrs[0].str);
8062 SAVEFREESV(data.substrs[1].str);
8063 SAVEFREESV(data.last_found);
8065 if (!RExC_rxi->regstclass) {
8066 ssc_init(pRExC_state, &ch_class);
8067 data.start_class = &ch_class;
8068 stclass_flag = SCF_DO_STCLASS_AND;
8069 } else /* XXXX Check for BOUND? */
8071 data.last_closep = &last_close;
8075 * MAIN ENTRY FOR study_chunk() FOR m/PATTERN/
8076 * (NO top level branches)
8078 minlen = study_chunk(pRExC_state, &first, &minlen, &fake,
8079 scan + RExC_size, /* Up to end */
8081 SCF_DO_SUBSTR | SCF_WHILEM_VISITED_POS | stclass_flag
8082 | (restudied ? SCF_TRIE_DOING_RESTUDY : 0),
8086 CHECK_RESTUDY_GOTO_butfirst(LEAVE_with_name("study_chunk"));
8089 if ( RExC_total_parens == 1 && !data.cur_is_floating
8090 && data.last_start_min == 0 && data.last_end > 0
8091 && !RExC_seen_zerolen
8092 && !(RExC_seen & REG_VERBARG_SEEN)
8093 && !(RExC_seen & REG_GPOS_SEEN)
8095 RExC_rx->extflags |= RXf_CHECK_ALL;
8097 scan_commit(pRExC_state, &data,&minlen, 0);
8100 /* XXX this is done in reverse order because that's the way the
8101 * code was before it was parameterised. Don't know whether it
8102 * actually needs doing in reverse order. DAPM */
8103 for (i = 1; i >= 0; i--) {
8104 longest_length[i] = CHR_SVLEN(data.substrs[i].str);
8107 && SvCUR(data.substrs[0].str) /* ok to leave SvCUR */
8108 && data.substrs[0].min_offset
8109 == data.substrs[1].min_offset
8110 && SvCUR(data.substrs[0].str)
8111 == SvCUR(data.substrs[1].str)
8113 && S_setup_longest (aTHX_ pRExC_state,
8114 &(RExC_rx->substrs->data[i]),
8118 RExC_rx->substrs->data[i].min_offset =
8119 data.substrs[i].min_offset - data.substrs[i].lookbehind;
8121 RExC_rx->substrs->data[i].max_offset = data.substrs[i].max_offset;
8122 /* Don't offset infinity */
8123 if (data.substrs[i].max_offset < SSize_t_MAX)
8124 RExC_rx->substrs->data[i].max_offset -= data.substrs[i].lookbehind;
8125 SvREFCNT_inc_simple_void_NN(data.substrs[i].str);
8128 RExC_rx->substrs->data[i].substr = NULL;
8129 RExC_rx->substrs->data[i].utf8_substr = NULL;
8130 longest_length[i] = 0;
8134 LEAVE_with_name("study_chunk");
8136 if (RExC_rxi->regstclass
8137 && (OP(RExC_rxi->regstclass) == REG_ANY || OP(RExC_rxi->regstclass) == SANY))
8138 RExC_rxi->regstclass = NULL;
8140 if ((!(RExC_rx->substrs->data[0].substr || RExC_rx->substrs->data[0].utf8_substr)
8141 || RExC_rx->substrs->data[0].min_offset)
8143 && ! (ANYOF_FLAGS(data.start_class) & SSC_MATCHES_EMPTY_STRING)
8144 && is_ssc_worth_it(pRExC_state, data.start_class))
8146 const U32 n = add_data(pRExC_state, STR_WITH_LEN("f"));
8148 ssc_finalize(pRExC_state, data.start_class);
8150 Newx(RExC_rxi->data->data[n], 1, regnode_ssc);
8151 StructCopy(data.start_class,
8152 (regnode_ssc*)RExC_rxi->data->data[n],
8154 RExC_rxi->regstclass = (regnode*)RExC_rxi->data->data[n];
8155 RExC_rx->intflags &= ~PREGf_SKIP; /* Used in find_byclass(). */
8156 DEBUG_COMPILE_r({ SV *sv = sv_newmortal();
8157 regprop(RExC_rx, sv, (regnode*)data.start_class, NULL, pRExC_state);
8158 Perl_re_printf( aTHX_
8159 "synthetic stclass \"%s\".\n",
8160 SvPVX_const(sv));});
8161 data.start_class = NULL;
8164 /* A temporary algorithm prefers floated substr to fixed one of
8165 * same length to dig more info. */
8166 i = (longest_length[0] <= longest_length[1]);
8167 RExC_rx->substrs->check_ix = i;
8168 RExC_rx->check_end_shift = RExC_rx->substrs->data[i].end_shift;
8169 RExC_rx->check_substr = RExC_rx->substrs->data[i].substr;
8170 RExC_rx->check_utf8 = RExC_rx->substrs->data[i].utf8_substr;
8171 RExC_rx->check_offset_min = RExC_rx->substrs->data[i].min_offset;
8172 RExC_rx->check_offset_max = RExC_rx->substrs->data[i].max_offset;
8173 if (!i && (RExC_rx->intflags & (PREGf_ANCH_SBOL|PREGf_ANCH_GPOS)))
8174 RExC_rx->intflags |= PREGf_NOSCAN;
8176 if ((RExC_rx->check_substr || RExC_rx->check_utf8) ) {
8177 RExC_rx->extflags |= RXf_USE_INTUIT;
8178 if (SvTAIL(RExC_rx->check_substr ? RExC_rx->check_substr : RExC_rx->check_utf8))
8179 RExC_rx->extflags |= RXf_INTUIT_TAIL;
8182 /* XXX Unneeded? dmq (shouldn't as this is handled elsewhere)
8183 if ( (STRLEN)minlen < longest_length[1] )
8184 minlen= longest_length[1];
8185 if ( (STRLEN)minlen < longest_length[0] )
8186 minlen= longest_length[0];
8190 /* Several toplevels. Best we can is to set minlen. */
8192 regnode_ssc ch_class;
8193 SSize_t last_close = 0;
8195 DEBUG_PARSE_r(Perl_re_printf( aTHX_ "\nMulti Top Level\n"));
8197 scan = RExC_rxi->program + 1;
8198 ssc_init(pRExC_state, &ch_class);
8199 data.start_class = &ch_class;
8200 data.last_closep = &last_close;
8204 * MAIN ENTRY FOR study_chunk() FOR m/P1|P2|.../
8205 * (patterns WITH top level branches)
8207 minlen = study_chunk(pRExC_state,
8208 &scan, &minlen, &fake, scan + RExC_size, &data, -1, 0, NULL,
8209 SCF_DO_STCLASS_AND|SCF_WHILEM_VISITED_POS|(restudied
8210 ? SCF_TRIE_DOING_RESTUDY
8214 CHECK_RESTUDY_GOTO_butfirst(NOOP);
8216 RExC_rx->check_substr = NULL;
8217 RExC_rx->check_utf8 = NULL;
8218 RExC_rx->substrs->data[0].substr = NULL;
8219 RExC_rx->substrs->data[0].utf8_substr = NULL;
8220 RExC_rx->substrs->data[1].substr = NULL;
8221 RExC_rx->substrs->data[1].utf8_substr = NULL;
8223 if (! (ANYOF_FLAGS(data.start_class) & SSC_MATCHES_EMPTY_STRING)
8224 && is_ssc_worth_it(pRExC_state, data.start_class))
8226 const U32 n = add_data(pRExC_state, STR_WITH_LEN("f"));
8228 ssc_finalize(pRExC_state, data.start_class);
8230 Newx(RExC_rxi->data->data[n], 1, regnode_ssc);
8231 StructCopy(data.start_class,
8232 (regnode_ssc*)RExC_rxi->data->data[n],
8234 RExC_rxi->regstclass = (regnode*)RExC_rxi->data->data[n];
8235 RExC_rx->intflags &= ~PREGf_SKIP; /* Used in find_byclass(). */
8236 DEBUG_COMPILE_r({ SV* sv = sv_newmortal();
8237 regprop(RExC_rx, sv, (regnode*)data.start_class, NULL, pRExC_state);
8238 Perl_re_printf( aTHX_
8239 "synthetic stclass \"%s\".\n",
8240 SvPVX_const(sv));});
8241 data.start_class = NULL;
8245 if (RExC_seen & REG_UNBOUNDED_QUANTIFIER_SEEN) {
8246 RExC_rx->extflags |= RXf_UNBOUNDED_QUANTIFIER_SEEN;
8247 RExC_rx->maxlen = REG_INFTY;
8250 RExC_rx->maxlen = RExC_maxlen;
8253 /* Guard against an embedded (?=) or (?<=) with a longer minlen than
8254 the "real" pattern. */
8256 Perl_re_printf( aTHX_ "minlen: %" IVdf " RExC_rx->minlen:%" IVdf " maxlen:%" IVdf "\n",
8257 (IV)minlen, (IV)RExC_rx->minlen, (IV)RExC_maxlen);
8259 RExC_rx->minlenret = minlen;
8260 if (RExC_rx->minlen < minlen)
8261 RExC_rx->minlen = minlen;
8263 if (RExC_seen & REG_RECURSE_SEEN ) {
8264 RExC_rx->intflags |= PREGf_RECURSE_SEEN;
8265 Newx(RExC_rx->recurse_locinput, RExC_rx->nparens + 1, char *);
8267 if (RExC_seen & REG_GPOS_SEEN)
8268 RExC_rx->intflags |= PREGf_GPOS_SEEN;
8269 if (RExC_seen & REG_LOOKBEHIND_SEEN)
8270 RExC_rx->extflags |= RXf_NO_INPLACE_SUBST; /* inplace might break the
8272 if (pRExC_state->code_blocks)
8273 RExC_rx->extflags |= RXf_EVAL_SEEN;
8274 if (RExC_seen & REG_VERBARG_SEEN)
8276 RExC_rx->intflags |= PREGf_VERBARG_SEEN;
8277 RExC_rx->extflags |= RXf_NO_INPLACE_SUBST; /* don't understand this! Yves */
8279 if (RExC_seen & REG_CUTGROUP_SEEN)
8280 RExC_rx->intflags |= PREGf_CUTGROUP_SEEN;
8281 if (pm_flags & PMf_USE_RE_EVAL)
8282 RExC_rx->intflags |= PREGf_USE_RE_EVAL;
8283 if (RExC_paren_names)
8284 RXp_PAREN_NAMES(RExC_rx) = MUTABLE_HV(SvREFCNT_inc(RExC_paren_names));
8286 RXp_PAREN_NAMES(RExC_rx) = NULL;
8288 /* If we have seen an anchor in our pattern then we set the extflag RXf_IS_ANCHORED
8289 * so it can be used in pp.c */
8290 if (RExC_rx->intflags & PREGf_ANCH)
8291 RExC_rx->extflags |= RXf_IS_ANCHORED;
8295 /* this is used to identify "special" patterns that might result
8296 * in Perl NOT calling the regex engine and instead doing the match "itself",
8297 * particularly special cases in split//. By having the regex compiler
8298 * do this pattern matching at a regop level (instead of by inspecting the pattern)
8299 * we avoid weird issues with equivalent patterns resulting in different behavior,
8300 * AND we allow non Perl engines to get the same optimizations by the setting the
8301 * flags appropriately - Yves */
8302 regnode *first = RExC_rxi->program + 1;
8304 regnode *next = regnext(first);
8307 if (PL_regkind[fop] == NOTHING && nop == END)
8308 RExC_rx->extflags |= RXf_NULL;
8309 else if ((fop == MBOL || (fop == SBOL && !first->flags)) && nop == END)
8310 /* when fop is SBOL first->flags will be true only when it was
8311 * produced by parsing /\A/, and not when parsing /^/. This is
8312 * very important for the split code as there we want to
8313 * treat /^/ as /^/m, but we do not want to treat /\A/ as /^/m.
8314 * See rt #122761 for more details. -- Yves */
8315 RExC_rx->extflags |= RXf_START_ONLY;
8316 else if (fop == PLUS
8317 && PL_regkind[nop] == POSIXD && FLAGS(next) == _CC_SPACE
8319 RExC_rx->extflags |= RXf_WHITE;
8320 else if ( RExC_rx->extflags & RXf_SPLIT
8321 && (fop == EXACT || fop == EXACT_ONLY8 || fop == EXACTL)
8322 && STR_LEN(first) == 1
8323 && *(STRING(first)) == ' '
8325 RExC_rx->extflags |= (RXf_SKIPWHITE|RXf_WHITE);
8329 if (RExC_contains_locale) {
8330 RXp_EXTFLAGS(RExC_rx) |= RXf_TAINTED;
8334 if (RExC_paren_names) {
8335 RExC_rxi->name_list_idx = add_data( pRExC_state, STR_WITH_LEN("a"));
8336 RExC_rxi->data->data[RExC_rxi->name_list_idx]
8337 = (void*)SvREFCNT_inc(RExC_paren_name_list);
8340 RExC_rxi->name_list_idx = 0;
8342 while ( RExC_recurse_count > 0 ) {
8343 const regnode *scan = RExC_recurse[ --RExC_recurse_count ];
8345 * This data structure is set up in study_chunk() and is used
8346 * to calculate the distance between a GOSUB regopcode and
8347 * the OPEN/CURLYM (CURLYM's are special and can act like OPEN's)
8350 * If for some reason someone writes code that optimises
8351 * away a GOSUB opcode then the assert should be changed to
8352 * an if(scan) to guard the ARG2L_SET() - Yves
8355 assert(scan && OP(scan) == GOSUB);
8356 ARG2L_SET( scan, RExC_open_parens[ARG(scan)] - REGNODE_OFFSET(scan));
8359 Newxz(RExC_rx->offs, RExC_total_parens, regexp_paren_pair);
8360 /* assume we don't need to swap parens around before we match */
8362 Perl_re_printf( aTHX_ "study_chunk_recursed_count: %lu\n",
8363 (unsigned long)RExC_study_chunk_recursed_count);
8367 Perl_re_printf( aTHX_ "Final program:\n");
8371 if (RExC_open_parens) {
8372 Safefree(RExC_open_parens);
8373 RExC_open_parens = NULL;
8375 if (RExC_close_parens) {
8376 Safefree(RExC_close_parens);
8377 RExC_close_parens = NULL;
8381 /* under ithreads the ?pat? PMf_USED flag on the pmop is simulated
8382 * by setting the regexp SV to readonly-only instead. If the
8383 * pattern's been recompiled, the USEDness should remain. */
8384 if (old_re && SvREADONLY(old_re))
8392 Perl_reg_named_buff(pTHX_ REGEXP * const rx, SV * const key, SV * const value,
8395 PERL_ARGS_ASSERT_REG_NAMED_BUFF;
8397 PERL_UNUSED_ARG(value);
8399 if (flags & RXapif_FETCH) {
8400 return reg_named_buff_fetch(rx, key, flags);
8401 } else if (flags & (RXapif_STORE | RXapif_DELETE | RXapif_CLEAR)) {
8402 Perl_croak_no_modify();
8404 } else if (flags & RXapif_EXISTS) {
8405 return reg_named_buff_exists(rx, key, flags)
8408 } else if (flags & RXapif_REGNAMES) {
8409 return reg_named_buff_all(rx, flags);
8410 } else if (flags & (RXapif_SCALAR | RXapif_REGNAMES_COUNT)) {
8411 return reg_named_buff_scalar(rx, flags);
8413 Perl_croak(aTHX_ "panic: Unknown flags %d in named_buff", (int)flags);
8419 Perl_reg_named_buff_iter(pTHX_ REGEXP * const rx, const SV * const lastkey,
8422 PERL_ARGS_ASSERT_REG_NAMED_BUFF_ITER;
8423 PERL_UNUSED_ARG(lastkey);
8425 if (flags & RXapif_FIRSTKEY)
8426 return reg_named_buff_firstkey(rx, flags);
8427 else if (flags & RXapif_NEXTKEY)
8428 return reg_named_buff_nextkey(rx, flags);
8430 Perl_croak(aTHX_ "panic: Unknown flags %d in named_buff_iter",
8437 Perl_reg_named_buff_fetch(pTHX_ REGEXP * const r, SV * const namesv,
8441 struct regexp *const rx = ReANY(r);
8443 PERL_ARGS_ASSERT_REG_NAMED_BUFF_FETCH;
8445 if (rx && RXp_PAREN_NAMES(rx)) {
8446 HE *he_str = hv_fetch_ent( RXp_PAREN_NAMES(rx), namesv, 0, 0 );
8449 SV* sv_dat=HeVAL(he_str);
8450 I32 *nums=(I32*)SvPVX(sv_dat);
8451 AV * const retarray = (flags & RXapif_ALL) ? newAV() : NULL;
8452 for ( i=0; i<SvIVX(sv_dat); i++ ) {
8453 if ((I32)(rx->nparens) >= nums[i]
8454 && rx->offs[nums[i]].start != -1
8455 && rx->offs[nums[i]].end != -1)
8458 CALLREG_NUMBUF_FETCH(r, nums[i], ret);
8463 ret = newSVsv(&PL_sv_undef);
8466 av_push(retarray, ret);
8469 return newRV_noinc(MUTABLE_SV(retarray));
8476 Perl_reg_named_buff_exists(pTHX_ REGEXP * const r, SV * const key,
8479 struct regexp *const rx = ReANY(r);
8481 PERL_ARGS_ASSERT_REG_NAMED_BUFF_EXISTS;
8483 if (rx && RXp_PAREN_NAMES(rx)) {
8484 if (flags & RXapif_ALL) {
8485 return hv_exists_ent(RXp_PAREN_NAMES(rx), key, 0);
8487 SV *sv = CALLREG_NAMED_BUFF_FETCH(r, key, flags);
8489 SvREFCNT_dec_NN(sv);
8501 Perl_reg_named_buff_firstkey(pTHX_ REGEXP * const r, const U32 flags)
8503 struct regexp *const rx = ReANY(r);
8505 PERL_ARGS_ASSERT_REG_NAMED_BUFF_FIRSTKEY;
8507 if ( rx && RXp_PAREN_NAMES(rx) ) {
8508 (void)hv_iterinit(RXp_PAREN_NAMES(rx));
8510 return CALLREG_NAMED_BUFF_NEXTKEY(r, NULL, flags & ~RXapif_FIRSTKEY);
8517 Perl_reg_named_buff_nextkey(pTHX_ REGEXP * const r, const U32 flags)
8519 struct regexp *const rx = ReANY(r);
8520 GET_RE_DEBUG_FLAGS_DECL;
8522 PERL_ARGS_ASSERT_REG_NAMED_BUFF_NEXTKEY;
8524 if (rx && RXp_PAREN_NAMES(rx)) {
8525 HV *hv = RXp_PAREN_NAMES(rx);
8527 while ( (temphe = hv_iternext_flags(hv, 0)) ) {
8530 SV* sv_dat = HeVAL(temphe);
8531 I32 *nums = (I32*)SvPVX(sv_dat);
8532 for ( i = 0; i < SvIVX(sv_dat); i++ ) {
8533 if ((I32)(rx->lastparen) >= nums[i] &&
8534 rx->offs[nums[i]].start != -1 &&
8535 rx->offs[nums[i]].end != -1)
8541 if (parno || flags & RXapif_ALL) {
8542 return newSVhek(HeKEY_hek(temphe));
8550 Perl_reg_named_buff_scalar(pTHX_ REGEXP * const r, const U32 flags)
8555 struct regexp *const rx = ReANY(r);
8557 PERL_ARGS_ASSERT_REG_NAMED_BUFF_SCALAR;
8559 if (rx && RXp_PAREN_NAMES(rx)) {
8560 if (flags & (RXapif_ALL | RXapif_REGNAMES_COUNT)) {
8561 return newSViv(HvTOTALKEYS(RXp_PAREN_NAMES(rx)));
8562 } else if (flags & RXapif_ONE) {
8563 ret = CALLREG_NAMED_BUFF_ALL(r, (flags | RXapif_REGNAMES));
8564 av = MUTABLE_AV(SvRV(ret));
8565 length = av_tindex(av);
8566 SvREFCNT_dec_NN(ret);
8567 return newSViv(length + 1);
8569 Perl_croak(aTHX_ "panic: Unknown flags %d in named_buff_scalar",
8574 return &PL_sv_undef;
8578 Perl_reg_named_buff_all(pTHX_ REGEXP * const r, const U32 flags)
8580 struct regexp *const rx = ReANY(r);
8583 PERL_ARGS_ASSERT_REG_NAMED_BUFF_ALL;
8585 if (rx && RXp_PAREN_NAMES(rx)) {
8586 HV *hv= RXp_PAREN_NAMES(rx);
8588 (void)hv_iterinit(hv);
8589 while ( (temphe = hv_iternext_flags(hv, 0)) ) {
8592 SV* sv_dat = HeVAL(temphe);
8593 I32 *nums = (I32*)SvPVX(sv_dat);
8594 for ( i = 0; i < SvIVX(sv_dat); i++ ) {
8595 if ((I32)(rx->lastparen) >= nums[i] &&
8596 rx->offs[nums[i]].start != -1 &&
8597 rx->offs[nums[i]].end != -1)
8603 if (parno || flags & RXapif_ALL) {
8604 av_push(av, newSVhek(HeKEY_hek(temphe)));
8609 return newRV_noinc(MUTABLE_SV(av));
8613 Perl_reg_numbered_buff_fetch(pTHX_ REGEXP * const r, const I32 paren,
8616 struct regexp *const rx = ReANY(r);
8622 PERL_ARGS_ASSERT_REG_NUMBERED_BUFF_FETCH;
8624 if ( n == RX_BUFF_IDX_CARET_PREMATCH
8625 || n == RX_BUFF_IDX_CARET_FULLMATCH
8626 || n == RX_BUFF_IDX_CARET_POSTMATCH
8629 bool keepcopy = cBOOL(rx->extflags & RXf_PMf_KEEPCOPY);
8631 /* on something like
8634 * the KEEPCOPY is set on the PMOP rather than the regex */
8635 if (PL_curpm && r == PM_GETRE(PL_curpm))
8636 keepcopy = cBOOL(PL_curpm->op_pmflags & PMf_KEEPCOPY);
8645 if (n == RX_BUFF_IDX_CARET_FULLMATCH)
8646 /* no need to distinguish between them any more */
8647 n = RX_BUFF_IDX_FULLMATCH;
8649 if ((n == RX_BUFF_IDX_PREMATCH || n == RX_BUFF_IDX_CARET_PREMATCH)
8650 && rx->offs[0].start != -1)
8652 /* $`, ${^PREMATCH} */
8653 i = rx->offs[0].start;
8657 if ((n == RX_BUFF_IDX_POSTMATCH || n == RX_BUFF_IDX_CARET_POSTMATCH)
8658 && rx->offs[0].end != -1)
8660 /* $', ${^POSTMATCH} */
8661 s = rx->subbeg - rx->suboffset + rx->offs[0].end;
8662 i = rx->sublen + rx->suboffset - rx->offs[0].end;
8665 if ( 0 <= n && n <= (I32)rx->nparens &&
8666 (s1 = rx->offs[n].start) != -1 &&
8667 (t1 = rx->offs[n].end) != -1)
8669 /* $&, ${^MATCH}, $1 ... */
8671 s = rx->subbeg + s1 - rx->suboffset;
8676 assert(s >= rx->subbeg);
8677 assert((STRLEN)rx->sublen >= (STRLEN)((s - rx->subbeg) + i) );
8679 #ifdef NO_TAINT_SUPPORT
8680 sv_setpvn(sv, s, i);
8682 const int oldtainted = TAINT_get;
8684 sv_setpvn(sv, s, i);
8685 TAINT_set(oldtainted);
8687 if (RXp_MATCH_UTF8(rx))
8692 if (RXp_MATCH_TAINTED(rx)) {
8693 if (SvTYPE(sv) >= SVt_PVMG) {
8694 MAGIC* const mg = SvMAGIC(sv);
8697 SvMAGIC_set(sv, mg->mg_moremagic);
8699 if ((mgt = SvMAGIC(sv))) {
8700 mg->mg_moremagic = mgt;
8701 SvMAGIC_set(sv, mg);
8718 Perl_reg_numbered_buff_store(pTHX_ REGEXP * const rx, const I32 paren,
8719 SV const * const value)
8721 PERL_ARGS_ASSERT_REG_NUMBERED_BUFF_STORE;
8723 PERL_UNUSED_ARG(rx);
8724 PERL_UNUSED_ARG(paren);
8725 PERL_UNUSED_ARG(value);
8728 Perl_croak_no_modify();
8732 Perl_reg_numbered_buff_length(pTHX_ REGEXP * const r, const SV * const sv,
8735 struct regexp *const rx = ReANY(r);
8739 PERL_ARGS_ASSERT_REG_NUMBERED_BUFF_LENGTH;
8741 if ( paren == RX_BUFF_IDX_CARET_PREMATCH
8742 || paren == RX_BUFF_IDX_CARET_FULLMATCH
8743 || paren == RX_BUFF_IDX_CARET_POSTMATCH
8746 bool keepcopy = cBOOL(rx->extflags & RXf_PMf_KEEPCOPY);
8748 /* on something like
8751 * the KEEPCOPY is set on the PMOP rather than the regex */
8752 if (PL_curpm && r == PM_GETRE(PL_curpm))
8753 keepcopy = cBOOL(PL_curpm->op_pmflags & PMf_KEEPCOPY);
8759 /* Some of this code was originally in C<Perl_magic_len> in F<mg.c> */
8761 case RX_BUFF_IDX_CARET_PREMATCH: /* ${^PREMATCH} */
8762 case RX_BUFF_IDX_PREMATCH: /* $` */
8763 if (rx->offs[0].start != -1) {
8764 i = rx->offs[0].start;
8773 case RX_BUFF_IDX_CARET_POSTMATCH: /* ${^POSTMATCH} */
8774 case RX_BUFF_IDX_POSTMATCH: /* $' */
8775 if (rx->offs[0].end != -1) {
8776 i = rx->sublen - rx->offs[0].end;
8778 s1 = rx->offs[0].end;
8785 default: /* $& / ${^MATCH}, $1, $2, ... */
8786 if (paren <= (I32)rx->nparens &&
8787 (s1 = rx->offs[paren].start) != -1 &&
8788 (t1 = rx->offs[paren].end) != -1)
8794 if (ckWARN(WARN_UNINITIALIZED))
8795 report_uninit((const SV *)sv);
8800 if (i > 0 && RXp_MATCH_UTF8(rx)) {
8801 const char * const s = rx->subbeg - rx->suboffset + s1;
8806 if (is_utf8_string_loclen((U8*)s, i, &ep, &el))
8813 Perl_reg_qr_package(pTHX_ REGEXP * const rx)
8815 PERL_ARGS_ASSERT_REG_QR_PACKAGE;
8816 PERL_UNUSED_ARG(rx);
8820 return newSVpvs("Regexp");
8823 /* Scans the name of a named buffer from the pattern.
8824 * If flags is REG_RSN_RETURN_NULL returns null.
8825 * If flags is REG_RSN_RETURN_NAME returns an SV* containing the name
8826 * If flags is REG_RSN_RETURN_DATA returns the data SV* corresponding
8827 * to the parsed name as looked up in the RExC_paren_names hash.
8828 * If there is an error throws a vFAIL().. type exception.
8831 #define REG_RSN_RETURN_NULL 0
8832 #define REG_RSN_RETURN_NAME 1
8833 #define REG_RSN_RETURN_DATA 2
8836 S_reg_scan_name(pTHX_ RExC_state_t *pRExC_state, U32 flags)
8838 char *name_start = RExC_parse;
8841 PERL_ARGS_ASSERT_REG_SCAN_NAME;
8843 assert (RExC_parse <= RExC_end);
8844 if (RExC_parse == RExC_end) NOOP;
8845 else if (isIDFIRST_lazy_if_safe(RExC_parse, RExC_end, UTF)) {
8846 /* Note that the code here assumes well-formed UTF-8. Skip IDFIRST by
8847 * using do...while */
8850 RExC_parse += UTF8SKIP(RExC_parse);
8851 } while ( RExC_parse < RExC_end
8852 && isWORDCHAR_utf8_safe((U8*)RExC_parse, (U8*) RExC_end));
8856 } while (RExC_parse < RExC_end && isWORDCHAR(*RExC_parse));
8858 RExC_parse++; /* so the <- from the vFAIL is after the offending
8860 vFAIL("Group name must start with a non-digit word character");
8862 sv_name = newSVpvn_flags(name_start, (int)(RExC_parse - name_start),
8863 SVs_TEMP | (UTF ? SVf_UTF8 : 0));
8864 if ( flags == REG_RSN_RETURN_NAME)
8866 else if (flags==REG_RSN_RETURN_DATA) {
8869 if ( ! sv_name ) /* should not happen*/
8870 Perl_croak(aTHX_ "panic: no svname in reg_scan_name");
8871 if (RExC_paren_names)
8872 he_str = hv_fetch_ent( RExC_paren_names, sv_name, 0, 0 );
8874 sv_dat = HeVAL(he_str);
8875 if ( ! sv_dat ) { /* Didn't find group */
8877 /* It might be a forward reference; we can't fail until we
8878 * know, by completing the parse to get all the groups, and
8880 if (ALL_PARENS_COUNTED) {
8881 vFAIL("Reference to nonexistent named group");
8884 REQUIRE_PARENS_PASS;
8890 Perl_croak(aTHX_ "panic: bad flag %lx in reg_scan_name",
8891 (unsigned long) flags);
8894 #define DEBUG_PARSE_MSG(funcname) DEBUG_PARSE_r({ \
8895 if (RExC_lastparse!=RExC_parse) { \
8896 Perl_re_printf( aTHX_ "%s", \
8897 Perl_pv_pretty(aTHX_ RExC_mysv1, RExC_parse, \
8898 RExC_end - RExC_parse, 16, \
8900 PERL_PV_ESCAPE_UNI_DETECT | \
8901 PERL_PV_PRETTY_ELLIPSES | \
8902 PERL_PV_PRETTY_LTGT | \
8903 PERL_PV_ESCAPE_RE | \
8904 PERL_PV_PRETTY_EXACTSIZE \
8908 Perl_re_printf( aTHX_ "%16s",""); \
8910 if (RExC_lastnum!=RExC_emit) \
8911 Perl_re_printf( aTHX_ "|%4d", RExC_emit); \
8913 Perl_re_printf( aTHX_ "|%4s",""); \
8914 Perl_re_printf( aTHX_ "|%*s%-4s", \
8915 (int)((depth*2)), "", \
8918 RExC_lastnum=RExC_emit; \
8919 RExC_lastparse=RExC_parse; \
8924 #define DEBUG_PARSE(funcname) DEBUG_PARSE_r({ \
8925 DEBUG_PARSE_MSG((funcname)); \
8926 Perl_re_printf( aTHX_ "%4s","\n"); \
8928 #define DEBUG_PARSE_FMT(funcname,fmt,args) DEBUG_PARSE_r({\
8929 DEBUG_PARSE_MSG((funcname)); \
8930 Perl_re_printf( aTHX_ fmt "\n",args); \
8933 /* This section of code defines the inversion list object and its methods. The
8934 * interfaces are highly subject to change, so as much as possible is static to
8935 * this file. An inversion list is here implemented as a malloc'd C UV array
8936 * as an SVt_INVLIST scalar.
8938 * An inversion list for Unicode is an array of code points, sorted by ordinal
8939 * number. Each element gives the code point that begins a range that extends
8940 * up-to but not including the code point given by the next element. The final
8941 * element gives the first code point of a range that extends to the platform's
8942 * infinity. The even-numbered elements (invlist[0], invlist[2], invlist[4],
8943 * ...) give ranges whose code points are all in the inversion list. We say
8944 * that those ranges are in the set. The odd-numbered elements give ranges
8945 * whose code points are not in the inversion list, and hence not in the set.
8946 * Thus, element [0] is the first code point in the list. Element [1]
8947 * is the first code point beyond that not in the list; and element [2] is the
8948 * first code point beyond that that is in the list. In other words, the first
8949 * range is invlist[0]..(invlist[1]-1), and all code points in that range are
8950 * in the inversion list. The second range is invlist[1]..(invlist[2]-1), and
8951 * all code points in that range are not in the inversion list. The third
8952 * range invlist[2]..(invlist[3]-1) gives code points that are in the inversion
8953 * list, and so forth. Thus every element whose index is divisible by two
8954 * gives the beginning of a range that is in the list, and every element whose
8955 * index is not divisible by two gives the beginning of a range not in the
8956 * list. If the final element's index is divisible by two, the inversion list
8957 * extends to the platform's infinity; otherwise the highest code point in the
8958 * inversion list is the contents of that element minus 1.
8960 * A range that contains just a single code point N will look like
8962 * invlist[i+1] == N+1
8964 * If N is UV_MAX (the highest representable code point on the machine), N+1 is
8965 * impossible to represent, so element [i+1] is omitted. The single element
8967 * invlist[0] == UV_MAX
8968 * contains just UV_MAX, but is interpreted as matching to infinity.
8970 * Taking the complement (inverting) an inversion list is quite simple, if the
8971 * first element is 0, remove it; otherwise add a 0 element at the beginning.
8972 * This implementation reserves an element at the beginning of each inversion
8973 * list to always contain 0; there is an additional flag in the header which
8974 * indicates if the list begins at the 0, or is offset to begin at the next
8975 * element. This means that the inversion list can be inverted without any
8976 * copying; just flip the flag.
8978 * More about inversion lists can be found in "Unicode Demystified"
8979 * Chapter 13 by Richard Gillam, published by Addison-Wesley.
8981 * The inversion list data structure is currently implemented as an SV pointing
8982 * to an array of UVs that the SV thinks are bytes. This allows us to have an
8983 * array of UV whose memory management is automatically handled by the existing
8984 * facilities for SV's.
8986 * Some of the methods should always be private to the implementation, and some
8987 * should eventually be made public */
8989 /* The header definitions are in F<invlist_inline.h> */
8991 #ifndef PERL_IN_XSUB_RE
8993 PERL_STATIC_INLINE UV*
8994 S__invlist_array_init(SV* const invlist, const bool will_have_0)
8996 /* Returns a pointer to the first element in the inversion list's array.
8997 * This is called upon initialization of an inversion list. Where the
8998 * array begins depends on whether the list has the code point U+0000 in it
8999 * or not. The other parameter tells it whether the code that follows this
9000 * call is about to put a 0 in the inversion list or not. The first
9001 * element is either the element reserved for 0, if TRUE, or the element
9002 * after it, if FALSE */
9004 bool* offset = get_invlist_offset_addr(invlist);
9005 UV* zero_addr = (UV *) SvPVX(invlist);
9007 PERL_ARGS_ASSERT__INVLIST_ARRAY_INIT;
9010 assert(! _invlist_len(invlist));
9014 /* 1^1 = 0; 1^0 = 1 */
9015 *offset = 1 ^ will_have_0;
9016 return zero_addr + *offset;
9019 PERL_STATIC_INLINE void
9020 S_invlist_set_len(pTHX_ SV* const invlist, const UV len, const bool offset)
9022 /* Sets the current number of elements stored in the inversion list.
9023 * Updates SvCUR correspondingly */
9024 PERL_UNUSED_CONTEXT;
9025 PERL_ARGS_ASSERT_INVLIST_SET_LEN;
9027 assert(is_invlist(invlist));
9032 : TO_INTERNAL_SIZE(len + offset));
9033 assert(SvLEN(invlist) == 0 || SvCUR(invlist) <= SvLEN(invlist));
9037 S_invlist_replace_list_destroys_src(pTHX_ SV * dest, SV * src)
9039 /* Replaces the inversion list in 'dest' with the one from 'src'. It
9040 * steals the list from 'src', so 'src' is made to have a NULL list. This
9041 * is similar to what SvSetMagicSV() would do, if it were implemented on
9042 * inversion lists, though this routine avoids a copy */
9044 const UV src_len = _invlist_len(src);
9045 const bool src_offset = *get_invlist_offset_addr(src);
9046 const STRLEN src_byte_len = SvLEN(src);
9047 char * array = SvPVX(src);
9049 const int oldtainted = TAINT_get;
9051 PERL_ARGS_ASSERT_INVLIST_REPLACE_LIST_DESTROYS_SRC;
9053 assert(is_invlist(src));
9054 assert(is_invlist(dest));
9055 assert(! invlist_is_iterating(src));
9056 assert(SvCUR(src) == 0 || SvCUR(src) < SvLEN(src));
9058 /* Make sure it ends in the right place with a NUL, as our inversion list
9059 * manipulations aren't careful to keep this true, but sv_usepvn_flags()
9061 array[src_byte_len - 1] = '\0';
9063 TAINT_NOT; /* Otherwise it breaks */
9064 sv_usepvn_flags(dest,
9068 /* This flag is documented to cause a copy to be avoided */
9069 SV_HAS_TRAILING_NUL);
9070 TAINT_set(oldtainted);
9075 /* Finish up copying over the other fields in an inversion list */
9076 *get_invlist_offset_addr(dest) = src_offset;
9077 invlist_set_len(dest, src_len, src_offset);
9078 *get_invlist_previous_index_addr(dest) = 0;
9079 invlist_iterfinish(dest);
9082 PERL_STATIC_INLINE IV*
9083 S_get_invlist_previous_index_addr(SV* invlist)
9085 /* Return the address of the IV that is reserved to hold the cached index
9087 PERL_ARGS_ASSERT_GET_INVLIST_PREVIOUS_INDEX_ADDR;
9089 assert(is_invlist(invlist));
9091 return &(((XINVLIST*) SvANY(invlist))->prev_index);
9094 PERL_STATIC_INLINE IV
9095 S_invlist_previous_index(SV* const invlist)
9097 /* Returns cached index of previous search */
9099 PERL_ARGS_ASSERT_INVLIST_PREVIOUS_INDEX;
9101 return *get_invlist_previous_index_addr(invlist);
9104 PERL_STATIC_INLINE void
9105 S_invlist_set_previous_index(SV* const invlist, const IV index)
9107 /* Caches <index> for later retrieval */
9109 PERL_ARGS_ASSERT_INVLIST_SET_PREVIOUS_INDEX;
9111 assert(index == 0 || index < (int) _invlist_len(invlist));
9113 *get_invlist_previous_index_addr(invlist) = index;
9116 PERL_STATIC_INLINE void
9117 S_invlist_trim(SV* invlist)
9119 /* Free the not currently-being-used space in an inversion list */
9121 /* But don't free up the space needed for the 0 UV that is always at the
9122 * beginning of the list, nor the trailing NUL */
9123 const UV min_size = TO_INTERNAL_SIZE(1) + 1;
9125 PERL_ARGS_ASSERT_INVLIST_TRIM;
9127 assert(is_invlist(invlist));
9129 SvPV_renew(invlist, MAX(min_size, SvCUR(invlist) + 1));
9132 PERL_STATIC_INLINE void
9133 S_invlist_clear(pTHX_ SV* invlist) /* Empty the inversion list */
9135 PERL_ARGS_ASSERT_INVLIST_CLEAR;
9137 assert(is_invlist(invlist));
9139 invlist_set_len(invlist, 0, 0);
9140 invlist_trim(invlist);
9143 #endif /* ifndef PERL_IN_XSUB_RE */
9145 PERL_STATIC_INLINE bool
9146 S_invlist_is_iterating(SV* const invlist)
9148 PERL_ARGS_ASSERT_INVLIST_IS_ITERATING;
9150 return *(get_invlist_iter_addr(invlist)) < (STRLEN) UV_MAX;
9153 #ifndef PERL_IN_XSUB_RE
9155 PERL_STATIC_INLINE UV
9156 S_invlist_max(SV* const invlist)
9158 /* Returns the maximum number of elements storable in the inversion list's
9159 * array, without having to realloc() */
9161 PERL_ARGS_ASSERT_INVLIST_MAX;
9163 assert(is_invlist(invlist));
9165 /* Assumes worst case, in which the 0 element is not counted in the
9166 * inversion list, so subtracts 1 for that */
9167 return SvLEN(invlist) == 0 /* This happens under _new_invlist_C_array */
9168 ? FROM_INTERNAL_SIZE(SvCUR(invlist)) - 1
9169 : FROM_INTERNAL_SIZE(SvLEN(invlist)) - 1;
9173 S_initialize_invlist_guts(pTHX_ SV* invlist, const Size_t initial_size)
9175 PERL_ARGS_ASSERT_INITIALIZE_INVLIST_GUTS;
9177 /* First 1 is in case the zero element isn't in the list; second 1 is for
9179 SvGROW(invlist, TO_INTERNAL_SIZE(initial_size + 1) + 1);
9180 invlist_set_len(invlist, 0, 0);
9182 /* Force iterinit() to be used to get iteration to work */
9183 invlist_iterfinish(invlist);
9185 *get_invlist_previous_index_addr(invlist) = 0;
9189 Perl__new_invlist(pTHX_ IV initial_size)
9192 /* Return a pointer to a newly constructed inversion list, with enough
9193 * space to store 'initial_size' elements. If that number is negative, a
9194 * system default is used instead */
9198 if (initial_size < 0) {
9202 new_list = newSV_type(SVt_INVLIST);
9203 initialize_invlist_guts(new_list, initial_size);
9209 Perl__new_invlist_C_array(pTHX_ const UV* const list)
9211 /* Return a pointer to a newly constructed inversion list, initialized to
9212 * point to <list>, which has to be in the exact correct inversion list
9213 * form, including internal fields. Thus this is a dangerous routine that
9214 * should not be used in the wrong hands. The passed in 'list' contains
9215 * several header fields at the beginning that are not part of the
9216 * inversion list body proper */
9218 const STRLEN length = (STRLEN) list[0];
9219 const UV version_id = list[1];
9220 const bool offset = cBOOL(list[2]);
9221 #define HEADER_LENGTH 3
9222 /* If any of the above changes in any way, you must change HEADER_LENGTH
9223 * (if appropriate) and regenerate INVLIST_VERSION_ID by running
9224 * perl -E 'say int(rand 2**31-1)'
9226 #define INVLIST_VERSION_ID 148565664 /* This is a combination of a version and
9227 data structure type, so that one being
9228 passed in can be validated to be an
9229 inversion list of the correct vintage.
9232 SV* invlist = newSV_type(SVt_INVLIST);
9234 PERL_ARGS_ASSERT__NEW_INVLIST_C_ARRAY;
9236 if (version_id != INVLIST_VERSION_ID) {
9237 Perl_croak(aTHX_ "panic: Incorrect version for previously generated inversion list");
9240 /* The generated array passed in includes header elements that aren't part
9241 * of the list proper, so start it just after them */
9242 SvPV_set(invlist, (char *) (list + HEADER_LENGTH));
9244 SvLEN_set(invlist, 0); /* Means we own the contents, and the system
9245 shouldn't touch it */
9247 *(get_invlist_offset_addr(invlist)) = offset;
9249 /* The 'length' passed to us is the physical number of elements in the
9250 * inversion list. But if there is an offset the logical number is one
9252 invlist_set_len(invlist, length - offset, offset);
9254 invlist_set_previous_index(invlist, 0);
9256 /* Initialize the iteration pointer. */
9257 invlist_iterfinish(invlist);
9259 SvREADONLY_on(invlist);
9265 S_invlist_extend(pTHX_ SV* const invlist, const UV new_max)
9267 /* Grow the maximum size of an inversion list */
9269 PERL_ARGS_ASSERT_INVLIST_EXTEND;
9271 assert(is_invlist(invlist));
9273 /* Add one to account for the zero element at the beginning which may not
9274 * be counted by the calling parameters */
9275 SvGROW((SV *)invlist, TO_INTERNAL_SIZE(new_max + 1));
9279 S__append_range_to_invlist(pTHX_ SV* const invlist,
9280 const UV start, const UV end)
9282 /* Subject to change or removal. Append the range from 'start' to 'end' at
9283 * the end of the inversion list. The range must be above any existing
9287 UV max = invlist_max(invlist);
9288 UV len = _invlist_len(invlist);
9291 PERL_ARGS_ASSERT__APPEND_RANGE_TO_INVLIST;
9293 if (len == 0) { /* Empty lists must be initialized */
9294 offset = start != 0;
9295 array = _invlist_array_init(invlist, ! offset);
9298 /* Here, the existing list is non-empty. The current max entry in the
9299 * list is generally the first value not in the set, except when the
9300 * set extends to the end of permissible values, in which case it is
9301 * the first entry in that final set, and so this call is an attempt to
9302 * append out-of-order */
9304 UV final_element = len - 1;
9305 array = invlist_array(invlist);
9306 if ( array[final_element] > start
9307 || ELEMENT_RANGE_MATCHES_INVLIST(final_element))
9309 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",
9310 array[final_element], start,
9311 ELEMENT_RANGE_MATCHES_INVLIST(final_element) ? 't' : 'f');
9314 /* Here, it is a legal append. If the new range begins 1 above the end
9315 * of the range below it, it is extending the range below it, so the
9316 * new first value not in the set is one greater than the newly
9317 * extended range. */
9318 offset = *get_invlist_offset_addr(invlist);
9319 if (array[final_element] == start) {
9320 if (end != UV_MAX) {
9321 array[final_element] = end + 1;
9324 /* But if the end is the maximum representable on the machine,
9325 * assume that infinity was actually what was meant. Just let
9326 * the range that this would extend to have no end */
9327 invlist_set_len(invlist, len - 1, offset);
9333 /* Here the new range doesn't extend any existing set. Add it */
9335 len += 2; /* Includes an element each for the start and end of range */
9337 /* If wll overflow the existing space, extend, which may cause the array to
9340 invlist_extend(invlist, len);
9342 /* Have to set len here to avoid assert failure in invlist_array() */
9343 invlist_set_len(invlist, len, offset);
9345 array = invlist_array(invlist);
9348 invlist_set_len(invlist, len, offset);
9351 /* The next item on the list starts the range, the one after that is
9352 * one past the new range. */
9353 array[len - 2] = start;
9354 if (end != UV_MAX) {
9355 array[len - 1] = end + 1;
9358 /* But if the end is the maximum representable on the machine, just let
9359 * the range have no end */
9360 invlist_set_len(invlist, len - 1, offset);
9365 Perl__invlist_search(SV* const invlist, const UV cp)
9367 /* Searches the inversion list for the entry that contains the input code
9368 * point <cp>. If <cp> is not in the list, -1 is returned. Otherwise, the
9369 * return value is the index into the list's array of the range that
9370 * contains <cp>, that is, 'i' such that
9371 * array[i] <= cp < array[i+1]
9376 IV high = _invlist_len(invlist);
9377 const IV highest_element = high - 1;
9380 PERL_ARGS_ASSERT__INVLIST_SEARCH;
9382 /* If list is empty, return failure. */
9387 /* (We can't get the array unless we know the list is non-empty) */
9388 array = invlist_array(invlist);
9390 mid = invlist_previous_index(invlist);
9392 if (mid > highest_element) {
9393 mid = highest_element;
9396 /* <mid> contains the cache of the result of the previous call to this
9397 * function (0 the first time). See if this call is for the same result,
9398 * or if it is for mid-1. This is under the theory that calls to this
9399 * function will often be for related code points that are near each other.
9400 * And benchmarks show that caching gives better results. We also test
9401 * here if the code point is within the bounds of the list. These tests
9402 * replace others that would have had to be made anyway to make sure that
9403 * the array bounds were not exceeded, and these give us extra information
9404 * at the same time */
9405 if (cp >= array[mid]) {
9406 if (cp >= array[highest_element]) {
9407 return highest_element;
9410 /* Here, array[mid] <= cp < array[highest_element]. This means that
9411 * the final element is not the answer, so can exclude it; it also
9412 * means that <mid> is not the final element, so can refer to 'mid + 1'
9414 if (cp < array[mid + 1]) {
9420 else { /* cp < aray[mid] */
9421 if (cp < array[0]) { /* Fail if outside the array */
9425 if (cp >= array[mid - 1]) {
9430 /* Binary search. What we are looking for is <i> such that
9431 * array[i] <= cp < array[i+1]
9432 * The loop below converges on the i+1. Note that there may not be an
9433 * (i+1)th element in the array, and things work nonetheless */
9434 while (low < high) {
9435 mid = (low + high) / 2;
9436 assert(mid <= highest_element);
9437 if (array[mid] <= cp) { /* cp >= array[mid] */
9440 /* We could do this extra test to exit the loop early.
9441 if (cp < array[low]) {
9446 else { /* cp < array[mid] */
9453 invlist_set_previous_index(invlist, high);
9458 Perl__invlist_union_maybe_complement_2nd(pTHX_ SV* const a, SV* const b,
9459 const bool complement_b, SV** output)
9461 /* Take the union of two inversion lists and point '*output' to it. On
9462 * input, '*output' MUST POINT TO NULL OR TO AN SV* INVERSION LIST (possibly
9463 * even 'a' or 'b'). If to an inversion list, the contents of the original
9464 * list will be replaced by the union. The first list, 'a', may be
9465 * NULL, in which case a copy of the second list is placed in '*output'.
9466 * If 'complement_b' is TRUE, the union is taken of the complement
9467 * (inversion) of 'b' instead of b itself.
9469 * The basis for this comes from "Unicode Demystified" Chapter 13 by
9470 * Richard Gillam, published by Addison-Wesley, and explained at some
9471 * length there. The preface says to incorporate its examples into your
9472 * code at your own risk.
9474 * The algorithm is like a merge sort. */
9476 const UV* array_a; /* a's array */
9478 UV len_a; /* length of a's array */
9481 SV* u; /* the resulting union */
9485 UV i_a = 0; /* current index into a's array */
9489 /* running count, as explained in the algorithm source book; items are
9490 * stopped accumulating and are output when the count changes to/from 0.
9491 * The count is incremented when we start a range that's in an input's set,
9492 * and decremented when we start a range that's not in a set. So this
9493 * variable can be 0, 1, or 2. When it is 0 neither input is in their set,
9494 * and hence nothing goes into the union; 1, just one of the inputs is in
9495 * its set (and its current range gets added to the union); and 2 when both
9496 * inputs are in their sets. */
9499 PERL_ARGS_ASSERT__INVLIST_UNION_MAYBE_COMPLEMENT_2ND;
9501 assert(*output == NULL || is_invlist(*output));
9503 len_b = _invlist_len(b);
9506 /* Here, 'b' is empty, hence it's complement is all possible code
9507 * points. So if the union includes the complement of 'b', it includes
9508 * everything, and we need not even look at 'a'. It's easiest to
9509 * create a new inversion list that matches everything. */
9511 SV* everything = _add_range_to_invlist(NULL, 0, UV_MAX);
9513 if (*output == NULL) { /* If the output didn't exist, just point it
9515 *output = everything;
9517 else { /* Otherwise, replace its contents with the new list */
9518 invlist_replace_list_destroys_src(*output, everything);
9519 SvREFCNT_dec_NN(everything);
9525 /* Here, we don't want the complement of 'b', and since 'b' is empty,
9526 * the union will come entirely from 'a'. If 'a' is NULL or empty, the
9527 * output will be empty */
9529 if (a == NULL || _invlist_len(a) == 0) {
9530 if (*output == NULL) {
9531 *output = _new_invlist(0);
9534 invlist_clear(*output);
9539 /* Here, 'a' is not empty, but 'b' is, so 'a' entirely determines the
9540 * union. We can just return a copy of 'a' if '*output' doesn't point
9541 * to an existing list */
9542 if (*output == NULL) {
9543 *output = invlist_clone(a, NULL);
9547 /* If the output is to overwrite 'a', we have a no-op, as it's
9553 /* Here, '*output' is to be overwritten by 'a' */
9554 u = invlist_clone(a, NULL);
9555 invlist_replace_list_destroys_src(*output, u);
9561 /* Here 'b' is not empty. See about 'a' */
9563 if (a == NULL || ((len_a = _invlist_len(a)) == 0)) {
9565 /* Here, 'a' is empty (and b is not). That means the union will come
9566 * entirely from 'b'. If '*output' is NULL, we can directly return a
9567 * clone of 'b'. Otherwise, we replace the contents of '*output' with
9570 SV ** dest = (*output == NULL) ? output : &u;
9571 *dest = invlist_clone(b, NULL);
9573 _invlist_invert(*dest);
9577 invlist_replace_list_destroys_src(*output, u);
9584 /* Here both lists exist and are non-empty */
9585 array_a = invlist_array(a);
9586 array_b = invlist_array(b);
9588 /* If are to take the union of 'a' with the complement of b, set it
9589 * up so are looking at b's complement. */
9592 /* To complement, we invert: if the first element is 0, remove it. To
9593 * do this, we just pretend the array starts one later */
9594 if (array_b[0] == 0) {
9600 /* But if the first element is not zero, we pretend the list starts
9601 * at the 0 that is always stored immediately before the array. */
9607 /* Size the union for the worst case: that the sets are completely
9609 u = _new_invlist(len_a + len_b);
9611 /* Will contain U+0000 if either component does */
9612 array_u = _invlist_array_init(u, ( len_a > 0 && array_a[0] == 0)
9613 || (len_b > 0 && array_b[0] == 0));
9615 /* Go through each input list item by item, stopping when have exhausted
9617 while (i_a < len_a && i_b < len_b) {
9618 UV cp; /* The element to potentially add to the union's array */
9619 bool cp_in_set; /* is it in the the input list's set or not */
9621 /* We need to take one or the other of the two inputs for the union.
9622 * Since we are merging two sorted lists, we take the smaller of the
9623 * next items. In case of a tie, we take first the one that is in its
9624 * set. If we first took the one not in its set, it would decrement
9625 * the count, possibly to 0 which would cause it to be output as ending
9626 * the range, and the next time through we would take the same number,
9627 * and output it again as beginning the next range. By doing it the
9628 * opposite way, there is no possibility that the count will be
9629 * momentarily decremented to 0, and thus the two adjoining ranges will
9630 * be seamlessly merged. (In a tie and both are in the set or both not
9631 * in the set, it doesn't matter which we take first.) */
9632 if ( array_a[i_a] < array_b[i_b]
9633 || ( array_a[i_a] == array_b[i_b]
9634 && ELEMENT_RANGE_MATCHES_INVLIST(i_a)))
9636 cp_in_set = ELEMENT_RANGE_MATCHES_INVLIST(i_a);
9637 cp = array_a[i_a++];
9640 cp_in_set = ELEMENT_RANGE_MATCHES_INVLIST(i_b);
9641 cp = array_b[i_b++];
9644 /* Here, have chosen which of the two inputs to look at. Only output
9645 * if the running count changes to/from 0, which marks the
9646 * beginning/end of a range that's in the set */
9649 array_u[i_u++] = cp;
9656 array_u[i_u++] = cp;
9662 /* The loop above increments the index into exactly one of the input lists
9663 * each iteration, and ends when either index gets to its list end. That
9664 * means the other index is lower than its end, and so something is
9665 * remaining in that one. We decrement 'count', as explained below, if
9666 * that list is in its set. (i_a and i_b each currently index the element
9667 * beyond the one we care about.) */
9668 if ( (i_a != len_a && PREV_RANGE_MATCHES_INVLIST(i_a))
9669 || (i_b != len_b && PREV_RANGE_MATCHES_INVLIST(i_b)))
9674 /* Above we decremented 'count' if the list that had unexamined elements in
9675 * it was in its set. This has made it so that 'count' being non-zero
9676 * means there isn't anything left to output; and 'count' equal to 0 means
9677 * that what is left to output is precisely that which is left in the
9678 * non-exhausted input list.
9680 * To see why, note first that the exhausted input obviously has nothing
9681 * left to add to the union. If it was in its set at its end, that means
9682 * the set extends from here to the platform's infinity, and hence so does
9683 * the union and the non-exhausted set is irrelevant. The exhausted set
9684 * also contributed 1 to 'count'. If 'count' was 2, it got decremented to
9685 * 1, but if it was 1, the non-exhausted set wasn't in its set, and so
9686 * 'count' remains at 1. This is consistent with the decremented 'count'
9687 * != 0 meaning there's nothing left to add to the union.
9689 * But if the exhausted input wasn't in its set, it contributed 0 to
9690 * 'count', and the rest of the union will be whatever the other input is.
9691 * If 'count' was 0, neither list was in its set, and 'count' remains 0;
9692 * otherwise it gets decremented to 0. This is consistent with 'count'
9693 * == 0 meaning the remainder of the union is whatever is left in the
9694 * non-exhausted list. */
9699 IV copy_count = len_a - i_a;
9700 if (copy_count > 0) { /* The non-exhausted input is 'a' */
9701 Copy(array_a + i_a, array_u + i_u, copy_count, UV);
9703 else { /* The non-exhausted input is b */
9704 copy_count = len_b - i_b;
9705 Copy(array_b + i_b, array_u + i_u, copy_count, UV);
9707 len_u = i_u + copy_count;
9710 /* Set the result to the final length, which can change the pointer to
9711 * array_u, so re-find it. (Note that it is unlikely that this will
9712 * change, as we are shrinking the space, not enlarging it) */
9713 if (len_u != _invlist_len(u)) {
9714 invlist_set_len(u, len_u, *get_invlist_offset_addr(u));
9716 array_u = invlist_array(u);
9719 if (*output == NULL) { /* Simply return the new inversion list */
9723 /* Otherwise, overwrite the inversion list that was in '*output'. We
9724 * could instead free '*output', and then set it to 'u', but experience
9725 * has shown [perl #127392] that if the input is a mortal, we can get a
9726 * huge build-up of these during regex compilation before they get
9728 invlist_replace_list_destroys_src(*output, u);
9736 Perl__invlist_intersection_maybe_complement_2nd(pTHX_ SV* const a, SV* const b,
9737 const bool complement_b, SV** i)
9739 /* Take the intersection of two inversion lists and point '*i' to it. On
9740 * input, '*i' MUST POINT TO NULL OR TO AN SV* INVERSION LIST (possibly
9741 * even 'a' or 'b'). If to an inversion list, the contents of the original
9742 * list will be replaced by the intersection. The first list, 'a', may be
9743 * NULL, in which case '*i' will be an empty list. If 'complement_b' is
9744 * TRUE, the result will be the intersection of 'a' and the complement (or
9745 * inversion) of 'b' instead of 'b' directly.
9747 * The basis for this comes from "Unicode Demystified" Chapter 13 by
9748 * Richard Gillam, published by Addison-Wesley, and explained at some
9749 * length there. The preface says to incorporate its examples into your
9750 * code at your own risk. In fact, it had bugs
9752 * The algorithm is like a merge sort, and is essentially the same as the
9756 const UV* array_a; /* a's array */
9758 UV len_a; /* length of a's array */
9761 SV* r; /* the resulting intersection */
9765 UV i_a = 0; /* current index into a's array */
9769 /* running count of how many of the two inputs are postitioned at ranges
9770 * that are in their sets. As explained in the algorithm source book,
9771 * items are stopped accumulating and are output when the count changes
9772 * to/from 2. The count is incremented when we start a range that's in an
9773 * input's set, and decremented when we start a range that's not in a set.
9774 * Only when it is 2 are we in the intersection. */
9777 PERL_ARGS_ASSERT__INVLIST_INTERSECTION_MAYBE_COMPLEMENT_2ND;
9779 assert(*i == NULL || is_invlist(*i));
9781 /* Special case if either one is empty */
9782 len_a = (a == NULL) ? 0 : _invlist_len(a);
9783 if ((len_a == 0) || ((len_b = _invlist_len(b)) == 0)) {
9784 if (len_a != 0 && complement_b) {
9786 /* Here, 'a' is not empty, therefore from the enclosing 'if', 'b'
9787 * must be empty. Here, also we are using 'b's complement, which
9788 * hence must be every possible code point. Thus the intersection
9791 if (*i == a) { /* No-op */
9796 *i = invlist_clone(a, NULL);
9800 r = invlist_clone(a, NULL);
9801 invlist_replace_list_destroys_src(*i, r);
9806 /* Here, 'a' or 'b' is empty and not using the complement of 'b'. The
9807 * intersection must be empty */
9809 *i = _new_invlist(0);
9817 /* Here both lists exist and are non-empty */
9818 array_a = invlist_array(a);
9819 array_b = invlist_array(b);
9821 /* If are to take the intersection of 'a' with the complement of b, set it
9822 * up so are looking at b's complement. */
9825 /* To complement, we invert: if the first element is 0, remove it. To
9826 * do this, we just pretend the array starts one later */
9827 if (array_b[0] == 0) {
9833 /* But if the first element is not zero, we pretend the list starts
9834 * at the 0 that is always stored immediately before the array. */
9840 /* Size the intersection for the worst case: that the intersection ends up
9841 * fragmenting everything to be completely disjoint */
9842 r= _new_invlist(len_a + len_b);
9844 /* Will contain U+0000 iff both components do */
9845 array_r = _invlist_array_init(r, len_a > 0 && array_a[0] == 0
9846 && len_b > 0 && array_b[0] == 0);
9848 /* Go through each list item by item, stopping when have exhausted one of
9850 while (i_a < len_a && i_b < len_b) {
9851 UV cp; /* The element to potentially add to the intersection's
9853 bool cp_in_set; /* Is it in the input list's set or not */
9855 /* We need to take one or the other of the two inputs for the
9856 * intersection. Since we are merging two sorted lists, we take the
9857 * smaller of the next items. In case of a tie, we take first the one
9858 * that is not in its set (a difference from the union algorithm). If
9859 * we first took the one in its set, it would increment the count,
9860 * possibly to 2 which would cause it to be output as starting a range
9861 * in the intersection, and the next time through we would take that
9862 * same number, and output it again as ending the set. By doing the
9863 * opposite of this, there is no possibility that the count will be
9864 * momentarily incremented to 2. (In a tie and both are in the set or
9865 * both not in the set, it doesn't matter which we take first.) */
9866 if ( array_a[i_a] < array_b[i_b]
9867 || ( array_a[i_a] == array_b[i_b]
9868 && ! ELEMENT_RANGE_MATCHES_INVLIST(i_a)))
9870 cp_in_set = ELEMENT_RANGE_MATCHES_INVLIST(i_a);
9871 cp = array_a[i_a++];
9874 cp_in_set = ELEMENT_RANGE_MATCHES_INVLIST(i_b);
9878 /* Here, have chosen which of the two inputs to look at. Only output
9879 * if the running count changes to/from 2, which marks the
9880 * beginning/end of a range that's in the intersection */
9884 array_r[i_r++] = cp;
9889 array_r[i_r++] = cp;
9896 /* The loop above increments the index into exactly one of the input lists
9897 * each iteration, and ends when either index gets to its list end. That
9898 * means the other index is lower than its end, and so something is
9899 * remaining in that one. We increment 'count', as explained below, if the
9900 * exhausted list was in its set. (i_a and i_b each currently index the
9901 * element beyond the one we care about.) */
9902 if ( (i_a == len_a && PREV_RANGE_MATCHES_INVLIST(i_a))
9903 || (i_b == len_b && PREV_RANGE_MATCHES_INVLIST(i_b)))
9908 /* Above we incremented 'count' if the exhausted list was in its set. This
9909 * has made it so that 'count' being below 2 means there is nothing left to
9910 * output; otheriwse what's left to add to the intersection is precisely
9911 * that which is left in the non-exhausted input list.
9913 * To see why, note first that the exhausted input obviously has nothing
9914 * left to affect the intersection. If it was in its set at its end, that
9915 * means the set extends from here to the platform's infinity, and hence
9916 * anything in the non-exhausted's list will be in the intersection, and
9917 * anything not in it won't be. Hence, the rest of the intersection is
9918 * precisely what's in the non-exhausted list The exhausted set also
9919 * contributed 1 to 'count', meaning 'count' was at least 1. Incrementing
9920 * it means 'count' is now at least 2. This is consistent with the
9921 * incremented 'count' being >= 2 means to add the non-exhausted list to
9924 * But if the exhausted input wasn't in its set, it contributed 0 to
9925 * 'count', and the intersection can't include anything further; the
9926 * non-exhausted set is irrelevant. 'count' was at most 1, and doesn't get
9927 * incremented. This is consistent with 'count' being < 2 meaning nothing
9928 * further to add to the intersection. */
9929 if (count < 2) { /* Nothing left to put in the intersection. */
9932 else { /* copy the non-exhausted list, unchanged. */
9933 IV copy_count = len_a - i_a;
9934 if (copy_count > 0) { /* a is the one with stuff left */
9935 Copy(array_a + i_a, array_r + i_r, copy_count, UV);
9937 else { /* b is the one with stuff left */
9938 copy_count = len_b - i_b;
9939 Copy(array_b + i_b, array_r + i_r, copy_count, UV);
9941 len_r = i_r + copy_count;
9944 /* Set the result to the final length, which can change the pointer to
9945 * array_r, so re-find it. (Note that it is unlikely that this will
9946 * change, as we are shrinking the space, not enlarging it) */
9947 if (len_r != _invlist_len(r)) {
9948 invlist_set_len(r, len_r, *get_invlist_offset_addr(r));
9950 array_r = invlist_array(r);
9953 if (*i == NULL) { /* Simply return the calculated intersection */
9956 else { /* Otherwise, replace the existing inversion list in '*i'. We could
9957 instead free '*i', and then set it to 'r', but experience has
9958 shown [perl #127392] that if the input is a mortal, we can get a
9959 huge build-up of these during regex compilation before they get
9962 invlist_replace_list_destroys_src(*i, r);
9974 Perl__add_range_to_invlist(pTHX_ SV* invlist, UV start, UV end)
9976 /* Add the range from 'start' to 'end' inclusive to the inversion list's
9977 * set. A pointer to the inversion list is returned. This may actually be
9978 * a new list, in which case the passed in one has been destroyed. The
9979 * passed-in inversion list can be NULL, in which case a new one is created
9980 * with just the one range in it. The new list is not necessarily
9981 * NUL-terminated. Space is not freed if the inversion list shrinks as a
9982 * result of this function. The gain would not be large, and in many
9983 * cases, this is called multiple times on a single inversion list, so
9984 * anything freed may almost immediately be needed again.
9986 * This used to mostly call the 'union' routine, but that is much more
9987 * heavyweight than really needed for a single range addition */
9989 UV* array; /* The array implementing the inversion list */
9990 UV len; /* How many elements in 'array' */
9991 SSize_t i_s; /* index into the invlist array where 'start'
9993 SSize_t i_e = 0; /* And the index where 'end' should go */
9994 UV cur_highest; /* The highest code point in the inversion list
9995 upon entry to this function */
9997 /* This range becomes the whole inversion list if none already existed */
9998 if (invlist == NULL) {
9999 invlist = _new_invlist(2);
10000 _append_range_to_invlist(invlist, start, end);
10004 /* Likewise, if the inversion list is currently empty */
10005 len = _invlist_len(invlist);
10007 _append_range_to_invlist(invlist, start, end);
10011 /* Starting here, we have to know the internals of the list */
10012 array = invlist_array(invlist);
10014 /* If the new range ends higher than the current highest ... */
10015 cur_highest = invlist_highest(invlist);
10016 if (end > cur_highest) {
10018 /* If the whole range is higher, we can just append it */
10019 if (start > cur_highest) {
10020 _append_range_to_invlist(invlist, start, end);
10024 /* Otherwise, add the portion that is higher ... */
10025 _append_range_to_invlist(invlist, cur_highest + 1, end);
10027 /* ... and continue on below to handle the rest. As a result of the
10028 * above append, we know that the index of the end of the range is the
10029 * final even numbered one of the array. Recall that the final element
10030 * always starts a range that extends to infinity. If that range is in
10031 * the set (meaning the set goes from here to infinity), it will be an
10032 * even index, but if it isn't in the set, it's odd, and the final
10033 * range in the set is one less, which is even. */
10034 if (end == UV_MAX) {
10042 /* We have dealt with appending, now see about prepending. If the new
10043 * range starts lower than the current lowest ... */
10044 if (start < array[0]) {
10046 /* Adding something which has 0 in it is somewhat tricky, and uncommon.
10047 * Let the union code handle it, rather than having to know the
10048 * trickiness in two code places. */
10049 if (UNLIKELY(start == 0)) {
10052 range_invlist = _new_invlist(2);
10053 _append_range_to_invlist(range_invlist, start, end);
10055 _invlist_union(invlist, range_invlist, &invlist);
10057 SvREFCNT_dec_NN(range_invlist);
10062 /* If the whole new range comes before the first entry, and doesn't
10063 * extend it, we have to insert it as an additional range */
10064 if (end < array[0] - 1) {
10066 goto splice_in_new_range;
10069 /* Here the new range adjoins the existing first range, extending it
10073 /* And continue on below to handle the rest. We know that the index of
10074 * the beginning of the range is the first one of the array */
10077 else { /* Not prepending any part of the new range to the existing list.
10078 * Find where in the list it should go. This finds i_s, such that:
10079 * invlist[i_s] <= start < array[i_s+1]
10081 i_s = _invlist_search(invlist, start);
10084 /* At this point, any extending before the beginning of the inversion list
10085 * and/or after the end has been done. This has made it so that, in the
10086 * code below, each endpoint of the new range is either in a range that is
10087 * in the set, or is in a gap between two ranges that are. This means we
10088 * don't have to worry about exceeding the array bounds.
10090 * Find where in the list the new range ends (but we can skip this if we
10091 * have already determined what it is, or if it will be the same as i_s,
10092 * which we already have computed) */
10094 i_e = (start == end)
10096 : _invlist_search(invlist, end);
10099 /* Here generally invlist[i_e] <= end < array[i_e+1]. But if invlist[i_e]
10100 * is a range that goes to infinity there is no element at invlist[i_e+1],
10101 * so only the first relation holds. */
10103 if ( ! ELEMENT_RANGE_MATCHES_INVLIST(i_s)) {
10105 /* Here, the ranges on either side of the beginning of the new range
10106 * are in the set, and this range starts in the gap between them.
10108 * The new range extends the range above it downwards if the new range
10109 * ends at or above that range's start */
10110 const bool extends_the_range_above = ( end == UV_MAX
10111 || end + 1 >= array[i_s+1]);
10113 /* The new range extends the range below it upwards if it begins just
10114 * after where that range ends */
10115 if (start == array[i_s]) {
10117 /* If the new range fills the entire gap between the other ranges,
10118 * they will get merged together. Other ranges may also get
10119 * merged, depending on how many of them the new range spans. In
10120 * the general case, we do the merge later, just once, after we
10121 * figure out how many to merge. But in the case where the new
10122 * range exactly spans just this one gap (possibly extending into
10123 * the one above), we do the merge here, and an early exit. This
10124 * is done here to avoid having to special case later. */
10125 if (i_e - i_s <= 1) {
10127 /* If i_e - i_s == 1, it means that the new range terminates
10128 * within the range above, and hence 'extends_the_range_above'
10129 * must be true. (If the range above it extends to infinity,
10130 * 'i_s+2' will be above the array's limit, but 'len-i_s-2'
10131 * will be 0, so no harm done.) */
10132 if (extends_the_range_above) {
10133 Move(array + i_s + 2, array + i_s, len - i_s - 2, UV);
10134 invlist_set_len(invlist,
10136 *(get_invlist_offset_addr(invlist)));
10140 /* Here, i_e must == i_s. We keep them in sync, as they apply
10141 * to the same range, and below we are about to decrement i_s
10146 /* Here, the new range is adjacent to the one below. (It may also
10147 * span beyond the range above, but that will get resolved later.)
10148 * Extend the range below to include this one. */
10149 array[i_s] = (end == UV_MAX) ? UV_MAX : end + 1;
10151 start = array[i_s];
10153 else if (extends_the_range_above) {
10155 /* Here the new range only extends the range above it, but not the
10156 * one below. It merges with the one above. Again, we keep i_e
10157 * and i_s in sync if they point to the same range */
10162 array[i_s] = start;
10166 /* Here, we've dealt with the new range start extending any adjoining
10169 * If the new range extends to infinity, it is now the final one,
10170 * regardless of what was there before */
10171 if (UNLIKELY(end == UV_MAX)) {
10172 invlist_set_len(invlist, i_s + 1, *(get_invlist_offset_addr(invlist)));
10176 /* If i_e started as == i_s, it has also been dealt with,
10177 * and been updated to the new i_s, which will fail the following if */
10178 if (! ELEMENT_RANGE_MATCHES_INVLIST(i_e)) {
10180 /* Here, the ranges on either side of the end of the new range are in
10181 * the set, and this range ends in the gap between them.
10183 * If this range is adjacent to (hence extends) the range above it, it
10184 * becomes part of that range; likewise if it extends the range below,
10185 * it becomes part of that range */
10186 if (end + 1 == array[i_e+1]) {
10188 array[i_e] = start;
10190 else if (start <= array[i_e]) {
10191 array[i_e] = end + 1;
10198 /* If the range fits entirely in an existing range (as possibly already
10199 * extended above), it doesn't add anything new */
10200 if (ELEMENT_RANGE_MATCHES_INVLIST(i_s)) {
10204 /* Here, no part of the range is in the list. Must add it. It will
10205 * occupy 2 more slots */
10206 splice_in_new_range:
10208 invlist_extend(invlist, len + 2);
10209 array = invlist_array(invlist);
10210 /* Move the rest of the array down two slots. Don't include any
10212 Move(array + i_e + 1, array + i_e + 3, len - i_e - 1, UV);
10214 /* Do the actual splice */
10215 array[i_e+1] = start;
10216 array[i_e+2] = end + 1;
10217 invlist_set_len(invlist, len + 2, *(get_invlist_offset_addr(invlist)));
10221 /* Here the new range crossed the boundaries of a pre-existing range. The
10222 * code above has adjusted things so that both ends are in ranges that are
10223 * in the set. This means everything in between must also be in the set.
10224 * Just squash things together */
10225 Move(array + i_e + 1, array + i_s + 1, len - i_e - 1, UV);
10226 invlist_set_len(invlist,
10228 *(get_invlist_offset_addr(invlist)));
10234 Perl__setup_canned_invlist(pTHX_ const STRLEN size, const UV element0,
10235 UV** other_elements_ptr)
10237 /* Create and return an inversion list whose contents are to be populated
10238 * by the caller. The caller gives the number of elements (in 'size') and
10239 * the very first element ('element0'). This function will set
10240 * '*other_elements_ptr' to an array of UVs, where the remaining elements
10241 * are to be placed.
10243 * Obviously there is some trust involved that the caller will properly
10244 * fill in the other elements of the array.
10246 * (The first element needs to be passed in, as the underlying code does
10247 * things differently depending on whether it is zero or non-zero) */
10249 SV* invlist = _new_invlist(size);
10252 PERL_ARGS_ASSERT__SETUP_CANNED_INVLIST;
10254 invlist = add_cp_to_invlist(invlist, element0);
10255 offset = *get_invlist_offset_addr(invlist);
10257 invlist_set_len(invlist, size, offset);
10258 *other_elements_ptr = invlist_array(invlist) + 1;
10264 PERL_STATIC_INLINE SV*
10265 S_add_cp_to_invlist(pTHX_ SV* invlist, const UV cp) {
10266 return _add_range_to_invlist(invlist, cp, cp);
10269 #ifndef PERL_IN_XSUB_RE
10271 Perl__invlist_invert(pTHX_ SV* const invlist)
10273 /* Complement the input inversion list. This adds a 0 if the list didn't
10274 * have a zero; removes it otherwise. As described above, the data
10275 * structure is set up so that this is very efficient */
10277 PERL_ARGS_ASSERT__INVLIST_INVERT;
10279 assert(! invlist_is_iterating(invlist));
10281 /* The inverse of matching nothing is matching everything */
10282 if (_invlist_len(invlist) == 0) {
10283 _append_range_to_invlist(invlist, 0, UV_MAX);
10287 *get_invlist_offset_addr(invlist) = ! *get_invlist_offset_addr(invlist);
10291 Perl_invlist_clone(pTHX_ SV* const invlist, SV* new_invlist)
10293 /* Return a new inversion list that is a copy of the input one, which is
10294 * unchanged. The new list will not be mortal even if the old one was. */
10296 const STRLEN nominal_length = _invlist_len(invlist);
10297 const STRLEN physical_length = SvCUR(invlist);
10298 const bool offset = *(get_invlist_offset_addr(invlist));
10300 PERL_ARGS_ASSERT_INVLIST_CLONE;
10302 if (new_invlist == NULL) {
10303 new_invlist = _new_invlist(nominal_length);
10306 sv_upgrade(new_invlist, SVt_INVLIST);
10307 initialize_invlist_guts(new_invlist, nominal_length);
10310 *(get_invlist_offset_addr(new_invlist)) = offset;
10311 invlist_set_len(new_invlist, nominal_length, offset);
10312 Copy(SvPVX(invlist), SvPVX(new_invlist), physical_length, char);
10314 return new_invlist;
10319 PERL_STATIC_INLINE STRLEN*
10320 S_get_invlist_iter_addr(SV* invlist)
10322 /* Return the address of the UV that contains the current iteration
10325 PERL_ARGS_ASSERT_GET_INVLIST_ITER_ADDR;
10327 assert(is_invlist(invlist));
10329 return &(((XINVLIST*) SvANY(invlist))->iterator);
10332 PERL_STATIC_INLINE void
10333 S_invlist_iterinit(SV* invlist) /* Initialize iterator for invlist */
10335 PERL_ARGS_ASSERT_INVLIST_ITERINIT;
10337 *get_invlist_iter_addr(invlist) = 0;
10340 PERL_STATIC_INLINE void
10341 S_invlist_iterfinish(SV* invlist)
10343 /* Terminate iterator for invlist. This is to catch development errors.
10344 * Any iteration that is interrupted before completed should call this
10345 * function. Functions that add code points anywhere else but to the end
10346 * of an inversion list assert that they are not in the middle of an
10347 * iteration. If they were, the addition would make the iteration
10348 * problematical: if the iteration hadn't reached the place where things
10349 * were being added, it would be ok */
10351 PERL_ARGS_ASSERT_INVLIST_ITERFINISH;
10353 *get_invlist_iter_addr(invlist) = (STRLEN) UV_MAX;
10357 S_invlist_iternext(SV* invlist, UV* start, UV* end)
10359 /* An C<invlist_iterinit> call on <invlist> must be used to set this up.
10360 * This call sets in <*start> and <*end>, the next range in <invlist>.
10361 * Returns <TRUE> if successful and the next call will return the next
10362 * range; <FALSE> if was already at the end of the list. If the latter,
10363 * <*start> and <*end> are unchanged, and the next call to this function
10364 * will start over at the beginning of the list */
10366 STRLEN* pos = get_invlist_iter_addr(invlist);
10367 UV len = _invlist_len(invlist);
10370 PERL_ARGS_ASSERT_INVLIST_ITERNEXT;
10373 *pos = (STRLEN) UV_MAX; /* Force iterinit() to be required next time */
10377 array = invlist_array(invlist);
10379 *start = array[(*pos)++];
10385 *end = array[(*pos)++] - 1;
10391 PERL_STATIC_INLINE UV
10392 S_invlist_highest(SV* const invlist)
10394 /* Returns the highest code point that matches an inversion list. This API
10395 * has an ambiguity, as it returns 0 under either the highest is actually
10396 * 0, or if the list is empty. If this distinction matters to you, check
10397 * for emptiness before calling this function */
10399 UV len = _invlist_len(invlist);
10402 PERL_ARGS_ASSERT_INVLIST_HIGHEST;
10408 array = invlist_array(invlist);
10410 /* The last element in the array in the inversion list always starts a
10411 * range that goes to infinity. That range may be for code points that are
10412 * matched in the inversion list, or it may be for ones that aren't
10413 * matched. In the latter case, the highest code point in the set is one
10414 * less than the beginning of this range; otherwise it is the final element
10415 * of this range: infinity */
10416 return (ELEMENT_RANGE_MATCHES_INVLIST(len - 1))
10418 : array[len - 1] - 1;
10422 S_invlist_contents(pTHX_ SV* const invlist, const bool traditional_style)
10424 /* Get the contents of an inversion list into a string SV so that they can
10425 * be printed out. If 'traditional_style' is TRUE, it uses the format
10426 * traditionally done for debug tracing; otherwise it uses a format
10427 * suitable for just copying to the output, with blanks between ranges and
10428 * a dash between range components */
10432 const char intra_range_delimiter = (traditional_style ? '\t' : '-');
10433 const char inter_range_delimiter = (traditional_style ? '\n' : ' ');
10435 if (traditional_style) {
10436 output = newSVpvs("\n");
10439 output = newSVpvs("");
10442 PERL_ARGS_ASSERT_INVLIST_CONTENTS;
10444 assert(! invlist_is_iterating(invlist));
10446 invlist_iterinit(invlist);
10447 while (invlist_iternext(invlist, &start, &end)) {
10448 if (end == UV_MAX) {
10449 Perl_sv_catpvf(aTHX_ output, "%04" UVXf "%cINFTY%c",
10450 start, intra_range_delimiter,
10451 inter_range_delimiter);
10453 else if (end != start) {
10454 Perl_sv_catpvf(aTHX_ output, "%04" UVXf "%c%04" UVXf "%c",
10456 intra_range_delimiter,
10457 end, inter_range_delimiter);
10460 Perl_sv_catpvf(aTHX_ output, "%04" UVXf "%c",
10461 start, inter_range_delimiter);
10465 if (SvCUR(output) && ! traditional_style) {/* Get rid of trailing blank */
10466 SvCUR_set(output, SvCUR(output) - 1);
10472 #ifndef PERL_IN_XSUB_RE
10474 Perl__invlist_dump(pTHX_ PerlIO *file, I32 level,
10475 const char * const indent, SV* const invlist)
10477 /* Designed to be called only by do_sv_dump(). Dumps out the ranges of the
10478 * inversion list 'invlist' to 'file' at 'level' Each line is prefixed by
10479 * the string 'indent'. The output looks like this:
10480 [0] 0x000A .. 0x000D
10482 [4] 0x2028 .. 0x2029
10483 [6] 0x3104 .. INFTY
10484 * This means that the first range of code points matched by the list are
10485 * 0xA through 0xD; the second range contains only the single code point
10486 * 0x85, etc. An inversion list is an array of UVs. Two array elements
10487 * are used to define each range (except if the final range extends to
10488 * infinity, only a single element is needed). The array index of the
10489 * first element for the corresponding range is given in brackets. */
10494 PERL_ARGS_ASSERT__INVLIST_DUMP;
10496 if (invlist_is_iterating(invlist)) {
10497 Perl_dump_indent(aTHX_ level, file,
10498 "%sCan't dump inversion list because is in middle of iterating\n",
10503 invlist_iterinit(invlist);
10504 while (invlist_iternext(invlist, &start, &end)) {
10505 if (end == UV_MAX) {
10506 Perl_dump_indent(aTHX_ level, file,
10507 "%s[%" UVuf "] 0x%04" UVXf " .. INFTY\n",
10508 indent, (UV)count, start);
10510 else if (end != start) {
10511 Perl_dump_indent(aTHX_ level, file,
10512 "%s[%" UVuf "] 0x%04" UVXf " .. 0x%04" UVXf "\n",
10513 indent, (UV)count, start, end);
10516 Perl_dump_indent(aTHX_ level, file, "%s[%" UVuf "] 0x%04" UVXf "\n",
10517 indent, (UV)count, start);
10525 #if defined(PERL_ARGS_ASSERT__INVLISTEQ) && !defined(PERL_IN_XSUB_RE)
10527 Perl__invlistEQ(pTHX_ SV* const a, SV* const b, const bool complement_b)
10529 /* Return a boolean as to if the two passed in inversion lists are
10530 * identical. The final argument, if TRUE, says to take the complement of
10531 * the second inversion list before doing the comparison */
10533 const UV len_a = _invlist_len(a);
10534 UV len_b = _invlist_len(b);
10536 const UV* array_a = NULL;
10537 const UV* array_b = NULL;
10539 PERL_ARGS_ASSERT__INVLISTEQ;
10541 /* This code avoids accessing the arrays unless it knows the length is
10546 return ! complement_b;
10550 array_a = invlist_array(a);
10554 array_b = invlist_array(b);
10557 /* If are to compare 'a' with the complement of b, set it
10558 * up so are looking at b's complement. */
10559 if (complement_b) {
10561 /* The complement of nothing is everything, so <a> would have to have
10562 * just one element, starting at zero (ending at infinity) */
10564 return (len_a == 1 && array_a[0] == 0);
10566 if (array_b[0] == 0) {
10568 /* Otherwise, to complement, we invert. Here, the first element is
10569 * 0, just remove it. To do this, we just pretend the array starts
10577 /* But if the first element is not zero, we pretend the list starts
10578 * at the 0 that is always stored immediately before the array. */
10584 return len_a == len_b
10585 && memEQ(array_a, array_b, len_a * sizeof(array_a[0]));
10591 * As best we can, determine the characters that can match the start of
10592 * the given EXACTF-ish node. This is for use in creating ssc nodes, so there
10593 * can be false positive matches
10595 * Returns the invlist as a new SV*; it is the caller's responsibility to
10596 * call SvREFCNT_dec() when done with it.
10599 S__make_exactf_invlist(pTHX_ RExC_state_t *pRExC_state, regnode *node)
10602 const U8 * s = (U8*)STRING(node);
10603 SSize_t bytelen = STR_LEN(node);
10605 /* Start out big enough for 2 separate code points */
10606 SV* invlist = _new_invlist(4);
10608 PERL_ARGS_ASSERT__MAKE_EXACTF_INVLIST;
10613 /* We punt and assume can match anything if the node begins
10614 * with a multi-character fold. Things are complicated. For
10615 * example, /ffi/i could match any of:
10616 * "\N{LATIN SMALL LIGATURE FFI}"
10617 * "\N{LATIN SMALL LIGATURE FF}I"
10618 * "F\N{LATIN SMALL LIGATURE FI}"
10619 * plus several other things; and making sure we have all the
10620 * possibilities is hard. */
10621 if (is_MULTI_CHAR_FOLD_latin1_safe(s, s + bytelen)) {
10622 invlist = _add_range_to_invlist(invlist, 0, UV_MAX);
10625 /* Any Latin1 range character can potentially match any
10626 * other depending on the locale, and in Turkic locales, U+130 and
10628 if (OP(node) == EXACTFL) {
10629 _invlist_union(invlist, PL_Latin1, &invlist);
10630 invlist = add_cp_to_invlist(invlist,
10631 LATIN_SMALL_LETTER_DOTLESS_I);
10632 invlist = add_cp_to_invlist(invlist,
10633 LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE);
10636 /* But otherwise, it matches at least itself. We can
10637 * quickly tell if it has a distinct fold, and if so,
10638 * it matches that as well */
10639 invlist = add_cp_to_invlist(invlist, uc);
10640 if (IS_IN_SOME_FOLD_L1(uc))
10641 invlist = add_cp_to_invlist(invlist, PL_fold_latin1[uc]);
10644 /* Some characters match above-Latin1 ones under /i. This
10645 * is true of EXACTFL ones when the locale is UTF-8 */
10646 if (HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE(uc)
10647 && (! isASCII(uc) || (OP(node) != EXACTFAA
10648 && OP(node) != EXACTFAA_NO_TRIE)))
10650 add_above_Latin1_folds(pRExC_state, (U8) uc, &invlist);
10654 else { /* Pattern is UTF-8 */
10655 U8 folded[UTF8_MAX_FOLD_CHAR_EXPAND * UTF8_MAXBYTES_CASE + 1] = { '\0' };
10656 const U8* e = s + bytelen;
10659 fc = uc = utf8_to_uvchr_buf(s, s + bytelen, NULL);
10661 /* The only code points that aren't folded in a UTF EXACTFish
10662 * node are are the problematic ones in EXACTFL nodes */
10663 if (OP(node) == EXACTFL && is_PROBLEMATIC_LOCALE_FOLDEDS_START_cp(uc)) {
10664 /* We need to check for the possibility that this EXACTFL
10665 * node begins with a multi-char fold. Therefore we fold
10666 * the first few characters of it so that we can make that
10672 for (i = 0; i < UTF8_MAX_FOLD_CHAR_EXPAND && s < e; i++) {
10674 *(d++) = (U8) toFOLD(*s);
10675 if (fc < 0) { /* Save the first fold */
10682 UV fold = toFOLD_utf8_safe(s, e, d, &len);
10683 if (fc < 0) { /* Save the first fold */
10691 /* And set up so the code below that looks in this folded
10692 * buffer instead of the node's string */
10697 /* When we reach here 's' points to the fold of the first
10698 * character(s) of the node; and 'e' points to far enough along
10699 * the folded string to be just past any possible multi-char
10702 * Unlike the non-UTF-8 case, the macro for determining if a
10703 * string is a multi-char fold requires all the characters to
10704 * already be folded. This is because of all the complications
10705 * if not. Note that they are folded anyway, except in EXACTFL
10706 * nodes. Like the non-UTF case above, we punt if the node
10707 * begins with a multi-char fold */
10709 if (is_MULTI_CHAR_FOLD_utf8_safe(s, e)) {
10710 invlist = _add_range_to_invlist(invlist, 0, UV_MAX);
10712 else { /* Single char fold */
10714 unsigned int first_fold;
10715 const unsigned int * remaining_folds;
10716 Size_t folds_count;
10718 /* It matches itself */
10719 invlist = add_cp_to_invlist(invlist, fc);
10721 /* ... plus all the things that fold to it, which are found in
10722 * PL_utf8_foldclosures */
10723 folds_count = _inverse_folds(fc, &first_fold,
10725 for (k = 0; k < folds_count; k++) {
10726 UV c = (k == 0) ? first_fold : remaining_folds[k-1];
10728 /* /aa doesn't allow folds between ASCII and non- */
10729 if ( (OP(node) == EXACTFAA || OP(node) == EXACTFAA_NO_TRIE)
10730 && isASCII(c) != isASCII(fc))
10735 invlist = add_cp_to_invlist(invlist, c);
10738 if (OP(node) == EXACTFL) {
10740 /* If either [iI] are present in an EXACTFL node the above code
10741 * should have added its normal case pair, but under a Turkish
10742 * locale they could match instead the case pairs from it. Add
10743 * those as potential matches as well */
10744 if (isALPHA_FOLD_EQ(fc, 'I')) {
10745 invlist = add_cp_to_invlist(invlist,
10746 LATIN_SMALL_LETTER_DOTLESS_I);
10747 invlist = add_cp_to_invlist(invlist,
10748 LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE);
10750 else if (fc == LATIN_SMALL_LETTER_DOTLESS_I) {
10751 invlist = add_cp_to_invlist(invlist, 'I');
10753 else if (fc == LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE) {
10754 invlist = add_cp_to_invlist(invlist, 'i');
10763 #undef HEADER_LENGTH
10764 #undef TO_INTERNAL_SIZE
10765 #undef FROM_INTERNAL_SIZE
10766 #undef INVLIST_VERSION_ID
10768 /* End of inversion list object */
10771 S_parse_lparen_question_flags(pTHX_ RExC_state_t *pRExC_state)
10773 /* This parses the flags that are in either the '(?foo)' or '(?foo:bar)'
10774 * constructs, and updates RExC_flags with them. On input, RExC_parse
10775 * should point to the first flag; it is updated on output to point to the
10776 * final ')' or ':'. There needs to be at least one flag, or this will
10779 /* for (?g), (?gc), and (?o) warnings; warning
10780 about (?c) will warn about (?g) -- japhy */
10782 #define WASTED_O 0x01
10783 #define WASTED_G 0x02
10784 #define WASTED_C 0x04
10785 #define WASTED_GC (WASTED_G|WASTED_C)
10786 I32 wastedflags = 0x00;
10787 U32 posflags = 0, negflags = 0;
10788 U32 *flagsp = &posflags;
10789 char has_charset_modifier = '\0';
10791 bool has_use_defaults = FALSE;
10792 const char* const seqstart = RExC_parse - 1; /* Point to the '?' */
10793 int x_mod_count = 0;
10795 PERL_ARGS_ASSERT_PARSE_LPAREN_QUESTION_FLAGS;
10797 /* '^' as an initial flag sets certain defaults */
10798 if (UCHARAT(RExC_parse) == '^') {
10800 has_use_defaults = TRUE;
10801 STD_PMMOD_FLAGS_CLEAR(&RExC_flags);
10802 cs = (RExC_uni_semantics)
10803 ? REGEX_UNICODE_CHARSET
10804 : REGEX_DEPENDS_CHARSET;
10805 set_regex_charset(&RExC_flags, cs);
10808 cs = get_regex_charset(RExC_flags);
10809 if ( cs == REGEX_DEPENDS_CHARSET
10810 && RExC_uni_semantics)
10812 cs = REGEX_UNICODE_CHARSET;
10816 while (RExC_parse < RExC_end) {
10817 /* && strchr("iogcmsx", *RExC_parse) */
10818 /* (?g), (?gc) and (?o) are useless here
10819 and must be globally applied -- japhy */
10820 switch (*RExC_parse) {
10822 /* Code for the imsxn flags */
10823 CASE_STD_PMMOD_FLAGS_PARSE_SET(flagsp, x_mod_count);
10825 case LOCALE_PAT_MOD:
10826 if (has_charset_modifier) {
10827 goto excess_modifier;
10829 else if (flagsp == &negflags) {
10832 cs = REGEX_LOCALE_CHARSET;
10833 has_charset_modifier = LOCALE_PAT_MOD;
10835 case UNICODE_PAT_MOD:
10836 if (has_charset_modifier) {
10837 goto excess_modifier;
10839 else if (flagsp == &negflags) {
10842 cs = REGEX_UNICODE_CHARSET;
10843 has_charset_modifier = UNICODE_PAT_MOD;
10845 case ASCII_RESTRICT_PAT_MOD:
10846 if (flagsp == &negflags) {
10849 if (has_charset_modifier) {
10850 if (cs != REGEX_ASCII_RESTRICTED_CHARSET) {
10851 goto excess_modifier;
10853 /* Doubled modifier implies more restricted */
10854 cs = REGEX_ASCII_MORE_RESTRICTED_CHARSET;
10857 cs = REGEX_ASCII_RESTRICTED_CHARSET;
10859 has_charset_modifier = ASCII_RESTRICT_PAT_MOD;
10861 case DEPENDS_PAT_MOD:
10862 if (has_use_defaults) {
10863 goto fail_modifiers;
10865 else if (flagsp == &negflags) {
10868 else if (has_charset_modifier) {
10869 goto excess_modifier;
10872 /* The dual charset means unicode semantics if the
10873 * pattern (or target, not known until runtime) are
10874 * utf8, or something in the pattern indicates unicode
10876 cs = (RExC_uni_semantics)
10877 ? REGEX_UNICODE_CHARSET
10878 : REGEX_DEPENDS_CHARSET;
10879 has_charset_modifier = DEPENDS_PAT_MOD;
10883 if (has_charset_modifier == ASCII_RESTRICT_PAT_MOD) {
10884 vFAIL2("Regexp modifier \"%c\" may appear a maximum of twice", ASCII_RESTRICT_PAT_MOD);
10886 else if (has_charset_modifier == *(RExC_parse - 1)) {
10887 vFAIL2("Regexp modifier \"%c\" may not appear twice",
10888 *(RExC_parse - 1));
10891 vFAIL3("Regexp modifiers \"%c\" and \"%c\" are mutually exclusive", has_charset_modifier, *(RExC_parse - 1));
10893 NOT_REACHED; /*NOTREACHED*/
10896 vFAIL2("Regexp modifier \"%c\" may not appear after the \"-\"",
10897 *(RExC_parse - 1));
10898 NOT_REACHED; /*NOTREACHED*/
10899 case ONCE_PAT_MOD: /* 'o' */
10900 case GLOBAL_PAT_MOD: /* 'g' */
10901 if (ckWARN(WARN_REGEXP)) {
10902 const I32 wflagbit = *RExC_parse == 'o'
10905 if (! (wastedflags & wflagbit) ) {
10906 wastedflags |= wflagbit;
10907 /* diag_listed_as: Useless (?-%s) - don't use /%s modifier in regex; marked by <-- HERE in m/%s/ */
10910 "Useless (%s%c) - %suse /%c modifier",
10911 flagsp == &negflags ? "?-" : "?",
10913 flagsp == &negflags ? "don't " : "",
10920 case CONTINUE_PAT_MOD: /* 'c' */
10921 if (ckWARN(WARN_REGEXP)) {
10922 if (! (wastedflags & WASTED_C) ) {
10923 wastedflags |= WASTED_GC;
10924 /* diag_listed_as: Useless (?-%s) - don't use /%s modifier in regex; marked by <-- HERE in m/%s/ */
10927 "Useless (%sc) - %suse /gc modifier",
10928 flagsp == &negflags ? "?-" : "?",
10929 flagsp == &negflags ? "don't " : ""
10934 case KEEPCOPY_PAT_MOD: /* 'p' */
10935 if (flagsp == &negflags) {
10936 ckWARNreg(RExC_parse + 1,"Useless use of (?-p)");
10938 *flagsp |= RXf_PMf_KEEPCOPY;
10942 /* A flag is a default iff it is following a minus, so
10943 * if there is a minus, it means will be trying to
10944 * re-specify a default which is an error */
10945 if (has_use_defaults || flagsp == &negflags) {
10946 goto fail_modifiers;
10948 flagsp = &negflags;
10949 wastedflags = 0; /* reset so (?g-c) warns twice */
10955 if ((posflags & (RXf_PMf_EXTENDED|RXf_PMf_EXTENDED_MORE)) == RXf_PMf_EXTENDED) {
10956 negflags |= RXf_PMf_EXTENDED_MORE;
10958 RExC_flags |= posflags;
10960 if (negflags & RXf_PMf_EXTENDED) {
10961 negflags |= RXf_PMf_EXTENDED_MORE;
10963 RExC_flags &= ~negflags;
10964 set_regex_charset(&RExC_flags, cs);
10969 RExC_parse += SKIP_IF_CHAR(RExC_parse, RExC_end);
10970 /* diag_listed_as: Sequence (?%s...) not recognized in regex; marked by <-- HERE in m/%s/ */
10971 vFAIL2utf8f("Sequence (%" UTF8f "...) not recognized",
10972 UTF8fARG(UTF, RExC_parse-seqstart, seqstart));
10973 NOT_REACHED; /*NOTREACHED*/
10976 RExC_parse += UTF ? UTF8SKIP(RExC_parse) : 1;
10979 vFAIL("Sequence (?... not terminated");
10983 - reg - regular expression, i.e. main body or parenthesized thing
10985 * Caller must absorb opening parenthesis.
10987 * Combining parenthesis handling with the base level of regular expression
10988 * is a trifle forced, but the need to tie the tails of the branches to what
10989 * follows makes it hard to avoid.
10991 #define REGTAIL(x,y,z) regtail((x),(y),(z),depth+1)
10993 #define REGTAIL_STUDY(x,y,z) regtail_study((x),(y),(z),depth+1)
10995 #define REGTAIL_STUDY(x,y,z) regtail((x),(y),(z),depth+1)
10998 PERL_STATIC_INLINE regnode_offset
10999 S_handle_named_backref(pTHX_ RExC_state_t *pRExC_state,
11001 char * parse_start,
11005 regnode_offset ret;
11006 char* name_start = RExC_parse;
11008 SV *sv_dat = reg_scan_name(pRExC_state, REG_RSN_RETURN_DATA);
11009 GET_RE_DEBUG_FLAGS_DECL;
11011 PERL_ARGS_ASSERT_HANDLE_NAMED_BACKREF;
11013 if (RExC_parse == name_start || *RExC_parse != ch) {
11014 /* diag_listed_as: Sequence \%s... not terminated in regex; marked by <-- HERE in m/%s/ */
11015 vFAIL2("Sequence %.3s... not terminated", parse_start);
11019 num = add_data( pRExC_state, STR_WITH_LEN("S"));
11020 RExC_rxi->data->data[num]=(void*)sv_dat;
11021 SvREFCNT_inc_simple_void_NN(sv_dat);
11024 ret = reganode(pRExC_state,
11027 : (ASCII_FOLD_RESTRICTED)
11029 : (AT_LEAST_UNI_SEMANTICS)
11035 *flagp |= HASWIDTH;
11037 Set_Node_Offset(REGNODE_p(ret), parse_start+1);
11038 Set_Node_Cur_Length(REGNODE_p(ret), parse_start);
11040 nextchar(pRExC_state);
11044 /* On success, returns the offset at which any next node should be placed into
11045 * the regex engine program being compiled.
11047 * Returns 0 otherwise, with *flagp set to indicate why:
11048 * TRYAGAIN at the end of (?) that only sets flags.
11049 * RESTART_PARSE if the parse needs to be restarted, or'd with
11050 * NEED_UTF8 if the pattern needs to be upgraded to UTF-8.
11051 * Otherwise would only return 0 if regbranch() returns 0, which cannot
11053 STATIC regnode_offset
11054 S_reg(pTHX_ RExC_state_t *pRExC_state, I32 paren, I32 *flagp, U32 depth)
11055 /* paren: Parenthesized? 0=top; 1,2=inside '(': changed to letter.
11056 * 2 is like 1, but indicates that nextchar() has been called to advance
11057 * RExC_parse beyond the '('. Things like '(?' are indivisible tokens, and
11058 * this flag alerts us to the need to check for that */
11060 regnode_offset ret = 0; /* Will be the head of the group. */
11062 regnode_offset lastbr;
11063 regnode_offset ender = 0;
11066 U32 oregflags = RExC_flags;
11067 bool have_branch = 0;
11069 I32 freeze_paren = 0;
11070 I32 after_freeze = 0;
11071 I32 num; /* numeric backreferences */
11072 SV * max_open; /* Max number of unclosed parens */
11074 char * parse_start = RExC_parse; /* MJD */
11075 char * const oregcomp_parse = RExC_parse;
11077 GET_RE_DEBUG_FLAGS_DECL;
11079 PERL_ARGS_ASSERT_REG;
11080 DEBUG_PARSE("reg ");
11083 max_open = get_sv(RE_COMPILE_RECURSION_LIMIT, GV_ADD);
11085 if (!SvIOK(max_open)) {
11086 sv_setiv(max_open, RE_COMPILE_RECURSION_INIT);
11088 if (depth > 4 * (UV) SvIV(max_open)) { /* We increase depth by 4 for each
11090 vFAIL("Too many nested open parens");
11093 *flagp = 0; /* Tentatively. */
11095 if (RExC_in_lookbehind) {
11096 RExC_in_lookbehind++;
11098 if (RExC_in_lookahead) {
11099 RExC_in_lookahead++;
11102 /* Having this true makes it feasible to have a lot fewer tests for the
11103 * parse pointer being in scope. For example, we can write
11104 * while(isFOO(*RExC_parse)) RExC_parse++;
11106 * while(RExC_parse < RExC_end && isFOO(*RExC_parse)) RExC_parse++;
11108 assert(*RExC_end == '\0');
11110 /* Make an OPEN node, if parenthesized. */
11113 /* Under /x, space and comments can be gobbled up between the '(' and
11114 * here (if paren ==2). The forms '(*VERB' and '(?...' disallow such
11115 * intervening space, as the sequence is a token, and a token should be
11117 bool has_intervening_patws = (paren == 2)
11118 && *(RExC_parse - 1) != '(';
11120 if (RExC_parse >= RExC_end) {
11121 vFAIL("Unmatched (");
11124 if (paren == 'r') { /* Atomic script run */
11128 else if ( *RExC_parse == '*') { /* (*VERB:ARG), (*construct:...) */
11129 char *start_verb = RExC_parse + 1;
11131 char *start_arg = NULL;
11132 unsigned char op = 0;
11133 int arg_required = 0;
11134 int internal_argval = -1; /* if >-1 we are not allowed an argument*/
11135 bool has_upper = FALSE;
11137 if (has_intervening_patws) {
11138 RExC_parse++; /* past the '*' */
11140 /* For strict backwards compatibility, don't change the message
11141 * now that we also have lowercase operands */
11142 if (isUPPER(*RExC_parse)) {
11143 vFAIL("In '(*VERB...)', the '(' and '*' must be adjacent");
11146 vFAIL("In '(*...)', the '(' and '*' must be adjacent");
11149 while (RExC_parse < RExC_end && *RExC_parse != ')' ) {
11150 if ( *RExC_parse == ':' ) {
11151 start_arg = RExC_parse + 1;
11155 if (isUPPER(*RExC_parse)) {
11161 RExC_parse += UTF8SKIP(RExC_parse);
11164 verb_len = RExC_parse - start_verb;
11166 if (RExC_parse >= RExC_end) {
11167 goto unterminated_verb_pattern;
11170 RExC_parse += UTF ? UTF8SKIP(RExC_parse) : 1;
11171 while ( RExC_parse < RExC_end && *RExC_parse != ')' ) {
11172 RExC_parse += UTF ? UTF8SKIP(RExC_parse) : 1;
11174 if ( RExC_parse >= RExC_end || *RExC_parse != ')' ) {
11175 unterminated_verb_pattern:
11177 vFAIL("Unterminated verb pattern argument");
11180 vFAIL("Unterminated '(*...' argument");
11184 if ( RExC_parse >= RExC_end || *RExC_parse != ')' ) {
11186 vFAIL("Unterminated verb pattern");
11189 vFAIL("Unterminated '(*...' construct");
11194 /* Here, we know that RExC_parse < RExC_end */
11196 switch ( *start_verb ) {
11197 case 'A': /* (*ACCEPT) */
11198 if ( memEQs(start_verb, verb_len,"ACCEPT") ) {
11200 internal_argval = RExC_nestroot;
11203 case 'C': /* (*COMMIT) */
11204 if ( memEQs(start_verb, verb_len,"COMMIT") )
11207 case 'F': /* (*FAIL) */
11208 if ( verb_len==1 || memEQs(start_verb, verb_len,"FAIL") ) {
11212 case ':': /* (*:NAME) */
11213 case 'M': /* (*MARK:NAME) */
11214 if ( verb_len==0 || memEQs(start_verb, verb_len,"MARK") ) {
11219 case 'P': /* (*PRUNE) */
11220 if ( memEQs(start_verb, verb_len,"PRUNE") )
11223 case 'S': /* (*SKIP) */
11224 if ( memEQs(start_verb, verb_len,"SKIP") )
11227 case 'T': /* (*THEN) */
11228 /* [19:06] <TimToady> :: is then */
11229 if ( memEQs(start_verb, verb_len,"THEN") ) {
11231 RExC_seen |= REG_CUTGROUP_SEEN;
11235 if ( memEQs(start_verb, verb_len, "asr")
11236 || memEQs(start_verb, verb_len, "atomic_script_run"))
11238 paren = 'r'; /* Mnemonic: recursed run */
11241 else if (memEQs(start_verb, verb_len, "atomic")) {
11242 paren = 't'; /* AtOMIC */
11243 goto alpha_assertions;
11247 if ( memEQs(start_verb, verb_len, "plb")
11248 || memEQs(start_verb, verb_len, "positive_lookbehind"))
11251 goto lookbehind_alpha_assertions;
11253 else if ( memEQs(start_verb, verb_len, "pla")
11254 || memEQs(start_verb, verb_len, "positive_lookahead"))
11257 goto alpha_assertions;
11261 if ( memEQs(start_verb, verb_len, "nlb")
11262 || memEQs(start_verb, verb_len, "negative_lookbehind"))
11265 goto lookbehind_alpha_assertions;
11267 else if ( memEQs(start_verb, verb_len, "nla")
11268 || memEQs(start_verb, verb_len, "negative_lookahead"))
11271 goto alpha_assertions;
11275 if ( memEQs(start_verb, verb_len, "sr")
11276 || memEQs(start_verb, verb_len, "script_run"))
11278 regnode_offset atomic;
11284 /* This indicates Unicode rules. */
11285 REQUIRE_UNI_RULES(flagp, 0);
11291 RExC_parse = start_arg;
11293 if (RExC_in_script_run) {
11295 /* Nested script runs are treated as no-ops, because
11296 * if the nested one fails, the outer one must as
11297 * well. It could fail sooner, and avoid (??{} with
11298 * side effects, but that is explicitly documented as
11299 * undefined behavior. */
11303 if (paren == 's') {
11308 /* But, the atomic part of a nested atomic script run
11309 * isn't a no-op, but can be treated just like a '(?>'
11315 /* By doing this here, we avoid extra warnings for nested
11317 ckWARNexperimental(RExC_parse,
11318 WARN_EXPERIMENTAL__SCRIPT_RUN,
11319 "The script_run feature is experimental");
11321 if (paren == 's') {
11322 /* Here, we're starting a new regular script run */
11323 ret = reg_node(pRExC_state, SROPEN);
11324 RExC_in_script_run = 1;
11329 /* Here, we are starting an atomic script run. This is
11330 * handled by recursing to deal with the atomic portion
11331 * separately, enclosed in SROPEN ... SRCLOSE nodes */
11333 ret = reg_node(pRExC_state, SROPEN);
11335 RExC_in_script_run = 1;
11337 atomic = reg(pRExC_state, 'r', &flags, depth);
11338 if (flags & (RESTART_PARSE|NEED_UTF8)) {
11339 *flagp = flags & (RESTART_PARSE|NEED_UTF8);
11343 if (! REGTAIL(pRExC_state, ret, atomic)) {
11344 REQUIRE_BRANCHJ(flagp, 0);
11347 if (! REGTAIL(pRExC_state, atomic, reg_node(pRExC_state,
11350 REQUIRE_BRANCHJ(flagp, 0);
11353 RExC_in_script_run = 0;
11359 lookbehind_alpha_assertions:
11360 RExC_seen |= REG_LOOKBEHIND_SEEN;
11361 RExC_in_lookbehind++;
11365 ckWARNexperimental(RExC_parse,
11366 WARN_EXPERIMENTAL__ALPHA_ASSERTIONS,
11367 "The alpha_assertions feature is experimental");
11369 RExC_seen_zerolen++;
11375 /* An empty negative lookahead assertion simply is failure */
11376 if (paren == 'A' && RExC_parse == start_arg) {
11377 ret=reganode(pRExC_state, OPFAIL, 0);
11378 nextchar(pRExC_state);
11382 RExC_parse = start_arg;
11387 "'(*%" UTF8f "' requires a terminating ':'",
11388 UTF8fARG(UTF, verb_len, start_verb));
11389 NOT_REACHED; /*NOTREACHED*/
11391 } /* End of switch */
11394 ? UTF8_SAFE_SKIP(RExC_parse, RExC_end)
11396 if (has_upper || verb_len == 0) {
11398 "Unknown verb pattern '%" UTF8f "'",
11399 UTF8fARG(UTF, verb_len, start_verb));
11403 "Unknown '(*...)' construct '%" UTF8f "'",
11404 UTF8fARG(UTF, verb_len, start_verb));
11407 if ( RExC_parse == start_arg ) {
11410 if ( arg_required && !start_arg ) {
11411 vFAIL3("Verb pattern '%.*s' has a mandatory argument",
11412 verb_len, start_verb);
11414 if (internal_argval == -1) {
11415 ret = reganode(pRExC_state, op, 0);
11417 ret = reg2Lanode(pRExC_state, op, 0, internal_argval);
11419 RExC_seen |= REG_VERBARG_SEEN;
11421 SV *sv = newSVpvn( start_arg,
11422 RExC_parse - start_arg);
11423 ARG(REGNODE_p(ret)) = add_data( pRExC_state,
11424 STR_WITH_LEN("S"));
11425 RExC_rxi->data->data[ARG(REGNODE_p(ret))]=(void*)sv;
11426 FLAGS(REGNODE_p(ret)) = 1;
11428 FLAGS(REGNODE_p(ret)) = 0;
11430 if ( internal_argval != -1 )
11431 ARG2L_SET(REGNODE_p(ret), internal_argval);
11432 nextchar(pRExC_state);
11435 else if (*RExC_parse == '?') { /* (?...) */
11436 bool is_logical = 0;
11437 const char * const seqstart = RExC_parse;
11438 const char * endptr;
11439 if (has_intervening_patws) {
11441 vFAIL("In '(?...)', the '(' and '?' must be adjacent");
11444 RExC_parse++; /* past the '?' */
11445 paren = *RExC_parse; /* might be a trailing NUL, if not
11447 RExC_parse += UTF ? UTF8SKIP(RExC_parse) : 1;
11448 if (RExC_parse > RExC_end) {
11451 ret = 0; /* For look-ahead/behind. */
11454 case 'P': /* (?P...) variants for those used to PCRE/Python */
11455 paren = *RExC_parse;
11456 if ( paren == '<') { /* (?P<...>) named capture */
11458 if (RExC_parse >= RExC_end) {
11459 vFAIL("Sequence (?P<... not terminated");
11461 goto named_capture;
11463 else if (paren == '>') { /* (?P>name) named recursion */
11465 if (RExC_parse >= RExC_end) {
11466 vFAIL("Sequence (?P>... not terminated");
11468 goto named_recursion;
11470 else if (paren == '=') { /* (?P=...) named backref */
11472 return handle_named_backref(pRExC_state, flagp,
11475 RExC_parse += SKIP_IF_CHAR(RExC_parse, RExC_end);
11476 /* diag_listed_as: Sequence (?%s...) not recognized in regex; marked by <-- HERE in m/%s/ */
11477 vFAIL3("Sequence (%.*s...) not recognized",
11478 RExC_parse-seqstart, seqstart);
11479 NOT_REACHED; /*NOTREACHED*/
11480 case '<': /* (?<...) */
11481 if (*RExC_parse == '!')
11483 else if (*RExC_parse != '=')
11490 case '\'': /* (?'...') */
11491 name_start = RExC_parse;
11492 svname = reg_scan_name(pRExC_state, REG_RSN_RETURN_NAME);
11493 if ( RExC_parse == name_start
11494 || RExC_parse >= RExC_end
11495 || *RExC_parse != paren)
11497 vFAIL2("Sequence (?%c... not terminated",
11498 paren=='>' ? '<' : paren);
11503 if (!svname) /* shouldn't happen */
11505 "panic: reg_scan_name returned NULL");
11506 if (!RExC_paren_names) {
11507 RExC_paren_names= newHV();
11508 sv_2mortal(MUTABLE_SV(RExC_paren_names));
11510 RExC_paren_name_list= newAV();
11511 sv_2mortal(MUTABLE_SV(RExC_paren_name_list));
11514 he_str = hv_fetch_ent( RExC_paren_names, svname, 1, 0 );
11516 sv_dat = HeVAL(he_str);
11518 /* croak baby croak */
11520 "panic: paren_name hash element allocation failed");
11521 } else if ( SvPOK(sv_dat) ) {
11522 /* (?|...) can mean we have dupes so scan to check
11523 its already been stored. Maybe a flag indicating
11524 we are inside such a construct would be useful,
11525 but the arrays are likely to be quite small, so
11526 for now we punt -- dmq */
11527 IV count = SvIV(sv_dat);
11528 I32 *pv = (I32*)SvPVX(sv_dat);
11530 for ( i = 0 ; i < count ; i++ ) {
11531 if ( pv[i] == RExC_npar ) {
11537 pv = (I32*)SvGROW(sv_dat,
11538 SvCUR(sv_dat) + sizeof(I32)+1);
11539 SvCUR_set(sv_dat, SvCUR(sv_dat) + sizeof(I32));
11540 pv[count] = RExC_npar;
11541 SvIV_set(sv_dat, SvIVX(sv_dat) + 1);
11544 (void)SvUPGRADE(sv_dat, SVt_PVNV);
11545 sv_setpvn(sv_dat, (char *)&(RExC_npar),
11548 SvIV_set(sv_dat, 1);
11551 /* Yes this does cause a memory leak in debugging Perls
11553 if (!av_store(RExC_paren_name_list,
11554 RExC_npar, SvREFCNT_inc_NN(svname)))
11555 SvREFCNT_dec_NN(svname);
11558 /*sv_dump(sv_dat);*/
11560 nextchar(pRExC_state);
11562 goto capturing_parens;
11565 RExC_seen |= REG_LOOKBEHIND_SEEN;
11566 RExC_in_lookbehind++;
11568 if (RExC_parse >= RExC_end) {
11569 vFAIL("Sequence (?... not terminated");
11571 RExC_seen_zerolen++;
11573 case '=': /* (?=...) */
11574 RExC_seen_zerolen++;
11575 RExC_in_lookahead++;
11577 case '!': /* (?!...) */
11578 RExC_seen_zerolen++;
11579 /* check if we're really just a "FAIL" assertion */
11580 skip_to_be_ignored_text(pRExC_state, &RExC_parse,
11581 FALSE /* Don't force to /x */ );
11582 if (*RExC_parse == ')') {
11583 ret=reganode(pRExC_state, OPFAIL, 0);
11584 nextchar(pRExC_state);
11588 case '|': /* (?|...) */
11589 /* branch reset, behave like a (?:...) except that
11590 buffers in alternations share the same numbers */
11592 after_freeze = freeze_paren = RExC_npar;
11594 /* XXX This construct currently requires an extra pass.
11595 * Investigation would be required to see if that could be
11597 REQUIRE_PARENS_PASS;
11599 case ':': /* (?:...) */
11600 case '>': /* (?>...) */
11602 case '$': /* (?$...) */
11603 case '@': /* (?@...) */
11604 vFAIL2("Sequence (?%c...) not implemented", (int)paren);
11606 case '0' : /* (?0) */
11607 case 'R' : /* (?R) */
11608 if (RExC_parse == RExC_end || *RExC_parse != ')')
11609 FAIL("Sequence (?R) not terminated");
11611 RExC_seen |= REG_RECURSE_SEEN;
11613 /* XXX These constructs currently require an extra pass.
11614 * It probably could be changed */
11615 REQUIRE_PARENS_PASS;
11617 *flagp |= POSTPONED;
11618 goto gen_recurse_regop;
11620 /* named and numeric backreferences */
11621 case '&': /* (?&NAME) */
11622 parse_start = RExC_parse - 1;
11625 SV *sv_dat = reg_scan_name(pRExC_state,
11626 REG_RSN_RETURN_DATA);
11627 num = sv_dat ? *((I32 *)SvPVX(sv_dat)) : 0;
11629 if (RExC_parse >= RExC_end || *RExC_parse != ')')
11630 vFAIL("Sequence (?&... not terminated");
11631 goto gen_recurse_regop;
11634 if (! inRANGE(RExC_parse[0], '1', '9')) {
11636 vFAIL("Illegal pattern");
11638 goto parse_recursion;
11640 case '-': /* (?-1) */
11641 if (! inRANGE(RExC_parse[0], '1', '9')) {
11642 RExC_parse--; /* rewind to let it be handled later */
11646 case '1': case '2': case '3': case '4': /* (?1) */
11647 case '5': case '6': case '7': case '8': case '9':
11648 RExC_parse = (char *) seqstart + 1; /* Point to the digit */
11651 bool is_neg = FALSE;
11653 parse_start = RExC_parse - 1; /* MJD */
11654 if (*RExC_parse == '-') {
11659 if (grok_atoUV(RExC_parse, &unum, &endptr)
11663 RExC_parse = (char*)endptr;
11667 /* Some limit for num? */
11671 if (*RExC_parse!=')')
11672 vFAIL("Expecting close bracket");
11675 if ( paren == '-' ) {
11677 Diagram of capture buffer numbering.
11678 Top line is the normal capture buffer numbers
11679 Bottom line is the negative indexing as from
11683 /(a(x)y)(a(b(c(?-2)d)e)f)(g(h))/
11687 num = RExC_npar + num;
11690 /* It might be a forward reference; we can't fail until
11691 * we know, by completing the parse to get all the
11692 * groups, and then reparsing */
11693 if (ALL_PARENS_COUNTED) {
11695 vFAIL("Reference to nonexistent group");
11698 REQUIRE_PARENS_PASS;
11701 } else if ( paren == '+' ) {
11702 num = RExC_npar + num - 1;
11704 /* We keep track how many GOSUB items we have produced.
11705 To start off the ARG2L() of the GOSUB holds its "id",
11706 which is used later in conjunction with RExC_recurse
11707 to calculate the offset we need to jump for the GOSUB,
11708 which it will store in the final representation.
11709 We have to defer the actual calculation until much later
11710 as the regop may move.
11713 ret = reg2Lanode(pRExC_state, GOSUB, num, RExC_recurse_count);
11714 if (num >= RExC_npar) {
11716 /* It might be a forward reference; we can't fail until we
11717 * know, by completing the parse to get all the groups, and
11718 * then reparsing */
11719 if (ALL_PARENS_COUNTED) {
11720 if (num >= RExC_total_parens) {
11722 vFAIL("Reference to nonexistent group");
11726 REQUIRE_PARENS_PASS;
11729 RExC_recurse_count++;
11730 DEBUG_OPTIMISE_MORE_r(Perl_re_printf( aTHX_
11731 "%*s%*s Recurse #%" UVuf " to %" IVdf "\n",
11732 22, "| |", (int)(depth * 2 + 1), "",
11733 (UV)ARG(REGNODE_p(ret)),
11734 (IV)ARG2L(REGNODE_p(ret))));
11735 RExC_seen |= REG_RECURSE_SEEN;
11737 Set_Node_Length(REGNODE_p(ret),
11738 1 + regarglen[OP(REGNODE_p(ret))]); /* MJD */
11739 Set_Node_Offset(REGNODE_p(ret), parse_start); /* MJD */
11741 *flagp |= POSTPONED;
11742 assert(*RExC_parse == ')');
11743 nextchar(pRExC_state);
11748 case '?': /* (??...) */
11750 if (*RExC_parse != '{') {
11751 RExC_parse += SKIP_IF_CHAR(RExC_parse, RExC_end);
11752 /* diag_listed_as: Sequence (?%s...) not recognized in regex; marked by <-- HERE in m/%s/ */
11754 "Sequence (%" UTF8f "...) not recognized",
11755 UTF8fARG(UTF, RExC_parse-seqstart, seqstart));
11756 NOT_REACHED; /*NOTREACHED*/
11758 *flagp |= POSTPONED;
11762 case '{': /* (?{...}) */
11765 struct reg_code_block *cb;
11768 RExC_seen_zerolen++;
11770 if ( !pRExC_state->code_blocks
11771 || pRExC_state->code_index
11772 >= pRExC_state->code_blocks->count
11773 || pRExC_state->code_blocks->cb[pRExC_state->code_index].start
11774 != (STRLEN)((RExC_parse -3 - (is_logical ? 1 : 0))
11777 if (RExC_pm_flags & PMf_USE_RE_EVAL)
11778 FAIL("panic: Sequence (?{...}): no code block found\n");
11779 FAIL("Eval-group not allowed at runtime, use re 'eval'");
11781 /* this is a pre-compiled code block (?{...}) */
11782 cb = &pRExC_state->code_blocks->cb[pRExC_state->code_index];
11783 RExC_parse = RExC_start + cb->end;
11785 if (cb->src_regex) {
11786 n = add_data(pRExC_state, STR_WITH_LEN("rl"));
11787 RExC_rxi->data->data[n] =
11788 (void*)SvREFCNT_inc((SV*)cb->src_regex);
11789 RExC_rxi->data->data[n+1] = (void*)o;
11792 n = add_data(pRExC_state,
11793 (RExC_pm_flags & PMf_HAS_CV) ? "L" : "l", 1);
11794 RExC_rxi->data->data[n] = (void*)o;
11796 pRExC_state->code_index++;
11797 nextchar(pRExC_state);
11800 regnode_offset eval;
11801 ret = reg_node(pRExC_state, LOGICAL);
11803 eval = reg2Lanode(pRExC_state, EVAL,
11806 /* for later propagation into (??{})
11808 RExC_flags & RXf_PMf_COMPILETIME
11810 FLAGS(REGNODE_p(ret)) = 2;
11811 if (! REGTAIL(pRExC_state, ret, eval)) {
11812 REQUIRE_BRANCHJ(flagp, 0);
11814 /* deal with the length of this later - MJD */
11817 ret = reg2Lanode(pRExC_state, EVAL, n, 0);
11818 Set_Node_Length(REGNODE_p(ret), RExC_parse - parse_start + 1);
11819 Set_Node_Offset(REGNODE_p(ret), parse_start);
11822 case '(': /* (?(?{...})...) and (?(?=...)...) */
11825 const int DEFINE_len = sizeof("DEFINE") - 1;
11826 if ( RExC_parse < RExC_end - 1
11827 && ( ( RExC_parse[0] == '?' /* (?(?...)) */
11828 && ( RExC_parse[1] == '='
11829 || RExC_parse[1] == '!'
11830 || RExC_parse[1] == '<'
11831 || RExC_parse[1] == '{'))
11832 || ( RExC_parse[0] == '*' /* (?(*...)) */
11833 && ( memBEGINs(RExC_parse + 1,
11834 (Size_t) (RExC_end - (RExC_parse + 1)),
11836 || memBEGINs(RExC_parse + 1,
11837 (Size_t) (RExC_end - (RExC_parse + 1)),
11839 || memBEGINs(RExC_parse + 1,
11840 (Size_t) (RExC_end - (RExC_parse + 1)),
11842 || memBEGINs(RExC_parse + 1,
11843 (Size_t) (RExC_end - (RExC_parse + 1)),
11845 || memBEGINs(RExC_parse + 1,
11846 (Size_t) (RExC_end - (RExC_parse + 1)),
11847 "positive_lookahead:")
11848 || memBEGINs(RExC_parse + 1,
11849 (Size_t) (RExC_end - (RExC_parse + 1)),
11850 "positive_lookbehind:")
11851 || memBEGINs(RExC_parse + 1,
11852 (Size_t) (RExC_end - (RExC_parse + 1)),
11853 "negative_lookahead:")
11854 || memBEGINs(RExC_parse + 1,
11855 (Size_t) (RExC_end - (RExC_parse + 1)),
11856 "negative_lookbehind:"))))
11857 ) { /* Lookahead or eval. */
11859 regnode_offset tail;
11861 ret = reg_node(pRExC_state, LOGICAL);
11862 FLAGS(REGNODE_p(ret)) = 1;
11864 tail = reg(pRExC_state, 1, &flag, depth+1);
11865 RETURN_FAIL_ON_RESTART(flag, flagp);
11866 if (! REGTAIL(pRExC_state, ret, tail)) {
11867 REQUIRE_BRANCHJ(flagp, 0);
11871 else if ( RExC_parse[0] == '<' /* (?(<NAME>)...) */
11872 || RExC_parse[0] == '\'' ) /* (?('NAME')...) */
11874 char ch = RExC_parse[0] == '<' ? '>' : '\'';
11875 char *name_start= RExC_parse++;
11877 SV *sv_dat=reg_scan_name(pRExC_state, REG_RSN_RETURN_DATA);
11878 if ( RExC_parse == name_start
11879 || RExC_parse >= RExC_end
11880 || *RExC_parse != ch)
11882 vFAIL2("Sequence (?(%c... not terminated",
11883 (ch == '>' ? '<' : ch));
11887 num = add_data( pRExC_state, STR_WITH_LEN("S"));
11888 RExC_rxi->data->data[num]=(void*)sv_dat;
11889 SvREFCNT_inc_simple_void_NN(sv_dat);
11891 ret = reganode(pRExC_state, GROUPPN, num);
11892 goto insert_if_check_paren;
11894 else if (memBEGINs(RExC_parse,
11895 (STRLEN) (RExC_end - RExC_parse),
11898 ret = reganode(pRExC_state, DEFINEP, 0);
11899 RExC_parse += DEFINE_len;
11901 goto insert_if_check_paren;
11903 else if (RExC_parse[0] == 'R') {
11905 /* parno == 0 => /(?(R)YES|NO)/ "in any form of recursion OR eval"
11906 * parno == 1 => /(?(R0)YES|NO)/ "in GOSUB (?0) / (?R)"
11907 * parno == 2 => /(?(R1)YES|NO)/ "in GOSUB (?1) (parno-1)"
11910 if (RExC_parse[0] == '0') {
11914 else if (inRANGE(RExC_parse[0], '1', '9')) {
11917 if (grok_atoUV(RExC_parse, &uv, &endptr)
11920 parno = (I32)uv + 1;
11921 RExC_parse = (char*)endptr;
11923 /* else "Switch condition not recognized" below */
11924 } else if (RExC_parse[0] == '&') {
11927 sv_dat = reg_scan_name(pRExC_state,
11928 REG_RSN_RETURN_DATA);
11930 parno = 1 + *((I32 *)SvPVX(sv_dat));
11932 ret = reganode(pRExC_state, INSUBP, parno);
11933 goto insert_if_check_paren;
11935 else if (inRANGE(RExC_parse[0], '1', '9')) {
11940 if (grok_atoUV(RExC_parse, &uv, &endptr)
11944 RExC_parse = (char*)endptr;
11947 vFAIL("panic: grok_atoUV returned FALSE");
11949 ret = reganode(pRExC_state, GROUPP, parno);
11951 insert_if_check_paren:
11952 if (UCHARAT(RExC_parse) != ')') {
11954 ? UTF8_SAFE_SKIP(RExC_parse, RExC_end)
11956 vFAIL("Switch condition not recognized");
11958 nextchar(pRExC_state);
11960 if (! REGTAIL(pRExC_state, ret, reganode(pRExC_state,
11963 REQUIRE_BRANCHJ(flagp, 0);
11965 br = regbranch(pRExC_state, &flags, 1, depth+1);
11967 RETURN_FAIL_ON_RESTART(flags,flagp);
11968 FAIL2("panic: regbranch returned failure, flags=%#" UVxf,
11971 if (! REGTAIL(pRExC_state, br, reganode(pRExC_state,
11974 REQUIRE_BRANCHJ(flagp, 0);
11976 c = UCHARAT(RExC_parse);
11977 nextchar(pRExC_state);
11978 if (flags&HASWIDTH)
11979 *flagp |= HASWIDTH;
11982 vFAIL("(?(DEFINE)....) does not allow branches");
11984 /* Fake one for optimizer. */
11985 lastbr = reganode(pRExC_state, IFTHEN, 0);
11987 if (!regbranch(pRExC_state, &flags, 1, depth+1)) {
11988 RETURN_FAIL_ON_RESTART(flags, flagp);
11989 FAIL2("panic: regbranch returned failure, flags=%#" UVxf,
11992 if (! REGTAIL(pRExC_state, ret, lastbr)) {
11993 REQUIRE_BRANCHJ(flagp, 0);
11995 if (flags&HASWIDTH)
11996 *flagp |= HASWIDTH;
11997 c = UCHARAT(RExC_parse);
11998 nextchar(pRExC_state);
12003 if (RExC_parse >= RExC_end)
12004 vFAIL("Switch (?(condition)... not terminated");
12006 vFAIL("Switch (?(condition)... contains too many branches");
12008 ender = reg_node(pRExC_state, TAIL);
12009 if (! REGTAIL(pRExC_state, br, ender)) {
12010 REQUIRE_BRANCHJ(flagp, 0);
12013 if (! REGTAIL(pRExC_state, lastbr, ender)) {
12014 REQUIRE_BRANCHJ(flagp, 0);
12016 if (! REGTAIL(pRExC_state,
12019 NEXTOPER(REGNODE_p(lastbr)))),
12022 REQUIRE_BRANCHJ(flagp, 0);
12026 if (! REGTAIL(pRExC_state, ret, ender)) {
12027 REQUIRE_BRANCHJ(flagp, 0);
12029 #if 0 /* Removing this doesn't cause failures in the test suite -- khw */
12030 RExC_size++; /* XXX WHY do we need this?!!
12031 For large programs it seems to be required
12032 but I can't figure out why. -- dmq*/
12037 ? UTF8_SAFE_SKIP(RExC_parse, RExC_end)
12039 vFAIL("Unknown switch condition (?(...))");
12041 case '[': /* (?[ ... ]) */
12042 return handle_regex_sets(pRExC_state, NULL, flagp, depth+1,
12044 case 0: /* A NUL */
12045 RExC_parse--; /* for vFAIL to print correctly */
12046 vFAIL("Sequence (? incomplete");
12050 if (RExC_strict) { /* [perl #132851] */
12051 ckWARNreg(RExC_parse, "Empty (?) without any modifiers");
12054 default: /* e.g., (?i) */
12055 RExC_parse = (char *) seqstart + 1;
12057 parse_lparen_question_flags(pRExC_state);
12058 if (UCHARAT(RExC_parse) != ':') {
12059 if (RExC_parse < RExC_end)
12060 nextchar(pRExC_state);
12065 nextchar(pRExC_state);
12071 if (*RExC_parse == '{') {
12072 ckWARNregdep(RExC_parse + 1,
12073 "Unescaped left brace in regex is "
12074 "deprecated here (and will be fatal "
12075 "in Perl 5.32), passed through");
12077 /* Not bothering to indent here, as the above 'else' is temporary
12079 if (!(RExC_flags & RXf_PMf_NOCAPTURE)) { /* (...) */
12083 if (! ALL_PARENS_COUNTED) {
12084 /* If we are in our first pass through (and maybe only pass),
12085 * we need to allocate memory for the capturing parentheses
12089 if (!RExC_parens_buf_size) {
12090 /* first guess at number of parens we might encounter */
12091 RExC_parens_buf_size = 10;
12093 /* setup RExC_open_parens, which holds the address of each
12094 * OPEN tag, and to make things simpler for the 0 index the
12095 * start of the program - this is used later for offsets */
12096 Newxz(RExC_open_parens, RExC_parens_buf_size,
12098 RExC_open_parens[0] = 1; /* +1 for REG_MAGIC */
12100 /* setup RExC_close_parens, which holds the address of each
12101 * CLOSE tag, and to make things simpler for the 0 index
12102 * the end of the program - this is used later for offsets
12104 Newxz(RExC_close_parens, RExC_parens_buf_size,
12106 /* we dont know where end op starts yet, so we dont need to
12107 * set RExC_close_parens[0] like we do RExC_open_parens[0]
12110 else if (RExC_npar > RExC_parens_buf_size) {
12111 I32 old_size = RExC_parens_buf_size;
12113 RExC_parens_buf_size *= 2;
12115 Renew(RExC_open_parens, RExC_parens_buf_size,
12117 Zero(RExC_open_parens + old_size,
12118 RExC_parens_buf_size - old_size, regnode_offset);
12120 Renew(RExC_close_parens, RExC_parens_buf_size,
12122 Zero(RExC_close_parens + old_size,
12123 RExC_parens_buf_size - old_size, regnode_offset);
12127 ret = reganode(pRExC_state, OPEN, parno);
12128 if (!RExC_nestroot)
12129 RExC_nestroot = parno;
12130 if (RExC_open_parens && !RExC_open_parens[parno])
12132 DEBUG_OPTIMISE_MORE_r(Perl_re_printf( aTHX_
12133 "%*s%*s Setting open paren #%" IVdf " to %d\n",
12134 22, "| |", (int)(depth * 2 + 1), "",
12136 RExC_open_parens[parno]= ret;
12139 Set_Node_Length(REGNODE_p(ret), 1); /* MJD */
12140 Set_Node_Offset(REGNODE_p(ret), RExC_parse); /* MJD */
12143 /* with RXf_PMf_NOCAPTURE treat (...) as (?:...) */
12153 /* Pick up the branches, linking them together. */
12154 parse_start = RExC_parse; /* MJD */
12155 br = regbranch(pRExC_state, &flags, 1, depth+1);
12157 /* branch_len = (paren != 0); */
12160 RETURN_FAIL_ON_RESTART(flags, flagp);
12161 FAIL2("panic: regbranch returned failure, flags=%#" UVxf, (UV) flags);
12163 if (*RExC_parse == '|') {
12164 if (RExC_use_BRANCHJ) {
12165 reginsert(pRExC_state, BRANCHJ, br, depth+1);
12168 reginsert(pRExC_state, BRANCH, br, depth+1);
12169 Set_Node_Length(REGNODE_p(br), paren != 0);
12170 Set_Node_Offset_To_R(br, parse_start-RExC_start);
12174 else if (paren == ':') {
12175 *flagp |= flags&SIMPLE;
12177 if (is_open) { /* Starts with OPEN. */
12178 if (! REGTAIL(pRExC_state, ret, br)) { /* OPEN -> first. */
12179 REQUIRE_BRANCHJ(flagp, 0);
12182 else if (paren != '?') /* Not Conditional */
12184 *flagp |= flags & (SPSTART | HASWIDTH | POSTPONED);
12186 while (*RExC_parse == '|') {
12187 if (RExC_use_BRANCHJ) {
12190 ender = reganode(pRExC_state, LONGJMP, 0);
12192 /* Append to the previous. */
12193 shut_gcc_up = REGTAIL(pRExC_state,
12194 REGNODE_OFFSET(NEXTOPER(NEXTOPER(REGNODE_p(lastbr)))),
12196 PERL_UNUSED_VAR(shut_gcc_up);
12198 nextchar(pRExC_state);
12199 if (freeze_paren) {
12200 if (RExC_npar > after_freeze)
12201 after_freeze = RExC_npar;
12202 RExC_npar = freeze_paren;
12204 br = regbranch(pRExC_state, &flags, 0, depth+1);
12207 RETURN_FAIL_ON_RESTART(flags, flagp);
12208 FAIL2("panic: regbranch returned failure, flags=%#" UVxf, (UV) flags);
12210 if (! REGTAIL(pRExC_state, lastbr, br)) { /* BRANCH -> BRANCH. */
12211 REQUIRE_BRANCHJ(flagp, 0);
12214 *flagp |= flags & (SPSTART | HASWIDTH | POSTPONED);
12217 if (have_branch || paren != ':') {
12220 /* Make a closing node, and hook it on the end. */
12223 ender = reg_node(pRExC_state, TAIL);
12226 ender = reganode(pRExC_state, CLOSE, parno);
12227 if ( RExC_close_parens ) {
12228 DEBUG_OPTIMISE_MORE_r(Perl_re_printf( aTHX_
12229 "%*s%*s Setting close paren #%" IVdf " to %d\n",
12230 22, "| |", (int)(depth * 2 + 1), "",
12231 (IV)parno, ender));
12232 RExC_close_parens[parno]= ender;
12233 if (RExC_nestroot == parno)
12236 Set_Node_Offset(REGNODE_p(ender), RExC_parse+1); /* MJD */
12237 Set_Node_Length(REGNODE_p(ender), 1); /* MJD */
12240 ender = reg_node(pRExC_state, SRCLOSE);
12241 RExC_in_script_run = 0;
12251 *flagp &= ~HASWIDTH;
12253 case 't': /* aTomic */
12255 ender = reg_node(pRExC_state, SUCCEED);
12258 ender = reg_node(pRExC_state, END);
12259 assert(!RExC_end_op); /* there can only be one! */
12260 RExC_end_op = REGNODE_p(ender);
12261 if (RExC_close_parens) {
12262 DEBUG_OPTIMISE_MORE_r(Perl_re_printf( aTHX_
12263 "%*s%*s Setting close paren #0 (END) to %d\n",
12264 22, "| |", (int)(depth * 2 + 1), "",
12267 RExC_close_parens[0]= ender;
12272 DEBUG_PARSE_MSG("lsbr");
12273 regprop(RExC_rx, RExC_mysv1, REGNODE_p(lastbr), NULL, pRExC_state);
12274 regprop(RExC_rx, RExC_mysv2, REGNODE_p(ender), NULL, pRExC_state);
12275 Perl_re_printf( aTHX_ "~ tying lastbr %s (%" IVdf ") to ender %s (%" IVdf ") offset %" IVdf "\n",
12276 SvPV_nolen_const(RExC_mysv1),
12278 SvPV_nolen_const(RExC_mysv2),
12280 (IV)(ender - lastbr)
12283 if (! REGTAIL(pRExC_state, lastbr, ender)) {
12284 REQUIRE_BRANCHJ(flagp, 0);
12288 char is_nothing= 1;
12290 RExC_seen |= REG_TOP_LEVEL_BRANCHES_SEEN;
12292 /* Hook the tails of the branches to the closing node. */
12293 for (br = REGNODE_p(ret); br; br = regnext(br)) {
12294 const U8 op = PL_regkind[OP(br)];
12295 if (op == BRANCH) {
12296 if (! REGTAIL_STUDY(pRExC_state,
12297 REGNODE_OFFSET(NEXTOPER(br)),
12300 REQUIRE_BRANCHJ(flagp, 0);
12302 if ( OP(NEXTOPER(br)) != NOTHING
12303 || regnext(NEXTOPER(br)) != REGNODE_p(ender))
12306 else if (op == BRANCHJ) {
12307 bool shut_gcc_up = REGTAIL_STUDY(pRExC_state,
12308 REGNODE_OFFSET(NEXTOPER(NEXTOPER(br))),
12310 PERL_UNUSED_VAR(shut_gcc_up);
12311 /* for now we always disable this optimisation * /
12312 if ( OP(NEXTOPER(NEXTOPER(br))) != NOTHING
12313 || regnext(NEXTOPER(NEXTOPER(br))) != REGNODE_p(ender))
12319 regnode * ret_as_regnode = REGNODE_p(ret);
12320 br= PL_regkind[OP(ret_as_regnode)] != BRANCH
12321 ? regnext(ret_as_regnode)
12324 DEBUG_PARSE_MSG("NADA");
12325 regprop(RExC_rx, RExC_mysv1, ret_as_regnode,
12326 NULL, pRExC_state);
12327 regprop(RExC_rx, RExC_mysv2, REGNODE_p(ender),
12328 NULL, pRExC_state);
12329 Perl_re_printf( aTHX_ "~ converting ret %s (%" IVdf ") to ender %s (%" IVdf ") offset %" IVdf "\n",
12330 SvPV_nolen_const(RExC_mysv1),
12331 (IV)REG_NODE_NUM(ret_as_regnode),
12332 SvPV_nolen_const(RExC_mysv2),
12338 if (OP(REGNODE_p(ender)) == TAIL) {
12340 RExC_emit= REGNODE_OFFSET(br) + 1;
12343 for ( opt= br + 1; opt < REGNODE_p(ender) ; opt++ )
12344 OP(opt)= OPTIMIZED;
12345 NEXT_OFF(br)= REGNODE_p(ender) - br;
12353 /* Even/odd or x=don't care: 010101x10x */
12354 static const char parens[] = "=!aA<,>Bbt";
12355 /* flag below is set to 0 up through 'A'; 1 for larger */
12357 if (paren && (p = strchr(parens, paren))) {
12358 U8 node = ((p - parens) % 2) ? UNLESSM : IFMATCH;
12359 int flag = (p - parens) > 3;
12361 if (paren == '>' || paren == 't') {
12362 node = SUSPEND, flag = 0;
12365 reginsert(pRExC_state, node, ret, depth+1);
12366 Set_Node_Cur_Length(REGNODE_p(ret), parse_start);
12367 Set_Node_Offset(REGNODE_p(ret), parse_start + 1);
12368 FLAGS(REGNODE_p(ret)) = flag;
12369 if (! REGTAIL_STUDY(pRExC_state, ret, reg_node(pRExC_state, TAIL)))
12371 REQUIRE_BRANCHJ(flagp, 0);
12376 /* Check for proper termination. */
12378 /* restore original flags, but keep (?p) and, if we've encountered
12379 * something in the parse that changes /d rules into /u, keep the /u */
12380 RExC_flags = oregflags | (RExC_flags & RXf_PMf_KEEPCOPY);
12381 if (DEPENDS_SEMANTICS && RExC_uni_semantics) {
12382 set_regex_charset(&RExC_flags, REGEX_UNICODE_CHARSET);
12384 if (RExC_parse >= RExC_end || UCHARAT(RExC_parse) != ')') {
12385 RExC_parse = oregcomp_parse;
12386 vFAIL("Unmatched (");
12388 nextchar(pRExC_state);
12390 else if (!paren && RExC_parse < RExC_end) {
12391 if (*RExC_parse == ')') {
12393 vFAIL("Unmatched )");
12396 FAIL("Junk on end of regexp"); /* "Can't happen". */
12397 NOT_REACHED; /* NOTREACHED */
12400 if (RExC_in_lookbehind) {
12401 RExC_in_lookbehind--;
12403 if (RExC_in_lookahead) {
12404 RExC_in_lookahead--;
12406 if (after_freeze > RExC_npar)
12407 RExC_npar = after_freeze;
12412 - regbranch - one alternative of an | operator
12414 * Implements the concatenation operator.
12416 * On success, returns the offset at which any next node should be placed into
12417 * the regex engine program being compiled.
12419 * Returns 0 otherwise, setting flagp to RESTART_PARSE if the parse needs
12420 * to be restarted, or'd with NEED_UTF8 if the pattern needs to be upgraded to
12423 STATIC regnode_offset
12424 S_regbranch(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, I32 first, U32 depth)
12426 regnode_offset ret;
12427 regnode_offset chain = 0;
12428 regnode_offset latest;
12429 I32 flags = 0, c = 0;
12430 GET_RE_DEBUG_FLAGS_DECL;
12432 PERL_ARGS_ASSERT_REGBRANCH;
12434 DEBUG_PARSE("brnc");
12439 if (RExC_use_BRANCHJ)
12440 ret = reganode(pRExC_state, BRANCHJ, 0);
12442 ret = reg_node(pRExC_state, BRANCH);
12443 Set_Node_Length(REGNODE_p(ret), 1);
12447 *flagp = WORST; /* Tentatively. */
12449 skip_to_be_ignored_text(pRExC_state, &RExC_parse,
12450 FALSE /* Don't force to /x */ );
12451 while (RExC_parse < RExC_end && *RExC_parse != '|' && *RExC_parse != ')') {
12452 flags &= ~TRYAGAIN;
12453 latest = regpiece(pRExC_state, &flags, depth+1);
12455 if (flags & TRYAGAIN)
12457 RETURN_FAIL_ON_RESTART(flags, flagp);
12458 FAIL2("panic: regpiece returned failure, flags=%#" UVxf, (UV) flags);
12462 *flagp |= flags&(HASWIDTH|POSTPONED);
12463 if (chain == 0) /* First piece. */
12464 *flagp |= flags&SPSTART;
12466 /* FIXME adding one for every branch after the first is probably
12467 * excessive now we have TRIE support. (hv) */
12469 if (! REGTAIL(pRExC_state, chain, latest)) {
12470 /* XXX We could just redo this branch, but figuring out what
12471 * bookkeeping needs to be reset is a pain, and it's likely
12472 * that other branches that goto END will also be too large */
12473 REQUIRE_BRANCHJ(flagp, 0);
12479 if (chain == 0) { /* Loop ran zero times. */
12480 chain = reg_node(pRExC_state, NOTHING);
12485 *flagp |= flags&SIMPLE;
12492 - regpiece - something followed by possible quantifier * + ? {n,m}
12494 * Note that the branching code sequences used for ? and the general cases
12495 * of * and + are somewhat optimized: they use the same NOTHING node as
12496 * both the endmarker for their branch list and the body of the last branch.
12497 * It might seem that this node could be dispensed with entirely, but the
12498 * endmarker role is not redundant.
12500 * On success, returns the offset at which any next node should be placed into
12501 * the regex engine program being compiled.
12503 * Returns 0 otherwise, with *flagp set to indicate why:
12504 * TRYAGAIN if regatom() returns 0 with TRYAGAIN.
12505 * RESTART_PARSE if the parse needs to be restarted, or'd with
12506 * NEED_UTF8 if the pattern needs to be upgraded to UTF-8.
12508 STATIC regnode_offset
12509 S_regpiece(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, U32 depth)
12511 regnode_offset ret;
12515 const char * const origparse = RExC_parse;
12517 I32 max = REG_INFTY;
12518 #ifdef RE_TRACK_PATTERN_OFFSETS
12521 const char *maxpos = NULL;
12524 /* Save the original in case we change the emitted regop to a FAIL. */
12525 const regnode_offset orig_emit = RExC_emit;
12527 GET_RE_DEBUG_FLAGS_DECL;
12529 PERL_ARGS_ASSERT_REGPIECE;
12531 DEBUG_PARSE("piec");
12533 ret = regatom(pRExC_state, &flags, depth+1);
12535 RETURN_FAIL_ON_RESTART_OR_FLAGS(flags, flagp, TRYAGAIN);
12536 FAIL2("panic: regatom returned failure, flags=%#" UVxf, (UV) flags);
12541 if (op == '{' && regcurly(RExC_parse)) {
12543 #ifdef RE_TRACK_PATTERN_OFFSETS
12544 parse_start = RExC_parse; /* MJD */
12546 next = RExC_parse + 1;
12547 while (isDIGIT(*next) || *next == ',') {
12548 if (*next == ',') {
12556 if (*next == '}') { /* got one */
12557 const char* endptr;
12561 if (isDIGIT(*RExC_parse)) {
12563 if (!grok_atoUV(RExC_parse, &uv, &endptr))
12564 vFAIL("Invalid quantifier in {,}");
12565 if (uv >= REG_INFTY)
12566 vFAIL2("Quantifier in {,} bigger than %d", REG_INFTY - 1);
12571 if (*maxpos == ',')
12574 maxpos = RExC_parse;
12575 if (isDIGIT(*maxpos)) {
12577 if (!grok_atoUV(maxpos, &uv, &endptr))
12578 vFAIL("Invalid quantifier in {,}");
12579 if (uv >= REG_INFTY)
12580 vFAIL2("Quantifier in {,} bigger than %d", REG_INFTY - 1);
12583 max = REG_INFTY; /* meaning "infinity" */
12586 nextchar(pRExC_state);
12587 if (max < min) { /* If can't match, warn and optimize to fail
12589 reginsert(pRExC_state, OPFAIL, orig_emit, depth+1);
12590 ckWARNreg(RExC_parse, "Quantifier {n,m} with n > m can't match");
12591 NEXT_OFF(REGNODE_p(orig_emit)) =
12592 regarglen[OPFAIL] + NODE_STEP_REGNODE;
12595 else if (min == max && *RExC_parse == '?')
12597 ckWARN2reg(RExC_parse + 1,
12598 "Useless use of greediness modifier '%c'",
12603 if ((flags&SIMPLE)) {
12604 if (min == 0 && max == REG_INFTY) {
12605 reginsert(pRExC_state, STAR, ret, depth+1);
12607 RExC_seen |= REG_UNBOUNDED_QUANTIFIER_SEEN;
12610 if (min == 1 && max == REG_INFTY) {
12611 reginsert(pRExC_state, PLUS, ret, depth+1);
12613 RExC_seen |= REG_UNBOUNDED_QUANTIFIER_SEEN;
12616 MARK_NAUGHTY_EXP(2, 2);
12617 reginsert(pRExC_state, CURLY, ret, depth+1);
12618 Set_Node_Offset(REGNODE_p(ret), parse_start+1); /* MJD */
12619 Set_Node_Cur_Length(REGNODE_p(ret), parse_start);
12622 const regnode_offset w = reg_node(pRExC_state, WHILEM);
12624 FLAGS(REGNODE_p(w)) = 0;
12625 if (! REGTAIL(pRExC_state, ret, w)) {
12626 REQUIRE_BRANCHJ(flagp, 0);
12628 if (RExC_use_BRANCHJ) {
12629 reginsert(pRExC_state, LONGJMP, ret, depth+1);
12630 reginsert(pRExC_state, NOTHING, ret, depth+1);
12631 NEXT_OFF(REGNODE_p(ret)) = 3; /* Go over LONGJMP. */
12633 reginsert(pRExC_state, CURLYX, ret, depth+1);
12635 Set_Node_Offset(REGNODE_p(ret), parse_start+1);
12636 Set_Node_Length(REGNODE_p(ret),
12637 op == '{' ? (RExC_parse - parse_start) : 1);
12639 if (RExC_use_BRANCHJ)
12640 NEXT_OFF(REGNODE_p(ret)) = 3; /* Go over NOTHING to
12642 if (! REGTAIL(pRExC_state, ret, reg_node(pRExC_state,
12645 REQUIRE_BRANCHJ(flagp, 0);
12647 RExC_whilem_seen++;
12648 MARK_NAUGHTY_EXP(1, 4); /* compound interest */
12650 FLAGS(REGNODE_p(ret)) = 0;
12655 *flagp |= HASWIDTH;
12656 ARG1_SET(REGNODE_p(ret), (U16)min);
12657 ARG2_SET(REGNODE_p(ret), (U16)max);
12658 if (max == REG_INFTY)
12659 RExC_seen |= REG_UNBOUNDED_QUANTIFIER_SEEN;
12665 if (!ISMULT1(op)) {
12670 #if 0 /* Now runtime fix should be reliable. */
12672 /* if this is reinstated, don't forget to put this back into perldiag:
12674 =item Regexp *+ operand could be empty at {#} in regex m/%s/
12676 (F) The part of the regexp subject to either the * or + quantifier
12677 could match an empty string. The {#} shows in the regular
12678 expression about where the problem was discovered.
12682 if (!(flags&HASWIDTH) && op != '?')
12683 vFAIL("Regexp *+ operand could be empty");
12686 #ifdef RE_TRACK_PATTERN_OFFSETS
12687 parse_start = RExC_parse;
12689 nextchar(pRExC_state);
12691 *flagp = (op != '+') ? (WORST|SPSTART|HASWIDTH) : (WORST|HASWIDTH);
12697 else if (op == '+') {
12701 else if (op == '?') {
12706 if (!(flags&(HASWIDTH|POSTPONED)) && max > REG_INFTY/3) {
12707 ckWARN2reg(RExC_parse,
12708 "%" UTF8f " matches null string many times",
12709 UTF8fARG(UTF, (RExC_parse >= origparse
12710 ? RExC_parse - origparse
12715 if (*RExC_parse == '?') {
12716 nextchar(pRExC_state);
12717 reginsert(pRExC_state, MINMOD, ret, depth+1);
12718 if (! REGTAIL(pRExC_state, ret, ret + NODE_STEP_REGNODE)) {
12719 REQUIRE_BRANCHJ(flagp, 0);
12722 else if (*RExC_parse == '+') {
12723 regnode_offset ender;
12724 nextchar(pRExC_state);
12725 ender = reg_node(pRExC_state, SUCCEED);
12726 if (! REGTAIL(pRExC_state, ret, ender)) {
12727 REQUIRE_BRANCHJ(flagp, 0);
12729 reginsert(pRExC_state, SUSPEND, ret, depth+1);
12730 ender = reg_node(pRExC_state, TAIL);
12731 if (! REGTAIL(pRExC_state, ret, ender)) {
12732 REQUIRE_BRANCHJ(flagp, 0);
12736 if (ISMULT2(RExC_parse)) {
12738 vFAIL("Nested quantifiers");
12745 S_grok_bslash_N(pTHX_ RExC_state_t *pRExC_state,
12746 regnode_offset * node_p,
12754 /* This routine teases apart the various meanings of \N and returns
12755 * accordingly. The input parameters constrain which meaning(s) is/are valid
12756 * in the current context.
12758 * Exactly one of <node_p> and <code_point_p> must be non-NULL.
12760 * If <code_point_p> is not NULL, the context is expecting the result to be a
12761 * single code point. If this \N instance turns out to a single code point,
12762 * the function returns TRUE and sets *code_point_p to that code point.
12764 * If <node_p> is not NULL, the context is expecting the result to be one of
12765 * the things representable by a regnode. If this \N instance turns out to be
12766 * one such, the function generates the regnode, returns TRUE and sets *node_p
12767 * to point to the offset of that regnode into the regex engine program being
12770 * If this instance of \N isn't legal in any context, this function will
12771 * generate a fatal error and not return.
12773 * On input, RExC_parse should point to the first char following the \N at the
12774 * time of the call. On successful return, RExC_parse will have been updated
12775 * to point to just after the sequence identified by this routine. Also
12776 * *flagp has been updated as needed.
12778 * When there is some problem with the current context and this \N instance,
12779 * the function returns FALSE, without advancing RExC_parse, nor setting
12780 * *node_p, nor *code_point_p, nor *flagp.
12782 * If <cp_count> is not NULL, the caller wants to know the length (in code
12783 * points) that this \N sequence matches. This is set, and the input is
12784 * parsed for errors, even if the function returns FALSE, as detailed below.
12786 * There are 6 possibilities here, as detailed in the next 6 paragraphs.
12788 * Probably the most common case is for the \N to specify a single code point.
12789 * *cp_count will be set to 1, and *code_point_p will be set to that code
12792 * Another possibility is for the input to be an empty \N{}. This is no
12793 * longer accepted, and will generate a fatal error.
12795 * Another possibility is for a custom charnames handler to be in effect which
12796 * translates the input name to an empty string. *cp_count will be set to 0.
12797 * *node_p will be set to a generated NOTHING node.
12799 * Still another possibility is for the \N to mean [^\n]. *cp_count will be
12800 * set to 0. *node_p will be set to a generated REG_ANY node.
12802 * The fifth possibility is that \N resolves to a sequence of more than one
12803 * code points. *cp_count will be set to the number of code points in the
12804 * sequence. *node_p will be set to a generated node returned by this
12805 * function calling S_reg().
12807 * The final possibility is that it is premature to be calling this function;
12808 * the parse needs to be restarted. This can happen when this changes from
12809 * /d to /u rules, or when the pattern needs to be upgraded to UTF-8. The
12810 * latter occurs only when the fifth possibility would otherwise be in
12811 * effect, and is because one of those code points requires the pattern to be
12812 * recompiled as UTF-8. The function returns FALSE, and sets the
12813 * RESTART_PARSE and NEED_UTF8 flags in *flagp, as appropriate. When this
12814 * happens, the caller needs to desist from continuing parsing, and return
12815 * this information to its caller. This is not set for when there is only one
12816 * code point, as this can be called as part of an ANYOF node, and they can
12817 * store above-Latin1 code points without the pattern having to be in UTF-8.
12819 * For non-single-quoted regexes, the tokenizer has resolved character and
12820 * sequence names inside \N{...} into their Unicode values, normalizing the
12821 * result into what we should see here: '\N{U+c1.c2...}', where c1... are the
12822 * hex-represented code points in the sequence. This is done there because
12823 * the names can vary based on what charnames pragma is in scope at the time,
12824 * so we need a way to take a snapshot of what they resolve to at the time of
12825 * the original parse. [perl #56444].
12827 * That parsing is skipped for single-quoted regexes, so here we may get
12828 * '\N{NAME}', which is parsed now. If the single-quoted regex is something
12829 * like '\N{U+41}', that code point is Unicode, and has to be translated into
12830 * the native character set for non-ASCII platforms. The other possibilities
12831 * are already native, so no translation is done. */
12833 char * endbrace; /* points to '}' following the name */
12834 char* p = RExC_parse; /* Temporary */
12836 SV * substitute_parse = NULL;
12841 GET_RE_DEBUG_FLAGS_DECL;
12843 PERL_ARGS_ASSERT_GROK_BSLASH_N;
12845 GET_RE_DEBUG_FLAGS;
12847 assert(cBOOL(node_p) ^ cBOOL(code_point_p)); /* Exactly one should be set */
12848 assert(! (node_p && cp_count)); /* At most 1 should be set */
12850 if (cp_count) { /* Initialize return for the most common case */
12854 /* The [^\n] meaning of \N ignores spaces and comments under the /x
12855 * modifier. The other meanings do not, so use a temporary until we find
12856 * out which we are being called with */
12857 skip_to_be_ignored_text(pRExC_state, &p,
12858 FALSE /* Don't force to /x */ );
12860 /* Disambiguate between \N meaning a named character versus \N meaning
12861 * [^\n]. The latter is assumed when the {...} following the \N is a legal
12862 * quantifier, or if there is no '{' at all */
12863 if (*p != '{' || regcurly(p)) {
12873 *node_p = reg_node(pRExC_state, REG_ANY);
12874 *flagp |= HASWIDTH|SIMPLE;
12876 Set_Node_Length(REGNODE_p(*(node_p)), 1); /* MJD */
12880 /* The test above made sure that the next real character is a '{', but
12881 * under the /x modifier, it could be separated by space (or a comment and
12882 * \n) and this is not allowed (for consistency with \x{...} and the
12883 * tokenizer handling of \N{NAME}). */
12884 if (*RExC_parse != '{') {
12885 vFAIL("Missing braces on \\N{}");
12888 RExC_parse++; /* Skip past the '{' */
12890 endbrace = (char *) memchr(RExC_parse, '}', RExC_end - RExC_parse);
12891 if (! endbrace) { /* no trailing brace */
12892 vFAIL2("Missing right brace on \\%c{}", 'N');
12895 /* Here, we have decided it should be a named character or sequence. These
12896 * imply Unicode semantics */
12897 REQUIRE_UNI_RULES(flagp, FALSE);
12899 /* \N{_} is what toke.c returns to us to indicate a name that evaluates to
12900 * nothing at all (not allowed under strict) */
12901 if (endbrace - RExC_parse == 1 && *RExC_parse == '_') {
12902 RExC_parse = endbrace;
12904 RExC_parse++; /* Position after the "}" */
12905 vFAIL("Zero length \\N{}");
12911 nextchar(pRExC_state);
12916 *node_p = reg_node(pRExC_state, NOTHING);
12920 if (endbrace - RExC_parse < 2 || ! strBEGINs(RExC_parse, "U+")) {
12922 /* Here, the name isn't of the form U+.... This can happen if the
12923 * pattern is single-quoted, so didn't get evaluated in toke.c. Now
12924 * is the time to find out what the name means */
12926 const STRLEN name_len = endbrace - RExC_parse;
12927 SV * value_sv; /* What does this name evaluate to */
12929 const U8 * value; /* string of name's value */
12930 STRLEN value_len; /* and its length */
12932 /* RExC_unlexed_names is a hash of names that weren't evaluated by
12933 * toke.c, and their values. Make sure is initialized */
12934 if (! RExC_unlexed_names) {
12935 RExC_unlexed_names = newHV();
12938 /* If we have already seen this name in this pattern, use that. This
12939 * allows us to only call the charnames handler once per name per
12940 * pattern. A broken or malicious handler could return something
12941 * different each time, which could cause the results to vary depending
12942 * on if something gets added or subtracted from the pattern that
12943 * causes the number of passes to change, for example */
12944 if ((value_svp = hv_fetch(RExC_unlexed_names, RExC_parse,
12947 value_sv = *value_svp;
12949 else { /* Otherwise we have to go out and get the name */
12950 const char * error_msg = NULL;
12951 value_sv = get_and_check_backslash_N_name(RExC_parse, endbrace,
12955 RExC_parse = endbrace;
12959 /* If no error message, should have gotten a valid return */
12962 /* Save the name's meaning for later use */
12963 if (! hv_store(RExC_unlexed_names, RExC_parse, name_len,
12966 Perl_croak(aTHX_ "panic: hv_store() unexpectedly failed");
12970 /* Here, we have the value the name evaluates to in 'value_sv' */
12971 value = (U8 *) SvPV(value_sv, value_len);
12973 /* See if the result is one code point vs 0 or multiple */
12974 if (value_len > 0 && value_len <= (UV) ((SvUTF8(value_sv))
12978 /* Here, exactly one code point. If that isn't what is wanted,
12980 if (! code_point_p) {
12985 /* Convert from string to numeric code point */
12986 *code_point_p = (SvUTF8(value_sv))
12987 ? valid_utf8_to_uvchr(value, NULL)
12990 /* Have parsed this entire single code point \N{...}. *cp_count
12991 * has already been set to 1, so don't do it again. */
12992 RExC_parse = endbrace;
12993 nextchar(pRExC_state);
12995 } /* End of is a single code point */
12997 /* Count the code points, if caller desires. The API says to do this
12998 * even if we will later return FALSE */
13002 *cp_count = (SvUTF8(value_sv))
13003 ? utf8_length(value, value + value_len)
13007 /* Fail if caller doesn't want to handle a multi-code-point sequence.
13008 * But don't back the pointer up if the caller wants to know how many
13009 * code points there are (they need to handle it themselves in this
13018 /* Convert this to a sub-pattern of the form "(?: ... )", and then call
13019 * reg recursively to parse it. That way, it retains its atomicness,
13020 * while not having to worry about any special handling that some code
13021 * points may have. */
13023 substitute_parse = newSVpvs("?:");
13024 sv_catsv(substitute_parse, value_sv);
13025 sv_catpv(substitute_parse, ")");
13027 /* The value should already be native, so no need to convert on EBCDIC
13029 assert(! RExC_recode_x_to_native);
13032 else { /* \N{U+...} */
13033 Size_t count = 0; /* code point count kept internally */
13035 /* We can get to here when the input is \N{U+...} or when toke.c has
13036 * converted a name to the \N{U+...} form. This include changing a
13037 * name that evaluates to multiple code points to \N{U+c1.c2.c3 ...} */
13039 RExC_parse += 2; /* Skip past the 'U+' */
13041 /* Code points are separated by dots. The '}' terminates the whole
13044 do { /* Loop until the ending brace */
13046 char * start_digit; /* The first of the current code point */
13047 if (! isXDIGIT(*RExC_parse)) {
13049 vFAIL("Invalid hexadecimal number in \\N{U+...}");
13052 start_digit = RExC_parse;
13055 /* Loop through the hex digits of the current code point */
13057 /* Adding this digit will shift the result 4 bits. If that
13058 * result would be above the legal max, it's overflow */
13059 if (cp > MAX_LEGAL_CP >> 4) {
13061 /* Find the end of the code point */
13064 } while (isXDIGIT(*RExC_parse) || *RExC_parse == '_');
13066 /* Be sure to synchronize this message with the similar one
13068 vFAIL4("Use of code point 0x%.*s is not allowed; the"
13069 " permissible max is 0x%" UVxf,
13070 (int) (RExC_parse - start_digit), start_digit,
13074 /* Accumulate this (valid) digit into the running total */
13075 cp = (cp << 4) + READ_XDIGIT(RExC_parse);
13077 /* READ_XDIGIT advanced the input pointer. Ignore a single
13078 * underscore separator */
13079 if (*RExC_parse == '_' && isXDIGIT(RExC_parse[1])) {
13082 } while (isXDIGIT(*RExC_parse));
13084 /* Here, have accumulated the next code point */
13085 if (RExC_parse >= endbrace) { /* If done ... */
13090 /* Here, is a single code point; fail if doesn't want that */
13091 if (! code_point_p) {
13096 /* A single code point is easy to handle; just return it */
13097 *code_point_p = UNI_TO_NATIVE(cp);
13098 RExC_parse = endbrace;
13099 nextchar(pRExC_state);
13103 /* Here, the only legal thing would be a multiple character
13104 * sequence (of the form "\N{U+c1.c2. ... }". So the next
13105 * character must be a dot (and the one after that can't be the
13106 * endbrace, or we'd have something like \N{U+100.} ) */
13107 if (*RExC_parse != '.' || RExC_parse + 1 >= endbrace) {
13108 RExC_parse += (RExC_orig_utf8) /* point to after 1st invalid */
13109 ? UTF8SKIP(RExC_parse)
13111 if (RExC_parse >= endbrace) { /* Guard against malformed utf8 */
13112 RExC_parse = endbrace;
13114 vFAIL("Invalid hexadecimal number in \\N{U+...}");
13117 /* Here, looks like its really a multiple character sequence. Fail
13118 * if that's not what the caller wants. But continue with counting
13119 * and error checking if they still want a count */
13120 if (! node_p && ! cp_count) {
13124 /* What is done here is to convert this to a sub-pattern of the
13125 * form \x{char1}\x{char2}... and then call reg recursively to
13126 * parse it (enclosing in "(?: ... )" ). That way, it retains its
13127 * atomicness, while not having to worry about special handling
13128 * that some code points may have. We don't create a subpattern,
13129 * but go through the motions of code point counting and error
13130 * checking, if the caller doesn't want a node returned. */
13132 if (node_p && count == 1) {
13133 substitute_parse = newSVpvs("?:");
13139 /* Convert to notation the rest of the code understands */
13140 sv_catpvs(substitute_parse, "\\x{");
13141 sv_catpvn(substitute_parse, start_digit,
13142 RExC_parse - start_digit);
13143 sv_catpvs(substitute_parse, "}");
13146 /* Move to after the dot (or ending brace the final time through.)
13151 } while (RExC_parse < endbrace);
13153 if (! node_p) { /* Doesn't want the node */
13160 sv_catpvs(substitute_parse, ")");
13162 /* The values are Unicode, and therefore have to be converted to native
13163 * on a non-Unicode (meaning non-ASCII) platform. */
13164 SET_recode_x_to_native(1);
13167 /* Here, we have the string the name evaluates to, ready to be parsed,
13168 * stored in 'substitute_parse' as a series of valid "\x{...}\x{...}"
13169 * constructs. This can be called from within a substitute parse already.
13170 * The error reporting mechanism doesn't work for 2 levels of this, but the
13171 * code above has validated this new construct, so there should be no
13172 * errors generated by the below. And this isn' an exact copy, so the
13173 * mechanism to seamlessly deal with this won't work, so turn off warnings
13175 save_start = RExC_start;
13176 orig_end = RExC_end;
13178 RExC_parse = RExC_start = SvPVX(substitute_parse);
13179 RExC_end = RExC_parse + SvCUR(substitute_parse);
13180 TURN_OFF_WARNINGS_IN_SUBSTITUTE_PARSE;
13182 *node_p = reg(pRExC_state, 1, &flags, depth+1);
13184 /* Restore the saved values */
13186 RExC_start = save_start;
13187 RExC_parse = endbrace;
13188 RExC_end = orig_end;
13189 SET_recode_x_to_native(0);
13191 SvREFCNT_dec_NN(substitute_parse);
13194 RETURN_FAIL_ON_RESTART(flags, flagp);
13195 FAIL2("panic: reg returned failure to grok_bslash_N, flags=%#" UVxf,
13198 *flagp |= flags&(HASWIDTH|SPSTART|SIMPLE|POSTPONED);
13200 nextchar(pRExC_state);
13206 PERL_STATIC_INLINE U8
13207 S_compute_EXACTish(RExC_state_t *pRExC_state)
13211 PERL_ARGS_ASSERT_COMPUTE_EXACTISH;
13219 op = get_regex_charset(RExC_flags);
13220 if (op >= REGEX_ASCII_RESTRICTED_CHARSET) {
13221 op--; /* /a is same as /u, and map /aa's offset to what /a's would have
13222 been, so there is no hole */
13225 return op + EXACTF;
13229 S_new_regcurly(const char *s, const char *e)
13231 /* This is a temporary function designed to match the most lenient form of
13232 * a {m,n} quantifier we ever envision, with either number omitted, and
13233 * spaces anywhere between/before/after them.
13235 * If this function fails, then the string it matches is very unlikely to
13236 * ever be considered a valid quantifier, so we can allow the '{' that
13237 * begins it to be considered as a literal */
13239 bool has_min = FALSE;
13240 bool has_max = FALSE;
13242 PERL_ARGS_ASSERT_NEW_REGCURLY;
13244 if (s >= e || *s++ != '{')
13247 while (s < e && isSPACE(*s)) {
13250 while (s < e && isDIGIT(*s)) {
13254 while (s < e && isSPACE(*s)) {
13260 while (s < e && isSPACE(*s)) {
13263 while (s < e && isDIGIT(*s)) {
13267 while (s < e && isSPACE(*s)) {
13272 return s < e && *s == '}' && (has_min || has_max);
13275 /* Parse backref decimal value, unless it's too big to sensibly be a backref,
13276 * in which case return I32_MAX (rather than possibly 32-bit wrapping) */
13279 S_backref_value(char *p, char *e)
13281 const char* endptr = e;
13283 if (grok_atoUV(p, &val, &endptr) && val <= I32_MAX)
13290 - regatom - the lowest level
13292 Try to identify anything special at the start of the current parse position.
13293 If there is, then handle it as required. This may involve generating a
13294 single regop, such as for an assertion; or it may involve recursing, such as
13295 to handle a () structure.
13297 If the string doesn't start with something special then we gobble up
13298 as much literal text as we can. If we encounter a quantifier, we have to
13299 back off the final literal character, as that quantifier applies to just it
13300 and not to the whole string of literals.
13302 Once we have been able to handle whatever type of thing started the
13303 sequence, we return the offset into the regex engine program being compiled
13304 at which any next regnode should be placed.
13306 Returns 0, setting *flagp to TRYAGAIN if reg() returns 0 with TRYAGAIN.
13307 Returns 0, setting *flagp to RESTART_PARSE if the parse needs to be
13308 restarted, or'd with NEED_UTF8 if the pattern needs to be upgraded to UTF-8
13309 Otherwise does not return 0.
13311 Note: we have to be careful with escapes, as they can be both literal
13312 and special, and in the case of \10 and friends, context determines which.
13314 A summary of the code structure is:
13316 switch (first_byte) {
13317 cases for each special:
13318 handle this special;
13321 switch (2nd byte) {
13322 cases for each unambiguous special:
13323 handle this special;
13325 cases for each ambigous special/literal:
13327 if (special) handle here
13329 default: // unambiguously literal:
13332 default: // is a literal char
13335 create EXACTish node for literal;
13336 while (more input and node isn't full) {
13337 switch (input_byte) {
13338 cases for each special;
13339 make sure parse pointer is set so that the next call to
13340 regatom will see this special first
13341 goto loopdone; // EXACTish node terminated by prev. char
13343 append char to EXACTISH node;
13345 get next input byte;
13349 return the generated node;
13351 Specifically there are two separate switches for handling
13352 escape sequences, with the one for handling literal escapes requiring
13353 a dummy entry for all of the special escapes that are actually handled
13358 STATIC regnode_offset
13359 S_regatom(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, U32 depth)
13362 regnode_offset ret = 0;
13369 GET_RE_DEBUG_FLAGS_DECL;
13371 *flagp = WORST; /* Tentatively. */
13373 DEBUG_PARSE("atom");
13375 PERL_ARGS_ASSERT_REGATOM;
13378 parse_start = RExC_parse;
13379 assert(RExC_parse < RExC_end);
13380 switch ((U8)*RExC_parse) {
13382 RExC_seen_zerolen++;
13383 nextchar(pRExC_state);
13384 if (RExC_flags & RXf_PMf_MULTILINE)
13385 ret = reg_node(pRExC_state, MBOL);
13387 ret = reg_node(pRExC_state, SBOL);
13388 Set_Node_Length(REGNODE_p(ret), 1); /* MJD */
13391 nextchar(pRExC_state);
13393 RExC_seen_zerolen++;
13394 if (RExC_flags & RXf_PMf_MULTILINE)
13395 ret = reg_node(pRExC_state, MEOL);
13397 ret = reg_node(pRExC_state, SEOL);
13398 Set_Node_Length(REGNODE_p(ret), 1); /* MJD */
13401 nextchar(pRExC_state);
13402 if (RExC_flags & RXf_PMf_SINGLELINE)
13403 ret = reg_node(pRExC_state, SANY);
13405 ret = reg_node(pRExC_state, REG_ANY);
13406 *flagp |= HASWIDTH|SIMPLE;
13408 Set_Node_Length(REGNODE_p(ret), 1); /* MJD */
13412 char * const oregcomp_parse = ++RExC_parse;
13413 ret = regclass(pRExC_state, flagp, depth+1,
13414 FALSE, /* means parse the whole char class */
13415 TRUE, /* allow multi-char folds */
13416 FALSE, /* don't silence non-portable warnings. */
13417 (bool) RExC_strict,
13418 TRUE, /* Allow an optimized regnode result */
13421 RETURN_FAIL_ON_RESTART_FLAGP(flagp);
13422 FAIL2("panic: regclass returned failure to regatom, flags=%#" UVxf,
13425 if (*RExC_parse != ']') {
13426 RExC_parse = oregcomp_parse;
13427 vFAIL("Unmatched [");
13429 nextchar(pRExC_state);
13430 Set_Node_Length(REGNODE_p(ret), RExC_parse - oregcomp_parse + 1); /* MJD */
13434 nextchar(pRExC_state);
13435 ret = reg(pRExC_state, 2, &flags, depth+1);
13437 if (flags & TRYAGAIN) {
13438 if (RExC_parse >= RExC_end) {
13439 /* Make parent create an empty node if needed. */
13440 *flagp |= TRYAGAIN;
13445 RETURN_FAIL_ON_RESTART(flags, flagp);
13446 FAIL2("panic: reg returned failure to regatom, flags=%#" UVxf,
13449 *flagp |= flags&(HASWIDTH|SPSTART|SIMPLE|POSTPONED);
13453 if (flags & TRYAGAIN) {
13454 *flagp |= TRYAGAIN;
13457 vFAIL("Internal urp");
13458 /* Supposed to be caught earlier. */
13464 vFAIL("Quantifier follows nothing");
13469 This switch handles escape sequences that resolve to some kind
13470 of special regop and not to literal text. Escape sequences that
13471 resolve to literal text are handled below in the switch marked
13474 Every entry in this switch *must* have a corresponding entry
13475 in the literal escape switch. However, the opposite is not
13476 required, as the default for this switch is to jump to the
13477 literal text handling code.
13480 switch ((U8)*RExC_parse) {
13481 /* Special Escapes */
13483 RExC_seen_zerolen++;
13484 ret = reg_node(pRExC_state, SBOL);
13485 /* SBOL is shared with /^/ so we set the flags so we can tell
13486 * /\A/ from /^/ in split. */
13487 FLAGS(REGNODE_p(ret)) = 1;
13489 goto finish_meta_pat;
13491 ret = reg_node(pRExC_state, GPOS);
13492 RExC_seen |= REG_GPOS_SEEN;
13494 goto finish_meta_pat;
13496 if (!RExC_in_lookbehind && !RExC_in_lookahead) {
13497 RExC_seen_zerolen++;
13498 ret = reg_node(pRExC_state, KEEPS);
13500 /* XXX:dmq : disabling in-place substitution seems to
13501 * be necessary here to avoid cases of memory corruption, as
13502 * with: C<$_="x" x 80; s/x\K/y/> -- rgs
13504 RExC_seen |= REG_LOOKBEHIND_SEEN;
13505 goto finish_meta_pat;
13508 ++RExC_parse; /* advance past the 'K' */
13509 vFAIL("\\K not permitted in lookahead/lookbehind");
13512 ret = reg_node(pRExC_state, SEOL);
13514 RExC_seen_zerolen++; /* Do not optimize RE away */
13515 goto finish_meta_pat;
13517 ret = reg_node(pRExC_state, EOS);
13519 RExC_seen_zerolen++; /* Do not optimize RE away */
13520 goto finish_meta_pat;
13522 vFAIL("\\C no longer supported");
13524 ret = reg_node(pRExC_state, CLUMP);
13525 *flagp |= HASWIDTH;
13526 goto finish_meta_pat;
13532 arg = ANYOF_WORDCHAR;
13541 regex_charset charset = get_regex_charset(RExC_flags);
13543 RExC_seen_zerolen++;
13544 RExC_seen |= REG_LOOKBEHIND_SEEN;
13545 op = BOUND + charset;
13547 if (RExC_parse >= RExC_end || *(RExC_parse + 1) != '{') {
13548 flags = TRADITIONAL_BOUND;
13549 if (op > BOUNDA) { /* /aa is same as /a */
13555 char name = *RExC_parse;
13556 char * endbrace = NULL;
13558 endbrace = (char *) memchr(RExC_parse, '}', RExC_end - RExC_parse);
13561 vFAIL2("Missing right brace on \\%c{}", name);
13563 /* XXX Need to decide whether to take spaces or not. Should be
13564 * consistent with \p{}, but that currently is SPACE, which
13565 * means vertical too, which seems wrong
13566 * while (isBLANK(*RExC_parse)) {
13569 if (endbrace == RExC_parse) {
13570 RExC_parse++; /* After the '}' */
13571 vFAIL2("Empty \\%c{}", name);
13573 length = endbrace - RExC_parse;
13574 /*while (isBLANK(*(RExC_parse + length - 1))) {
13577 switch (*RExC_parse) {
13580 && (memNEs(RExC_parse + 1, length - 1, "cb")))
13582 goto bad_bound_type;
13587 if (length != 2 || *(RExC_parse + 1) != 'b') {
13588 goto bad_bound_type;
13593 if (length != 2 || *(RExC_parse + 1) != 'b') {
13594 goto bad_bound_type;
13599 if (length != 2 || *(RExC_parse + 1) != 'b') {
13600 goto bad_bound_type;
13606 RExC_parse = endbrace;
13608 "'%" UTF8f "' is an unknown bound type",
13609 UTF8fARG(UTF, length, endbrace - length));
13610 NOT_REACHED; /*NOTREACHED*/
13612 RExC_parse = endbrace;
13613 REQUIRE_UNI_RULES(flagp, 0);
13618 else if (op >= BOUNDA) { /* /aa is same as /a */
13622 /* Don't have to worry about UTF-8, in this message because
13623 * to get here the contents of the \b must be ASCII */
13624 ckWARN4reg(RExC_parse + 1, /* Include the '}' in msg */
13625 "Using /u for '%.*s' instead of /%s",
13627 endbrace - length + 1,
13628 (charset == REGEX_ASCII_RESTRICTED_CHARSET)
13629 ? ASCII_RESTRICT_PAT_MODS
13630 : ASCII_MORE_RESTRICT_PAT_MODS);
13635 RExC_seen_d_op = TRUE;
13637 else if (op == BOUNDL) {
13638 RExC_contains_locale = 1;
13642 op += NBOUND - BOUND;
13645 ret = reg_node(pRExC_state, op);
13646 FLAGS(REGNODE_p(ret)) = flags;
13650 goto finish_meta_pat;
13658 if (! DEPENDS_SEMANTICS) {
13662 /* \d doesn't have any matches in the upper Latin1 range, hence /d
13663 * is equivalent to /u. Changing to /u saves some branches at
13666 goto join_posix_op_known;
13669 ret = reg_node(pRExC_state, LNBREAK);
13670 *flagp |= HASWIDTH|SIMPLE;
13671 goto finish_meta_pat;
13679 goto join_posix_op_known;
13685 arg = ANYOF_VERTWS;
13687 goto join_posix_op_known;
13697 op = POSIXD + get_regex_charset(RExC_flags);
13698 if (op > POSIXA) { /* /aa is same as /a */
13701 else if (op == POSIXL) {
13702 RExC_contains_locale = 1;
13704 else if (op == POSIXD) {
13705 RExC_seen_d_op = TRUE;
13708 join_posix_op_known:
13711 op += NPOSIXD - POSIXD;
13714 ret = reg_node(pRExC_state, op);
13715 FLAGS(REGNODE_p(ret)) = namedclass_to_classnum(arg);
13717 *flagp |= HASWIDTH|SIMPLE;
13721 if ( UCHARAT(RExC_parse + 1) == '{'
13722 && UNLIKELY(! new_regcurly(RExC_parse + 1, RExC_end)))
13725 vFAIL("Unescaped left brace in regex is illegal here");
13727 nextchar(pRExC_state);
13728 Set_Node_Length(REGNODE_p(ret), 2); /* MJD */
13734 ret = regclass(pRExC_state, flagp, depth+1,
13735 TRUE, /* means just parse this element */
13736 FALSE, /* don't allow multi-char folds */
13737 FALSE, /* don't silence non-portable warnings. It
13738 would be a bug if these returned
13740 (bool) RExC_strict,
13741 TRUE, /* Allow an optimized regnode result */
13743 RETURN_FAIL_ON_RESTART_FLAGP(flagp);
13744 /* regclass() can only return RESTART_PARSE and NEED_UTF8 if
13745 * multi-char folds are allowed. */
13747 FAIL2("panic: regclass returned failure to regatom, flags=%#" UVxf,
13752 Set_Node_Offset(REGNODE_p(ret), parse_start);
13753 Set_Node_Cur_Length(REGNODE_p(ret), parse_start - 2);
13754 nextchar(pRExC_state);
13757 /* Handle \N, \N{} and \N{NAMED SEQUENCE} (the latter meaning the
13758 * \N{...} evaluates to a sequence of more than one code points).
13759 * The function call below returns a regnode, which is our result.
13760 * The parameters cause it to fail if the \N{} evaluates to a
13761 * single code point; we handle those like any other literal. The
13762 * reason that the multicharacter case is handled here and not as
13763 * part of the EXACtish code is because of quantifiers. In
13764 * /\N{BLAH}+/, the '+' applies to the whole thing, and doing it
13765 * this way makes that Just Happen. dmq.
13766 * join_exact() will join this up with adjacent EXACTish nodes
13767 * later on, if appropriate. */
13769 if (grok_bslash_N(pRExC_state,
13770 &ret, /* Want a regnode returned */
13771 NULL, /* Fail if evaluates to a single code
13773 NULL, /* Don't need a count of how many code
13782 RETURN_FAIL_ON_RESTART_FLAGP(flagp);
13784 /* Here, evaluates to a single code point. Go get that */
13785 RExC_parse = parse_start;
13788 case 'k': /* Handle \k<NAME> and \k'NAME' */
13792 if ( RExC_parse >= RExC_end - 1
13793 || (( ch = RExC_parse[1]) != '<'
13798 /* diag_listed_as: Sequence \%s... not terminated in regex; marked by <-- HERE in m/%s/ */
13799 vFAIL2("Sequence %.2s... not terminated", parse_start);
13802 ret = handle_named_backref(pRExC_state,
13814 case '1': case '2': case '3': case '4':
13815 case '5': case '6': case '7': case '8': case '9':
13820 if (*RExC_parse == 'g') {
13824 if (*RExC_parse == '{') {
13828 if (*RExC_parse == '-') {
13832 if (hasbrace && !isDIGIT(*RExC_parse)) {
13833 if (isrel) RExC_parse--;
13835 goto parse_named_seq;
13838 if (RExC_parse >= RExC_end) {
13839 goto unterminated_g;
13841 num = S_backref_value(RExC_parse, RExC_end);
13843 vFAIL("Reference to invalid group 0");
13844 else if (num == I32_MAX) {
13845 if (isDIGIT(*RExC_parse))
13846 vFAIL("Reference to nonexistent group");
13849 vFAIL("Unterminated \\g... pattern");
13853 num = RExC_npar - num;
13855 vFAIL("Reference to nonexistent or unclosed group");
13859 num = S_backref_value(RExC_parse, RExC_end);
13860 /* bare \NNN might be backref or octal - if it is larger
13861 * than or equal RExC_npar then it is assumed to be an
13862 * octal escape. Note RExC_npar is +1 from the actual
13863 * number of parens. */
13864 /* Note we do NOT check if num == I32_MAX here, as that is
13865 * handled by the RExC_npar check */
13868 /* any numeric escape < 10 is always a backref */
13870 /* any numeric escape < RExC_npar is a backref */
13871 && num >= RExC_npar
13872 /* cannot be an octal escape if it starts with 8 */
13873 && *RExC_parse != '8'
13874 /* cannot be an octal escape if it starts with 9 */
13875 && *RExC_parse != '9'
13877 /* Probably not meant to be a backref, instead likely
13878 * to be an octal character escape, e.g. \35 or \777.
13879 * The above logic should make it obvious why using
13880 * octal escapes in patterns is problematic. - Yves */
13881 RExC_parse = parse_start;
13886 /* At this point RExC_parse points at a numeric escape like
13887 * \12 or \88 or something similar, which we should NOT treat
13888 * as an octal escape. It may or may not be a valid backref
13889 * escape. For instance \88888888 is unlikely to be a valid
13891 while (isDIGIT(*RExC_parse))
13894 if (*RExC_parse != '}')
13895 vFAIL("Unterminated \\g{...} pattern");
13898 if (num >= (I32)RExC_npar) {
13900 /* It might be a forward reference; we can't fail until we
13901 * know, by completing the parse to get all the groups, and
13902 * then reparsing */
13903 if (ALL_PARENS_COUNTED) {
13904 if (num >= RExC_total_parens) {
13905 vFAIL("Reference to nonexistent group");
13909 REQUIRE_PARENS_PASS;
13913 ret = reganode(pRExC_state,
13916 : (ASCII_FOLD_RESTRICTED)
13918 : (AT_LEAST_UNI_SEMANTICS)
13924 if (OP(REGNODE_p(ret)) == REFF) {
13925 RExC_seen_d_op = TRUE;
13927 *flagp |= HASWIDTH;
13929 /* override incorrect value set in reganode MJD */
13930 Set_Node_Offset(REGNODE_p(ret), parse_start);
13931 Set_Node_Cur_Length(REGNODE_p(ret), parse_start-1);
13932 skip_to_be_ignored_text(pRExC_state, &RExC_parse,
13933 FALSE /* Don't force to /x */ );
13937 if (RExC_parse >= RExC_end)
13938 FAIL("Trailing \\");
13941 /* Do not generate "unrecognized" warnings here, we fall
13942 back into the quick-grab loop below */
13943 RExC_parse = parse_start;
13945 } /* end of switch on a \foo sequence */
13950 /* '#' comments should have been spaced over before this function was
13952 assert((RExC_flags & RXf_PMf_EXTENDED) == 0);
13954 if (RExC_flags & RXf_PMf_EXTENDED) {
13955 RExC_parse = reg_skipcomment( pRExC_state, RExC_parse );
13956 if (RExC_parse < RExC_end)
13966 /* Here, we have determined that the next thing is probably a
13967 * literal character. RExC_parse points to the first byte of its
13968 * definition. (It still may be an escape sequence that evaluates
13969 * to a single character) */
13976 /* This allows us to fill a node with just enough spare so that if the final
13977 * character folds, its expansion is guaranteed to fit */
13978 #define MAX_NODE_STRING_SIZE (255-UTF8_MAXBYTES_CASE)
13981 U8 upper_parse = MAX_NODE_STRING_SIZE;
13983 /* We start out as an EXACT node, even if under /i, until we find a
13984 * character which is in a fold. The algorithm now segregates into
13985 * separate nodes, characters that fold from those that don't under
13986 * /i. (This hopefully will create nodes that are fixed strings
13987 * even under /i, giving the optimizer something to grab on to.)
13988 * So, if a node has something in it and the next character is in
13989 * the opposite category, that node is closed up, and the function
13990 * returns. Then regatom is called again, and a new node is
13991 * created for the new category. */
13992 U8 node_type = EXACT;
13994 /* Assume the node will be fully used; the excess is given back at
13995 * the end. We can't make any other length assumptions, as a byte
13996 * input sequence could shrink down. */
13997 Ptrdiff_t initial_size = STR_SZ(256);
13999 bool next_is_quantifier;
14000 char * oldp = NULL;
14002 /* We can convert EXACTF nodes to EXACTFU if they contain only
14003 * characters that match identically regardless of the target
14004 * string's UTF8ness. The reason to do this is that EXACTF is not
14005 * trie-able, EXACTFU is, and EXACTFU requires fewer operations at
14008 * Similarly, we can convert EXACTFL nodes to EXACTFLU8 if they
14009 * contain only above-Latin1 characters (hence must be in UTF8),
14010 * which don't participate in folds with Latin1-range characters,
14011 * as the latter's folds aren't known until runtime. */
14012 bool maybe_exactfu = FOLD && (DEPENDS_SEMANTICS || LOC);
14014 /* Single-character EXACTish nodes are almost always SIMPLE. This
14015 * allows us to override this as encountered */
14016 U8 maybe_SIMPLE = SIMPLE;
14018 /* Does this node contain something that can't match unless the
14019 * target string is (also) in UTF-8 */
14020 bool requires_utf8_target = FALSE;
14022 /* The sequence 'ss' is problematic in non-UTF-8 patterns. */
14023 bool has_ss = FALSE;
14025 /* So is the MICRO SIGN */
14026 bool has_micro_sign = FALSE;
14028 /* Allocate an EXACT node. The node_type may change below to
14029 * another EXACTish node, but since the size of the node doesn't
14030 * change, it works */
14031 ret = regnode_guts(pRExC_state, node_type, initial_size, "exact");
14032 FILL_NODE(ret, node_type);
14035 s = STRING(REGNODE_p(ret));
14041 /* This breaks under rare circumstances. If folding, we do not
14042 * want to split a node at a character that is a non-final in a
14043 * multi-char fold, as an input string could just happen to want to
14044 * match across the node boundary. The code at the end of the loop
14045 * looks for this, and backs off until it finds not such a
14046 * character, but it is possible (though extremely, extremely
14047 * unlikely) for all characters in the node to be non-final fold
14048 * ones, in which case we just leave the node fully filled, and
14049 * hope that it doesn't match the string in just the wrong place */
14051 assert( ! UTF /* Is at the beginning of a character */
14052 || UTF8_IS_INVARIANT(UCHARAT(RExC_parse))
14053 || UTF8_IS_START(UCHARAT(RExC_parse)));
14055 /* Here, we have a literal character. Find the maximal string of
14056 * them in the input that we can fit into a single EXACTish node.
14057 * We quit at the first non-literal or when the node gets full, or
14058 * under /i the categorization of folding/non-folding character
14060 for (p = RExC_parse; len < upper_parse && p < RExC_end; ) {
14062 /* In most cases each iteration adds one byte to the output.
14063 * The exceptions override this */
14064 Size_t added_len = 1;
14068 /* White space has already been ignored */
14069 assert( (RExC_flags & RXf_PMf_EXTENDED) == 0
14070 || ! is_PATWS_safe((p), RExC_end, UTF));
14082 /* Literal Escapes Switch
14084 This switch is meant to handle escape sequences that
14085 resolve to a literal character.
14087 Every escape sequence that represents something
14088 else, like an assertion or a char class, is handled
14089 in the switch marked 'Special Escapes' above in this
14090 routine, but also has an entry here as anything that
14091 isn't explicitly mentioned here will be treated as
14092 an unescaped equivalent literal.
14095 switch ((U8)*++p) {
14097 /* These are all the special escapes. */
14098 case 'A': /* Start assertion */
14099 case 'b': case 'B': /* Word-boundary assertion*/
14100 case 'C': /* Single char !DANGEROUS! */
14101 case 'd': case 'D': /* digit class */
14102 case 'g': case 'G': /* generic-backref, pos assertion */
14103 case 'h': case 'H': /* HORIZWS */
14104 case 'k': case 'K': /* named backref, keep marker */
14105 case 'p': case 'P': /* Unicode property */
14106 case 'R': /* LNBREAK */
14107 case 's': case 'S': /* space class */
14108 case 'v': case 'V': /* VERTWS */
14109 case 'w': case 'W': /* word class */
14110 case 'X': /* eXtended Unicode "combining
14111 character sequence" */
14112 case 'z': case 'Z': /* End of line/string assertion */
14116 /* Anything after here is an escape that resolves to a
14117 literal. (Except digits, which may or may not)
14123 case 'N': /* Handle a single-code point named character. */
14124 RExC_parse = p + 1;
14125 if (! grok_bslash_N(pRExC_state,
14126 NULL, /* Fail if evaluates to
14127 anything other than a
14128 single code point */
14129 &ender, /* The returned single code
14131 NULL, /* Don't need a count of
14132 how many code points */
14137 if (*flagp & NEED_UTF8)
14138 FAIL("panic: grok_bslash_N set NEED_UTF8");
14139 RETURN_FAIL_ON_RESTART_FLAGP(flagp);
14141 /* Here, it wasn't a single code point. Go close
14142 * up this EXACTish node. The switch() prior to
14143 * this switch handles the other cases */
14144 RExC_parse = p = oldp;
14148 RExC_parse = parse_start;
14150 /* The \N{} means the pattern, if previously /d,
14151 * becomes /u. That means it can't be an EXACTF node,
14152 * but an EXACTFU */
14153 if (node_type == EXACTF) {
14154 node_type = EXACTFU;
14156 /* If the node already contains something that
14157 * differs between EXACTF and EXACTFU, reparse it
14159 if (! maybe_exactfu) {
14180 ender = ESC_NATIVE;
14190 const char* error_msg;
14192 bool valid = grok_bslash_o(&p,
14196 TO_OUTPUT_WARNINGS(p),
14197 (bool) RExC_strict,
14198 TRUE, /* Output warnings
14203 RExC_parse = p; /* going to die anyway; point
14204 to exact spot of failure */
14207 UPDATE_WARNINGS_LOC(p - 1);
14213 UV result = UV_MAX; /* initialize to erroneous
14215 const char* error_msg;
14217 bool valid = grok_bslash_x(&p,
14221 TO_OUTPUT_WARNINGS(p),
14222 (bool) RExC_strict,
14223 TRUE, /* Silence warnings
14228 RExC_parse = p; /* going to die anyway; point
14229 to exact spot of failure */
14232 UPDATE_WARNINGS_LOC(p - 1);
14236 if (ender < 0x100) {
14237 if (RExC_recode_x_to_native) {
14238 ender = LATIN1_TO_NATIVE(ender);
14246 ender = grok_bslash_c(*p, TO_OUTPUT_WARNINGS(p));
14247 UPDATE_WARNINGS_LOC(p);
14250 case '8': case '9': /* must be a backreference */
14252 /* we have an escape like \8 which cannot be an octal escape
14253 * so we exit the loop, and let the outer loop handle this
14254 * escape which may or may not be a legitimate backref. */
14256 case '1': case '2': case '3':case '4':
14257 case '5': case '6': case '7':
14258 /* When we parse backslash escapes there is ambiguity
14259 * between backreferences and octal escapes. Any escape
14260 * from \1 - \9 is a backreference, any multi-digit
14261 * escape which does not start with 0 and which when
14262 * evaluated as decimal could refer to an already
14263 * parsed capture buffer is a back reference. Anything
14266 * Note this implies that \118 could be interpreted as
14267 * 118 OR as "\11" . "8" depending on whether there
14268 * were 118 capture buffers defined already in the
14271 /* NOTE, RExC_npar is 1 more than the actual number of
14272 * parens we have seen so far, hence the "<" as opposed
14274 if ( !isDIGIT(p[1]) || S_backref_value(p, RExC_end) < RExC_npar)
14275 { /* Not to be treated as an octal constant, go
14283 I32 flags = PERL_SCAN_SILENT_ILLDIGIT;
14285 ender = grok_oct(p, &numlen, &flags, NULL);
14287 if ( isDIGIT(*p) /* like \08, \178 */
14288 && ckWARN(WARN_REGEXP)
14291 reg_warn_non_literal_string(
14293 form_short_octal_warning(p, numlen));
14299 FAIL("Trailing \\");
14302 if (isALPHANUMERIC(*p)) {
14303 /* An alpha followed by '{' is going to fail next
14304 * iteration, so don't output this warning in that
14306 if (! isALPHA(*p) || *(p + 1) != '{') {
14307 ckWARN2reg(p + 1, "Unrecognized escape \\%.1s"
14308 " passed through", p);
14311 goto normal_default;
14312 } /* End of switch on '\' */
14315 /* Trying to gain new uses for '{' without breaking too
14316 * much existing code is hard. The solution currently
14318 * 1) If there is no ambiguity that a '{' should always
14319 * be taken literally, at the start of a construct, we
14321 * 2) If the literal '{' conflicts with our desired use
14322 * of it as a metacharacter, we die. The deprecation
14323 * cycles for this have come and gone.
14324 * 3) If there is ambiguity, we raise a simple warning.
14325 * This could happen, for example, if the user
14326 * intended it to introduce a quantifier, but slightly
14327 * misspelled the quantifier. Without this warning,
14328 * the quantifier would silently be taken as a literal
14329 * string of characters instead of a meta construct */
14330 if (len || (p > RExC_start && isALPHA_A(*(p - 1)))) {
14332 || ( p > parse_start + 1
14333 && isALPHA_A(*(p - 1))
14334 && *(p - 2) == '\\')
14335 || new_regcurly(p, RExC_end))
14337 RExC_parse = p + 1;
14338 vFAIL("Unescaped left brace in regex is "
14341 ckWARNreg(p + 1, "Unescaped left brace in regex is"
14342 " passed through");
14344 goto normal_default;
14347 if (p > RExC_parse && RExC_strict) {
14348 ckWARN2reg(p + 1, "Unescaped literal '%c'", *p);
14351 default: /* A literal character */
14353 if (! UTF8_IS_INVARIANT(*p) && UTF) {
14355 ender = utf8n_to_uvchr((U8*)p, RExC_end - p,
14356 &numlen, UTF8_ALLOW_DEFAULT);
14362 } /* End of switch on the literal */
14364 /* Here, have looked at the literal character, and <ender>
14365 * contains its ordinal; <p> points to the character after it.
14369 REQUIRE_UTF8(flagp);
14372 /* We need to check if the next non-ignored thing is a
14373 * quantifier. Move <p> to after anything that should be
14374 * ignored, which, as a side effect, positions <p> for the next
14375 * loop iteration */
14376 skip_to_be_ignored_text(pRExC_state, &p,
14377 FALSE /* Don't force to /x */ );
14379 /* If the next thing is a quantifier, it applies to this
14380 * character only, which means that this character has to be in
14381 * its own node and can't just be appended to the string in an
14382 * existing node, so if there are already other characters in
14383 * the node, close the node with just them, and set up to do
14384 * this character again next time through, when it will be the
14385 * only thing in its new node */
14387 next_is_quantifier = LIKELY(p < RExC_end)
14388 && UNLIKELY(ISMULT2(p));
14390 if (next_is_quantifier && LIKELY(len)) {
14395 /* Ready to add 'ender' to the node */
14397 if (! FOLD) { /* The simple case, just append the literal */
14400 if (UVCHR_IS_INVARIANT(ender) || ! UTF) {
14401 *(s++) = (char) ender;
14404 U8 * new_s = uvchr_to_utf8((U8*)s, ender);
14405 added_len = (char *) new_s - s;
14406 s = (char *) new_s;
14409 requires_utf8_target = TRUE;
14413 else if (LOC && is_PROBLEMATIC_LOCALE_FOLD_cp(ender)) {
14415 /* Here are folding under /l, and the code point is
14416 * problematic. If this is the first character in the
14417 * node, change the node type to folding. Otherwise, if
14418 * this is the first problematic character, close up the
14419 * existing node, so can start a new node with this one */
14421 node_type = EXACTFL;
14422 RExC_contains_locale = 1;
14424 else if (node_type == EXACT) {
14429 /* This problematic code point means we can't simplify
14431 maybe_exactfu = FALSE;
14433 /* Here, we are adding a problematic fold character.
14434 * "Problematic" in this context means that its fold isn't
14435 * known until runtime. (The non-problematic code points
14436 * are the above-Latin1 ones that fold to also all
14437 * above-Latin1. Their folds don't vary no matter what the
14438 * locale is.) But here we have characters whose fold
14439 * depends on the locale. We just add in the unfolded
14440 * character, and wait until runtime to fold it */
14441 goto not_fold_common;
14443 else /* regular fold; see if actually is in a fold */
14444 if ( (ender < 256 && ! IS_IN_SOME_FOLD_L1(ender))
14446 && ! _invlist_contains_cp(PL_in_some_fold, ender)))
14448 /* Here, folding, but the character isn't in a fold.
14450 * Start a new node if previous characters in the node were
14452 if (len && node_type != EXACT) {
14457 /* Here, continuing a node with non-folded characters. Add
14459 goto not_fold_common;
14461 else { /* Here, does participate in some fold */
14463 /* If this is the first character in the node, change its
14464 * type to folding. Otherwise, if this is the first
14465 * folding character in the node, close up the existing
14466 * node, so can start a new node with this one. */
14468 node_type = compute_EXACTish(pRExC_state);
14470 else if (node_type == EXACT) {
14475 if (UTF) { /* Use the folded value */
14476 if (UVCHR_IS_INVARIANT(ender)) {
14477 *(s)++ = (U8) toFOLD(ender);
14480 ender = _to_uni_fold_flags(
14484 FOLD_FLAGS_FULL | ((ASCII_FOLD_RESTRICTED)
14485 ? FOLD_FLAGS_NOMIX_ASCII
14490 && LIKELY(ender != GREEK_SMALL_LETTER_MU))
14492 /* U+B5 folds to the MU, so its possible for a
14493 * non-UTF-8 target to match it */
14494 requires_utf8_target = TRUE;
14500 /* Here is non-UTF8. First, see if the character's
14501 * fold differs between /d and /u. */
14502 if (PL_fold[ender] != PL_fold_latin1[ender]) {
14503 maybe_exactfu = FALSE;
14506 #if UNICODE_MAJOR_VERSION > 3 /* no multifolds in early Unicode */ \
14507 || (UNICODE_MAJOR_VERSION == 3 && ( UNICODE_DOT_VERSION > 0) \
14508 || UNICODE_DOT_DOT_VERSION > 0)
14510 /* On non-ancient Unicode versions, this includes the
14511 * multi-char fold SHARP S to 'ss' */
14513 if ( UNLIKELY(ender == LATIN_SMALL_LETTER_SHARP_S)
14514 || ( isALPHA_FOLD_EQ(ender, 's')
14516 && isALPHA_FOLD_EQ(*(s-1), 's')))
14518 /* Here, we have one of the following:
14519 * a) a SHARP S. This folds to 'ss' only under
14520 * /u rules. If we are in that situation,
14521 * fold the SHARP S to 'ss'. See the comments
14522 * for join_exact() as to why we fold this
14523 * non-UTF at compile time, and no others.
14524 * b) 'ss'. When under /u, there's nothing
14525 * special needed to be done here. The
14526 * previous iteration handled the first 's',
14527 * and this iteration will handle the second.
14528 * If, on the otherhand it's not /u, we have
14529 * to exclude the possibility of moving to /u,
14530 * so that we won't generate an unwanted
14531 * match, unless, at runtime, the target
14532 * string is in UTF-8.
14536 maybe_exactfu = FALSE; /* Can't generate an
14537 EXACTFU node (unless we
14538 already are in one) */
14539 if (UNLIKELY(ender == LATIN_SMALL_LETTER_SHARP_S)) {
14541 if (node_type == EXACTFU) {
14544 /* Let the code below add in the extra 's' */
14552 else if (UNLIKELY(ender == MICRO_SIGN)) {
14553 has_micro_sign = TRUE;
14556 *(s++) = (DEPENDS_SEMANTICS)
14557 ? (char) toFOLD(ender)
14559 /* Under /u, the fold of any character in
14560 * the 0-255 range happens to be its
14561 * lowercase equivalent, except for LATIN
14562 * SMALL LETTER SHARP S, which was handled
14563 * above, and the MICRO SIGN, whose fold
14564 * requires UTF-8 to represent. */
14565 : (char) toLOWER_L1(ender);
14567 } /* End of adding current character to the node */
14571 if (next_is_quantifier) {
14573 /* Here, the next input is a quantifier, and to get here,
14574 * the current character is the only one in the node. */
14578 } /* End of loop through literal characters */
14580 /* Here we have either exhausted the input or ran out of room in
14581 * the node. (If we encountered a character that can't be in the
14582 * node, transfer is made directly to <loopdone>, and so we
14583 * wouldn't have fallen off the end of the loop.) In the latter
14584 * case, we artificially have to split the node into two, because
14585 * we just don't have enough space to hold everything. This
14586 * creates a problem if the final character participates in a
14587 * multi-character fold in the non-final position, as a match that
14588 * should have occurred won't, due to the way nodes are matched,
14589 * and our artificial boundary. So back off until we find a non-
14590 * problematic character -- one that isn't at the beginning or
14591 * middle of such a fold. (Either it doesn't participate in any
14592 * folds, or appears only in the final position of all the folds it
14593 * does participate in.) A better solution with far fewer false
14594 * positives, and that would fill the nodes more completely, would
14595 * be to actually have available all the multi-character folds to
14596 * test against, and to back-off only far enough to be sure that
14597 * this node isn't ending with a partial one. <upper_parse> is set
14598 * further below (if we need to reparse the node) to include just
14599 * up through that final non-problematic character that this code
14600 * identifies, so when it is set to less than the full node, we can
14601 * skip the rest of this */
14602 if (FOLD && p < RExC_end && upper_parse == MAX_NODE_STRING_SIZE) {
14603 PERL_UINT_FAST8_T backup_count = 0;
14605 const STRLEN full_len = len;
14607 assert(len >= MAX_NODE_STRING_SIZE);
14609 /* Here, <s> points to just beyond where we have output the
14610 * final character of the node. Look backwards through the
14611 * string until find a non- problematic character */
14615 /* This has no multi-char folds to non-UTF characters */
14616 if (ASCII_FOLD_RESTRICTED) {
14620 while (--s >= s0 && IS_NON_FINAL_FOLD(*s)) {
14627 /* Point to the first byte of the final character */
14628 s = (char *) utf8_hop_back((U8 *) s, -1, (U8 *) s0);
14630 while (s >= s0) { /* Search backwards until find
14631 a non-problematic char */
14632 if (UTF8_IS_INVARIANT(*s)) {
14634 /* There are no ascii characters that participate
14635 * in multi-char folds under /aa. In EBCDIC, the
14636 * non-ascii invariants are all control characters,
14637 * so don't ever participate in any folds. */
14638 if (ASCII_FOLD_RESTRICTED
14639 || ! IS_NON_FINAL_FOLD(*s))
14644 else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
14645 if (! IS_NON_FINAL_FOLD(EIGHT_BIT_UTF8_TO_NATIVE(
14651 else if (! _invlist_contains_cp(
14653 valid_utf8_to_uvchr((U8 *) s, NULL)))
14658 /* Here, the current character is problematic in that
14659 * it does occur in the non-final position of some
14660 * fold, so try the character before it, but have to
14661 * special case the very first byte in the string, so
14662 * we don't read outside the string */
14663 s = (s == s0) ? s -1 : (char *) utf8_hop((U8 *) s, -1);
14665 } /* End of loop backwards through the string */
14667 /* If there were only problematic characters in the string,
14668 * <s> will point to before s0, in which case the length
14669 * should be 0, otherwise include the length of the
14670 * non-problematic character just found */
14671 len = (s < s0) ? 0 : s - s0 + UTF8SKIP(s);
14674 /* Here, have found the final character, if any, that is
14675 * non-problematic as far as ending the node without splitting
14676 * it across a potential multi-char fold. <len> contains the
14677 * number of bytes in the node up-to and including that
14678 * character, or is 0 if there is no such character, meaning
14679 * the whole node contains only problematic characters. In
14680 * this case, give up and just take the node as-is. We can't
14687 /* Here, the node does contain some characters that aren't
14688 * problematic. If we didn't have to backup any, then the
14689 * final character in the node is non-problematic, and we
14690 * can take the node as-is */
14691 if (backup_count == 0) {
14694 else if (backup_count == 1) {
14696 /* If the final character is problematic, but the
14697 * penultimate is not, back-off that last character to
14698 * later start a new node with it */
14703 /* Here, the final non-problematic character is earlier
14704 * in the input than the penultimate character. What we do
14705 * is reparse from the beginning, going up only as far as
14706 * this final ok one, thus guaranteeing that the node ends
14707 * in an acceptable character. The reason we reparse is
14708 * that we know how far in the character is, but we don't
14709 * know how to correlate its position with the input parse.
14710 * An alternate implementation would be to build that
14711 * correlation as we go along during the original parse,
14712 * but that would entail extra work for every node, whereas
14713 * this code gets executed only when the string is too
14714 * large for the node, and the final two characters are
14715 * problematic, an infrequent occurrence. Yet another
14716 * possible strategy would be to save the tail of the
14717 * string, and the next time regatom is called, initialize
14718 * with that. The problem with this is that unless you
14719 * back off one more character, you won't be guaranteed
14720 * regatom will get called again, unless regbranch,
14721 * regpiece ... are also changed. If you do back off that
14722 * extra character, so that there is input guaranteed to
14723 * force calling regatom, you can't handle the case where
14724 * just the first character in the node is acceptable. I
14725 * (khw) decided to try this method which doesn't have that
14726 * pitfall; if performance issues are found, we can do a
14727 * combination of the current approach plus that one */
14733 } /* End of verifying node ends with an appropriate char */
14735 loopdone: /* Jumped to when encounters something that shouldn't be
14738 /* Free up any over-allocated space; cast is to silence bogus
14739 * warning in MS VC */
14740 change_engine_size(pRExC_state,
14741 - (Ptrdiff_t) (initial_size - STR_SZ(len)));
14743 /* I (khw) don't know if you can get here with zero length, but the
14744 * old code handled this situation by creating a zero-length EXACT
14745 * node. Might as well be NOTHING instead */
14747 OP(REGNODE_p(ret)) = NOTHING;
14751 /* If the node type is EXACT here, check to see if it
14752 * should be EXACTL, or EXACT_ONLY8. */
14753 if (node_type == EXACT) {
14755 node_type = EXACTL;
14757 else if (requires_utf8_target) {
14758 node_type = EXACT_ONLY8;
14761 if ( UNLIKELY(has_micro_sign || has_ss)
14762 && (node_type == EXACTFU || ( node_type == EXACTF
14763 && maybe_exactfu)))
14764 { /* These two conditions are problematic in non-UTF-8
14767 node_type = EXACTFUP;
14769 else if (node_type == EXACTFL) {
14771 /* 'maybe_exactfu' is deliberately set above to
14772 * indicate this node type, where all code points in it
14774 if (maybe_exactfu) {
14775 node_type = EXACTFLU8;
14778 else if (node_type == EXACTF) { /* Means is /di */
14780 /* If 'maybe_exactfu' is clear, then we need to stay
14781 * /di. If it is set, it means there are no code
14782 * points that match differently depending on UTF8ness
14783 * of the target string, so it can become an EXACTFU
14785 if (! maybe_exactfu) {
14786 RExC_seen_d_op = TRUE;
14788 else if ( isALPHA_FOLD_EQ(* STRING(REGNODE_p(ret)), 's')
14789 || isALPHA_FOLD_EQ(ender, 's'))
14791 /* But, if the node begins or ends in an 's' we
14792 * have to defer changing it into an EXACTFU, as
14793 * the node could later get joined with another one
14794 * that ends or begins with 's' creating an 'ss'
14795 * sequence which would then wrongly match the
14796 * sharp s without the target being UTF-8. We
14797 * create a special node that we resolve later when
14798 * we join nodes together */
14800 node_type = EXACTFU_S_EDGE;
14803 node_type = EXACTFU;
14807 if (requires_utf8_target && node_type == EXACTFU) {
14808 node_type = EXACTFU_ONLY8;
14812 OP(REGNODE_p(ret)) = node_type;
14813 STR_LEN(REGNODE_p(ret)) = len;
14814 RExC_emit += STR_SZ(len);
14816 /* If the node isn't a single character, it can't be SIMPLE */
14817 if (len > (Size_t) ((UTF) ? UVCHR_SKIP(ender) : 1)) {
14821 *flagp |= HASWIDTH | maybe_SIMPLE;
14824 Set_Node_Length(REGNODE_p(ret), p - parse_start - 1);
14828 /* len is STRLEN which is unsigned, need to copy to signed */
14831 vFAIL("Internal disaster");
14834 } /* End of label 'defchar:' */
14836 } /* End of giant switch on input character */
14838 /* Position parse to next real character */
14839 skip_to_be_ignored_text(pRExC_state, &RExC_parse,
14840 FALSE /* Don't force to /x */ );
14841 if ( *RExC_parse == '{'
14842 && OP(REGNODE_p(ret)) != SBOL && ! regcurly(RExC_parse))
14844 if (RExC_strict || new_regcurly(RExC_parse, RExC_end)) {
14846 vFAIL("Unescaped left brace in regex is illegal here");
14848 ckWARNreg(RExC_parse + 1, "Unescaped left brace in regex is"
14849 " passed through");
14857 S_populate_ANYOF_from_invlist(pTHX_ regnode *node, SV** invlist_ptr)
14859 /* Uses the inversion list '*invlist_ptr' to populate the ANYOF 'node'. It
14860 * sets up the bitmap and any flags, removing those code points from the
14861 * inversion list, setting it to NULL should it become completely empty */
14865 PERL_ARGS_ASSERT_POPULATE_ANYOF_FROM_INVLIST;
14866 assert(PL_regkind[OP(node)] == ANYOF);
14868 /* There is no bitmap for this node type */
14869 if (inRANGE(OP(node), ANYOFH, ANYOFHr)) {
14873 ANYOF_BITMAP_ZERO(node);
14874 if (*invlist_ptr) {
14876 /* This gets set if we actually need to modify things */
14877 bool change_invlist = FALSE;
14881 /* Start looking through *invlist_ptr */
14882 invlist_iterinit(*invlist_ptr);
14883 while (invlist_iternext(*invlist_ptr, &start, &end)) {
14887 if (end == UV_MAX && start <= NUM_ANYOF_CODE_POINTS) {
14888 ANYOF_FLAGS(node) |= ANYOF_MATCHES_ALL_ABOVE_BITMAP;
14891 /* Quit if are above what we should change */
14892 if (start >= NUM_ANYOF_CODE_POINTS) {
14896 change_invlist = TRUE;
14898 /* Set all the bits in the range, up to the max that we are doing */
14899 high = (end < NUM_ANYOF_CODE_POINTS - 1)
14901 : NUM_ANYOF_CODE_POINTS - 1;
14902 for (i = start; i <= (int) high; i++) {
14903 if (! ANYOF_BITMAP_TEST(node, i)) {
14904 ANYOF_BITMAP_SET(node, i);
14908 invlist_iterfinish(*invlist_ptr);
14910 /* Done with loop; remove any code points that are in the bitmap from
14911 * *invlist_ptr; similarly for code points above the bitmap if we have
14912 * a flag to match all of them anyways */
14913 if (change_invlist) {
14914 _invlist_subtract(*invlist_ptr, PL_InBitmap, invlist_ptr);
14916 if (ANYOF_FLAGS(node) & ANYOF_MATCHES_ALL_ABOVE_BITMAP) {
14917 _invlist_intersection(*invlist_ptr, PL_InBitmap, invlist_ptr);
14920 /* If have completely emptied it, remove it completely */
14921 if (_invlist_len(*invlist_ptr) == 0) {
14922 SvREFCNT_dec_NN(*invlist_ptr);
14923 *invlist_ptr = NULL;
14928 /* Parse POSIX character classes: [[:foo:]], [[=foo=]], [[.foo.]].
14929 Character classes ([:foo:]) can also be negated ([:^foo:]).
14930 Returns a named class id (ANYOF_XXX) if successful, -1 otherwise.
14931 Equivalence classes ([=foo=]) and composites ([.foo.]) are parsed,
14932 but trigger failures because they are currently unimplemented. */
14934 #define POSIXCC_DONE(c) ((c) == ':')
14935 #define POSIXCC_NOTYET(c) ((c) == '=' || (c) == '.')
14936 #define POSIXCC(c) (POSIXCC_DONE(c) || POSIXCC_NOTYET(c))
14937 #define MAYBE_POSIXCC(c) (POSIXCC(c) || (c) == '^' || (c) == ';')
14939 #define WARNING_PREFIX "Assuming NOT a POSIX class since "
14940 #define NO_BLANKS_POSIX_WARNING "no blanks are allowed in one"
14941 #define SEMI_COLON_POSIX_WARNING "a semi-colon was found instead of a colon"
14943 #define NOT_MEANT_TO_BE_A_POSIX_CLASS (OOB_NAMEDCLASS - 1)
14945 /* 'posix_warnings' and 'warn_text' are names of variables in the following
14947 #define ADD_POSIX_WARNING(p, text) STMT_START { \
14948 if (posix_warnings) { \
14949 if (! RExC_warn_text ) RExC_warn_text = \
14950 (AV *) sv_2mortal((SV *) newAV()); \
14951 av_push(RExC_warn_text, Perl_newSVpvf(aTHX_ \
14955 REPORT_LOCATION_ARGS(p))); \
14958 #define CLEAR_POSIX_WARNINGS() \
14960 if (posix_warnings && RExC_warn_text) \
14961 av_clear(RExC_warn_text); \
14964 #define CLEAR_POSIX_WARNINGS_AND_RETURN(ret) \
14966 CLEAR_POSIX_WARNINGS(); \
14971 S_handle_possible_posix(pTHX_ RExC_state_t *pRExC_state,
14973 const char * const s, /* Where the putative posix class begins.
14974 Normally, this is one past the '['. This
14975 parameter exists so it can be somewhere
14976 besides RExC_parse. */
14977 char ** updated_parse_ptr, /* Where to set the updated parse pointer, or
14979 AV ** posix_warnings, /* Where to place any generated warnings, or
14981 const bool check_only /* Don't die if error */
14984 /* This parses what the caller thinks may be one of the three POSIX
14986 * 1) a character class, like [:blank:]
14987 * 2) a collating symbol, like [. .]
14988 * 3) an equivalence class, like [= =]
14989 * In the latter two cases, it croaks if it finds a syntactically legal
14990 * one, as these are not handled by Perl.
14992 * The main purpose is to look for a POSIX character class. It returns:
14993 * a) the class number
14994 * if it is a completely syntactically and semantically legal class.
14995 * 'updated_parse_ptr', if not NULL, is set to point to just after the
14996 * closing ']' of the class
14997 * b) OOB_NAMEDCLASS
14998 * if it appears that one of the three POSIX constructs was meant, but
14999 * its specification was somehow defective. 'updated_parse_ptr', if
15000 * not NULL, is set to point to the character just after the end
15001 * character of the class. See below for handling of warnings.
15002 * c) NOT_MEANT_TO_BE_A_POSIX_CLASS
15003 * if it doesn't appear that a POSIX construct was intended.
15004 * 'updated_parse_ptr' is not changed. No warnings nor errors are
15007 * In b) there may be errors or warnings generated. If 'check_only' is
15008 * TRUE, then any errors are discarded. Warnings are returned to the
15009 * caller via an AV* created into '*posix_warnings' if it is not NULL. If
15010 * instead it is NULL, warnings are suppressed.
15012 * The reason for this function, and its complexity is that a bracketed
15013 * character class can contain just about anything. But it's easy to
15014 * mistype the very specific posix class syntax but yielding a valid
15015 * regular bracketed class, so it silently gets compiled into something
15016 * quite unintended.
15018 * The solution adopted here maintains backward compatibility except that
15019 * it adds a warning if it looks like a posix class was intended but
15020 * improperly specified. The warning is not raised unless what is input
15021 * very closely resembles one of the 14 legal posix classes. To do this,
15022 * it uses fuzzy parsing. It calculates how many single-character edits it
15023 * would take to transform what was input into a legal posix class. Only
15024 * if that number is quite small does it think that the intention was a
15025 * posix class. Obviously these are heuristics, and there will be cases
15026 * where it errs on one side or another, and they can be tweaked as
15027 * experience informs.
15029 * The syntax for a legal posix class is:
15031 * qr/(?xa: \[ : \^? [[:lower:]]{4,6} : \] )/
15033 * What this routine considers syntactically to be an intended posix class
15034 * is this (the comments indicate some restrictions that the pattern
15037 * qr/(?x: \[? # The left bracket, possibly
15039 * \h* # possibly followed by blanks
15040 * (?: \^ \h* )? # possibly a misplaced caret
15041 * [:;]? # The opening class character,
15042 * # possibly omitted. A typo
15043 * # semi-colon can also be used.
15045 * \^? # possibly a correctly placed
15046 * # caret, but not if there was also
15047 * # a misplaced one
15049 * .{3,15} # The class name. If there are
15050 * # deviations from the legal syntax,
15051 * # its edit distance must be close
15052 * # to a real class name in order
15053 * # for it to be considered to be
15054 * # an intended posix class.
15056 * [[:punct:]]? # The closing class character,
15057 * # possibly omitted. If not a colon
15058 * # nor semi colon, the class name
15059 * # must be even closer to a valid
15062 * \]? # The right bracket, possibly
15066 * In the above, \h must be ASCII-only.
15068 * These are heuristics, and can be tweaked as field experience dictates.
15069 * There will be cases when someone didn't intend to specify a posix class
15070 * that this warns as being so. The goal is to minimize these, while
15071 * maximizing the catching of things intended to be a posix class that
15072 * aren't parsed as such.
15076 const char * const e = RExC_end;
15077 unsigned complement = 0; /* If to complement the class */
15078 bool found_problem = FALSE; /* Assume OK until proven otherwise */
15079 bool has_opening_bracket = FALSE;
15080 bool has_opening_colon = FALSE;
15081 int class_number = OOB_NAMEDCLASS; /* Out-of-bounds until find
15083 const char * possible_end = NULL; /* used for a 2nd parse pass */
15084 const char* name_start; /* ptr to class name first char */
15086 /* If the number of single-character typos the input name is away from a
15087 * legal name is no more than this number, it is considered to have meant
15088 * the legal name */
15089 int max_distance = 2;
15091 /* to store the name. The size determines the maximum length before we
15092 * decide that no posix class was intended. Should be at least
15093 * sizeof("alphanumeric") */
15095 STATIC_ASSERT_DECL(C_ARRAY_LENGTH(input_text) >= sizeof "alphanumeric");
15097 PERL_ARGS_ASSERT_HANDLE_POSSIBLE_POSIX;
15099 CLEAR_POSIX_WARNINGS();
15102 return NOT_MEANT_TO_BE_A_POSIX_CLASS;
15105 if (*(p - 1) != '[') {
15106 ADD_POSIX_WARNING(p, "it doesn't start with a '['");
15107 found_problem = TRUE;
15110 has_opening_bracket = TRUE;
15113 /* They could be confused and think you can put spaces between the
15116 found_problem = TRUE;
15120 } while (p < e && isBLANK(*p));
15122 ADD_POSIX_WARNING(p, NO_BLANKS_POSIX_WARNING);
15125 /* For [. .] and [= =]. These are quite different internally from [: :],
15126 * so they are handled separately. */
15127 if (POSIXCC_NOTYET(*p) && p < e - 3) /* 1 for the close, and 1 for the ']'
15128 and 1 for at least one char in it
15131 const char open_char = *p;
15132 const char * temp_ptr = p + 1;
15134 /* These two constructs are not handled by perl, and if we find a
15135 * syntactically valid one, we croak. khw, who wrote this code, finds
15136 * this explanation of them very unclear:
15137 * http://pubs.opengroup.org/onlinepubs/009696899/basedefs/xbd_chap09.html
15138 * And searching the rest of the internet wasn't very helpful either.
15139 * It looks like just about any byte can be in these constructs,
15140 * depending on the locale. But unless the pattern is being compiled
15141 * under /l, which is very rare, Perl runs under the C or POSIX locale.
15142 * In that case, it looks like [= =] isn't allowed at all, and that
15143 * [. .] could be any single code point, but for longer strings the
15144 * constituent characters would have to be the ASCII alphabetics plus
15145 * the minus-hyphen. Any sensible locale definition would limit itself
15146 * to these. And any portable one definitely should. Trying to parse
15147 * the general case is a nightmare (see [perl #127604]). So, this code
15148 * looks only for interiors of these constructs that match:
15150 * Using \w relaxes the apparent rules a little, without adding much
15151 * danger of mistaking something else for one of these constructs.
15153 * [. .] in some implementations described on the internet is usable to
15154 * escape a character that otherwise is special in bracketed character
15155 * classes. For example [.].] means a literal right bracket instead of
15156 * the ending of the class
15158 * [= =] can legitimately contain a [. .] construct, but we don't
15159 * handle this case, as that [. .] construct will later get parsed
15160 * itself and croak then. And [= =] is checked for even when not under
15161 * /l, as Perl has long done so.
15163 * The code below relies on there being a trailing NUL, so it doesn't
15164 * have to keep checking if the parse ptr < e.
15166 if (temp_ptr[1] == open_char) {
15169 else while ( temp_ptr < e
15170 && (isWORDCHAR(*temp_ptr) || *temp_ptr == '-'))
15175 if (*temp_ptr == open_char) {
15177 if (*temp_ptr == ']') {
15179 if (! found_problem && ! check_only) {
15180 RExC_parse = (char *) temp_ptr;
15181 vFAIL3("POSIX syntax [%c %c] is reserved for future "
15182 "extensions", open_char, open_char);
15185 /* Here, the syntax wasn't completely valid, or else the call
15186 * is to check-only */
15187 if (updated_parse_ptr) {
15188 *updated_parse_ptr = (char *) temp_ptr;
15191 CLEAR_POSIX_WARNINGS_AND_RETURN(OOB_NAMEDCLASS);
15195 /* If we find something that started out to look like one of these
15196 * constructs, but isn't, we continue below so that it can be checked
15197 * for being a class name with a typo of '.' or '=' instead of a colon.
15201 /* Here, we think there is a possibility that a [: :] class was meant, and
15202 * we have the first real character. It could be they think the '^' comes
15205 found_problem = TRUE;
15206 ADD_POSIX_WARNING(p + 1, "the '^' must come after the colon");
15211 found_problem = TRUE;
15215 } while (p < e && isBLANK(*p));
15217 ADD_POSIX_WARNING(p, NO_BLANKS_POSIX_WARNING);
15221 /* But the first character should be a colon, which they could have easily
15222 * mistyped on a qwerty keyboard as a semi-colon (and which may be hard to
15223 * distinguish from a colon, so treat that as a colon). */
15226 has_opening_colon = TRUE;
15228 else if (*p == ';') {
15229 found_problem = TRUE;
15231 ADD_POSIX_WARNING(p, SEMI_COLON_POSIX_WARNING);
15232 has_opening_colon = TRUE;
15235 found_problem = TRUE;
15236 ADD_POSIX_WARNING(p, "there must be a starting ':'");
15238 /* Consider an initial punctuation (not one of the recognized ones) to
15239 * be a left terminator */
15240 if (*p != '^' && *p != ']' && isPUNCT(*p)) {
15245 /* They may think that you can put spaces between the components */
15247 found_problem = TRUE;
15251 } while (p < e && isBLANK(*p));
15253 ADD_POSIX_WARNING(p, NO_BLANKS_POSIX_WARNING);
15258 /* We consider something like [^:^alnum:]] to not have been intended to
15259 * be a posix class, but XXX maybe we should */
15261 CLEAR_POSIX_WARNINGS_AND_RETURN(NOT_MEANT_TO_BE_A_POSIX_CLASS);
15268 /* Again, they may think that you can put spaces between the components */
15270 found_problem = TRUE;
15274 } while (p < e && isBLANK(*p));
15276 ADD_POSIX_WARNING(p, NO_BLANKS_POSIX_WARNING);
15281 /* XXX This ']' may be a typo, and something else was meant. But
15282 * treating it as such creates enough complications, that that
15283 * possibility isn't currently considered here. So we assume that the
15284 * ']' is what is intended, and if we've already found an initial '[',
15285 * this leaves this construct looking like [:] or [:^], which almost
15286 * certainly weren't intended to be posix classes */
15287 if (has_opening_bracket) {
15288 CLEAR_POSIX_WARNINGS_AND_RETURN(NOT_MEANT_TO_BE_A_POSIX_CLASS);
15291 /* But this function can be called when we parse the colon for
15292 * something like qr/[alpha:]]/, so we back up to look for the
15297 found_problem = TRUE;
15298 ADD_POSIX_WARNING(p, SEMI_COLON_POSIX_WARNING);
15300 else if (*p != ':') {
15302 /* XXX We are currently very restrictive here, so this code doesn't
15303 * consider the possibility that, say, /[alpha.]]/ was intended to
15304 * be a posix class. */
15305 CLEAR_POSIX_WARNINGS_AND_RETURN(NOT_MEANT_TO_BE_A_POSIX_CLASS);
15308 /* Here we have something like 'foo:]'. There was no initial colon,
15309 * and we back up over 'foo. XXX Unlike the going forward case, we
15310 * don't handle typos of non-word chars in the middle */
15311 has_opening_colon = FALSE;
15314 while (p > RExC_start && isWORDCHAR(*p)) {
15319 /* Here, we have positioned ourselves to where we think the first
15320 * character in the potential class is */
15323 /* Now the interior really starts. There are certain key characters that
15324 * can end the interior, or these could just be typos. To catch both
15325 * cases, we may have to do two passes. In the first pass, we keep on
15326 * going unless we come to a sequence that matches
15327 * qr/ [[:punct:]] [[:blank:]]* \] /xa
15328 * This means it takes a sequence to end the pass, so two typos in a row if
15329 * that wasn't what was intended. If the class is perfectly formed, just
15330 * this one pass is needed. We also stop if there are too many characters
15331 * being accumulated, but this number is deliberately set higher than any
15332 * real class. It is set high enough so that someone who thinks that
15333 * 'alphanumeric' is a correct name would get warned that it wasn't.
15334 * While doing the pass, we keep track of where the key characters were in
15335 * it. If we don't find an end to the class, and one of the key characters
15336 * was found, we redo the pass, but stop when we get to that character.
15337 * Thus the key character was considered a typo in the first pass, but a
15338 * terminator in the second. If two key characters are found, we stop at
15339 * the second one in the first pass. Again this can miss two typos, but
15340 * catches a single one
15342 * In the first pass, 'possible_end' starts as NULL, and then gets set to
15343 * point to the first key character. For the second pass, it starts as -1.
15349 bool has_blank = FALSE;
15350 bool has_upper = FALSE;
15351 bool has_terminating_colon = FALSE;
15352 bool has_terminating_bracket = FALSE;
15353 bool has_semi_colon = FALSE;
15354 unsigned int name_len = 0;
15355 int punct_count = 0;
15359 /* Squeeze out blanks when looking up the class name below */
15360 if (isBLANK(*p) ) {
15362 found_problem = TRUE;
15367 /* The name will end with a punctuation */
15369 const char * peek = p + 1;
15371 /* Treat any non-']' punctuation followed by a ']' (possibly
15372 * with intervening blanks) as trying to terminate the class.
15373 * ']]' is very likely to mean a class was intended (but
15374 * missing the colon), but the warning message that gets
15375 * generated shows the error position better if we exit the
15376 * loop at the bottom (eventually), so skip it here. */
15378 if (peek < e && isBLANK(*peek)) {
15380 found_problem = TRUE;
15383 } while (peek < e && isBLANK(*peek));
15386 if (peek < e && *peek == ']') {
15387 has_terminating_bracket = TRUE;
15389 has_terminating_colon = TRUE;
15391 else if (*p == ';') {
15392 has_semi_colon = TRUE;
15393 has_terminating_colon = TRUE;
15396 found_problem = TRUE;
15403 /* Here we have punctuation we thought didn't end the class.
15404 * Keep track of the position of the key characters that are
15405 * more likely to have been class-enders */
15406 if (*p == ']' || *p == '[' || *p == ':' || *p == ';') {
15408 /* Allow just one such possible class-ender not actually
15409 * ending the class. */
15410 if (possible_end) {
15416 /* If we have too many punctuation characters, no use in
15418 if (++punct_count > max_distance) {
15422 /* Treat the punctuation as a typo. */
15423 input_text[name_len++] = *p;
15426 else if (isUPPER(*p)) { /* Use lowercase for lookup */
15427 input_text[name_len++] = toLOWER(*p);
15429 found_problem = TRUE;
15431 } else if (! UTF || UTF8_IS_INVARIANT(*p)) {
15432 input_text[name_len++] = *p;
15436 input_text[name_len++] = utf8_to_uvchr_buf((U8 *) p, e, NULL);
15440 /* The declaration of 'input_text' is how long we allow a potential
15441 * class name to be, before saying they didn't mean a class name at
15443 if (name_len >= C_ARRAY_LENGTH(input_text)) {
15448 /* We get to here when the possible class name hasn't been properly
15449 * terminated before:
15450 * 1) we ran off the end of the pattern; or
15451 * 2) found two characters, each of which might have been intended to
15452 * be the name's terminator
15453 * 3) found so many punctuation characters in the purported name,
15454 * that the edit distance to a valid one is exceeded
15455 * 4) we decided it was more characters than anyone could have
15456 * intended to be one. */
15458 found_problem = TRUE;
15460 /* In the final two cases, we know that looking up what we've
15461 * accumulated won't lead to a match, even a fuzzy one. */
15462 if ( name_len >= C_ARRAY_LENGTH(input_text)
15463 || punct_count > max_distance)
15465 /* If there was an intermediate key character that could have been
15466 * an intended end, redo the parse, but stop there */
15467 if (possible_end && possible_end != (char *) -1) {
15468 possible_end = (char *) -1; /* Special signal value to say
15469 we've done a first pass */
15474 /* Otherwise, it can't have meant to have been a class */
15475 CLEAR_POSIX_WARNINGS_AND_RETURN(NOT_MEANT_TO_BE_A_POSIX_CLASS);
15478 /* If we ran off the end, and the final character was a punctuation
15479 * one, back up one, to look at that final one just below. Later, we
15480 * will restore the parse pointer if appropriate */
15481 if (name_len && p == e && isPUNCT(*(p-1))) {
15486 if (p < e && isPUNCT(*p)) {
15488 has_terminating_bracket = TRUE;
15490 /* If this is a 2nd ']', and the first one is just below this
15491 * one, consider that to be the real terminator. This gives a
15492 * uniform and better positioning for the warning message */
15494 && possible_end != (char *) -1
15495 && *possible_end == ']'
15496 && name_len && input_text[name_len - 1] == ']')
15501 /* And this is actually equivalent to having done the 2nd
15502 * pass now, so set it to not try again */
15503 possible_end = (char *) -1;
15508 has_terminating_colon = TRUE;
15510 else if (*p == ';') {
15511 has_semi_colon = TRUE;
15512 has_terminating_colon = TRUE;
15520 /* Here, we have a class name to look up. We can short circuit the
15521 * stuff below for short names that can't possibly be meant to be a
15522 * class name. (We can do this on the first pass, as any second pass
15523 * will yield an even shorter name) */
15524 if (name_len < 3) {
15525 CLEAR_POSIX_WARNINGS_AND_RETURN(NOT_MEANT_TO_BE_A_POSIX_CLASS);
15528 /* Find which class it is. Initially switch on the length of the name.
15530 switch (name_len) {
15532 if (memEQs(name_start, 4, "word")) {
15533 /* this is not POSIX, this is the Perl \w */
15534 class_number = ANYOF_WORDCHAR;
15538 /* Names all of length 5: alnum alpha ascii blank cntrl digit
15539 * graph lower print punct space upper
15540 * Offset 4 gives the best switch position. */
15541 switch (name_start[4]) {
15543 if (memBEGINs(name_start, 5, "alph")) /* alpha */
15544 class_number = ANYOF_ALPHA;
15547 if (memBEGINs(name_start, 5, "spac")) /* space */
15548 class_number = ANYOF_SPACE;
15551 if (memBEGINs(name_start, 5, "grap")) /* graph */
15552 class_number = ANYOF_GRAPH;
15555 if (memBEGINs(name_start, 5, "asci")) /* ascii */
15556 class_number = ANYOF_ASCII;
15559 if (memBEGINs(name_start, 5, "blan")) /* blank */
15560 class_number = ANYOF_BLANK;
15563 if (memBEGINs(name_start, 5, "cntr")) /* cntrl */
15564 class_number = ANYOF_CNTRL;
15567 if (memBEGINs(name_start, 5, "alnu")) /* alnum */
15568 class_number = ANYOF_ALPHANUMERIC;
15571 if (memBEGINs(name_start, 5, "lowe")) /* lower */
15572 class_number = (FOLD) ? ANYOF_CASED : ANYOF_LOWER;
15573 else if (memBEGINs(name_start, 5, "uppe")) /* upper */
15574 class_number = (FOLD) ? ANYOF_CASED : ANYOF_UPPER;
15577 if (memBEGINs(name_start, 5, "digi")) /* digit */
15578 class_number = ANYOF_DIGIT;
15579 else if (memBEGINs(name_start, 5, "prin")) /* print */
15580 class_number = ANYOF_PRINT;
15581 else if (memBEGINs(name_start, 5, "punc")) /* punct */
15582 class_number = ANYOF_PUNCT;
15587 if (memEQs(name_start, 6, "xdigit"))
15588 class_number = ANYOF_XDIGIT;
15592 /* If the name exactly matches a posix class name the class number will
15593 * here be set to it, and the input almost certainly was meant to be a
15594 * posix class, so we can skip further checking. If instead the syntax
15595 * is exactly correct, but the name isn't one of the legal ones, we
15596 * will return that as an error below. But if neither of these apply,
15597 * it could be that no posix class was intended at all, or that one
15598 * was, but there was a typo. We tease these apart by doing fuzzy
15599 * matching on the name */
15600 if (class_number == OOB_NAMEDCLASS && found_problem) {
15601 const UV posix_names[][6] = {
15602 { 'a', 'l', 'n', 'u', 'm' },
15603 { 'a', 'l', 'p', 'h', 'a' },
15604 { 'a', 's', 'c', 'i', 'i' },
15605 { 'b', 'l', 'a', 'n', 'k' },
15606 { 'c', 'n', 't', 'r', 'l' },
15607 { 'd', 'i', 'g', 'i', 't' },
15608 { 'g', 'r', 'a', 'p', 'h' },
15609 { 'l', 'o', 'w', 'e', 'r' },
15610 { 'p', 'r', 'i', 'n', 't' },
15611 { 'p', 'u', 'n', 'c', 't' },
15612 { 's', 'p', 'a', 'c', 'e' },
15613 { 'u', 'p', 'p', 'e', 'r' },
15614 { 'w', 'o', 'r', 'd' },
15615 { 'x', 'd', 'i', 'g', 'i', 't' }
15617 /* The names of the above all have added NULs to make them the same
15618 * size, so we need to also have the real lengths */
15619 const UV posix_name_lengths[] = {
15620 sizeof("alnum") - 1,
15621 sizeof("alpha") - 1,
15622 sizeof("ascii") - 1,
15623 sizeof("blank") - 1,
15624 sizeof("cntrl") - 1,
15625 sizeof("digit") - 1,
15626 sizeof("graph") - 1,
15627 sizeof("lower") - 1,
15628 sizeof("print") - 1,
15629 sizeof("punct") - 1,
15630 sizeof("space") - 1,
15631 sizeof("upper") - 1,
15632 sizeof("word") - 1,
15633 sizeof("xdigit")- 1
15636 int temp_max = max_distance; /* Use a temporary, so if we
15637 reparse, we haven't changed the
15640 /* Use a smaller max edit distance if we are missing one of the
15642 if ( has_opening_bracket + has_opening_colon < 2
15643 || has_terminating_bracket + has_terminating_colon < 2)
15648 /* See if the input name is close to a legal one */
15649 for (i = 0; i < C_ARRAY_LENGTH(posix_names); i++) {
15651 /* Short circuit call if the lengths are too far apart to be
15653 if (abs( (int) (name_len - posix_name_lengths[i]))
15659 if (edit_distance(input_text,
15662 posix_name_lengths[i],
15666 { /* If it is close, it probably was intended to be a class */
15667 goto probably_meant_to_be;
15671 /* Here the input name is not close enough to a valid class name
15672 * for us to consider it to be intended to be a posix class. If
15673 * we haven't already done so, and the parse found a character that
15674 * could have been terminators for the name, but which we absorbed
15675 * as typos during the first pass, repeat the parse, signalling it
15676 * to stop at that character */
15677 if (possible_end && possible_end != (char *) -1) {
15678 possible_end = (char *) -1;
15683 /* Here neither pass found a close-enough class name */
15684 CLEAR_POSIX_WARNINGS_AND_RETURN(NOT_MEANT_TO_BE_A_POSIX_CLASS);
15687 probably_meant_to_be:
15689 /* Here we think that a posix specification was intended. Update any
15691 if (updated_parse_ptr) {
15692 *updated_parse_ptr = (char *) p;
15695 /* If a posix class name was intended but incorrectly specified, we
15696 * output or return the warnings */
15697 if (found_problem) {
15699 /* We set flags for these issues in the parse loop above instead of
15700 * adding them to the list of warnings, because we can parse it
15701 * twice, and we only want one warning instance */
15703 ADD_POSIX_WARNING(p, "the name must be all lowercase letters");
15706 ADD_POSIX_WARNING(p, NO_BLANKS_POSIX_WARNING);
15708 if (has_semi_colon) {
15709 ADD_POSIX_WARNING(p, SEMI_COLON_POSIX_WARNING);
15711 else if (! has_terminating_colon) {
15712 ADD_POSIX_WARNING(p, "there is no terminating ':'");
15714 if (! has_terminating_bracket) {
15715 ADD_POSIX_WARNING(p, "there is no terminating ']'");
15718 if ( posix_warnings
15720 && av_top_index(RExC_warn_text) > -1)
15722 *posix_warnings = RExC_warn_text;
15725 else if (class_number != OOB_NAMEDCLASS) {
15726 /* If it is a known class, return the class. The class number
15727 * #defines are structured so each complement is +1 to the normal
15729 CLEAR_POSIX_WARNINGS_AND_RETURN(class_number + complement);
15731 else if (! check_only) {
15733 /* Here, it is an unrecognized class. This is an error (unless the
15734 * call is to check only, which we've already handled above) */
15735 const char * const complement_string = (complement)
15738 RExC_parse = (char *) p;
15739 vFAIL3utf8f("POSIX class [:%s%" UTF8f ":] unknown",
15741 UTF8fARG(UTF, RExC_parse - name_start - 2, name_start));
15745 return OOB_NAMEDCLASS;
15747 #undef ADD_POSIX_WARNING
15749 STATIC unsigned int
15750 S_regex_set_precedence(const U8 my_operator) {
15752 /* Returns the precedence in the (?[...]) construct of the input operator,
15753 * specified by its character representation. The precedence follows
15754 * general Perl rules, but it extends this so that ')' and ']' have (low)
15755 * precedence even though they aren't really operators */
15757 switch (my_operator) {
15773 NOT_REACHED; /* NOTREACHED */
15774 return 0; /* Silence compiler warning */
15777 STATIC regnode_offset
15778 S_handle_regex_sets(pTHX_ RExC_state_t *pRExC_state, SV** return_invlist,
15779 I32 *flagp, U32 depth,
15780 char * const oregcomp_parse)
15782 /* Handle the (?[...]) construct to do set operations */
15784 U8 curchar; /* Current character being parsed */
15785 UV start, end; /* End points of code point ranges */
15786 SV* final = NULL; /* The end result inversion list */
15787 SV* result_string; /* 'final' stringified */
15788 AV* stack; /* stack of operators and operands not yet
15790 AV* fence_stack = NULL; /* A stack containing the positions in
15791 'stack' of where the undealt-with left
15792 parens would be if they were actually
15794 /* The 'volatile' is a workaround for an optimiser bug
15795 * in Solaris Studio 12.3. See RT #127455 */
15796 volatile IV fence = 0; /* Position of where most recent undealt-
15797 with left paren in stack is; -1 if none.
15799 STRLEN len; /* Temporary */
15800 regnode_offset node; /* Temporary, and final regnode returned by
15802 const bool save_fold = FOLD; /* Temporary */
15803 char *save_end, *save_parse; /* Temporaries */
15804 const bool in_locale = LOC; /* we turn off /l during processing */
15806 GET_RE_DEBUG_FLAGS_DECL;
15808 PERL_ARGS_ASSERT_HANDLE_REGEX_SETS;
15810 DEBUG_PARSE("xcls");
15813 set_regex_charset(&RExC_flags, REGEX_UNICODE_CHARSET);
15816 /* The use of this operator implies /u. This is required so that the
15817 * compile time values are valid in all runtime cases */
15818 REQUIRE_UNI_RULES(flagp, 0);
15820 ckWARNexperimental(RExC_parse,
15821 WARN_EXPERIMENTAL__REGEX_SETS,
15822 "The regex_sets feature is experimental");
15824 /* Everything in this construct is a metacharacter. Operands begin with
15825 * either a '\' (for an escape sequence), or a '[' for a bracketed
15826 * character class. Any other character should be an operator, or
15827 * parenthesis for grouping. Both types of operands are handled by calling
15828 * regclass() to parse them. It is called with a parameter to indicate to
15829 * return the computed inversion list. The parsing here is implemented via
15830 * a stack. Each entry on the stack is a single character representing one
15831 * of the operators; or else a pointer to an operand inversion list. */
15833 #define IS_OPERATOR(a) SvIOK(a)
15834 #define IS_OPERAND(a) (! IS_OPERATOR(a))
15836 /* The stack is kept in Łukasiewicz order. (That's pronounced similar
15837 * to luke-a-shave-itch (or -itz), but people who didn't want to bother
15838 * with pronouncing it called it Reverse Polish instead, but now that YOU
15839 * know how to pronounce it you can use the correct term, thus giving due
15840 * credit to the person who invented it, and impressing your geek friends.
15841 * Wikipedia says that the pronounciation of "Ł" has been changing so that
15842 * it is now more like an English initial W (as in wonk) than an L.)
15844 * This means that, for example, 'a | b & c' is stored on the stack as
15852 * where the numbers in brackets give the stack [array] element number.
15853 * In this implementation, parentheses are not stored on the stack.
15854 * Instead a '(' creates a "fence" so that the part of the stack below the
15855 * fence is invisible except to the corresponding ')' (this allows us to
15856 * replace testing for parens, by using instead subtraction of the fence
15857 * position). As new operands are processed they are pushed onto the stack
15858 * (except as noted in the next paragraph). New operators of higher
15859 * precedence than the current final one are inserted on the stack before
15860 * the lhs operand (so that when the rhs is pushed next, everything will be
15861 * in the correct positions shown above. When an operator of equal or
15862 * lower precedence is encountered in parsing, all the stacked operations
15863 * of equal or higher precedence are evaluated, leaving the result as the
15864 * top entry on the stack. This makes higher precedence operations
15865 * evaluate before lower precedence ones, and causes operations of equal
15866 * precedence to left associate.
15868 * The only unary operator '!' is immediately pushed onto the stack when
15869 * encountered. When an operand is encountered, if the top of the stack is
15870 * a '!", the complement is immediately performed, and the '!' popped. The
15871 * resulting value is treated as a new operand, and the logic in the
15872 * previous paragraph is executed. Thus in the expression
15874 * the stack looks like
15880 * as 'b' gets parsed, the latter gets evaluated to '!b', and the stack
15887 * A ')' is treated as an operator with lower precedence than all the
15888 * aforementioned ones, which causes all operations on the stack above the
15889 * corresponding '(' to be evaluated down to a single resultant operand.
15890 * Then the fence for the '(' is removed, and the operand goes through the
15891 * algorithm above, without the fence.
15893 * A separate stack is kept of the fence positions, so that the position of
15894 * the latest so-far unbalanced '(' is at the top of it.
15896 * The ']' ending the construct is treated as the lowest operator of all,
15897 * so that everything gets evaluated down to a single operand, which is the
15900 sv_2mortal((SV *)(stack = newAV()));
15901 sv_2mortal((SV *)(fence_stack = newAV()));
15903 while (RExC_parse < RExC_end) {
15904 I32 top_index; /* Index of top-most element in 'stack' */
15905 SV** top_ptr; /* Pointer to top 'stack' element */
15906 SV* current = NULL; /* To contain the current inversion list
15908 SV* only_to_avoid_leaks;
15910 skip_to_be_ignored_text(pRExC_state, &RExC_parse,
15911 TRUE /* Force /x */ );
15912 if (RExC_parse >= RExC_end) { /* Fail */
15916 curchar = UCHARAT(RExC_parse);
15920 #ifdef ENABLE_REGEX_SETS_DEBUGGING
15921 /* Enable with -Accflags=-DENABLE_REGEX_SETS_DEBUGGING */
15922 DEBUG_U(dump_regex_sets_structures(pRExC_state,
15923 stack, fence, fence_stack));
15926 top_index = av_tindex_skip_len_mg(stack);
15929 SV** stacked_ptr; /* Ptr to something already on 'stack' */
15930 char stacked_operator; /* The topmost operator on the 'stack'. */
15931 SV* lhs; /* Operand to the left of the operator */
15932 SV* rhs; /* Operand to the right of the operator */
15933 SV* fence_ptr; /* Pointer to top element of the fence
15938 if ( RExC_parse < RExC_end - 2
15939 && UCHARAT(RExC_parse + 1) == '?'
15940 && UCHARAT(RExC_parse + 2) == '^')
15942 /* If is a '(?', could be an embedded '(?^flags:(?[...])'.
15943 * This happens when we have some thing like
15945 * my $thai_or_lao = qr/(?[ \p{Thai} + \p{Lao} ])/;
15947 * qr/(?[ \p{Digit} & $thai_or_lao ])/;
15949 * Here we would be handling the interpolated
15950 * '$thai_or_lao'. We handle this by a recursive call to
15951 * ourselves which returns the inversion list the
15952 * interpolated expression evaluates to. We use the flags
15953 * from the interpolated pattern. */
15954 U32 save_flags = RExC_flags;
15955 const char * save_parse;
15957 RExC_parse += 2; /* Skip past the '(?' */
15958 save_parse = RExC_parse;
15960 /* Parse the flags for the '(?'. We already know the first
15961 * flag to parse is a '^' */
15962 parse_lparen_question_flags(pRExC_state);
15964 if ( RExC_parse >= RExC_end - 4
15965 || UCHARAT(RExC_parse) != ':'
15966 || UCHARAT(++RExC_parse) != '('
15967 || UCHARAT(++RExC_parse) != '?'
15968 || UCHARAT(++RExC_parse) != '[')
15971 /* In combination with the above, this moves the
15972 * pointer to the point just after the first erroneous
15974 if (RExC_parse >= RExC_end - 4) {
15975 RExC_parse = RExC_end;
15977 else if (RExC_parse != save_parse) {
15978 RExC_parse += (UTF)
15979 ? UTF8_SAFE_SKIP(RExC_parse, RExC_end)
15982 vFAIL("Expecting '(?flags:(?[...'");
15985 /* Recurse, with the meat of the embedded expression */
15987 (void) handle_regex_sets(pRExC_state, ¤t, flagp,
15988 depth+1, oregcomp_parse);
15990 /* Here, 'current' contains the embedded expression's
15991 * inversion list, and RExC_parse points to the trailing
15992 * ']'; the next character should be the ')' */
15994 if (UCHARAT(RExC_parse) != ')')
15995 vFAIL("Expecting close paren for nested extended charclass");
15997 /* Then the ')' matching the original '(' handled by this
15998 * case: statement */
16000 if (UCHARAT(RExC_parse) != ')')
16001 vFAIL("Expecting close paren for wrapper for nested extended charclass");
16003 RExC_flags = save_flags;
16004 goto handle_operand;
16007 /* A regular '('. Look behind for illegal syntax */
16008 if (top_index - fence >= 0) {
16009 /* If the top entry on the stack is an operator, it had
16010 * better be a '!', otherwise the entry below the top
16011 * operand should be an operator */
16012 if ( ! (top_ptr = av_fetch(stack, top_index, FALSE))
16013 || (IS_OPERATOR(*top_ptr) && SvUV(*top_ptr) != '!')
16014 || ( IS_OPERAND(*top_ptr)
16015 && ( top_index - fence < 1
16016 || ! (stacked_ptr = av_fetch(stack,
16019 || ! IS_OPERATOR(*stacked_ptr))))
16022 vFAIL("Unexpected '(' with no preceding operator");
16026 /* Stack the position of this undealt-with left paren */
16027 av_push(fence_stack, newSViv(fence));
16028 fence = top_index + 1;
16032 /* regclass() can only return RESTART_PARSE and NEED_UTF8 if
16033 * multi-char folds are allowed. */
16034 if (!regclass(pRExC_state, flagp, depth+1,
16035 TRUE, /* means parse just the next thing */
16036 FALSE, /* don't allow multi-char folds */
16037 FALSE, /* don't silence non-portable warnings. */
16039 FALSE, /* Require return to be an ANYOF */
16042 goto regclass_failed;
16045 /* regclass() will return with parsing just the \ sequence,
16046 * leaving the parse pointer at the next thing to parse */
16048 goto handle_operand;
16050 case '[': /* Is a bracketed character class */
16052 /* See if this is a [:posix:] class. */
16053 bool is_posix_class = (OOB_NAMEDCLASS
16054 < handle_possible_posix(pRExC_state,
16058 TRUE /* checking only */));
16059 /* If it is a posix class, leave the parse pointer at the '['
16060 * to fool regclass() into thinking it is part of a
16061 * '[[:posix:]]'. */
16062 if (! is_posix_class) {
16066 /* regclass() can only return RESTART_PARSE and NEED_UTF8 if
16067 * multi-char folds are allowed. */
16068 if (!regclass(pRExC_state, flagp, depth+1,
16069 is_posix_class, /* parse the whole char
16070 class only if not a
16072 FALSE, /* don't allow multi-char folds */
16073 TRUE, /* silence non-portable warnings. */
16075 FALSE, /* Require return to be an ANYOF */
16078 goto regclass_failed;
16085 /* function call leaves parse pointing to the ']', except if we
16087 if (is_posix_class) {
16091 goto handle_operand;
16095 if (top_index >= 1) {
16096 goto join_operators;
16099 /* Only a single operand on the stack: are done */
16103 if (av_tindex_skip_len_mg(fence_stack) < 0) {
16104 if (UCHARAT(RExC_parse - 1) == ']') {
16108 vFAIL("Unexpected ')'");
16111 /* If nothing after the fence, is missing an operand */
16112 if (top_index - fence < 0) {
16116 /* If at least two things on the stack, treat this as an
16118 if (top_index - fence >= 1) {
16119 goto join_operators;
16122 /* Here only a single thing on the fenced stack, and there is a
16123 * fence. Get rid of it */
16124 fence_ptr = av_pop(fence_stack);
16126 fence = SvIV(fence_ptr);
16127 SvREFCNT_dec_NN(fence_ptr);
16134 /* Having gotten rid of the fence, we pop the operand at the
16135 * stack top and process it as a newly encountered operand */
16136 current = av_pop(stack);
16137 if (IS_OPERAND(current)) {
16138 goto handle_operand;
16150 /* These binary operators should have a left operand already
16152 if ( top_index - fence < 0
16153 || top_index - fence == 1
16154 || ( ! (top_ptr = av_fetch(stack, top_index, FALSE)))
16155 || ! IS_OPERAND(*top_ptr))
16157 goto unexpected_binary;
16160 /* If only the one operand is on the part of the stack visible
16161 * to us, we just place this operator in the proper position */
16162 if (top_index - fence < 2) {
16164 /* Place the operator before the operand */
16166 SV* lhs = av_pop(stack);
16167 av_push(stack, newSVuv(curchar));
16168 av_push(stack, lhs);
16172 /* But if there is something else on the stack, we need to
16173 * process it before this new operator if and only if the
16174 * stacked operation has equal or higher precedence than the
16179 /* The operator on the stack is supposed to be below both its
16181 if ( ! (stacked_ptr = av_fetch(stack, top_index - 2, FALSE))
16182 || IS_OPERAND(*stacked_ptr))
16184 /* But if not, it's legal and indicates we are completely
16185 * done if and only if we're currently processing a ']',
16186 * which should be the final thing in the expression */
16187 if (curchar == ']') {
16193 vFAIL2("Unexpected binary operator '%c' with no "
16194 "preceding operand", curchar);
16196 stacked_operator = (char) SvUV(*stacked_ptr);
16198 if (regex_set_precedence(curchar)
16199 > regex_set_precedence(stacked_operator))
16201 /* Here, the new operator has higher precedence than the
16202 * stacked one. This means we need to add the new one to
16203 * the stack to await its rhs operand (and maybe more
16204 * stuff). We put it before the lhs operand, leaving
16205 * untouched the stacked operator and everything below it
16207 lhs = av_pop(stack);
16208 assert(IS_OPERAND(lhs));
16210 av_push(stack, newSVuv(curchar));
16211 av_push(stack, lhs);
16215 /* Here, the new operator has equal or lower precedence than
16216 * what's already there. This means the operation already
16217 * there should be performed now, before the new one. */
16219 rhs = av_pop(stack);
16220 if (! IS_OPERAND(rhs)) {
16222 /* This can happen when a ! is not followed by an operand,
16223 * like in /(?[\t &!])/ */
16227 lhs = av_pop(stack);
16229 if (! IS_OPERAND(lhs)) {
16231 /* This can happen when there is an empty (), like in
16232 * /(?[[0]+()+])/ */
16236 switch (stacked_operator) {
16238 _invlist_intersection(lhs, rhs, &rhs);
16243 _invlist_union(lhs, rhs, &rhs);
16247 _invlist_subtract(lhs, rhs, &rhs);
16250 case '^': /* The union minus the intersection */
16255 _invlist_union(lhs, rhs, &u);
16256 _invlist_intersection(lhs, rhs, &i);
16257 _invlist_subtract(u, i, &rhs);
16258 SvREFCNT_dec_NN(i);
16259 SvREFCNT_dec_NN(u);
16265 /* Here, the higher precedence operation has been done, and the
16266 * result is in 'rhs'. We overwrite the stacked operator with
16267 * the result. Then we redo this code to either push the new
16268 * operator onto the stack or perform any higher precedence
16269 * stacked operation */
16270 only_to_avoid_leaks = av_pop(stack);
16271 SvREFCNT_dec(only_to_avoid_leaks);
16272 av_push(stack, rhs);
16275 case '!': /* Highest priority, right associative */
16277 /* If what's already at the top of the stack is another '!",
16278 * they just cancel each other out */
16279 if ( (top_ptr = av_fetch(stack, top_index, FALSE))
16280 && (IS_OPERATOR(*top_ptr) && SvUV(*top_ptr) == '!'))
16282 only_to_avoid_leaks = av_pop(stack);
16283 SvREFCNT_dec(only_to_avoid_leaks);
16285 else { /* Otherwise, since it's right associative, just push
16287 av_push(stack, newSVuv(curchar));
16292 RExC_parse += (UTF) ? UTF8SKIP(RExC_parse) : 1;
16293 if (RExC_parse >= RExC_end) {
16296 vFAIL("Unexpected character");
16300 /* Here 'current' is the operand. If something is already on the
16301 * stack, we have to check if it is a !. But first, the code above
16302 * may have altered the stack in the time since we earlier set
16305 top_index = av_tindex_skip_len_mg(stack);
16306 if (top_index - fence >= 0) {
16307 /* If the top entry on the stack is an operator, it had better
16308 * be a '!', otherwise the entry below the top operand should
16309 * be an operator */
16310 top_ptr = av_fetch(stack, top_index, FALSE);
16312 if (IS_OPERATOR(*top_ptr)) {
16314 /* The only permissible operator at the top of the stack is
16315 * '!', which is applied immediately to this operand. */
16316 curchar = (char) SvUV(*top_ptr);
16317 if (curchar != '!') {
16318 SvREFCNT_dec(current);
16319 vFAIL2("Unexpected binary operator '%c' with no "
16320 "preceding operand", curchar);
16323 _invlist_invert(current);
16325 only_to_avoid_leaks = av_pop(stack);
16326 SvREFCNT_dec(only_to_avoid_leaks);
16328 /* And we redo with the inverted operand. This allows
16329 * handling multiple ! in a row */
16330 goto handle_operand;
16332 /* Single operand is ok only for the non-binary ')'
16334 else if ((top_index - fence == 0 && curchar != ')')
16335 || (top_index - fence > 0
16336 && (! (stacked_ptr = av_fetch(stack,
16339 || IS_OPERAND(*stacked_ptr))))
16341 SvREFCNT_dec(current);
16342 vFAIL("Operand with no preceding operator");
16346 /* Here there was nothing on the stack or the top element was
16347 * another operand. Just add this new one */
16348 av_push(stack, current);
16350 } /* End of switch on next parse token */
16352 RExC_parse += (UTF) ? UTF8SKIP(RExC_parse) : 1;
16353 } /* End of loop parsing through the construct */
16355 vFAIL("Syntax error in (?[...])");
16359 if (RExC_parse >= RExC_end || RExC_parse[1] != ')') {
16360 if (RExC_parse < RExC_end) {
16364 vFAIL("Unexpected ']' with no following ')' in (?[...");
16367 if (av_tindex_skip_len_mg(fence_stack) >= 0) {
16368 vFAIL("Unmatched (");
16371 if (av_tindex_skip_len_mg(stack) < 0 /* Was empty */
16372 || ((final = av_pop(stack)) == NULL)
16373 || ! IS_OPERAND(final)
16374 || ! is_invlist(final)
16375 || av_tindex_skip_len_mg(stack) >= 0) /* More left on stack */
16378 SvREFCNT_dec(final);
16379 vFAIL("Incomplete expression within '(?[ ])'");
16382 /* Here, 'final' is the resultant inversion list from evaluating the
16383 * expression. Return it if so requested */
16384 if (return_invlist) {
16385 *return_invlist = final;
16389 /* Otherwise generate a resultant node, based on 'final'. regclass() is
16390 * expecting a string of ranges and individual code points */
16391 invlist_iterinit(final);
16392 result_string = newSVpvs("");
16393 while (invlist_iternext(final, &start, &end)) {
16394 if (start == end) {
16395 Perl_sv_catpvf(aTHX_ result_string, "\\x{%" UVXf "}", start);
16398 Perl_sv_catpvf(aTHX_ result_string, "\\x{%" UVXf "}-\\x{%" UVXf "}",
16403 /* About to generate an ANYOF (or similar) node from the inversion list we
16404 * have calculated */
16405 save_parse = RExC_parse;
16406 RExC_parse = SvPV(result_string, len);
16407 save_end = RExC_end;
16408 RExC_end = RExC_parse + len;
16409 TURN_OFF_WARNINGS_IN_SUBSTITUTE_PARSE;
16411 /* We turn off folding around the call, as the class we have constructed
16412 * already has all folding taken into consideration, and we don't want
16413 * regclass() to add to that */
16414 RExC_flags &= ~RXf_PMf_FOLD;
16415 /* regclass() can only return RESTART_PARSE and NEED_UTF8 if multi-char
16416 * folds are allowed. */
16417 node = regclass(pRExC_state, flagp, depth+1,
16418 FALSE, /* means parse the whole char class */
16419 FALSE, /* don't allow multi-char folds */
16420 TRUE, /* silence non-portable warnings. The above may very
16421 well have generated non-portable code points, but
16422 they're valid on this machine */
16423 FALSE, /* similarly, no need for strict */
16424 FALSE, /* Require return to be an ANYOF */
16429 RExC_parse = save_parse + 1;
16430 RExC_end = save_end;
16431 SvREFCNT_dec_NN(final);
16432 SvREFCNT_dec_NN(result_string);
16435 RExC_flags |= RXf_PMf_FOLD;
16439 goto regclass_failed;
16441 /* Fix up the node type if we are in locale. (We have pretended we are
16442 * under /u for the purposes of regclass(), as this construct will only
16443 * work under UTF-8 locales. But now we change the opcode to be ANYOFL (so
16444 * as to cause any warnings about bad locales to be output in regexec.c),
16445 * and add the flag that indicates to check if not in a UTF-8 locale. The
16446 * reason we above forbid optimization into something other than an ANYOF
16447 * node is simply to minimize the number of code changes in regexec.c.
16448 * Otherwise we would have to create new EXACTish node types and deal with
16449 * them. This decision could be revisited should this construct become
16452 * (One might think we could look at the resulting ANYOF node and suppress
16453 * the flag if everything is above 255, as those would be UTF-8 only,
16454 * but this isn't true, as the components that led to that result could
16455 * have been locale-affected, and just happen to cancel each other out
16456 * under UTF-8 locales.) */
16458 set_regex_charset(&RExC_flags, REGEX_LOCALE_CHARSET);
16460 assert(OP(REGNODE_p(node)) == ANYOF);
16462 OP(REGNODE_p(node)) = ANYOFL;
16463 ANYOF_FLAGS(REGNODE_p(node))
16464 |= ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD;
16467 nextchar(pRExC_state);
16468 Set_Node_Length(REGNODE_p(node), RExC_parse - oregcomp_parse + 1); /* MJD */
16472 FAIL2("panic: regclass returned failure to handle_sets, " "flags=%#" UVxf,
16476 #ifdef ENABLE_REGEX_SETS_DEBUGGING
16479 S_dump_regex_sets_structures(pTHX_ RExC_state_t *pRExC_state,
16480 AV * stack, const IV fence, AV * fence_stack)
16481 { /* Dumps the stacks in handle_regex_sets() */
16483 const SSize_t stack_top = av_tindex_skip_len_mg(stack);
16484 const SSize_t fence_stack_top = av_tindex_skip_len_mg(fence_stack);
16487 PERL_ARGS_ASSERT_DUMP_REGEX_SETS_STRUCTURES;
16489 PerlIO_printf(Perl_debug_log, "\nParse position is:%s\n", RExC_parse);
16491 if (stack_top < 0) {
16492 PerlIO_printf(Perl_debug_log, "Nothing on stack\n");
16495 PerlIO_printf(Perl_debug_log, "Stack: (fence=%d)\n", (int) fence);
16496 for (i = stack_top; i >= 0; i--) {
16497 SV ** element_ptr = av_fetch(stack, i, FALSE);
16498 if (! element_ptr) {
16501 if (IS_OPERATOR(*element_ptr)) {
16502 PerlIO_printf(Perl_debug_log, "[%d]: %c\n",
16503 (int) i, (int) SvIV(*element_ptr));
16506 PerlIO_printf(Perl_debug_log, "[%d] ", (int) i);
16507 sv_dump(*element_ptr);
16512 if (fence_stack_top < 0) {
16513 PerlIO_printf(Perl_debug_log, "Nothing on fence_stack\n");
16516 PerlIO_printf(Perl_debug_log, "Fence_stack: \n");
16517 for (i = fence_stack_top; i >= 0; i--) {
16518 SV ** element_ptr = av_fetch(fence_stack, i, FALSE);
16519 if (! element_ptr) {
16522 PerlIO_printf(Perl_debug_log, "[%d]: %d\n",
16523 (int) i, (int) SvIV(*element_ptr));
16534 S_add_above_Latin1_folds(pTHX_ RExC_state_t *pRExC_state, const U8 cp, SV** invlist)
16536 /* This adds the Latin1/above-Latin1 folding rules.
16538 * This should be called only for a Latin1-range code points, cp, which is
16539 * known to be involved in a simple fold with other code points above
16540 * Latin1. It would give false results if /aa has been specified.
16541 * Multi-char folds are outside the scope of this, and must be handled
16544 PERL_ARGS_ASSERT_ADD_ABOVE_LATIN1_FOLDS;
16546 assert(HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE(cp));
16548 /* The rules that are valid for all Unicode versions are hard-coded in */
16553 add_cp_to_invlist(*invlist, KELVIN_SIGN);
16557 *invlist = add_cp_to_invlist(*invlist, LATIN_SMALL_LETTER_LONG_S);
16560 *invlist = add_cp_to_invlist(*invlist, GREEK_CAPITAL_LETTER_MU);
16561 *invlist = add_cp_to_invlist(*invlist, GREEK_SMALL_LETTER_MU);
16563 case LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE:
16564 case LATIN_SMALL_LETTER_A_WITH_RING_ABOVE:
16565 *invlist = add_cp_to_invlist(*invlist, ANGSTROM_SIGN);
16567 case LATIN_SMALL_LETTER_Y_WITH_DIAERESIS:
16568 *invlist = add_cp_to_invlist(*invlist,
16569 LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS);
16572 default: /* Other code points are checked against the data for the
16573 current Unicode version */
16575 Size_t folds_count;
16576 unsigned int first_fold;
16577 const unsigned int * remaining_folds;
16581 folded_cp = toFOLD(cp);
16584 U8 dummy_fold[UTF8_MAXBYTES_CASE+1];
16586 folded_cp = _to_fold_latin1(cp, dummy_fold, &dummy_len, 0);
16589 if (folded_cp > 255) {
16590 *invlist = add_cp_to_invlist(*invlist, folded_cp);
16593 folds_count = _inverse_folds(folded_cp, &first_fold,
16595 if (folds_count == 0) {
16597 /* Use deprecated warning to increase the chances of this being
16599 ckWARN2reg_d(RExC_parse,
16600 "Perl folding rules are not up-to-date for 0x%02X;"
16601 " please use the perlbug utility to report;", cp);
16606 if (first_fold > 255) {
16607 *invlist = add_cp_to_invlist(*invlist, first_fold);
16609 for (i = 0; i < folds_count - 1; i++) {
16610 if (remaining_folds[i] > 255) {
16611 *invlist = add_cp_to_invlist(*invlist,
16612 remaining_folds[i]);
16622 S_output_posix_warnings(pTHX_ RExC_state_t *pRExC_state, AV* posix_warnings)
16624 /* Output the elements of the array given by '*posix_warnings' as REGEXP
16628 const bool first_is_fatal = ckDEAD(packWARN(WARN_REGEXP));
16630 PERL_ARGS_ASSERT_OUTPUT_POSIX_WARNINGS;
16632 if (! TO_OUTPUT_WARNINGS(RExC_parse)) {
16636 while ((msg = av_shift(posix_warnings)) != &PL_sv_undef) {
16637 if (first_is_fatal) { /* Avoid leaking this */
16638 av_undef(posix_warnings); /* This isn't necessary if the
16639 array is mortal, but is a
16641 (void) sv_2mortal(msg);
16644 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s", SvPVX(msg));
16645 SvREFCNT_dec_NN(msg);
16648 UPDATE_WARNINGS_LOC(RExC_parse);
16652 S_add_multi_match(pTHX_ AV* multi_char_matches, SV* multi_string, const STRLEN cp_count)
16654 /* This adds the string scalar <multi_string> to the array
16655 * <multi_char_matches>. <multi_string> is known to have exactly
16656 * <cp_count> code points in it. This is used when constructing a
16657 * bracketed character class and we find something that needs to match more
16658 * than a single character.
16660 * <multi_char_matches> is actually an array of arrays. Each top-level
16661 * element is an array that contains all the strings known so far that are
16662 * the same length. And that length (in number of code points) is the same
16663 * as the index of the top-level array. Hence, the [2] element is an
16664 * array, each element thereof is a string containing TWO code points;
16665 * while element [3] is for strings of THREE characters, and so on. Since
16666 * this is for multi-char strings there can never be a [0] nor [1] element.
16668 * When we rewrite the character class below, we will do so such that the
16669 * longest strings are written first, so that it prefers the longest
16670 * matching strings first. This is done even if it turns out that any
16671 * quantifier is non-greedy, out of this programmer's (khw) laziness. Tom
16672 * Christiansen has agreed that this is ok. This makes the test for the
16673 * ligature 'ffi' come before the test for 'ff', for example */
16676 AV** this_array_ptr;
16678 PERL_ARGS_ASSERT_ADD_MULTI_MATCH;
16680 if (! multi_char_matches) {
16681 multi_char_matches = newAV();
16684 if (av_exists(multi_char_matches, cp_count)) {
16685 this_array_ptr = (AV**) av_fetch(multi_char_matches, cp_count, FALSE);
16686 this_array = *this_array_ptr;
16689 this_array = newAV();
16690 av_store(multi_char_matches, cp_count,
16693 av_push(this_array, multi_string);
16695 return multi_char_matches;
16698 /* The names of properties whose definitions are not known at compile time are
16699 * stored in this SV, after a constant heading. So if the length has been
16700 * changed since initialization, then there is a run-time definition. */
16701 #define HAS_NONLOCALE_RUNTIME_PROPERTY_DEFINITION \
16702 (SvCUR(listsv) != initial_listsv_len)
16704 /* There is a restricted set of white space characters that are legal when
16705 * ignoring white space in a bracketed character class. This generates the
16706 * code to skip them.
16708 * There is a line below that uses the same white space criteria but is outside
16709 * this macro. Both here and there must use the same definition */
16710 #define SKIP_BRACKETED_WHITE_SPACE(do_skip, p) \
16713 while (isBLANK_A(UCHARAT(p))) \
16720 STATIC regnode_offset
16721 S_regclass(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, U32 depth,
16722 const bool stop_at_1, /* Just parse the next thing, don't
16723 look for a full character class */
16724 bool allow_mutiple_chars,
16725 const bool silence_non_portable, /* Don't output warnings
16729 bool optimizable, /* ? Allow a non-ANYOF return
16731 SV** ret_invlist /* Return an inversion list, not a node */
16734 /* parse a bracketed class specification. Most of these will produce an
16735 * ANYOF node; but something like [a] will produce an EXACT node; [aA], an
16736 * EXACTFish node; [[:ascii:]], a POSIXA node; etc. It is more complex
16737 * under /i with multi-character folds: it will be rewritten following the
16738 * paradigm of this example, where the <multi-fold>s are characters which
16739 * fold to multiple character sequences:
16740 * /[abc\x{multi-fold1}def\x{multi-fold2}ghi]/i
16741 * gets effectively rewritten as:
16742 * /(?:\x{multi-fold1}|\x{multi-fold2}|[abcdefghi]/i
16743 * reg() gets called (recursively) on the rewritten version, and this
16744 * function will return what it constructs. (Actually the <multi-fold>s
16745 * aren't physically removed from the [abcdefghi], it's just that they are
16746 * ignored in the recursion by means of a flag:
16747 * <RExC_in_multi_char_class>.)
16749 * ANYOF nodes contain a bit map for the first NUM_ANYOF_CODE_POINTS
16750 * characters, with the corresponding bit set if that character is in the
16751 * list. For characters above this, an inversion list is used. There
16752 * are extra bits for \w, etc. in locale ANYOFs, as what these match is not
16753 * determinable at compile time
16755 * On success, returns the offset at which any next node should be placed
16756 * into the regex engine program being compiled.
16758 * Returns 0 otherwise, setting flagp to RESTART_PARSE if the parse needs
16759 * to be restarted, or'd with NEED_UTF8 if the pattern needs to be upgraded to
16764 UV prevvalue = OOB_UNICODE, save_prevvalue = OOB_UNICODE;
16766 UV value = OOB_UNICODE, save_value = OOB_UNICODE;
16767 regnode_offset ret = -1; /* Initialized to an illegal value */
16769 int namedclass = OOB_NAMEDCLASS;
16770 char *rangebegin = NULL;
16771 SV *listsv = NULL; /* List of \p{user-defined} whose definitions
16772 aren't available at the time this was called */
16773 STRLEN initial_listsv_len = 0; /* Kind of a kludge to see if it is more
16774 than just initialized. */
16775 SV* properties = NULL; /* Code points that match \p{} \P{} */
16776 SV* posixes = NULL; /* Code points that match classes like [:word:],
16777 extended beyond the Latin1 range. These have to
16778 be kept separate from other code points for much
16779 of this function because their handling is
16780 different under /i, and for most classes under
16782 SV* nposixes = NULL; /* Similarly for [:^word:]. These are kept
16783 separate for a while from the non-complemented
16784 versions because of complications with /d
16786 SV* simple_posixes = NULL; /* But under some conditions, the classes can be
16787 treated more simply than the general case,
16788 leading to less compilation and execution
16790 UV element_count = 0; /* Number of distinct elements in the class.
16791 Optimizations may be possible if this is tiny */
16792 AV * multi_char_matches = NULL; /* Code points that fold to more than one
16793 character; used under /i */
16795 char * stop_ptr = RExC_end; /* where to stop parsing */
16797 /* ignore unescaped whitespace? */
16798 const bool skip_white = cBOOL( ret_invlist
16799 || (RExC_flags & RXf_PMf_EXTENDED_MORE));
16801 /* inversion list of code points this node matches only when the target
16802 * string is in UTF-8. These are all non-ASCII, < 256. (Because is under
16804 SV* upper_latin1_only_utf8_matches = NULL;
16806 /* Inversion list of code points this node matches regardless of things
16807 * like locale, folding, utf8ness of the target string */
16808 SV* cp_list = NULL;
16810 /* Like cp_list, but code points on this list need to be checked for things
16811 * that fold to/from them under /i */
16812 SV* cp_foldable_list = NULL;
16814 /* Like cp_list, but code points on this list are valid only when the
16815 * runtime locale is UTF-8 */
16816 SV* only_utf8_locale_list = NULL;
16818 /* In a range, if one of the endpoints is non-character-set portable,
16819 * meaning that it hard-codes a code point that may mean a different
16820 * charactger in ASCII vs. EBCDIC, as opposed to, say, a literal 'A' or a
16821 * mnemonic '\t' which each mean the same character no matter which
16822 * character set the platform is on. */
16823 unsigned int non_portable_endpoint = 0;
16825 /* Is the range unicode? which means on a platform that isn't 1-1 native
16826 * to Unicode (i.e. non-ASCII), each code point in it should be considered
16827 * to be a Unicode value. */
16828 bool unicode_range = FALSE;
16829 bool invert = FALSE; /* Is this class to be complemented */
16831 bool warn_super = ALWAYS_WARN_SUPER;
16833 const char * orig_parse = RExC_parse;
16835 /* This variable is used to mark where the end in the input is of something
16836 * that looks like a POSIX construct but isn't. During the parse, when
16837 * something looks like it could be such a construct is encountered, it is
16838 * checked for being one, but not if we've already checked this area of the
16839 * input. Only after this position is reached do we check again */
16840 char *not_posix_region_end = RExC_parse - 1;
16842 AV* posix_warnings = NULL;
16843 const bool do_posix_warnings = ckWARN(WARN_REGEXP);
16844 U8 op = END; /* The returned node-type, initialized to an impossible
16846 U8 anyof_flags = 0; /* flag bits if the node is an ANYOF-type */
16847 U32 posixl = 0; /* bit field of posix classes matched under /l */
16850 /* Flags as to what things aren't knowable until runtime. (Note that these are
16851 * mutually exclusive.) */
16852 #define HAS_USER_DEFINED_PROPERTY 0x01 /* /u any user-defined properties that
16853 haven't been defined as of yet */
16854 #define HAS_D_RUNTIME_DEPENDENCY 0x02 /* /d if the target being matched is
16856 #define HAS_L_RUNTIME_DEPENDENCY 0x04 /* /l what the posix classes match and
16857 what gets folded */
16858 U32 has_runtime_dependency = 0; /* OR of the above flags */
16860 GET_RE_DEBUG_FLAGS_DECL;
16862 PERL_ARGS_ASSERT_REGCLASS;
16864 PERL_UNUSED_ARG(depth);
16868 /* If wants an inversion list returned, we can't optimize to something
16871 optimizable = FALSE;
16874 DEBUG_PARSE("clas");
16876 #if UNICODE_MAJOR_VERSION < 3 /* no multifolds in early Unicode */ \
16877 || (UNICODE_MAJOR_VERSION == 3 && UNICODE_DOT_VERSION == 0 \
16878 && UNICODE_DOT_DOT_VERSION == 0)
16879 allow_mutiple_chars = FALSE;
16882 /* We include the /i status at the beginning of this so that we can
16883 * know it at runtime */
16884 listsv = sv_2mortal(Perl_newSVpvf(aTHX_ "#%d\n", cBOOL(FOLD)));
16885 initial_listsv_len = SvCUR(listsv);
16886 SvTEMP_off(listsv); /* Grr, TEMPs and mortals are conflated. */
16888 SKIP_BRACKETED_WHITE_SPACE(skip_white, RExC_parse);
16890 assert(RExC_parse <= RExC_end);
16892 if (UCHARAT(RExC_parse) == '^') { /* Complement the class */
16895 allow_mutiple_chars = FALSE;
16897 SKIP_BRACKETED_WHITE_SPACE(skip_white, RExC_parse);
16900 /* Check that they didn't say [:posix:] instead of [[:posix:]] */
16901 if (! ret_invlist && MAYBE_POSIXCC(UCHARAT(RExC_parse))) {
16902 int maybe_class = handle_possible_posix(pRExC_state,
16904 ¬_posix_region_end,
16906 TRUE /* checking only */);
16907 if (maybe_class >= OOB_NAMEDCLASS && do_posix_warnings) {
16908 ckWARN4reg(not_posix_region_end,
16909 "POSIX syntax [%c %c] belongs inside character classes%s",
16910 *RExC_parse, *RExC_parse,
16911 (maybe_class == OOB_NAMEDCLASS)
16912 ? ((POSIXCC_NOTYET(*RExC_parse))
16913 ? " (but this one isn't implemented)"
16914 : " (but this one isn't fully valid)")
16920 /* If the caller wants us to just parse a single element, accomplish this
16921 * by faking the loop ending condition */
16922 if (stop_at_1 && RExC_end > RExC_parse) {
16923 stop_ptr = RExC_parse + 1;
16926 /* allow 1st char to be ']' (allowing it to be '-' is dealt with later) */
16927 if (UCHARAT(RExC_parse) == ']')
16928 goto charclassloop;
16932 if ( posix_warnings
16933 && av_tindex_skip_len_mg(posix_warnings) >= 0
16934 && RExC_parse > not_posix_region_end)
16936 /* Warnings about posix class issues are considered tentative until
16937 * we are far enough along in the parse that we can no longer
16938 * change our mind, at which point we output them. This is done
16939 * each time through the loop so that a later class won't zap them
16940 * before they have been dealt with. */
16941 output_posix_warnings(pRExC_state, posix_warnings);
16944 if (RExC_parse >= stop_ptr) {
16948 SKIP_BRACKETED_WHITE_SPACE(skip_white, RExC_parse);
16950 if (UCHARAT(RExC_parse) == ']') {
16956 namedclass = OOB_NAMEDCLASS; /* initialize as illegal */
16957 save_value = value;
16958 save_prevvalue = prevvalue;
16961 rangebegin = RExC_parse;
16963 non_portable_endpoint = 0;
16965 if (UTF && ! UTF8_IS_INVARIANT(* RExC_parse)) {
16966 value = utf8n_to_uvchr((U8*)RExC_parse,
16967 RExC_end - RExC_parse,
16968 &numlen, UTF8_ALLOW_DEFAULT);
16969 RExC_parse += numlen;
16972 value = UCHARAT(RExC_parse++);
16974 if (value == '[') {
16975 char * posix_class_end;
16976 namedclass = handle_possible_posix(pRExC_state,
16979 do_posix_warnings ? &posix_warnings : NULL,
16980 FALSE /* die if error */);
16981 if (namedclass > OOB_NAMEDCLASS) {
16983 /* If there was an earlier attempt to parse this particular
16984 * posix class, and it failed, it was a false alarm, as this
16985 * successful one proves */
16986 if ( posix_warnings
16987 && av_tindex_skip_len_mg(posix_warnings) >= 0
16988 && not_posix_region_end >= RExC_parse
16989 && not_posix_region_end <= posix_class_end)
16991 av_undef(posix_warnings);
16994 RExC_parse = posix_class_end;
16996 else if (namedclass == OOB_NAMEDCLASS) {
16997 not_posix_region_end = posix_class_end;
17000 namedclass = OOB_NAMEDCLASS;
17003 else if ( RExC_parse - 1 > not_posix_region_end
17004 && MAYBE_POSIXCC(value))
17006 (void) handle_possible_posix(
17008 RExC_parse - 1, /* -1 because parse has already been
17010 ¬_posix_region_end,
17011 do_posix_warnings ? &posix_warnings : NULL,
17012 TRUE /* checking only */);
17014 else if ( strict && ! skip_white
17015 && ( _generic_isCC(value, _CC_VERTSPACE)
17016 || is_VERTWS_cp_high(value)))
17018 vFAIL("Literal vertical space in [] is illegal except under /x");
17020 else if (value == '\\') {
17021 /* Is a backslash; get the code point of the char after it */
17023 if (RExC_parse >= RExC_end) {
17024 vFAIL("Unmatched [");
17027 if (UTF && ! UTF8_IS_INVARIANT(UCHARAT(RExC_parse))) {
17028 value = utf8n_to_uvchr((U8*)RExC_parse,
17029 RExC_end - RExC_parse,
17030 &numlen, UTF8_ALLOW_DEFAULT);
17031 RExC_parse += numlen;
17034 value = UCHARAT(RExC_parse++);
17036 /* Some compilers cannot handle switching on 64-bit integer
17037 * values, therefore value cannot be an UV. Yes, this will
17038 * be a problem later if we want switch on Unicode.
17039 * A similar issue a little bit later when switching on
17040 * namedclass. --jhi */
17042 /* If the \ is escaping white space when white space is being
17043 * skipped, it means that that white space is wanted literally, and
17044 * is already in 'value'. Otherwise, need to translate the escape
17045 * into what it signifies. */
17046 if (! skip_white || ! isBLANK_A(value)) switch ((I32)value) {
17048 case 'w': namedclass = ANYOF_WORDCHAR; break;
17049 case 'W': namedclass = ANYOF_NWORDCHAR; break;
17050 case 's': namedclass = ANYOF_SPACE; break;
17051 case 'S': namedclass = ANYOF_NSPACE; break;
17052 case 'd': namedclass = ANYOF_DIGIT; break;
17053 case 'D': namedclass = ANYOF_NDIGIT; break;
17054 case 'v': namedclass = ANYOF_VERTWS; break;
17055 case 'V': namedclass = ANYOF_NVERTWS; break;
17056 case 'h': namedclass = ANYOF_HORIZWS; break;
17057 case 'H': namedclass = ANYOF_NHORIZWS; break;
17058 case 'N': /* Handle \N{NAME} in class */
17060 const char * const backslash_N_beg = RExC_parse - 2;
17063 if (! grok_bslash_N(pRExC_state,
17064 NULL, /* No regnode */
17065 &value, /* Yes single value */
17066 &cp_count, /* Multiple code pt count */
17072 if (*flagp & NEED_UTF8)
17073 FAIL("panic: grok_bslash_N set NEED_UTF8");
17075 RETURN_FAIL_ON_RESTART_FLAGP(flagp);
17077 if (cp_count < 0) {
17078 vFAIL("\\N in a character class must be a named character: \\N{...}");
17080 else if (cp_count == 0) {
17081 ckWARNreg(RExC_parse,
17082 "Ignoring zero length \\N{} in character class");
17084 else { /* cp_count > 1 */
17085 assert(cp_count > 1);
17086 if (! RExC_in_multi_char_class) {
17087 if ( ! allow_mutiple_chars
17090 || *RExC_parse == '-')
17094 vFAIL("\\N{} here is restricted to one character");
17096 ckWARNreg(RExC_parse, "Using just the first character returned by \\N{} in character class");
17097 break; /* <value> contains the first code
17098 point. Drop out of the switch to
17102 SV * multi_char_N = newSVpvn(backslash_N_beg,
17103 RExC_parse - backslash_N_beg);
17105 = add_multi_match(multi_char_matches,
17110 } /* End of cp_count != 1 */
17112 /* This element should not be processed further in this
17115 value = save_value;
17116 prevvalue = save_prevvalue;
17117 continue; /* Back to top of loop to get next char */
17120 /* Here, is a single code point, and <value> contains it */
17121 unicode_range = TRUE; /* \N{} are Unicode */
17129 /* \p means they want Unicode semantics */
17130 REQUIRE_UNI_RULES(flagp, 0);
17132 if (RExC_parse >= RExC_end)
17133 vFAIL2("Empty \\%c", (U8)value);
17134 if (*RExC_parse == '{') {
17135 const U8 c = (U8)value;
17136 e = (char *) memchr(RExC_parse, '}', RExC_end - RExC_parse);
17139 vFAIL2("Missing right brace on \\%c{}", c);
17144 /* White space is allowed adjacent to the braces and after
17145 * any '^', even when not under /x */
17146 while (isSPACE(*RExC_parse)) {
17150 if (UCHARAT(RExC_parse) == '^') {
17152 /* toggle. (The rhs xor gets the single bit that
17153 * differs between P and p; the other xor inverts just
17155 value ^= 'P' ^ 'p';
17158 while (isSPACE(*RExC_parse)) {
17163 if (e == RExC_parse)
17164 vFAIL2("Empty \\%c{}", c);
17166 n = e - RExC_parse;
17167 while (isSPACE(*(RExC_parse + n - 1)))
17170 } /* The \p isn't immediately followed by a '{' */
17171 else if (! isALPHA(*RExC_parse)) {
17172 RExC_parse += (UTF)
17173 ? UTF8_SAFE_SKIP(RExC_parse, RExC_end)
17175 vFAIL2("Character following \\%c must be '{' or a "
17176 "single-character Unicode property name",
17184 char* name = RExC_parse;
17186 /* Any message returned about expanding the definition */
17187 SV* msg = newSVpvs_flags("", SVs_TEMP);
17189 /* If set TRUE, the property is user-defined as opposed to
17190 * official Unicode */
17191 bool user_defined = FALSE;
17193 SV * prop_definition = parse_uniprop_string(
17194 name, n, UTF, FOLD,
17195 FALSE, /* This is compile-time */
17197 /* We can't defer this defn when
17198 * the full result is required in
17200 ! cBOOL(ret_invlist),
17206 if (SvCUR(msg)) { /* Assumes any error causes a msg */
17207 assert(prop_definition == NULL);
17208 RExC_parse = e + 1;
17209 if (SvUTF8(msg)) { /* msg being UTF-8 makes the whole
17210 thing so, or else the display is
17214 /* diag_listed_as: Can't find Unicode property definition "%s" in regex; marked by <-- HERE in m/%s/ */
17215 vFAIL2utf8f("%" UTF8f, UTF8fARG(SvUTF8(msg),
17216 SvCUR(msg), SvPVX(msg)));
17219 if (! is_invlist(prop_definition)) {
17221 /* Here, the definition isn't known, so we have gotten
17222 * returned a string that will be evaluated if and when
17223 * encountered at runtime. We add it to the list of
17224 * such properties, along with whether it should be
17225 * complemented or not */
17226 if (value == 'P') {
17227 sv_catpvs(listsv, "!");
17230 sv_catpvs(listsv, "+");
17232 sv_catsv(listsv, prop_definition);
17234 has_runtime_dependency |= HAS_USER_DEFINED_PROPERTY;
17236 /* We don't know yet what this matches, so have to flag
17238 anyof_flags |= ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP;
17241 assert (prop_definition && is_invlist(prop_definition));
17243 /* Here we do have the complete property definition
17245 * Temporary workaround for [perl #133136]. For this
17246 * precise input that is in the .t that is failing,
17247 * load utf8.pm, which is what the test wants, so that
17248 * that .t passes */
17249 if ( memEQs(RExC_start, e + 1 - RExC_start,
17251 && ! hv_common(GvHVn(PL_incgv),
17253 "utf8.pm", sizeof("utf8.pm") - 1,
17254 0, HV_FETCH_ISEXISTS, NULL, 0))
17256 require_pv("utf8.pm");
17259 if (! user_defined &&
17260 /* We warn on matching an above-Unicode code point
17261 * if the match would return true, except don't
17262 * warn for \p{All}, which has exactly one element
17264 (_invlist_contains_cp(prop_definition, 0x110000)
17265 && (! (_invlist_len(prop_definition) == 1
17266 && *invlist_array(prop_definition) == 0))))
17271 /* Invert if asking for the complement */
17272 if (value == 'P') {
17273 _invlist_union_complement_2nd(properties,
17278 _invlist_union(properties, prop_definition, &properties);
17283 RExC_parse = e + 1;
17284 namedclass = ANYOF_UNIPROP; /* no official name, but it's
17288 case 'n': value = '\n'; break;
17289 case 'r': value = '\r'; break;
17290 case 't': value = '\t'; break;
17291 case 'f': value = '\f'; break;
17292 case 'b': value = '\b'; break;
17293 case 'e': value = ESC_NATIVE; break;
17294 case 'a': value = '\a'; break;
17296 RExC_parse--; /* function expects to be pointed at the 'o' */
17298 const char* error_msg;
17299 bool valid = grok_bslash_o(&RExC_parse,
17303 TO_OUTPUT_WARNINGS(RExC_parse),
17305 silence_non_portable,
17310 UPDATE_WARNINGS_LOC(RExC_parse - 1);
17312 non_portable_endpoint++;
17315 RExC_parse--; /* function expects to be pointed at the 'x' */
17317 const char* error_msg;
17318 bool valid = grok_bslash_x(&RExC_parse,
17322 TO_OUTPUT_WARNINGS(RExC_parse),
17324 silence_non_portable,
17329 UPDATE_WARNINGS_LOC(RExC_parse - 1);
17331 non_portable_endpoint++;
17334 value = grok_bslash_c(*RExC_parse, TO_OUTPUT_WARNINGS(RExC_parse));
17335 UPDATE_WARNINGS_LOC(RExC_parse);
17337 non_portable_endpoint++;
17339 case '0': case '1': case '2': case '3': case '4':
17340 case '5': case '6': case '7':
17342 /* Take 1-3 octal digits */
17343 I32 flags = PERL_SCAN_SILENT_ILLDIGIT;
17344 numlen = (strict) ? 4 : 3;
17345 value = grok_oct(--RExC_parse, &numlen, &flags, NULL);
17346 RExC_parse += numlen;
17349 RExC_parse += (UTF)
17350 ? UTF8_SAFE_SKIP(RExC_parse, RExC_end)
17352 vFAIL("Need exactly 3 octal digits");
17354 else if ( numlen < 3 /* like \08, \178 */
17355 && RExC_parse < RExC_end
17356 && isDIGIT(*RExC_parse)
17357 && ckWARN(WARN_REGEXP))
17359 reg_warn_non_literal_string(
17361 form_short_octal_warning(RExC_parse, numlen));
17364 non_portable_endpoint++;
17368 /* Allow \_ to not give an error */
17369 if (isWORDCHAR(value) && value != '_') {
17371 vFAIL2("Unrecognized escape \\%c in character class",
17375 ckWARN2reg(RExC_parse,
17376 "Unrecognized escape \\%c in character class passed through",
17381 } /* End of switch on char following backslash */
17382 } /* end of handling backslash escape sequences */
17384 /* Here, we have the current token in 'value' */
17386 if (namedclass > OOB_NAMEDCLASS) { /* this is a named class \blah */
17389 /* a bad range like a-\d, a-[:digit:]. The '-' is taken as a
17390 * literal, as is the character that began the false range, i.e.
17391 * the 'a' in the examples */
17393 const int w = (RExC_parse >= rangebegin)
17394 ? RExC_parse - rangebegin
17398 "False [] range \"%" UTF8f "\"",
17399 UTF8fARG(UTF, w, rangebegin));
17402 ckWARN2reg(RExC_parse,
17403 "False [] range \"%" UTF8f "\"",
17404 UTF8fARG(UTF, w, rangebegin));
17405 cp_list = add_cp_to_invlist(cp_list, '-');
17406 cp_foldable_list = add_cp_to_invlist(cp_foldable_list,
17410 range = 0; /* this was not a true range */
17411 element_count += 2; /* So counts for three values */
17414 classnum = namedclass_to_classnum(namedclass);
17416 if (LOC && namedclass < ANYOF_POSIXL_MAX
17417 #ifndef HAS_ISASCII
17418 && classnum != _CC_ASCII
17421 SV* scratch_list = NULL;
17423 /* What the Posix classes (like \w, [:space:]) match isn't
17424 * generally knowable under locale until actual match time. A
17425 * special node is used for these which has extra space for a
17426 * bitmap, with a bit reserved for each named class that is to
17427 * be matched against. (This isn't needed for \p{} and
17428 * pseudo-classes, as they are not affected by locale, and
17429 * hence are dealt with separately.) However, if a named class
17430 * and its complement are both present, then it matches
17431 * everything, and there is no runtime dependency. Odd numbers
17432 * are the complements of the next lower number, so xor works.
17433 * (Note that something like [\w\D] should match everything,
17434 * because \d should be a proper subset of \w. But rather than
17435 * trust that the locale is well behaved, we leave this to
17436 * runtime to sort out) */
17437 if (POSIXL_TEST(posixl, namedclass ^ 1)) {
17438 cp_list = _add_range_to_invlist(cp_list, 0, UV_MAX);
17439 POSIXL_ZERO(posixl);
17440 has_runtime_dependency &= ~HAS_L_RUNTIME_DEPENDENCY;
17441 anyof_flags &= ~ANYOF_MATCHES_POSIXL;
17442 continue; /* We could ignore the rest of the class, but
17443 best to parse it for any errors */
17445 else { /* Here, isn't the complement of any already parsed
17447 POSIXL_SET(posixl, namedclass);
17448 has_runtime_dependency |= HAS_L_RUNTIME_DEPENDENCY;
17449 anyof_flags |= ANYOF_MATCHES_POSIXL;
17451 /* The above-Latin1 characters are not subject to locale
17452 * rules. Just add them to the unconditionally-matched
17455 /* Get the list of the above-Latin1 code points this
17457 _invlist_intersection_maybe_complement_2nd(PL_AboveLatin1,
17458 PL_XPosix_ptrs[classnum],
17460 /* Odd numbers are complements,
17461 * like NDIGIT, NASCII, ... */
17462 namedclass % 2 != 0,
17464 /* Checking if 'cp_list' is NULL first saves an extra
17465 * clone. Its reference count will be decremented at the
17466 * next union, etc, or if this is the only instance, at the
17467 * end of the routine */
17469 cp_list = scratch_list;
17472 _invlist_union(cp_list, scratch_list, &cp_list);
17473 SvREFCNT_dec_NN(scratch_list);
17475 continue; /* Go get next character */
17480 /* Here, is not /l, or is a POSIX class for which /l doesn't
17481 * matter (or is a Unicode property, which is skipped here). */
17482 if (namedclass >= ANYOF_POSIXL_MAX) { /* If a special class */
17483 if (namedclass != ANYOF_UNIPROP) { /* UNIPROP = \p and \P */
17485 /* Here, should be \h, \H, \v, or \V. None of /d, /i
17486 * nor /l make a difference in what these match,
17487 * therefore we just add what they match to cp_list. */
17488 if (classnum != _CC_VERTSPACE) {
17489 assert( namedclass == ANYOF_HORIZWS
17490 || namedclass == ANYOF_NHORIZWS);
17492 /* It turns out that \h is just a synonym for
17494 classnum = _CC_BLANK;
17497 _invlist_union_maybe_complement_2nd(
17499 PL_XPosix_ptrs[classnum],
17500 namedclass % 2 != 0, /* Complement if odd
17501 (NHORIZWS, NVERTWS)
17506 else if ( AT_LEAST_UNI_SEMANTICS
17507 || classnum == _CC_ASCII
17508 || (DEPENDS_SEMANTICS && ( classnum == _CC_DIGIT
17509 || classnum == _CC_XDIGIT)))
17511 /* We usually have to worry about /d affecting what POSIX
17512 * classes match, with special code needed because we won't
17513 * know until runtime what all matches. But there is no
17514 * extra work needed under /u and /a; and [:ascii:] is
17515 * unaffected by /d; and :digit: and :xdigit: don't have
17516 * runtime differences under /d. So we can special case
17517 * these, and avoid some extra work below, and at runtime.
17519 _invlist_union_maybe_complement_2nd(
17521 ((AT_LEAST_ASCII_RESTRICTED)
17522 ? PL_Posix_ptrs[classnum]
17523 : PL_XPosix_ptrs[classnum]),
17524 namedclass % 2 != 0,
17527 else { /* Garden variety class. If is NUPPER, NALPHA, ...
17528 complement and use nposixes */
17529 SV** posixes_ptr = namedclass % 2 == 0
17532 _invlist_union_maybe_complement_2nd(
17534 PL_XPosix_ptrs[classnum],
17535 namedclass % 2 != 0,
17539 } /* end of namedclass \blah */
17541 SKIP_BRACKETED_WHITE_SPACE(skip_white, RExC_parse);
17543 /* If 'range' is set, 'value' is the ending of a range--check its
17544 * validity. (If value isn't a single code point in the case of a
17545 * range, we should have figured that out above in the code that
17546 * catches false ranges). Later, we will handle each individual code
17547 * point in the range. If 'range' isn't set, this could be the
17548 * beginning of a range, so check for that by looking ahead to see if
17549 * the next real character to be processed is the range indicator--the
17554 /* For unicode ranges, we have to test that the Unicode as opposed
17555 * to the native values are not decreasing. (Above 255, there is
17556 * no difference between native and Unicode) */
17557 if (unicode_range && prevvalue < 255 && value < 255) {
17558 if (NATIVE_TO_LATIN1(prevvalue) > NATIVE_TO_LATIN1(value)) {
17559 goto backwards_range;
17564 if (prevvalue > value) /* b-a */ {
17569 w = RExC_parse - rangebegin;
17571 "Invalid [] range \"%" UTF8f "\"",
17572 UTF8fARG(UTF, w, rangebegin));
17573 NOT_REACHED; /* NOTREACHED */
17577 prevvalue = value; /* save the beginning of the potential range */
17578 if (! stop_at_1 /* Can't be a range if parsing just one thing */
17579 && *RExC_parse == '-')
17581 char* next_char_ptr = RExC_parse + 1;
17583 /* Get the next real char after the '-' */
17584 SKIP_BRACKETED_WHITE_SPACE(skip_white, next_char_ptr);
17586 /* If the '-' is at the end of the class (just before the ']',
17587 * it is a literal minus; otherwise it is a range */
17588 if (next_char_ptr < RExC_end && *next_char_ptr != ']') {
17589 RExC_parse = next_char_ptr;
17591 /* a bad range like \w-, [:word:]- ? */
17592 if (namedclass > OOB_NAMEDCLASS) {
17593 if (strict || ckWARN(WARN_REGEXP)) {
17594 const int w = RExC_parse >= rangebegin
17595 ? RExC_parse - rangebegin
17598 vFAIL4("False [] range \"%*.*s\"",
17603 "False [] range \"%*.*s\"",
17607 cp_list = add_cp_to_invlist(cp_list, '-');
17610 range = 1; /* yeah, it's a range! */
17611 continue; /* but do it the next time */
17616 if (namedclass > OOB_NAMEDCLASS) {
17620 /* Here, we have a single value this time through the loop, and
17621 * <prevvalue> is the beginning of the range, if any; or <value> if
17624 /* non-Latin1 code point implies unicode semantics. */
17626 REQUIRE_UNI_RULES(flagp, 0);
17629 /* Ready to process either the single value, or the completed range.
17630 * For single-valued non-inverted ranges, we consider the possibility
17631 * of multi-char folds. (We made a conscious decision to not do this
17632 * for the other cases because it can often lead to non-intuitive
17633 * results. For example, you have the peculiar case that:
17634 * "s s" =~ /^[^\xDF]+$/i => Y
17635 * "ss" =~ /^[^\xDF]+$/i => N
17637 * See [perl #89750] */
17638 if (FOLD && allow_mutiple_chars && value == prevvalue) {
17639 if ( value == LATIN_SMALL_LETTER_SHARP_S
17640 || (value > 255 && _invlist_contains_cp(PL_HasMultiCharFold,
17643 /* Here <value> is indeed a multi-char fold. Get what it is */
17645 U8 foldbuf[UTF8_MAXBYTES_CASE+1];
17648 UV folded = _to_uni_fold_flags(
17652 FOLD_FLAGS_FULL | (ASCII_FOLD_RESTRICTED
17653 ? FOLD_FLAGS_NOMIX_ASCII
17657 /* Here, <folded> should be the first character of the
17658 * multi-char fold of <value>, with <foldbuf> containing the
17659 * whole thing. But, if this fold is not allowed (because of
17660 * the flags), <fold> will be the same as <value>, and should
17661 * be processed like any other character, so skip the special
17663 if (folded != value) {
17665 /* Skip if we are recursed, currently parsing the class
17666 * again. Otherwise add this character to the list of
17667 * multi-char folds. */
17668 if (! RExC_in_multi_char_class) {
17669 STRLEN cp_count = utf8_length(foldbuf,
17670 foldbuf + foldlen);
17671 SV* multi_fold = sv_2mortal(newSVpvs(""));
17673 Perl_sv_catpvf(aTHX_ multi_fold, "\\x{%" UVXf "}", value);
17676 = add_multi_match(multi_char_matches,
17682 /* This element should not be processed further in this
17685 value = save_value;
17686 prevvalue = save_prevvalue;
17692 if (strict && ckWARN(WARN_REGEXP)) {
17695 /* If the range starts above 255, everything is portable and
17696 * likely to be so for any forseeable character set, so don't
17698 if (unicode_range && non_portable_endpoint && prevvalue < 256) {
17699 vWARN(RExC_parse, "Both or neither range ends should be Unicode");
17701 else if (prevvalue != value) {
17703 /* Under strict, ranges that stop and/or end in an ASCII
17704 * printable should have each end point be a portable value
17705 * for it (preferably like 'A', but we don't warn if it is
17706 * a (portable) Unicode name or code point), and the range
17707 * must be be all digits or all letters of the same case.
17708 * Otherwise, the range is non-portable and unclear as to
17709 * what it contains */
17710 if ( (isPRINT_A(prevvalue) || isPRINT_A(value))
17711 && ( non_portable_endpoint
17712 || ! ( (isDIGIT_A(prevvalue) && isDIGIT_A(value))
17713 || (isLOWER_A(prevvalue) && isLOWER_A(value))
17714 || (isUPPER_A(prevvalue) && isUPPER_A(value))
17716 vWARN(RExC_parse, "Ranges of ASCII printables should"
17717 " be some subset of \"0-9\","
17718 " \"A-Z\", or \"a-z\"");
17720 else if (prevvalue >= FIRST_NON_ASCII_DECIMAL_DIGIT) {
17721 SSize_t index_start;
17722 SSize_t index_final;
17724 /* But the nature of Unicode and languages mean we
17725 * can't do the same checks for above-ASCII ranges,
17726 * except in the case of digit ones. These should
17727 * contain only digits from the same group of 10. The
17728 * ASCII case is handled just above. Hence here, the
17729 * range could be a range of digits. First some
17730 * unlikely special cases. Grandfather in that a range
17731 * ending in 19DA (NEW TAI LUE THAM DIGIT ONE) is bad
17732 * if its starting value is one of the 10 digits prior
17733 * to it. This is because it is an alternate way of
17734 * writing 19D1, and some people may expect it to be in
17735 * that group. But it is bad, because it won't give
17736 * the expected results. In Unicode 5.2 it was
17737 * considered to be in that group (of 11, hence), but
17738 * this was fixed in the next version */
17740 if (UNLIKELY(value == 0x19DA && prevvalue >= 0x19D0)) {
17741 goto warn_bad_digit_range;
17743 else if (UNLIKELY( prevvalue >= 0x1D7CE
17744 && value <= 0x1D7FF))
17746 /* This is the only other case currently in Unicode
17747 * where the algorithm below fails. The code
17748 * points just above are the end points of a single
17749 * range containing only decimal digits. It is 5
17750 * different series of 0-9. All other ranges of
17751 * digits currently in Unicode are just a single
17752 * series. (And mktables will notify us if a later
17753 * Unicode version breaks this.)
17755 * If the range being checked is at most 9 long,
17756 * and the digit values represented are in
17757 * numerical order, they are from the same series.
17759 if ( value - prevvalue > 9
17760 || ((( value - 0x1D7CE) % 10)
17761 <= (prevvalue - 0x1D7CE) % 10))
17763 goto warn_bad_digit_range;
17768 /* For all other ranges of digits in Unicode, the
17769 * algorithm is just to check if both end points
17770 * are in the same series, which is the same range.
17772 index_start = _invlist_search(
17773 PL_XPosix_ptrs[_CC_DIGIT],
17776 /* Warn if the range starts and ends with a digit,
17777 * and they are not in the same group of 10. */
17778 if ( index_start >= 0
17779 && ELEMENT_RANGE_MATCHES_INVLIST(index_start)
17781 _invlist_search(PL_XPosix_ptrs[_CC_DIGIT],
17782 value)) != index_start
17783 && index_final >= 0
17784 && ELEMENT_RANGE_MATCHES_INVLIST(index_final))
17786 warn_bad_digit_range:
17787 vWARN(RExC_parse, "Ranges of digits should be"
17788 " from the same group of"
17795 if ((! range || prevvalue == value) && non_portable_endpoint) {
17796 if (isPRINT_A(value)) {
17799 if (isBACKSLASHED_PUNCT(value)) {
17800 literal[d++] = '\\';
17802 literal[d++] = (char) value;
17803 literal[d++] = '\0';
17806 "\"%.*s\" is more clearly written simply as \"%s\"",
17807 (int) (RExC_parse - rangebegin),
17812 else if (isMNEMONIC_CNTRL(value)) {
17814 "\"%.*s\" is more clearly written simply as \"%s\"",
17815 (int) (RExC_parse - rangebegin),
17817 cntrl_to_mnemonic((U8) value)
17823 /* Deal with this element of the class */
17826 cp_foldable_list = _add_range_to_invlist(cp_foldable_list,
17829 /* On non-ASCII platforms, for ranges that span all of 0..255, and ones
17830 * that don't require special handling, we can just add the range like
17831 * we do for ASCII platforms */
17832 if ((UNLIKELY(prevvalue == 0) && value >= 255)
17833 || ! (prevvalue < 256
17835 || (! non_portable_endpoint
17836 && ((isLOWER_A(prevvalue) && isLOWER_A(value))
17837 || (isUPPER_A(prevvalue)
17838 && isUPPER_A(value)))))))
17840 cp_foldable_list = _add_range_to_invlist(cp_foldable_list,
17844 /* Here, requires special handling. This can be because it is a
17845 * range whose code points are considered to be Unicode, and so
17846 * must be individually translated into native, or because its a
17847 * subrange of 'A-Z' or 'a-z' which each aren't contiguous in
17848 * EBCDIC, but we have defined them to include only the "expected"
17849 * upper or lower case ASCII alphabetics. Subranges above 255 are
17850 * the same in native and Unicode, so can be added as a range */
17851 U8 start = NATIVE_TO_LATIN1(prevvalue);
17853 U8 end = (value < 256) ? NATIVE_TO_LATIN1(value) : 255;
17854 for (j = start; j <= end; j++) {
17855 cp_foldable_list = add_cp_to_invlist(cp_foldable_list, LATIN1_TO_NATIVE(j));
17858 cp_foldable_list = _add_range_to_invlist(cp_foldable_list,
17864 range = 0; /* this range (if it was one) is done now */
17865 } /* End of loop through all the text within the brackets */
17867 if ( posix_warnings && av_tindex_skip_len_mg(posix_warnings) >= 0) {
17868 output_posix_warnings(pRExC_state, posix_warnings);
17871 /* If anything in the class expands to more than one character, we have to
17872 * deal with them by building up a substitute parse string, and recursively
17873 * calling reg() on it, instead of proceeding */
17874 if (multi_char_matches) {
17875 SV * substitute_parse = newSVpvn_flags("?:", 2, SVs_TEMP);
17878 char *save_end = RExC_end;
17879 char *save_parse = RExC_parse;
17880 char *save_start = RExC_start;
17881 Size_t constructed_prefix_len = 0; /* This gives the length of the
17882 constructed portion of the
17883 substitute parse. */
17884 bool first_time = TRUE; /* First multi-char occurrence doesn't get
17889 /* Only one level of recursion allowed */
17890 assert(RExC_copy_start_in_constructed == RExC_precomp);
17892 #if 0 /* Have decided not to deal with multi-char folds in inverted classes,
17893 because too confusing */
17895 sv_catpvs(substitute_parse, "(?:");
17899 /* Look at the longest folds first */
17900 for (cp_count = av_tindex_skip_len_mg(multi_char_matches);
17905 if (av_exists(multi_char_matches, cp_count)) {
17906 AV** this_array_ptr;
17909 this_array_ptr = (AV**) av_fetch(multi_char_matches,
17911 while ((this_sequence = av_pop(*this_array_ptr)) !=
17914 if (! first_time) {
17915 sv_catpvs(substitute_parse, "|");
17917 first_time = FALSE;
17919 sv_catpv(substitute_parse, SvPVX(this_sequence));
17924 /* If the character class contains anything else besides these
17925 * multi-character folds, have to include it in recursive parsing */
17926 if (element_count) {
17927 sv_catpvs(substitute_parse, "|[");
17928 constructed_prefix_len = SvCUR(substitute_parse);
17929 sv_catpvn(substitute_parse, orig_parse, RExC_parse - orig_parse);
17931 /* Put in a closing ']' only if not going off the end, as otherwise
17932 * we are adding something that really isn't there */
17933 if (RExC_parse < RExC_end) {
17934 sv_catpvs(substitute_parse, "]");
17938 sv_catpvs(substitute_parse, ")");
17941 /* This is a way to get the parse to skip forward a whole named
17942 * sequence instead of matching the 2nd character when it fails the
17944 sv_catpvs(substitute_parse, "(*THEN)(*SKIP)(*FAIL)|.)");
17948 /* Set up the data structure so that any errors will be properly
17949 * reported. See the comments at the definition of
17950 * REPORT_LOCATION_ARGS for details */
17951 RExC_copy_start_in_input = (char *) orig_parse;
17952 RExC_start = RExC_parse = SvPV(substitute_parse, len);
17953 RExC_copy_start_in_constructed = RExC_start + constructed_prefix_len;
17954 RExC_end = RExC_parse + len;
17955 RExC_in_multi_char_class = 1;
17957 ret = reg(pRExC_state, 1, ®_flags, depth+1);
17959 *flagp |= reg_flags & (HASWIDTH|SIMPLE|SPSTART|POSTPONED|RESTART_PARSE|NEED_UTF8);
17961 /* And restore so can parse the rest of the pattern */
17962 RExC_parse = save_parse;
17963 RExC_start = RExC_copy_start_in_constructed = RExC_copy_start_in_input = save_start;
17964 RExC_end = save_end;
17965 RExC_in_multi_char_class = 0;
17966 SvREFCNT_dec_NN(multi_char_matches);
17970 /* If folding, we calculate all characters that could fold to or from the
17971 * ones already on the list */
17972 if (cp_foldable_list) {
17974 UV start, end; /* End points of code point ranges */
17976 SV* fold_intersection = NULL;
17979 /* Our calculated list will be for Unicode rules. For locale
17980 * matching, we have to keep a separate list that is consulted at
17981 * runtime only when the locale indicates Unicode rules (and we
17982 * don't include potential matches in the ASCII/Latin1 range, as
17983 * any code point could fold to any other, based on the run-time
17984 * locale). For non-locale, we just use the general list */
17986 use_list = &only_utf8_locale_list;
17989 use_list = &cp_list;
17992 /* Only the characters in this class that participate in folds need
17993 * be checked. Get the intersection of this class and all the
17994 * possible characters that are foldable. This can quickly narrow
17995 * down a large class */
17996 _invlist_intersection(PL_in_some_fold, cp_foldable_list,
17997 &fold_intersection);
17999 /* Now look at the foldable characters in this class individually */
18000 invlist_iterinit(fold_intersection);
18001 while (invlist_iternext(fold_intersection, &start, &end)) {
18005 /* Look at every character in the range */
18006 for (j = start; j <= end; j++) {
18007 U8 foldbuf[UTF8_MAXBYTES_CASE+1];
18010 Size_t folds_count;
18011 unsigned int first_fold;
18012 const unsigned int * remaining_folds;
18016 /* Under /l, we don't know what code points below 256
18017 * fold to, except we do know the MICRO SIGN folds to
18018 * an above-255 character if the locale is UTF-8, so we
18019 * add it to the special list (in *use_list) Otherwise
18020 * we know now what things can match, though some folds
18021 * are valid under /d only if the target is UTF-8.
18022 * Those go in a separate list */
18023 if ( IS_IN_SOME_FOLD_L1(j)
18024 && ! (LOC && j != MICRO_SIGN))
18027 /* ASCII is always matched; non-ASCII is matched
18028 * only under Unicode rules (which could happen
18029 * under /l if the locale is a UTF-8 one */
18030 if (isASCII(j) || ! DEPENDS_SEMANTICS) {
18031 *use_list = add_cp_to_invlist(*use_list,
18032 PL_fold_latin1[j]);
18034 else if (j != PL_fold_latin1[j]) {
18035 upper_latin1_only_utf8_matches
18036 = add_cp_to_invlist(
18037 upper_latin1_only_utf8_matches,
18038 PL_fold_latin1[j]);
18042 if (HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE(j)
18043 && (! isASCII(j) || ! ASCII_FOLD_RESTRICTED))
18045 add_above_Latin1_folds(pRExC_state,
18052 /* Here is an above Latin1 character. We don't have the
18053 * rules hard-coded for it. First, get its fold. This is
18054 * the simple fold, as the multi-character folds have been
18055 * handled earlier and separated out */
18056 folded = _to_uni_fold_flags(j, foldbuf, &foldlen,
18057 (ASCII_FOLD_RESTRICTED)
18058 ? FOLD_FLAGS_NOMIX_ASCII
18061 /* Single character fold of above Latin1. Add everything
18062 * in its fold closure to the list that this node should
18064 folds_count = _inverse_folds(folded, &first_fold,
18066 for (k = 0; k <= folds_count; k++) {
18067 UV c = (k == 0) /* First time through use itself */
18069 : (k == 1) /* 2nd time use, the first fold */
18072 /* Then the remaining ones */
18073 : remaining_folds[k-2];
18075 /* /aa doesn't allow folds between ASCII and non- */
18076 if (( ASCII_FOLD_RESTRICTED
18077 && (isASCII(c) != isASCII(j))))
18082 /* Folds under /l which cross the 255/256 boundary are
18083 * added to a separate list. (These are valid only
18084 * when the locale is UTF-8.) */
18085 if (c < 256 && LOC) {
18086 *use_list = add_cp_to_invlist(*use_list, c);
18090 if (isASCII(c) || c > 255 || AT_LEAST_UNI_SEMANTICS)
18092 cp_list = add_cp_to_invlist(cp_list, c);
18095 /* Similarly folds involving non-ascii Latin1
18096 * characters under /d are added to their list */
18097 upper_latin1_only_utf8_matches
18098 = add_cp_to_invlist(
18099 upper_latin1_only_utf8_matches,
18105 SvREFCNT_dec_NN(fold_intersection);
18108 /* Now that we have finished adding all the folds, there is no reason
18109 * to keep the foldable list separate */
18110 _invlist_union(cp_list, cp_foldable_list, &cp_list);
18111 SvREFCNT_dec_NN(cp_foldable_list);
18114 /* And combine the result (if any) with any inversion lists from posix
18115 * classes. The lists are kept separate up to now because we don't want to
18116 * fold the classes */
18117 if (simple_posixes) { /* These are the classes known to be unaffected by
18120 _invlist_union(cp_list, simple_posixes, &cp_list);
18121 SvREFCNT_dec_NN(simple_posixes);
18124 cp_list = simple_posixes;
18127 if (posixes || nposixes) {
18128 if (! DEPENDS_SEMANTICS) {
18130 /* For everything but /d, we can just add the current 'posixes' and
18131 * 'nposixes' to the main list */
18134 _invlist_union(cp_list, posixes, &cp_list);
18135 SvREFCNT_dec_NN(posixes);
18143 _invlist_union(cp_list, nposixes, &cp_list);
18144 SvREFCNT_dec_NN(nposixes);
18147 cp_list = nposixes;
18152 /* Under /d, things like \w match upper Latin1 characters only if
18153 * the target string is in UTF-8. But things like \W match all the
18154 * upper Latin1 characters if the target string is not in UTF-8.
18156 * Handle the case with something like \W separately */
18158 SV* only_non_utf8_list = invlist_clone(PL_UpperLatin1, NULL);
18160 /* A complemented posix class matches all upper Latin1
18161 * characters if not in UTF-8. And it matches just certain
18162 * ones when in UTF-8. That means those certain ones are
18163 * matched regardless, so can just be added to the
18164 * unconditional list */
18166 _invlist_union(cp_list, nposixes, &cp_list);
18167 SvREFCNT_dec_NN(nposixes);
18171 cp_list = nposixes;
18174 /* Likewise for 'posixes' */
18175 _invlist_union(posixes, cp_list, &cp_list);
18177 /* Likewise for anything else in the range that matched only
18179 if (upper_latin1_only_utf8_matches) {
18180 _invlist_union(cp_list,
18181 upper_latin1_only_utf8_matches,
18183 SvREFCNT_dec_NN(upper_latin1_only_utf8_matches);
18184 upper_latin1_only_utf8_matches = NULL;
18187 /* If we don't match all the upper Latin1 characters regardless
18188 * of UTF-8ness, we have to set a flag to match the rest when
18190 _invlist_subtract(only_non_utf8_list, cp_list,
18191 &only_non_utf8_list);
18192 if (_invlist_len(only_non_utf8_list) != 0) {
18193 anyof_flags |= ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER;
18195 SvREFCNT_dec_NN(only_non_utf8_list);
18198 /* Here there were no complemented posix classes. That means
18199 * the upper Latin1 characters in 'posixes' match only when the
18200 * target string is in UTF-8. So we have to add them to the
18201 * list of those types of code points, while adding the
18202 * remainder to the unconditional list.
18204 * First calculate what they are */
18205 SV* nonascii_but_latin1_properties = NULL;
18206 _invlist_intersection(posixes, PL_UpperLatin1,
18207 &nonascii_but_latin1_properties);
18209 /* And add them to the final list of such characters. */
18210 _invlist_union(upper_latin1_only_utf8_matches,
18211 nonascii_but_latin1_properties,
18212 &upper_latin1_only_utf8_matches);
18214 /* Remove them from what now becomes the unconditional list */
18215 _invlist_subtract(posixes, nonascii_but_latin1_properties,
18218 /* And add those unconditional ones to the final list */
18220 _invlist_union(cp_list, posixes, &cp_list);
18221 SvREFCNT_dec_NN(posixes);
18228 SvREFCNT_dec(nonascii_but_latin1_properties);
18230 /* Get rid of any characters from the conditional list that we
18231 * now know are matched unconditionally, which may make that
18233 _invlist_subtract(upper_latin1_only_utf8_matches,
18235 &upper_latin1_only_utf8_matches);
18236 if (_invlist_len(upper_latin1_only_utf8_matches) == 0) {
18237 SvREFCNT_dec_NN(upper_latin1_only_utf8_matches);
18238 upper_latin1_only_utf8_matches = NULL;
18244 /* And combine the result (if any) with any inversion list from properties.
18245 * The lists are kept separate up to now so that we can distinguish the two
18246 * in regards to matching above-Unicode. A run-time warning is generated
18247 * if a Unicode property is matched against a non-Unicode code point. But,
18248 * we allow user-defined properties to match anything, without any warning,
18249 * and we also suppress the warning if there is a portion of the character
18250 * class that isn't a Unicode property, and which matches above Unicode, \W
18251 * or [\x{110000}] for example.
18252 * (Note that in this case, unlike the Posix one above, there is no
18253 * <upper_latin1_only_utf8_matches>, because having a Unicode property
18254 * forces Unicode semantics */
18258 /* If it matters to the final outcome, see if a non-property
18259 * component of the class matches above Unicode. If so, the
18260 * warning gets suppressed. This is true even if just a single
18261 * such code point is specified, as, though not strictly correct if
18262 * another such code point is matched against, the fact that they
18263 * are using above-Unicode code points indicates they should know
18264 * the issues involved */
18266 warn_super = ! (invert
18267 ^ (invlist_highest(cp_list) > PERL_UNICODE_MAX));
18270 _invlist_union(properties, cp_list, &cp_list);
18271 SvREFCNT_dec_NN(properties);
18274 cp_list = properties;
18279 |= ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER;
18281 /* Because an ANYOF node is the only one that warns, this node
18282 * can't be optimized into something else */
18283 optimizable = FALSE;
18287 /* Here, we have calculated what code points should be in the character
18290 * Now we can see about various optimizations. Fold calculation (which we
18291 * did above) needs to take place before inversion. Otherwise /[^k]/i
18292 * would invert to include K, which under /i would match k, which it
18293 * shouldn't. Therefore we can't invert folded locale now, as it won't be
18294 * folded until runtime */
18296 /* If we didn't do folding, it's because some information isn't available
18297 * until runtime; set the run-time fold flag for these We know to set the
18298 * flag if we have a non-NULL list for UTF-8 locales, or the class matches
18299 * at least one 0-255 range code point */
18302 /* Some things on the list might be unconditionally included because of
18303 * other components. Remove them, and clean up the list if it goes to
18305 if (only_utf8_locale_list && cp_list) {
18306 _invlist_subtract(only_utf8_locale_list, cp_list,
18307 &only_utf8_locale_list);
18309 if (_invlist_len(only_utf8_locale_list) == 0) {
18310 SvREFCNT_dec_NN(only_utf8_locale_list);
18311 only_utf8_locale_list = NULL;
18314 if ( only_utf8_locale_list
18315 || (cp_list && ( _invlist_contains_cp(cp_list, LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE)
18316 || _invlist_contains_cp(cp_list, LATIN_SMALL_LETTER_DOTLESS_I))))
18318 has_runtime_dependency |= HAS_L_RUNTIME_DEPENDENCY;
18321 | ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD;
18323 else if (cp_list) { /* Look to see if a 0-255 code point is in list */
18325 invlist_iterinit(cp_list);
18326 if (invlist_iternext(cp_list, &start, &end) && start < 256) {
18327 anyof_flags |= ANYOFL_FOLD;
18328 has_runtime_dependency |= HAS_L_RUNTIME_DEPENDENCY;
18330 invlist_iterfinish(cp_list);
18333 else if ( DEPENDS_SEMANTICS
18334 && ( upper_latin1_only_utf8_matches
18335 || (anyof_flags & ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER)))
18337 RExC_seen_d_op = TRUE;
18338 has_runtime_dependency |= HAS_D_RUNTIME_DEPENDENCY;
18341 /* Optimize inverted patterns (e.g. [^a-z]) when everything is known at
18345 && ! has_runtime_dependency)
18347 _invlist_invert(cp_list);
18349 /* Clear the invert flag since have just done it here */
18354 *ret_invlist = cp_list;
18359 /* All possible optimizations below still have these characteristics.
18360 * (Multi-char folds aren't SIMPLE, but they don't get this far in this
18362 *flagp |= HASWIDTH|SIMPLE;
18364 if (anyof_flags & ANYOF_LOCALE_FLAGS) {
18365 RExC_contains_locale = 1;
18368 /* Some character classes are equivalent to other nodes. Such nodes take
18369 * up less room, and some nodes require fewer operations to execute, than
18370 * ANYOF nodes. EXACTish nodes may be joinable with adjacent nodes to
18371 * improve efficiency. */
18374 PERL_UINT_FAST8_T i;
18375 Size_t partial_cp_count = 0;
18376 UV start[MAX_FOLD_FROMS+1] = { 0 }; /* +1 for the folded-to char */
18377 UV end[MAX_FOLD_FROMS+1] = { 0 };
18379 if (cp_list) { /* Count the code points in enough ranges that we would
18380 see all the ones possible in any fold in this version
18383 invlist_iterinit(cp_list);
18384 for (i = 0; i <= MAX_FOLD_FROMS; i++) {
18385 if (! invlist_iternext(cp_list, &start[i], &end[i])) {
18388 partial_cp_count += end[i] - start[i] + 1;
18391 invlist_iterfinish(cp_list);
18394 /* If we know at compile time that this matches every possible code
18395 * point, any run-time dependencies don't matter */
18396 if (start[0] == 0 && end[0] == UV_MAX) {
18398 ret = reganode(pRExC_state, OPFAIL, 0);
18401 ret = reg_node(pRExC_state, SANY);
18407 /* Similarly, for /l posix classes, if both a class and its
18408 * complement match, any run-time dependencies don't matter */
18410 for (namedclass = 0; namedclass < ANYOF_POSIXL_MAX;
18413 if ( POSIXL_TEST(posixl, namedclass) /* class */
18414 && POSIXL_TEST(posixl, namedclass + 1)) /* its complement */
18417 ret = reganode(pRExC_state, OPFAIL, 0);
18420 ret = reg_node(pRExC_state, SANY);
18426 /* For well-behaved locales, some classes are subsets of others,
18427 * so complementing the subset and including the non-complemented
18428 * superset should match everything, like [\D[:alnum:]], and
18429 * [[:^alpha:][:alnum:]], but some implementations of locales are
18430 * buggy, and khw thinks its a bad idea to have optimization change
18431 * behavior, even if it avoids an OS bug in a given case */
18433 #define isSINGLE_BIT_SET(n) isPOWER_OF_2(n)
18435 /* If is a single posix /l class, can optimize to just that op.
18436 * Such a node will not match anything in the Latin1 range, as that
18437 * is not determinable until runtime, but will match whatever the
18438 * class does outside that range. (Note that some classes won't
18439 * match anything outside the range, like [:ascii:]) */
18440 if ( isSINGLE_BIT_SET(posixl)
18441 && (partial_cp_count == 0 || start[0] > 255))
18444 SV * class_above_latin1 = NULL;
18445 bool already_inverted;
18446 bool are_equivalent;
18448 /* Compute which bit is set, which is the same thing as, e.g.,
18449 * ANYOF_CNTRL. From
18450 * https://graphics.stanford.edu/~seander/bithacks.html#IntegerLogDeBruijn
18452 static const int MultiplyDeBruijnBitPosition2[32] =
18454 0, 1, 28, 2, 29, 14, 24, 3, 30, 22, 20, 15, 25, 17, 4, 8,
18455 31, 27, 13, 23, 21, 19, 16, 7, 26, 12, 18, 6, 11, 5, 10, 9
18458 namedclass = MultiplyDeBruijnBitPosition2[(posixl
18459 * 0x077CB531U) >> 27];
18460 classnum = namedclass_to_classnum(namedclass);
18462 /* The named classes are such that the inverted number is one
18463 * larger than the non-inverted one */
18464 already_inverted = namedclass
18465 - classnum_to_namedclass(classnum);
18467 /* Create an inversion list of the official property, inverted
18468 * if the constructed node list is inverted, and restricted to
18469 * only the above latin1 code points, which are the only ones
18470 * known at compile time */
18471 _invlist_intersection_maybe_complement_2nd(
18473 PL_XPosix_ptrs[classnum],
18475 &class_above_latin1);
18476 are_equivalent = _invlistEQ(class_above_latin1, cp_list,
18478 SvREFCNT_dec_NN(class_above_latin1);
18480 if (are_equivalent) {
18482 /* Resolve the run-time inversion flag with this possibly
18483 * inverted class */
18484 invert = invert ^ already_inverted;
18486 ret = reg_node(pRExC_state,
18487 POSIXL + invert * (NPOSIXL - POSIXL));
18488 FLAGS(REGNODE_p(ret)) = classnum;
18494 /* khw can't think of any other possible transformation involving
18496 if (has_runtime_dependency & HAS_USER_DEFINED_PROPERTY) {
18500 if (! has_runtime_dependency) {
18502 /* If the list is empty, nothing matches. This happens, for
18503 * example, when a Unicode property that doesn't match anything is
18504 * the only element in the character class (perluniprops.pod notes
18505 * such properties). */
18506 if (partial_cp_count == 0) {
18508 ret = reg_node(pRExC_state, SANY);
18511 ret = reganode(pRExC_state, OPFAIL, 0);
18517 /* If matches everything but \n */
18518 if ( start[0] == 0 && end[0] == '\n' - 1
18519 && start[1] == '\n' + 1 && end[1] == UV_MAX)
18522 ret = reg_node(pRExC_state, REG_ANY);
18528 /* Next see if can optimize classes that contain just a few code points
18529 * into an EXACTish node. The reason to do this is to let the
18530 * optimizer join this node with adjacent EXACTish ones.
18532 * An EXACTFish node can be generated even if not under /i, and vice
18533 * versa. But care must be taken. An EXACTFish node has to be such
18534 * that it only matches precisely the code points in the class, but we
18535 * want to generate the least restrictive one that does that, to
18536 * increase the odds of being able to join with an adjacent node. For
18537 * example, if the class contains [kK], we have to make it an EXACTFAA
18538 * node to prevent the KELVIN SIGN from matching. Whether we are under
18539 * /i or not is irrelevant in this case. Less obvious is the pattern
18540 * qr/[\x{02BC}]n/i. U+02BC is MODIFIER LETTER APOSTROPHE. That is
18541 * supposed to match the single character U+0149 LATIN SMALL LETTER N
18542 * PRECEDED BY APOSTROPHE. And so even though there is no simple fold
18543 * that includes \X{02BC}, there is a multi-char fold that does, and so
18544 * the node generated for it must be an EXACTFish one. On the other
18545 * hand qr/:/i should generate a plain EXACT node since the colon
18546 * participates in no fold whatsoever, and having it EXACT tells the
18547 * optimizer the target string cannot match unless it has a colon in
18550 * We don't typically generate an EXACTish node if doing so would
18551 * require changing the pattern to UTF-8, as that affects /d and
18552 * otherwise is slower. However, under /i, not changing to UTF-8 can
18553 * miss some potential multi-character folds. We calculate the
18554 * EXACTish node, and then decide if something would be missed if we
18559 /* Only try if there are no more code points in the class than
18560 * in the max possible fold */
18561 && partial_cp_count > 0 && partial_cp_count <= MAX_FOLD_FROMS + 1
18563 && (start[0] < 256 || UTF || FOLD))
18565 if (partial_cp_count == 1 && ! upper_latin1_only_utf8_matches)
18567 /* We can always make a single code point class into an
18568 * EXACTish node. */
18572 /* Here is /l: Use EXACTL, except /li indicates EXACTFL,
18573 * as that means there is a fold not known until runtime so
18574 * shows as only a single code point here. */
18575 op = (FOLD) ? EXACTFL : EXACTL;
18577 else if (! FOLD) { /* Not /l and not /i */
18578 op = (start[0] < 256) ? EXACT : EXACT_ONLY8;
18580 else if (start[0] < 256) { /* /i, not /l, and the code point is
18583 /* Under /i, it gets a little tricky. A code point that
18584 * doesn't participate in a fold should be an EXACT node.
18585 * We know this one isn't the result of a simple fold, or
18586 * there'd be more than one code point in the list, but it
18587 * could be part of a multi- character fold. In that case
18588 * we better not create an EXACT node, as we would wrongly
18589 * be telling the optimizer that this code point must be in
18590 * the target string, and that is wrong. This is because
18591 * if the sequence around this code point forms a
18592 * multi-char fold, what needs to be in the string could be
18593 * the code point that folds to the sequence.
18595 * This handles the case of below-255 code points, as we
18596 * have an easy look up for those. The next clause handles
18597 * the above-256 one */
18598 op = IS_IN_SOME_FOLD_L1(start[0])
18602 else { /* /i, larger code point. Since we are under /i, and
18603 have just this code point, we know that it can't
18604 fold to something else, so PL_InMultiCharFold
18606 op = _invlist_contains_cp(PL_InMultiCharFold,
18614 else if ( ! (has_runtime_dependency & ~HAS_D_RUNTIME_DEPENDENCY)
18615 && _invlist_contains_cp(PL_in_some_fold, start[0]))
18617 /* Here, the only runtime dependency, if any, is from /d, and
18618 * the class matches more than one code point, and the lowest
18619 * code point participates in some fold. It might be that the
18620 * other code points are /i equivalent to this one, and hence
18621 * they would representable by an EXACTFish node. Above, we
18622 * eliminated classes that contain too many code points to be
18623 * EXACTFish, with the test for MAX_FOLD_FROMS
18625 * First, special case the ASCII fold pairs, like 'B' and 'b'.
18626 * We do this because we have EXACTFAA at our disposal for the
18628 if (partial_cp_count == 2 && isASCII(start[0])) {
18630 /* The only ASCII characters that participate in folds are
18632 assert(isALPHA(start[0]));
18633 if ( end[0] == start[0] /* First range is a single
18634 character, so 2nd exists */
18635 && isALPHA_FOLD_EQ(start[0], start[1]))
18638 /* Here, is part of an ASCII fold pair */
18640 if ( ASCII_FOLD_RESTRICTED
18641 || HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE(start[0]))
18643 /* If the second clause just above was true, it
18644 * means we can't be under /i, or else the list
18645 * would have included more than this fold pair.
18646 * Therefore we have to exclude the possibility of
18647 * whatever else it is that folds to these, by
18648 * using EXACTFAA */
18651 else if (HAS_NONLATIN1_FOLD_CLOSURE(start[0])) {
18653 /* Here, there's no simple fold that start[0] is part
18654 * of, but there is a multi-character one. If we
18655 * are not under /i, we want to exclude that
18656 * possibility; if under /i, we want to include it
18658 op = (FOLD) ? EXACTFU : EXACTFAA;
18662 /* Here, the only possible fold start[0] particpates in
18663 * is with start[1]. /i or not isn't relevant */
18667 value = toFOLD(start[0]);
18670 else if ( ! upper_latin1_only_utf8_matches
18671 || ( _invlist_len(upper_latin1_only_utf8_matches)
18674 invlist_highest(upper_latin1_only_utf8_matches)]
18677 /* Here, the smallest character is non-ascii or there are
18678 * more than 2 code points matched by this node. Also, we
18679 * either don't have /d UTF-8 dependent matches, or if we
18680 * do, they look like they could be a single character that
18681 * is the fold of the lowest one in the always-match list.
18682 * This test quickly excludes most of the false positives
18683 * when there are /d UTF-8 depdendent matches. These are
18684 * like LATIN CAPITAL LETTER A WITH GRAVE matching LATIN
18685 * SMALL LETTER A WITH GRAVE iff the target string is
18686 * UTF-8. (We don't have to worry above about exceeding
18687 * the array bounds of PL_fold_latin1[] because any code
18688 * point in 'upper_latin1_only_utf8_matches' is below 256.)
18690 * EXACTFAA would apply only to pairs (hence exactly 2 code
18691 * points) in the ASCII range, so we can't use it here to
18692 * artificially restrict the fold domain, so we check if
18693 * the class does or does not match some EXACTFish node.
18694 * Further, if we aren't under /i, and and the folded-to
18695 * character is part of a multi-character fold, we can't do
18696 * this optimization, as the sequence around it could be
18697 * that multi-character fold, and we don't here know the
18698 * context, so we have to assume it is that multi-char
18699 * fold, to prevent potential bugs.
18701 * To do the general case, we first find the fold of the
18702 * lowest code point (which may be higher than the lowest
18703 * one), then find everything that folds to it. (The data
18704 * structure we have only maps from the folded code points,
18705 * so we have to do the earlier step.) */
18708 U8 foldbuf[UTF8_MAXBYTES_CASE];
18709 UV folded = _to_uni_fold_flags(start[0],
18710 foldbuf, &foldlen, 0);
18711 unsigned int first_fold;
18712 const unsigned int * remaining_folds;
18713 Size_t folds_to_this_cp_count = _inverse_folds(
18717 Size_t folds_count = folds_to_this_cp_count + 1;
18718 SV * fold_list = _new_invlist(folds_count);
18721 /* If there are UTF-8 dependent matches, create a temporary
18722 * list of what this node matches, including them. */
18723 SV * all_cp_list = NULL;
18724 SV ** use_this_list = &cp_list;
18726 if (upper_latin1_only_utf8_matches) {
18727 all_cp_list = _new_invlist(0);
18728 use_this_list = &all_cp_list;
18729 _invlist_union(cp_list,
18730 upper_latin1_only_utf8_matches,
18734 /* Having gotten everything that participates in the fold
18735 * containing the lowest code point, we turn that into an
18736 * inversion list, making sure everything is included. */
18737 fold_list = add_cp_to_invlist(fold_list, start[0]);
18738 fold_list = add_cp_to_invlist(fold_list, folded);
18739 if (folds_to_this_cp_count > 0) {
18740 fold_list = add_cp_to_invlist(fold_list, first_fold);
18741 for (i = 0; i + 1 < folds_to_this_cp_count; i++) {
18742 fold_list = add_cp_to_invlist(fold_list,
18743 remaining_folds[i]);
18747 /* If the fold list is identical to what's in this ANYOF
18748 * node, the node can be represented by an EXACTFish one
18750 if (_invlistEQ(*use_this_list, fold_list,
18751 0 /* Don't complement */ )
18754 /* But, we have to be careful, as mentioned above.
18755 * Just the right sequence of characters could match
18756 * this if it is part of a multi-character fold. That
18757 * IS what we want if we are under /i. But it ISN'T
18758 * what we want if not under /i, as it could match when
18759 * it shouldn't. So, when we aren't under /i and this
18760 * character participates in a multi-char fold, we
18761 * don't optimize into an EXACTFish node. So, for each
18762 * case below we have to check if we are folding
18763 * and if not, if it is not part of a multi-char fold.
18765 if (start[0] > 255) { /* Highish code point */
18766 if (FOLD || ! _invlist_contains_cp(
18767 PL_InMultiCharFold, folded))
18771 : (ASCII_FOLD_RESTRICTED)
18776 } /* Below, the lowest code point < 256 */
18779 && DEPENDS_SEMANTICS)
18780 { /* An EXACTF node containing a single character
18781 's', can be an EXACTFU if it doesn't get
18782 joined with an adjacent 's' */
18783 op = EXACTFU_S_EDGE;
18787 || ! HAS_NONLATIN1_FOLD_CLOSURE(start[0]))
18789 if (upper_latin1_only_utf8_matches) {
18792 /* We can't use the fold, as that only matches
18796 else if ( UNLIKELY(start[0] == MICRO_SIGN)
18798 { /* EXACTFUP is a special node for this
18800 op = (ASCII_FOLD_RESTRICTED)
18803 value = MICRO_SIGN;
18805 else if ( ASCII_FOLD_RESTRICTED
18806 && ! isASCII(start[0]))
18807 { /* For ASCII under /iaa, we can use EXACTFU
18819 SvREFCNT_dec_NN(fold_list);
18820 SvREFCNT_dec(all_cp_list);
18826 /* Here, we have calculated what EXACTish node we would use.
18827 * But we don't use it if it would require converting the
18828 * pattern to UTF-8, unless not using it could cause us to miss
18829 * some folds (hence be buggy) */
18831 if (! UTF && value > 255) {
18832 SV * in_multis = NULL;
18836 /* If there is no code point that is part of a multi-char
18837 * fold, then there aren't any matches, so we don't do this
18838 * optimization. Otherwise, it could match depending on
18839 * the context around us, so we do upgrade */
18840 _invlist_intersection(PL_InMultiCharFold, cp_list, &in_multis);
18841 if (UNLIKELY(_invlist_len(in_multis) != 0)) {
18842 REQUIRE_UTF8(flagp);
18850 U8 len = (UTF) ? UVCHR_SKIP(value) : 1;
18852 ret = regnode_guts(pRExC_state, op, len, "exact");
18853 FILL_NODE(ret, op);
18854 RExC_emit += 1 + STR_SZ(len);
18855 STR_LEN(REGNODE_p(ret)) = len;
18857 *STRING(REGNODE_p(ret)) = (U8) value;
18860 uvchr_to_utf8((U8 *) STRING(REGNODE_p(ret)), value);
18867 if (! has_runtime_dependency) {
18869 /* See if this can be turned into an ANYOFM node. Think about the
18870 * bit patterns in two different bytes. In some positions, the
18871 * bits in each will be 1; and in other positions both will be 0;
18872 * and in some positions the bit will be 1 in one byte, and 0 in
18873 * the other. Let 'n' be the number of positions where the bits
18874 * differ. We create a mask which has exactly 'n' 0 bits, each in
18875 * a position where the two bytes differ. Now take the set of all
18876 * bytes that when ANDed with the mask yield the same result. That
18877 * set has 2**n elements, and is representable by just two 8 bit
18878 * numbers: the result and the mask. Importantly, matching the set
18879 * can be vectorized by creating a word full of the result bytes,
18880 * and a word full of the mask bytes, yielding a significant speed
18881 * up. Here, see if this node matches such a set. As a concrete
18882 * example consider [01], and the byte representing '0' which is
18883 * 0x30 on ASCII machines. It has the bits 0011 0000. Take the
18884 * mask 1111 1110. If we AND 0x31 and 0x30 with that mask we get
18885 * 0x30. Any other bytes ANDed yield something else. So [01],
18886 * which is a common usage, is optimizable into ANYOFM, and can
18887 * benefit from the speed up. We can only do this on UTF-8
18888 * invariant bytes, because they have the same bit patterns under
18890 PERL_UINT_FAST8_T inverted = 0;
18892 const PERL_UINT_FAST8_T max_permissible = 0xFF;
18894 const PERL_UINT_FAST8_T max_permissible = 0x7F;
18896 /* If doesn't fit the criteria for ANYOFM, invert and try again.
18897 * If that works we will instead later generate an NANYOFM, and
18898 * invert back when through */
18899 if (invlist_highest(cp_list) > max_permissible) {
18900 _invlist_invert(cp_list);
18904 if (invlist_highest(cp_list) <= max_permissible) {
18905 UV this_start, this_end;
18906 UV lowest_cp = UV_MAX; /* inited to suppress compiler warn */
18907 U8 bits_differing = 0;
18908 Size_t full_cp_count = 0;
18909 bool first_time = TRUE;
18911 /* Go through the bytes and find the bit positions that differ
18913 invlist_iterinit(cp_list);
18914 while (invlist_iternext(cp_list, &this_start, &this_end)) {
18915 unsigned int i = this_start;
18918 if (! UVCHR_IS_INVARIANT(i)) {
18922 first_time = FALSE;
18923 lowest_cp = this_start;
18925 /* We have set up the code point to compare with.
18926 * Don't compare it with itself */
18930 /* Find the bit positions that differ from the lowest code
18931 * point in the node. Keep track of all such positions by
18933 for (; i <= this_end; i++) {
18934 if (! UVCHR_IS_INVARIANT(i)) {
18938 bits_differing |= i ^ lowest_cp;
18941 full_cp_count += this_end - this_start + 1;
18944 /* At the end of the loop, we count how many bits differ from
18945 * the bits in lowest code point, call the count 'd'. If the
18946 * set we found contains 2**d elements, it is the closure of
18947 * all code points that differ only in those bit positions. To
18948 * convince yourself of that, first note that the number in the
18949 * closure must be a power of 2, which we test for. The only
18950 * way we could have that count and it be some differing set,
18951 * is if we got some code points that don't differ from the
18952 * lowest code point in any position, but do differ from each
18953 * other in some other position. That means one code point has
18954 * a 1 in that position, and another has a 0. But that would
18955 * mean that one of them differs from the lowest code point in
18956 * that position, which possibility we've already excluded. */
18957 if ( (inverted || full_cp_count > 1)
18958 && full_cp_count == 1U << PL_bitcount[bits_differing])
18962 op = ANYOFM + inverted;;
18964 /* We need to make the bits that differ be 0's */
18965 ANYOFM_mask = ~ bits_differing; /* This goes into FLAGS */
18967 /* The argument is the lowest code point */
18968 ret = reganode(pRExC_state, op, lowest_cp);
18969 FLAGS(REGNODE_p(ret)) = ANYOFM_mask;
18973 invlist_iterfinish(cp_list);
18977 _invlist_invert(cp_list);
18985 if (! (anyof_flags & ANYOF_LOCALE_FLAGS)) {
18986 PERL_UINT_FAST8_T type;
18987 SV * intersection = NULL;
18988 SV* d_invlist = NULL;
18990 /* See if this matches any of the POSIX classes. The POSIXA and
18991 * POSIXD ones are about the same speed as ANYOF ops, but take less
18992 * room; the ones that have above-Latin1 code point matches are
18993 * somewhat faster than ANYOF. */
18995 for (type = POSIXA; type >= POSIXD; type--) {
18998 if (type == POSIXL) { /* But not /l posix classes */
19002 for (posix_class = 0;
19003 posix_class <= _HIGHEST_REGCOMP_DOT_H_SYNC;
19006 SV** our_code_points = &cp_list;
19007 SV** official_code_points;
19010 if (type == POSIXA) {
19011 official_code_points = &PL_Posix_ptrs[posix_class];
19014 official_code_points = &PL_XPosix_ptrs[posix_class];
19017 /* Skip non-existent classes of this type. e.g. \v only
19018 * has an entry in PL_XPosix_ptrs */
19019 if (! *official_code_points) {
19023 /* Try both the regular class, and its inversion */
19024 for (try_inverted = 0; try_inverted < 2; try_inverted++) {
19025 bool this_inverted = invert ^ try_inverted;
19027 if (type != POSIXD) {
19029 /* This class that isn't /d can't match if we have
19030 * /d dependencies */
19031 if (has_runtime_dependency
19032 & HAS_D_RUNTIME_DEPENDENCY)
19037 else /* is /d */ if (! this_inverted) {
19039 /* /d classes don't match anything non-ASCII below
19040 * 256 unconditionally (which cp_list contains) */
19041 _invlist_intersection(cp_list, PL_UpperLatin1,
19043 if (_invlist_len(intersection) != 0) {
19047 SvREFCNT_dec(d_invlist);
19048 d_invlist = invlist_clone(cp_list, NULL);
19050 /* But under UTF-8 it turns into using /u rules.
19051 * Add the things it matches under these conditions
19052 * so that we check below that these are identical
19053 * to what the tested class should match */
19054 if (upper_latin1_only_utf8_matches) {
19057 upper_latin1_only_utf8_matches,
19060 our_code_points = &d_invlist;
19062 else { /* POSIXD, inverted. If this doesn't have this
19063 flag set, it isn't /d. */
19064 if (! (anyof_flags & ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER))
19068 our_code_points = &cp_list;
19071 /* Here, have weeded out some things. We want to see
19072 * if the list of characters this node contains
19073 * ('*our_code_points') precisely matches those of the
19074 * class we are currently checking against
19075 * ('*official_code_points'). */
19076 if (_invlistEQ(*our_code_points,
19077 *official_code_points,
19080 /* Here, they precisely match. Optimize this ANYOF
19081 * node into its equivalent POSIX one of the
19082 * correct type, possibly inverted */
19083 ret = reg_node(pRExC_state, (try_inverted)
19087 FLAGS(REGNODE_p(ret)) = posix_class;
19088 SvREFCNT_dec(d_invlist);
19089 SvREFCNT_dec(intersection);
19095 SvREFCNT_dec(d_invlist);
19096 SvREFCNT_dec(intersection);
19099 /* If didn't find an optimization and there is no need for a bitmap,
19100 * optimize to indicate that */
19101 if ( start[0] >= NUM_ANYOF_CODE_POINTS
19103 && ! upper_latin1_only_utf8_matches
19104 && anyof_flags == 0)
19106 U8 low_utf8[UTF8_MAXBYTES+1];
19107 UV highest_cp = invlist_highest(cp_list);
19111 /* Currently the maximum allowed code point by the system is
19112 * IV_MAX. Higher ones are reserved for future internal use. This
19113 * particular regnode can be used for higher ones, but we can't
19114 * calculate the code point of those. IV_MAX suffices though, as
19115 * it will be a large first byte */
19116 (void) uvchr_to_utf8(low_utf8, MIN(start[0], IV_MAX));
19118 /* We store the lowest possible first byte of the UTF-8
19119 * representation, using the flags field. This allows for quick
19120 * ruling out of some inputs without having to convert from UTF-8
19121 * to code point. For EBCDIC, this has to be I8. */
19122 anyof_flags = NATIVE_UTF8_TO_I8(low_utf8[0]);
19124 /* If the first UTF-8 start byte for the highest code point in the
19125 * range is suitably small, we may be able to get an upper bound as
19127 if (highest_cp <= IV_MAX) {
19128 U8 high_utf8[UTF8_MAXBYTES+1];
19130 (void) uvchr_to_utf8(high_utf8, highest_cp);
19132 /* If the lowest and highest are the same, we can get an exact
19133 * first byte instead of a just minimum. We signal this with a
19134 * different regnode */
19135 if (low_utf8[0] == high_utf8[0]) {
19137 /* No need to convert to I8 for EBCDIC as this is an exact
19139 anyof_flags = low_utf8[0];
19142 else if (NATIVE_UTF8_TO_I8(high_utf8[0]) <= MAX_ANYOF_HRx_BYTE)
19145 /* Here, the high byte is not the same as the low, but is
19146 * small enough that its reasonable to have a loose upper
19147 * bound, which is packed in with the strict lower bound.
19148 * See comments at the definition of MAX_ANYOF_HRx_BYTE.
19149 * On EBCDIC platforms, I8 is used. On ASCII platforms I8
19150 * is the same thing as UTF-8 */
19153 U8 max_range_diff = MAX_ANYOF_HRx_BYTE - anyof_flags;
19154 U8 range_diff = NATIVE_UTF8_TO_I8(high_utf8[0])
19157 if (range_diff <= max_range_diff / 8) {
19160 else if (range_diff <= max_range_diff / 4) {
19163 else if (range_diff <= max_range_diff / 2) {
19166 anyof_flags = (anyof_flags - 0xC0) << 2 | bits;
19171 goto done_finding_op;
19173 } /* End of seeing if can optimize it into a different node */
19175 is_anyof: /* It's going to be an ANYOF node. */
19176 op = (has_runtime_dependency & HAS_D_RUNTIME_DEPENDENCY)
19186 ret = regnode_guts(pRExC_state, op, regarglen[op], "anyof");
19187 FILL_NODE(ret, op); /* We set the argument later */
19188 RExC_emit += 1 + regarglen[op];
19189 ANYOF_FLAGS(REGNODE_p(ret)) = anyof_flags;
19191 /* Here, <cp_list> contains all the code points we can determine at
19192 * compile time that match under all conditions. Go through it, and
19193 * for things that belong in the bitmap, put them there, and delete from
19194 * <cp_list>. While we are at it, see if everything above 255 is in the
19195 * list, and if so, set a flag to speed up execution */
19197 populate_ANYOF_from_invlist(REGNODE_p(ret), &cp_list);
19200 ANYOF_POSIXL_SET_TO_BITMAP(REGNODE_p(ret), posixl);
19204 ANYOF_FLAGS(REGNODE_p(ret)) |= ANYOF_INVERT;
19207 /* Here, the bitmap has been populated with all the Latin1 code points that
19208 * always match. Can now add to the overall list those that match only
19209 * when the target string is UTF-8 (<upper_latin1_only_utf8_matches>).
19211 if (upper_latin1_only_utf8_matches) {
19213 _invlist_union(cp_list,
19214 upper_latin1_only_utf8_matches,
19216 SvREFCNT_dec_NN(upper_latin1_only_utf8_matches);
19219 cp_list = upper_latin1_only_utf8_matches;
19221 ANYOF_FLAGS(REGNODE_p(ret)) |= ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP;
19224 set_ANYOF_arg(pRExC_state, REGNODE_p(ret), cp_list,
19225 (HAS_NONLOCALE_RUNTIME_PROPERTY_DEFINITION)
19227 only_utf8_locale_list);
19232 /* Here, the node is getting optimized into something that's not an ANYOF
19233 * one. Finish up. */
19235 Set_Node_Offset_Length(REGNODE_p(ret), orig_parse - RExC_start,
19236 RExC_parse - orig_parse);;
19237 SvREFCNT_dec(cp_list);;
19241 #undef HAS_NONLOCALE_RUNTIME_PROPERTY_DEFINITION
19244 S_set_ANYOF_arg(pTHX_ RExC_state_t* const pRExC_state,
19245 regnode* const node,
19247 SV* const runtime_defns,
19248 SV* const only_utf8_locale_list)
19250 /* Sets the arg field of an ANYOF-type node 'node', using information about
19251 * the node passed-in. If there is nothing outside the node's bitmap, the
19252 * arg is set to ANYOF_ONLY_HAS_BITMAP. Otherwise, it sets the argument to
19253 * the count returned by add_data(), having allocated and stored an array,
19256 * av[0] stores the inversion list defining this class as far as known at
19257 * this time, or PL_sv_undef if nothing definite is now known.
19258 * av[1] stores the inversion list of code points that match only if the
19259 * current locale is UTF-8, or if none, PL_sv_undef if there is an
19260 * av[2], or no entry otherwise.
19261 * av[2] stores the list of user-defined properties whose subroutine
19262 * definitions aren't known at this time, or no entry if none. */
19266 PERL_ARGS_ASSERT_SET_ANYOF_ARG;
19268 if (! cp_list && ! runtime_defns && ! only_utf8_locale_list) {
19269 assert(! (ANYOF_FLAGS(node)
19270 & ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP));
19271 ARG_SET(node, ANYOF_ONLY_HAS_BITMAP);
19274 AV * const av = newAV();
19278 av_store(av, INVLIST_INDEX, cp_list);
19281 if (only_utf8_locale_list) {
19282 av_store(av, ONLY_LOCALE_MATCHES_INDEX, only_utf8_locale_list);
19285 if (runtime_defns) {
19286 av_store(av, DEFERRED_USER_DEFINED_INDEX, SvREFCNT_inc(runtime_defns));
19289 rv = newRV_noinc(MUTABLE_SV(av));
19290 n = add_data(pRExC_state, STR_WITH_LEN("s"));
19291 RExC_rxi->data->data[n] = (void*)rv;
19296 #if !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION)
19298 Perl__get_regclass_nonbitmap_data(pTHX_ const regexp *prog,
19299 const regnode* node,
19302 SV** only_utf8_locale_ptr,
19303 SV** output_invlist)
19306 /* For internal core use only.
19307 * Returns the inversion list for the input 'node' in the regex 'prog'.
19308 * If <doinit> is 'true', will attempt to create the inversion list if not
19310 * If <listsvp> is non-null, will return the printable contents of the
19311 * property definition. This can be used to get debugging information
19312 * even before the inversion list exists, by calling this function with
19313 * 'doinit' set to false, in which case the components that will be used
19314 * to eventually create the inversion list are returned (in a printable
19316 * If <only_utf8_locale_ptr> is not NULL, it is where this routine is to
19317 * store an inversion list of code points that should match only if the
19318 * execution-time locale is a UTF-8 one.
19319 * If <output_invlist> is not NULL, it is where this routine is to store an
19320 * inversion list of the code points that would be instead returned in
19321 * <listsvp> if this were NULL. Thus, what gets output in <listsvp>
19322 * when this parameter is used, is just the non-code point data that
19323 * will go into creating the inversion list. This currently should be just
19324 * user-defined properties whose definitions were not known at compile
19325 * time. Using this parameter allows for easier manipulation of the
19326 * inversion list's data by the caller. It is illegal to call this
19327 * function with this parameter set, but not <listsvp>
19329 * Tied intimately to how S_set_ANYOF_arg sets up the data structure. Note
19330 * that, in spite of this function's name, the inversion list it returns
19331 * may include the bitmap data as well */
19333 SV *si = NULL; /* Input initialization string */
19334 SV* invlist = NULL;
19336 RXi_GET_DECL(prog, progi);
19337 const struct reg_data * const data = prog ? progi->data : NULL;
19339 PERL_ARGS_ASSERT__GET_REGCLASS_NONBITMAP_DATA;
19340 assert(! output_invlist || listsvp);
19342 if (data && data->count) {
19343 const U32 n = ARG(node);
19345 if (data->what[n] == 's') {
19346 SV * const rv = MUTABLE_SV(data->data[n]);
19347 AV * const av = MUTABLE_AV(SvRV(rv));
19348 SV **const ary = AvARRAY(av);
19350 invlist = ary[INVLIST_INDEX];
19352 if (av_tindex_skip_len_mg(av) >= ONLY_LOCALE_MATCHES_INDEX) {
19353 *only_utf8_locale_ptr = ary[ONLY_LOCALE_MATCHES_INDEX];
19356 if (av_tindex_skip_len_mg(av) >= DEFERRED_USER_DEFINED_INDEX) {
19357 si = ary[DEFERRED_USER_DEFINED_INDEX];
19360 if (doinit && (si || invlist)) {
19363 SV * msg = newSVpvs_flags("", SVs_TEMP);
19365 SV * prop_definition = handle_user_defined_property(
19366 "", 0, FALSE, /* There is no \p{}, \P{} */
19367 SvPVX_const(si)[1] - '0', /* /i or not has been
19368 stored here for just
19370 TRUE, /* run time */
19371 FALSE, /* This call must find the defn */
19372 si, /* The property definition */
19375 0 /* base level call */
19379 assert(prop_definition == NULL);
19381 Perl_croak(aTHX_ "%" UTF8f,
19382 UTF8fARG(SvUTF8(msg), SvCUR(msg), SvPVX(msg)));
19386 _invlist_union(invlist, prop_definition, &invlist);
19387 SvREFCNT_dec_NN(prop_definition);
19390 invlist = prop_definition;
19393 STATIC_ASSERT_STMT(ONLY_LOCALE_MATCHES_INDEX == 1 + INVLIST_INDEX);
19394 STATIC_ASSERT_STMT(DEFERRED_USER_DEFINED_INDEX == 1 + ONLY_LOCALE_MATCHES_INDEX);
19396 av_store(av, INVLIST_INDEX, invlist);
19397 av_fill(av, (ary[ONLY_LOCALE_MATCHES_INDEX])
19398 ? ONLY_LOCALE_MATCHES_INDEX:
19406 /* If requested, return a printable version of what this ANYOF node matches
19409 SV* matches_string = NULL;
19411 /* This function can be called at compile-time, before everything gets
19412 * resolved, in which case we return the currently best available
19413 * information, which is the string that will eventually be used to do
19414 * that resolving, 'si' */
19416 /* Here, we only have 'si' (and possibly some passed-in data in
19417 * 'invlist', which is handled below) If the caller only wants
19418 * 'si', use that. */
19419 if (! output_invlist) {
19420 matches_string = newSVsv(si);
19423 /* But if the caller wants an inversion list of the node, we
19424 * need to parse 'si' and place as much as possible in the
19425 * desired output inversion list, making 'matches_string' only
19426 * contain the currently unresolvable things */
19427 const char *si_string = SvPVX(si);
19428 STRLEN remaining = SvCUR(si);
19432 /* Ignore everything before the first new-line */
19433 while (*si_string != '\n' && remaining > 0) {
19437 assert(remaining > 0);
19442 while (remaining > 0) {
19444 /* The data consists of just strings defining user-defined
19445 * property names, but in prior incarnations, and perhaps
19446 * somehow from pluggable regex engines, it could still
19447 * hold hex code point definitions. Each component of a
19448 * range would be separated by a tab, and each range by a
19449 * new-line. If these are found, instead add them to the
19450 * inversion list */
19451 I32 grok_flags = PERL_SCAN_SILENT_ILLDIGIT
19452 |PERL_SCAN_SILENT_NON_PORTABLE;
19453 STRLEN len = remaining;
19454 UV cp = grok_hex(si_string, &len, &grok_flags, NULL);
19456 /* If the hex decode routine found something, it should go
19457 * up to the next \n */
19458 if ( *(si_string + len) == '\n') {
19459 if (count) { /* 2nd code point on line */
19460 *output_invlist = _add_range_to_invlist(*output_invlist, prev_cp, cp);
19463 *output_invlist = add_cp_to_invlist(*output_invlist, cp);
19466 goto prepare_for_next_iteration;
19469 /* If the hex decode was instead for the lower range limit,
19470 * save it, and go parse the upper range limit */
19471 if (*(si_string + len) == '\t') {
19472 assert(count == 0);
19476 prepare_for_next_iteration:
19477 si_string += len + 1;
19478 remaining -= len + 1;
19482 /* Here, didn't find a legal hex number. Just add it from
19483 * here to the next \n */
19486 while (*(si_string + len) != '\n' && remaining > 0) {
19490 if (*(si_string + len) == '\n') {
19494 if (matches_string) {
19495 sv_catpvn(matches_string, si_string, len - 1);
19498 matches_string = newSVpvn(si_string, len - 1);
19501 sv_catpvs(matches_string, " ");
19502 } /* end of loop through the text */
19504 assert(matches_string);
19505 if (SvCUR(matches_string)) { /* Get rid of trailing blank */
19506 SvCUR_set(matches_string, SvCUR(matches_string) - 1);
19508 } /* end of has an 'si' */
19511 /* Add the stuff that's already known */
19514 /* Again, if the caller doesn't want the output inversion list, put
19515 * everything in 'matches-string' */
19516 if (! output_invlist) {
19517 if ( ! matches_string) {
19518 matches_string = newSVpvs("\n");
19520 sv_catsv(matches_string, invlist_contents(invlist,
19521 TRUE /* traditional style */
19524 else if (! *output_invlist) {
19525 *output_invlist = invlist_clone(invlist, NULL);
19528 _invlist_union(*output_invlist, invlist, output_invlist);
19532 *listsvp = matches_string;
19537 #endif /* !defined(PERL_IN_XSUB_RE) || defined(PLUGGABLE_RE_EXTENSION) */
19539 /* reg_skipcomment()
19541 Absorbs an /x style # comment from the input stream,
19542 returning a pointer to the first character beyond the comment, or if the
19543 comment terminates the pattern without anything following it, this returns
19544 one past the final character of the pattern (in other words, RExC_end) and
19545 sets the REG_RUN_ON_COMMENT_SEEN flag.
19547 Note it's the callers responsibility to ensure that we are
19548 actually in /x mode
19552 PERL_STATIC_INLINE char*
19553 S_reg_skipcomment(RExC_state_t *pRExC_state, char* p)
19555 PERL_ARGS_ASSERT_REG_SKIPCOMMENT;
19559 while (p < RExC_end) {
19560 if (*(++p) == '\n') {
19565 /* we ran off the end of the pattern without ending the comment, so we have
19566 * to add an \n when wrapping */
19567 RExC_seen |= REG_RUN_ON_COMMENT_SEEN;
19572 S_skip_to_be_ignored_text(pTHX_ RExC_state_t *pRExC_state,
19574 const bool force_to_xmod
19577 /* If the text at the current parse position '*p' is a '(?#...)' comment,
19578 * or if we are under /x or 'force_to_xmod' is TRUE, and the text at '*p'
19579 * is /x whitespace, advance '*p' so that on exit it points to the first
19580 * byte past all such white space and comments */
19582 const bool use_xmod = force_to_xmod || (RExC_flags & RXf_PMf_EXTENDED);
19584 PERL_ARGS_ASSERT_SKIP_TO_BE_IGNORED_TEXT;
19586 assert( ! UTF || UTF8_IS_INVARIANT(**p) || UTF8_IS_START(**p));
19589 if (RExC_end - (*p) >= 3
19591 && *(*p + 1) == '?'
19592 && *(*p + 2) == '#')
19594 while (*(*p) != ')') {
19595 if ((*p) == RExC_end)
19596 FAIL("Sequence (?#... not terminated");
19604 const char * save_p = *p;
19605 while ((*p) < RExC_end) {
19607 if ((len = is_PATWS_safe((*p), RExC_end, UTF))) {
19610 else if (*(*p) == '#') {
19611 (*p) = reg_skipcomment(pRExC_state, (*p));
19617 if (*p != save_p) {
19630 Advances the parse position by one byte, unless that byte is the beginning
19631 of a '(?#...)' style comment, or is /x whitespace and /x is in effect. In
19632 those two cases, the parse position is advanced beyond all such comments and
19635 This is the UTF, (?#...), and /x friendly way of saying RExC_parse++.
19639 S_nextchar(pTHX_ RExC_state_t *pRExC_state)
19641 PERL_ARGS_ASSERT_NEXTCHAR;
19643 if (RExC_parse < RExC_end) {
19645 || UTF8_IS_INVARIANT(*RExC_parse)
19646 || UTF8_IS_START(*RExC_parse));
19648 RExC_parse += (UTF)
19649 ? UTF8_SAFE_SKIP(RExC_parse, RExC_end)
19652 skip_to_be_ignored_text(pRExC_state, &RExC_parse,
19653 FALSE /* Don't force /x */ );
19658 S_change_engine_size(pTHX_ RExC_state_t *pRExC_state, const Ptrdiff_t size)
19660 /* 'size' is the delta to add or subtract from the current memory allocated
19661 * to the regex engine being constructed */
19663 PERL_ARGS_ASSERT_CHANGE_ENGINE_SIZE;
19668 sizeof(regexp_internal) + (RExC_size + 1) * sizeof(regnode),
19669 /* +1 for REG_MAGIC */
19672 if ( RExC_rxi == NULL )
19673 FAIL("Regexp out of space");
19674 RXi_SET(RExC_rx, RExC_rxi);
19676 RExC_emit_start = RExC_rxi->program;
19678 Zero(REGNODE_p(RExC_emit), size, regnode);
19681 #ifdef RE_TRACK_PATTERN_OFFSETS
19682 Renew(RExC_offsets, 2*RExC_size+1, U32);
19684 Zero(RExC_offsets + 2*(RExC_size - size) + 1, 2 * size, U32);
19686 RExC_offsets[0] = RExC_size;
19690 STATIC regnode_offset
19691 S_regnode_guts(pTHX_ RExC_state_t *pRExC_state, const U8 op, const STRLEN extra_size, const char* const name)
19693 /* Allocate a regnode for 'op', with 'extra_size' extra space. It aligns
19694 * and increments RExC_size and RExC_emit
19696 * It returns the regnode's offset into the regex engine program */
19698 const regnode_offset ret = RExC_emit;
19700 GET_RE_DEBUG_FLAGS_DECL;
19702 PERL_ARGS_ASSERT_REGNODE_GUTS;
19704 SIZE_ALIGN(RExC_size);
19705 change_engine_size(pRExC_state, (Ptrdiff_t) 1 + extra_size);
19706 NODE_ALIGN_FILL(REGNODE_p(ret));
19707 #ifndef RE_TRACK_PATTERN_OFFSETS
19708 PERL_UNUSED_ARG(name);
19709 PERL_UNUSED_ARG(op);
19711 assert(extra_size >= regarglen[op] || PL_regkind[op] == ANYOF);
19713 if (RExC_offsets) { /* MJD */
19715 ("%s:%d: (op %s) %s %" UVuf " (len %" UVuf ") (max %" UVuf ").\n",
19718 (UV)(RExC_emit) > RExC_offsets[0]
19719 ? "Overwriting end of array!\n" : "OK",
19721 (UV)(RExC_parse - RExC_start),
19722 (UV)RExC_offsets[0]));
19723 Set_Node_Offset(REGNODE_p(RExC_emit), RExC_parse + (op == END));
19730 - reg_node - emit a node
19732 STATIC regnode_offset /* Location. */
19733 S_reg_node(pTHX_ RExC_state_t *pRExC_state, U8 op)
19735 const regnode_offset ret = regnode_guts(pRExC_state, op, regarglen[op], "reg_node");
19736 regnode_offset ptr = ret;
19738 PERL_ARGS_ASSERT_REG_NODE;
19740 assert(regarglen[op] == 0);
19742 FILL_ADVANCE_NODE(ptr, op);
19748 - reganode - emit a node with an argument
19750 STATIC regnode_offset /* Location. */
19751 S_reganode(pTHX_ RExC_state_t *pRExC_state, U8 op, U32 arg)
19753 const regnode_offset ret = regnode_guts(pRExC_state, op, regarglen[op], "reganode");
19754 regnode_offset ptr = ret;
19756 PERL_ARGS_ASSERT_REGANODE;
19758 /* ANYOF are special cased to allow non-length 1 args */
19759 assert(regarglen[op] == 1);
19761 FILL_ADVANCE_NODE_ARG(ptr, op, arg);
19766 STATIC regnode_offset
19767 S_reg2Lanode(pTHX_ RExC_state_t *pRExC_state, const U8 op, const U32 arg1, const I32 arg2)
19769 /* emit a node with U32 and I32 arguments */
19771 const regnode_offset ret = regnode_guts(pRExC_state, op, regarglen[op], "reg2Lanode");
19772 regnode_offset ptr = ret;
19774 PERL_ARGS_ASSERT_REG2LANODE;
19776 assert(regarglen[op] == 2);
19778 FILL_ADVANCE_NODE_2L_ARG(ptr, op, arg1, arg2);
19784 - reginsert - insert an operator in front of already-emitted operand
19786 * That means that on exit 'operand' is the offset of the newly inserted
19787 * operator, and the original operand has been relocated.
19789 * IMPORTANT NOTE - it is the *callers* responsibility to correctly
19790 * set up NEXT_OFF() of the inserted node if needed. Something like this:
19792 * reginsert(pRExC, OPFAIL, orig_emit, depth+1);
19793 * NEXT_OFF(orig_emit) = regarglen[OPFAIL] + NODE_STEP_REGNODE;
19795 * ALSO NOTE - FLAGS(newly-inserted-operator) will be set to 0 as well.
19798 S_reginsert(pTHX_ RExC_state_t *pRExC_state, const U8 op,
19799 const regnode_offset operand, const U32 depth)
19804 const int offset = regarglen[(U8)op];
19805 const int size = NODE_STEP_REGNODE + offset;
19806 GET_RE_DEBUG_FLAGS_DECL;
19808 PERL_ARGS_ASSERT_REGINSERT;
19809 PERL_UNUSED_CONTEXT;
19810 PERL_UNUSED_ARG(depth);
19811 /* (PL_regkind[(U8)op] == CURLY ? EXTRA_STEP_2ARGS : 0); */
19812 DEBUG_PARSE_FMT("inst"," - %s", PL_reg_name[op]);
19813 assert(!RExC_study_started); /* I believe we should never use reginsert once we have started
19814 studying. If this is wrong then we need to adjust RExC_recurse
19815 below like we do with RExC_open_parens/RExC_close_parens. */
19816 change_engine_size(pRExC_state, (Ptrdiff_t) size);
19817 src = REGNODE_p(RExC_emit);
19819 dst = REGNODE_p(RExC_emit);
19821 /* If we are in a "count the parentheses" pass, the numbers are unreliable,
19822 * and [perl #133871] shows this can lead to problems, so skip this
19823 * realignment of parens until a later pass when they are reliable */
19824 if (! IN_PARENS_PASS && RExC_open_parens) {
19826 /*DEBUG_PARSE_FMT("inst"," - %" IVdf, (IV)RExC_npar);*/
19827 /* remember that RExC_npar is rex->nparens + 1,
19828 * iow it is 1 more than the number of parens seen in
19829 * the pattern so far. */
19830 for ( paren=0 ; paren < RExC_npar ; paren++ ) {
19831 /* note, RExC_open_parens[0] is the start of the
19832 * regex, it can't move. RExC_close_parens[0] is the end
19833 * of the regex, it *can* move. */
19834 if ( paren && RExC_open_parens[paren] >= operand ) {
19835 /*DEBUG_PARSE_FMT("open"," - %d", size);*/
19836 RExC_open_parens[paren] += size;
19838 /*DEBUG_PARSE_FMT("open"," - %s","ok");*/
19840 if ( RExC_close_parens[paren] >= operand ) {
19841 /*DEBUG_PARSE_FMT("close"," - %d", size);*/
19842 RExC_close_parens[paren] += size;
19844 /*DEBUG_PARSE_FMT("close"," - %s","ok");*/
19849 RExC_end_op += size;
19851 while (src > REGNODE_p(operand)) {
19852 StructCopy(--src, --dst, regnode);
19853 #ifdef RE_TRACK_PATTERN_OFFSETS
19854 if (RExC_offsets) { /* MJD 20010112 */
19856 ("%s(%d): (op %s) %s copy %" UVuf " -> %" UVuf " (max %" UVuf ").\n",
19860 (UV)(REGNODE_OFFSET(dst)) > RExC_offsets[0]
19861 ? "Overwriting end of array!\n" : "OK",
19862 (UV)REGNODE_OFFSET(src),
19863 (UV)REGNODE_OFFSET(dst),
19864 (UV)RExC_offsets[0]));
19865 Set_Node_Offset_To_R(REGNODE_OFFSET(dst), Node_Offset(src));
19866 Set_Node_Length_To_R(REGNODE_OFFSET(dst), Node_Length(src));
19871 place = REGNODE_p(operand); /* Op node, where operand used to be. */
19872 #ifdef RE_TRACK_PATTERN_OFFSETS
19873 if (RExC_offsets) { /* MJD */
19875 ("%s(%d): (op %s) %s %" UVuf " <- %" UVuf " (max %" UVuf ").\n",
19879 (UV)REGNODE_OFFSET(place) > RExC_offsets[0]
19880 ? "Overwriting end of array!\n" : "OK",
19881 (UV)REGNODE_OFFSET(place),
19882 (UV)(RExC_parse - RExC_start),
19883 (UV)RExC_offsets[0]));
19884 Set_Node_Offset(place, RExC_parse);
19885 Set_Node_Length(place, 1);
19888 src = NEXTOPER(place);
19890 FILL_NODE(operand, op);
19892 /* Zero out any arguments in the new node */
19893 Zero(src, offset, regnode);
19897 - regtail - set the next-pointer at the end of a node chain of p to val. If
19898 that value won't fit in the space available, instead returns FALSE.
19899 (Except asserts if we can't fit in the largest space the regex
19900 engine is designed for.)
19901 - SEE ALSO: regtail_study
19904 S_regtail(pTHX_ RExC_state_t * pRExC_state,
19905 const regnode_offset p,
19906 const regnode_offset val,
19909 regnode_offset scan;
19910 GET_RE_DEBUG_FLAGS_DECL;
19912 PERL_ARGS_ASSERT_REGTAIL;
19914 PERL_UNUSED_ARG(depth);
19917 /* Find last node. */
19918 scan = (regnode_offset) p;
19920 regnode * const temp = regnext(REGNODE_p(scan));
19922 DEBUG_PARSE_MSG((scan==p ? "tail" : ""));
19923 regprop(RExC_rx, RExC_mysv, REGNODE_p(scan), NULL, pRExC_state);
19924 Perl_re_printf( aTHX_ "~ %s (%d) %s %s\n",
19925 SvPV_nolen_const(RExC_mysv), scan,
19926 (temp == NULL ? "->" : ""),
19927 (temp == NULL ? PL_reg_name[OP(REGNODE_p(val))] : "")
19932 scan = REGNODE_OFFSET(temp);
19935 if (reg_off_by_arg[OP(REGNODE_p(scan))]) {
19936 assert((UV) (val - scan) <= U32_MAX);
19937 ARG_SET(REGNODE_p(scan), val - scan);
19940 if (val - scan > U16_MAX) {
19941 /* Populate this with something that won't loop and will likely
19942 * lead to a crash if the caller ignores the failure return, and
19943 * execution continues */
19944 NEXT_OFF(REGNODE_p(scan)) = U16_MAX;
19947 NEXT_OFF(REGNODE_p(scan)) = val - scan;
19955 - regtail_study - set the next-pointer at the end of a node chain of p to val.
19956 - Look for optimizable sequences at the same time.
19957 - currently only looks for EXACT chains.
19959 This is experimental code. The idea is to use this routine to perform
19960 in place optimizations on branches and groups as they are constructed,
19961 with the long term intention of removing optimization from study_chunk so
19962 that it is purely analytical.
19964 Currently only used when in DEBUG mode. The macro REGTAIL_STUDY() is used
19965 to control which is which.
19967 This used to return a value that was ignored. It was a problem that it is
19968 #ifdef'd to be another function that didn't return a value. khw has changed it
19969 so both currently return a pass/fail return.
19972 /* TODO: All four parms should be const */
19975 S_regtail_study(pTHX_ RExC_state_t *pRExC_state, regnode_offset p,
19976 const regnode_offset val, U32 depth)
19978 regnode_offset scan;
19980 #ifdef EXPERIMENTAL_INPLACESCAN
19983 GET_RE_DEBUG_FLAGS_DECL;
19985 PERL_ARGS_ASSERT_REGTAIL_STUDY;
19988 /* Find last node. */
19992 regnode * const temp = regnext(REGNODE_p(scan));
19993 #ifdef EXPERIMENTAL_INPLACESCAN
19994 if (PL_regkind[OP(REGNODE_p(scan))] == EXACT) {
19995 bool unfolded_multi_char; /* Unexamined in this routine */
19996 if (join_exact(pRExC_state, scan, &min,
19997 &unfolded_multi_char, 1, REGNODE_p(val), depth+1))
19998 return TRUE; /* Was return EXACT */
20002 switch (OP(REGNODE_p(scan))) {
20007 case EXACTFU_S_EDGE:
20008 case EXACTFAA_NO_TRIE:
20011 case EXACTFU_ONLY8:
20015 if( exact == PSEUDO )
20016 exact= OP(REGNODE_p(scan));
20017 else if ( exact != OP(REGNODE_p(scan)) )
20026 DEBUG_PARSE_MSG((scan==p ? "tsdy" : ""));
20027 regprop(RExC_rx, RExC_mysv, REGNODE_p(scan), NULL, pRExC_state);
20028 Perl_re_printf( aTHX_ "~ %s (%d) -> %s\n",
20029 SvPV_nolen_const(RExC_mysv),
20031 PL_reg_name[exact]);
20035 scan = REGNODE_OFFSET(temp);
20038 DEBUG_PARSE_MSG("");
20039 regprop(RExC_rx, RExC_mysv, REGNODE_p(val), NULL, pRExC_state);
20040 Perl_re_printf( aTHX_
20041 "~ attach to %s (%" IVdf ") offset to %" IVdf "\n",
20042 SvPV_nolen_const(RExC_mysv),
20047 if (reg_off_by_arg[OP(REGNODE_p(scan))]) {
20048 assert((UV) (val - scan) <= U32_MAX);
20049 ARG_SET(REGNODE_p(scan), val - scan);
20052 if (val - scan > U16_MAX) {
20053 /* Populate this with something that won't loop and will likely
20054 * lead to a crash if the caller ignores the failure return, and
20055 * execution continues */
20056 NEXT_OFF(REGNODE_p(scan)) = U16_MAX;
20059 NEXT_OFF(REGNODE_p(scan)) = val - scan;
20062 return TRUE; /* Was 'return exact' */
20067 S_get_ANYOFM_contents(pTHX_ const regnode * n) {
20069 /* Returns an inversion list of all the code points matched by the
20070 * ANYOFM/NANYOFM node 'n' */
20072 SV * cp_list = _new_invlist(-1);
20073 const U8 lowest = (U8) ARG(n);
20076 U8 needed = 1U << PL_bitcount[ (U8) ~ FLAGS(n)];
20078 PERL_ARGS_ASSERT_GET_ANYOFM_CONTENTS;
20080 /* Starting with the lowest code point, any code point that ANDed with the
20081 * mask yields the lowest code point is in the set */
20082 for (i = lowest; i <= 0xFF; i++) {
20083 if ((i & FLAGS(n)) == ARG(n)) {
20084 cp_list = add_cp_to_invlist(cp_list, i);
20087 /* We know how many code points (a power of two) that are in the
20088 * set. No use looking once we've got that number */
20089 if (count >= needed) break;
20093 if (OP(n) == NANYOFM) {
20094 _invlist_invert(cp_list);
20100 - regdump - dump a regexp onto Perl_debug_log in vaguely comprehensible form
20105 S_regdump_intflags(pTHX_ const char *lead, const U32 flags)
20110 ASSUME(REG_INTFLAGS_NAME_SIZE <= sizeof(flags)*8);
20112 for (bit=0; bit<REG_INTFLAGS_NAME_SIZE; bit++) {
20113 if (flags & (1<<bit)) {
20114 if (!set++ && lead)
20115 Perl_re_printf( aTHX_ "%s", lead);
20116 Perl_re_printf( aTHX_ "%s ", PL_reg_intflags_name[bit]);
20121 Perl_re_printf( aTHX_ "\n");
20123 Perl_re_printf( aTHX_ "%s[none-set]\n", lead);
20128 S_regdump_extflags(pTHX_ const char *lead, const U32 flags)
20134 ASSUME(REG_EXTFLAGS_NAME_SIZE <= sizeof(flags)*8);
20136 for (bit=0; bit<REG_EXTFLAGS_NAME_SIZE; bit++) {
20137 if (flags & (1<<bit)) {
20138 if ((1<<bit) & RXf_PMf_CHARSET) { /* Output separately, below */
20141 if (!set++ && lead)
20142 Perl_re_printf( aTHX_ "%s", lead);
20143 Perl_re_printf( aTHX_ "%s ", PL_reg_extflags_name[bit]);
20146 if ((cs = get_regex_charset(flags)) != REGEX_DEPENDS_CHARSET) {
20147 if (!set++ && lead) {
20148 Perl_re_printf( aTHX_ "%s", lead);
20151 case REGEX_UNICODE_CHARSET:
20152 Perl_re_printf( aTHX_ "UNICODE");
20154 case REGEX_LOCALE_CHARSET:
20155 Perl_re_printf( aTHX_ "LOCALE");
20157 case REGEX_ASCII_RESTRICTED_CHARSET:
20158 Perl_re_printf( aTHX_ "ASCII-RESTRICTED");
20160 case REGEX_ASCII_MORE_RESTRICTED_CHARSET:
20161 Perl_re_printf( aTHX_ "ASCII-MORE_RESTRICTED");
20164 Perl_re_printf( aTHX_ "UNKNOWN CHARACTER SET");
20170 Perl_re_printf( aTHX_ "\n");
20172 Perl_re_printf( aTHX_ "%s[none-set]\n", lead);
20178 Perl_regdump(pTHX_ const regexp *r)
20182 SV * const sv = sv_newmortal();
20183 SV *dsv= sv_newmortal();
20184 RXi_GET_DECL(r, ri);
20185 GET_RE_DEBUG_FLAGS_DECL;
20187 PERL_ARGS_ASSERT_REGDUMP;
20189 (void)dumpuntil(r, ri->program, ri->program + 1, NULL, NULL, sv, 0, 0);
20191 /* Header fields of interest. */
20192 for (i = 0; i < 2; i++) {
20193 if (r->substrs->data[i].substr) {
20194 RE_PV_QUOTED_DECL(s, 0, dsv,
20195 SvPVX_const(r->substrs->data[i].substr),
20196 RE_SV_DUMPLEN(r->substrs->data[i].substr),
20197 PL_dump_re_max_len);
20198 Perl_re_printf( aTHX_
20199 "%s %s%s at %" IVdf "..%" UVuf " ",
20200 i ? "floating" : "anchored",
20202 RE_SV_TAIL(r->substrs->data[i].substr),
20203 (IV)r->substrs->data[i].min_offset,
20204 (UV)r->substrs->data[i].max_offset);
20206 else if (r->substrs->data[i].utf8_substr) {
20207 RE_PV_QUOTED_DECL(s, 1, dsv,
20208 SvPVX_const(r->substrs->data[i].utf8_substr),
20209 RE_SV_DUMPLEN(r->substrs->data[i].utf8_substr),
20211 Perl_re_printf( aTHX_
20212 "%s utf8 %s%s at %" IVdf "..%" UVuf " ",
20213 i ? "floating" : "anchored",
20215 RE_SV_TAIL(r->substrs->data[i].utf8_substr),
20216 (IV)r->substrs->data[i].min_offset,
20217 (UV)r->substrs->data[i].max_offset);
20221 if (r->check_substr || r->check_utf8)
20222 Perl_re_printf( aTHX_
20224 ( r->check_substr == r->substrs->data[1].substr
20225 && r->check_utf8 == r->substrs->data[1].utf8_substr
20226 ? "(checking floating" : "(checking anchored"));
20227 if (r->intflags & PREGf_NOSCAN)
20228 Perl_re_printf( aTHX_ " noscan");
20229 if (r->extflags & RXf_CHECK_ALL)
20230 Perl_re_printf( aTHX_ " isall");
20231 if (r->check_substr || r->check_utf8)
20232 Perl_re_printf( aTHX_ ") ");
20234 if (ri->regstclass) {
20235 regprop(r, sv, ri->regstclass, NULL, NULL);
20236 Perl_re_printf( aTHX_ "stclass %s ", SvPVX_const(sv));
20238 if (r->intflags & PREGf_ANCH) {
20239 Perl_re_printf( aTHX_ "anchored");
20240 if (r->intflags & PREGf_ANCH_MBOL)
20241 Perl_re_printf( aTHX_ "(MBOL)");
20242 if (r->intflags & PREGf_ANCH_SBOL)
20243 Perl_re_printf( aTHX_ "(SBOL)");
20244 if (r->intflags & PREGf_ANCH_GPOS)
20245 Perl_re_printf( aTHX_ "(GPOS)");
20246 Perl_re_printf( aTHX_ " ");
20248 if (r->intflags & PREGf_GPOS_SEEN)
20249 Perl_re_printf( aTHX_ "GPOS:%" UVuf " ", (UV)r->gofs);
20250 if (r->intflags & PREGf_SKIP)
20251 Perl_re_printf( aTHX_ "plus ");
20252 if (r->intflags & PREGf_IMPLICIT)
20253 Perl_re_printf( aTHX_ "implicit ");
20254 Perl_re_printf( aTHX_ "minlen %" IVdf " ", (IV)r->minlen);
20255 if (r->extflags & RXf_EVAL_SEEN)
20256 Perl_re_printf( aTHX_ "with eval ");
20257 Perl_re_printf( aTHX_ "\n");
20259 regdump_extflags("r->extflags: ", r->extflags);
20260 regdump_intflags("r->intflags: ", r->intflags);
20263 PERL_ARGS_ASSERT_REGDUMP;
20264 PERL_UNUSED_CONTEXT;
20265 PERL_UNUSED_ARG(r);
20266 #endif /* DEBUGGING */
20269 /* Should be synchronized with ANYOF_ #defines in regcomp.h */
20272 # if _CC_WORDCHAR != 0 || _CC_DIGIT != 1 || _CC_ALPHA != 2 \
20273 || _CC_LOWER != 3 || _CC_UPPER != 4 || _CC_PUNCT != 5 \
20274 || _CC_PRINT != 6 || _CC_ALPHANUMERIC != 7 || _CC_GRAPH != 8 \
20275 || _CC_CASED != 9 || _CC_SPACE != 10 || _CC_BLANK != 11 \
20276 || _CC_XDIGIT != 12 || _CC_CNTRL != 13 || _CC_ASCII != 14 \
20277 || _CC_VERTSPACE != 15
20278 # error Need to adjust order of anyofs[]
20280 static const char * const anyofs[] = {
20317 - regprop - printable representation of opcode, with run time support
20321 Perl_regprop(pTHX_ const regexp *prog, SV *sv, const regnode *o, const regmatch_info *reginfo, const RExC_state_t *pRExC_state)
20326 RXi_GET_DECL(prog, progi);
20327 GET_RE_DEBUG_FLAGS_DECL;
20329 PERL_ARGS_ASSERT_REGPROP;
20333 if (OP(o) > REGNODE_MAX) { /* regnode.type is unsigned */
20334 if (pRExC_state) { /* This gives more info, if we have it */
20335 FAIL3("panic: corrupted regexp opcode %d > %d",
20336 (int)OP(o), (int)REGNODE_MAX);
20339 Perl_croak(aTHX_ "panic: corrupted regexp opcode %d > %d",
20340 (int)OP(o), (int)REGNODE_MAX);
20343 sv_catpv(sv, PL_reg_name[OP(o)]); /* Take off const! */
20345 k = PL_regkind[OP(o)];
20348 sv_catpvs(sv, " ");
20349 /* Using is_utf8_string() (via PERL_PV_UNI_DETECT)
20350 * is a crude hack but it may be the best for now since
20351 * we have no flag "this EXACTish node was UTF-8"
20353 pv_pretty(sv, STRING(o), STR_LEN(o), PL_dump_re_max_len,
20354 PL_colors[0], PL_colors[1],
20355 PERL_PV_ESCAPE_UNI_DETECT |
20356 PERL_PV_ESCAPE_NONASCII |
20357 PERL_PV_PRETTY_ELLIPSES |
20358 PERL_PV_PRETTY_LTGT |
20359 PERL_PV_PRETTY_NOCLEAR
20361 } else if (k == TRIE) {
20362 /* print the details of the trie in dumpuntil instead, as
20363 * progi->data isn't available here */
20364 const char op = OP(o);
20365 const U32 n = ARG(o);
20366 const reg_ac_data * const ac = IS_TRIE_AC(op) ?
20367 (reg_ac_data *)progi->data->data[n] :
20369 const reg_trie_data * const trie
20370 = (reg_trie_data*)progi->data->data[!IS_TRIE_AC(op) ? n : ac->trie];
20372 Perl_sv_catpvf(aTHX_ sv, "-%s", PL_reg_name[o->flags]);
20373 DEBUG_TRIE_COMPILE_r({
20375 sv_catpvs(sv, "(JUMP)");
20376 Perl_sv_catpvf(aTHX_ sv,
20377 "<S:%" UVuf "/%" IVdf " W:%" UVuf " L:%" UVuf "/%" UVuf " C:%" UVuf "/%" UVuf ">",
20378 (UV)trie->startstate,
20379 (IV)trie->statecount-1, /* -1 because of the unused 0 element */
20380 (UV)trie->wordcount,
20383 (UV)TRIE_CHARCOUNT(trie),
20384 (UV)trie->uniquecharcount
20387 if ( IS_ANYOF_TRIE(op) || trie->bitmap ) {
20388 sv_catpvs(sv, "[");
20389 (void) put_charclass_bitmap_innards(sv,
20390 ((IS_ANYOF_TRIE(op))
20392 : TRIE_BITMAP(trie)),
20398 sv_catpvs(sv, "]");
20400 } else if (k == CURLY) {
20401 U32 lo = ARG1(o), hi = ARG2(o);
20402 if (OP(o) == CURLYM || OP(o) == CURLYN || OP(o) == CURLYX)
20403 Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* Parenth number */
20404 Perl_sv_catpvf(aTHX_ sv, "{%u,", (unsigned) lo);
20405 if (hi == REG_INFTY)
20406 sv_catpvs(sv, "INFTY");
20408 Perl_sv_catpvf(aTHX_ sv, "%u", (unsigned) hi);
20409 sv_catpvs(sv, "}");
20411 else if (k == WHILEM && o->flags) /* Ordinal/of */
20412 Perl_sv_catpvf(aTHX_ sv, "[%d/%d]", o->flags & 0xf, o->flags>>4);
20413 else if (k == REF || k == OPEN || k == CLOSE
20414 || k == GROUPP || OP(o)==ACCEPT)
20416 AV *name_list= NULL;
20417 U32 parno= OP(o) == ACCEPT ? (U32)ARG2L(o) : ARG(o);
20418 Perl_sv_catpvf(aTHX_ sv, "%" UVuf, (UV)parno); /* Parenth number */
20419 if ( RXp_PAREN_NAMES(prog) ) {
20420 name_list= MUTABLE_AV(progi->data->data[progi->name_list_idx]);
20421 } else if ( pRExC_state ) {
20422 name_list= RExC_paren_name_list;
20425 if ( k != REF || (OP(o) < REFN)) {
20426 SV **name= av_fetch(name_list, parno, 0 );
20428 Perl_sv_catpvf(aTHX_ sv, " '%" SVf "'", SVfARG(*name));
20431 SV *sv_dat= MUTABLE_SV(progi->data->data[ parno ]);
20432 I32 *nums=(I32*)SvPVX(sv_dat);
20433 SV **name= av_fetch(name_list, nums[0], 0 );
20436 for ( n=0; n<SvIVX(sv_dat); n++ ) {
20437 Perl_sv_catpvf(aTHX_ sv, "%s%" IVdf,
20438 (n ? "," : ""), (IV)nums[n]);
20440 Perl_sv_catpvf(aTHX_ sv, " '%" SVf "'", SVfARG(*name));
20444 if ( k == REF && reginfo) {
20445 U32 n = ARG(o); /* which paren pair */
20446 I32 ln = prog->offs[n].start;
20447 if (prog->lastparen < n || ln == -1 || prog->offs[n].end == -1)
20448 Perl_sv_catpvf(aTHX_ sv, ": FAIL");
20449 else if (ln == prog->offs[n].end)
20450 Perl_sv_catpvf(aTHX_ sv, ": ACCEPT - EMPTY STRING");
20452 const char *s = reginfo->strbeg + ln;
20453 Perl_sv_catpvf(aTHX_ sv, ": ");
20454 Perl_pv_pretty( aTHX_ sv, s, prog->offs[n].end - prog->offs[n].start, 32, 0, 0,
20455 PERL_PV_ESCAPE_UNI_DETECT|PERL_PV_PRETTY_NOCLEAR|PERL_PV_PRETTY_ELLIPSES|PERL_PV_PRETTY_QUOTE );
20458 } else if (k == GOSUB) {
20459 AV *name_list= NULL;
20460 if ( RXp_PAREN_NAMES(prog) ) {
20461 name_list= MUTABLE_AV(progi->data->data[progi->name_list_idx]);
20462 } else if ( pRExC_state ) {
20463 name_list= RExC_paren_name_list;
20466 /* Paren and offset */
20467 Perl_sv_catpvf(aTHX_ sv, "%d[%+d:%d]", (int)ARG(o),(int)ARG2L(o),
20468 (int)((o + (int)ARG2L(o)) - progi->program) );
20470 SV **name= av_fetch(name_list, ARG(o), 0 );
20472 Perl_sv_catpvf(aTHX_ sv, " '%" SVf "'", SVfARG(*name));
20475 else if (k == LOGICAL)
20476 /* 2: embedded, otherwise 1 */
20477 Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags);
20478 else if (k == ANYOF) {
20479 const U8 flags = inRANGE(OP(o), ANYOFH, ANYOFHr)
20482 bool do_sep = FALSE; /* Do we need to separate various components of
20484 /* Set if there is still an unresolved user-defined property */
20485 SV *unresolved = NULL;
20487 /* Things that are ignored except when the runtime locale is UTF-8 */
20488 SV *only_utf8_locale_invlist = NULL;
20490 /* Code points that don't fit in the bitmap */
20491 SV *nonbitmap_invlist = NULL;
20493 /* And things that aren't in the bitmap, but are small enough to be */
20494 SV* bitmap_range_not_in_bitmap = NULL;
20496 const bool inverted = flags & ANYOF_INVERT;
20498 if (OP(o) == ANYOFL || OP(o) == ANYOFPOSIXL) {
20499 if (ANYOFL_UTF8_LOCALE_REQD(flags)) {
20500 sv_catpvs(sv, "{utf8-locale-reqd}");
20502 if (flags & ANYOFL_FOLD) {
20503 sv_catpvs(sv, "{i}");
20507 /* If there is stuff outside the bitmap, get it */
20508 if (ARG(o) != ANYOF_ONLY_HAS_BITMAP) {
20509 (void) _get_regclass_nonbitmap_data(prog, o, FALSE,
20511 &only_utf8_locale_invlist,
20512 &nonbitmap_invlist);
20513 /* The non-bitmap data may contain stuff that could fit in the
20514 * bitmap. This could come from a user-defined property being
20515 * finally resolved when this call was done; or much more likely
20516 * because there are matches that require UTF-8 to be valid, and so
20517 * aren't in the bitmap. This is teased apart later */
20518 _invlist_intersection(nonbitmap_invlist,
20520 &bitmap_range_not_in_bitmap);
20521 /* Leave just the things that don't fit into the bitmap */
20522 _invlist_subtract(nonbitmap_invlist,
20524 &nonbitmap_invlist);
20527 /* Obey this flag to add all above-the-bitmap code points */
20528 if (flags & ANYOF_MATCHES_ALL_ABOVE_BITMAP) {
20529 nonbitmap_invlist = _add_range_to_invlist(nonbitmap_invlist,
20530 NUM_ANYOF_CODE_POINTS,
20534 /* Ready to start outputting. First, the initial left bracket */
20535 Perl_sv_catpvf(aTHX_ sv, "[%s", PL_colors[0]);
20537 if (! inRANGE(OP(o), ANYOFH, ANYOFHr)) {
20538 /* Then all the things that could fit in the bitmap */
20539 do_sep = put_charclass_bitmap_innards(sv,
20541 bitmap_range_not_in_bitmap,
20542 only_utf8_locale_invlist,
20545 /* Can't try inverting for a
20546 * better display if there
20547 * are things that haven't
20549 unresolved != NULL);
20550 SvREFCNT_dec(bitmap_range_not_in_bitmap);
20552 /* If there are user-defined properties which haven't been defined
20553 * yet, output them. If the result is not to be inverted, it is
20554 * clearest to output them in a separate [] from the bitmap range
20555 * stuff. If the result is to be complemented, we have to show
20556 * everything in one [], as the inversion applies to the whole
20557 * thing. Use {braces} to separate them from anything in the
20558 * bitmap and anything above the bitmap. */
20561 if (! do_sep) { /* If didn't output anything in the bitmap
20563 sv_catpvs(sv, "^");
20565 sv_catpvs(sv, "{");
20568 Perl_sv_catpvf(aTHX_ sv,"%s][%s", PL_colors[1],
20571 sv_catsv(sv, unresolved);
20573 sv_catpvs(sv, "}");
20575 do_sep = ! inverted;
20579 /* And, finally, add the above-the-bitmap stuff */
20580 if (nonbitmap_invlist && _invlist_len(nonbitmap_invlist)) {
20583 /* See if truncation size is overridden */
20584 const STRLEN dump_len = (PL_dump_re_max_len > 256)
20585 ? PL_dump_re_max_len
20588 /* This is output in a separate [] */
20590 Perl_sv_catpvf(aTHX_ sv,"%s][%s", PL_colors[1], PL_colors[0]);
20593 /* And, for easy of understanding, it is shown in the
20594 * uncomplemented form if possible. The one exception being if
20595 * there are unresolved items, where the inversion has to be
20596 * delayed until runtime */
20597 if (inverted && ! unresolved) {
20598 _invlist_invert(nonbitmap_invlist);
20599 _invlist_subtract(nonbitmap_invlist, PL_InBitmap, &nonbitmap_invlist);
20602 contents = invlist_contents(nonbitmap_invlist,
20603 FALSE /* output suitable for catsv */
20606 /* If the output is shorter than the permissible maximum, just do it. */
20607 if (SvCUR(contents) <= dump_len) {
20608 sv_catsv(sv, contents);
20611 const char * contents_string = SvPVX(contents);
20612 STRLEN i = dump_len;
20614 /* Otherwise, start at the permissible max and work back to the
20615 * first break possibility */
20616 while (i > 0 && contents_string[i] != ' ') {
20619 if (i == 0) { /* Fail-safe. Use the max if we couldn't
20620 find a legal break */
20624 sv_catpvn(sv, contents_string, i);
20625 sv_catpvs(sv, "...");
20628 SvREFCNT_dec_NN(contents);
20629 SvREFCNT_dec_NN(nonbitmap_invlist);
20632 /* And finally the matching, closing ']' */
20633 Perl_sv_catpvf(aTHX_ sv, "%s]", PL_colors[1]);
20635 if (inRANGE(OP(o), ANYOFH, ANYOFHr)) {
20636 U8 lowest = (OP(o) != ANYOFHr)
20638 : LOWEST_ANYOF_HRx_BYTE(FLAGS(o));
20639 U8 highest = (OP(o) == ANYOFHb)
20643 : HIGHEST_ANYOF_HRx_BYTE(FLAGS(o));
20644 Perl_sv_catpvf(aTHX_ sv, " (First UTF-8 byte=%02X", lowest);
20645 if (lowest != highest) {
20646 Perl_sv_catpvf(aTHX_ sv, "-%02X", highest);
20648 Perl_sv_catpvf(aTHX_ sv, ")");
20651 SvREFCNT_dec(unresolved);
20653 else if (k == ANYOFM) {
20654 SV * cp_list = get_ANYOFM_contents(o);
20656 Perl_sv_catpvf(aTHX_ sv, "[%s", PL_colors[0]);
20657 if (OP(o) == NANYOFM) {
20658 _invlist_invert(cp_list);
20661 put_charclass_bitmap_innards(sv, NULL, cp_list, NULL, NULL, TRUE);
20662 Perl_sv_catpvf(aTHX_ sv, "%s]", PL_colors[1]);
20664 SvREFCNT_dec(cp_list);
20666 else if (k == POSIXD || k == NPOSIXD) {
20667 U8 index = FLAGS(o) * 2;
20668 if (index < C_ARRAY_LENGTH(anyofs)) {
20669 if (*anyofs[index] != '[') {
20670 sv_catpvs(sv, "[");
20672 sv_catpv(sv, anyofs[index]);
20673 if (*anyofs[index] != '[') {
20674 sv_catpvs(sv, "]");
20678 Perl_sv_catpvf(aTHX_ sv, "[illegal type=%d])", index);
20681 else if (k == BOUND || k == NBOUND) {
20682 /* Must be synced with order of 'bound_type' in regcomp.h */
20683 const char * const bounds[] = {
20684 "", /* Traditional */
20690 assert(FLAGS(o) < C_ARRAY_LENGTH(bounds));
20691 sv_catpv(sv, bounds[FLAGS(o)]);
20693 else if (k == BRANCHJ && (OP(o) == UNLESSM || OP(o) == IFMATCH)) {
20694 Perl_sv_catpvf(aTHX_ sv, "[%d", -(o->flags));
20696 Perl_sv_catpvf(aTHX_ sv, "..-%d", o->flags - o->next_off);
20698 Perl_sv_catpvf(aTHX_ sv, "]");
20700 else if (OP(o) == SBOL)
20701 Perl_sv_catpvf(aTHX_ sv, " /%s/", o->flags ? "\\A" : "^");
20703 /* add on the verb argument if there is one */
20704 if ( ( k == VERB || OP(o) == ACCEPT || OP(o) == OPFAIL ) && o->flags) {
20706 Perl_sv_catpvf(aTHX_ sv, ":%" SVf,
20707 SVfARG((MUTABLE_SV(progi->data->data[ ARG( o ) ]))));
20709 sv_catpvs(sv, ":NULL");
20712 PERL_UNUSED_CONTEXT;
20713 PERL_UNUSED_ARG(sv);
20714 PERL_UNUSED_ARG(o);
20715 PERL_UNUSED_ARG(prog);
20716 PERL_UNUSED_ARG(reginfo);
20717 PERL_UNUSED_ARG(pRExC_state);
20718 #endif /* DEBUGGING */
20724 Perl_re_intuit_string(pTHX_ REGEXP * const r)
20725 { /* Assume that RE_INTUIT is set */
20726 struct regexp *const prog = ReANY(r);
20727 GET_RE_DEBUG_FLAGS_DECL;
20729 PERL_ARGS_ASSERT_RE_INTUIT_STRING;
20730 PERL_UNUSED_CONTEXT;
20734 const char * const s = SvPV_nolen_const(RX_UTF8(r)
20735 ? prog->check_utf8 : prog->check_substr);
20737 if (!PL_colorset) reginitcolors();
20738 Perl_re_printf( aTHX_
20739 "%sUsing REx %ssubstr:%s \"%s%.60s%s%s\"\n",
20741 RX_UTF8(r) ? "utf8 " : "",
20742 PL_colors[5], PL_colors[0],
20745 (strlen(s) > PL_dump_re_max_len ? "..." : ""));
20748 /* use UTF8 check substring if regexp pattern itself is in UTF8 */
20749 return RX_UTF8(r) ? prog->check_utf8 : prog->check_substr;
20755 handles refcounting and freeing the perl core regexp structure. When
20756 it is necessary to actually free the structure the first thing it
20757 does is call the 'free' method of the regexp_engine associated to
20758 the regexp, allowing the handling of the void *pprivate; member
20759 first. (This routine is not overridable by extensions, which is why
20760 the extensions free is called first.)
20762 See regdupe and regdupe_internal if you change anything here.
20764 #ifndef PERL_IN_XSUB_RE
20766 Perl_pregfree(pTHX_ REGEXP *r)
20772 Perl_pregfree2(pTHX_ REGEXP *rx)
20774 struct regexp *const r = ReANY(rx);
20775 GET_RE_DEBUG_FLAGS_DECL;
20777 PERL_ARGS_ASSERT_PREGFREE2;
20782 if (r->mother_re) {
20783 ReREFCNT_dec(r->mother_re);
20785 CALLREGFREE_PVT(rx); /* free the private data */
20786 SvREFCNT_dec(RXp_PAREN_NAMES(r));
20790 for (i = 0; i < 2; i++) {
20791 SvREFCNT_dec(r->substrs->data[i].substr);
20792 SvREFCNT_dec(r->substrs->data[i].utf8_substr);
20794 Safefree(r->substrs);
20796 RX_MATCH_COPY_FREE(rx);
20797 #ifdef PERL_ANY_COW
20798 SvREFCNT_dec(r->saved_copy);
20801 SvREFCNT_dec(r->qr_anoncv);
20802 if (r->recurse_locinput)
20803 Safefree(r->recurse_locinput);
20809 Copy ssv to dsv, both of which should of type SVt_REGEXP or SVt_PVLV,
20810 except that dsv will be created if NULL.
20812 This function is used in two main ways. First to implement
20813 $r = qr/....; $s = $$r;
20815 Secondly, it is used as a hacky workaround to the structural issue of
20817 being stored in the regexp structure which is in turn stored in
20818 PL_curpm/PL_reg_curpm. The problem is that due to qr// the pattern
20819 could be PL_curpm in multiple contexts, and could require multiple
20820 result sets being associated with the pattern simultaneously, such
20821 as when doing a recursive match with (??{$qr})
20823 The solution is to make a lightweight copy of the regexp structure
20824 when a qr// is returned from the code executed by (??{$qr}) this
20825 lightweight copy doesn't actually own any of its data except for
20826 the starp/end and the actual regexp structure itself.
20832 Perl_reg_temp_copy(pTHX_ REGEXP *dsv, REGEXP *ssv)
20834 struct regexp *drx;
20835 struct regexp *const srx = ReANY(ssv);
20836 const bool islv = dsv && SvTYPE(dsv) == SVt_PVLV;
20838 PERL_ARGS_ASSERT_REG_TEMP_COPY;
20841 dsv = (REGEXP*) newSV_type(SVt_REGEXP);
20843 assert(SvTYPE(dsv) == SVt_REGEXP || (SvTYPE(dsv) == SVt_PVLV));
20845 /* our only valid caller, sv_setsv_flags(), should have done
20846 * a SV_CHECK_THINKFIRST_COW_DROP() by now */
20847 assert(!SvOOK(dsv));
20848 assert(!SvIsCOW(dsv));
20849 assert(!SvROK(dsv));
20851 if (SvPVX_const(dsv)) {
20853 Safefree(SvPVX(dsv));
20858 SvOK_off((SV *)dsv);
20861 /* For PVLVs, the head (sv_any) points to an XPVLV, while
20862 * the LV's xpvlenu_rx will point to a regexp body, which
20863 * we allocate here */
20864 REGEXP *temp = (REGEXP *)newSV_type(SVt_REGEXP);
20865 assert(!SvPVX(dsv));
20866 ((XPV*)SvANY(dsv))->xpv_len_u.xpvlenu_rx = temp->sv_any;
20867 temp->sv_any = NULL;
20868 SvFLAGS(temp) = (SvFLAGS(temp) & ~SVTYPEMASK) | SVt_NULL;
20869 SvREFCNT_dec_NN(temp);
20870 /* SvCUR still resides in the xpvlv struct, so the regexp copy-
20871 ing below will not set it. */
20872 SvCUR_set(dsv, SvCUR(ssv));
20875 /* This ensures that SvTHINKFIRST(sv) is true, and hence that
20876 sv_force_normal(sv) is called. */
20880 SvFLAGS(dsv) |= SvFLAGS(ssv) & (SVf_POK|SVp_POK|SVf_UTF8);
20881 SvPV_set(dsv, RX_WRAPPED(ssv));
20882 /* We share the same string buffer as the original regexp, on which we
20883 hold a reference count, incremented when mother_re is set below.
20884 The string pointer is copied here, being part of the regexp struct.
20886 memcpy(&(drx->xpv_cur), &(srx->xpv_cur),
20887 sizeof(regexp) - STRUCT_OFFSET(regexp, xpv_cur));
20891 const I32 npar = srx->nparens+1;
20892 Newx(drx->offs, npar, regexp_paren_pair);
20893 Copy(srx->offs, drx->offs, npar, regexp_paren_pair);
20895 if (srx->substrs) {
20897 Newx(drx->substrs, 1, struct reg_substr_data);
20898 StructCopy(srx->substrs, drx->substrs, struct reg_substr_data);
20900 for (i = 0; i < 2; i++) {
20901 SvREFCNT_inc_void(drx->substrs->data[i].substr);
20902 SvREFCNT_inc_void(drx->substrs->data[i].utf8_substr);
20905 /* check_substr and check_utf8, if non-NULL, point to either their
20906 anchored or float namesakes, and don't hold a second reference. */
20908 RX_MATCH_COPIED_off(dsv);
20909 #ifdef PERL_ANY_COW
20910 drx->saved_copy = NULL;
20912 drx->mother_re = ReREFCNT_inc(srx->mother_re ? srx->mother_re : ssv);
20913 SvREFCNT_inc_void(drx->qr_anoncv);
20914 if (srx->recurse_locinput)
20915 Newx(drx->recurse_locinput, srx->nparens + 1, char *);
20922 /* regfree_internal()
20924 Free the private data in a regexp. This is overloadable by
20925 extensions. Perl takes care of the regexp structure in pregfree(),
20926 this covers the *pprivate pointer which technically perl doesn't
20927 know about, however of course we have to handle the
20928 regexp_internal structure when no extension is in use.
20930 Note this is called before freeing anything in the regexp
20935 Perl_regfree_internal(pTHX_ REGEXP * const rx)
20937 struct regexp *const r = ReANY(rx);
20938 RXi_GET_DECL(r, ri);
20939 GET_RE_DEBUG_FLAGS_DECL;
20941 PERL_ARGS_ASSERT_REGFREE_INTERNAL;
20951 SV *dsv= sv_newmortal();
20952 RE_PV_QUOTED_DECL(s, RX_UTF8(rx),
20953 dsv, RX_PRECOMP(rx), RX_PRELEN(rx), PL_dump_re_max_len);
20954 Perl_re_printf( aTHX_ "%sFreeing REx:%s %s\n",
20955 PL_colors[4], PL_colors[5], s);
20959 #ifdef RE_TRACK_PATTERN_OFFSETS
20961 Safefree(ri->u.offsets); /* 20010421 MJD */
20963 if (ri->code_blocks)
20964 S_free_codeblocks(aTHX_ ri->code_blocks);
20967 int n = ri->data->count;
20970 /* If you add a ->what type here, update the comment in regcomp.h */
20971 switch (ri->data->what[n]) {
20977 SvREFCNT_dec(MUTABLE_SV(ri->data->data[n]));
20980 Safefree(ri->data->data[n]);
20986 { /* Aho Corasick add-on structure for a trie node.
20987 Used in stclass optimization only */
20989 reg_ac_data *aho=(reg_ac_data*)ri->data->data[n];
20990 #ifdef USE_ITHREADS
20994 refcount = --aho->refcount;
20997 PerlMemShared_free(aho->states);
20998 PerlMemShared_free(aho->fail);
20999 /* do this last!!!! */
21000 PerlMemShared_free(ri->data->data[n]);
21001 /* we should only ever get called once, so
21002 * assert as much, and also guard the free
21003 * which /might/ happen twice. At the least
21004 * it will make code anlyzers happy and it
21005 * doesn't cost much. - Yves */
21006 assert(ri->regstclass);
21007 if (ri->regstclass) {
21008 PerlMemShared_free(ri->regstclass);
21009 ri->regstclass = 0;
21016 /* trie structure. */
21018 reg_trie_data *trie=(reg_trie_data*)ri->data->data[n];
21019 #ifdef USE_ITHREADS
21023 refcount = --trie->refcount;
21026 PerlMemShared_free(trie->charmap);
21027 PerlMemShared_free(trie->states);
21028 PerlMemShared_free(trie->trans);
21030 PerlMemShared_free(trie->bitmap);
21032 PerlMemShared_free(trie->jump);
21033 PerlMemShared_free(trie->wordinfo);
21034 /* do this last!!!! */
21035 PerlMemShared_free(ri->data->data[n]);
21040 Perl_croak(aTHX_ "panic: regfree data code '%c'",
21041 ri->data->what[n]);
21044 Safefree(ri->data->what);
21045 Safefree(ri->data);
21051 #define av_dup_inc(s, t) MUTABLE_AV(sv_dup_inc((const SV *)s, t))
21052 #define hv_dup_inc(s, t) MUTABLE_HV(sv_dup_inc((const SV *)s, t))
21053 #define SAVEPVN(p, n) ((p) ? savepvn(p, n) : NULL)
21056 re_dup_guts - duplicate a regexp.
21058 This routine is expected to clone a given regexp structure. It is only
21059 compiled under USE_ITHREADS.
21061 After all of the core data stored in struct regexp is duplicated
21062 the regexp_engine.dupe method is used to copy any private data
21063 stored in the *pprivate pointer. This allows extensions to handle
21064 any duplication it needs to do.
21066 See pregfree() and regfree_internal() if you change anything here.
21068 #if defined(USE_ITHREADS)
21069 #ifndef PERL_IN_XSUB_RE
21071 Perl_re_dup_guts(pTHX_ const REGEXP *sstr, REGEXP *dstr, CLONE_PARAMS *param)
21075 const struct regexp *r = ReANY(sstr);
21076 struct regexp *ret = ReANY(dstr);
21078 PERL_ARGS_ASSERT_RE_DUP_GUTS;
21080 npar = r->nparens+1;
21081 Newx(ret->offs, npar, regexp_paren_pair);
21082 Copy(r->offs, ret->offs, npar, regexp_paren_pair);
21084 if (ret->substrs) {
21085 /* Do it this way to avoid reading from *r after the StructCopy().
21086 That way, if any of the sv_dup_inc()s dislodge *r from the L1
21087 cache, it doesn't matter. */
21089 const bool anchored = r->check_substr
21090 ? r->check_substr == r->substrs->data[0].substr
21091 : r->check_utf8 == r->substrs->data[0].utf8_substr;
21092 Newx(ret->substrs, 1, struct reg_substr_data);
21093 StructCopy(r->substrs, ret->substrs, struct reg_substr_data);
21095 for (i = 0; i < 2; i++) {
21096 ret->substrs->data[i].substr =
21097 sv_dup_inc(ret->substrs->data[i].substr, param);
21098 ret->substrs->data[i].utf8_substr =
21099 sv_dup_inc(ret->substrs->data[i].utf8_substr, param);
21102 /* check_substr and check_utf8, if non-NULL, point to either their
21103 anchored or float namesakes, and don't hold a second reference. */
21105 if (ret->check_substr) {
21107 assert(r->check_utf8 == r->substrs->data[0].utf8_substr);
21109 ret->check_substr = ret->substrs->data[0].substr;
21110 ret->check_utf8 = ret->substrs->data[0].utf8_substr;
21112 assert(r->check_substr == r->substrs->data[1].substr);
21113 assert(r->check_utf8 == r->substrs->data[1].utf8_substr);
21115 ret->check_substr = ret->substrs->data[1].substr;
21116 ret->check_utf8 = ret->substrs->data[1].utf8_substr;
21118 } else if (ret->check_utf8) {
21120 ret->check_utf8 = ret->substrs->data[0].utf8_substr;
21122 ret->check_utf8 = ret->substrs->data[1].utf8_substr;
21127 RXp_PAREN_NAMES(ret) = hv_dup_inc(RXp_PAREN_NAMES(ret), param);
21128 ret->qr_anoncv = MUTABLE_CV(sv_dup_inc((const SV *)ret->qr_anoncv, param));
21129 if (r->recurse_locinput)
21130 Newx(ret->recurse_locinput, r->nparens + 1, char *);
21133 RXi_SET(ret, CALLREGDUPE_PVT(dstr, param));
21135 if (RX_MATCH_COPIED(dstr))
21136 ret->subbeg = SAVEPVN(ret->subbeg, ret->sublen);
21138 ret->subbeg = NULL;
21139 #ifdef PERL_ANY_COW
21140 ret->saved_copy = NULL;
21143 /* Whether mother_re be set or no, we need to copy the string. We
21144 cannot refrain from copying it when the storage points directly to
21145 our mother regexp, because that's
21146 1: a buffer in a different thread
21147 2: something we no longer hold a reference on
21148 so we need to copy it locally. */
21149 RX_WRAPPED(dstr) = SAVEPVN(RX_WRAPPED_const(sstr), SvCUR(sstr)+1);
21150 /* set malloced length to a non-zero value so it will be freed
21151 * (otherwise in combination with SVf_FAKE it looks like an alien
21152 * buffer). It doesn't have to be the actual malloced size, since it
21153 * should never be grown */
21154 SvLEN_set(dstr, SvCUR(sstr)+1);
21155 ret->mother_re = NULL;
21157 #endif /* PERL_IN_XSUB_RE */
21162 This is the internal complement to regdupe() which is used to copy
21163 the structure pointed to by the *pprivate pointer in the regexp.
21164 This is the core version of the extension overridable cloning hook.
21165 The regexp structure being duplicated will be copied by perl prior
21166 to this and will be provided as the regexp *r argument, however
21167 with the /old/ structures pprivate pointer value. Thus this routine
21168 may override any copying normally done by perl.
21170 It returns a pointer to the new regexp_internal structure.
21174 Perl_regdupe_internal(pTHX_ REGEXP * const rx, CLONE_PARAMS *param)
21177 struct regexp *const r = ReANY(rx);
21178 regexp_internal *reti;
21180 RXi_GET_DECL(r, ri);
21182 PERL_ARGS_ASSERT_REGDUPE_INTERNAL;
21186 Newxc(reti, sizeof(regexp_internal) + len*sizeof(regnode),
21187 char, regexp_internal);
21188 Copy(ri->program, reti->program, len+1, regnode);
21191 if (ri->code_blocks) {
21193 Newx(reti->code_blocks, 1, struct reg_code_blocks);
21194 Newx(reti->code_blocks->cb, ri->code_blocks->count,
21195 struct reg_code_block);
21196 Copy(ri->code_blocks->cb, reti->code_blocks->cb,
21197 ri->code_blocks->count, struct reg_code_block);
21198 for (n = 0; n < ri->code_blocks->count; n++)
21199 reti->code_blocks->cb[n].src_regex = (REGEXP*)
21200 sv_dup_inc((SV*)(ri->code_blocks->cb[n].src_regex), param);
21201 reti->code_blocks->count = ri->code_blocks->count;
21202 reti->code_blocks->refcnt = 1;
21205 reti->code_blocks = NULL;
21207 reti->regstclass = NULL;
21210 struct reg_data *d;
21211 const int count = ri->data->count;
21214 Newxc(d, sizeof(struct reg_data) + count*sizeof(void *),
21215 char, struct reg_data);
21216 Newx(d->what, count, U8);
21219 for (i = 0; i < count; i++) {
21220 d->what[i] = ri->data->what[i];
21221 switch (d->what[i]) {
21222 /* see also regcomp.h and regfree_internal() */
21223 case 'a': /* actually an AV, but the dup function is identical.
21224 values seem to be "plain sv's" generally. */
21225 case 'r': /* a compiled regex (but still just another SV) */
21226 case 's': /* an RV (currently only used for an RV to an AV by the ANYOF code)
21227 this use case should go away, the code could have used
21228 'a' instead - see S_set_ANYOF_arg() for array contents. */
21229 case 'S': /* actually an SV, but the dup function is identical. */
21230 case 'u': /* actually an HV, but the dup function is identical.
21231 values are "plain sv's" */
21232 d->data[i] = sv_dup_inc((const SV *)ri->data->data[i], param);
21235 /* Synthetic Start Class - "Fake" charclass we generate to optimize
21236 * patterns which could start with several different things. Pre-TRIE
21237 * this was more important than it is now, however this still helps
21238 * in some places, for instance /x?a+/ might produce a SSC equivalent
21239 * to [xa]. This is used by Perl_re_intuit_start() and S_find_byclass()
21242 /* This is cheating. */
21243 Newx(d->data[i], 1, regnode_ssc);
21244 StructCopy(ri->data->data[i], d->data[i], regnode_ssc);
21245 reti->regstclass = (regnode*)d->data[i];
21248 /* AHO-CORASICK fail table */
21249 /* Trie stclasses are readonly and can thus be shared
21250 * without duplication. We free the stclass in pregfree
21251 * when the corresponding reg_ac_data struct is freed.
21253 reti->regstclass= ri->regstclass;
21256 /* TRIE transition table */
21258 ((reg_trie_data*)ri->data->data[i])->refcount++;
21261 case 'l': /* (?{...}) or (??{ ... }) code (cb->block) */
21262 case 'L': /* same when RExC_pm_flags & PMf_HAS_CV and code
21263 is not from another regexp */
21264 d->data[i] = ri->data->data[i];
21267 Perl_croak(aTHX_ "panic: re_dup_guts unknown data code '%c'",
21268 ri->data->what[i]);
21277 reti->name_list_idx = ri->name_list_idx;
21279 #ifdef RE_TRACK_PATTERN_OFFSETS
21280 if (ri->u.offsets) {
21281 Newx(reti->u.offsets, 2*len+1, U32);
21282 Copy(ri->u.offsets, reti->u.offsets, 2*len+1, U32);
21285 SetProgLen(reti, len);
21288 return (void*)reti;
21291 #endif /* USE_ITHREADS */
21293 #ifndef PERL_IN_XSUB_RE
21296 - regnext - dig the "next" pointer out of a node
21299 Perl_regnext(pTHX_ regnode *p)
21306 if (OP(p) > REGNODE_MAX) { /* regnode.type is unsigned */
21307 Perl_croak(aTHX_ "Corrupted regexp opcode %d > %d",
21308 (int)OP(p), (int)REGNODE_MAX);
21311 offset = (reg_off_by_arg[OP(p)] ? ARG(p) : NEXT_OFF(p));
21321 S_re_croak2(pTHX_ bool utf8, const char* pat1, const char* pat2,...)
21324 STRLEN l1 = strlen(pat1);
21325 STRLEN l2 = strlen(pat2);
21328 const char *message;
21330 PERL_ARGS_ASSERT_RE_CROAK2;
21336 Copy(pat1, buf, l1 , char);
21337 Copy(pat2, buf + l1, l2 , char);
21338 buf[l1 + l2] = '\n';
21339 buf[l1 + l2 + 1] = '\0';
21340 va_start(args, pat2);
21341 msv = vmess(buf, &args);
21343 message = SvPV_const(msv, l1);
21346 Copy(message, buf, l1 , char);
21347 /* l1-1 to avoid \n */
21348 Perl_croak(aTHX_ "%" UTF8f, UTF8fARG(utf8, l1-1, buf));
21351 /* XXX Here's a total kludge. But we need to re-enter for swash routines. */
21353 #ifndef PERL_IN_XSUB_RE
21355 Perl_save_re_context(pTHX)
21360 /* Save $1..$n (#18107: UTF-8 s/(\w+)/uc($1)/e); AMS 20021106. */
21363 const REGEXP * const rx = PM_GETRE(PL_curpm);
21365 nparens = RX_NPARENS(rx);
21368 /* RT #124109. This is a complete hack; in the SWASHNEW case we know
21369 * that PL_curpm will be null, but that utf8.pm and the modules it
21370 * loads will only use $1..$3.
21371 * The t/porting/re_context.t test file checks this assumption.
21376 for (i = 1; i <= nparens; i++) {
21377 char digits[TYPE_CHARS(long)];
21378 const STRLEN len = my_snprintf(digits, sizeof(digits),
21380 GV *const *const gvp
21381 = (GV**)hv_fetch(PL_defstash, digits, len, 0);
21384 GV * const gv = *gvp;
21385 if (SvTYPE(gv) == SVt_PVGV && GvSV(gv))
21395 S_put_code_point(pTHX_ SV *sv, UV c)
21397 PERL_ARGS_ASSERT_PUT_CODE_POINT;
21400 Perl_sv_catpvf(aTHX_ sv, "\\x{%04" UVXf "}", c);
21402 else if (isPRINT(c)) {
21403 const char string = (char) c;
21405 /* We use {phrase} as metanotation in the class, so also escape literal
21407 if (isBACKSLASHED_PUNCT(c) || c == '{' || c == '}')
21408 sv_catpvs(sv, "\\");
21409 sv_catpvn(sv, &string, 1);
21411 else if (isMNEMONIC_CNTRL(c)) {
21412 Perl_sv_catpvf(aTHX_ sv, "%s", cntrl_to_mnemonic((U8) c));
21415 Perl_sv_catpvf(aTHX_ sv, "\\x%02X", (U8) c);
21419 #define MAX_PRINT_A MAX_PRINT_A_FOR_USE_ONLY_BY_REGCOMP_DOT_C
21422 S_put_range(pTHX_ SV *sv, UV start, const UV end, const bool allow_literals)
21424 /* Appends to 'sv' a displayable version of the range of code points from
21425 * 'start' to 'end'. Mnemonics (like '\r') are used for the few controls
21426 * that have them, when they occur at the beginning or end of the range.
21427 * It uses hex to output the remaining code points, unless 'allow_literals'
21428 * is true, in which case the printable ASCII ones are output as-is (though
21429 * some of these will be escaped by put_code_point()).
21431 * NOTE: This is designed only for printing ranges of code points that fit
21432 * inside an ANYOF bitmap. Higher code points are simply suppressed
21435 const unsigned int min_range_count = 3;
21437 assert(start <= end);
21439 PERL_ARGS_ASSERT_PUT_RANGE;
21441 while (start <= end) {
21443 const char * format;
21445 if (end - start < min_range_count) {
21447 /* Output chars individually when they occur in short ranges */
21448 for (; start <= end; start++) {
21449 put_code_point(sv, start);
21454 /* If permitted by the input options, and there is a possibility that
21455 * this range contains a printable literal, look to see if there is
21457 if (allow_literals && start <= MAX_PRINT_A) {
21459 /* If the character at the beginning of the range isn't an ASCII
21460 * printable, effectively split the range into two parts:
21461 * 1) the portion before the first such printable,
21463 * and output them separately. */
21464 if (! isPRINT_A(start)) {
21465 UV temp_end = start + 1;
21467 /* There is no point looking beyond the final possible
21468 * printable, in MAX_PRINT_A */
21469 UV max = MIN(end, MAX_PRINT_A);
21471 while (temp_end <= max && ! isPRINT_A(temp_end)) {
21475 /* Here, temp_end points to one beyond the first printable if
21476 * found, or to one beyond 'max' if not. If none found, make
21477 * sure that we use the entire range */
21478 if (temp_end > MAX_PRINT_A) {
21479 temp_end = end + 1;
21482 /* Output the first part of the split range: the part that
21483 * doesn't have printables, with the parameter set to not look
21484 * for literals (otherwise we would infinitely recurse) */
21485 put_range(sv, start, temp_end - 1, FALSE);
21487 /* The 2nd part of the range (if any) starts here. */
21490 /* We do a continue, instead of dropping down, because even if
21491 * the 2nd part is non-empty, it could be so short that we want
21492 * to output it as individual characters, as tested for at the
21493 * top of this loop. */
21497 /* Here, 'start' is a printable ASCII. If it is an alphanumeric,
21498 * output a sub-range of just the digits or letters, then process
21499 * the remaining portion as usual. */
21500 if (isALPHANUMERIC_A(start)) {
21501 UV mask = (isDIGIT_A(start))
21506 UV temp_end = start + 1;
21508 /* Find the end of the sub-range that includes just the
21509 * characters in the same class as the first character in it */
21510 while (temp_end <= end && _generic_isCC_A(temp_end, mask)) {
21515 /* For short ranges, don't duplicate the code above to output
21516 * them; just call recursively */
21517 if (temp_end - start < min_range_count) {
21518 put_range(sv, start, temp_end, FALSE);
21520 else { /* Output as a range */
21521 put_code_point(sv, start);
21522 sv_catpvs(sv, "-");
21523 put_code_point(sv, temp_end);
21525 start = temp_end + 1;
21529 /* We output any other printables as individual characters */
21530 if (isPUNCT_A(start) || isSPACE_A(start)) {
21531 while (start <= end && (isPUNCT_A(start)
21532 || isSPACE_A(start)))
21534 put_code_point(sv, start);
21539 } /* End of looking for literals */
21541 /* Here is not to output as a literal. Some control characters have
21542 * mnemonic names. Split off any of those at the beginning and end of
21543 * the range to print mnemonically. It isn't possible for many of
21544 * these to be in a row, so this won't overwhelm with output */
21546 && (isMNEMONIC_CNTRL(start) || isMNEMONIC_CNTRL(end)))
21548 while (isMNEMONIC_CNTRL(start) && start <= end) {
21549 put_code_point(sv, start);
21553 /* If this didn't take care of the whole range ... */
21554 if (start <= end) {
21556 /* Look backwards from the end to find the final non-mnemonic
21559 while (isMNEMONIC_CNTRL(temp_end)) {
21563 /* And separately output the interior range that doesn't start
21564 * or end with mnemonics */
21565 put_range(sv, start, temp_end, FALSE);
21567 /* Then output the mnemonic trailing controls */
21568 start = temp_end + 1;
21569 while (start <= end) {
21570 put_code_point(sv, start);
21577 /* As a final resort, output the range or subrange as hex. */
21579 this_end = (end < NUM_ANYOF_CODE_POINTS)
21581 : NUM_ANYOF_CODE_POINTS - 1;
21582 #if NUM_ANYOF_CODE_POINTS > 256
21583 format = (this_end < 256)
21584 ? "\\x%02" UVXf "-\\x%02" UVXf
21585 : "\\x{%04" UVXf "}-\\x{%04" UVXf "}";
21587 format = "\\x%02" UVXf "-\\x%02" UVXf;
21589 GCC_DIAG_IGNORE_STMT(-Wformat-nonliteral);
21590 Perl_sv_catpvf(aTHX_ sv, format, start, this_end);
21591 GCC_DIAG_RESTORE_STMT;
21597 S_put_charclass_bitmap_innards_invlist(pTHX_ SV *sv, SV* invlist)
21599 /* Concatenate onto the PV in 'sv' a displayable form of the inversion list
21603 bool allow_literals = TRUE;
21605 PERL_ARGS_ASSERT_PUT_CHARCLASS_BITMAP_INNARDS_INVLIST;
21607 /* Generally, it is more readable if printable characters are output as
21608 * literals, but if a range (nearly) spans all of them, it's best to output
21609 * it as a single range. This code will use a single range if all but 2
21610 * ASCII printables are in it */
21611 invlist_iterinit(invlist);
21612 while (invlist_iternext(invlist, &start, &end)) {
21614 /* If the range starts beyond the final printable, it doesn't have any
21616 if (start > MAX_PRINT_A) {
21620 /* In both ASCII and EBCDIC, a SPACE is the lowest printable. To span
21621 * all but two, the range must start and end no later than 2 from
21623 if (start < ' ' + 2 && end > MAX_PRINT_A - 2) {
21624 if (end > MAX_PRINT_A) {
21630 if (end - start >= MAX_PRINT_A - ' ' - 2) {
21631 allow_literals = FALSE;
21636 invlist_iterfinish(invlist);
21638 /* Here we have figured things out. Output each range */
21639 invlist_iterinit(invlist);
21640 while (invlist_iternext(invlist, &start, &end)) {
21641 if (start >= NUM_ANYOF_CODE_POINTS) {
21644 put_range(sv, start, end, allow_literals);
21646 invlist_iterfinish(invlist);
21652 S_put_charclass_bitmap_innards_common(pTHX_
21653 SV* invlist, /* The bitmap */
21654 SV* posixes, /* Under /l, things like [:word:], \S */
21655 SV* only_utf8, /* Under /d, matches iff the target is UTF-8 */
21656 SV* not_utf8, /* /d, matches iff the target isn't UTF-8 */
21657 SV* only_utf8_locale, /* Under /l, matches if the locale is UTF-8 */
21658 const bool invert /* Is the result to be inverted? */
21661 /* Create and return an SV containing a displayable version of the bitmap
21662 * and associated information determined by the input parameters. If the
21663 * output would have been only the inversion indicator '^', NULL is instead
21669 PERL_ARGS_ASSERT_PUT_CHARCLASS_BITMAP_INNARDS_COMMON;
21672 output = newSVpvs("^");
21675 output = newSVpvs("");
21678 /* First, the code points in the bitmap that are unconditionally there */
21679 put_charclass_bitmap_innards_invlist(output, invlist);
21681 /* Traditionally, these have been placed after the main code points */
21683 sv_catsv(output, posixes);
21686 if (only_utf8 && _invlist_len(only_utf8)) {
21687 Perl_sv_catpvf(aTHX_ output, "%s{utf8}%s", PL_colors[1], PL_colors[0]);
21688 put_charclass_bitmap_innards_invlist(output, only_utf8);
21691 if (not_utf8 && _invlist_len(not_utf8)) {
21692 Perl_sv_catpvf(aTHX_ output, "%s{not utf8}%s", PL_colors[1], PL_colors[0]);
21693 put_charclass_bitmap_innards_invlist(output, not_utf8);
21696 if (only_utf8_locale && _invlist_len(only_utf8_locale)) {
21697 Perl_sv_catpvf(aTHX_ output, "%s{utf8 locale}%s", PL_colors[1], PL_colors[0]);
21698 put_charclass_bitmap_innards_invlist(output, only_utf8_locale);
21700 /* This is the only list in this routine that can legally contain code
21701 * points outside the bitmap range. The call just above to
21702 * 'put_charclass_bitmap_innards_invlist' will simply suppress them, so
21703 * output them here. There's about a half-dozen possible, and none in
21704 * contiguous ranges longer than 2 */
21705 if (invlist_highest(only_utf8_locale) >= NUM_ANYOF_CODE_POINTS) {
21707 SV* above_bitmap = NULL;
21709 _invlist_subtract(only_utf8_locale, PL_InBitmap, &above_bitmap);
21711 invlist_iterinit(above_bitmap);
21712 while (invlist_iternext(above_bitmap, &start, &end)) {
21715 for (i = start; i <= end; i++) {
21716 put_code_point(output, i);
21719 invlist_iterfinish(above_bitmap);
21720 SvREFCNT_dec_NN(above_bitmap);
21724 if (invert && SvCUR(output) == 1) {
21732 S_put_charclass_bitmap_innards(pTHX_ SV *sv,
21734 SV *nonbitmap_invlist,
21735 SV *only_utf8_locale_invlist,
21736 const regnode * const node,
21737 const bool force_as_is_display)
21739 /* Appends to 'sv' a displayable version of the innards of the bracketed
21740 * character class defined by the other arguments:
21741 * 'bitmap' points to the bitmap, or NULL if to ignore that.
21742 * 'nonbitmap_invlist' is an inversion list of the code points that are in
21743 * the bitmap range, but for some reason aren't in the bitmap; NULL if
21744 * none. The reasons for this could be that they require some
21745 * condition such as the target string being or not being in UTF-8
21746 * (under /d), or because they came from a user-defined property that
21747 * was not resolved at the time of the regex compilation (under /u)
21748 * 'only_utf8_locale_invlist' is an inversion list of the code points that
21749 * are valid only if the runtime locale is a UTF-8 one; NULL if none
21750 * 'node' is the regex pattern ANYOF node. It is needed only when the
21751 * above two parameters are not null, and is passed so that this
21752 * routine can tease apart the various reasons for them.
21753 * 'force_as_is_display' is TRUE if this routine should definitely NOT try
21754 * to invert things to see if that leads to a cleaner display. If
21755 * FALSE, this routine is free to use its judgment about doing this.
21757 * It returns TRUE if there was actually something output. (It may be that
21758 * the bitmap, etc is empty.)
21760 * When called for outputting the bitmap of a non-ANYOF node, just pass the
21761 * bitmap, with the succeeding parameters set to NULL, and the final one to
21765 /* In general, it tries to display the 'cleanest' representation of the
21766 * innards, choosing whether to display them inverted or not, regardless of
21767 * whether the class itself is to be inverted. However, there are some
21768 * cases where it can't try inverting, as what actually matches isn't known
21769 * until runtime, and hence the inversion isn't either. */
21772 bool inverting_allowed = ! force_as_is_display;
21775 STRLEN orig_sv_cur = SvCUR(sv);
21777 SV* invlist; /* Inversion list we accumulate of code points that
21778 are unconditionally matched */
21779 SV* only_utf8 = NULL; /* Under /d, list of matches iff the target is
21781 SV* not_utf8 = NULL; /* /d, list of matches iff the target isn't UTF-8
21783 SV* posixes = NULL; /* Under /l, string of things like [:word:], \D */
21784 SV* only_utf8_locale = NULL; /* Under /l, list of matches if the locale
21787 SV* as_is_display; /* The output string when we take the inputs
21789 SV* inverted_display; /* The output string when we invert the inputs */
21791 U8 flags = (node) ? ANYOF_FLAGS(node) : 0;
21793 bool invert = cBOOL(flags & ANYOF_INVERT); /* Is the input to be inverted
21795 /* We are biased in favor of displaying things without them being inverted,
21796 * as that is generally easier to understand */
21797 const int bias = 5;
21799 PERL_ARGS_ASSERT_PUT_CHARCLASS_BITMAP_INNARDS;
21801 /* Start off with whatever code points are passed in. (We clone, so we
21802 * don't change the caller's list) */
21803 if (nonbitmap_invlist) {
21804 assert(invlist_highest(nonbitmap_invlist) < NUM_ANYOF_CODE_POINTS);
21805 invlist = invlist_clone(nonbitmap_invlist, NULL);
21807 else { /* Worst case size is every other code point is matched */
21808 invlist = _new_invlist(NUM_ANYOF_CODE_POINTS / 2);
21812 if (OP(node) == ANYOFD) {
21814 /* This flag indicates that the code points below 0x100 in the
21815 * nonbitmap list are precisely the ones that match only when the
21816 * target is UTF-8 (they should all be non-ASCII). */
21817 if (flags & ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP)
21819 _invlist_intersection(invlist, PL_UpperLatin1, &only_utf8);
21820 _invlist_subtract(invlist, only_utf8, &invlist);
21823 /* And this flag for matching all non-ASCII 0xFF and below */
21824 if (flags & ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER)
21826 not_utf8 = invlist_clone(PL_UpperLatin1, NULL);
21829 else if (OP(node) == ANYOFL || OP(node) == ANYOFPOSIXL) {
21831 /* If either of these flags are set, what matches isn't
21832 * determinable except during execution, so don't know enough here
21834 if (flags & (ANYOFL_FOLD|ANYOF_MATCHES_POSIXL)) {
21835 inverting_allowed = FALSE;
21838 /* What the posix classes match also varies at runtime, so these
21839 * will be output symbolically. */
21840 if (ANYOF_POSIXL_TEST_ANY_SET(node)) {
21843 posixes = newSVpvs("");
21844 for (i = 0; i < ANYOF_POSIXL_MAX; i++) {
21845 if (ANYOF_POSIXL_TEST(node, i)) {
21846 sv_catpv(posixes, anyofs[i]);
21853 /* Accumulate the bit map into the unconditional match list */
21855 for (i = 0; i < NUM_ANYOF_CODE_POINTS; i++) {
21856 if (BITMAP_TEST(bitmap, i)) {
21859 i < NUM_ANYOF_CODE_POINTS && BITMAP_TEST(bitmap, i);
21862 invlist = _add_range_to_invlist(invlist, start, i-1);
21867 /* Make sure that the conditional match lists don't have anything in them
21868 * that match unconditionally; otherwise the output is quite confusing.
21869 * This could happen if the code that populates these misses some
21872 _invlist_subtract(only_utf8, invlist, &only_utf8);
21875 _invlist_subtract(not_utf8, invlist, ¬_utf8);
21878 if (only_utf8_locale_invlist) {
21880 /* Since this list is passed in, we have to make a copy before
21882 only_utf8_locale = invlist_clone(only_utf8_locale_invlist, NULL);
21884 _invlist_subtract(only_utf8_locale, invlist, &only_utf8_locale);
21886 /* And, it can get really weird for us to try outputting an inverted
21887 * form of this list when it has things above the bitmap, so don't even
21889 if (invlist_highest(only_utf8_locale) >= NUM_ANYOF_CODE_POINTS) {
21890 inverting_allowed = FALSE;
21894 /* Calculate what the output would be if we take the input as-is */
21895 as_is_display = put_charclass_bitmap_innards_common(invlist,
21902 /* If have to take the output as-is, just do that */
21903 if (! inverting_allowed) {
21904 if (as_is_display) {
21905 sv_catsv(sv, as_is_display);
21906 SvREFCNT_dec_NN(as_is_display);
21909 else { /* But otherwise, create the output again on the inverted input, and
21910 use whichever version is shorter */
21912 int inverted_bias, as_is_bias;
21914 /* We will apply our bias to whichever of the the results doesn't have
21924 inverted_bias = bias;
21927 /* Now invert each of the lists that contribute to the output,
21928 * excluding from the result things outside the possible range */
21930 /* For the unconditional inversion list, we have to add in all the
21931 * conditional code points, so that when inverted, they will be gone
21933 _invlist_union(only_utf8, invlist, &invlist);
21934 _invlist_union(not_utf8, invlist, &invlist);
21935 _invlist_union(only_utf8_locale, invlist, &invlist);
21936 _invlist_invert(invlist);
21937 _invlist_intersection(invlist, PL_InBitmap, &invlist);
21940 _invlist_invert(only_utf8);
21941 _invlist_intersection(only_utf8, PL_UpperLatin1, &only_utf8);
21943 else if (not_utf8) {
21945 /* If a code point matches iff the target string is not in UTF-8,
21946 * then complementing the result has it not match iff not in UTF-8,
21947 * which is the same thing as matching iff it is UTF-8. */
21948 only_utf8 = not_utf8;
21952 if (only_utf8_locale) {
21953 _invlist_invert(only_utf8_locale);
21954 _invlist_intersection(only_utf8_locale,
21956 &only_utf8_locale);
21959 inverted_display = put_charclass_bitmap_innards_common(
21964 only_utf8_locale, invert);
21966 /* Use the shortest representation, taking into account our bias
21967 * against showing it inverted */
21968 if ( inverted_display
21969 && ( ! as_is_display
21970 || ( SvCUR(inverted_display) + inverted_bias
21971 < SvCUR(as_is_display) + as_is_bias)))
21973 sv_catsv(sv, inverted_display);
21975 else if (as_is_display) {
21976 sv_catsv(sv, as_is_display);
21979 SvREFCNT_dec(as_is_display);
21980 SvREFCNT_dec(inverted_display);
21983 SvREFCNT_dec_NN(invlist);
21984 SvREFCNT_dec(only_utf8);
21985 SvREFCNT_dec(not_utf8);
21986 SvREFCNT_dec(posixes);
21987 SvREFCNT_dec(only_utf8_locale);
21989 return SvCUR(sv) > orig_sv_cur;
21992 #define CLEAR_OPTSTART \
21993 if (optstart) STMT_START { \
21994 DEBUG_OPTIMISE_r(Perl_re_printf( aTHX_ \
21995 " (%" IVdf " nodes)\n", (IV)(node - optstart))); \
21999 #define DUMPUNTIL(b,e) \
22001 node=dumpuntil(r,start,(b),(e),last,sv,indent+1,depth+1);
22003 STATIC const regnode *
22004 S_dumpuntil(pTHX_ const regexp *r, const regnode *start, const regnode *node,
22005 const regnode *last, const regnode *plast,
22006 SV* sv, I32 indent, U32 depth)
22008 U8 op = PSEUDO; /* Arbitrary non-END op. */
22009 const regnode *next;
22010 const regnode *optstart= NULL;
22012 RXi_GET_DECL(r, ri);
22013 GET_RE_DEBUG_FLAGS_DECL;
22015 PERL_ARGS_ASSERT_DUMPUNTIL;
22017 #ifdef DEBUG_DUMPUNTIL
22018 Perl_re_printf( aTHX_ "--- %d : %d - %d - %d\n", indent, node-start,
22019 last ? last-start : 0, plast ? plast-start : 0);
22022 if (plast && plast < last)
22025 while (PL_regkind[op] != END && (!last || node < last)) {
22027 /* While that wasn't END last time... */
22030 if (op == CLOSE || op == SRCLOSE || op == WHILEM)
22032 next = regnext((regnode *)node);
22035 if (OP(node) == OPTIMIZED) {
22036 if (!optstart && RE_DEBUG_FLAG(RE_DEBUG_COMPILE_OPTIMISE))
22043 regprop(r, sv, node, NULL, NULL);
22044 Perl_re_printf( aTHX_ "%4" IVdf ":%*s%s", (IV)(node - start),
22045 (int)(2*indent + 1), "", SvPVX_const(sv));
22047 if (OP(node) != OPTIMIZED) {
22048 if (next == NULL) /* Next ptr. */
22049 Perl_re_printf( aTHX_ " (0)");
22050 else if (PL_regkind[(U8)op] == BRANCH
22051 && PL_regkind[OP(next)] != BRANCH )
22052 Perl_re_printf( aTHX_ " (FAIL)");
22054 Perl_re_printf( aTHX_ " (%" IVdf ")", (IV)(next - start));
22055 Perl_re_printf( aTHX_ "\n");
22059 if (PL_regkind[(U8)op] == BRANCHJ) {
22062 const regnode *nnode = (OP(next) == LONGJMP
22063 ? regnext((regnode *)next)
22065 if (last && nnode > last)
22067 DUMPUNTIL(NEXTOPER(NEXTOPER(node)), nnode);
22070 else if (PL_regkind[(U8)op] == BRANCH) {
22072 DUMPUNTIL(NEXTOPER(node), next);
22074 else if ( PL_regkind[(U8)op] == TRIE ) {
22075 const regnode *this_trie = node;
22076 const char op = OP(node);
22077 const U32 n = ARG(node);
22078 const reg_ac_data * const ac = op>=AHOCORASICK ?
22079 (reg_ac_data *)ri->data->data[n] :
22081 const reg_trie_data * const trie =
22082 (reg_trie_data*)ri->data->data[op<AHOCORASICK ? n : ac->trie];
22084 AV *const trie_words
22085 = MUTABLE_AV(ri->data->data[n + TRIE_WORDS_OFFSET]);
22087 const regnode *nextbranch= NULL;
22090 for (word_idx= 0; word_idx < (I32)trie->wordcount; word_idx++) {
22091 SV ** const elem_ptr = av_fetch(trie_words, word_idx, 0);
22093 Perl_re_indentf( aTHX_ "%s ",
22096 ? pv_pretty(sv, SvPV_nolen_const(*elem_ptr),
22097 SvCUR(*elem_ptr), PL_dump_re_max_len,
22098 PL_colors[0], PL_colors[1],
22100 ? PERL_PV_ESCAPE_UNI
22102 | PERL_PV_PRETTY_ELLIPSES
22103 | PERL_PV_PRETTY_LTGT
22108 U16 dist= trie->jump[word_idx+1];
22109 Perl_re_printf( aTHX_ "(%" UVuf ")\n",
22110 (UV)((dist ? this_trie + dist : next) - start));
22113 nextbranch= this_trie + trie->jump[0];
22114 DUMPUNTIL(this_trie + dist, nextbranch);
22116 if (nextbranch && PL_regkind[OP(nextbranch)]==BRANCH)
22117 nextbranch= regnext((regnode *)nextbranch);
22119 Perl_re_printf( aTHX_ "\n");
22122 if (last && next > last)
22127 else if ( op == CURLY ) { /* "next" might be very big: optimizer */
22128 DUMPUNTIL(NEXTOPER(node) + EXTRA_STEP_2ARGS,
22129 NEXTOPER(node) + EXTRA_STEP_2ARGS + 1);
22131 else if (PL_regkind[(U8)op] == CURLY && op != CURLYX) {
22133 DUMPUNTIL(NEXTOPER(node) + EXTRA_STEP_2ARGS, next);
22135 else if ( op == PLUS || op == STAR) {
22136 DUMPUNTIL(NEXTOPER(node), NEXTOPER(node) + 1);
22138 else if (PL_regkind[(U8)op] == EXACT) {
22139 /* Literal string, where present. */
22140 node += NODE_SZ_STR(node) - 1;
22141 node = NEXTOPER(node);
22144 node = NEXTOPER(node);
22145 node += regarglen[(U8)op];
22147 if (op == CURLYX || op == OPEN || op == SROPEN)
22151 #ifdef DEBUG_DUMPUNTIL
22152 Perl_re_printf( aTHX_ "--- %d\n", (int)indent);
22157 #endif /* DEBUGGING */
22159 #ifndef PERL_IN_XSUB_RE
22161 #include "uni_keywords.h"
22164 Perl_init_uniprops(pTHX)
22168 PL_user_def_props = newHV();
22170 #ifdef USE_ITHREADS
22172 HvSHAREKEYS_off(PL_user_def_props);
22173 PL_user_def_props_aTHX = aTHX;
22177 /* Set up the inversion list global variables */
22179 PL_XPosix_ptrs[_CC_ASCII] = _new_invlist_C_array(uni_prop_ptrs[UNI_ASCII]);
22180 PL_XPosix_ptrs[_CC_ALPHANUMERIC] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXALNUM]);
22181 PL_XPosix_ptrs[_CC_ALPHA] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXALPHA]);
22182 PL_XPosix_ptrs[_CC_BLANK] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXBLANK]);
22183 PL_XPosix_ptrs[_CC_CASED] = _new_invlist_C_array(uni_prop_ptrs[UNI_CASED]);
22184 PL_XPosix_ptrs[_CC_CNTRL] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXCNTRL]);
22185 PL_XPosix_ptrs[_CC_DIGIT] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXDIGIT]);
22186 PL_XPosix_ptrs[_CC_GRAPH] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXGRAPH]);
22187 PL_XPosix_ptrs[_CC_LOWER] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXLOWER]);
22188 PL_XPosix_ptrs[_CC_PRINT] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXPRINT]);
22189 PL_XPosix_ptrs[_CC_PUNCT] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXPUNCT]);
22190 PL_XPosix_ptrs[_CC_SPACE] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXSPACE]);
22191 PL_XPosix_ptrs[_CC_UPPER] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXUPPER]);
22192 PL_XPosix_ptrs[_CC_VERTSPACE] = _new_invlist_C_array(uni_prop_ptrs[UNI_VERTSPACE]);
22193 PL_XPosix_ptrs[_CC_WORDCHAR] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXWORD]);
22194 PL_XPosix_ptrs[_CC_XDIGIT] = _new_invlist_C_array(uni_prop_ptrs[UNI_XPOSIXXDIGIT]);
22196 PL_Posix_ptrs[_CC_ASCII] = _new_invlist_C_array(uni_prop_ptrs[UNI_ASCII]);
22197 PL_Posix_ptrs[_CC_ALPHANUMERIC] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXALNUM]);
22198 PL_Posix_ptrs[_CC_ALPHA] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXALPHA]);
22199 PL_Posix_ptrs[_CC_BLANK] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXBLANK]);
22200 PL_Posix_ptrs[_CC_CASED] = PL_Posix_ptrs[_CC_ALPHA];
22201 PL_Posix_ptrs[_CC_CNTRL] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXCNTRL]);
22202 PL_Posix_ptrs[_CC_DIGIT] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXDIGIT]);
22203 PL_Posix_ptrs[_CC_GRAPH] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXGRAPH]);
22204 PL_Posix_ptrs[_CC_LOWER] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXLOWER]);
22205 PL_Posix_ptrs[_CC_PRINT] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXPRINT]);
22206 PL_Posix_ptrs[_CC_PUNCT] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXPUNCT]);
22207 PL_Posix_ptrs[_CC_SPACE] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXSPACE]);
22208 PL_Posix_ptrs[_CC_UPPER] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXUPPER]);
22209 PL_Posix_ptrs[_CC_VERTSPACE] = NULL;
22210 PL_Posix_ptrs[_CC_WORDCHAR] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXWORD]);
22211 PL_Posix_ptrs[_CC_XDIGIT] = _new_invlist_C_array(uni_prop_ptrs[UNI_POSIXXDIGIT]);
22213 PL_GCB_invlist = _new_invlist_C_array(_Perl_GCB_invlist);
22214 PL_SB_invlist = _new_invlist_C_array(_Perl_SB_invlist);
22215 PL_WB_invlist = _new_invlist_C_array(_Perl_WB_invlist);
22216 PL_LB_invlist = _new_invlist_C_array(_Perl_LB_invlist);
22217 PL_SCX_invlist = _new_invlist_C_array(_Perl_SCX_invlist);
22219 PL_AboveLatin1 = _new_invlist_C_array(AboveLatin1_invlist);
22220 PL_Latin1 = _new_invlist_C_array(Latin1_invlist);
22221 PL_UpperLatin1 = _new_invlist_C_array(UpperLatin1_invlist);
22223 PL_Assigned_invlist = _new_invlist_C_array(uni_prop_ptrs[UNI_ASSIGNED]);
22225 PL_utf8_perl_idstart = _new_invlist_C_array(uni_prop_ptrs[UNI__PERL_IDSTART]);
22226 PL_utf8_perl_idcont = _new_invlist_C_array(uni_prop_ptrs[UNI__PERL_IDCONT]);
22228 PL_utf8_charname_begin = _new_invlist_C_array(uni_prop_ptrs[UNI__PERL_CHARNAME_BEGIN]);
22229 PL_utf8_charname_continue = _new_invlist_C_array(uni_prop_ptrs[UNI__PERL_CHARNAME_CONTINUE]);
22231 PL_in_some_fold = _new_invlist_C_array(uni_prop_ptrs[UNI__PERL_ANY_FOLDS]);
22232 PL_HasMultiCharFold = _new_invlist_C_array(uni_prop_ptrs[
22233 UNI__PERL_FOLDS_TO_MULTI_CHAR]);
22234 PL_InMultiCharFold = _new_invlist_C_array(uni_prop_ptrs[
22235 UNI__PERL_IS_IN_MULTI_CHAR_FOLD]);
22236 PL_NonFinalFold = _new_invlist_C_array(uni_prop_ptrs[
22237 UNI__PERL_NON_FINAL_FOLDS]);
22239 PL_utf8_toupper = _new_invlist_C_array(Uppercase_Mapping_invlist);
22240 PL_utf8_tolower = _new_invlist_C_array(Lowercase_Mapping_invlist);
22241 PL_utf8_totitle = _new_invlist_C_array(Titlecase_Mapping_invlist);
22242 PL_utf8_tofold = _new_invlist_C_array(Case_Folding_invlist);
22243 PL_utf8_tosimplefold = _new_invlist_C_array(Simple_Case_Folding_invlist);
22244 PL_utf8_foldclosures = _new_invlist_C_array(_Perl_IVCF_invlist);
22245 PL_utf8_mark = _new_invlist_C_array(uni_prop_ptrs[UNI_M]);
22246 PL_CCC_non0_non230 = _new_invlist_C_array(_Perl_CCC_non0_non230_invlist);
22247 PL_Private_Use = _new_invlist_C_array(uni_prop_ptrs[UNI_CO]);
22250 /* The below are used only by deprecated functions. They could be removed */
22251 PL_utf8_xidcont = _new_invlist_C_array(uni_prop_ptrs[UNI_XIDC]);
22252 PL_utf8_idcont = _new_invlist_C_array(uni_prop_ptrs[UNI_IDC]);
22253 PL_utf8_xidstart = _new_invlist_C_array(uni_prop_ptrs[UNI_XIDS]);
22259 This code was mainly added for backcompat to give a warning for non-portable
22260 code points in user-defined properties. But experiments showed that the
22261 warning in earlier perls were only omitted on overflow, which should be an
22262 error, so there really isnt a backcompat issue, and actually adding the
22263 warning when none was present before might cause breakage, for little gain. So
22264 khw left this code in, but not enabled. Tests were never added.
22267 Ei |const char *|get_extended_utf8_msg|const UV cp
22269 PERL_STATIC_INLINE const char *
22270 S_get_extended_utf8_msg(pTHX_ const UV cp)
22272 U8 dummy[UTF8_MAXBYTES + 1];
22276 uvchr_to_utf8_flags_msgs(dummy, cp, UNICODE_WARN_PERL_EXTENDED,
22279 msg = hv_fetchs(msgs, "text", 0);
22282 (void) sv_2mortal((SV *) msgs);
22284 return SvPVX(*msg);
22290 Perl_handle_user_defined_property(pTHX_
22292 /* Parses the contents of a user-defined property definition; returning the
22293 * expanded definition if possible. If so, the return is an inversion
22296 * If there are subroutines that are part of the expansion and which aren't
22297 * known at the time of the call to this function, this returns what
22298 * parse_uniprop_string() returned for the first one encountered.
22300 * If an error was found, NULL is returned, and 'msg' gets a suitable
22301 * message appended to it. (Appending allows the back trace of how we got
22302 * to the faulty definition to be displayed through nested calls of
22303 * user-defined subs.)
22305 * The caller IS responsible for freeing any returned SV.
22307 * The syntax of the contents is pretty much described in perlunicode.pod,
22308 * but we also allow comments on each line */
22310 const char * name, /* Name of property */
22311 const STRLEN name_len, /* The name's length in bytes */
22312 const bool is_utf8, /* ? Is 'name' encoded in UTF-8 */
22313 const bool to_fold, /* ? Is this under /i */
22314 const bool runtime, /* ? Are we in compile- or run-time */
22315 const bool deferrable, /* Is it ok for this property's full definition
22316 to be deferred until later? */
22317 SV* contents, /* The property's definition */
22318 bool *user_defined_ptr, /* This will be set TRUE as we wouldn't be
22319 getting called unless this is thought to be
22320 a user-defined property */
22321 SV * msg, /* Any error or warning msg(s) are appended to
22323 const STRLEN level) /* Recursion level of this call */
22326 const char * string = SvPV_const(contents, len);
22327 const char * const e = string + len;
22328 const bool is_contents_utf8 = cBOOL(SvUTF8(contents));
22329 const STRLEN msgs_length_on_entry = SvCUR(msg);
22331 const char * s0 = string; /* Points to first byte in the current line
22332 being parsed in 'string' */
22333 const char overflow_msg[] = "Code point too large in \"";
22334 SV* running_definition = NULL;
22336 PERL_ARGS_ASSERT_HANDLE_USER_DEFINED_PROPERTY;
22338 *user_defined_ptr = TRUE;
22340 /* Look at each line */
22342 const char * s; /* Current byte */
22343 char op = '+'; /* Default operation is 'union' */
22344 IV min = 0; /* range begin code point */
22345 IV max = -1; /* and range end */
22346 SV* this_definition;
22348 /* Skip comment lines */
22350 s0 = strchr(s0, '\n');
22358 /* For backcompat, allow an empty first line */
22364 /* First character in the line may optionally be the operation */
22373 /* If the line is one or two hex digits separated by blank space, its
22374 * a range; otherwise it is either another user-defined property or an
22379 if (! isXDIGIT(*s)) {
22380 goto check_if_property;
22383 do { /* Each new hex digit will add 4 bits. */
22384 if (min > ( (IV) MAX_LEGAL_CP >> 4)) {
22385 s = strchr(s, '\n');
22389 if (SvCUR(msg) > 0) sv_catpvs(msg, "; ");
22390 sv_catpv(msg, overflow_msg);
22391 Perl_sv_catpvf(aTHX_ msg, "%" UTF8f,
22392 UTF8fARG(is_contents_utf8, s - s0, s0));
22393 sv_catpvs(msg, "\"");
22394 goto return_failure;
22397 /* Accumulate this digit into the value */
22398 min = (min << 4) + READ_XDIGIT(s);
22399 } while (isXDIGIT(*s));
22401 while (isBLANK(*s)) { s++; }
22403 /* We allow comments at the end of the line */
22405 s = strchr(s, '\n');
22411 else if (s < e && *s != '\n') {
22412 if (! isXDIGIT(*s)) {
22413 goto check_if_property;
22416 /* Look for the high point of the range */
22419 if (max > ( (IV) MAX_LEGAL_CP >> 4)) {
22420 s = strchr(s, '\n');
22424 if (SvCUR(msg) > 0) sv_catpvs(msg, "; ");
22425 sv_catpv(msg, overflow_msg);
22426 Perl_sv_catpvf(aTHX_ msg, "%" UTF8f,
22427 UTF8fARG(is_contents_utf8, s - s0, s0));
22428 sv_catpvs(msg, "\"");
22429 goto return_failure;
22432 max = (max << 4) + READ_XDIGIT(s);
22433 } while (isXDIGIT(*s));
22435 while (isBLANK(*s)) { s++; }
22438 s = strchr(s, '\n');
22443 else if (s < e && *s != '\n') {
22444 goto check_if_property;
22448 if (max == -1) { /* The line only had one entry */
22451 else if (max < min) {
22452 if (SvCUR(msg) > 0) sv_catpvs(msg, "; ");
22453 sv_catpvs(msg, "Illegal range in \"");
22454 Perl_sv_catpvf(aTHX_ msg, "%" UTF8f,
22455 UTF8fARG(is_contents_utf8, s - s0, s0));
22456 sv_catpvs(msg, "\"");
22457 goto return_failure;
22460 #if 0 /* See explanation at definition above of get_extended_utf8_msg() */
22462 if ( UNICODE_IS_PERL_EXTENDED(min)
22463 || UNICODE_IS_PERL_EXTENDED(max))
22465 if (SvCUR(msg) > 0) sv_catpvs(msg, "; ");
22467 /* If both code points are non-portable, warn only on the lower
22469 sv_catpv(msg, get_extended_utf8_msg(
22470 (UNICODE_IS_PERL_EXTENDED(min))
22472 sv_catpvs(msg, " in \"");
22473 Perl_sv_catpvf(aTHX_ msg, "%" UTF8f,
22474 UTF8fARG(is_contents_utf8, s - s0, s0));
22475 sv_catpvs(msg, "\"");
22480 /* Here, this line contains a legal range */
22481 this_definition = sv_2mortal(_new_invlist(2));
22482 this_definition = _add_range_to_invlist(this_definition, min, max);
22487 /* Here it isn't a legal range line. See if it is a legal property
22488 * line. First find the end of the meat of the line */
22489 s = strpbrk(s, "#\n");
22494 /* Ignore trailing blanks in keeping with the requirements of
22495 * parse_uniprop_string() */
22497 while (s > s0 && isBLANK_A(*s)) {
22502 this_definition = parse_uniprop_string(s0, s - s0,
22503 is_utf8, to_fold, runtime,
22505 user_defined_ptr, msg,
22507 ? level /* Don't increase level
22508 if input is empty */
22511 if (this_definition == NULL) {
22512 goto return_failure; /* 'msg' should have had the reason
22513 appended to it by the above call */
22516 if (! is_invlist(this_definition)) { /* Unknown at this time */
22517 return newSVsv(this_definition);
22521 s = strchr(s, '\n');
22531 _invlist_union(running_definition, this_definition,
22532 &running_definition);
22535 _invlist_subtract(running_definition, this_definition,
22536 &running_definition);
22539 _invlist_intersection(running_definition, this_definition,
22540 &running_definition);
22543 _invlist_union_complement_2nd(running_definition,
22544 this_definition, &running_definition);
22547 Perl_croak(aTHX_ "panic: %s: %d: Unexpected operation %d",
22548 __FILE__, __LINE__, op);
22552 /* Position past the '\n' */
22554 } /* End of loop through the lines of 'contents' */
22556 /* Here, we processed all the lines in 'contents' without error. If we
22557 * didn't add any warnings, simply return success */
22558 if (msgs_length_on_entry == SvCUR(msg)) {
22560 /* If the expansion was empty, the answer isn't nothing: its an empty
22561 * inversion list */
22562 if (running_definition == NULL) {
22563 running_definition = _new_invlist(1);
22566 return running_definition;
22569 /* Otherwise, add some explanatory text, but we will return success */
22573 running_definition = NULL;
22577 if (name_len > 0) {
22578 sv_catpvs(msg, " in expansion of ");
22579 Perl_sv_catpvf(aTHX_ msg, "%" UTF8f, UTF8fARG(is_utf8, name_len, name));
22582 return running_definition;
22585 /* As explained below, certain operations need to take place in the first
22586 * thread created. These macros switch contexts */
22587 #ifdef USE_ITHREADS
22588 # define DECLARATION_FOR_GLOBAL_CONTEXT \
22589 PerlInterpreter * save_aTHX = aTHX;
22590 # define SWITCH_TO_GLOBAL_CONTEXT \
22591 PERL_SET_CONTEXT((aTHX = PL_user_def_props_aTHX))
22592 # define RESTORE_CONTEXT PERL_SET_CONTEXT((aTHX = save_aTHX));
22593 # define CUR_CONTEXT aTHX
22594 # define ORIGINAL_CONTEXT save_aTHX
22596 # define DECLARATION_FOR_GLOBAL_CONTEXT
22597 # define SWITCH_TO_GLOBAL_CONTEXT NOOP
22598 # define RESTORE_CONTEXT NOOP
22599 # define CUR_CONTEXT NULL
22600 # define ORIGINAL_CONTEXT NULL
22604 S_delete_recursion_entry(pTHX_ void *key)
22606 /* Deletes the entry used to detect recursion when expanding user-defined
22607 * properties. This is a function so it can be set up to be called even if
22608 * the program unexpectedly quits */
22611 SV ** current_entry;
22612 const STRLEN key_len = strlen((const char *) key);
22613 DECLARATION_FOR_GLOBAL_CONTEXT;
22615 SWITCH_TO_GLOBAL_CONTEXT;
22617 /* If the entry is one of these types, it is a permanent entry, and not the
22618 * one used to detect recursions. This function should delete only the
22619 * recursion entry */
22620 current_entry = hv_fetch(PL_user_def_props, (const char *) key, key_len, 0);
22622 && ! is_invlist(*current_entry)
22623 && ! SvPOK(*current_entry))
22625 (void) hv_delete(PL_user_def_props, (const char *) key, key_len,
22633 S_get_fq_name(pTHX_
22634 const char * const name, /* The first non-blank in the \p{}, \P{} */
22635 const Size_t name_len, /* Its length in bytes, not including any trailing space */
22636 const bool is_utf8, /* ? Is 'name' encoded in UTF-8 */
22637 const bool has_colon_colon
22640 /* Returns a mortal SV containing the fully qualified version of the input
22645 fq_name = newSVpvs_flags("", SVs_TEMP);
22647 /* Use the current package if it wasn't included in our input */
22648 if (! has_colon_colon) {
22649 const HV * pkg = (IN_PERL_COMPILETIME)
22651 : CopSTASH(PL_curcop);
22652 const char* pkgname = HvNAME(pkg);
22654 Perl_sv_catpvf(aTHX_ fq_name, "%" UTF8f,
22655 UTF8fARG(is_utf8, strlen(pkgname), pkgname));
22656 sv_catpvs(fq_name, "::");
22659 Perl_sv_catpvf(aTHX_ fq_name, "%" UTF8f,
22660 UTF8fARG(is_utf8, name_len, name));
22665 Perl_parse_uniprop_string(pTHX_
22667 /* Parse the interior of a \p{}, \P{}. Returns its definition if knowable
22668 * now. If so, the return is an inversion list.
22670 * If the property is user-defined, it is a subroutine, which in turn
22671 * may call other subroutines. This function will call the whole nest of
22672 * them to get the definition they return; if some aren't known at the time
22673 * of the call to this function, the fully qualified name of the highest
22674 * level sub is returned. It is an error to call this function at runtime
22675 * without every sub defined.
22677 * If an error was found, NULL is returned, and 'msg' gets a suitable
22678 * message appended to it. (Appending allows the back trace of how we got
22679 * to the faulty definition to be displayed through nested calls of
22680 * user-defined subs.)
22682 * The caller should NOT try to free any returned inversion list.
22684 * Other parameters will be set on return as described below */
22686 const char * const name, /* The first non-blank in the \p{}, \P{} */
22687 const Size_t name_len, /* Its length in bytes, not including any
22689 const bool is_utf8, /* ? Is 'name' encoded in UTF-8 */
22690 const bool to_fold, /* ? Is this under /i */
22691 const bool runtime, /* TRUE if this is being called at run time */
22692 const bool deferrable, /* TRUE if it's ok for the definition to not be
22693 known at this call */
22694 bool *user_defined_ptr, /* Upon return from this function it will be
22695 set to TRUE if any component is a
22696 user-defined property */
22697 SV * msg, /* Any error or warning msg(s) are appended to
22699 const STRLEN level) /* Recursion level of this call */
22702 char* lookup_name; /* normalized name for lookup in our tables */
22703 unsigned lookup_len; /* Its length */
22704 bool stricter = FALSE; /* Some properties have stricter name
22705 normalization rules, which we decide upon
22706 based on parsing */
22708 /* nv= or numeric_value=, or possibly one of the cjk numeric properties
22709 * (though it requires extra effort to download them from Unicode and
22710 * compile perl to know about them) */
22711 bool is_nv_type = FALSE;
22713 unsigned int i, j = 0;
22714 int equals_pos = -1; /* Where the '=' is found, or negative if none */
22715 int slash_pos = -1; /* Where the '/' is found, or negative if none */
22716 int table_index = 0; /* The entry number for this property in the table
22717 of all Unicode property names */
22718 bool starts_with_Is = FALSE; /* ? Does the name start with 'Is' */
22719 Size_t lookup_offset = 0; /* Used to ignore the first few characters of
22720 the normalized name in certain situations */
22721 Size_t non_pkg_begin = 0; /* Offset of first byte in 'name' that isn't
22722 part of a package name */
22723 bool could_be_user_defined = TRUE; /* ? Could this be a user-defined
22724 property rather than a Unicode
22726 SV * prop_definition = NULL; /* The returned definition of 'name' or NULL
22727 if an error. If it is an inversion list,
22728 it is the definition. Otherwise it is a
22729 string containing the fully qualified sub
22731 SV * fq_name = NULL; /* For user-defined properties, the fully
22733 bool invert_return = FALSE; /* ? Do we need to complement the result before
22736 PERL_ARGS_ASSERT_PARSE_UNIPROP_STRING;
22738 /* The input will be normalized into 'lookup_name' */
22739 Newx(lookup_name, name_len, char);
22740 SAVEFREEPV(lookup_name);
22742 /* Parse the input. */
22743 for (i = 0; i < name_len; i++) {
22744 char cur = name[i];
22746 /* Most of the characters in the input will be of this ilk, being parts
22748 if (isIDCONT_A(cur)) {
22750 /* Case differences are ignored. Our lookup routine assumes
22751 * everything is lowercase, so normalize to that */
22752 if (isUPPER_A(cur)) {
22753 lookup_name[j++] = toLOWER_A(cur);
22757 if (cur == '_') { /* Don't include these in the normalized name */
22761 lookup_name[j++] = cur;
22763 /* The first character in a user-defined name must be of this type.
22765 if (i - non_pkg_begin == 0 && ! isIDFIRST_A(cur)) {
22766 could_be_user_defined = FALSE;
22772 /* Here, the character is not something typically in a name, But these
22773 * two types of characters (and the '_' above) can be freely ignored in
22774 * most situations. Later it may turn out we shouldn't have ignored
22775 * them, and we have to reparse, but we don't have enough information
22776 * yet to make that decision */
22777 if (cur == '-' || isSPACE_A(cur)) {
22778 could_be_user_defined = FALSE;
22782 /* An equals sign or single colon mark the end of the first part of
22783 * the property name */
22785 || (cur == ':' && (i >= name_len - 1 || name[i+1] != ':')))
22787 lookup_name[j++] = '='; /* Treat the colon as an '=' */
22788 equals_pos = j; /* Note where it occurred in the input */
22789 could_be_user_defined = FALSE;
22793 /* Otherwise, this character is part of the name. */
22794 lookup_name[j++] = cur;
22796 /* Here it isn't a single colon, so if it is a colon, it must be a
22800 /* A double colon should be a package qualifier. We note its
22801 * position and continue. Note that one could have
22802 * pkg1::pkg2::...::foo
22803 * so that the position at the end of the loop will be just after
22804 * the final qualifier */
22807 non_pkg_begin = i + 1;
22808 lookup_name[j++] = ':';
22810 else { /* Only word chars (and '::') can be in a user-defined name */
22811 could_be_user_defined = FALSE;
22813 } /* End of parsing through the lhs of the property name (or all of it if
22816 #define STRLENs(s) (sizeof("" s "") - 1)
22818 /* If there is a single package name 'utf8::', it is ambiguous. It could
22819 * be for a user-defined property, or it could be a Unicode property, as
22820 * all of them are considered to be for that package. For the purposes of
22821 * parsing the rest of the property, strip it off */
22822 if (non_pkg_begin == STRLENs("utf8::") && memBEGINPs(name, name_len, "utf8::")) {
22823 lookup_name += STRLENs("utf8::");
22824 j -= STRLENs("utf8::");
22825 equals_pos -= STRLENs("utf8::");
22828 /* Here, we are either done with the whole property name, if it was simple;
22829 * or are positioned just after the '=' if it is compound. */
22831 if (equals_pos >= 0) {
22832 assert(! stricter); /* We shouldn't have set this yet */
22834 /* Space immediately after the '=' is ignored */
22836 for (; i < name_len; i++) {
22837 if (! isSPACE_A(name[i])) {
22842 /* Most punctuation after the equals indicates a subpattern, like
22844 if ( isPUNCT_A(name[i])
22850 /* Find the property. The table includes the equals sign, so we
22852 table_index = match_uniprop((U8 *) lookup_name, j);
22854 const char * const * prop_values
22855 = UNI_prop_value_ptrs[table_index];
22857 Size_t subpattern_len;
22858 REGEXP * subpattern_re;
22859 char open = name[i++];
22861 const char * pos_in_brackets;
22864 /* A backslash means the real delimitter is the next character.
22866 if (open == '\\') {
22871 /* This data structure is constructed so that the matching
22872 * closing bracket is 3 past its matching opening. The second
22873 * set of closing is so that if the opening is something like
22874 * ']', the closing will be that as well. Something similar is
22875 * done in toke.c */
22876 pos_in_brackets = strchr("([<)]>)]>", open);
22877 close = (pos_in_brackets) ? pos_in_brackets[3] : open;
22880 || name[name_len-1] != close
22881 || (escaped && name[name_len-2] != '\\'))
22883 sv_catpvs(msg, "Unicode property wildcard not terminated");
22884 goto append_name_to_msg;
22887 Perl_ck_warner_d(aTHX_
22888 packWARN(WARN_EXPERIMENTAL__UNIPROP_WILDCARDS),
22889 "The Unicode property wildcards feature is experimental");
22891 /* Now create and compile the wildcard subpattern. Use /iaa
22892 * because nothing outside of ASCII will match, and it the
22893 * property values should all match /i. Note that when the
22894 * pattern fails to compile, our added text to the user's
22895 * pattern will be displayed to the user, which is not so
22897 subpattern_len = name_len - i - 1 - escaped;
22898 subpattern = Perl_newSVpvf(aTHX_ "(?iaa:%.*s)",
22899 (unsigned) subpattern_len,
22901 subpattern = sv_2mortal(subpattern);
22902 subpattern_re = re_compile(subpattern, 0);
22903 assert(subpattern_re); /* Should have died if didn't compile
22906 /* For each legal property value, see if the supplied pattern
22908 while (*prop_values) {
22909 const char * const entry = *prop_values;
22910 const Size_t len = strlen(entry);
22911 SV* entry_sv = newSVpvn_flags(entry, len, SVs_TEMP);
22913 if (pregexec(subpattern_re,
22915 (char *) entry + len,
22919 { /* Here, matched. Add to the returned list */
22920 Size_t total_len = j + len;
22921 SV * sub_invlist = NULL;
22922 char * this_string;
22924 /* We know this is a legal \p{property=value}. Call
22925 * the function to return the list of code points that
22927 Newxz(this_string, total_len + 1, char);
22928 Copy(lookup_name, this_string, j, char);
22929 my_strlcat(this_string, entry, total_len + 1);
22930 SAVEFREEPV(this_string);
22931 sub_invlist = parse_uniprop_string(this_string,
22940 _invlist_union(prop_definition, sub_invlist,
22944 prop_values++; /* Next iteration, look at next propvalue */
22945 } /* End of looking through property values; (the data
22946 structure is terminated by a NULL ptr) */
22948 SvREFCNT_dec_NN(subpattern_re);
22950 if (prop_definition) {
22951 return prop_definition;
22954 sv_catpvs(msg, "No Unicode property value wildcard matches:");
22955 goto append_name_to_msg;
22958 /* Here's how khw thinks we should proceed to handle the properties
22959 * not yet done: Bidi Mirroring Glyph
22960 Bidi Paired Bracket
22961 Case Folding (both full and simple)
22962 Decomposition Mapping
22963 Equivalent Unified Ideograph
22966 Lowercase Mapping (both full and simple)
22968 Titlecase Mapping (both full and simple)
22969 Uppercase Mapping (both full and simple)
22970 * Move the part that looks at the property values into a perl
22971 * script, like utf8_heavy.pl is done. This makes things somewhat
22972 * easier, but most importantly, it avoids always adding all these
22973 * strings to the memory usage when the feature is little-used.
22975 * The property values would all be concatenated into a single
22976 * string per property with each value on a separate line, and the
22977 * code point it's for on alternating lines. Then we match the
22978 * user's input pattern m//mg, without having to worry about their
22979 * uses of '^' and '$'. Only the values that aren't the default
22980 * would be in the strings. Code points would be in UTF-8. The
22981 * search pattern that we would construct would look like
22982 * (?: \n (code-point_re) \n (?aam: user-re ) \n )
22983 * And so $1 would contain the code point that matched the user-re.
22984 * For properties where the default is the code point itself, such
22985 * as any of the case changing mappings, the string would otherwise
22986 * consist of all Unicode code points in UTF-8 strung together.
22987 * This would be impractical. So instead, examine their compiled
22988 * pattern, looking at the ssc. If none, reject the pattern as an
22989 * error. Otherwise run the pattern against every code point in
22990 * the ssc. The ssc is kind of like tr18's 3.9 Possible Match Sets
22991 * And it might be good to create an API to return the ssc.
22993 * For the name properties, a new function could be created in
22994 * charnames which essentially does the same thing as above,
22995 * sharing Name.pl with the other charname functions. Don't know
22996 * about loose name matching, or algorithmically determined names.
22997 * Decomposition.pl similarly.
22999 * It might be that a new pattern modifier would have to be
23000 * created, like /t for resTricTed, which changed the behavior of
23001 * some constructs in their subpattern, like \A. */
23002 } /* End of is a wildcard subppattern */
23005 /* Certain properties whose values are numeric need special handling.
23006 * They may optionally be prefixed by 'is'. Ignore that prefix for the
23007 * purposes of checking if this is one of those properties */
23008 if (memBEGINPs(lookup_name, j, "is")) {
23012 /* Then check if it is one of these specially-handled properties. The
23013 * possibilities are hard-coded because easier this way, and the list
23014 * is unlikely to change.
23016 * All numeric value type properties are of this ilk, and are also
23017 * special in a different way later on. So find those first. There
23018 * are several numeric value type properties in the Unihan DB (which is
23019 * unlikely to be compiled with perl, but we handle it here in case it
23020 * does get compiled). They all end with 'numeric'. The interiors
23021 * aren't checked for the precise property. This would stop working if
23022 * a cjk property were to be created that ended with 'numeric' and
23023 * wasn't a numeric type */
23024 is_nv_type = memEQs(lookup_name + lookup_offset,
23025 j - 1 - lookup_offset, "numericvalue")
23026 || memEQs(lookup_name + lookup_offset,
23027 j - 1 - lookup_offset, "nv")
23028 || ( memENDPs(lookup_name + lookup_offset,
23029 j - 1 - lookup_offset, "numeric")
23030 && ( memBEGINPs(lookup_name + lookup_offset,
23031 j - 1 - lookup_offset, "cjk")
23032 || memBEGINPs(lookup_name + lookup_offset,
23033 j - 1 - lookup_offset, "k")));
23035 || memEQs(lookup_name + lookup_offset,
23036 j - 1 - lookup_offset, "canonicalcombiningclass")
23037 || memEQs(lookup_name + lookup_offset,
23038 j - 1 - lookup_offset, "ccc")
23039 || memEQs(lookup_name + lookup_offset,
23040 j - 1 - lookup_offset, "age")
23041 || memEQs(lookup_name + lookup_offset,
23042 j - 1 - lookup_offset, "in")
23043 || memEQs(lookup_name + lookup_offset,
23044 j - 1 - lookup_offset, "presentin"))
23048 /* Since the stuff after the '=' is a number, we can't throw away
23049 * '-' willy-nilly, as those could be a minus sign. Other stricter
23050 * rules also apply. However, these properties all can have the
23051 * rhs not be a number, in which case they contain at least one
23052 * alphabetic. In those cases, the stricter rules don't apply.
23053 * But the numeric type properties can have the alphas [Ee] to
23054 * signify an exponent, and it is still a number with stricter
23055 * rules. So look for an alpha that signifies not-strict */
23057 for (k = i; k < name_len; k++) {
23058 if ( isALPHA_A(name[k])
23059 && (! is_nv_type || ! isALPHA_FOLD_EQ(name[k], 'E')))
23069 /* A number may have a leading '+' or '-'. The latter is retained
23071 if (name[i] == '+') {
23074 else if (name[i] == '-') {
23075 lookup_name[j++] = '-';
23079 /* Skip leading zeros including single underscores separating the
23080 * zeros, or between the final leading zero and the first other
23082 for (; i < name_len - 1; i++) {
23083 if ( name[i] != '0'
23084 && (name[i] != '_' || ! isDIGIT_A(name[i+1])))
23091 else { /* No '=' */
23093 /* Only a few properties without an '=' should be parsed with stricter
23094 * rules. The list is unlikely to change. */
23095 if ( memBEGINPs(lookup_name, j, "perl")
23096 && memNEs(lookup_name + 4, j - 4, "space")
23097 && memNEs(lookup_name + 4, j - 4, "word"))
23101 /* We set the inputs back to 0 and the code below will reparse,
23107 /* Here, we have either finished the property, or are positioned to parse
23108 * the remainder, and we know if stricter rules apply. Finish out, if not
23110 for (; i < name_len; i++) {
23111 char cur = name[i];
23113 /* In all instances, case differences are ignored, and we normalize to
23115 if (isUPPER_A(cur)) {
23116 lookup_name[j++] = toLOWER(cur);
23120 /* An underscore is skipped, but not under strict rules unless it
23121 * separates two digits */
23124 && ( i == 0 || (int) i == equals_pos || i == name_len- 1
23125 || ! isDIGIT_A(name[i-1]) || ! isDIGIT_A(name[i+1])))
23127 lookup_name[j++] = '_';
23132 /* Hyphens are skipped except under strict */
23133 if (cur == '-' && ! stricter) {
23137 /* XXX Bug in documentation. It says white space skipped adjacent to
23138 * non-word char. Maybe we should, but shouldn't skip it next to a dot
23140 if (isSPACE_A(cur) && ! stricter) {
23144 lookup_name[j++] = cur;
23146 /* Unless this is a non-trailing slash, we are done with it */
23147 if (i >= name_len - 1 || cur != '/') {
23153 /* A slash in the 'numeric value' property indicates that what follows
23154 * is a denominator. It can have a leading '+' and '0's that should be
23155 * skipped. But we have never allowed a negative denominator, so treat
23156 * a minus like every other character. (No need to rule out a second
23157 * '/', as that won't match anything anyway */
23160 if (i < name_len && name[i] == '+') {
23164 /* Skip leading zeros including underscores separating digits */
23165 for (; i < name_len - 1; i++) {
23166 if ( name[i] != '0'
23167 && (name[i] != '_' || ! isDIGIT_A(name[i+1])))
23173 /* Store the first real character in the denominator */
23174 if (i < name_len) {
23175 lookup_name[j++] = name[i];
23180 /* Here are completely done parsing the input 'name', and 'lookup_name'
23181 * contains a copy, normalized.
23183 * This special case is grandfathered in: 'L_' and 'GC=L_' are accepted and
23184 * different from without the underscores. */
23185 if ( ( UNLIKELY(memEQs(lookup_name, j, "l"))
23186 || UNLIKELY(memEQs(lookup_name, j, "gc=l")))
23187 && UNLIKELY(name[name_len-1] == '_'))
23189 lookup_name[j++] = '&';
23192 /* If the original input began with 'In' or 'Is', it could be a subroutine
23193 * call to a user-defined property instead of a Unicode property name. */
23194 if ( name_len - non_pkg_begin > 2
23195 && name[non_pkg_begin+0] == 'I'
23196 && (name[non_pkg_begin+1] == 'n' || name[non_pkg_begin+1] == 's'))
23198 /* Names that start with In have different characterstics than those
23199 * that start with Is */
23200 if (name[non_pkg_begin+1] == 's') {
23201 starts_with_Is = TRUE;
23205 could_be_user_defined = FALSE;
23208 if (could_be_user_defined) {
23211 /* If the user defined property returns the empty string, it could
23212 * easily be because the pattern is being compiled before the data it
23213 * actually needs to compile is available. This could be argued to be
23214 * a bug in the perl code, but this is a change of behavior for Perl,
23215 * so we handle it. This means that intentionally returning nothing
23216 * will not be resolved until runtime */
23217 bool empty_return = FALSE;
23219 /* Here, the name could be for a user defined property, which are
23220 * implemented as subs. */
23221 user_sub = get_cvn_flags(name, name_len, 0);
23223 const char insecure[] = "Insecure user-defined property";
23225 /* Here, there is a sub by the correct name. Normally we call it
23226 * to get the property definition */
23228 SV * user_sub_sv = MUTABLE_SV(user_sub);
23229 SV * error; /* Any error returned by calling 'user_sub' */
23230 SV * key; /* The key into the hash of user defined sub names
23233 SV ** saved_user_prop_ptr; /* Hash entry for this property */
23235 /* How many times to retry when another thread is in the middle of
23236 * expanding the same definition we want */
23237 PERL_INT_FAST8_T retry_countdown = 10;
23239 DECLARATION_FOR_GLOBAL_CONTEXT;
23241 /* If we get here, we know this property is user-defined */
23242 *user_defined_ptr = TRUE;
23244 /* We refuse to call a potentially tainted subroutine; returning an
23247 if (SvCUR(msg) > 0) sv_catpvs(msg, "; ");
23248 sv_catpvn(msg, insecure, sizeof(insecure) - 1);
23249 goto append_name_to_msg;
23252 /* In principal, we only call each subroutine property definition
23253 * once during the life of the program. This guarantees that the
23254 * property definition never changes. The results of the single
23255 * sub call are stored in a hash, which is used instead for future
23256 * references to this property. The property definition is thus
23257 * immutable. But, to allow the user to have a /i-dependent
23258 * definition, we call the sub once for non-/i, and once for /i,
23259 * should the need arise, passing the /i status as a parameter.
23261 * We start by constructing the hash key name, consisting of the
23262 * fully qualified subroutine name, preceded by the /i status, so
23263 * that there is a key for /i and a different key for non-/i */
23264 key = newSVpvn(((to_fold) ? "1" : "0"), 1);
23265 fq_name = S_get_fq_name(aTHX_ name, name_len, is_utf8,
23266 non_pkg_begin != 0);
23267 sv_catsv(key, fq_name);
23270 /* We only call the sub once throughout the life of the program
23271 * (with the /i, non-/i exception noted above). That means the
23272 * hash must be global and accessible to all threads. It is
23273 * created at program start-up, before any threads are created, so
23274 * is accessible to all children. But this creates some
23277 * 1) The keys can't be shared, or else problems arise; sharing is
23278 * turned off at hash creation time
23279 * 2) All SVs in it are there for the remainder of the life of the
23280 * program, and must be created in the same interpreter context
23281 * as the hash, or else they will be freed from the wrong pool
23282 * at global destruction time. This is handled by switching to
23283 * the hash's context to create each SV going into it, and then
23284 * immediately switching back
23285 * 3) All accesses to the hash must be controlled by a mutex, to
23286 * prevent two threads from getting an unstable state should
23287 * they simultaneously be accessing it. The code below is
23288 * crafted so that the mutex is locked whenever there is an
23289 * access and unlocked only when the next stable state is
23292 * The hash stores either the definition of the property if it was
23293 * valid, or, if invalid, the error message that was raised. We
23294 * use the type of SV to distinguish.
23296 * There's also the need to guard against the definition expansion
23297 * from infinitely recursing. This is handled by storing the aTHX
23298 * of the expanding thread during the expansion. Again the SV type
23299 * is used to distinguish this from the other two cases. If we
23300 * come to here and the hash entry for this property is our aTHX,
23301 * it means we have recursed, and the code assumes that we would
23302 * infinitely recurse, so instead stops and raises an error.
23303 * (Any recursion has always been treated as infinite recursion in
23306 * If instead, the entry is for a different aTHX, it means that
23307 * that thread has gotten here first, and hasn't finished expanding
23308 * the definition yet. We just have to wait until it is done. We
23309 * sleep and retry a few times, returning an error if the other
23310 * thread doesn't complete. */
23313 USER_PROP_MUTEX_LOCK;
23315 /* If we have an entry for this key, the subroutine has already
23316 * been called once with this /i status. */
23317 saved_user_prop_ptr = hv_fetch(PL_user_def_props,
23318 SvPVX(key), SvCUR(key), 0);
23319 if (saved_user_prop_ptr) {
23321 /* If the saved result is an inversion list, it is the valid
23322 * definition of this property */
23323 if (is_invlist(*saved_user_prop_ptr)) {
23324 prop_definition = *saved_user_prop_ptr;
23326 /* The SV in the hash won't be removed until global
23327 * destruction, so it is stable and we can unlock */
23328 USER_PROP_MUTEX_UNLOCK;
23330 /* The caller shouldn't try to free this SV */
23331 return prop_definition;
23334 /* Otherwise, if it is a string, it is the error message
23335 * that was returned when we first tried to evaluate this
23336 * property. Fail, and append the message */
23337 if (SvPOK(*saved_user_prop_ptr)) {
23338 if (SvCUR(msg) > 0) sv_catpvs(msg, "; ");
23339 sv_catsv(msg, *saved_user_prop_ptr);
23341 /* The SV in the hash won't be removed until global
23342 * destruction, so it is stable and we can unlock */
23343 USER_PROP_MUTEX_UNLOCK;
23348 assert(SvIOK(*saved_user_prop_ptr));
23350 /* Here, we have an unstable entry in the hash. Either another
23351 * thread is in the middle of expanding the property's
23352 * definition, or we are ourselves recursing. We use the aTHX
23353 * in it to distinguish */
23354 if (SvIV(*saved_user_prop_ptr) != PTR2IV(CUR_CONTEXT)) {
23356 /* Here, it's another thread doing the expanding. We've
23357 * looked as much as we are going to at the contents of the
23358 * hash entry. It's safe to unlock. */
23359 USER_PROP_MUTEX_UNLOCK;
23361 /* Retry a few times */
23362 if (retry_countdown-- > 0) {
23367 if (SvCUR(msg) > 0) sv_catpvs(msg, "; ");
23368 sv_catpvs(msg, "Timeout waiting for another thread to "
23370 goto append_name_to_msg;
23373 /* Here, we are recursing; don't dig any deeper */
23374 USER_PROP_MUTEX_UNLOCK;
23376 if (SvCUR(msg) > 0) sv_catpvs(msg, "; ");
23378 "Infinite recursion in user-defined property");
23379 goto append_name_to_msg;
23382 /* Here, this thread has exclusive control, and there is no entry
23383 * for this property in the hash. So we have the go ahead to
23384 * expand the definition ourselves. */
23386 PUSHSTACKi(PERLSI_MAGIC);
23389 /* Create a temporary placeholder in the hash to detect recursion
23391 SWITCH_TO_GLOBAL_CONTEXT;
23392 placeholder= newSVuv(PTR2IV(ORIGINAL_CONTEXT));
23393 (void) hv_store_ent(PL_user_def_props, key, placeholder, 0);
23396 /* Now that we have a placeholder, we can let other threads
23398 USER_PROP_MUTEX_UNLOCK;
23400 /* Make sure the placeholder always gets destroyed */
23401 SAVEDESTRUCTOR_X(S_delete_recursion_entry, SvPVX(key));
23406 /* Call the user's function, with the /i status as a parameter.
23407 * Note that we have gone to a lot of trouble to keep this call
23408 * from being within the locked mutex region. */
23409 XPUSHs(boolSV(to_fold));
23412 /* The following block was taken from swash_init(). Presumably
23413 * they apply to here as well, though we no longer use a swash --
23417 /* We might get here via a subroutine signature which uses a utf8
23418 * parameter name, at which point PL_subname will have been set
23419 * but not yet used. */
23420 save_item(PL_subname);
23422 (void) call_sv(user_sub_sv, G_EVAL|G_SCALAR);
23427 if (TAINT_get || SvTRUE(error)) {
23428 if (SvCUR(msg) > 0) sv_catpvs(msg, "; ");
23429 if (SvTRUE(error)) {
23430 sv_catpvs(msg, "Error \"");
23431 sv_catsv(msg, error);
23432 sv_catpvs(msg, "\"");
23435 if (SvTRUE(error)) sv_catpvs(msg, "; ");
23436 sv_catpvn(msg, insecure, sizeof(insecure) - 1);
23439 if (name_len > 0) {
23440 sv_catpvs(msg, " in expansion of ");
23441 Perl_sv_catpvf(aTHX_ msg, "%" UTF8f, UTF8fARG(is_utf8,
23447 prop_definition = NULL;
23449 else { /* G_SCALAR guarantees a single return value */
23450 SV * contents = POPs;
23452 /* The contents is supposed to be the expansion of the property
23453 * definition. If the definition is deferrable, and we got an
23454 * empty string back, set a flag to later defer it (after clean
23457 && (! SvPOK(contents) || SvCUR(contents) == 0))
23459 empty_return = TRUE;
23461 else { /* Otherwise, call a function to check for valid syntax,
23464 prop_definition = handle_user_defined_property(
23466 is_utf8, to_fold, runtime,
23468 contents, user_defined_ptr,
23474 /* Here, we have the results of the expansion. Delete the
23475 * placeholder, and if the definition is now known, replace it with
23476 * that definition. We need exclusive access to the hash, and we
23477 * can't let anyone else in, between when we delete the placeholder
23478 * and add the permanent entry */
23479 USER_PROP_MUTEX_LOCK;
23481 S_delete_recursion_entry(aTHX_ SvPVX(key));
23483 if ( ! empty_return
23484 && (! prop_definition || is_invlist(prop_definition)))
23486 /* If we got success we use the inversion list defining the
23487 * property; otherwise use the error message */
23488 SWITCH_TO_GLOBAL_CONTEXT;
23489 (void) hv_store_ent(PL_user_def_props,
23492 ? newSVsv(prop_definition)
23498 /* All done, and the hash now has a permanent entry for this
23499 * property. Give up exclusive control */
23500 USER_PROP_MUTEX_UNLOCK;
23506 if (empty_return) {
23507 goto definition_deferred;
23510 if (prop_definition) {
23512 /* If the definition is for something not known at this time,
23513 * we toss it, and go return the main property name, as that's
23514 * the one the user will be aware of */
23515 if (! is_invlist(prop_definition)) {
23516 SvREFCNT_dec_NN(prop_definition);
23517 goto definition_deferred;
23520 sv_2mortal(prop_definition);
23524 return prop_definition;
23526 } /* End of calling the subroutine for the user-defined property */
23527 } /* End of it could be a user-defined property */
23529 /* Here it wasn't a user-defined property that is known at this time. See
23530 * if it is a Unicode property */
23532 lookup_len = j; /* This is a more mnemonic name than 'j' */
23534 /* Get the index into our pointer table of the inversion list corresponding
23535 * to the property */
23536 table_index = match_uniprop((U8 *) lookup_name, lookup_len);
23538 /* If it didn't find the property ... */
23539 if (table_index == 0) {
23541 /* Try again stripping off any initial 'Is'. This is because we
23542 * promise that an initial Is is optional. The same isn't true of
23543 * names that start with 'In'. Those can match only blocks, and the
23544 * lookup table already has those accounted for. */
23545 if (starts_with_Is) {
23551 table_index = match_uniprop((U8 *) lookup_name, lookup_len);
23554 if (table_index == 0) {
23557 /* Here, we didn't find it. If not a numeric type property, and
23558 * can't be a user-defined one, it isn't a legal property */
23559 if (! is_nv_type) {
23560 if (! could_be_user_defined) {
23564 /* Here, the property name is legal as a user-defined one. At
23565 * compile time, it might just be that the subroutine for that
23566 * property hasn't been encountered yet, but at runtime, it's
23567 * an error to try to use an undefined one */
23568 if (! deferrable) {
23569 if (SvCUR(msg) > 0) sv_catpvs(msg, "; ");
23570 sv_catpvs(msg, "Unknown user-defined property name");
23571 goto append_name_to_msg;
23574 goto definition_deferred;
23575 } /* End of isn't a numeric type property */
23577 /* The numeric type properties need more work to decide. What we
23578 * do is make sure we have the number in canonical form and look
23581 if (slash_pos < 0) { /* No slash */
23583 /* When it isn't a rational, take the input, convert it to a
23584 * NV, then create a canonical string representation of that
23588 SSize_t value_len = lookup_len - equals_pos;
23590 /* Get the value */
23591 if ( value_len <= 0
23592 || my_atof3(lookup_name + equals_pos, &value,
23594 != lookup_name + lookup_len)
23599 /* If the value is an integer, the canonical value is integral
23601 if (Perl_ceil(value) == value) {
23602 canonical = Perl_form(aTHX_ "%.*s%.0" NVff,
23603 equals_pos, lookup_name, value);
23605 else { /* Otherwise, it is %e with a known precision */
23608 canonical = Perl_form(aTHX_ "%.*s%.*" NVef,
23609 equals_pos, lookup_name,
23610 PL_E_FORMAT_PRECISION, value);
23612 /* The exponent generated is expecting two digits, whereas
23613 * %e on some systems will generate three. Remove leading
23614 * zeros in excess of 2 from the exponent. We start
23615 * looking for them after the '=' */
23616 exp_ptr = strchr(canonical + equals_pos, 'e');
23618 char * cur_ptr = exp_ptr + 2; /* past the 'e[+-]' */
23619 SSize_t excess_exponent_len = strlen(cur_ptr) - 2;
23621 assert(*(cur_ptr - 1) == '-' || *(cur_ptr - 1) == '+');
23623 if (excess_exponent_len > 0) {
23624 SSize_t leading_zeros = strspn(cur_ptr, "0");
23625 SSize_t excess_leading_zeros
23626 = MIN(leading_zeros, excess_exponent_len);
23627 if (excess_leading_zeros > 0) {
23628 Move(cur_ptr + excess_leading_zeros,
23630 strlen(cur_ptr) - excess_leading_zeros
23631 + 1, /* Copy the NUL as well */
23638 else { /* Has a slash. Create a rational in canonical form */
23639 UV numerator, denominator, gcd, trial;
23640 const char * end_ptr;
23641 const char * sign = "";
23643 /* We can't just find the numerator, denominator, and do the
23644 * division, then use the method above, because that is
23645 * inexact. And the input could be a rational that is within
23646 * epsilon (given our precision) of a valid rational, and would
23647 * then incorrectly compare valid.
23649 * We're only interested in the part after the '=' */
23650 const char * this_lookup_name = lookup_name + equals_pos;
23651 lookup_len -= equals_pos;
23652 slash_pos -= equals_pos;
23654 /* Handle any leading minus */
23655 if (this_lookup_name[0] == '-') {
23657 this_lookup_name++;
23662 /* Convert the numerator to numeric */
23663 end_ptr = this_lookup_name + slash_pos;
23664 if (! grok_atoUV(this_lookup_name, &numerator, &end_ptr)) {
23668 /* It better have included all characters before the slash */
23669 if (*end_ptr != '/') {
23673 /* Set to look at just the denominator */
23674 this_lookup_name += slash_pos;
23675 lookup_len -= slash_pos;
23676 end_ptr = this_lookup_name + lookup_len;
23678 /* Convert the denominator to numeric */
23679 if (! grok_atoUV(this_lookup_name, &denominator, &end_ptr)) {
23683 /* It better be the rest of the characters, and don't divide by
23685 if ( end_ptr != this_lookup_name + lookup_len
23686 || denominator == 0)
23691 /* Get the greatest common denominator using
23692 http://en.wikipedia.org/wiki/Euclidean_algorithm */
23694 trial = denominator;
23695 while (trial != 0) {
23697 trial = gcd % trial;
23701 /* If already in lowest possible terms, we have already tried
23702 * looking this up */
23707 /* Reduce the rational, which should put it in canonical form
23710 denominator /= gcd;
23712 canonical = Perl_form(aTHX_ "%.*s%s%" UVuf "/%" UVuf,
23713 equals_pos, lookup_name, sign, numerator, denominator);
23716 /* Here, we have the number in canonical form. Try that */
23717 table_index = match_uniprop((U8 *) canonical, strlen(canonical));
23718 if (table_index == 0) {
23721 } /* End of still didn't find the property in our table */
23722 } /* End of didn't find the property in our table */
23724 /* Here, we have a non-zero return, which is an index into a table of ptrs.
23725 * A negative return signifies that the real index is the absolute value,
23726 * but the result needs to be inverted */
23727 if (table_index < 0) {
23728 invert_return = TRUE;
23729 table_index = -table_index;
23732 /* Out-of band indices indicate a deprecated property. The proper index is
23733 * modulo it with the table size. And dividing by the table size yields
23734 * an offset into a table constructed by regen/mk_invlists.pl to contain
23735 * the corresponding warning message */
23736 if (table_index > MAX_UNI_KEYWORD_INDEX) {
23737 Size_t warning_offset = table_index / MAX_UNI_KEYWORD_INDEX;
23738 table_index %= MAX_UNI_KEYWORD_INDEX;
23739 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED),
23740 "Use of '%.*s' in \\p{} or \\P{} is deprecated because: %s",
23741 (int) name_len, name, deprecated_property_msgs[warning_offset]);
23744 /* In a few properties, a different property is used under /i. These are
23745 * unlikely to change, so are hard-coded here. */
23747 if ( table_index == UNI_XPOSIXUPPER
23748 || table_index == UNI_XPOSIXLOWER
23749 || table_index == UNI_TITLE)
23751 table_index = UNI_CASED;
23753 else if ( table_index == UNI_UPPERCASELETTER
23754 || table_index == UNI_LOWERCASELETTER
23755 # ifdef UNI_TITLECASELETTER /* Missing from early Unicodes */
23756 || table_index == UNI_TITLECASELETTER
23759 table_index = UNI_CASEDLETTER;
23761 else if ( table_index == UNI_POSIXUPPER
23762 || table_index == UNI_POSIXLOWER)
23764 table_index = UNI_POSIXALPHA;
23768 /* Create and return the inversion list */
23769 prop_definition =_new_invlist_C_array(uni_prop_ptrs[table_index]);
23770 sv_2mortal(prop_definition);
23773 /* See if there is a private use override to add to this definition */
23775 COPHH * hinthash = (IN_PERL_COMPILETIME)
23776 ? CopHINTHASH_get(&PL_compiling)
23777 : CopHINTHASH_get(PL_curcop);
23778 SV * pu_overrides = cophh_fetch_pv(hinthash, "private_use", 0, 0);
23780 if (UNLIKELY(pu_overrides && SvPOK(pu_overrides))) {
23782 /* See if there is an element in the hints hash for this table */
23783 SV * pu_lookup = Perl_newSVpvf(aTHX_ "%d=", table_index);
23784 const char * pos = strstr(SvPVX(pu_overrides), SvPVX(pu_lookup));
23788 SV * pu_definition;
23790 SV * expanded_prop_definition =
23791 sv_2mortal(invlist_clone(prop_definition, NULL));
23793 /* If so, it's definition is the string from here to the next
23794 * \a character. And its format is the same as a user-defined
23796 pos += SvCUR(pu_lookup);
23797 pu_definition = newSVpvn(pos, strchr(pos, '\a') - pos);
23798 pu_invlist = handle_user_defined_property(lookup_name,
23801 0, /* Not folded */
23809 if (SvCUR(msg) > 0) sv_catpvs(msg, "; ");
23810 sv_catpvs(msg, "Insecure private-use override");
23811 goto append_name_to_msg;
23814 /* For now, as a safety measure, make sure that it doesn't
23815 * override non-private use code points */
23816 _invlist_intersection(pu_invlist, PL_Private_Use, &pu_invlist);
23818 /* Add it to the list to be returned */
23819 _invlist_union(prop_definition, pu_invlist,
23820 &expanded_prop_definition);
23821 prop_definition = expanded_prop_definition;
23822 Perl_ck_warner_d(aTHX_ packWARN(WARN_EXPERIMENTAL__PRIVATE_USE), "The private_use feature is experimental");
23827 if (invert_return) {
23828 _invlist_invert(prop_definition);
23830 return prop_definition;
23834 if (non_pkg_begin != 0) {
23835 if (SvCUR(msg) > 0) sv_catpvs(msg, "; ");
23836 sv_catpvs(msg, "Illegal user-defined property name");
23839 if (SvCUR(msg) > 0) sv_catpvs(msg, "; ");
23840 sv_catpvs(msg, "Can't find Unicode property definition");
23844 append_name_to_msg:
23846 const char * prefix = (runtime && level == 0) ? " \\p{" : " \"";
23847 const char * suffix = (runtime && level == 0) ? "}" : "\"";
23849 sv_catpv(msg, prefix);
23850 Perl_sv_catpvf(aTHX_ msg, "%" UTF8f, UTF8fARG(is_utf8, name_len, name));
23851 sv_catpv(msg, suffix);
23856 definition_deferred:
23858 /* Here it could yet to be defined, so defer evaluation of this
23859 * until its needed at runtime. We need the fully qualified property name
23860 * to avoid ambiguity, and a trailing newline */
23862 fq_name = S_get_fq_name(aTHX_ name, name_len, is_utf8,
23863 non_pkg_begin != 0 /* If has "::" */
23866 sv_catpvs(fq_name, "\n");
23868 *user_defined_ptr = TRUE;
23875 * ex: set ts=8 sts=4 sw=4 et: