1 /* Extended regular expression matching and search library.
2 Copyright (C) 2002, 2003, 2004, 2005, 2006 Free Software Foundation, Inc.
3 This file is part of the GNU C Library.
4 Contributed by Isamu Hasegawa <isamu@yamato.ibm.com>.
6 This program is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2, or (at your option)
11 This program is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License along
17 with this program; if not, write to the Free Software Foundation,
18 Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. */
20 static reg_errcode_t
match_ctx_init (re_match_context_t
*cache
, int eflags
,
21 Idx n
) internal_function
;
22 static void match_ctx_clean (re_match_context_t
*mctx
) internal_function
;
23 static void match_ctx_free (re_match_context_t
*cache
) internal_function
;
24 static reg_errcode_t
match_ctx_add_entry (re_match_context_t
*cache
, Idx node
,
25 Idx str_idx
, Idx from
, Idx to
)
27 static Idx
search_cur_bkref_entry (const re_match_context_t
*mctx
, Idx str_idx
)
29 static reg_errcode_t
match_ctx_add_subtop (re_match_context_t
*mctx
, Idx node
,
30 Idx str_idx
) internal_function
;
31 static re_sub_match_last_t
* match_ctx_add_sublast (re_sub_match_top_t
*subtop
,
32 Idx node
, Idx str_idx
)
34 static void sift_ctx_init (re_sift_context_t
*sctx
, re_dfastate_t
**sifted_sts
,
35 re_dfastate_t
**limited_sts
, Idx last_node
,
38 static reg_errcode_t
re_search_internal (const regex_t
*preg
,
39 const char *string
, Idx length
,
40 Idx start
, Idx last_start
, Idx stop
,
41 size_t nmatch
, regmatch_t pmatch
[],
42 int eflags
) internal_function
;
43 static regoff_t
re_search_2_stub (struct re_pattern_buffer
*bufp
,
44 const char *string1
, Idx length1
,
45 const char *string2
, Idx length2
,
46 Idx start
, regoff_t range
,
47 struct re_registers
*regs
,
48 Idx stop
, bool ret_len
) internal_function
;
49 static regoff_t
re_search_stub (struct re_pattern_buffer
*bufp
,
50 const char *string
, Idx length
, Idx start
,
51 regoff_t range
, Idx stop
,
52 struct re_registers
*regs
,
53 bool ret_len
) internal_function
;
54 static unsigned int re_copy_regs (struct re_registers
*regs
, regmatch_t
*pmatch
,
55 Idx nregs
, int regs_allocated
)
57 static reg_errcode_t
prune_impossible_nodes (re_match_context_t
*mctx
)
59 static Idx
check_matching (re_match_context_t
*mctx
, bool fl_longest_match
,
60 Idx
*p_match_first
) internal_function
;
61 static Idx
check_halt_state_context (const re_match_context_t
*mctx
,
62 const re_dfastate_t
*state
, Idx idx
)
64 static void update_regs (const re_dfa_t
*dfa
, regmatch_t
*pmatch
,
65 regmatch_t
*prev_idx_match
, Idx cur_node
,
66 Idx cur_idx
, Idx nmatch
) internal_function
;
67 static reg_errcode_t
push_fail_stack (struct re_fail_stack_t
*fs
,
68 Idx str_idx
, Idx dest_node
, Idx nregs
,
70 re_node_set
*eps_via_nodes
)
72 static reg_errcode_t
set_regs (const regex_t
*preg
,
73 const re_match_context_t
*mctx
,
74 size_t nmatch
, regmatch_t
*pmatch
,
75 bool fl_backtrack
) internal_function
;
76 static reg_errcode_t
free_fail_stack_return (struct re_fail_stack_t
*fs
)
80 static int sift_states_iter_mb (const re_match_context_t
*mctx
,
81 re_sift_context_t
*sctx
,
82 Idx node_idx
, Idx str_idx
, Idx max_str_idx
)
84 #endif /* RE_ENABLE_I18N */
85 static reg_errcode_t
sift_states_backward (const re_match_context_t
*mctx
,
86 re_sift_context_t
*sctx
)
88 static reg_errcode_t
build_sifted_states (const re_match_context_t
*mctx
,
89 re_sift_context_t
*sctx
, Idx str_idx
,
90 re_node_set
*cur_dest
)
92 static reg_errcode_t
update_cur_sifted_state (const re_match_context_t
*mctx
,
93 re_sift_context_t
*sctx
,
95 re_node_set
*dest_nodes
)
97 static reg_errcode_t
add_epsilon_src_nodes (const re_dfa_t
*dfa
,
98 re_node_set
*dest_nodes
,
99 const re_node_set
*candidates
)
101 static bool check_dst_limits (const re_match_context_t
*mctx
,
102 const re_node_set
*limits
,
103 Idx dst_node
, Idx dst_idx
, Idx src_node
,
104 Idx src_idx
) internal_function
;
105 static int check_dst_limits_calc_pos_1 (const re_match_context_t
*mctx
,
106 int boundaries
, Idx subexp_idx
,
107 Idx from_node
, Idx bkref_idx
)
109 static int check_dst_limits_calc_pos (const re_match_context_t
*mctx
,
110 Idx limit
, Idx subexp_idx
,
111 Idx node
, Idx str_idx
,
112 Idx bkref_idx
) internal_function
;
113 static reg_errcode_t
check_subexp_limits (const re_dfa_t
*dfa
,
114 re_node_set
*dest_nodes
,
115 const re_node_set
*candidates
,
117 struct re_backref_cache_entry
*bkref_ents
,
118 Idx str_idx
) internal_function
;
119 static reg_errcode_t
sift_states_bkref (const re_match_context_t
*mctx
,
120 re_sift_context_t
*sctx
,
121 Idx str_idx
, const re_node_set
*candidates
)
123 static reg_errcode_t
merge_state_array (const re_dfa_t
*dfa
,
125 re_dfastate_t
**src
, Idx num
)
127 static re_dfastate_t
*find_recover_state (reg_errcode_t
*err
,
128 re_match_context_t
*mctx
) internal_function
;
129 static re_dfastate_t
*transit_state (reg_errcode_t
*err
,
130 re_match_context_t
*mctx
,
131 re_dfastate_t
*state
) internal_function
;
132 static re_dfastate_t
*merge_state_with_log (reg_errcode_t
*err
,
133 re_match_context_t
*mctx
,
134 re_dfastate_t
*next_state
)
136 static reg_errcode_t
check_subexp_matching_top (re_match_context_t
*mctx
,
137 re_node_set
*cur_nodes
,
138 Idx str_idx
) internal_function
;
140 static re_dfastate_t
*transit_state_sb (reg_errcode_t
*err
,
141 re_match_context_t
*mctx
,
142 re_dfastate_t
*pstate
)
145 #ifdef RE_ENABLE_I18N
146 static reg_errcode_t
transit_state_mb (re_match_context_t
*mctx
,
147 re_dfastate_t
*pstate
)
149 #endif /* RE_ENABLE_I18N */
150 static reg_errcode_t
transit_state_bkref (re_match_context_t
*mctx
,
151 const re_node_set
*nodes
)
153 static reg_errcode_t
get_subexp (re_match_context_t
*mctx
,
154 Idx bkref_node
, Idx bkref_str_idx
)
156 static reg_errcode_t
get_subexp_sub (re_match_context_t
*mctx
,
157 const re_sub_match_top_t
*sub_top
,
158 re_sub_match_last_t
*sub_last
,
159 Idx bkref_node
, Idx bkref_str
)
161 static Idx
find_subexp_node (const re_dfa_t
*dfa
, const re_node_set
*nodes
,
162 Idx subexp_idx
, int type
) internal_function
;
163 static reg_errcode_t
check_arrival (re_match_context_t
*mctx
,
164 state_array_t
*path
, Idx top_node
,
165 Idx top_str
, Idx last_node
, Idx last_str
,
166 int type
) internal_function
;
167 static reg_errcode_t
check_arrival_add_next_nodes (re_match_context_t
*mctx
,
169 re_node_set
*cur_nodes
,
170 re_node_set
*next_nodes
)
172 static reg_errcode_t
check_arrival_expand_ecl (const re_dfa_t
*dfa
,
173 re_node_set
*cur_nodes
,
174 Idx ex_subexp
, int type
)
176 static reg_errcode_t
check_arrival_expand_ecl_sub (const re_dfa_t
*dfa
,
177 re_node_set
*dst_nodes
,
178 Idx target
, Idx ex_subexp
,
179 int type
) internal_function
;
180 static reg_errcode_t
expand_bkref_cache (re_match_context_t
*mctx
,
181 re_node_set
*cur_nodes
, Idx cur_str
,
182 Idx subexp_num
, int type
)
184 static bool build_trtable (const re_dfa_t
*dfa
,
185 re_dfastate_t
*state
) internal_function
;
186 #ifdef RE_ENABLE_I18N
187 static int check_node_accept_bytes (const re_dfa_t
*dfa
, Idx node_idx
,
188 const re_string_t
*input
, Idx idx
)
191 static unsigned int find_collation_sequence_value (const unsigned char *mbs
,
195 #endif /* RE_ENABLE_I18N */
196 static Idx
group_nodes_into_DFAstates (const re_dfa_t
*dfa
,
197 const re_dfastate_t
*state
,
198 re_node_set
*states_node
,
199 bitset_t
*states_ch
) internal_function
;
200 static bool check_node_accept (const re_match_context_t
*mctx
,
201 const re_token_t
*node
, Idx idx
)
203 static reg_errcode_t
extend_buffers (re_match_context_t
*mctx
)
206 /* Entry point for POSIX code. */
208 /* regexec searches for a given pattern, specified by PREG, in the
211 If NMATCH is zero or REG_NOSUB was set in the cflags argument to
212 `regcomp', we ignore PMATCH. Otherwise, we assume PMATCH has at
213 least NMATCH elements, and we set them to the offsets of the
214 corresponding matched substrings.
216 EFLAGS specifies `execution flags' which affect matching: if
217 REG_NOTBOL is set, then ^ does not match at the beginning of the
218 string; if REG_NOTEOL is set, then $ does not match at the end.
220 We return 0 if we find a match and REG_NOMATCH if not. */
223 regexec (preg
, string
, nmatch
, pmatch
, eflags
)
224 const regex_t
*__restrict preg
;
225 const char *__restrict string
;
233 re_dfa_t
*dfa
= (re_dfa_t
*) preg
->buffer
;
236 if (eflags
& ~(REG_NOTBOL
| REG_NOTEOL
| REG_STARTEND
))
239 if (eflags
& REG_STARTEND
)
241 start
= pmatch
[0].rm_so
;
242 length
= pmatch
[0].rm_eo
;
247 length
= strlen (string
);
250 __libc_lock_lock (dfa
->lock
);
252 err
= re_search_internal (preg
, string
, length
, start
, length
,
253 length
, 0, NULL
, eflags
);
255 err
= re_search_internal (preg
, string
, length
, start
, length
,
256 length
, nmatch
, pmatch
, eflags
);
257 __libc_lock_unlock (dfa
->lock
);
258 return err
!= REG_NOERROR
;
262 # include <shlib-compat.h>
263 versioned_symbol (libc
, __regexec
, regexec
, GLIBC_2_3_4
);
265 # if SHLIB_COMPAT (libc, GLIBC_2_0, GLIBC_2_3_4)
266 __typeof__ (__regexec
) __compat_regexec
;
269 attribute_compat_text_section
270 __compat_regexec (const regex_t
*__restrict preg
,
271 const char *__restrict string
, size_t nmatch
,
272 regmatch_t pmatch
[], int eflags
)
274 return regexec (preg
, string
, nmatch
, pmatch
,
275 eflags
& (REG_NOTBOL
| REG_NOTEOL
));
277 compat_symbol (libc
, __compat_regexec
, regexec
, GLIBC_2_0
);
281 /* Entry points for GNU code. */
283 /* re_match, re_search, re_match_2, re_search_2
285 The former two functions operate on STRING with length LENGTH,
286 while the later two operate on concatenation of STRING1 and STRING2
287 with lengths LENGTH1 and LENGTH2, respectively.
289 re_match() matches the compiled pattern in BUFP against the string,
290 starting at index START.
292 re_search() first tries matching at index START, then it tries to match
293 starting from index START + 1, and so on. The last start position tried
294 is START + RANGE. (Thus RANGE = 0 forces re_search to operate the same
297 The parameter STOP of re_{match,search}_2 specifies that no match exceeding
298 the first STOP characters of the concatenation of the strings should be
301 If REGS is not NULL, and BUFP->no_sub is not set, the offsets of the match
302 and all groups is stored in REGS. (For the "_2" variants, the offsets are
303 computed relative to the concatenation, not relative to the individual
306 On success, re_match* functions return the length of the match, re_search*
307 return the position of the start of the match. Return value -1 means no
308 match was found and -2 indicates an internal error. */
311 re_match (bufp
, string
, length
, start
, regs
)
312 struct re_pattern_buffer
*bufp
;
315 struct re_registers
*regs
;
317 return re_search_stub (bufp
, string
, length
, start
, 0, length
, regs
, true);
320 weak_alias (__re_match
, re_match
)
324 re_search (bufp
, string
, length
, start
, range
, regs
)
325 struct re_pattern_buffer
*bufp
;
329 struct re_registers
*regs
;
331 return re_search_stub (bufp
, string
, length
, start
, range
, length
, regs
,
335 weak_alias (__re_search
, re_search
)
339 re_match_2 (bufp
, string1
, length1
, string2
, length2
, start
, regs
, stop
)
340 struct re_pattern_buffer
*bufp
;
341 const char *string1
, *string2
;
342 Idx length1
, length2
, start
, stop
;
343 struct re_registers
*regs
;
345 return re_search_2_stub (bufp
, string1
, length1
, string2
, length2
,
346 start
, 0, regs
, stop
, true);
349 weak_alias (__re_match_2
, re_match_2
)
353 re_search_2 (bufp
, string1
, length1
, string2
, length2
, start
, range
, regs
, stop
)
354 struct re_pattern_buffer
*bufp
;
355 const char *string1
, *string2
;
356 Idx length1
, length2
, start
, stop
;
358 struct re_registers
*regs
;
360 return re_search_2_stub (bufp
, string1
, length1
, string2
, length2
,
361 start
, range
, regs
, stop
, false);
364 weak_alias (__re_search_2
, re_search_2
)
369 re_search_2_stub (struct re_pattern_buffer
*bufp
,
370 const char *string1
, Idx length1
,
371 const char *string2
, Idx length2
,
372 Idx start
, regoff_t range
, struct re_registers
*regs
,
373 Idx stop
, bool ret_len
)
377 Idx len
= length1
+ length2
;
380 if (BE (length1
< 0 || length2
< 0 || stop
< 0 || len
< length1
, 0))
383 /* Concatenate the strings. */
387 s
= re_malloc (char, len
);
389 if (BE (s
== NULL
, 0))
392 memcpy (__mempcpy (s
, string1
, length1
), string2
, length2
);
394 memcpy (s
, string1
, length1
);
395 memcpy (s
+ length1
, string2
, length2
);
404 rval
= re_search_stub (bufp
, str
, len
, start
, range
, stop
, regs
,
410 /* The parameters have the same meaning as those of re_search.
411 Additional parameters:
412 If RET_LEN is true the length of the match is returned (re_match style);
413 otherwise the position of the match is returned. */
417 re_search_stub (struct re_pattern_buffer
*bufp
,
418 const char *string
, Idx length
,
419 Idx start
, regoff_t range
, Idx stop
, struct re_registers
*regs
,
422 reg_errcode_t result
;
428 re_dfa_t
*dfa
= (re_dfa_t
*) bufp
->buffer
;
430 Idx last_start
= start
+ range
;
432 /* Check for out-of-range. */
433 if (BE (start
< 0 || start
> length
, 0))
435 if (BE (length
< last_start
|| (0 <= range
&& last_start
< start
), 0))
437 else if (BE (last_start
< 0 || (range
< 0 && start
<= last_start
), 0))
440 __libc_lock_lock (dfa
->lock
);
442 eflags
|= (bufp
->not_bol
) ? REG_NOTBOL
: 0;
443 eflags
|= (bufp
->not_eol
) ? REG_NOTEOL
: 0;
445 /* Compile fastmap if we haven't yet. */
446 if (start
< last_start
&& bufp
->fastmap
!= NULL
&& !bufp
->fastmap_accurate
)
447 re_compile_fastmap (bufp
);
449 if (BE (bufp
->no_sub
, 0))
452 /* We need at least 1 register. */
455 else if (BE (bufp
->regs_allocated
== REGS_FIXED
456 && regs
->num_regs
<= bufp
->re_nsub
, 0))
458 nregs
= regs
->num_regs
;
459 if (BE (nregs
< 1, 0))
461 /* Nothing can be copied to regs. */
467 nregs
= bufp
->re_nsub
+ 1;
468 pmatch
= re_malloc (regmatch_t
, nregs
);
469 if (BE (pmatch
== NULL
, 0))
475 result
= re_search_internal (bufp
, string
, length
, start
, last_start
, stop
,
476 nregs
, pmatch
, eflags
);
480 /* I hope we needn't fill ther regs with -1's when no match was found. */
481 if (result
!= REG_NOERROR
)
483 else if (regs
!= NULL
)
485 /* If caller wants register contents data back, copy them. */
486 bufp
->regs_allocated
= re_copy_regs (regs
, pmatch
, nregs
,
487 bufp
->regs_allocated
);
488 if (BE (bufp
->regs_allocated
== REGS_UNALLOCATED
, 0))
492 if (BE (rval
== 0, 1))
496 assert (pmatch
[0].rm_so
== start
);
497 rval
= pmatch
[0].rm_eo
- start
;
500 rval
= pmatch
[0].rm_so
;
504 __libc_lock_unlock (dfa
->lock
);
510 re_copy_regs (struct re_registers
*regs
, regmatch_t
*pmatch
, Idx nregs
,
513 int rval
= REGS_REALLOCATE
;
515 Idx need_regs
= nregs
+ 1;
516 /* We need one extra element beyond `num_regs' for the `-1' marker GNU code
519 /* Have the register data arrays been allocated? */
520 if (regs_allocated
== REGS_UNALLOCATED
)
521 { /* No. So allocate them with malloc. */
522 regs
->start
= re_malloc (regoff_t
, need_regs
);
523 if (BE (regs
->start
== NULL
, 0))
524 return REGS_UNALLOCATED
;
525 regs
->end
= re_malloc (regoff_t
, need_regs
);
526 if (BE (regs
->end
== NULL
, 0))
528 re_free (regs
->start
);
529 return REGS_UNALLOCATED
;
531 regs
->num_regs
= need_regs
;
533 else if (regs_allocated
== REGS_REALLOCATE
)
534 { /* Yes. If we need more elements than were already
535 allocated, reallocate them. If we need fewer, just
537 if (BE (need_regs
> regs
->num_regs
, 0))
539 regoff_t
*new_start
= re_realloc (regs
->start
, regoff_t
, need_regs
);
541 if (BE (new_start
== NULL
, 0))
542 return REGS_UNALLOCATED
;
543 new_end
= re_realloc (regs
->end
, regoff_t
, need_regs
);
544 if (BE (new_end
== NULL
, 0))
547 return REGS_UNALLOCATED
;
549 regs
->start
= new_start
;
551 regs
->num_regs
= need_regs
;
556 assert (regs_allocated
== REGS_FIXED
);
557 /* This function may not be called with REGS_FIXED and nregs too big. */
558 assert (regs
->num_regs
>= nregs
);
563 for (i
= 0; i
< nregs
; ++i
)
565 regs
->start
[i
] = pmatch
[i
].rm_so
;
566 regs
->end
[i
] = pmatch
[i
].rm_eo
;
568 for ( ; i
< regs
->num_regs
; ++i
)
569 regs
->start
[i
] = regs
->end
[i
] = -1;
574 /* Set REGS to hold NUM_REGS registers, storing them in STARTS and
575 ENDS. Subsequent matches using PATTERN_BUFFER and REGS will use
576 this memory for recording register information. STARTS and ENDS
577 must be allocated using the malloc library routine, and must each
578 be at least NUM_REGS * sizeof (regoff_t) bytes long.
580 If NUM_REGS == 0, then subsequent matches should allocate their own
583 Unless this function is called, the first search or match using
584 PATTERN_BUFFER will allocate its own register data, without
585 freeing the old data. */
588 re_set_registers (bufp
, regs
, num_regs
, starts
, ends
)
589 struct re_pattern_buffer
*bufp
;
590 struct re_registers
*regs
;
591 __re_size_t num_regs
;
592 regoff_t
*starts
, *ends
;
596 bufp
->regs_allocated
= REGS_REALLOCATE
;
597 regs
->num_regs
= num_regs
;
598 regs
->start
= starts
;
603 bufp
->regs_allocated
= REGS_UNALLOCATED
;
605 regs
->start
= regs
->end
= NULL
;
609 weak_alias (__re_set_registers
, re_set_registers
)
612 /* Entry points compatible with 4.2 BSD regex library. We don't define
613 them unless specifically requested. */
615 #if defined _REGEX_RE_COMP || defined _LIBC
623 return 0 == regexec (&re_comp_buf
, s
, 0, NULL
, 0);
625 #endif /* _REGEX_RE_COMP */
627 /* Internal entry point. */
629 /* Searches for a compiled pattern PREG in the string STRING, whose
630 length is LENGTH. NMATCH, PMATCH, and EFLAGS have the same
631 meaning as with regexec. LAST_START is START + RANGE, where
632 START and RANGE have the same meaning as with re_search.
633 Return REG_NOERROR if we find a match, and REG_NOMATCH if not,
634 otherwise return the error code.
635 Note: We assume front end functions already check ranges.
636 (0 <= LAST_START && LAST_START <= LENGTH) */
640 re_search_internal (const regex_t
*preg
,
641 const char *string
, Idx length
,
642 Idx start
, Idx last_start
, Idx stop
,
643 size_t nmatch
, regmatch_t pmatch
[],
647 const re_dfa_t
*dfa
= (const re_dfa_t
*) preg
->buffer
;
648 Idx left_lim
, right_lim
;
650 bool fl_longest_match
;
653 Idx match_last
= REG_MISSING
;
657 #if defined _LIBC || (defined __STDC_VERSION__ && __STDC_VERSION__ >= 199901L)
658 re_match_context_t mctx
= { .dfa
= dfa
};
660 re_match_context_t mctx
;
662 char *fastmap
= ((preg
->fastmap
!= NULL
&& preg
->fastmap_accurate
663 && start
!= last_start
&& !preg
->can_be_null
)
664 ? preg
->fastmap
: NULL
);
665 RE_TRANSLATE_TYPE t
= preg
->translate
;
667 #if !(defined _LIBC || (defined __STDC_VERSION__ && __STDC_VERSION__ >= 199901L))
668 memset (&mctx
, '\0', sizeof (re_match_context_t
));
672 extra_nmatch
= (nmatch
> preg
->re_nsub
) ? nmatch
- (preg
->re_nsub
+ 1) : 0;
673 nmatch
-= extra_nmatch
;
675 /* Check if the DFA haven't been compiled. */
676 if (BE (preg
->used
== 0 || dfa
->init_state
== NULL
677 || dfa
->init_state_word
== NULL
|| dfa
->init_state_nl
== NULL
678 || dfa
->init_state_begbuf
== NULL
, 0))
682 /* We assume front-end functions already check them. */
683 assert (0 <= last_start
&& last_start
<= length
);
686 /* If initial states with non-begbuf contexts have no elements,
687 the regex must be anchored. If preg->newline_anchor is set,
688 we'll never use init_state_nl, so do not check it. */
689 if (dfa
->init_state
->nodes
.nelem
== 0
690 && dfa
->init_state_word
->nodes
.nelem
== 0
691 && (dfa
->init_state_nl
->nodes
.nelem
== 0
692 || !preg
->newline_anchor
))
694 if (start
!= 0 && last_start
!= 0)
696 start
= last_start
= 0;
699 /* We must check the longest matching, if nmatch > 0. */
700 fl_longest_match
= (nmatch
!= 0 || dfa
->nbackref
);
702 err
= re_string_allocate (&mctx
.input
, string
, length
, dfa
->nodes_len
+ 1,
703 preg
->translate
, preg
->syntax
& RE_ICASE
, dfa
);
704 if (BE (err
!= REG_NOERROR
, 0))
706 mctx
.input
.stop
= stop
;
707 mctx
.input
.raw_stop
= stop
;
708 mctx
.input
.newline_anchor
= preg
->newline_anchor
;
710 err
= match_ctx_init (&mctx
, eflags
, dfa
->nbackref
* 2);
711 if (BE (err
!= REG_NOERROR
, 0))
714 /* We will log all the DFA states through which the dfa pass,
715 if nmatch > 1, or this dfa has "multibyte node", which is a
716 back-reference or a node which can accept multibyte character or
717 multi character collating element. */
718 if (nmatch
> 1 || dfa
->has_mb_node
)
720 /* Avoid overflow. */
721 if (BE (SIZE_MAX
/ sizeof (re_dfastate_t
*) <= mctx
.input
.bufs_len
, 0))
727 mctx
.state_log
= re_malloc (re_dfastate_t
*, mctx
.input
.bufs_len
+ 1);
728 if (BE (mctx
.state_log
== NULL
, 0))
735 mctx
.state_log
= NULL
;
738 mctx
.input
.tip_context
= (eflags
& REG_NOTBOL
) ? CONTEXT_BEGBUF
739 : CONTEXT_NEWLINE
| CONTEXT_BEGBUF
;
741 /* Check incrementally whether of not the input string match. */
742 incr
= (last_start
< start
) ? -1 : 1;
743 left_lim
= (last_start
< start
) ? last_start
: start
;
744 right_lim
= (last_start
< start
) ? start
: last_start
;
745 sb
= dfa
->mb_cur_max
== 1;
748 ? ((sb
|| !(preg
->syntax
& RE_ICASE
|| t
) ? 4 : 0)
749 | (start
<= last_start
? 2 : 0)
750 | (t
!= NULL
? 1 : 0))
753 for (;; match_first
+= incr
)
756 if (match_first
< left_lim
|| right_lim
< match_first
)
759 /* Advance as rapidly as possible through the string, until we
760 find a plausible place to start matching. This may be done
761 with varying efficiency, so there are various possibilities:
762 only the most common of them are specialized, in order to
763 save on code size. We use a switch statement for speed. */
771 /* Fastmap with single-byte translation, match forward. */
772 while (BE (match_first
< right_lim
, 1)
773 && !fastmap
[t
[(unsigned char) string
[match_first
]]])
775 goto forward_match_found_start_or_reached_end
;
778 /* Fastmap without translation, match forward. */
779 while (BE (match_first
< right_lim
, 1)
780 && !fastmap
[(unsigned char) string
[match_first
]])
783 forward_match_found_start_or_reached_end
:
784 if (BE (match_first
== right_lim
, 0))
786 ch
= match_first
>= length
787 ? 0 : (unsigned char) string
[match_first
];
788 if (!fastmap
[t
? t
[ch
] : ch
])
795 /* Fastmap without multi-byte translation, match backwards. */
796 while (match_first
>= left_lim
)
798 ch
= match_first
>= length
799 ? 0 : (unsigned char) string
[match_first
];
800 if (fastmap
[t
? t
[ch
] : ch
])
804 if (match_first
< left_lim
)
809 /* In this case, we can't determine easily the current byte,
810 since it might be a component byte of a multibyte
811 character. Then we use the constructed buffer instead. */
814 /* If MATCH_FIRST is out of the valid range, reconstruct the
816 __re_size_t offset
= match_first
- mctx
.input
.raw_mbs_idx
;
817 if (BE (offset
>= (__re_size_t
) mctx
.input
.valid_raw_len
, 0))
819 err
= re_string_reconstruct (&mctx
.input
, match_first
,
821 if (BE (err
!= REG_NOERROR
, 0))
824 offset
= match_first
- mctx
.input
.raw_mbs_idx
;
826 /* If MATCH_FIRST is out of the buffer, leave it as '\0'.
827 Note that MATCH_FIRST must not be smaller than 0. */
828 ch
= (match_first
>= length
829 ? 0 : re_string_byte_at (&mctx
.input
, offset
));
833 if (match_first
< left_lim
|| match_first
> right_lim
)
842 /* Reconstruct the buffers so that the matcher can assume that
843 the matching starts from the beginning of the buffer. */
844 err
= re_string_reconstruct (&mctx
.input
, match_first
, eflags
);
845 if (BE (err
!= REG_NOERROR
, 0))
848 #ifdef RE_ENABLE_I18N
849 /* Don't consider this char as a possible match start if it part,
850 yet isn't the head, of a multibyte character. */
851 if (!sb
&& !re_string_first_byte (&mctx
.input
, 0))
855 /* It seems to be appropriate one, then use the matcher. */
856 /* We assume that the matching starts from 0. */
857 mctx
.state_log_top
= mctx
.nbkref_ents
= mctx
.max_mb_elem_len
= 0;
858 match_last
= check_matching (&mctx
, fl_longest_match
,
859 start
<= last_start
? &match_first
: NULL
);
860 if (match_last
!= REG_MISSING
)
862 if (BE (match_last
== REG_ERROR
, 0))
869 mctx
.match_last
= match_last
;
870 if ((!preg
->no_sub
&& nmatch
> 1) || dfa
->nbackref
)
872 re_dfastate_t
*pstate
= mctx
.state_log
[match_last
];
873 mctx
.last_node
= check_halt_state_context (&mctx
, pstate
,
876 if ((!preg
->no_sub
&& nmatch
> 1 && dfa
->has_plural_match
)
879 err
= prune_impossible_nodes (&mctx
);
880 if (err
== REG_NOERROR
)
882 if (BE (err
!= REG_NOMATCH
, 0))
884 match_last
= REG_MISSING
;
887 break; /* We found a match. */
891 match_ctx_clean (&mctx
);
895 assert (match_last
!= REG_MISSING
);
896 assert (err
== REG_NOERROR
);
899 /* Set pmatch[] if we need. */
904 /* Initialize registers. */
905 for (reg_idx
= 1; reg_idx
< nmatch
; ++reg_idx
)
906 pmatch
[reg_idx
].rm_so
= pmatch
[reg_idx
].rm_eo
= -1;
908 /* Set the points where matching start/end. */
910 pmatch
[0].rm_eo
= mctx
.match_last
;
911 /* FIXME: This function should fail if mctx.match_last exceeds
912 the maximum possible regoff_t value. We need a new error
913 code REG_OVERFLOW. */
915 if (!preg
->no_sub
&& nmatch
> 1)
917 err
= set_regs (preg
, &mctx
, nmatch
, pmatch
,
918 dfa
->has_plural_match
&& dfa
->nbackref
> 0);
919 if (BE (err
!= REG_NOERROR
, 0))
923 /* At last, add the offset to the each registers, since we slided
924 the buffers so that we could assume that the matching starts
926 for (reg_idx
= 0; reg_idx
< nmatch
; ++reg_idx
)
927 if (pmatch
[reg_idx
].rm_so
!= -1)
929 #ifdef RE_ENABLE_I18N
930 if (BE (mctx
.input
.offsets_needed
!= 0, 0))
932 pmatch
[reg_idx
].rm_so
=
933 (pmatch
[reg_idx
].rm_so
== mctx
.input
.valid_len
934 ? mctx
.input
.valid_raw_len
935 : mctx
.input
.offsets
[pmatch
[reg_idx
].rm_so
]);
936 pmatch
[reg_idx
].rm_eo
=
937 (pmatch
[reg_idx
].rm_eo
== mctx
.input
.valid_len
938 ? mctx
.input
.valid_raw_len
939 : mctx
.input
.offsets
[pmatch
[reg_idx
].rm_eo
]);
942 assert (mctx
.input
.offsets_needed
== 0);
944 pmatch
[reg_idx
].rm_so
+= match_first
;
945 pmatch
[reg_idx
].rm_eo
+= match_first
;
947 for (reg_idx
= 0; reg_idx
< extra_nmatch
; ++reg_idx
)
949 pmatch
[nmatch
+ reg_idx
].rm_so
= -1;
950 pmatch
[nmatch
+ reg_idx
].rm_eo
= -1;
954 for (reg_idx
= 0; reg_idx
+ 1 < nmatch
; reg_idx
++)
955 if (dfa
->subexp_map
[reg_idx
] != reg_idx
)
957 pmatch
[reg_idx
+ 1].rm_so
958 = pmatch
[dfa
->subexp_map
[reg_idx
] + 1].rm_so
;
959 pmatch
[reg_idx
+ 1].rm_eo
960 = pmatch
[dfa
->subexp_map
[reg_idx
] + 1].rm_eo
;
965 re_free (mctx
.state_log
);
967 match_ctx_free (&mctx
);
968 re_string_destruct (&mctx
.input
);
974 prune_impossible_nodes (re_match_context_t
*mctx
)
976 const re_dfa_t
*const dfa
= mctx
->dfa
;
977 Idx halt_node
, match_last
;
979 re_dfastate_t
**sifted_states
;
980 re_dfastate_t
**lim_states
= NULL
;
981 re_sift_context_t sctx
;
983 assert (mctx
->state_log
!= NULL
);
985 match_last
= mctx
->match_last
;
986 halt_node
= mctx
->last_node
;
988 /* Avoid overflow. */
989 if (BE (SIZE_MAX
/ sizeof (re_dfastate_t
*) <= match_last
, 0))
992 sifted_states
= re_malloc (re_dfastate_t
*, match_last
+ 1);
993 if (BE (sifted_states
== NULL
, 0))
1000 lim_states
= re_malloc (re_dfastate_t
*, match_last
+ 1);
1001 if (BE (lim_states
== NULL
, 0))
1008 memset (lim_states
, '\0',
1009 sizeof (re_dfastate_t
*) * (match_last
+ 1));
1010 sift_ctx_init (&sctx
, sifted_states
, lim_states
, halt_node
,
1012 ret
= sift_states_backward (mctx
, &sctx
);
1013 re_node_set_free (&sctx
.limits
);
1014 if (BE (ret
!= REG_NOERROR
, 0))
1016 if (sifted_states
[0] != NULL
|| lim_states
[0] != NULL
)
1021 if (! REG_VALID_INDEX (match_last
))
1026 } while (mctx
->state_log
[match_last
] == NULL
1027 || !mctx
->state_log
[match_last
]->halt
);
1028 halt_node
= check_halt_state_context (mctx
,
1029 mctx
->state_log
[match_last
],
1032 ret
= merge_state_array (dfa
, sifted_states
, lim_states
,
1034 re_free (lim_states
);
1036 if (BE (ret
!= REG_NOERROR
, 0))
1041 sift_ctx_init (&sctx
, sifted_states
, lim_states
, halt_node
, match_last
);
1042 ret
= sift_states_backward (mctx
, &sctx
);
1043 re_node_set_free (&sctx
.limits
);
1044 if (BE (ret
!= REG_NOERROR
, 0))
1047 re_free (mctx
->state_log
);
1048 mctx
->state_log
= sifted_states
;
1049 sifted_states
= NULL
;
1050 mctx
->last_node
= halt_node
;
1051 mctx
->match_last
= match_last
;
1054 re_free (sifted_states
);
1055 re_free (lim_states
);
1059 /* Acquire an initial state and return it.
1060 We must select appropriate initial state depending on the context,
1061 since initial states may have constraints like "\<", "^", etc.. */
1063 static inline re_dfastate_t
*
1064 __attribute ((always_inline
)) internal_function
1065 acquire_init_state_context (reg_errcode_t
*err
, const re_match_context_t
*mctx
,
1068 const re_dfa_t
*const dfa
= mctx
->dfa
;
1069 if (dfa
->init_state
->has_constraint
)
1071 unsigned int context
;
1072 context
= re_string_context_at (&mctx
->input
, idx
- 1, mctx
->eflags
);
1073 if (IS_WORD_CONTEXT (context
))
1074 return dfa
->init_state_word
;
1075 else if (IS_ORDINARY_CONTEXT (context
))
1076 return dfa
->init_state
;
1077 else if (IS_BEGBUF_CONTEXT (context
) && IS_NEWLINE_CONTEXT (context
))
1078 return dfa
->init_state_begbuf
;
1079 else if (IS_NEWLINE_CONTEXT (context
))
1080 return dfa
->init_state_nl
;
1081 else if (IS_BEGBUF_CONTEXT (context
))
1083 /* It is relatively rare case, then calculate on demand. */
1084 return re_acquire_state_context (err
, dfa
,
1085 dfa
->init_state
->entrance_nodes
,
1089 /* Must not happen? */
1090 return dfa
->init_state
;
1093 return dfa
->init_state
;
1096 /* Check whether the regular expression match input string INPUT or not,
1097 and return the index where the matching end. Return REG_MISSING if
1098 there is no match, and return REG_ERROR in case of an error.
1099 FL_LONGEST_MATCH means we want the POSIX longest matching.
1100 If P_MATCH_FIRST is not NULL, and the match fails, it is set to the
1101 next place where we may want to try matching.
1102 Note that the matcher assume that the maching starts from the current
1103 index of the buffer. */
1107 check_matching (re_match_context_t
*mctx
, bool fl_longest_match
,
1110 const re_dfa_t
*const dfa
= mctx
->dfa
;
1113 Idx match_last
= REG_MISSING
;
1114 Idx cur_str_idx
= re_string_cur_idx (&mctx
->input
);
1115 re_dfastate_t
*cur_state
;
1116 bool at_init_state
= p_match_first
!= NULL
;
1117 Idx next_start_idx
= cur_str_idx
;
1120 cur_state
= acquire_init_state_context (&err
, mctx
, cur_str_idx
);
1121 /* An initial state must not be NULL (invalid). */
1122 if (BE (cur_state
== NULL
, 0))
1124 assert (err
== REG_ESPACE
);
1128 if (mctx
->state_log
!= NULL
)
1130 mctx
->state_log
[cur_str_idx
] = cur_state
;
1132 /* Check OP_OPEN_SUBEXP in the initial state in case that we use them
1133 later. E.g. Processing back references. */
1134 if (BE (dfa
->nbackref
, 0))
1136 at_init_state
= false;
1137 err
= check_subexp_matching_top (mctx
, &cur_state
->nodes
, 0);
1138 if (BE (err
!= REG_NOERROR
, 0))
1141 if (cur_state
->has_backref
)
1143 err
= transit_state_bkref (mctx
, &cur_state
->nodes
);
1144 if (BE (err
!= REG_NOERROR
, 0))
1150 /* If the RE accepts NULL string. */
1151 if (BE (cur_state
->halt
, 0))
1153 if (!cur_state
->has_constraint
1154 || check_halt_state_context (mctx
, cur_state
, cur_str_idx
))
1156 if (!fl_longest_match
)
1160 match_last
= cur_str_idx
;
1166 while (!re_string_eoi (&mctx
->input
))
1168 re_dfastate_t
*old_state
= cur_state
;
1169 Idx next_char_idx
= re_string_cur_idx (&mctx
->input
) + 1;
1171 if (BE (next_char_idx
>= mctx
->input
.bufs_len
, 0)
1172 || (BE (next_char_idx
>= mctx
->input
.valid_len
, 0)
1173 && mctx
->input
.valid_len
< mctx
->input
.len
))
1175 err
= extend_buffers (mctx
);
1176 if (BE (err
!= REG_NOERROR
, 0))
1178 assert (err
== REG_ESPACE
);
1183 cur_state
= transit_state (&err
, mctx
, cur_state
);
1184 if (mctx
->state_log
!= NULL
)
1185 cur_state
= merge_state_with_log (&err
, mctx
, cur_state
);
1187 if (cur_state
== NULL
)
1189 /* Reached the invalid state or an error. Try to recover a valid
1190 state using the state log, if available and if we have not
1191 already found a valid (even if not the longest) match. */
1192 if (BE (err
!= REG_NOERROR
, 0))
1195 if (mctx
->state_log
== NULL
1196 || (match
&& !fl_longest_match
)
1197 || (cur_state
= find_recover_state (&err
, mctx
)) == NULL
)
1201 if (BE (at_init_state
, 0))
1203 if (old_state
== cur_state
)
1204 next_start_idx
= next_char_idx
;
1206 at_init_state
= false;
1209 if (cur_state
->halt
)
1211 /* Reached a halt state.
1212 Check the halt state can satisfy the current context. */
1213 if (!cur_state
->has_constraint
1214 || check_halt_state_context (mctx
, cur_state
,
1215 re_string_cur_idx (&mctx
->input
)))
1217 /* We found an appropriate halt state. */
1218 match_last
= re_string_cur_idx (&mctx
->input
);
1221 /* We found a match, do not modify match_first below. */
1222 p_match_first
= NULL
;
1223 if (!fl_longest_match
)
1230 *p_match_first
+= next_start_idx
;
1235 /* Check NODE match the current context. */
1239 check_halt_node_context (const re_dfa_t
*dfa
, Idx node
, unsigned int context
)
1241 re_token_type_t type
= dfa
->nodes
[node
].type
;
1242 unsigned int constraint
= dfa
->nodes
[node
].constraint
;
1243 if (type
!= END_OF_RE
)
1247 if (NOT_SATISFY_NEXT_CONSTRAINT (constraint
, context
))
1252 /* Check the halt state STATE match the current context.
1253 Return 0 if not match, if the node, STATE has, is a halt node and
1254 match the context, return the node. */
1258 check_halt_state_context (const re_match_context_t
*mctx
,
1259 const re_dfastate_t
*state
, Idx idx
)
1262 unsigned int context
;
1264 assert (state
->halt
);
1266 context
= re_string_context_at (&mctx
->input
, idx
, mctx
->eflags
);
1267 for (i
= 0; i
< state
->nodes
.nelem
; ++i
)
1268 if (check_halt_node_context (mctx
->dfa
, state
->nodes
.elems
[i
], context
))
1269 return state
->nodes
.elems
[i
];
1273 /* Compute the next node to which "NFA" transit from NODE("NFA" is a NFA
1274 corresponding to the DFA).
1275 Return the destination node, and update EPS_VIA_NODES;
1276 return REG_MISSING in case of errors. */
1280 proceed_next_node (const re_match_context_t
*mctx
, Idx nregs
, regmatch_t
*regs
,
1281 Idx
*pidx
, Idx node
, re_node_set
*eps_via_nodes
,
1282 struct re_fail_stack_t
*fs
)
1284 const re_dfa_t
*const dfa
= mctx
->dfa
;
1287 if (IS_EPSILON_NODE (dfa
->nodes
[node
].type
))
1289 re_node_set
*cur_nodes
= &mctx
->state_log
[*pidx
]->nodes
;
1290 re_node_set
*edests
= &dfa
->edests
[node
];
1292 ok
= re_node_set_insert (eps_via_nodes
, node
);
1295 /* Pick up a valid destination, or return REG_MISSING if none
1297 for (dest_node
= REG_MISSING
, i
= 0; i
< edests
->nelem
; ++i
)
1299 Idx candidate
= edests
->elems
[i
];
1300 if (!re_node_set_contains (cur_nodes
, candidate
))
1302 if (dest_node
== REG_MISSING
)
1303 dest_node
= candidate
;
1307 /* In order to avoid infinite loop like "(a*)*", return the second
1308 epsilon-transition if the first was already considered. */
1309 if (re_node_set_contains (eps_via_nodes
, dest_node
))
1312 /* Otherwise, push the second epsilon-transition on the fail stack. */
1314 && push_fail_stack (fs
, *pidx
, candidate
, nregs
, regs
,
1318 /* We know we are going to exit. */
1327 re_token_type_t type
= dfa
->nodes
[node
].type
;
1329 #ifdef RE_ENABLE_I18N
1330 if (dfa
->nodes
[node
].accept_mb
)
1331 naccepted
= check_node_accept_bytes (dfa
, node
, &mctx
->input
, *pidx
);
1333 #endif /* RE_ENABLE_I18N */
1334 if (type
== OP_BACK_REF
)
1336 Idx subexp_idx
= dfa
->nodes
[node
].opr
.idx
+ 1;
1337 naccepted
= regs
[subexp_idx
].rm_eo
- regs
[subexp_idx
].rm_so
;
1340 if (regs
[subexp_idx
].rm_so
== -1 || regs
[subexp_idx
].rm_eo
== -1)
1344 char *buf
= (char *) re_string_get_buffer (&mctx
->input
);
1345 if (memcmp (buf
+ regs
[subexp_idx
].rm_so
, buf
+ *pidx
,
1354 ok
= re_node_set_insert (eps_via_nodes
, node
);
1357 dest_node
= dfa
->edests
[node
].elems
[0];
1358 if (re_node_set_contains (&mctx
->state_log
[*pidx
]->nodes
,
1365 || check_node_accept (mctx
, dfa
->nodes
+ node
, *pidx
))
1367 Idx dest_node
= dfa
->nexts
[node
];
1368 *pidx
= (naccepted
== 0) ? *pidx
+ 1 : *pidx
+ naccepted
;
1369 if (fs
&& (*pidx
> mctx
->match_last
|| mctx
->state_log
[*pidx
] == NULL
1370 || !re_node_set_contains (&mctx
->state_log
[*pidx
]->nodes
,
1373 re_node_set_empty (eps_via_nodes
);
1380 static reg_errcode_t
1382 push_fail_stack (struct re_fail_stack_t
*fs
, Idx str_idx
, Idx dest_node
,
1383 Idx nregs
, regmatch_t
*regs
, re_node_set
*eps_via_nodes
)
1386 Idx num
= fs
->num
++;
1387 if (fs
->num
== fs
->alloc
)
1389 struct re_fail_stack_ent_t
*new_array
;
1390 new_array
= realloc (fs
->stack
, (sizeof (struct re_fail_stack_ent_t
)
1392 if (new_array
== NULL
)
1395 fs
->stack
= new_array
;
1397 fs
->stack
[num
].idx
= str_idx
;
1398 fs
->stack
[num
].node
= dest_node
;
1399 fs
->stack
[num
].regs
= re_malloc (regmatch_t
, nregs
);
1400 if (fs
->stack
[num
].regs
== NULL
)
1402 memcpy (fs
->stack
[num
].regs
, regs
, sizeof (regmatch_t
) * nregs
);
1403 err
= re_node_set_init_copy (&fs
->stack
[num
].eps_via_nodes
, eps_via_nodes
);
1409 pop_fail_stack (struct re_fail_stack_t
*fs
, Idx
*pidx
, Idx nregs
,
1410 regmatch_t
*regs
, re_node_set
*eps_via_nodes
)
1412 Idx num
= --fs
->num
;
1413 assert (REG_VALID_INDEX (num
));
1414 *pidx
= fs
->stack
[num
].idx
;
1415 memcpy (regs
, fs
->stack
[num
].regs
, sizeof (regmatch_t
) * nregs
);
1416 re_node_set_free (eps_via_nodes
);
1417 re_free (fs
->stack
[num
].regs
);
1418 *eps_via_nodes
= fs
->stack
[num
].eps_via_nodes
;
1419 return fs
->stack
[num
].node
;
1422 /* Set the positions where the subexpressions are starts/ends to registers
1424 Note: We assume that pmatch[0] is already set, and
1425 pmatch[i].rm_so == pmatch[i].rm_eo == -1 for 0 < i < nmatch. */
1427 static reg_errcode_t
1429 set_regs (const regex_t
*preg
, const re_match_context_t
*mctx
, size_t nmatch
,
1430 regmatch_t
*pmatch
, bool fl_backtrack
)
1432 const re_dfa_t
*dfa
= (const re_dfa_t
*) preg
->buffer
;
1434 re_node_set eps_via_nodes
;
1435 struct re_fail_stack_t
*fs
;
1436 struct re_fail_stack_t fs_body
= { 0, 2, NULL
};
1437 regmatch_t
*prev_idx_match
;
1438 bool prev_idx_match_malloced
= false;
1441 assert (nmatch
> 1);
1442 assert (mctx
->state_log
!= NULL
);
1447 fs
->stack
= re_malloc (struct re_fail_stack_ent_t
, fs
->alloc
);
1448 if (fs
->stack
== NULL
)
1454 cur_node
= dfa
->init_node
;
1455 re_node_set_init_empty (&eps_via_nodes
);
1457 if (__libc_use_alloca (nmatch
* sizeof (regmatch_t
)))
1458 prev_idx_match
= (regmatch_t
*) alloca (nmatch
* sizeof (regmatch_t
));
1461 prev_idx_match
= re_malloc (regmatch_t
, nmatch
);
1462 if (prev_idx_match
== NULL
)
1464 free_fail_stack_return (fs
);
1467 prev_idx_match_malloced
= true;
1469 memcpy (prev_idx_match
, pmatch
, sizeof (regmatch_t
) * nmatch
);
1471 for (idx
= pmatch
[0].rm_so
; idx
<= pmatch
[0].rm_eo
;)
1473 update_regs (dfa
, pmatch
, prev_idx_match
, cur_node
, idx
, nmatch
);
1475 if (idx
== pmatch
[0].rm_eo
&& cur_node
== mctx
->last_node
)
1480 for (reg_idx
= 0; reg_idx
< nmatch
; ++reg_idx
)
1481 if (pmatch
[reg_idx
].rm_so
> -1 && pmatch
[reg_idx
].rm_eo
== -1)
1483 if (reg_idx
== nmatch
)
1485 re_node_set_free (&eps_via_nodes
);
1486 if (prev_idx_match_malloced
)
1487 re_free (prev_idx_match
);
1488 return free_fail_stack_return (fs
);
1490 cur_node
= pop_fail_stack (fs
, &idx
, nmatch
, pmatch
,
1495 re_node_set_free (&eps_via_nodes
);
1496 if (prev_idx_match_malloced
)
1497 re_free (prev_idx_match
);
1502 /* Proceed to next node. */
1503 cur_node
= proceed_next_node (mctx
, nmatch
, pmatch
, &idx
, cur_node
,
1504 &eps_via_nodes
, fs
);
1506 if (BE (! REG_VALID_INDEX (cur_node
), 0))
1508 if (BE (cur_node
== REG_ERROR
, 0))
1510 re_node_set_free (&eps_via_nodes
);
1511 if (prev_idx_match_malloced
)
1512 re_free (prev_idx_match
);
1513 free_fail_stack_return (fs
);
1517 cur_node
= pop_fail_stack (fs
, &idx
, nmatch
, pmatch
,
1521 re_node_set_free (&eps_via_nodes
);
1522 if (prev_idx_match_malloced
)
1523 re_free (prev_idx_match
);
1528 re_node_set_free (&eps_via_nodes
);
1529 if (prev_idx_match_malloced
)
1530 re_free (prev_idx_match
);
1531 return free_fail_stack_return (fs
);
1534 static reg_errcode_t
1536 free_fail_stack_return (struct re_fail_stack_t
*fs
)
1541 for (fs_idx
= 0; fs_idx
< fs
->num
; ++fs_idx
)
1543 re_node_set_free (&fs
->stack
[fs_idx
].eps_via_nodes
);
1544 re_free (fs
->stack
[fs_idx
].regs
);
1546 re_free (fs
->stack
);
1553 update_regs (const re_dfa_t
*dfa
, regmatch_t
*pmatch
,
1554 regmatch_t
*prev_idx_match
, Idx cur_node
, Idx cur_idx
, Idx nmatch
)
1556 int type
= dfa
->nodes
[cur_node
].type
;
1557 if (type
== OP_OPEN_SUBEXP
)
1559 Idx reg_num
= dfa
->nodes
[cur_node
].opr
.idx
+ 1;
1561 /* We are at the first node of this sub expression. */
1562 if (reg_num
< nmatch
)
1564 pmatch
[reg_num
].rm_so
= cur_idx
;
1565 pmatch
[reg_num
].rm_eo
= -1;
1568 else if (type
== OP_CLOSE_SUBEXP
)
1570 Idx reg_num
= dfa
->nodes
[cur_node
].opr
.idx
+ 1;
1571 if (reg_num
< nmatch
)
1573 /* We are at the last node of this sub expression. */
1574 if (pmatch
[reg_num
].rm_so
< cur_idx
)
1576 pmatch
[reg_num
].rm_eo
= cur_idx
;
1577 /* This is a non-empty match or we are not inside an optional
1578 subexpression. Accept this right away. */
1579 memcpy (prev_idx_match
, pmatch
, sizeof (regmatch_t
) * nmatch
);
1583 if (dfa
->nodes
[cur_node
].opt_subexp
1584 && prev_idx_match
[reg_num
].rm_so
!= -1)
1585 /* We transited through an empty match for an optional
1586 subexpression, like (a?)*, and this is not the subexp's
1587 first match. Copy back the old content of the registers
1588 so that matches of an inner subexpression are undone as
1589 well, like in ((a?))*. */
1590 memcpy (pmatch
, prev_idx_match
, sizeof (regmatch_t
) * nmatch
);
1592 /* We completed a subexpression, but it may be part of
1593 an optional one, so do not update PREV_IDX_MATCH. */
1594 pmatch
[reg_num
].rm_eo
= cur_idx
;
1600 /* This function checks the STATE_LOG from the SCTX->last_str_idx to 0
1601 and sift the nodes in each states according to the following rules.
1602 Updated state_log will be wrote to STATE_LOG.
1604 Rules: We throw away the Node `a' in the STATE_LOG[STR_IDX] if...
1605 1. When STR_IDX == MATCH_LAST(the last index in the state_log):
1606 If `a' isn't the LAST_NODE and `a' can't epsilon transit to
1607 the LAST_NODE, we throw away the node `a'.
1608 2. When 0 <= STR_IDX < MATCH_LAST and `a' accepts
1609 string `s' and transit to `b':
1610 i. If 'b' isn't in the STATE_LOG[STR_IDX+strlen('s')], we throw
1612 ii. If 'b' is in the STATE_LOG[STR_IDX+strlen('s')] but 'b' is
1613 thrown away, we throw away the node `a'.
1614 3. When 0 <= STR_IDX < MATCH_LAST and 'a' epsilon transit to 'b':
1615 i. If 'b' isn't in the STATE_LOG[STR_IDX], we throw away the
1617 ii. If 'b' is in the STATE_LOG[STR_IDX] but 'b' is thrown away,
1618 we throw away the node `a'. */
1620 #define STATE_NODE_CONTAINS(state,node) \
1621 ((state) != NULL && re_node_set_contains (&(state)->nodes, node))
1623 static reg_errcode_t
1625 sift_states_backward (const re_match_context_t
*mctx
, re_sift_context_t
*sctx
)
1629 Idx str_idx
= sctx
->last_str_idx
;
1630 re_node_set cur_dest
;
1633 assert (mctx
->state_log
!= NULL
&& mctx
->state_log
[str_idx
] != NULL
);
1636 /* Build sifted state_log[str_idx]. It has the nodes which can epsilon
1637 transit to the last_node and the last_node itself. */
1638 err
= re_node_set_init_1 (&cur_dest
, sctx
->last_node
);
1639 if (BE (err
!= REG_NOERROR
, 0))
1641 err
= update_cur_sifted_state (mctx
, sctx
, str_idx
, &cur_dest
);
1642 if (BE (err
!= REG_NOERROR
, 0))
1645 /* Then check each states in the state_log. */
1648 /* Update counters. */
1649 null_cnt
= (sctx
->sifted_states
[str_idx
] == NULL
) ? null_cnt
+ 1 : 0;
1650 if (null_cnt
> mctx
->max_mb_elem_len
)
1652 memset (sctx
->sifted_states
, '\0',
1653 sizeof (re_dfastate_t
*) * str_idx
);
1654 re_node_set_free (&cur_dest
);
1657 re_node_set_empty (&cur_dest
);
1660 if (mctx
->state_log
[str_idx
])
1662 err
= build_sifted_states (mctx
, sctx
, str_idx
, &cur_dest
);
1663 if (BE (err
!= REG_NOERROR
, 0))
1667 /* Add all the nodes which satisfy the following conditions:
1668 - It can epsilon transit to a node in CUR_DEST.
1670 And update state_log. */
1671 err
= update_cur_sifted_state (mctx
, sctx
, str_idx
, &cur_dest
);
1672 if (BE (err
!= REG_NOERROR
, 0))
1677 re_node_set_free (&cur_dest
);
1681 static reg_errcode_t
1683 build_sifted_states (const re_match_context_t
*mctx
, re_sift_context_t
*sctx
,
1684 Idx str_idx
, re_node_set
*cur_dest
)
1686 const re_dfa_t
*const dfa
= mctx
->dfa
;
1687 const re_node_set
*cur_src
= &mctx
->state_log
[str_idx
]->non_eps_nodes
;
1690 /* Then build the next sifted state.
1691 We build the next sifted state on `cur_dest', and update
1692 `sifted_states[str_idx]' with `cur_dest'.
1694 `cur_dest' is the sifted state from `state_log[str_idx + 1]'.
1695 `cur_src' points the node_set of the old `state_log[str_idx]'
1696 (with the epsilon nodes pre-filtered out). */
1697 for (i
= 0; i
< cur_src
->nelem
; i
++)
1699 Idx prev_node
= cur_src
->elems
[i
];
1704 re_token_type_t type
= dfa
->nodes
[prev_node
].type
;
1705 assert (!IS_EPSILON_NODE (type
));
1707 #ifdef RE_ENABLE_I18N
1708 /* If the node may accept `multi byte'. */
1709 if (dfa
->nodes
[prev_node
].accept_mb
)
1710 naccepted
= sift_states_iter_mb (mctx
, sctx
, prev_node
,
1711 str_idx
, sctx
->last_str_idx
);
1712 #endif /* RE_ENABLE_I18N */
1714 /* We don't check backreferences here.
1715 See update_cur_sifted_state(). */
1717 && check_node_accept (mctx
, dfa
->nodes
+ prev_node
, str_idx
)
1718 && STATE_NODE_CONTAINS (sctx
->sifted_states
[str_idx
+ 1],
1719 dfa
->nexts
[prev_node
]))
1725 if (sctx
->limits
.nelem
)
1727 Idx to_idx
= str_idx
+ naccepted
;
1728 if (check_dst_limits (mctx
, &sctx
->limits
,
1729 dfa
->nexts
[prev_node
], to_idx
,
1730 prev_node
, str_idx
))
1733 ok
= re_node_set_insert (cur_dest
, prev_node
);
1741 /* Helper functions. */
1743 static reg_errcode_t
1745 clean_state_log_if_needed (re_match_context_t
*mctx
, Idx next_state_log_idx
)
1747 Idx top
= mctx
->state_log_top
;
1749 if (next_state_log_idx
>= mctx
->input
.bufs_len
1750 || (next_state_log_idx
>= mctx
->input
.valid_len
1751 && mctx
->input
.valid_len
< mctx
->input
.len
))
1754 err
= extend_buffers (mctx
);
1755 if (BE (err
!= REG_NOERROR
, 0))
1759 if (top
< next_state_log_idx
)
1761 memset (mctx
->state_log
+ top
+ 1, '\0',
1762 sizeof (re_dfastate_t
*) * (next_state_log_idx
- top
));
1763 mctx
->state_log_top
= next_state_log_idx
;
1768 static reg_errcode_t
1770 merge_state_array (const re_dfa_t
*dfa
, re_dfastate_t
**dst
,
1771 re_dfastate_t
**src
, Idx num
)
1775 for (st_idx
= 0; st_idx
< num
; ++st_idx
)
1777 if (dst
[st_idx
] == NULL
)
1778 dst
[st_idx
] = src
[st_idx
];
1779 else if (src
[st_idx
] != NULL
)
1781 re_node_set merged_set
;
1782 err
= re_node_set_init_union (&merged_set
, &dst
[st_idx
]->nodes
,
1783 &src
[st_idx
]->nodes
);
1784 if (BE (err
!= REG_NOERROR
, 0))
1786 dst
[st_idx
] = re_acquire_state (&err
, dfa
, &merged_set
);
1787 re_node_set_free (&merged_set
);
1788 if (BE (err
!= REG_NOERROR
, 0))
1795 static reg_errcode_t
1797 update_cur_sifted_state (const re_match_context_t
*mctx
,
1798 re_sift_context_t
*sctx
, Idx str_idx
,
1799 re_node_set
*dest_nodes
)
1801 const re_dfa_t
*const dfa
= mctx
->dfa
;
1802 reg_errcode_t err
= REG_NOERROR
;
1803 const re_node_set
*candidates
;
1804 candidates
= ((mctx
->state_log
[str_idx
] == NULL
) ? NULL
1805 : &mctx
->state_log
[str_idx
]->nodes
);
1807 if (dest_nodes
->nelem
== 0)
1808 sctx
->sifted_states
[str_idx
] = NULL
;
1813 /* At first, add the nodes which can epsilon transit to a node in
1815 err
= add_epsilon_src_nodes (dfa
, dest_nodes
, candidates
);
1816 if (BE (err
!= REG_NOERROR
, 0))
1819 /* Then, check the limitations in the current sift_context. */
1820 if (sctx
->limits
.nelem
)
1822 err
= check_subexp_limits (dfa
, dest_nodes
, candidates
, &sctx
->limits
,
1823 mctx
->bkref_ents
, str_idx
);
1824 if (BE (err
!= REG_NOERROR
, 0))
1829 sctx
->sifted_states
[str_idx
] = re_acquire_state (&err
, dfa
, dest_nodes
);
1830 if (BE (err
!= REG_NOERROR
, 0))
1834 if (candidates
&& mctx
->state_log
[str_idx
]->has_backref
)
1836 err
= sift_states_bkref (mctx
, sctx
, str_idx
, candidates
);
1837 if (BE (err
!= REG_NOERROR
, 0))
1843 static reg_errcode_t
1845 add_epsilon_src_nodes (const re_dfa_t
*dfa
, re_node_set
*dest_nodes
,
1846 const re_node_set
*candidates
)
1848 reg_errcode_t err
= REG_NOERROR
;
1851 re_dfastate_t
*state
= re_acquire_state (&err
, dfa
, dest_nodes
);
1852 if (BE (err
!= REG_NOERROR
, 0))
1855 if (!state
->inveclosure
.alloc
)
1857 err
= re_node_set_alloc (&state
->inveclosure
, dest_nodes
->nelem
);
1858 if (BE (err
!= REG_NOERROR
, 0))
1860 for (i
= 0; i
< dest_nodes
->nelem
; i
++)
1861 re_node_set_merge (&state
->inveclosure
,
1862 dfa
->inveclosures
+ dest_nodes
->elems
[i
]);
1864 return re_node_set_add_intersect (dest_nodes
, candidates
,
1865 &state
->inveclosure
);
1868 static reg_errcode_t
1870 sub_epsilon_src_nodes (const re_dfa_t
*dfa
, Idx node
, re_node_set
*dest_nodes
,
1871 const re_node_set
*candidates
)
1875 re_node_set
*inv_eclosure
= dfa
->inveclosures
+ node
;
1876 re_node_set except_nodes
;
1877 re_node_set_init_empty (&except_nodes
);
1878 for (ecl_idx
= 0; ecl_idx
< inv_eclosure
->nelem
; ++ecl_idx
)
1880 Idx cur_node
= inv_eclosure
->elems
[ecl_idx
];
1881 if (cur_node
== node
)
1883 if (IS_EPSILON_NODE (dfa
->nodes
[cur_node
].type
))
1885 Idx edst1
= dfa
->edests
[cur_node
].elems
[0];
1886 Idx edst2
= ((dfa
->edests
[cur_node
].nelem
> 1)
1887 ? dfa
->edests
[cur_node
].elems
[1] : REG_MISSING
);
1888 if ((!re_node_set_contains (inv_eclosure
, edst1
)
1889 && re_node_set_contains (dest_nodes
, edst1
))
1890 || (REG_VALID_NONZERO_INDEX (edst2
)
1891 && !re_node_set_contains (inv_eclosure
, edst2
)
1892 && re_node_set_contains (dest_nodes
, edst2
)))
1894 err
= re_node_set_add_intersect (&except_nodes
, candidates
,
1895 dfa
->inveclosures
+ cur_node
);
1896 if (BE (err
!= REG_NOERROR
, 0))
1898 re_node_set_free (&except_nodes
);
1904 for (ecl_idx
= 0; ecl_idx
< inv_eclosure
->nelem
; ++ecl_idx
)
1906 Idx cur_node
= inv_eclosure
->elems
[ecl_idx
];
1907 if (!re_node_set_contains (&except_nodes
, cur_node
))
1909 Idx idx
= re_node_set_contains (dest_nodes
, cur_node
) - 1;
1910 re_node_set_remove_at (dest_nodes
, idx
);
1913 re_node_set_free (&except_nodes
);
1919 check_dst_limits (const re_match_context_t
*mctx
, const re_node_set
*limits
,
1920 Idx dst_node
, Idx dst_idx
, Idx src_node
, Idx src_idx
)
1922 const re_dfa_t
*const dfa
= mctx
->dfa
;
1923 Idx lim_idx
, src_pos
, dst_pos
;
1925 Idx dst_bkref_idx
= search_cur_bkref_entry (mctx
, dst_idx
);
1926 Idx src_bkref_idx
= search_cur_bkref_entry (mctx
, src_idx
);
1927 for (lim_idx
= 0; lim_idx
< limits
->nelem
; ++lim_idx
)
1930 struct re_backref_cache_entry
*ent
;
1931 ent
= mctx
->bkref_ents
+ limits
->elems
[lim_idx
];
1932 subexp_idx
= dfa
->nodes
[ent
->node
].opr
.idx
;
1934 dst_pos
= check_dst_limits_calc_pos (mctx
, limits
->elems
[lim_idx
],
1935 subexp_idx
, dst_node
, dst_idx
,
1937 src_pos
= check_dst_limits_calc_pos (mctx
, limits
->elems
[lim_idx
],
1938 subexp_idx
, src_node
, src_idx
,
1942 <src> <dst> ( <subexp> )
1943 ( <subexp> ) <src> <dst>
1944 ( <subexp1> <src> <subexp2> <dst> <subexp3> ) */
1945 if (src_pos
== dst_pos
)
1946 continue; /* This is unrelated limitation. */
1955 check_dst_limits_calc_pos_1 (const re_match_context_t
*mctx
, int boundaries
,
1956 Idx subexp_idx
, Idx from_node
, Idx bkref_idx
)
1958 const re_dfa_t
*const dfa
= mctx
->dfa
;
1959 const re_node_set
*eclosures
= dfa
->eclosures
+ from_node
;
1962 /* Else, we are on the boundary: examine the nodes on the epsilon
1964 for (node_idx
= 0; node_idx
< eclosures
->nelem
; ++node_idx
)
1966 Idx node
= eclosures
->elems
[node_idx
];
1967 switch (dfa
->nodes
[node
].type
)
1970 if (bkref_idx
!= REG_MISSING
)
1972 struct re_backref_cache_entry
*ent
= mctx
->bkref_ents
+ bkref_idx
;
1978 if (ent
->node
!= node
)
1981 if (subexp_idx
< BITSET_WORD_BITS
1982 && !(ent
->eps_reachable_subexps_map
1983 & ((bitset_word_t
) 1 << subexp_idx
)))
1986 /* Recurse trying to reach the OP_OPEN_SUBEXP and
1987 OP_CLOSE_SUBEXP cases below. But, if the
1988 destination node is the same node as the source
1989 node, don't recurse because it would cause an
1990 infinite loop: a regex that exhibits this behavior
1992 dst
= dfa
->edests
[node
].elems
[0];
1993 if (dst
== from_node
)
1997 else /* if (boundaries & 2) */
2002 check_dst_limits_calc_pos_1 (mctx
, boundaries
, subexp_idx
,
2004 if (cpos
== -1 /* && (boundaries & 1) */)
2006 if (cpos
== 0 && (boundaries
& 2))
2009 if (subexp_idx
< BITSET_WORD_BITS
)
2010 ent
->eps_reachable_subexps_map
2011 &= ~((bitset_word_t
) 1 << subexp_idx
);
2013 while (ent
++->more
);
2017 case OP_OPEN_SUBEXP
:
2018 if ((boundaries
& 1) && subexp_idx
== dfa
->nodes
[node
].opr
.idx
)
2022 case OP_CLOSE_SUBEXP
:
2023 if ((boundaries
& 2) && subexp_idx
== dfa
->nodes
[node
].opr
.idx
)
2032 return (boundaries
& 2) ? 1 : 0;
2037 check_dst_limits_calc_pos (const re_match_context_t
*mctx
, Idx limit
,
2038 Idx subexp_idx
, Idx from_node
, Idx str_idx
,
2041 struct re_backref_cache_entry
*lim
= mctx
->bkref_ents
+ limit
;
2044 /* If we are outside the range of the subexpression, return -1 or 1. */
2045 if (str_idx
< lim
->subexp_from
)
2048 if (lim
->subexp_to
< str_idx
)
2051 /* If we are within the subexpression, return 0. */
2052 boundaries
= (str_idx
== lim
->subexp_from
);
2053 boundaries
|= (str_idx
== lim
->subexp_to
) << 1;
2054 if (boundaries
== 0)
2057 /* Else, examine epsilon closure. */
2058 return check_dst_limits_calc_pos_1 (mctx
, boundaries
, subexp_idx
,
2059 from_node
, bkref_idx
);
2062 /* Check the limitations of sub expressions LIMITS, and remove the nodes
2063 which are against limitations from DEST_NODES. */
2065 static reg_errcode_t
2067 check_subexp_limits (const re_dfa_t
*dfa
, re_node_set
*dest_nodes
,
2068 const re_node_set
*candidates
, re_node_set
*limits
,
2069 struct re_backref_cache_entry
*bkref_ents
, Idx str_idx
)
2072 Idx node_idx
, lim_idx
;
2074 for (lim_idx
= 0; lim_idx
< limits
->nelem
; ++lim_idx
)
2077 struct re_backref_cache_entry
*ent
;
2078 ent
= bkref_ents
+ limits
->elems
[lim_idx
];
2080 if (str_idx
<= ent
->subexp_from
|| ent
->str_idx
< str_idx
)
2081 continue; /* This is unrelated limitation. */
2083 subexp_idx
= dfa
->nodes
[ent
->node
].opr
.idx
;
2084 if (ent
->subexp_to
== str_idx
)
2086 Idx ops_node
= REG_MISSING
;
2087 Idx cls_node
= REG_MISSING
;
2088 for (node_idx
= 0; node_idx
< dest_nodes
->nelem
; ++node_idx
)
2090 Idx node
= dest_nodes
->elems
[node_idx
];
2091 re_token_type_t type
= dfa
->nodes
[node
].type
;
2092 if (type
== OP_OPEN_SUBEXP
2093 && subexp_idx
== dfa
->nodes
[node
].opr
.idx
)
2095 else if (type
== OP_CLOSE_SUBEXP
2096 && subexp_idx
== dfa
->nodes
[node
].opr
.idx
)
2100 /* Check the limitation of the open subexpression. */
2101 /* Note that (ent->subexp_to = str_idx != ent->subexp_from). */
2102 if (REG_VALID_INDEX (ops_node
))
2104 err
= sub_epsilon_src_nodes (dfa
, ops_node
, dest_nodes
,
2106 if (BE (err
!= REG_NOERROR
, 0))
2110 /* Check the limitation of the close subexpression. */
2111 if (REG_VALID_INDEX (cls_node
))
2112 for (node_idx
= 0; node_idx
< dest_nodes
->nelem
; ++node_idx
)
2114 Idx node
= dest_nodes
->elems
[node_idx
];
2115 if (!re_node_set_contains (dfa
->inveclosures
+ node
,
2117 && !re_node_set_contains (dfa
->eclosures
+ node
,
2120 /* It is against this limitation.
2121 Remove it form the current sifted state. */
2122 err
= sub_epsilon_src_nodes (dfa
, node
, dest_nodes
,
2124 if (BE (err
!= REG_NOERROR
, 0))
2130 else /* (ent->subexp_to != str_idx) */
2132 for (node_idx
= 0; node_idx
< dest_nodes
->nelem
; ++node_idx
)
2134 Idx node
= dest_nodes
->elems
[node_idx
];
2135 re_token_type_t type
= dfa
->nodes
[node
].type
;
2136 if (type
== OP_CLOSE_SUBEXP
|| type
== OP_OPEN_SUBEXP
)
2138 if (subexp_idx
!= dfa
->nodes
[node
].opr
.idx
)
2140 /* It is against this limitation.
2141 Remove it form the current sifted state. */
2142 err
= sub_epsilon_src_nodes (dfa
, node
, dest_nodes
,
2144 if (BE (err
!= REG_NOERROR
, 0))
2153 static reg_errcode_t
2155 sift_states_bkref (const re_match_context_t
*mctx
, re_sift_context_t
*sctx
,
2156 Idx str_idx
, const re_node_set
*candidates
)
2158 const re_dfa_t
*const dfa
= mctx
->dfa
;
2161 re_sift_context_t local_sctx
;
2162 Idx first_idx
= search_cur_bkref_entry (mctx
, str_idx
);
2164 if (first_idx
== REG_MISSING
)
2167 local_sctx
.sifted_states
= NULL
; /* Mark that it hasn't been initialized. */
2169 for (node_idx
= 0; node_idx
< candidates
->nelem
; ++node_idx
)
2172 re_token_type_t type
;
2173 struct re_backref_cache_entry
*entry
;
2174 node
= candidates
->elems
[node_idx
];
2175 type
= dfa
->nodes
[node
].type
;
2176 /* Avoid infinite loop for the REs like "()\1+". */
2177 if (node
== sctx
->last_node
&& str_idx
== sctx
->last_str_idx
)
2179 if (type
!= OP_BACK_REF
)
2182 entry
= mctx
->bkref_ents
+ first_idx
;
2183 enabled_idx
= first_idx
;
2190 re_dfastate_t
*cur_state
;
2192 if (entry
->node
!= node
)
2194 subexp_len
= entry
->subexp_to
- entry
->subexp_from
;
2195 to_idx
= str_idx
+ subexp_len
;
2196 dst_node
= (subexp_len
? dfa
->nexts
[node
]
2197 : dfa
->edests
[node
].elems
[0]);
2199 if (to_idx
> sctx
->last_str_idx
2200 || sctx
->sifted_states
[to_idx
] == NULL
2201 || !STATE_NODE_CONTAINS (sctx
->sifted_states
[to_idx
], dst_node
)
2202 || check_dst_limits (mctx
, &sctx
->limits
, node
,
2203 str_idx
, dst_node
, to_idx
))
2206 if (local_sctx
.sifted_states
== NULL
)
2209 err
= re_node_set_init_copy (&local_sctx
.limits
, &sctx
->limits
);
2210 if (BE (err
!= REG_NOERROR
, 0))
2213 local_sctx
.last_node
= node
;
2214 local_sctx
.last_str_idx
= str_idx
;
2215 ok
= re_node_set_insert (&local_sctx
.limits
, enabled_idx
);
2221 cur_state
= local_sctx
.sifted_states
[str_idx
];
2222 err
= sift_states_backward (mctx
, &local_sctx
);
2223 if (BE (err
!= REG_NOERROR
, 0))
2225 if (sctx
->limited_states
!= NULL
)
2227 err
= merge_state_array (dfa
, sctx
->limited_states
,
2228 local_sctx
.sifted_states
,
2230 if (BE (err
!= REG_NOERROR
, 0))
2233 local_sctx
.sifted_states
[str_idx
] = cur_state
;
2234 re_node_set_remove (&local_sctx
.limits
, enabled_idx
);
2236 /* mctx->bkref_ents may have changed, reload the pointer. */
2237 entry
= mctx
->bkref_ents
+ enabled_idx
;
2239 while (enabled_idx
++, entry
++->more
);
2243 if (local_sctx
.sifted_states
!= NULL
)
2245 re_node_set_free (&local_sctx
.limits
);
2252 #ifdef RE_ENABLE_I18N
2255 sift_states_iter_mb (const re_match_context_t
*mctx
, re_sift_context_t
*sctx
,
2256 Idx node_idx
, Idx str_idx
, Idx max_str_idx
)
2258 const re_dfa_t
*const dfa
= mctx
->dfa
;
2260 /* Check the node can accept `multi byte'. */
2261 naccepted
= check_node_accept_bytes (dfa
, node_idx
, &mctx
->input
, str_idx
);
2262 if (naccepted
> 0 && str_idx
+ naccepted
<= max_str_idx
&&
2263 !STATE_NODE_CONTAINS (sctx
->sifted_states
[str_idx
+ naccepted
],
2264 dfa
->nexts
[node_idx
]))
2265 /* The node can't accept the `multi byte', or the
2266 destination was already thrown away, then the node
2267 could't accept the current input `multi byte'. */
2269 /* Otherwise, it is sure that the node could accept
2270 `naccepted' bytes input. */
2273 #endif /* RE_ENABLE_I18N */
2276 /* Functions for state transition. */
2278 /* Return the next state to which the current state STATE will transit by
2279 accepting the current input byte, and update STATE_LOG if necessary.
2280 If STATE can accept a multibyte char/collating element/back reference
2281 update the destination of STATE_LOG. */
2283 static re_dfastate_t
*
2285 transit_state (reg_errcode_t
*err
, re_match_context_t
*mctx
,
2286 re_dfastate_t
*state
)
2288 re_dfastate_t
**trtable
;
2291 #ifdef RE_ENABLE_I18N
2292 /* If the current state can accept multibyte. */
2293 if (BE (state
->accept_mb
, 0))
2295 *err
= transit_state_mb (mctx
, state
);
2296 if (BE (*err
!= REG_NOERROR
, 0))
2299 #endif /* RE_ENABLE_I18N */
2301 /* Then decide the next state with the single byte. */
2304 /* don't use transition table */
2305 return transit_state_sb (err
, mctx
, state
);
2308 /* Use transition table */
2309 ch
= re_string_fetch_byte (&mctx
->input
);
2312 trtable
= state
->trtable
;
2313 if (BE (trtable
!= NULL
, 1))
2316 trtable
= state
->word_trtable
;
2317 if (BE (trtable
!= NULL
, 1))
2319 unsigned int context
;
2321 = re_string_context_at (&mctx
->input
,
2322 re_string_cur_idx (&mctx
->input
) - 1,
2324 if (IS_WORD_CONTEXT (context
))
2325 return trtable
[ch
+ SBC_MAX
];
2330 if (!build_trtable (mctx
->dfa
, state
))
2336 /* Retry, we now have a transition table. */
2340 /* Update the state_log if we need */
2343 merge_state_with_log (reg_errcode_t
*err
, re_match_context_t
*mctx
,
2344 re_dfastate_t
*next_state
)
2346 const re_dfa_t
*const dfa
= mctx
->dfa
;
2347 Idx cur_idx
= re_string_cur_idx (&mctx
->input
);
2349 if (cur_idx
> mctx
->state_log_top
)
2351 mctx
->state_log
[cur_idx
] = next_state
;
2352 mctx
->state_log_top
= cur_idx
;
2354 else if (mctx
->state_log
[cur_idx
] == 0)
2356 mctx
->state_log
[cur_idx
] = next_state
;
2360 re_dfastate_t
*pstate
;
2361 unsigned int context
;
2362 re_node_set next_nodes
, *log_nodes
, *table_nodes
= NULL
;
2363 /* If (state_log[cur_idx] != 0), it implies that cur_idx is
2364 the destination of a multibyte char/collating element/
2365 back reference. Then the next state is the union set of
2366 these destinations and the results of the transition table. */
2367 pstate
= mctx
->state_log
[cur_idx
];
2368 log_nodes
= pstate
->entrance_nodes
;
2369 if (next_state
!= NULL
)
2371 table_nodes
= next_state
->entrance_nodes
;
2372 *err
= re_node_set_init_union (&next_nodes
, table_nodes
,
2374 if (BE (*err
!= REG_NOERROR
, 0))
2378 next_nodes
= *log_nodes
;
2379 /* Note: We already add the nodes of the initial state,
2380 then we don't need to add them here. */
2382 context
= re_string_context_at (&mctx
->input
,
2383 re_string_cur_idx (&mctx
->input
) - 1,
2385 next_state
= mctx
->state_log
[cur_idx
]
2386 = re_acquire_state_context (err
, dfa
, &next_nodes
, context
);
2387 /* We don't need to check errors here, since the return value of
2388 this function is next_state and ERR is already set. */
2390 if (table_nodes
!= NULL
)
2391 re_node_set_free (&next_nodes
);
2394 if (BE (dfa
->nbackref
, 0) && next_state
!= NULL
)
2396 /* Check OP_OPEN_SUBEXP in the current state in case that we use them
2397 later. We must check them here, since the back references in the
2398 next state might use them. */
2399 *err
= check_subexp_matching_top (mctx
, &next_state
->nodes
,
2401 if (BE (*err
!= REG_NOERROR
, 0))
2404 /* If the next state has back references. */
2405 if (next_state
->has_backref
)
2407 *err
= transit_state_bkref (mctx
, &next_state
->nodes
);
2408 if (BE (*err
!= REG_NOERROR
, 0))
2410 next_state
= mctx
->state_log
[cur_idx
];
2417 /* Skip bytes in the input that correspond to part of a
2418 multi-byte match, then look in the log for a state
2419 from which to restart matching. */
2420 static re_dfastate_t
*
2422 find_recover_state (reg_errcode_t
*err
, re_match_context_t
*mctx
)
2424 re_dfastate_t
*cur_state
;
2427 Idx max
= mctx
->state_log_top
;
2428 Idx cur_str_idx
= re_string_cur_idx (&mctx
->input
);
2432 if (++cur_str_idx
> max
)
2434 re_string_skip_bytes (&mctx
->input
, 1);
2436 while (mctx
->state_log
[cur_str_idx
] == NULL
);
2438 cur_state
= merge_state_with_log (err
, mctx
, NULL
);
2440 while (*err
== REG_NOERROR
&& cur_state
== NULL
);
2444 /* Helper functions for transit_state. */
2446 /* From the node set CUR_NODES, pick up the nodes whose types are
2447 OP_OPEN_SUBEXP and which have corresponding back references in the regular
2448 expression. And register them to use them later for evaluating the
2449 correspoding back references. */
2451 static reg_errcode_t
2453 check_subexp_matching_top (re_match_context_t
*mctx
, re_node_set
*cur_nodes
,
2456 const re_dfa_t
*const dfa
= mctx
->dfa
;
2460 /* TODO: This isn't efficient.
2461 Because there might be more than one nodes whose types are
2462 OP_OPEN_SUBEXP and whose index is SUBEXP_IDX, we must check all
2465 for (node_idx
= 0; node_idx
< cur_nodes
->nelem
; ++node_idx
)
2467 Idx node
= cur_nodes
->elems
[node_idx
];
2468 if (dfa
->nodes
[node
].type
== OP_OPEN_SUBEXP
2469 && dfa
->nodes
[node
].opr
.idx
< BITSET_WORD_BITS
2470 && (dfa
->used_bkref_map
2471 & ((bitset_word_t
) 1 << dfa
->nodes
[node
].opr
.idx
)))
2473 err
= match_ctx_add_subtop (mctx
, node
, str_idx
);
2474 if (BE (err
!= REG_NOERROR
, 0))
2482 /* Return the next state to which the current state STATE will transit by
2483 accepting the current input byte. */
2485 static re_dfastate_t
*
2486 transit_state_sb (reg_errcode_t
*err
, re_match_context_t
*mctx
,
2487 re_dfastate_t
*state
)
2489 const re_dfa_t
*const dfa
= mctx
->dfa
;
2490 re_node_set next_nodes
;
2491 re_dfastate_t
*next_state
;
2492 Idx node_cnt
, cur_str_idx
= re_string_cur_idx (&mctx
->input
);
2493 unsigned int context
;
2495 *err
= re_node_set_alloc (&next_nodes
, state
->nodes
.nelem
+ 1);
2496 if (BE (*err
!= REG_NOERROR
, 0))
2498 for (node_cnt
= 0; node_cnt
< state
->nodes
.nelem
; ++node_cnt
)
2500 Idx cur_node
= state
->nodes
.elems
[node_cnt
];
2501 if (check_node_accept (mctx
, dfa
->nodes
+ cur_node
, cur_str_idx
))
2503 *err
= re_node_set_merge (&next_nodes
,
2504 dfa
->eclosures
+ dfa
->nexts
[cur_node
]);
2505 if (BE (*err
!= REG_NOERROR
, 0))
2507 re_node_set_free (&next_nodes
);
2512 context
= re_string_context_at (&mctx
->input
, cur_str_idx
, mctx
->eflags
);
2513 next_state
= re_acquire_state_context (err
, dfa
, &next_nodes
, context
);
2514 /* We don't need to check errors here, since the return value of
2515 this function is next_state and ERR is already set. */
2517 re_node_set_free (&next_nodes
);
2518 re_string_skip_bytes (&mctx
->input
, 1);
2523 #ifdef RE_ENABLE_I18N
2524 static reg_errcode_t
2526 transit_state_mb (re_match_context_t
*mctx
, re_dfastate_t
*pstate
)
2528 const re_dfa_t
*const dfa
= mctx
->dfa
;
2532 for (i
= 0; i
< pstate
->nodes
.nelem
; ++i
)
2534 re_node_set dest_nodes
, *new_nodes
;
2535 Idx cur_node_idx
= pstate
->nodes
.elems
[i
];
2538 unsigned int context
;
2539 re_dfastate_t
*dest_state
;
2541 if (!dfa
->nodes
[cur_node_idx
].accept_mb
)
2544 if (dfa
->nodes
[cur_node_idx
].constraint
)
2546 context
= re_string_context_at (&mctx
->input
,
2547 re_string_cur_idx (&mctx
->input
),
2549 if (NOT_SATISFY_NEXT_CONSTRAINT (dfa
->nodes
[cur_node_idx
].constraint
,
2554 /* How many bytes the node can accept? */
2555 naccepted
= check_node_accept_bytes (dfa
, cur_node_idx
, &mctx
->input
,
2556 re_string_cur_idx (&mctx
->input
));
2560 /* The node can accepts `naccepted' bytes. */
2561 dest_idx
= re_string_cur_idx (&mctx
->input
) + naccepted
;
2562 mctx
->max_mb_elem_len
= ((mctx
->max_mb_elem_len
< naccepted
) ? naccepted
2563 : mctx
->max_mb_elem_len
);
2564 err
= clean_state_log_if_needed (mctx
, dest_idx
);
2565 if (BE (err
!= REG_NOERROR
, 0))
2568 assert (dfa
->nexts
[cur_node_idx
] != REG_MISSING
);
2570 new_nodes
= dfa
->eclosures
+ dfa
->nexts
[cur_node_idx
];
2572 dest_state
= mctx
->state_log
[dest_idx
];
2573 if (dest_state
== NULL
)
2574 dest_nodes
= *new_nodes
;
2577 err
= re_node_set_init_union (&dest_nodes
,
2578 dest_state
->entrance_nodes
, new_nodes
);
2579 if (BE (err
!= REG_NOERROR
, 0))
2582 context
= re_string_context_at (&mctx
->input
, dest_idx
- 1,
2584 mctx
->state_log
[dest_idx
]
2585 = re_acquire_state_context (&err
, dfa
, &dest_nodes
, context
);
2586 if (dest_state
!= NULL
)
2587 re_node_set_free (&dest_nodes
);
2588 if (BE (mctx
->state_log
[dest_idx
] == NULL
&& err
!= REG_NOERROR
, 0))
2593 #endif /* RE_ENABLE_I18N */
2595 static reg_errcode_t
2597 transit_state_bkref (re_match_context_t
*mctx
, const re_node_set
*nodes
)
2599 const re_dfa_t
*const dfa
= mctx
->dfa
;
2602 Idx cur_str_idx
= re_string_cur_idx (&mctx
->input
);
2604 for (i
= 0; i
< nodes
->nelem
; ++i
)
2606 Idx dest_str_idx
, prev_nelem
, bkc_idx
;
2607 Idx node_idx
= nodes
->elems
[i
];
2608 unsigned int context
;
2609 const re_token_t
*node
= dfa
->nodes
+ node_idx
;
2610 re_node_set
*new_dest_nodes
;
2612 /* Check whether `node' is a backreference or not. */
2613 if (node
->type
!= OP_BACK_REF
)
2616 if (node
->constraint
)
2618 context
= re_string_context_at (&mctx
->input
, cur_str_idx
,
2620 if (NOT_SATISFY_NEXT_CONSTRAINT (node
->constraint
, context
))
2624 /* `node' is a backreference.
2625 Check the substring which the substring matched. */
2626 bkc_idx
= mctx
->nbkref_ents
;
2627 err
= get_subexp (mctx
, node_idx
, cur_str_idx
);
2628 if (BE (err
!= REG_NOERROR
, 0))
2631 /* And add the epsilon closures (which is `new_dest_nodes') of
2632 the backreference to appropriate state_log. */
2634 assert (dfa
->nexts
[node_idx
] != REG_MISSING
);
2636 for (; bkc_idx
< mctx
->nbkref_ents
; ++bkc_idx
)
2639 re_dfastate_t
*dest_state
;
2640 struct re_backref_cache_entry
*bkref_ent
;
2641 bkref_ent
= mctx
->bkref_ents
+ bkc_idx
;
2642 if (bkref_ent
->node
!= node_idx
|| bkref_ent
->str_idx
!= cur_str_idx
)
2644 subexp_len
= bkref_ent
->subexp_to
- bkref_ent
->subexp_from
;
2645 new_dest_nodes
= (subexp_len
== 0
2646 ? dfa
->eclosures
+ dfa
->edests
[node_idx
].elems
[0]
2647 : dfa
->eclosures
+ dfa
->nexts
[node_idx
]);
2648 dest_str_idx
= (cur_str_idx
+ bkref_ent
->subexp_to
2649 - bkref_ent
->subexp_from
);
2650 context
= re_string_context_at (&mctx
->input
, dest_str_idx
- 1,
2652 dest_state
= mctx
->state_log
[dest_str_idx
];
2653 prev_nelem
= ((mctx
->state_log
[cur_str_idx
] == NULL
) ? 0
2654 : mctx
->state_log
[cur_str_idx
]->nodes
.nelem
);
2655 /* Add `new_dest_node' to state_log. */
2656 if (dest_state
== NULL
)
2658 mctx
->state_log
[dest_str_idx
]
2659 = re_acquire_state_context (&err
, dfa
, new_dest_nodes
,
2661 if (BE (mctx
->state_log
[dest_str_idx
] == NULL
2662 && err
!= REG_NOERROR
, 0))
2667 re_node_set dest_nodes
;
2668 err
= re_node_set_init_union (&dest_nodes
,
2669 dest_state
->entrance_nodes
,
2671 if (BE (err
!= REG_NOERROR
, 0))
2673 re_node_set_free (&dest_nodes
);
2676 mctx
->state_log
[dest_str_idx
]
2677 = re_acquire_state_context (&err
, dfa
, &dest_nodes
, context
);
2678 re_node_set_free (&dest_nodes
);
2679 if (BE (mctx
->state_log
[dest_str_idx
] == NULL
2680 && err
!= REG_NOERROR
, 0))
2683 /* We need to check recursively if the backreference can epsilon
2686 && mctx
->state_log
[cur_str_idx
]->nodes
.nelem
> prev_nelem
)
2688 err
= check_subexp_matching_top (mctx
, new_dest_nodes
,
2690 if (BE (err
!= REG_NOERROR
, 0))
2692 err
= transit_state_bkref (mctx
, new_dest_nodes
);
2693 if (BE (err
!= REG_NOERROR
, 0))
2703 /* Enumerate all the candidates which the backreference BKREF_NODE can match
2704 at BKREF_STR_IDX, and register them by match_ctx_add_entry().
2705 Note that we might collect inappropriate candidates here.
2706 However, the cost of checking them strictly here is too high, then we
2707 delay these checking for prune_impossible_nodes(). */
2709 static reg_errcode_t
2711 get_subexp (re_match_context_t
*mctx
, Idx bkref_node
, Idx bkref_str_idx
)
2713 const re_dfa_t
*const dfa
= mctx
->dfa
;
2714 Idx subexp_num
, sub_top_idx
;
2715 const char *buf
= (const char *) re_string_get_buffer (&mctx
->input
);
2716 /* Return if we have already checked BKREF_NODE at BKREF_STR_IDX. */
2717 Idx cache_idx
= search_cur_bkref_entry (mctx
, bkref_str_idx
);
2718 if (cache_idx
!= REG_MISSING
)
2720 const struct re_backref_cache_entry
*entry
2721 = mctx
->bkref_ents
+ cache_idx
;
2723 if (entry
->node
== bkref_node
)
2724 return REG_NOERROR
; /* We already checked it. */
2725 while (entry
++->more
);
2728 subexp_num
= dfa
->nodes
[bkref_node
].opr
.idx
;
2730 /* For each sub expression */
2731 for (sub_top_idx
= 0; sub_top_idx
< mctx
->nsub_tops
; ++sub_top_idx
)
2734 re_sub_match_top_t
*sub_top
= mctx
->sub_tops
[sub_top_idx
];
2735 re_sub_match_last_t
*sub_last
;
2736 Idx sub_last_idx
, sl_str
, bkref_str_off
;
2738 if (dfa
->nodes
[sub_top
->node
].opr
.idx
!= subexp_num
)
2739 continue; /* It isn't related. */
2741 sl_str
= sub_top
->str_idx
;
2742 bkref_str_off
= bkref_str_idx
;
2743 /* At first, check the last node of sub expressions we already
2745 for (sub_last_idx
= 0; sub_last_idx
< sub_top
->nlasts
; ++sub_last_idx
)
2747 regoff_t sl_str_diff
;
2748 sub_last
= sub_top
->lasts
[sub_last_idx
];
2749 sl_str_diff
= sub_last
->str_idx
- sl_str
;
2750 /* The matched string by the sub expression match with the substring
2751 at the back reference? */
2752 if (sl_str_diff
> 0)
2754 if (BE (bkref_str_off
+ sl_str_diff
> mctx
->input
.valid_len
, 0))
2756 /* Not enough chars for a successful match. */
2757 if (bkref_str_off
+ sl_str_diff
> mctx
->input
.len
)
2760 err
= clean_state_log_if_needed (mctx
,
2763 if (BE (err
!= REG_NOERROR
, 0))
2765 buf
= (const char *) re_string_get_buffer (&mctx
->input
);
2767 if (memcmp (buf
+ bkref_str_off
, buf
+ sl_str
, sl_str_diff
) != 0)
2768 /* We don't need to search this sub expression any more. */
2771 bkref_str_off
+= sl_str_diff
;
2772 sl_str
+= sl_str_diff
;
2773 err
= get_subexp_sub (mctx
, sub_top
, sub_last
, bkref_node
,
2776 /* Reload buf, since the preceding call might have reallocated
2778 buf
= (const char *) re_string_get_buffer (&mctx
->input
);
2780 if (err
== REG_NOMATCH
)
2782 if (BE (err
!= REG_NOERROR
, 0))
2786 if (sub_last_idx
< sub_top
->nlasts
)
2788 if (sub_last_idx
> 0)
2790 /* Then, search for the other last nodes of the sub expression. */
2791 for (; sl_str
<= bkref_str_idx
; ++sl_str
)
2794 regoff_t sl_str_off
;
2795 const re_node_set
*nodes
;
2796 sl_str_off
= sl_str
- sub_top
->str_idx
;
2797 /* The matched string by the sub expression match with the substring
2798 at the back reference? */
2801 if (BE (bkref_str_off
>= mctx
->input
.valid_len
, 0))
2803 /* If we are at the end of the input, we cannot match. */
2804 if (bkref_str_off
>= mctx
->input
.len
)
2807 err
= extend_buffers (mctx
);
2808 if (BE (err
!= REG_NOERROR
, 0))
2811 buf
= (const char *) re_string_get_buffer (&mctx
->input
);
2813 if (buf
[bkref_str_off
++] != buf
[sl_str
- 1])
2814 break; /* We don't need to search this sub expression
2817 if (mctx
->state_log
[sl_str
] == NULL
)
2819 /* Does this state have a ')' of the sub expression? */
2820 nodes
= &mctx
->state_log
[sl_str
]->nodes
;
2821 cls_node
= find_subexp_node (dfa
, nodes
, subexp_num
,
2823 if (cls_node
== REG_MISSING
)
2825 if (sub_top
->path
== NULL
)
2827 sub_top
->path
= calloc (sizeof (state_array_t
),
2828 sl_str
- sub_top
->str_idx
+ 1);
2829 if (sub_top
->path
== NULL
)
2832 /* Can the OP_OPEN_SUBEXP node arrive the OP_CLOSE_SUBEXP node
2833 in the current context? */
2834 err
= check_arrival (mctx
, sub_top
->path
, sub_top
->node
,
2835 sub_top
->str_idx
, cls_node
, sl_str
,
2837 if (err
== REG_NOMATCH
)
2839 if (BE (err
!= REG_NOERROR
, 0))
2841 sub_last
= match_ctx_add_sublast (sub_top
, cls_node
, sl_str
);
2842 if (BE (sub_last
== NULL
, 0))
2844 err
= get_subexp_sub (mctx
, sub_top
, sub_last
, bkref_node
,
2846 if (err
== REG_NOMATCH
)
2853 /* Helper functions for get_subexp(). */
2855 /* Check SUB_LAST can arrive to the back reference BKREF_NODE at BKREF_STR.
2856 If it can arrive, register the sub expression expressed with SUB_TOP
2859 static reg_errcode_t
2861 get_subexp_sub (re_match_context_t
*mctx
, const re_sub_match_top_t
*sub_top
,
2862 re_sub_match_last_t
*sub_last
, Idx bkref_node
, Idx bkref_str
)
2866 /* Can the subexpression arrive the back reference? */
2867 err
= check_arrival (mctx
, &sub_last
->path
, sub_last
->node
,
2868 sub_last
->str_idx
, bkref_node
, bkref_str
,
2870 if (err
!= REG_NOERROR
)
2872 err
= match_ctx_add_entry (mctx
, bkref_node
, bkref_str
, sub_top
->str_idx
,
2874 if (BE (err
!= REG_NOERROR
, 0))
2876 to_idx
= bkref_str
+ sub_last
->str_idx
- sub_top
->str_idx
;
2877 return clean_state_log_if_needed (mctx
, to_idx
);
2880 /* Find the first node which is '(' or ')' and whose index is SUBEXP_IDX.
2881 Search '(' if FL_OPEN, or search ')' otherwise.
2882 TODO: This function isn't efficient...
2883 Because there might be more than one nodes whose types are
2884 OP_OPEN_SUBEXP and whose index is SUBEXP_IDX, we must check all
2890 find_subexp_node (const re_dfa_t
*dfa
, const re_node_set
*nodes
,
2891 Idx subexp_idx
, int type
)
2894 for (cls_idx
= 0; cls_idx
< nodes
->nelem
; ++cls_idx
)
2896 Idx cls_node
= nodes
->elems
[cls_idx
];
2897 const re_token_t
*node
= dfa
->nodes
+ cls_node
;
2898 if (node
->type
== type
2899 && node
->opr
.idx
== subexp_idx
)
2905 /* Check whether the node TOP_NODE at TOP_STR can arrive to the node
2906 LAST_NODE at LAST_STR. We record the path onto PATH since it will be
2908 Return REG_NOERROR if it can arrive, or REG_NOMATCH otherwise. */
2910 static reg_errcode_t
2912 check_arrival (re_match_context_t
*mctx
, state_array_t
*path
, Idx top_node
,
2913 Idx top_str
, Idx last_node
, Idx last_str
, int type
)
2915 const re_dfa_t
*const dfa
= mctx
->dfa
;
2916 reg_errcode_t err
= REG_NOERROR
;
2917 Idx subexp_num
, backup_cur_idx
, str_idx
, null_cnt
;
2918 re_dfastate_t
*cur_state
= NULL
;
2919 re_node_set
*cur_nodes
, next_nodes
;
2920 re_dfastate_t
**backup_state_log
;
2921 unsigned int context
;
2923 subexp_num
= dfa
->nodes
[top_node
].opr
.idx
;
2924 /* Extend the buffer if we need. */
2925 if (BE (path
->alloc
< last_str
+ mctx
->max_mb_elem_len
+ 1, 0))
2927 re_dfastate_t
**new_array
;
2928 Idx old_alloc
= path
->alloc
;
2929 Idx new_alloc
= old_alloc
+ last_str
+ mctx
->max_mb_elem_len
+ 1;
2930 if (BE (new_alloc
< old_alloc
, 0)
2931 || BE (SIZE_MAX
/ sizeof (re_dfastate_t
*) < new_alloc
, 0))
2933 new_array
= re_realloc (path
->array
, re_dfastate_t
*, new_alloc
);
2934 if (BE (new_array
== NULL
, 0))
2936 path
->array
= new_array
;
2937 path
->alloc
= new_alloc
;
2938 memset (new_array
+ old_alloc
, '\0',
2939 sizeof (re_dfastate_t
*) * (path
->alloc
- old_alloc
));
2942 str_idx
= path
->next_idx
? path
->next_idx
: top_str
;
2944 /* Temporary modify MCTX. */
2945 backup_state_log
= mctx
->state_log
;
2946 backup_cur_idx
= mctx
->input
.cur_idx
;
2947 mctx
->state_log
= path
->array
;
2948 mctx
->input
.cur_idx
= str_idx
;
2950 /* Setup initial node set. */
2951 context
= re_string_context_at (&mctx
->input
, str_idx
- 1, mctx
->eflags
);
2952 if (str_idx
== top_str
)
2954 err
= re_node_set_init_1 (&next_nodes
, top_node
);
2955 if (BE (err
!= REG_NOERROR
, 0))
2957 err
= check_arrival_expand_ecl (dfa
, &next_nodes
, subexp_num
, type
);
2958 if (BE (err
!= REG_NOERROR
, 0))
2960 re_node_set_free (&next_nodes
);
2966 cur_state
= mctx
->state_log
[str_idx
];
2967 if (cur_state
&& cur_state
->has_backref
)
2969 err
= re_node_set_init_copy (&next_nodes
, &cur_state
->nodes
);
2970 if (BE (err
!= REG_NOERROR
, 0))
2974 re_node_set_init_empty (&next_nodes
);
2976 if (str_idx
== top_str
|| (cur_state
&& cur_state
->has_backref
))
2978 if (next_nodes
.nelem
)
2980 err
= expand_bkref_cache (mctx
, &next_nodes
, str_idx
,
2982 if (BE (err
!= REG_NOERROR
, 0))
2984 re_node_set_free (&next_nodes
);
2988 cur_state
= re_acquire_state_context (&err
, dfa
, &next_nodes
, context
);
2989 if (BE (cur_state
== NULL
&& err
!= REG_NOERROR
, 0))
2991 re_node_set_free (&next_nodes
);
2994 mctx
->state_log
[str_idx
] = cur_state
;
2997 for (null_cnt
= 0; str_idx
< last_str
&& null_cnt
<= mctx
->max_mb_elem_len
;)
2999 re_node_set_empty (&next_nodes
);
3000 if (mctx
->state_log
[str_idx
+ 1])
3002 err
= re_node_set_merge (&next_nodes
,
3003 &mctx
->state_log
[str_idx
+ 1]->nodes
);
3004 if (BE (err
!= REG_NOERROR
, 0))
3006 re_node_set_free (&next_nodes
);
3012 err
= check_arrival_add_next_nodes (mctx
, str_idx
,
3013 &cur_state
->non_eps_nodes
,
3015 if (BE (err
!= REG_NOERROR
, 0))
3017 re_node_set_free (&next_nodes
);
3022 if (next_nodes
.nelem
)
3024 err
= check_arrival_expand_ecl (dfa
, &next_nodes
, subexp_num
, type
);
3025 if (BE (err
!= REG_NOERROR
, 0))
3027 re_node_set_free (&next_nodes
);
3030 err
= expand_bkref_cache (mctx
, &next_nodes
, str_idx
,
3032 if (BE (err
!= REG_NOERROR
, 0))
3034 re_node_set_free (&next_nodes
);
3038 context
= re_string_context_at (&mctx
->input
, str_idx
- 1, mctx
->eflags
);
3039 cur_state
= re_acquire_state_context (&err
, dfa
, &next_nodes
, context
);
3040 if (BE (cur_state
== NULL
&& err
!= REG_NOERROR
, 0))
3042 re_node_set_free (&next_nodes
);
3045 mctx
->state_log
[str_idx
] = cur_state
;
3046 null_cnt
= cur_state
== NULL
? null_cnt
+ 1 : 0;
3048 re_node_set_free (&next_nodes
);
3049 cur_nodes
= (mctx
->state_log
[last_str
] == NULL
? NULL
3050 : &mctx
->state_log
[last_str
]->nodes
);
3051 path
->next_idx
= str_idx
;
3054 mctx
->state_log
= backup_state_log
;
3055 mctx
->input
.cur_idx
= backup_cur_idx
;
3057 /* Then check the current node set has the node LAST_NODE. */
3058 if (cur_nodes
!= NULL
&& re_node_set_contains (cur_nodes
, last_node
))
3064 /* Helper functions for check_arrival. */
3066 /* Calculate the destination nodes of CUR_NODES at STR_IDX, and append them
3068 TODO: This function is similar to the functions transit_state*(),
3069 however this function has many additional works.
3070 Can't we unify them? */
3072 static reg_errcode_t
3074 check_arrival_add_next_nodes (re_match_context_t
*mctx
, Idx str_idx
,
3075 re_node_set
*cur_nodes
, re_node_set
*next_nodes
)
3077 const re_dfa_t
*const dfa
= mctx
->dfa
;
3080 reg_errcode_t err
= REG_NOERROR
;
3081 re_node_set union_set
;
3082 re_node_set_init_empty (&union_set
);
3083 for (cur_idx
= 0; cur_idx
< cur_nodes
->nelem
; ++cur_idx
)
3086 Idx cur_node
= cur_nodes
->elems
[cur_idx
];
3088 re_token_type_t type
= dfa
->nodes
[cur_node
].type
;
3089 assert (!IS_EPSILON_NODE (type
));
3091 #ifdef RE_ENABLE_I18N
3092 /* If the node may accept `multi byte'. */
3093 if (dfa
->nodes
[cur_node
].accept_mb
)
3095 naccepted
= check_node_accept_bytes (dfa
, cur_node
, &mctx
->input
,
3099 re_dfastate_t
*dest_state
;
3100 Idx next_node
= dfa
->nexts
[cur_node
];
3101 Idx next_idx
= str_idx
+ naccepted
;
3102 dest_state
= mctx
->state_log
[next_idx
];
3103 re_node_set_empty (&union_set
);
3106 err
= re_node_set_merge (&union_set
, &dest_state
->nodes
);
3107 if (BE (err
!= REG_NOERROR
, 0))
3109 re_node_set_free (&union_set
);
3113 ok
= re_node_set_insert (&union_set
, next_node
);
3116 re_node_set_free (&union_set
);
3119 mctx
->state_log
[next_idx
] = re_acquire_state (&err
, dfa
,
3121 if (BE (mctx
->state_log
[next_idx
] == NULL
3122 && err
!= REG_NOERROR
, 0))
3124 re_node_set_free (&union_set
);
3129 #endif /* RE_ENABLE_I18N */
3131 || check_node_accept (mctx
, dfa
->nodes
+ cur_node
, str_idx
))
3133 ok
= re_node_set_insert (next_nodes
, dfa
->nexts
[cur_node
]);
3136 re_node_set_free (&union_set
);
3141 re_node_set_free (&union_set
);
3145 /* For all the nodes in CUR_NODES, add the epsilon closures of them to
3146 CUR_NODES, however exclude the nodes which are:
3147 - inside the sub expression whose number is EX_SUBEXP, if FL_OPEN.
3148 - out of the sub expression whose number is EX_SUBEXP, if !FL_OPEN.
3151 static reg_errcode_t
3153 check_arrival_expand_ecl (const re_dfa_t
*dfa
, re_node_set
*cur_nodes
,
3154 Idx ex_subexp
, int type
)
3157 Idx idx
, outside_node
;
3158 re_node_set new_nodes
;
3160 assert (cur_nodes
->nelem
);
3162 err
= re_node_set_alloc (&new_nodes
, cur_nodes
->nelem
);
3163 if (BE (err
!= REG_NOERROR
, 0))
3165 /* Create a new node set NEW_NODES with the nodes which are epsilon
3166 closures of the node in CUR_NODES. */
3168 for (idx
= 0; idx
< cur_nodes
->nelem
; ++idx
)
3170 Idx cur_node
= cur_nodes
->elems
[idx
];
3171 const re_node_set
*eclosure
= dfa
->eclosures
+ cur_node
;
3172 outside_node
= find_subexp_node (dfa
, eclosure
, ex_subexp
, type
);
3173 if (outside_node
== REG_MISSING
)
3175 /* There are no problematic nodes, just merge them. */
3176 err
= re_node_set_merge (&new_nodes
, eclosure
);
3177 if (BE (err
!= REG_NOERROR
, 0))
3179 re_node_set_free (&new_nodes
);
3185 /* There are problematic nodes, re-calculate incrementally. */
3186 err
= check_arrival_expand_ecl_sub (dfa
, &new_nodes
, cur_node
,
3188 if (BE (err
!= REG_NOERROR
, 0))
3190 re_node_set_free (&new_nodes
);
3195 re_node_set_free (cur_nodes
);
3196 *cur_nodes
= new_nodes
;
3200 /* Helper function for check_arrival_expand_ecl.
3201 Check incrementally the epsilon closure of TARGET, and if it isn't
3202 problematic append it to DST_NODES. */
3204 static reg_errcode_t
3206 check_arrival_expand_ecl_sub (const re_dfa_t
*dfa
, re_node_set
*dst_nodes
,
3207 Idx target
, Idx ex_subexp
, int type
)
3210 for (cur_node
= target
; !re_node_set_contains (dst_nodes
, cur_node
);)
3214 if (dfa
->nodes
[cur_node
].type
== type
3215 && dfa
->nodes
[cur_node
].opr
.idx
== ex_subexp
)
3217 if (type
== OP_CLOSE_SUBEXP
)
3219 ok
= re_node_set_insert (dst_nodes
, cur_node
);
3225 ok
= re_node_set_insert (dst_nodes
, cur_node
);
3228 if (dfa
->edests
[cur_node
].nelem
== 0)
3230 if (dfa
->edests
[cur_node
].nelem
== 2)
3233 err
= check_arrival_expand_ecl_sub (dfa
, dst_nodes
,
3234 dfa
->edests
[cur_node
].elems
[1],
3236 if (BE (err
!= REG_NOERROR
, 0))
3239 cur_node
= dfa
->edests
[cur_node
].elems
[0];
3245 /* For all the back references in the current state, calculate the
3246 destination of the back references by the appropriate entry
3247 in MCTX->BKREF_ENTS. */
3249 static reg_errcode_t
3251 expand_bkref_cache (re_match_context_t
*mctx
, re_node_set
*cur_nodes
,
3252 Idx cur_str
, Idx subexp_num
, int type
)
3254 const re_dfa_t
*const dfa
= mctx
->dfa
;
3256 Idx cache_idx_start
= search_cur_bkref_entry (mctx
, cur_str
);
3257 struct re_backref_cache_entry
*ent
;
3259 if (cache_idx_start
== REG_MISSING
)
3263 ent
= mctx
->bkref_ents
+ cache_idx_start
;
3266 Idx to_idx
, next_node
;
3268 /* Is this entry ENT is appropriate? */
3269 if (!re_node_set_contains (cur_nodes
, ent
->node
))
3272 to_idx
= cur_str
+ ent
->subexp_to
- ent
->subexp_from
;
3273 /* Calculate the destination of the back reference, and append it
3274 to MCTX->STATE_LOG. */
3275 if (to_idx
== cur_str
)
3277 /* The backreference did epsilon transit, we must re-check all the
3278 node in the current state. */
3279 re_node_set new_dests
;
3280 reg_errcode_t err2
, err3
;
3281 next_node
= dfa
->edests
[ent
->node
].elems
[0];
3282 if (re_node_set_contains (cur_nodes
, next_node
))
3284 err
= re_node_set_init_1 (&new_dests
, next_node
);
3285 err2
= check_arrival_expand_ecl (dfa
, &new_dests
, subexp_num
, type
);
3286 err3
= re_node_set_merge (cur_nodes
, &new_dests
);
3287 re_node_set_free (&new_dests
);
3288 if (BE (err
!= REG_NOERROR
|| err2
!= REG_NOERROR
3289 || err3
!= REG_NOERROR
, 0))
3291 err
= (err
!= REG_NOERROR
? err
3292 : (err2
!= REG_NOERROR
? err2
: err3
));
3295 /* TODO: It is still inefficient... */
3300 re_node_set union_set
;
3301 next_node
= dfa
->nexts
[ent
->node
];
3302 if (mctx
->state_log
[to_idx
])
3305 if (re_node_set_contains (&mctx
->state_log
[to_idx
]->nodes
,
3308 err
= re_node_set_init_copy (&union_set
,
3309 &mctx
->state_log
[to_idx
]->nodes
);
3310 ok
= re_node_set_insert (&union_set
, next_node
);
3311 if (BE (err
!= REG_NOERROR
|| ! ok
, 0))
3313 re_node_set_free (&union_set
);
3314 err
= err
!= REG_NOERROR
? err
: REG_ESPACE
;
3320 err
= re_node_set_init_1 (&union_set
, next_node
);
3321 if (BE (err
!= REG_NOERROR
, 0))
3324 mctx
->state_log
[to_idx
] = re_acquire_state (&err
, dfa
, &union_set
);
3325 re_node_set_free (&union_set
);
3326 if (BE (mctx
->state_log
[to_idx
] == NULL
3327 && err
!= REG_NOERROR
, 0))
3331 while (ent
++->more
);
3335 /* Build transition table for the state.
3336 Return true if successful. */
3340 build_trtable (const re_dfa_t
*dfa
, re_dfastate_t
*state
)
3345 bool need_word_trtable
= false;
3346 bitset_word_t elem
, mask
;
3347 bool dests_node_malloced
= false;
3348 bool dest_states_malloced
= false;
3349 Idx ndests
; /* Number of the destination states from `state'. */
3350 re_dfastate_t
**trtable
;
3351 re_dfastate_t
**dest_states
= NULL
, **dest_states_word
, **dest_states_nl
;
3352 re_node_set follows
, *dests_node
;
3354 bitset_t acceptable
;
3358 re_node_set dests_node
[SBC_MAX
];
3359 bitset_t dests_ch
[SBC_MAX
];
3362 /* We build DFA states which corresponds to the destination nodes
3363 from `state'. `dests_node[i]' represents the nodes which i-th
3364 destination state contains, and `dests_ch[i]' represents the
3365 characters which i-th destination state accepts. */
3366 if (__libc_use_alloca (sizeof (struct dests_alloc
)))
3367 dests_alloc
= (struct dests_alloc
*) alloca (sizeof (struct dests_alloc
));
3370 dests_alloc
= re_malloc (struct dests_alloc
, 1);
3371 if (BE (dests_alloc
== NULL
, 0))
3373 dests_node_malloced
= true;
3375 dests_node
= dests_alloc
->dests_node
;
3376 dests_ch
= dests_alloc
->dests_ch
;
3378 /* Initialize transiton table. */
3379 state
->word_trtable
= state
->trtable
= NULL
;
3381 /* At first, group all nodes belonging to `state' into several
3383 ndests
= group_nodes_into_DFAstates (dfa
, state
, dests_node
, dests_ch
);
3384 if (BE (! REG_VALID_NONZERO_INDEX (ndests
), 0))
3386 if (dests_node_malloced
)
3390 state
->trtable
= (re_dfastate_t
**)
3391 calloc (sizeof (re_dfastate_t
*), SBC_MAX
);
3397 err
= re_node_set_alloc (&follows
, ndests
+ 1);
3398 if (BE (err
!= REG_NOERROR
, 0))
3401 /* Avoid arithmetic overflow in size calculation. */
3402 if (BE ((((SIZE_MAX
- (sizeof (re_node_set
) + sizeof (bitset_t
)) * SBC_MAX
)
3403 / (3 * sizeof (re_dfastate_t
*)))
3408 if (__libc_use_alloca ((sizeof (re_node_set
) + sizeof (bitset_t
)) * SBC_MAX
3409 + ndests
* 3 * sizeof (re_dfastate_t
*)))
3410 dest_states
= (re_dfastate_t
**)
3411 alloca (ndests
* 3 * sizeof (re_dfastate_t
*));
3414 dest_states
= (re_dfastate_t
**)
3415 malloc (ndests
* 3 * sizeof (re_dfastate_t
*));
3416 if (BE (dest_states
== NULL
, 0))
3419 if (dest_states_malloced
)
3421 re_node_set_free (&follows
);
3422 for (i
= 0; i
< ndests
; ++i
)
3423 re_node_set_free (dests_node
+ i
);
3424 if (dests_node_malloced
)
3428 dest_states_malloced
= true;
3430 dest_states_word
= dest_states
+ ndests
;
3431 dest_states_nl
= dest_states_word
+ ndests
;
3432 bitset_empty (acceptable
);
3434 /* Then build the states for all destinations. */
3435 for (i
= 0; i
< ndests
; ++i
)
3438 re_node_set_empty (&follows
);
3439 /* Merge the follows of this destination states. */
3440 for (j
= 0; j
< dests_node
[i
].nelem
; ++j
)
3442 next_node
= dfa
->nexts
[dests_node
[i
].elems
[j
]];
3443 if (next_node
!= REG_MISSING
)
3445 err
= re_node_set_merge (&follows
, dfa
->eclosures
+ next_node
);
3446 if (BE (err
!= REG_NOERROR
, 0))
3450 dest_states
[i
] = re_acquire_state_context (&err
, dfa
, &follows
, 0);
3451 if (BE (dest_states
[i
] == NULL
&& err
!= REG_NOERROR
, 0))
3453 /* If the new state has context constraint,
3454 build appropriate states for these contexts. */
3455 if (dest_states
[i
]->has_constraint
)
3457 dest_states_word
[i
] = re_acquire_state_context (&err
, dfa
, &follows
,
3459 if (BE (dest_states_word
[i
] == NULL
&& err
!= REG_NOERROR
, 0))
3462 if (dest_states
[i
] != dest_states_word
[i
] && dfa
->mb_cur_max
> 1)
3463 need_word_trtable
= true;
3465 dest_states_nl
[i
] = re_acquire_state_context (&err
, dfa
, &follows
,
3467 if (BE (dest_states_nl
[i
] == NULL
&& err
!= REG_NOERROR
, 0))
3472 dest_states_word
[i
] = dest_states
[i
];
3473 dest_states_nl
[i
] = dest_states
[i
];
3475 bitset_merge (acceptable
, dests_ch
[i
]);
3478 if (!BE (need_word_trtable
, 0))
3480 /* We don't care about whether the following character is a word
3481 character, or we are in a single-byte character set so we can
3482 discern by looking at the character code: allocate a
3483 256-entry transition table. */
3484 trtable
= state
->trtable
=
3485 (re_dfastate_t
**) calloc (sizeof (re_dfastate_t
*), SBC_MAX
);
3486 if (BE (trtable
== NULL
, 0))
3489 /* For all characters ch...: */
3490 for (i
= 0; i
< BITSET_WORDS
; ++i
)
3491 for (ch
= i
* BITSET_WORD_BITS
, elem
= acceptable
[i
], mask
= 1;
3493 mask
<<= 1, elem
>>= 1, ++ch
)
3494 if (BE (elem
& 1, 0))
3496 /* There must be exactly one destination which accepts
3497 character ch. See group_nodes_into_DFAstates. */
3498 for (j
= 0; (dests_ch
[j
][i
] & mask
) == 0; ++j
)
3501 /* j-th destination accepts the word character ch. */
3502 if (dfa
->word_char
[i
] & mask
)
3503 trtable
[ch
] = dest_states_word
[j
];
3505 trtable
[ch
] = dest_states
[j
];
3510 /* We care about whether the following character is a word
3511 character, and we are in a multi-byte character set: discern
3512 by looking at the character code: build two 256-entry
3513 transition tables, one starting at trtable[0] and one
3514 starting at trtable[SBC_MAX]. */
3515 trtable
= state
->word_trtable
=
3516 (re_dfastate_t
**) calloc (sizeof (re_dfastate_t
*), 2 * SBC_MAX
);
3517 if (BE (trtable
== NULL
, 0))
3520 /* For all characters ch...: */
3521 for (i
= 0; i
< BITSET_WORDS
; ++i
)
3522 for (ch
= i
* BITSET_WORD_BITS
, elem
= acceptable
[i
], mask
= 1;
3524 mask
<<= 1, elem
>>= 1, ++ch
)
3525 if (BE (elem
& 1, 0))
3527 /* There must be exactly one destination which accepts
3528 character ch. See group_nodes_into_DFAstates. */
3529 for (j
= 0; (dests_ch
[j
][i
] & mask
) == 0; ++j
)
3532 /* j-th destination accepts the word character ch. */
3533 trtable
[ch
] = dest_states
[j
];
3534 trtable
[ch
+ SBC_MAX
] = dest_states_word
[j
];
3539 if (bitset_contain (acceptable
, NEWLINE_CHAR
))
3541 /* The current state accepts newline character. */
3542 for (j
= 0; j
< ndests
; ++j
)
3543 if (bitset_contain (dests_ch
[j
], NEWLINE_CHAR
))
3545 /* k-th destination accepts newline character. */
3546 trtable
[NEWLINE_CHAR
] = dest_states_nl
[j
];
3547 if (need_word_trtable
)
3548 trtable
[NEWLINE_CHAR
+ SBC_MAX
] = dest_states_nl
[j
];
3549 /* There must be only one destination which accepts
3550 newline. See group_nodes_into_DFAstates. */
3555 if (dest_states_malloced
)
3558 re_node_set_free (&follows
);
3559 for (i
= 0; i
< ndests
; ++i
)
3560 re_node_set_free (dests_node
+ i
);
3562 if (dests_node_malloced
)
3568 /* Group all nodes belonging to STATE into several destinations.
3569 Then for all destinations, set the nodes belonging to the destination
3570 to DESTS_NODE[i] and set the characters accepted by the destination
3571 to DEST_CH[i]. This function return the number of destinations. */
3575 group_nodes_into_DFAstates (const re_dfa_t
*dfa
, const re_dfastate_t
*state
,
3576 re_node_set
*dests_node
, bitset_t
*dests_ch
)
3581 Idx ndests
; /* Number of the destinations from `state'. */
3582 bitset_t accepts
; /* Characters a node can accept. */
3583 const re_node_set
*cur_nodes
= &state
->nodes
;
3584 bitset_empty (accepts
);
3587 /* For all the nodes belonging to `state', */
3588 for (i
= 0; i
< cur_nodes
->nelem
; ++i
)
3590 re_token_t
*node
= &dfa
->nodes
[cur_nodes
->elems
[i
]];
3591 re_token_type_t type
= node
->type
;
3592 unsigned int constraint
= node
->constraint
;
3594 /* Enumerate all single byte character this node can accept. */
3595 if (type
== CHARACTER
)
3596 bitset_set (accepts
, node
->opr
.c
);
3597 else if (type
== SIMPLE_BRACKET
)
3599 bitset_merge (accepts
, node
->opr
.sbcset
);
3601 else if (type
== OP_PERIOD
)
3603 #ifdef RE_ENABLE_I18N
3604 if (dfa
->mb_cur_max
> 1)
3605 bitset_merge (accepts
, dfa
->sb_char
);
3608 bitset_set_all (accepts
);
3609 if (!(dfa
->syntax
& RE_DOT_NEWLINE
))
3610 bitset_clear (accepts
, '\n');
3611 if (dfa
->syntax
& RE_DOT_NOT_NULL
)
3612 bitset_clear (accepts
, '\0');
3614 #ifdef RE_ENABLE_I18N
3615 else if (type
== OP_UTF8_PERIOD
)
3617 if (ASCII_CHARS
% BITSET_WORD_BITS
== 0)
3618 memset (accepts
, -1, ASCII_CHARS
/ CHAR_BIT
);
3620 bitset_merge (accepts
, utf8_sb_map
);
3621 if (!(dfa
->syntax
& RE_DOT_NEWLINE
))
3622 bitset_clear (accepts
, '\n');
3623 if (dfa
->syntax
& RE_DOT_NOT_NULL
)
3624 bitset_clear (accepts
, '\0');
3630 /* Check the `accepts' and sift the characters which are not
3631 match it the context. */
3634 if (constraint
& NEXT_NEWLINE_CONSTRAINT
)
3636 bool accepts_newline
= bitset_contain (accepts
, NEWLINE_CHAR
);
3637 bitset_empty (accepts
);
3638 if (accepts_newline
)
3639 bitset_set (accepts
, NEWLINE_CHAR
);
3643 if (constraint
& NEXT_ENDBUF_CONSTRAINT
)
3645 bitset_empty (accepts
);
3649 if (constraint
& NEXT_WORD_CONSTRAINT
)
3651 bitset_word_t any_set
= 0;
3652 if (type
== CHARACTER
&& !node
->word_char
)
3654 bitset_empty (accepts
);
3657 #ifdef RE_ENABLE_I18N
3658 if (dfa
->mb_cur_max
> 1)
3659 for (j
= 0; j
< BITSET_WORDS
; ++j
)
3660 any_set
|= (accepts
[j
] &= (dfa
->word_char
[j
] | ~dfa
->sb_char
[j
]));
3663 for (j
= 0; j
< BITSET_WORDS
; ++j
)
3664 any_set
|= (accepts
[j
] &= dfa
->word_char
[j
]);
3668 if (constraint
& NEXT_NOTWORD_CONSTRAINT
)
3670 bitset_word_t any_set
= 0;
3671 if (type
== CHARACTER
&& node
->word_char
)
3673 bitset_empty (accepts
);
3676 #ifdef RE_ENABLE_I18N
3677 if (dfa
->mb_cur_max
> 1)
3678 for (j
= 0; j
< BITSET_WORDS
; ++j
)
3679 any_set
|= (accepts
[j
] &= ~(dfa
->word_char
[j
] & dfa
->sb_char
[j
]));
3682 for (j
= 0; j
< BITSET_WORDS
; ++j
)
3683 any_set
|= (accepts
[j
] &= ~dfa
->word_char
[j
]);
3689 /* Then divide `accepts' into DFA states, or create a new
3690 state. Above, we make sure that accepts is not empty. */
3691 for (j
= 0; j
< ndests
; ++j
)
3693 bitset_t intersec
; /* Intersection sets, see below. */
3695 /* Flags, see below. */
3696 bitset_word_t has_intersec
, not_subset
, not_consumed
;
3698 /* Optimization, skip if this state doesn't accept the character. */
3699 if (type
== CHARACTER
&& !bitset_contain (dests_ch
[j
], node
->opr
.c
))
3702 /* Enumerate the intersection set of this state and `accepts'. */
3704 for (k
= 0; k
< BITSET_WORDS
; ++k
)
3705 has_intersec
|= intersec
[k
] = accepts
[k
] & dests_ch
[j
][k
];
3706 /* And skip if the intersection set is empty. */
3710 /* Then check if this state is a subset of `accepts'. */
3711 not_subset
= not_consumed
= 0;
3712 for (k
= 0; k
< BITSET_WORDS
; ++k
)
3714 not_subset
|= remains
[k
] = ~accepts
[k
] & dests_ch
[j
][k
];
3715 not_consumed
|= accepts
[k
] = accepts
[k
] & ~dests_ch
[j
][k
];
3718 /* If this state isn't a subset of `accepts', create a
3719 new group state, which has the `remains'. */
3722 bitset_copy (dests_ch
[ndests
], remains
);
3723 bitset_copy (dests_ch
[j
], intersec
);
3724 err
= re_node_set_init_copy (dests_node
+ ndests
, &dests_node
[j
]);
3725 if (BE (err
!= REG_NOERROR
, 0))
3730 /* Put the position in the current group. */
3731 ok
= re_node_set_insert (&dests_node
[j
], cur_nodes
->elems
[i
]);
3735 /* If all characters are consumed, go to next node. */
3739 /* Some characters remain, create a new group. */
3742 bitset_copy (dests_ch
[ndests
], accepts
);
3743 err
= re_node_set_init_1 (dests_node
+ ndests
, cur_nodes
->elems
[i
]);
3744 if (BE (err
!= REG_NOERROR
, 0))
3747 bitset_empty (accepts
);
3752 for (j
= 0; j
< ndests
; ++j
)
3753 re_node_set_free (dests_node
+ j
);
3757 #ifdef RE_ENABLE_I18N
3758 /* Check how many bytes the node `dfa->nodes[node_idx]' accepts.
3759 Return the number of the bytes the node accepts.
3760 STR_IDX is the current index of the input string.
3762 This function handles the nodes which can accept one character, or
3763 one collating element like '.', '[a-z]', opposite to the other nodes
3764 can only accept one byte. */
3768 check_node_accept_bytes (const re_dfa_t
*dfa
, Idx node_idx
,
3769 const re_string_t
*input
, Idx str_idx
)
3771 const re_token_t
*node
= dfa
->nodes
+ node_idx
;
3772 int char_len
, elem_len
;
3775 if (BE (node
->type
== OP_UTF8_PERIOD
, 0))
3777 unsigned char c
= re_string_byte_at (input
, str_idx
), d
;
3778 if (BE (c
< 0xc2, 1))
3781 if (str_idx
+ 2 > input
->len
)
3784 d
= re_string_byte_at (input
, str_idx
+ 1);
3786 return (d
< 0x80 || d
> 0xbf) ? 0 : 2;
3790 if (c
== 0xe0 && d
< 0xa0)
3796 if (c
== 0xf0 && d
< 0x90)
3802 if (c
== 0xf8 && d
< 0x88)
3808 if (c
== 0xfc && d
< 0x84)
3814 if (str_idx
+ char_len
> input
->len
)
3817 for (i
= 1; i
< char_len
; ++i
)
3819 d
= re_string_byte_at (input
, str_idx
+ i
);
3820 if (d
< 0x80 || d
> 0xbf)
3826 char_len
= re_string_char_size_at (input
, str_idx
);
3827 if (node
->type
== OP_PERIOD
)
3831 /* FIXME: I don't think this if is needed, as both '\n'
3832 and '\0' are char_len == 1. */
3833 /* '.' accepts any one character except the following two cases. */
3834 if ((!(dfa
->syntax
& RE_DOT_NEWLINE
) &&
3835 re_string_byte_at (input
, str_idx
) == '\n') ||
3836 ((dfa
->syntax
& RE_DOT_NOT_NULL
) &&
3837 re_string_byte_at (input
, str_idx
) == '\0'))
3842 elem_len
= re_string_elem_size_at (input
, str_idx
);
3843 if ((elem_len
<= 1 && char_len
<= 1) || char_len
== 0)
3846 if (node
->type
== COMPLEX_BRACKET
)
3848 const re_charset_t
*cset
= node
->opr
.mbcset
;
3850 const unsigned char *pin
3851 = ((const unsigned char *) re_string_get_buffer (input
) + str_idx
);
3856 wchar_t wc
= ((cset
->nranges
|| cset
->nchar_classes
|| cset
->nmbchars
)
3857 ? re_string_wchar_at (input
, str_idx
) : 0);
3859 /* match with multibyte character? */
3860 for (i
= 0; i
< cset
->nmbchars
; ++i
)
3861 if (wc
== cset
->mbchars
[i
])
3863 match_len
= char_len
;
3864 goto check_node_accept_bytes_match
;
3866 /* match with character_class? */
3867 for (i
= 0; i
< cset
->nchar_classes
; ++i
)
3869 wctype_t wt
= cset
->char_classes
[i
];
3870 if (__iswctype (wc
, wt
))
3872 match_len
= char_len
;
3873 goto check_node_accept_bytes_match
;
3878 nrules
= _NL_CURRENT_WORD (LC_COLLATE
, _NL_COLLATE_NRULES
);
3881 unsigned int in_collseq
= 0;
3882 const int32_t *table
, *indirect
;
3883 const unsigned char *weights
, *extra
;
3884 const char *collseqwc
;
3886 /* This #include defines a local function! */
3887 # include <locale/weight.h>
3889 /* match with collating_symbol? */
3890 if (cset
->ncoll_syms
)
3891 extra
= (const unsigned char *)
3892 _NL_CURRENT (LC_COLLATE
, _NL_COLLATE_SYMB_EXTRAMB
);
3893 for (i
= 0; i
< cset
->ncoll_syms
; ++i
)
3895 const unsigned char *coll_sym
= extra
+ cset
->coll_syms
[i
];
3896 /* Compare the length of input collating element and
3897 the length of current collating element. */
3898 if (*coll_sym
!= elem_len
)
3900 /* Compare each bytes. */
3901 for (j
= 0; j
< *coll_sym
; j
++)
3902 if (pin
[j
] != coll_sym
[1 + j
])
3906 /* Match if every bytes is equal. */
3908 goto check_node_accept_bytes_match
;
3914 if (elem_len
<= char_len
)
3916 collseqwc
= _NL_CURRENT (LC_COLLATE
, _NL_COLLATE_COLLSEQWC
);
3917 in_collseq
= __collseq_table_lookup (collseqwc
, wc
);
3920 in_collseq
= find_collation_sequence_value (pin
, elem_len
);
3922 /* match with range expression? */
3923 for (i
= 0; i
< cset
->nranges
; ++i
)
3924 if (cset
->range_starts
[i
] <= in_collseq
3925 && in_collseq
<= cset
->range_ends
[i
])
3927 match_len
= elem_len
;
3928 goto check_node_accept_bytes_match
;
3931 /* match with equivalence_class? */
3932 if (cset
->nequiv_classes
)
3934 const unsigned char *cp
= pin
;
3935 table
= (const int32_t *)
3936 _NL_CURRENT (LC_COLLATE
, _NL_COLLATE_TABLEMB
);
3937 weights
= (const unsigned char *)
3938 _NL_CURRENT (LC_COLLATE
, _NL_COLLATE_WEIGHTMB
);
3939 extra
= (const unsigned char *)
3940 _NL_CURRENT (LC_COLLATE
, _NL_COLLATE_EXTRAMB
);
3941 indirect
= (const int32_t *)
3942 _NL_CURRENT (LC_COLLATE
, _NL_COLLATE_INDIRECTMB
);
3943 idx
= findidx (&cp
);
3945 for (i
= 0; i
< cset
->nequiv_classes
; ++i
)
3947 int32_t equiv_class_idx
= cset
->equiv_classes
[i
];
3948 size_t weight_len
= weights
[idx
];
3949 if (weight_len
== weights
[equiv_class_idx
])
3952 while (cnt
<= weight_len
3953 && (weights
[equiv_class_idx
+ 1 + cnt
]
3954 == weights
[idx
+ 1 + cnt
]))
3956 if (cnt
> weight_len
)
3958 match_len
= elem_len
;
3959 goto check_node_accept_bytes_match
;
3968 /* match with range expression? */
3969 #if __GNUC__ >= 2 && ! (__STDC_VERSION__ < 199901L && __STRICT_ANSI__)
3970 wchar_t cmp_buf
[] = {L
'\0', L
'\0', wc
, L
'\0', L
'\0', L
'\0'};
3972 wchar_t cmp_buf
[] = {L
'\0', L
'\0', L
'\0', L
'\0', L
'\0', L
'\0'};
3975 for (i
= 0; i
< cset
->nranges
; ++i
)
3977 cmp_buf
[0] = cset
->range_starts
[i
];
3978 cmp_buf
[4] = cset
->range_ends
[i
];
3979 if (wcscoll (cmp_buf
, cmp_buf
+ 2) <= 0
3980 && wcscoll (cmp_buf
+ 2, cmp_buf
+ 4) <= 0)
3982 match_len
= char_len
;
3983 goto check_node_accept_bytes_match
;
3987 check_node_accept_bytes_match
:
3988 if (!cset
->non_match
)
3995 return (elem_len
> char_len
) ? elem_len
: char_len
;
4004 find_collation_sequence_value (const unsigned char *mbs
, size_t mbs_len
)
4006 uint32_t nrules
= _NL_CURRENT_WORD (LC_COLLATE
, _NL_COLLATE_NRULES
);
4011 /* No valid character. Match it as a single byte character. */
4012 const unsigned char *collseq
= (const unsigned char *)
4013 _NL_CURRENT (LC_COLLATE
, _NL_COLLATE_COLLSEQMB
);
4014 return collseq
[mbs
[0]];
4021 const unsigned char *extra
= (const unsigned char *)
4022 _NL_CURRENT (LC_COLLATE
, _NL_COLLATE_SYMB_EXTRAMB
);
4023 int32_t extrasize
= (const unsigned char *)
4024 _NL_CURRENT (LC_COLLATE
, _NL_COLLATE_SYMB_EXTRAMB
+ 1) - extra
;
4026 for (idx
= 0; idx
< extrasize
;)
4030 int32_t elem_mbs_len
;
4031 /* Skip the name of collating element name. */
4032 idx
= idx
+ extra
[idx
] + 1;
4033 elem_mbs_len
= extra
[idx
++];
4034 if (mbs_len
== elem_mbs_len
)
4036 for (mbs_cnt
= 0; mbs_cnt
< elem_mbs_len
; ++mbs_cnt
)
4037 if (extra
[idx
+ mbs_cnt
] != mbs
[mbs_cnt
])
4039 if (mbs_cnt
== elem_mbs_len
)
4040 /* Found the entry. */
4043 /* Skip the byte sequence of the collating element. */
4044 idx
+= elem_mbs_len
;
4045 /* Adjust for the alignment. */
4046 idx
= (idx
+ 3) & ~3;
4047 /* Skip the collation sequence value. */
4048 idx
+= sizeof (uint32_t);
4049 /* Skip the wide char sequence of the collating element. */
4050 idx
= idx
+ sizeof (uint32_t) * (extra
[idx
] + 1);
4051 /* If we found the entry, return the sequence value. */
4053 return *(uint32_t *) (extra
+ idx
);
4054 /* Skip the collation sequence value. */
4055 idx
+= sizeof (uint32_t);
4061 #endif /* RE_ENABLE_I18N */
4063 /* Check whether the node accepts the byte which is IDX-th
4064 byte of the INPUT. */
4068 check_node_accept (const re_match_context_t
*mctx
, const re_token_t
*node
,
4072 ch
= re_string_byte_at (&mctx
->input
, idx
);
4076 if (node
->opr
.c
!= ch
)
4080 case SIMPLE_BRACKET
:
4081 if (!bitset_contain (node
->opr
.sbcset
, ch
))
4085 #ifdef RE_ENABLE_I18N
4086 case OP_UTF8_PERIOD
:
4087 if (ch
>= ASCII_CHARS
)
4092 if ((ch
== '\n' && !(mctx
->dfa
->syntax
& RE_DOT_NEWLINE
))
4093 || (ch
== '\0' && (mctx
->dfa
->syntax
& RE_DOT_NOT_NULL
)))
4101 if (node
->constraint
)
4103 /* The node has constraints. Check whether the current context
4104 satisfies the constraints. */
4105 unsigned int context
= re_string_context_at (&mctx
->input
, idx
,
4107 if (NOT_SATISFY_NEXT_CONSTRAINT (node
->constraint
, context
))
4114 /* Extend the buffers, if the buffers have run out. */
4116 static reg_errcode_t
4118 extend_buffers (re_match_context_t
*mctx
)
4121 re_string_t
*pstr
= &mctx
->input
;
4123 /* Avoid overflow. */
4124 if (BE (SIZE_MAX
/ 2 / sizeof (re_dfastate_t
*) <= pstr
->bufs_len
, 0))
4127 /* Double the lengthes of the buffers. */
4128 ret
= re_string_realloc_buffers (pstr
, pstr
->bufs_len
* 2);
4129 if (BE (ret
!= REG_NOERROR
, 0))
4132 if (mctx
->state_log
!= NULL
)
4134 /* And double the length of state_log. */
4135 /* XXX We have no indication of the size of this buffer. If this
4136 allocation fail we have no indication that the state_log array
4137 does not have the right size. */
4138 re_dfastate_t
**new_array
= re_realloc (mctx
->state_log
, re_dfastate_t
*,
4139 pstr
->bufs_len
+ 1);
4140 if (BE (new_array
== NULL
, 0))
4142 mctx
->state_log
= new_array
;
4145 /* Then reconstruct the buffers. */
4148 #ifdef RE_ENABLE_I18N
4149 if (pstr
->mb_cur_max
> 1)
4151 ret
= build_wcs_upper_buffer (pstr
);
4152 if (BE (ret
!= REG_NOERROR
, 0))
4156 #endif /* RE_ENABLE_I18N */
4157 build_upper_buffer (pstr
);
4161 #ifdef RE_ENABLE_I18N
4162 if (pstr
->mb_cur_max
> 1)
4163 build_wcs_buffer (pstr
);
4165 #endif /* RE_ENABLE_I18N */
4167 if (pstr
->trans
!= NULL
)
4168 re_string_translate_buffer (pstr
);
4175 /* Functions for matching context. */
4177 /* Initialize MCTX. */
4179 static reg_errcode_t
4181 match_ctx_init (re_match_context_t
*mctx
, int eflags
, Idx n
)
4183 mctx
->eflags
= eflags
;
4184 mctx
->match_last
= REG_MISSING
;
4187 /* Avoid overflow. */
4188 size_t max_object_size
=
4189 MAX (sizeof (struct re_backref_cache_entry
),
4190 sizeof (re_sub_match_top_t
*));
4191 if (BE (SIZE_MAX
/ max_object_size
< n
, 0))
4194 mctx
->bkref_ents
= re_malloc (struct re_backref_cache_entry
, n
);
4195 mctx
->sub_tops
= re_malloc (re_sub_match_top_t
*, n
);
4196 if (BE (mctx
->bkref_ents
== NULL
|| mctx
->sub_tops
== NULL
, 0))
4199 /* Already zero-ed by the caller.
4201 mctx->bkref_ents = NULL;
4202 mctx->nbkref_ents = 0;
4203 mctx->nsub_tops = 0; */
4204 mctx
->abkref_ents
= n
;
4205 mctx
->max_mb_elem_len
= 1;
4206 mctx
->asub_tops
= n
;
4210 /* Clean the entries which depend on the current input in MCTX.
4211 This function must be invoked when the matcher changes the start index
4212 of the input, or changes the input string. */
4216 match_ctx_clean (re_match_context_t
*mctx
)
4219 for (st_idx
= 0; st_idx
< mctx
->nsub_tops
; ++st_idx
)
4222 re_sub_match_top_t
*top
= mctx
->sub_tops
[st_idx
];
4223 for (sl_idx
= 0; sl_idx
< top
->nlasts
; ++sl_idx
)
4225 re_sub_match_last_t
*last
= top
->lasts
[sl_idx
];
4226 re_free (last
->path
.array
);
4229 re_free (top
->lasts
);
4232 re_free (top
->path
->array
);
4233 re_free (top
->path
);
4238 mctx
->nsub_tops
= 0;
4239 mctx
->nbkref_ents
= 0;
4242 /* Free all the memory associated with MCTX. */
4246 match_ctx_free (re_match_context_t
*mctx
)
4248 /* First, free all the memory associated with MCTX->SUB_TOPS. */
4249 match_ctx_clean (mctx
);
4250 re_free (mctx
->sub_tops
);
4251 re_free (mctx
->bkref_ents
);
4254 /* Add a new backreference entry to MCTX.
4255 Note that we assume that caller never call this function with duplicate
4256 entry, and call with STR_IDX which isn't smaller than any existing entry.
4259 static reg_errcode_t
4261 match_ctx_add_entry (re_match_context_t
*mctx
, Idx node
, Idx str_idx
, Idx from
,
4264 if (mctx
->nbkref_ents
>= mctx
->abkref_ents
)
4266 struct re_backref_cache_entry
* new_entry
;
4267 new_entry
= re_realloc (mctx
->bkref_ents
, struct re_backref_cache_entry
,
4268 mctx
->abkref_ents
* 2);
4269 if (BE (new_entry
== NULL
, 0))
4271 re_free (mctx
->bkref_ents
);
4274 mctx
->bkref_ents
= new_entry
;
4275 memset (mctx
->bkref_ents
+ mctx
->nbkref_ents
, '\0',
4276 sizeof (struct re_backref_cache_entry
) * mctx
->abkref_ents
);
4277 mctx
->abkref_ents
*= 2;
4279 if (mctx
->nbkref_ents
> 0
4280 && mctx
->bkref_ents
[mctx
->nbkref_ents
- 1].str_idx
== str_idx
)
4281 mctx
->bkref_ents
[mctx
->nbkref_ents
- 1].more
= 1;
4283 mctx
->bkref_ents
[mctx
->nbkref_ents
].node
= node
;
4284 mctx
->bkref_ents
[mctx
->nbkref_ents
].str_idx
= str_idx
;
4285 mctx
->bkref_ents
[mctx
->nbkref_ents
].subexp_from
= from
;
4286 mctx
->bkref_ents
[mctx
->nbkref_ents
].subexp_to
= to
;
4288 /* This is a cache that saves negative results of check_dst_limits_calc_pos.
4289 If bit N is clear, means that this entry won't epsilon-transition to
4290 an OP_OPEN_SUBEXP or OP_CLOSE_SUBEXP for the N+1-th subexpression. If
4291 it is set, check_dst_limits_calc_pos_1 will recurse and try to find one
4294 A backreference does not epsilon-transition unless it is empty, so set
4295 to all zeros if FROM != TO. */
4296 mctx
->bkref_ents
[mctx
->nbkref_ents
].eps_reachable_subexps_map
4297 = (from
== to
? -1 : 0);
4299 mctx
->bkref_ents
[mctx
->nbkref_ents
++].more
= 0;
4300 if (mctx
->max_mb_elem_len
< to
- from
)
4301 mctx
->max_mb_elem_len
= to
- from
;
4305 /* Return the first entry with the same str_idx, or REG_MISSING if none is
4306 found. Note that MCTX->BKREF_ENTS is already sorted by MCTX->STR_IDX. */
4310 search_cur_bkref_entry (const re_match_context_t
*mctx
, Idx str_idx
)
4312 Idx left
, right
, mid
, last
;
4313 last
= right
= mctx
->nbkref_ents
;
4314 for (left
= 0; left
< right
;)
4316 mid
= (left
+ right
) / 2;
4317 if (mctx
->bkref_ents
[mid
].str_idx
< str_idx
)
4322 if (left
< last
&& mctx
->bkref_ents
[left
].str_idx
== str_idx
)
4328 /* Register the node NODE, whose type is OP_OPEN_SUBEXP, and which matches
4331 static reg_errcode_t
4333 match_ctx_add_subtop (re_match_context_t
*mctx
, Idx node
, Idx str_idx
)
4336 assert (mctx
->sub_tops
!= NULL
);
4337 assert (mctx
->asub_tops
> 0);
4339 if (BE (mctx
->nsub_tops
== mctx
->asub_tops
, 0))
4341 Idx new_asub_tops
= mctx
->asub_tops
* 2;
4342 re_sub_match_top_t
**new_array
= re_realloc (mctx
->sub_tops
,
4343 re_sub_match_top_t
*,
4345 if (BE (new_array
== NULL
, 0))
4347 mctx
->sub_tops
= new_array
;
4348 mctx
->asub_tops
= new_asub_tops
;
4350 mctx
->sub_tops
[mctx
->nsub_tops
] = calloc (1, sizeof (re_sub_match_top_t
));
4351 if (BE (mctx
->sub_tops
[mctx
->nsub_tops
] == NULL
, 0))
4353 mctx
->sub_tops
[mctx
->nsub_tops
]->node
= node
;
4354 mctx
->sub_tops
[mctx
->nsub_tops
++]->str_idx
= str_idx
;
4358 /* Register the node NODE, whose type is OP_CLOSE_SUBEXP, and which matches
4359 at STR_IDX, whose corresponding OP_OPEN_SUBEXP is SUB_TOP. */
4361 static re_sub_match_last_t
*
4363 match_ctx_add_sublast (re_sub_match_top_t
*subtop
, Idx node
, Idx str_idx
)
4365 re_sub_match_last_t
*new_entry
;
4366 if (BE (subtop
->nlasts
== subtop
->alasts
, 0))
4368 Idx new_alasts
= 2 * subtop
->alasts
+ 1;
4369 re_sub_match_last_t
**new_array
= re_realloc (subtop
->lasts
,
4370 re_sub_match_last_t
*,
4372 if (BE (new_array
== NULL
, 0))
4374 subtop
->lasts
= new_array
;
4375 subtop
->alasts
= new_alasts
;
4377 new_entry
= calloc (1, sizeof (re_sub_match_last_t
));
4378 if (BE (new_entry
!= NULL
, 1))
4380 subtop
->lasts
[subtop
->nlasts
] = new_entry
;
4381 new_entry
->node
= node
;
4382 new_entry
->str_idx
= str_idx
;
4390 sift_ctx_init (re_sift_context_t
*sctx
, re_dfastate_t
**sifted_sts
,
4391 re_dfastate_t
**limited_sts
, Idx last_node
, Idx last_str_idx
)
4393 sctx
->sifted_states
= sifted_sts
;
4394 sctx
->limited_states
= limited_sts
;
4395 sctx
->last_node
= last_node
;
4396 sctx
->last_str_idx
= last_str_idx
;
4397 re_node_set_init_empty (&sctx
->limits
);