1 /* String search routines for GNU Emacs.
3 Copyright (C) 1985-1987, 1993-1994, 1997-1999, 2001-2014 Free Software
6 This file is part of GNU Emacs.
8 GNU Emacs is free software: you can redistribute it and/or modify
9 it under the terms of the GNU General Public License as published by
10 the Free Software Foundation, either version 3 of the License, or
11 (at your option) any later version.
13 GNU Emacs is distributed in the hope that it will be useful,
14 but WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 GNU General Public License for more details.
18 You should have received a copy of the GNU General Public License
19 along with GNU Emacs. If not, see <http://www.gnu.org/licenses/>. */
26 #include "character.h"
30 #include "region-cache.h"
32 #include "blockinput.h"
33 #include "intervals.h"
35 #include <sys/types.h>
38 #define REGEXP_CACHE_SIZE 20
40 /* If the regexp is non-nil, then the buffer contains the compiled form
41 of that regexp, suitable for searching. */
44 struct regexp_cache
*next
;
45 Lisp_Object regexp
, whitespace_regexp
;
46 /* Syntax table for which the regexp applies. We need this because
47 of character classes. If this is t, then the compiled pattern is valid
48 for any syntax-table. */
49 Lisp_Object syntax_table
;
50 struct re_pattern_buffer buf
;
52 /* True means regexp was compiled to do full POSIX backtracking. */
56 /* The instances of that struct. */
57 static struct regexp_cache searchbufs
[REGEXP_CACHE_SIZE
];
59 /* The head of the linked list; points to the most recently used buffer. */
60 static struct regexp_cache
*searchbuf_head
;
63 /* Every call to re_match, etc., must pass &search_regs as the regs
64 argument unless you can show it is unnecessary (i.e., if re_match
65 is certainly going to be called again before region-around-match
68 Since the registers are now dynamically allocated, we need to make
69 sure not to refer to the Nth register before checking that it has
70 been allocated by checking search_regs.num_regs.
72 The regex code keeps track of whether it has allocated the search
73 buffer using bits in the re_pattern_buffer. This means that whenever
74 you compile a new pattern, it completely forgets whether it has
75 allocated any registers, and will allocate new registers the next
76 time you call a searching or matching function. Therefore, we need
77 to call re_set_registers after compiling a new pattern or after
78 setting the match registers, so that the regex functions will be
79 able to free or re-allocate it properly. */
80 static struct re_registers search_regs
;
82 /* The buffer in which the last search was performed, or
83 Qt if the last search was done in a string;
84 Qnil if no searching has been done yet. */
85 static Lisp_Object last_thing_searched
;
87 /* Error condition signaled when regexp compile_pattern fails. */
88 static Lisp_Object Qinvalid_regexp
;
90 /* Error condition used for failing searches. */
91 static Lisp_Object Qsearch_failed
;
93 static void set_search_regs (ptrdiff_t, ptrdiff_t);
94 static void save_search_regs (void);
95 static EMACS_INT
simple_search (EMACS_INT
, unsigned char *, ptrdiff_t,
96 ptrdiff_t, Lisp_Object
, ptrdiff_t, ptrdiff_t,
97 ptrdiff_t, ptrdiff_t);
98 static EMACS_INT
boyer_moore (EMACS_INT
, unsigned char *, ptrdiff_t,
99 Lisp_Object
, Lisp_Object
, ptrdiff_t,
101 static EMACS_INT
search_buffer (Lisp_Object
, ptrdiff_t, ptrdiff_t,
102 ptrdiff_t, ptrdiff_t, EMACS_INT
, int,
103 Lisp_Object
, Lisp_Object
, bool);
105 static _Noreturn
void
106 matcher_overflow (void)
108 error ("Stack overflow in regexp matcher");
111 /* Compile a regexp and signal a Lisp error if anything goes wrong.
112 PATTERN is the pattern to compile.
113 CP is the place to put the result.
114 TRANSLATE is a translation table for ignoring case, or nil for none.
115 POSIX is true if we want full backtracking (POSIX style) for this pattern.
116 False means backtrack only enough to get a valid match.
118 The behavior also depends on Vsearch_spaces_regexp. */
121 compile_pattern_1 (struct regexp_cache
*cp
, Lisp_Object pattern
,
122 Lisp_Object translate
, bool posix
)
128 cp
->buf
.translate
= (! NILP (translate
) ? translate
: make_number (0));
130 cp
->buf
.multibyte
= STRING_MULTIBYTE (pattern
);
131 cp
->buf
.charset_unibyte
= charset_unibyte
;
132 if (STRINGP (Vsearch_spaces_regexp
))
133 cp
->whitespace_regexp
= Vsearch_spaces_regexp
;
135 cp
->whitespace_regexp
= Qnil
;
137 /* rms: I think BLOCK_INPUT is not needed here any more,
138 because regex.c defines malloc to call xmalloc.
139 Using BLOCK_INPUT here means the debugger won't run if an error occurs.
140 So let's turn it off. */
142 old
= re_set_syntax (RE_SYNTAX_EMACS
143 | (posix
? 0 : RE_NO_POSIX_BACKTRACKING
));
145 if (STRINGP (Vsearch_spaces_regexp
))
146 re_set_whitespace_regexp (SSDATA (Vsearch_spaces_regexp
));
148 re_set_whitespace_regexp (NULL
);
150 val
= (char *) re_compile_pattern (SSDATA (pattern
),
151 SBYTES (pattern
), &cp
->buf
);
153 /* If the compiled pattern hard codes some of the contents of the
154 syntax-table, it can only be reused with *this* syntax table. */
155 cp
->syntax_table
= cp
->buf
.used_syntax
? BVAR (current_buffer
, syntax_table
) : Qt
;
157 re_set_whitespace_regexp (NULL
);
160 /* unblock_input (); */
162 xsignal1 (Qinvalid_regexp
, build_string (val
));
164 cp
->regexp
= Fcopy_sequence (pattern
);
167 /* Shrink each compiled regexp buffer in the cache
168 to the size actually used right now.
169 This is called from garbage collection. */
172 shrink_regexp_cache (void)
174 struct regexp_cache
*cp
;
176 for (cp
= searchbuf_head
; cp
!= 0; cp
= cp
->next
)
178 cp
->buf
.allocated
= cp
->buf
.used
;
179 cp
->buf
.buffer
= xrealloc (cp
->buf
.buffer
, cp
->buf
.used
);
183 /* Clear the regexp cache w.r.t. a particular syntax table,
184 because it was changed.
185 There is no danger of memory leak here because re_compile_pattern
186 automagically manages the memory in each re_pattern_buffer struct,
187 based on its `allocated' and `buffer' values. */
189 clear_regexp_cache (void)
193 for (i
= 0; i
< REGEXP_CACHE_SIZE
; ++i
)
194 /* It's tempting to compare with the syntax-table we've actually changed,
195 but it's not sufficient because char-table inheritance means that
196 modifying one syntax-table can change others at the same time. */
197 if (!EQ (searchbufs
[i
].syntax_table
, Qt
))
198 searchbufs
[i
].regexp
= Qnil
;
201 /* Compile a regexp if necessary, but first check to see if there's one in
203 PATTERN is the pattern to compile.
204 TRANSLATE is a translation table for ignoring case, or nil for none.
205 REGP is the structure that says where to store the "register"
206 values that will result from matching this pattern.
207 If it is 0, we should compile the pattern not to record any
208 subexpression bounds.
209 POSIX is true if we want full backtracking (POSIX style) for this pattern.
210 False means backtrack only enough to get a valid match. */
212 struct re_pattern_buffer
*
213 compile_pattern (Lisp_Object pattern
, struct re_registers
*regp
,
214 Lisp_Object translate
, bool posix
, bool multibyte
)
216 struct regexp_cache
*cp
, **cpp
;
218 for (cpp
= &searchbuf_head
; ; cpp
= &cp
->next
)
221 /* Entries are initialized to nil, and may be set to nil by
222 compile_pattern_1 if the pattern isn't valid. Don't apply
223 string accessors in those cases. However, compile_pattern_1
224 is only applied to the cache entry we pick here to reuse. So
225 nil should never appear before a non-nil entry. */
226 if (NILP (cp
->regexp
))
228 if (SCHARS (cp
->regexp
) == SCHARS (pattern
)
229 && STRING_MULTIBYTE (cp
->regexp
) == STRING_MULTIBYTE (pattern
)
230 && !NILP (Fstring_equal (cp
->regexp
, pattern
))
231 && EQ (cp
->buf
.translate
, (! NILP (translate
) ? translate
: make_number (0)))
232 && cp
->posix
== posix
233 && (EQ (cp
->syntax_table
, Qt
)
234 || EQ (cp
->syntax_table
, BVAR (current_buffer
, syntax_table
)))
235 && !NILP (Fequal (cp
->whitespace_regexp
, Vsearch_spaces_regexp
))
236 && cp
->buf
.charset_unibyte
== charset_unibyte
)
239 /* If we're at the end of the cache, compile into the nil cell
240 we found, or the last (least recently used) cell with a
245 compile_pattern_1 (cp
, pattern
, translate
, posix
);
250 /* When we get here, cp (aka *cpp) contains the compiled pattern,
251 either because we found it in the cache or because we just compiled it.
252 Move it to the front of the queue to mark it as most recently used. */
254 cp
->next
= searchbuf_head
;
257 /* Advise the searching functions about the space we have allocated
258 for register data. */
260 re_set_registers (&cp
->buf
, regp
, regp
->num_regs
, regp
->start
, regp
->end
);
262 /* The compiled pattern can be used both for multibyte and unibyte
263 target. But, we have to tell which the pattern is used for. */
264 cp
->buf
.target_multibyte
= multibyte
;
271 looking_at_1 (Lisp_Object string
, bool posix
)
274 unsigned char *p1
, *p2
;
276 register ptrdiff_t i
;
277 struct re_pattern_buffer
*bufp
;
279 if (running_asynch_code
)
282 /* This is so set_image_of_range_1 in regex.c can find the EQV table. */
283 set_char_table_extras (BVAR (current_buffer
, case_canon_table
), 2,
284 BVAR (current_buffer
, case_eqv_table
));
286 CHECK_STRING (string
);
287 bufp
= compile_pattern (string
,
288 (NILP (Vinhibit_changing_match_data
)
289 ? &search_regs
: NULL
),
290 (!NILP (BVAR (current_buffer
, case_fold_search
))
291 ? BVAR (current_buffer
, case_canon_table
) : Qnil
),
293 !NILP (BVAR (current_buffer
, enable_multibyte_characters
)));
296 QUIT
; /* Do a pending quit right away, to avoid paradoxical behavior */
298 /* Get pointers and sizes of the two strings
299 that make up the visible portion of the buffer. */
302 s1
= GPT_BYTE
- BEGV_BYTE
;
304 s2
= ZV_BYTE
- GPT_BYTE
;
308 s2
= ZV_BYTE
- BEGV_BYTE
;
313 s1
= ZV_BYTE
- BEGV_BYTE
;
317 re_match_object
= Qnil
;
319 i
= re_match_2 (bufp
, (char *) p1
, s1
, (char *) p2
, s2
,
321 (NILP (Vinhibit_changing_match_data
)
322 ? &search_regs
: NULL
),
323 ZV_BYTE
- BEGV_BYTE
);
329 val
= (i
>= 0 ? Qt
: Qnil
);
330 if (NILP (Vinhibit_changing_match_data
) && i
>= 0)
332 for (i
= 0; i
< search_regs
.num_regs
; i
++)
333 if (search_regs
.start
[i
] >= 0)
336 = BYTE_TO_CHAR (search_regs
.start
[i
] + BEGV_BYTE
);
338 = BYTE_TO_CHAR (search_regs
.end
[i
] + BEGV_BYTE
);
340 /* Set last_thing_searched only when match data is changed. */
341 XSETBUFFER (last_thing_searched
, current_buffer
);
347 DEFUN ("looking-at", Flooking_at
, Slooking_at
, 1, 1, 0,
348 doc
: /* Return t if text after point matches regular expression REGEXP.
349 This function modifies the match data that `match-beginning',
350 `match-end' and `match-data' access; save and restore the match
351 data if you want to preserve them. */)
354 return looking_at_1 (regexp
, 0);
357 DEFUN ("posix-looking-at", Fposix_looking_at
, Sposix_looking_at
, 1, 1, 0,
358 doc
: /* Return t if text after point matches regular expression REGEXP.
359 Find the longest match, in accord with Posix regular expression rules.
360 This function modifies the match data that `match-beginning',
361 `match-end' and `match-data' access; save and restore the match
362 data if you want to preserve them. */)
365 return looking_at_1 (regexp
, 1);
369 string_match_1 (Lisp_Object regexp
, Lisp_Object string
, Lisp_Object start
,
373 struct re_pattern_buffer
*bufp
;
375 ptrdiff_t pos_byte
, i
;
377 if (running_asynch_code
)
380 CHECK_STRING (regexp
);
381 CHECK_STRING (string
);
384 pos
= 0, pos_byte
= 0;
387 ptrdiff_t len
= SCHARS (string
);
389 CHECK_NUMBER (start
);
391 if (pos
< 0 && -pos
<= len
)
393 else if (0 > pos
|| pos
> len
)
394 args_out_of_range (string
, start
);
395 pos_byte
= string_char_to_byte (string
, pos
);
398 /* This is so set_image_of_range_1 in regex.c can find the EQV table. */
399 set_char_table_extras (BVAR (current_buffer
, case_canon_table
), 2,
400 BVAR (current_buffer
, case_eqv_table
));
402 bufp
= compile_pattern (regexp
,
403 (NILP (Vinhibit_changing_match_data
)
404 ? &search_regs
: NULL
),
405 (!NILP (BVAR (current_buffer
, case_fold_search
))
406 ? BVAR (current_buffer
, case_canon_table
) : Qnil
),
408 STRING_MULTIBYTE (string
));
410 re_match_object
= string
;
412 val
= re_search (bufp
, SSDATA (string
),
413 SBYTES (string
), pos_byte
,
414 SBYTES (string
) - pos_byte
,
415 (NILP (Vinhibit_changing_match_data
)
416 ? &search_regs
: NULL
));
419 /* Set last_thing_searched only when match data is changed. */
420 if (NILP (Vinhibit_changing_match_data
))
421 last_thing_searched
= Qt
;
425 if (val
< 0) return Qnil
;
427 if (NILP (Vinhibit_changing_match_data
))
428 for (i
= 0; i
< search_regs
.num_regs
; i
++)
429 if (search_regs
.start
[i
] >= 0)
432 = string_byte_to_char (string
, search_regs
.start
[i
]);
434 = string_byte_to_char (string
, search_regs
.end
[i
]);
437 return make_number (string_byte_to_char (string
, val
));
440 DEFUN ("string-match", Fstring_match
, Sstring_match
, 2, 3, 0,
441 doc
: /* Return index of start of first match for REGEXP in STRING, or nil.
442 Matching ignores case if `case-fold-search' is non-nil.
443 If third arg START is non-nil, start search at that index in STRING.
444 For index of first char beyond the match, do (match-end 0).
445 `match-end' and `match-beginning' also give indices of substrings
446 matched by parenthesis constructs in the pattern.
448 You can use the function `match-string' to extract the substrings
449 matched by the parenthesis constructions in REGEXP. */)
450 (Lisp_Object regexp
, Lisp_Object string
, Lisp_Object start
)
452 return string_match_1 (regexp
, string
, start
, 0);
455 DEFUN ("posix-string-match", Fposix_string_match
, Sposix_string_match
, 2, 3, 0,
456 doc
: /* Return index of start of first match for REGEXP in STRING, or nil.
457 Find the longest match, in accord with Posix regular expression rules.
458 Case is ignored if `case-fold-search' is non-nil in the current buffer.
459 If third arg START is non-nil, start search at that index in STRING.
460 For index of first char beyond the match, do (match-end 0).
461 `match-end' and `match-beginning' also give indices of substrings
462 matched by parenthesis constructs in the pattern. */)
463 (Lisp_Object regexp
, Lisp_Object string
, Lisp_Object start
)
465 return string_match_1 (regexp
, string
, start
, 1);
468 /* Match REGEXP against STRING, searching all of STRING,
469 and return the index of the match, or negative on failure.
470 This does not clobber the match data. */
473 fast_string_match (Lisp_Object regexp
, Lisp_Object string
)
476 struct re_pattern_buffer
*bufp
;
478 bufp
= compile_pattern (regexp
, 0, Qnil
,
479 0, STRING_MULTIBYTE (string
));
481 re_match_object
= string
;
483 val
= re_search (bufp
, SSDATA (string
),
490 /* Match REGEXP against STRING, searching all of STRING ignoring case,
491 and return the index of the match, or negative on failure.
492 This does not clobber the match data.
493 We assume that STRING contains single-byte characters. */
496 fast_c_string_match_ignore_case (Lisp_Object regexp
,
497 const char *string
, ptrdiff_t len
)
500 struct re_pattern_buffer
*bufp
;
502 regexp
= string_make_unibyte (regexp
);
503 re_match_object
= Qt
;
504 bufp
= compile_pattern (regexp
, 0,
505 Vascii_canon_table
, 0,
508 val
= re_search (bufp
, string
, len
, 0, len
, 0);
513 /* Like fast_string_match but ignore case. */
516 fast_string_match_ignore_case (Lisp_Object regexp
, Lisp_Object string
)
519 struct re_pattern_buffer
*bufp
;
521 bufp
= compile_pattern (regexp
, 0, Vascii_canon_table
,
522 0, STRING_MULTIBYTE (string
));
524 re_match_object
= string
;
526 val
= re_search (bufp
, SSDATA (string
),
533 /* Match REGEXP against the characters after POS to LIMIT, and return
534 the number of matched characters. If STRING is non-nil, match
535 against the characters in it. In that case, POS and LIMIT are
536 indices into the string. This function doesn't modify the match
540 fast_looking_at (Lisp_Object regexp
, ptrdiff_t pos
, ptrdiff_t pos_byte
,
541 ptrdiff_t limit
, ptrdiff_t limit_byte
, Lisp_Object string
)
544 struct re_pattern_buffer
*buf
;
545 unsigned char *p1
, *p2
;
549 if (STRINGP (string
))
552 pos_byte
= string_char_to_byte (string
, pos
);
554 limit_byte
= string_char_to_byte (string
, limit
);
558 s2
= SBYTES (string
);
559 re_match_object
= string
;
560 multibyte
= STRING_MULTIBYTE (string
);
565 pos_byte
= CHAR_TO_BYTE (pos
);
567 limit_byte
= CHAR_TO_BYTE (limit
);
568 pos_byte
-= BEGV_BYTE
;
569 limit_byte
-= BEGV_BYTE
;
571 s1
= GPT_BYTE
- BEGV_BYTE
;
573 s2
= ZV_BYTE
- GPT_BYTE
;
577 s2
= ZV_BYTE
- BEGV_BYTE
;
582 s1
= ZV_BYTE
- BEGV_BYTE
;
585 re_match_object
= Qnil
;
586 multibyte
= ! NILP (BVAR (current_buffer
, enable_multibyte_characters
));
589 buf
= compile_pattern (regexp
, 0, Qnil
, 0, multibyte
);
591 len
= re_match_2 (buf
, (char *) p1
, s1
, (char *) p2
, s2
,
592 pos_byte
, NULL
, limit_byte
);
599 /* The newline cache: remembering which sections of text have no newlines. */
601 /* If the user has requested the long scans caching, make sure it's on.
602 Otherwise, make sure it's off.
603 This is our cheezy way of associating an action with the change of
604 state of a buffer-local variable. */
605 static struct region_cache
*
606 newline_cache_on_off (struct buffer
*buf
)
608 struct buffer
*base_buf
= buf
;
609 bool indirect_p
= false;
611 if (buf
->base_buffer
)
613 base_buf
= buf
->base_buffer
;
617 /* Don't turn on or off the cache in the base buffer, if the value
618 of cache-long-scans of the base buffer is inconsistent with that.
619 This is because doing so will just make the cache pure overhead,
620 since if we turn it on via indirect buffer, it will be
621 immediately turned off by its base buffer. */
622 if (NILP (BVAR (buf
, cache_long_scans
)))
625 || NILP (BVAR (base_buf
, cache_long_scans
)))
627 /* It should be off. */
628 if (base_buf
->newline_cache
)
630 free_region_cache (base_buf
->newline_cache
);
631 base_buf
->newline_cache
= 0;
639 || !NILP (BVAR (base_buf
, cache_long_scans
)))
641 /* It should be on. */
642 if (base_buf
->newline_cache
== 0)
643 base_buf
->newline_cache
= new_region_cache ();
645 return base_buf
->newline_cache
;
650 /* Search for COUNT newlines between START/START_BYTE and END/END_BYTE.
652 If COUNT is positive, search forwards; END must be >= START.
653 If COUNT is negative, search backwards for the -COUNTth instance;
654 END must be <= START.
655 If COUNT is zero, do anything you please; run rogue, for all I care.
657 If END is zero, use BEGV or ZV instead, as appropriate for the
658 direction indicated by COUNT.
660 If we find COUNT instances, set *SHORTAGE to zero, and return the
661 position past the COUNTth match. Note that for reverse motion
662 this is not the same as the usual convention for Emacs motion commands.
664 If we don't find COUNT instances before reaching END, set *SHORTAGE
665 to the number of newlines left unfound, and return END.
667 If BYTEPOS is not NULL, set *BYTEPOS to the byte position corresponding
668 to the returned character position.
670 If ALLOW_QUIT, set immediate_quit. That's good to do
671 except when inside redisplay. */
674 find_newline (ptrdiff_t start
, ptrdiff_t start_byte
, ptrdiff_t end
,
675 ptrdiff_t end_byte
, ptrdiff_t count
, ptrdiff_t *shortage
,
676 ptrdiff_t *bytepos
, bool allow_quit
)
678 struct region_cache
*newline_cache
;
680 struct buffer
*cache_buffer
;
686 end
= ZV
, end_byte
= ZV_BYTE
;
692 end
= BEGV
, end_byte
= BEGV_BYTE
;
695 end_byte
= CHAR_TO_BYTE (end
);
697 newline_cache
= newline_cache_on_off (current_buffer
);
698 if (current_buffer
->base_buffer
)
699 cache_buffer
= current_buffer
->base_buffer
;
701 cache_buffer
= current_buffer
;
706 immediate_quit
= allow_quit
;
711 /* Our innermost scanning loop is very simple; it doesn't know
712 about gaps, buffer ends, or the newline cache. ceiling is
713 the position of the last character before the next such
714 obstacle --- the last character the dumb search loop should
716 ptrdiff_t tem
, ceiling_byte
= end_byte
- 1;
718 /* If we're using the newline cache, consult it to see whether
719 we can avoid some scanning. */
722 ptrdiff_t next_change
;
726 while (start
< end
&& result
)
730 result
= region_cache_forward (cache_buffer
, newline_cache
,
731 start
, &next_change
);
735 lim1
= next_change
= end
;
738 lim1
= min (next_change
, end
);
740 /* The cache returned zero for this region; see if
741 this is because the region is known and includes
742 only newlines. While at that, count any newlines
743 we bump into, and exit if we found enough off them. */
744 start_byte
= CHAR_TO_BYTE (start
);
746 && FETCH_BYTE (start_byte
) == '\n')
753 *bytepos
= start_byte
;
757 /* If we found a non-newline character before hitting
758 position where the cache will again return non-zero
759 (i.e. no newlines beyond that position), it means
760 this region is not yet known to the cache, and we
761 must resort to the "dumb loop" method. */
762 if (start
< next_change
&& !result
)
769 start_byte
= end_byte
;
772 immediate_quit
= allow_quit
;
774 /* START should never be after END. */
775 if (start_byte
> ceiling_byte
)
776 start_byte
= ceiling_byte
;
778 /* Now the text after start is an unknown region, and
779 next_change is the position of the next known region. */
780 ceiling_byte
= min (CHAR_TO_BYTE (next_change
) - 1, ceiling_byte
);
782 else if (start_byte
== -1)
783 start_byte
= CHAR_TO_BYTE (start
);
785 /* The dumb loop can only scan text stored in contiguous
786 bytes. BUFFER_CEILING_OF returns the last character
787 position that is contiguous, so the ceiling is the
788 position after that. */
789 tem
= BUFFER_CEILING_OF (start_byte
);
790 ceiling_byte
= min (tem
, ceiling_byte
);
793 /* The termination address of the dumb loop. */
794 unsigned char *lim_addr
= BYTE_POS_ADDR (ceiling_byte
) + 1;
795 ptrdiff_t lim_byte
= ceiling_byte
+ 1;
797 /* Nonpositive offsets (relative to LIM_ADDR and LIM_BYTE)
798 of the base, the cursor, and the next line. */
799 ptrdiff_t base
= start_byte
- lim_byte
;
800 ptrdiff_t cursor
, next
;
802 for (cursor
= base
; cursor
< 0; cursor
= next
)
805 unsigned char *nl
= memchr (lim_addr
+ cursor
, '\n', - cursor
);
806 next
= nl
? nl
- lim_addr
: 0;
808 /* If we're using the newline cache, cache the fact that
809 the region we just traversed is free of newlines. */
810 if (newline_cache
&& cursor
!= next
)
812 know_region_cache (cache_buffer
, newline_cache
,
813 BYTE_TO_CHAR (lim_byte
+ cursor
),
814 BYTE_TO_CHAR (lim_byte
+ next
));
815 /* know_region_cache can relocate buffer text. */
816 lim_addr
= BYTE_POS_ADDR (ceiling_byte
) + 1;
827 *bytepos
= lim_byte
+ next
;
828 return BYTE_TO_CHAR (lim_byte
+ next
);
832 start_byte
= lim_byte
;
833 start
= BYTE_TO_CHAR (start_byte
);
839 /* The last character to check before the next obstacle. */
840 ptrdiff_t tem
, ceiling_byte
= end_byte
;
842 /* Consult the newline cache, if appropriate. */
845 ptrdiff_t next_change
;
849 while (start
> end
&& result
)
853 result
= region_cache_backward (cache_buffer
, newline_cache
,
854 start
, &next_change
);
858 lim1
= next_change
= end
;
861 lim1
= max (next_change
, end
);
862 start_byte
= CHAR_TO_BYTE (start
);
864 && FETCH_BYTE (start_byte
- 1) == '\n')
869 *bytepos
= start_byte
;
875 if (start
> next_change
&& !result
)
882 start_byte
= end_byte
;
885 immediate_quit
= allow_quit
;
887 /* Start should never be at or before end. */
888 if (start_byte
<= ceiling_byte
)
889 start_byte
= ceiling_byte
+ 1;
891 /* Now the text before start is an unknown region, and
892 next_change is the position of the next known region. */
893 ceiling_byte
= max (CHAR_TO_BYTE (next_change
), ceiling_byte
);
895 else if (start_byte
== -1)
896 start_byte
= CHAR_TO_BYTE (start
);
898 /* Stop scanning before the gap. */
899 tem
= BUFFER_FLOOR_OF (start_byte
- 1);
900 ceiling_byte
= max (tem
, ceiling_byte
);
903 /* The termination address of the dumb loop. */
904 unsigned char *ceiling_addr
= BYTE_POS_ADDR (ceiling_byte
);
906 /* Offsets (relative to CEILING_ADDR and CEILING_BYTE) of
907 the base, the cursor, and the previous line. These
908 offsets are at least -1. */
909 ptrdiff_t base
= start_byte
- ceiling_byte
;
910 ptrdiff_t cursor
, prev
;
912 for (cursor
= base
; 0 < cursor
; cursor
= prev
)
914 unsigned char *nl
= memrchr (ceiling_addr
, '\n', cursor
);
915 prev
= nl
? nl
- ceiling_addr
: -1;
917 /* If we're looking for newlines, cache the fact that
918 this line's region is free of them. */
919 if (newline_cache
&& cursor
!= prev
+ 1)
921 know_region_cache (cache_buffer
, newline_cache
,
922 BYTE_TO_CHAR (ceiling_byte
+ prev
+ 1),
923 BYTE_TO_CHAR (ceiling_byte
+ cursor
));
924 /* know_region_cache can relocate buffer text. */
925 ceiling_addr
= BYTE_POS_ADDR (ceiling_byte
);
935 *bytepos
= ceiling_byte
+ prev
+ 1;
936 return BYTE_TO_CHAR (ceiling_byte
+ prev
+ 1);
940 start_byte
= ceiling_byte
;
941 start
= BYTE_TO_CHAR (start_byte
);
947 *shortage
= count
* direction
;
950 *bytepos
= start_byte
== -1 ? CHAR_TO_BYTE (start
) : start_byte
;
951 eassert (*bytepos
== CHAR_TO_BYTE (start
));
956 /* Search for COUNT instances of a line boundary.
957 Start at START. If COUNT is negative, search backwards.
959 We report the resulting position by calling TEMP_SET_PT_BOTH.
961 If we find COUNT instances. we position after (always after,
962 even if scanning backwards) the COUNTth match, and return 0.
964 If we don't find COUNT instances before reaching the end of the
965 buffer (or the beginning, if scanning backwards), we return
966 the number of line boundaries left unfound, and position at
967 the limit we bumped up against.
969 If ALLOW_QUIT, set immediate_quit. That's good to do
970 except in special cases. */
973 scan_newline (ptrdiff_t start
, ptrdiff_t start_byte
,
974 ptrdiff_t limit
, ptrdiff_t limit_byte
,
975 ptrdiff_t count
, bool allow_quit
)
977 ptrdiff_t charpos
, bytepos
, shortage
;
979 charpos
= find_newline (start
, start_byte
, limit
, limit_byte
,
980 count
, &shortage
, &bytepos
, allow_quit
);
982 TEMP_SET_PT_BOTH (limit
, limit_byte
);
984 TEMP_SET_PT_BOTH (charpos
, bytepos
);
988 /* Like find_newline, but doesn't allow QUITting and doesn't return
991 find_newline_no_quit (ptrdiff_t from
, ptrdiff_t frombyte
,
992 ptrdiff_t cnt
, ptrdiff_t *bytepos
)
994 return find_newline (from
, frombyte
, 0, -1, cnt
, NULL
, bytepos
, 0);
997 /* Like find_newline, but returns position before the newline, not
998 after, and only search up to TO.
999 This isn't just find_newline_no_quit (...)-1, because you might hit TO. */
1002 find_before_next_newline (ptrdiff_t from
, ptrdiff_t to
,
1003 ptrdiff_t cnt
, ptrdiff_t *bytepos
)
1006 ptrdiff_t pos
= find_newline (from
, -1, to
, -1, cnt
, &shortage
, bytepos
, 1);
1011 DEC_BOTH (pos
, *bytepos
);
1018 /* Subroutines of Lisp buffer search functions. */
1021 search_command (Lisp_Object string
, Lisp_Object bound
, Lisp_Object noerror
,
1022 Lisp_Object count
, int direction
, int RE
, bool posix
)
1027 EMACS_INT n
= direction
;
1031 CHECK_NUMBER (count
);
1035 CHECK_STRING (string
);
1039 lim
= ZV
, lim_byte
= ZV_BYTE
;
1041 lim
= BEGV
, lim_byte
= BEGV_BYTE
;
1045 CHECK_NUMBER_COERCE_MARKER (bound
);
1047 if (n
> 0 ? lim
< PT
: lim
> PT
)
1048 error ("Invalid search bound (wrong side of point)");
1050 lim
= ZV
, lim_byte
= ZV_BYTE
;
1051 else if (lim
< BEGV
)
1052 lim
= BEGV
, lim_byte
= BEGV_BYTE
;
1054 lim_byte
= CHAR_TO_BYTE (lim
);
1057 /* This is so set_image_of_range_1 in regex.c can find the EQV table. */
1058 set_char_table_extras (BVAR (current_buffer
, case_canon_table
), 2,
1059 BVAR (current_buffer
, case_eqv_table
));
1061 np
= search_buffer (string
, PT
, PT_BYTE
, lim
, lim_byte
, n
, RE
,
1062 (!NILP (BVAR (current_buffer
, case_fold_search
))
1063 ? BVAR (current_buffer
, case_canon_table
)
1065 (!NILP (BVAR (current_buffer
, case_fold_search
))
1066 ? BVAR (current_buffer
, case_eqv_table
)
1072 xsignal1 (Qsearch_failed
, string
);
1074 if (!EQ (noerror
, Qt
))
1076 eassert (BEGV
<= lim
&& lim
<= ZV
);
1077 SET_PT_BOTH (lim
, lim_byte
);
1079 #if 0 /* This would be clean, but maybe programs depend on
1080 a value of nil here. */
1088 eassert (BEGV
<= np
&& np
<= ZV
);
1091 return make_number (np
);
1094 /* Return true if REGEXP it matches just one constant string. */
1097 trivial_regexp_p (Lisp_Object regexp
)
1099 ptrdiff_t len
= SBYTES (regexp
);
1100 unsigned char *s
= SDATA (regexp
);
1105 case '.': case '*': case '+': case '?': case '[': case '^': case '$':
1112 case '|': case '(': case ')': case '`': case '\'': case 'b':
1113 case 'B': case '<': case '>': case 'w': case 'W': case 's':
1114 case 'S': case '=': case '{': case '}': case '_':
1115 case 'c': case 'C': /* for categoryspec and notcategoryspec */
1116 case '1': case '2': case '3': case '4': case '5':
1117 case '6': case '7': case '8': case '9':
1125 /* Search for the n'th occurrence of STRING in the current buffer,
1126 starting at position POS and stopping at position LIM,
1127 treating STRING as a literal string if RE is false or as
1128 a regular expression if RE is true.
1130 If N is positive, searching is forward and LIM must be greater than POS.
1131 If N is negative, searching is backward and LIM must be less than POS.
1133 Returns -x if x occurrences remain to be found (x > 0),
1134 or else the position at the beginning of the Nth occurrence
1135 (if searching backward) or the end (if searching forward).
1137 POSIX is nonzero if we want full backtracking (POSIX style)
1138 for this pattern. 0 means backtrack only enough to get a valid match. */
1140 #define TRANSLATE(out, trt, d) \
1146 temp = Faref (trt, make_number (d)); \
1147 if (INTEGERP (temp)) \
1148 out = XINT (temp); \
1157 /* Only used in search_buffer, to record the end position of the match
1158 when searching regexps and SEARCH_REGS should not be changed
1159 (i.e. Vinhibit_changing_match_data is non-nil). */
1160 static struct re_registers search_regs_1
;
1163 search_buffer (Lisp_Object string
, ptrdiff_t pos
, ptrdiff_t pos_byte
,
1164 ptrdiff_t lim
, ptrdiff_t lim_byte
, EMACS_INT n
,
1165 int RE
, Lisp_Object trt
, Lisp_Object inverse_trt
, bool posix
)
1167 ptrdiff_t len
= SCHARS (string
);
1168 ptrdiff_t len_byte
= SBYTES (string
);
1169 register ptrdiff_t i
;
1171 if (running_asynch_code
)
1172 save_search_regs ();
1174 /* Searching 0 times means don't move. */
1175 /* Null string is found at starting position. */
1176 if (len
== 0 || n
== 0)
1178 set_search_regs (pos_byte
, 0);
1182 if (RE
&& !(trivial_regexp_p (string
) && NILP (Vsearch_spaces_regexp
)))
1184 unsigned char *p1
, *p2
;
1186 struct re_pattern_buffer
*bufp
;
1188 bufp
= compile_pattern (string
,
1189 (NILP (Vinhibit_changing_match_data
)
1190 ? &search_regs
: &search_regs_1
),
1192 !NILP (BVAR (current_buffer
, enable_multibyte_characters
)));
1194 immediate_quit
= 1; /* Quit immediately if user types ^G,
1195 because letting this function finish
1196 can take too long. */
1197 QUIT
; /* Do a pending quit right away,
1198 to avoid paradoxical behavior */
1199 /* Get pointers and sizes of the two strings
1200 that make up the visible portion of the buffer. */
1203 s1
= GPT_BYTE
- BEGV_BYTE
;
1205 s2
= ZV_BYTE
- GPT_BYTE
;
1209 s2
= ZV_BYTE
- BEGV_BYTE
;
1214 s1
= ZV_BYTE
- BEGV_BYTE
;
1217 re_match_object
= Qnil
;
1223 val
= re_search_2 (bufp
, (char *) p1
, s1
, (char *) p2
, s2
,
1224 pos_byte
- BEGV_BYTE
, lim_byte
- pos_byte
,
1225 (NILP (Vinhibit_changing_match_data
)
1226 ? &search_regs
: &search_regs_1
),
1227 /* Don't allow match past current point */
1228 pos_byte
- BEGV_BYTE
);
1231 matcher_overflow ();
1235 if (NILP (Vinhibit_changing_match_data
))
1237 pos_byte
= search_regs
.start
[0] + BEGV_BYTE
;
1238 for (i
= 0; i
< search_regs
.num_regs
; i
++)
1239 if (search_regs
.start
[i
] >= 0)
1241 search_regs
.start
[i
]
1242 = BYTE_TO_CHAR (search_regs
.start
[i
] + BEGV_BYTE
);
1244 = BYTE_TO_CHAR (search_regs
.end
[i
] + BEGV_BYTE
);
1246 XSETBUFFER (last_thing_searched
, current_buffer
);
1247 /* Set pos to the new position. */
1248 pos
= search_regs
.start
[0];
1252 pos_byte
= search_regs_1
.start
[0] + BEGV_BYTE
;
1253 /* Set pos to the new position. */
1254 pos
= BYTE_TO_CHAR (search_regs_1
.start
[0] + BEGV_BYTE
);
1268 val
= re_search_2 (bufp
, (char *) p1
, s1
, (char *) p2
, s2
,
1269 pos_byte
- BEGV_BYTE
, lim_byte
- pos_byte
,
1270 (NILP (Vinhibit_changing_match_data
)
1271 ? &search_regs
: &search_regs_1
),
1272 lim_byte
- BEGV_BYTE
);
1275 matcher_overflow ();
1279 if (NILP (Vinhibit_changing_match_data
))
1281 pos_byte
= search_regs
.end
[0] + BEGV_BYTE
;
1282 for (i
= 0; i
< search_regs
.num_regs
; i
++)
1283 if (search_regs
.start
[i
] >= 0)
1285 search_regs
.start
[i
]
1286 = BYTE_TO_CHAR (search_regs
.start
[i
] + BEGV_BYTE
);
1288 = BYTE_TO_CHAR (search_regs
.end
[i
] + BEGV_BYTE
);
1290 XSETBUFFER (last_thing_searched
, current_buffer
);
1291 pos
= search_regs
.end
[0];
1295 pos_byte
= search_regs_1
.end
[0] + BEGV_BYTE
;
1296 pos
= BYTE_TO_CHAR (search_regs_1
.end
[0] + BEGV_BYTE
);
1309 else /* non-RE case */
1311 unsigned char *raw_pattern
, *pat
;
1312 ptrdiff_t raw_pattern_size
;
1313 ptrdiff_t raw_pattern_size_byte
;
1314 unsigned char *patbuf
;
1315 bool multibyte
= !NILP (BVAR (current_buffer
, enable_multibyte_characters
));
1316 unsigned char *base_pat
;
1317 /* Set to positive if we find a non-ASCII char that need
1318 translation. Otherwise set to zero later. */
1320 bool boyer_moore_ok
= 1;
1323 /* MULTIBYTE says whether the text to be searched is multibyte.
1324 We must convert PATTERN to match that, or we will not really
1325 find things right. */
1327 if (multibyte
== STRING_MULTIBYTE (string
))
1329 raw_pattern
= SDATA (string
);
1330 raw_pattern_size
= SCHARS (string
);
1331 raw_pattern_size_byte
= SBYTES (string
);
1335 raw_pattern_size
= SCHARS (string
);
1336 raw_pattern_size_byte
1337 = count_size_as_multibyte (SDATA (string
),
1339 raw_pattern
= SAFE_ALLOCA (raw_pattern_size_byte
+ 1);
1340 copy_text (SDATA (string
), raw_pattern
,
1341 SCHARS (string
), 0, 1);
1345 /* Converting multibyte to single-byte.
1347 ??? Perhaps this conversion should be done in a special way
1348 by subtracting nonascii-insert-offset from each non-ASCII char,
1349 so that only the multibyte chars which really correspond to
1350 the chosen single-byte character set can possibly match. */
1351 raw_pattern_size
= SCHARS (string
);
1352 raw_pattern_size_byte
= SCHARS (string
);
1353 raw_pattern
= SAFE_ALLOCA (raw_pattern_size
+ 1);
1354 copy_text (SDATA (string
), raw_pattern
,
1355 SBYTES (string
), 1, 0);
1358 /* Copy and optionally translate the pattern. */
1359 len
= raw_pattern_size
;
1360 len_byte
= raw_pattern_size_byte
;
1361 SAFE_NALLOCA (patbuf
, MAX_MULTIBYTE_LENGTH
, len
);
1363 base_pat
= raw_pattern
;
1366 /* Fill patbuf by translated characters in STRING while
1367 checking if we can use boyer-moore search. If TRT is
1368 non-nil, we can use boyer-moore search only if TRT can be
1369 represented by the byte array of 256 elements. For that,
1370 all non-ASCII case-equivalents of all case-sensitive
1371 characters in STRING must belong to the same character
1372 group (two characters belong to the same group iff their
1373 multibyte forms are the same except for the last byte;
1374 i.e. every 64 characters form a group; U+0000..U+003F,
1375 U+0040..U+007F, U+0080..U+00BF, ...). */
1379 unsigned char str_base
[MAX_MULTIBYTE_LENGTH
], *str
;
1380 int c
, translated
, inverse
;
1381 int in_charlen
, charlen
;
1383 /* If we got here and the RE flag is set, it's because we're
1384 dealing with a regexp known to be trivial, so the backslash
1385 just quotes the next character. */
1386 if (RE
&& *base_pat
== '\\')
1394 c
= STRING_CHAR_AND_LENGTH (base_pat
, in_charlen
);
1399 charlen
= in_charlen
;
1403 /* Translate the character. */
1404 TRANSLATE (translated
, trt
, c
);
1405 charlen
= CHAR_STRING (translated
, str_base
);
1408 /* Check if C has any other case-equivalents. */
1409 TRANSLATE (inverse
, inverse_trt
, c
);
1410 /* If so, check if we can use boyer-moore. */
1411 if (c
!= inverse
&& boyer_moore_ok
)
1413 /* Check if all equivalents belong to the same
1414 group of characters. Note that the check of C
1415 itself is done by the last iteration. */
1416 int this_char_base
= -1;
1418 while (boyer_moore_ok
)
1420 if (ASCII_CHAR_P (inverse
))
1422 if (this_char_base
> 0)
1427 else if (CHAR_BYTE8_P (inverse
))
1428 /* Boyer-moore search can't handle a
1429 translation of an eight-bit
1432 else if (this_char_base
< 0)
1434 this_char_base
= inverse
& ~0x3F;
1436 char_base
= this_char_base
;
1437 else if (this_char_base
!= char_base
)
1440 else if ((inverse
& ~0x3F) != this_char_base
)
1444 TRANSLATE (inverse
, inverse_trt
, inverse
);
1449 /* Store this character into the translated pattern. */
1450 memcpy (pat
, str
, charlen
);
1452 base_pat
+= in_charlen
;
1453 len_byte
-= in_charlen
;
1456 /* If char_base is still negative we didn't find any translated
1457 non-ASCII characters. */
1463 /* Unibyte buffer. */
1467 int c
, translated
, inverse
;
1469 /* If we got here and the RE flag is set, it's because we're
1470 dealing with a regexp known to be trivial, so the backslash
1471 just quotes the next character. */
1472 if (RE
&& *base_pat
== '\\')
1479 TRANSLATE (translated
, trt
, c
);
1480 *pat
++ = translated
;
1481 /* Check that none of C's equivalents violates the
1482 assumptions of boyer_moore. */
1483 TRANSLATE (inverse
, inverse_trt
, c
);
1486 if (inverse
>= 0200)
1493 TRANSLATE (inverse
, inverse_trt
, inverse
);
1498 len_byte
= pat
- patbuf
;
1499 pat
= base_pat
= patbuf
;
1503 ? boyer_moore (n
, pat
, len_byte
, trt
, inverse_trt
,
1506 : simple_search (n
, pat
, raw_pattern_size
, len_byte
, trt
,
1507 pos
, pos_byte
, lim
, lim_byte
));
1513 /* Do a simple string search N times for the string PAT,
1514 whose length is LEN/LEN_BYTE,
1515 from buffer position POS/POS_BYTE until LIM/LIM_BYTE.
1516 TRT is the translation table.
1518 Return the character position where the match is found.
1519 Otherwise, if M matches remained to be found, return -M.
1521 This kind of search works regardless of what is in PAT and
1522 regardless of what is in TRT. It is used in cases where
1523 boyer_moore cannot work. */
1526 simple_search (EMACS_INT n
, unsigned char *pat
,
1527 ptrdiff_t len
, ptrdiff_t len_byte
, Lisp_Object trt
,
1528 ptrdiff_t pos
, ptrdiff_t pos_byte
,
1529 ptrdiff_t lim
, ptrdiff_t lim_byte
)
1531 bool multibyte
= ! NILP (BVAR (current_buffer
, enable_multibyte_characters
));
1532 bool forward
= n
> 0;
1533 /* Number of buffer bytes matched. Note that this may be different
1534 from len_byte in a multibyte buffer. */
1535 ptrdiff_t match_byte
= PTRDIFF_MIN
;
1537 if (lim
> pos
&& multibyte
)
1542 /* Try matching at position POS. */
1543 ptrdiff_t this_pos
= pos
;
1544 ptrdiff_t this_pos_byte
= pos_byte
;
1545 ptrdiff_t this_len
= len
;
1546 unsigned char *p
= pat
;
1547 if (pos
+ len
> lim
|| pos_byte
+ len_byte
> lim_byte
)
1550 while (this_len
> 0)
1552 int charlen
, buf_charlen
;
1555 pat_ch
= STRING_CHAR_AND_LENGTH (p
, charlen
);
1556 buf_ch
= STRING_CHAR_AND_LENGTH (BYTE_POS_ADDR (this_pos_byte
),
1558 TRANSLATE (buf_ch
, trt
, buf_ch
);
1560 if (buf_ch
!= pat_ch
)
1566 this_pos_byte
+= buf_charlen
;
1572 match_byte
= this_pos_byte
- pos_byte
;
1574 pos_byte
+= match_byte
;
1578 INC_BOTH (pos
, pos_byte
);
1588 /* Try matching at position POS. */
1589 ptrdiff_t this_pos
= pos
;
1590 ptrdiff_t this_len
= len
;
1591 unsigned char *p
= pat
;
1593 if (pos
+ len
> lim
)
1596 while (this_len
> 0)
1599 int buf_ch
= FETCH_BYTE (this_pos
);
1600 TRANSLATE (buf_ch
, trt
, buf_ch
);
1602 if (buf_ch
!= pat_ch
)
1621 /* Backwards search. */
1622 else if (lim
< pos
&& multibyte
)
1627 /* Try matching at position POS. */
1628 ptrdiff_t this_pos
= pos
;
1629 ptrdiff_t this_pos_byte
= pos_byte
;
1630 ptrdiff_t this_len
= len
;
1631 const unsigned char *p
= pat
+ len_byte
;
1633 if (this_pos
- len
< lim
|| (pos_byte
- len_byte
) < lim_byte
)
1636 while (this_len
> 0)
1640 DEC_BOTH (this_pos
, this_pos_byte
);
1641 PREV_CHAR_BOUNDARY (p
, pat
);
1642 pat_ch
= STRING_CHAR (p
);
1643 buf_ch
= STRING_CHAR (BYTE_POS_ADDR (this_pos_byte
));
1644 TRANSLATE (buf_ch
, trt
, buf_ch
);
1646 if (buf_ch
!= pat_ch
)
1654 match_byte
= pos_byte
- this_pos_byte
;
1656 pos_byte
= this_pos_byte
;
1660 DEC_BOTH (pos
, pos_byte
);
1670 /* Try matching at position POS. */
1671 ptrdiff_t this_pos
= pos
- len
;
1672 ptrdiff_t this_len
= len
;
1673 unsigned char *p
= pat
;
1678 while (this_len
> 0)
1681 int buf_ch
= FETCH_BYTE (this_pos
);
1682 TRANSLATE (buf_ch
, trt
, buf_ch
);
1684 if (buf_ch
!= pat_ch
)
1706 eassert (match_byte
!= PTRDIFF_MIN
);
1708 set_search_regs ((multibyte
? pos_byte
: pos
) - match_byte
, match_byte
);
1710 set_search_regs (multibyte
? pos_byte
: pos
, match_byte
);
1720 /* Do Boyer-Moore search N times for the string BASE_PAT,
1721 whose length is LEN_BYTE,
1722 from buffer position POS_BYTE until LIM_BYTE.
1723 DIRECTION says which direction we search in.
1724 TRT and INVERSE_TRT are translation tables.
1725 Characters in PAT are already translated by TRT.
1727 This kind of search works if all the characters in BASE_PAT that
1728 have nontrivial translation are the same aside from the last byte.
1729 This makes it possible to translate just the last byte of a
1730 character, and do so after just a simple test of the context.
1731 CHAR_BASE is nonzero if there is such a non-ASCII character.
1733 If that criterion is not satisfied, do not call this function. */
1736 boyer_moore (EMACS_INT n
, unsigned char *base_pat
,
1738 Lisp_Object trt
, Lisp_Object inverse_trt
,
1739 ptrdiff_t pos_byte
, ptrdiff_t lim_byte
,
1742 int direction
= ((n
> 0) ? 1 : -1);
1743 register ptrdiff_t dirlen
;
1745 int stride_for_teases
= 0;
1747 register unsigned char *cursor
, *p_limit
;
1748 register ptrdiff_t i
;
1750 unsigned char *pat
, *pat_end
;
1751 bool multibyte
= ! NILP (BVAR (current_buffer
, enable_multibyte_characters
));
1753 unsigned char simple_translate
[0400];
1754 /* These are set to the preceding bytes of a byte to be translated
1755 if char_base is nonzero. As the maximum byte length of a
1756 multibyte character is 5, we have to check at most four previous
1758 int translate_prev_byte1
= 0;
1759 int translate_prev_byte2
= 0;
1760 int translate_prev_byte3
= 0;
1762 /* The general approach is that we are going to maintain that we know
1763 the first (closest to the present position, in whatever direction
1764 we're searching) character that could possibly be the last
1765 (furthest from present position) character of a valid match. We
1766 advance the state of our knowledge by looking at that character
1767 and seeing whether it indeed matches the last character of the
1768 pattern. If it does, we take a closer look. If it does not, we
1769 move our pointer (to putative last characters) as far as is
1770 logically possible. This amount of movement, which I call a
1771 stride, will be the length of the pattern if the actual character
1772 appears nowhere in the pattern, otherwise it will be the distance
1773 from the last occurrence of that character to the end of the
1774 pattern. If the amount is zero we have a possible match. */
1776 /* Here we make a "mickey mouse" BM table. The stride of the search
1777 is determined only by the last character of the putative match.
1778 If that character does not match, we will stride the proper
1779 distance to propose a match that superimposes it on the last
1780 instance of a character that matches it (per trt), or misses
1781 it entirely if there is none. */
1783 dirlen
= len_byte
* direction
;
1785 /* Record position after the end of the pattern. */
1786 pat_end
= base_pat
+ len_byte
;
1787 /* BASE_PAT points to a character that we start scanning from.
1788 It is the first character in a forward search,
1789 the last character in a backward search. */
1791 base_pat
= pat_end
- 1;
1793 /* A character that does not appear in the pattern induces a
1794 stride equal to the pattern length. */
1795 for (i
= 0; i
< 0400; i
++)
1798 /* We use this for translation, instead of TRT itself.
1799 We fill this in to handle the characters that actually
1800 occur in the pattern. Others don't matter anyway! */
1801 for (i
= 0; i
< 0400; i
++)
1802 simple_translate
[i
] = i
;
1806 /* Setup translate_prev_byte1/2/3/4 from CHAR_BASE. Only a
1807 byte following them are the target of translation. */
1808 unsigned char str
[MAX_MULTIBYTE_LENGTH
];
1809 int cblen
= CHAR_STRING (char_base
, str
);
1811 translate_prev_byte1
= str
[cblen
- 2];
1814 translate_prev_byte2
= str
[cblen
- 3];
1816 translate_prev_byte3
= str
[cblen
- 4];
1823 unsigned char *ptr
= base_pat
+ i
;
1827 /* If the byte currently looking at is the last of a
1828 character to check case-equivalents, set CH to that
1829 character. An ASCII character and a non-ASCII character
1830 matching with CHAR_BASE are to be checked. */
1833 if (ASCII_CHAR_P (*ptr
) || ! multibyte
)
1836 && ((pat_end
- ptr
) == 1 || CHAR_HEAD_P (ptr
[1])))
1838 unsigned char *charstart
= ptr
- 1;
1840 while (! (CHAR_HEAD_P (*charstart
)))
1842 ch
= STRING_CHAR (charstart
);
1843 if (char_base
!= (ch
& ~0x3F))
1847 if (ch
>= 0200 && multibyte
)
1848 j
= (ch
& 0x3F) | 0200;
1853 stride_for_teases
= BM_tab
[j
];
1855 BM_tab
[j
] = dirlen
- i
;
1856 /* A translation table is accompanied by its inverse -- see
1857 comment following downcase_table for details. */
1860 int starting_ch
= ch
;
1865 TRANSLATE (ch
, inverse_trt
, ch
);
1866 if (ch
>= 0200 && multibyte
)
1867 j
= (ch
& 0x3F) | 0200;
1871 /* For all the characters that map into CH,
1872 set up simple_translate to map the last byte
1874 simple_translate
[j
] = starting_j
;
1875 if (ch
== starting_ch
)
1877 BM_tab
[j
] = dirlen
- i
;
1886 stride_for_teases
= BM_tab
[j
];
1887 BM_tab
[j
] = dirlen
- i
;
1889 /* stride_for_teases tells how much to stride if we get a
1890 match on the far character but are subsequently
1891 disappointed, by recording what the stride would have been
1892 for that character if the last character had been
1895 pos_byte
+= dirlen
- ((direction
> 0) ? direction
: 0);
1896 /* loop invariant - POS_BYTE points at where last char (first
1897 char if reverse) of pattern would align in a possible match. */
1901 unsigned char *tail_end_ptr
;
1903 /* It's been reported that some (broken) compiler thinks that
1904 Boolean expressions in an arithmetic context are unsigned.
1905 Using an explicit ?1:0 prevents this. */
1906 if ((lim_byte
- pos_byte
- ((direction
> 0) ? 1 : 0)) * direction
1908 return (n
* (0 - direction
));
1909 /* First we do the part we can by pointers (maybe nothing) */
1912 limit
= pos_byte
- dirlen
+ direction
;
1915 limit
= BUFFER_CEILING_OF (limit
);
1916 /* LIMIT is now the last (not beyond-last!) value POS_BYTE
1917 can take on without hitting edge of buffer or the gap. */
1918 limit
= min (limit
, pos_byte
+ 20000);
1919 limit
= min (limit
, lim_byte
- 1);
1923 limit
= BUFFER_FLOOR_OF (limit
);
1924 /* LIMIT is now the last (not beyond-last!) value POS_BYTE
1925 can take on without hitting edge of buffer or the gap. */
1926 limit
= max (limit
, pos_byte
- 20000);
1927 limit
= max (limit
, lim_byte
);
1929 tail_end
= BUFFER_CEILING_OF (pos_byte
) + 1;
1930 tail_end_ptr
= BYTE_POS_ADDR (tail_end
);
1932 if ((limit
- pos_byte
) * direction
> 20)
1936 p_limit
= BYTE_POS_ADDR (limit
);
1937 p2
= (cursor
= BYTE_POS_ADDR (pos_byte
));
1938 /* In this loop, pos + cursor - p2 is the surrogate for pos. */
1939 while (1) /* use one cursor setting as long as i can */
1941 if (direction
> 0) /* worth duplicating */
1943 while (cursor
<= p_limit
)
1945 if (BM_tab
[*cursor
] == 0)
1947 cursor
+= BM_tab
[*cursor
];
1952 while (cursor
>= p_limit
)
1954 if (BM_tab
[*cursor
] == 0)
1956 cursor
+= BM_tab
[*cursor
];
1959 /* If you are here, cursor is beyond the end of the
1960 searched region. You fail to match within the
1961 permitted region and would otherwise try a character
1962 beyond that region. */
1966 i
= dirlen
- direction
;
1969 while ((i
-= direction
) + direction
!= 0)
1972 cursor
-= direction
;
1973 /* Translate only the last byte of a character. */
1975 || ((cursor
== tail_end_ptr
1976 || CHAR_HEAD_P (cursor
[1]))
1977 && (CHAR_HEAD_P (cursor
[0])
1978 /* Check if this is the last byte of
1979 a translatable character. */
1980 || (translate_prev_byte1
== cursor
[-1]
1981 && (CHAR_HEAD_P (translate_prev_byte1
)
1982 || (translate_prev_byte2
== cursor
[-2]
1983 && (CHAR_HEAD_P (translate_prev_byte2
)
1984 || (translate_prev_byte3
== cursor
[-3]))))))))
1985 ch
= simple_translate
[*cursor
];
1994 while ((i
-= direction
) + direction
!= 0)
1996 cursor
-= direction
;
1997 if (pat
[i
] != *cursor
)
2001 cursor
+= dirlen
- i
- direction
; /* fix cursor */
2002 if (i
+ direction
== 0)
2004 ptrdiff_t position
, start
, end
;
2006 cursor
-= direction
;
2008 position
= pos_byte
+ cursor
- p2
+ ((direction
> 0)
2009 ? 1 - len_byte
: 0);
2010 set_search_regs (position
, len_byte
);
2012 if (NILP (Vinhibit_changing_match_data
))
2014 start
= search_regs
.start
[0];
2015 end
= search_regs
.end
[0];
2018 /* If Vinhibit_changing_match_data is non-nil,
2019 search_regs will not be changed. So let's
2020 compute start and end here. */
2022 start
= BYTE_TO_CHAR (position
);
2023 end
= BYTE_TO_CHAR (position
+ len_byte
);
2026 if ((n
-= direction
) != 0)
2027 cursor
+= dirlen
; /* to resume search */
2029 return direction
> 0 ? end
: start
;
2032 cursor
+= stride_for_teases
; /* <sigh> we lose - */
2034 pos_byte
+= cursor
- p2
;
2037 /* Now we'll pick up a clump that has to be done the hard
2038 way because it covers a discontinuity. */
2040 limit
= ((direction
> 0)
2041 ? BUFFER_CEILING_OF (pos_byte
- dirlen
+ 1)
2042 : BUFFER_FLOOR_OF (pos_byte
- dirlen
- 1));
2043 limit
= ((direction
> 0)
2044 ? min (limit
+ len_byte
, lim_byte
- 1)
2045 : max (limit
- len_byte
, lim_byte
));
2046 /* LIMIT is now the last value POS_BYTE can have
2047 and still be valid for a possible match. */
2050 /* This loop can be coded for space rather than
2051 speed because it will usually run only once.
2052 (the reach is at most len + 21, and typically
2053 does not exceed len). */
2054 while ((limit
- pos_byte
) * direction
>= 0)
2056 int ch
= FETCH_BYTE (pos_byte
);
2057 if (BM_tab
[ch
] == 0)
2059 pos_byte
+= BM_tab
[ch
];
2061 break; /* ran off the end */
2064 /* Found what might be a match. */
2065 i
= dirlen
- direction
;
2066 while ((i
-= direction
) + direction
!= 0)
2070 pos_byte
-= direction
;
2071 ptr
= BYTE_POS_ADDR (pos_byte
);
2072 /* Translate only the last byte of a character. */
2074 || ((ptr
== tail_end_ptr
2075 || CHAR_HEAD_P (ptr
[1]))
2076 && (CHAR_HEAD_P (ptr
[0])
2077 /* Check if this is the last byte of a
2078 translatable character. */
2079 || (translate_prev_byte1
== ptr
[-1]
2080 && (CHAR_HEAD_P (translate_prev_byte1
)
2081 || (translate_prev_byte2
== ptr
[-2]
2082 && (CHAR_HEAD_P (translate_prev_byte2
)
2083 || translate_prev_byte3
== ptr
[-3])))))))
2084 ch
= simple_translate
[*ptr
];
2090 /* Above loop has moved POS_BYTE part or all the way
2091 back to the first pos (last pos if reverse).
2092 Set it once again at the last (first if reverse) char. */
2093 pos_byte
+= dirlen
- i
- direction
;
2094 if (i
+ direction
== 0)
2096 ptrdiff_t position
, start
, end
;
2097 pos_byte
-= direction
;
2099 position
= pos_byte
+ ((direction
> 0) ? 1 - len_byte
: 0);
2100 set_search_regs (position
, len_byte
);
2102 if (NILP (Vinhibit_changing_match_data
))
2104 start
= search_regs
.start
[0];
2105 end
= search_regs
.end
[0];
2108 /* If Vinhibit_changing_match_data is non-nil,
2109 search_regs will not be changed. So let's
2110 compute start and end here. */
2112 start
= BYTE_TO_CHAR (position
);
2113 end
= BYTE_TO_CHAR (position
+ len_byte
);
2116 if ((n
-= direction
) != 0)
2117 pos_byte
+= dirlen
; /* to resume search */
2119 return direction
> 0 ? end
: start
;
2122 pos_byte
+= stride_for_teases
;
2125 /* We have done one clump. Can we continue? */
2126 if ((lim_byte
- pos_byte
) * direction
< 0)
2127 return ((0 - n
) * direction
);
2129 return BYTE_TO_CHAR (pos_byte
);
2132 /* Record beginning BEG_BYTE and end BEG_BYTE + NBYTES
2133 for the overall match just found in the current buffer.
2134 Also clear out the match data for registers 1 and up. */
2137 set_search_regs (ptrdiff_t beg_byte
, ptrdiff_t nbytes
)
2141 if (!NILP (Vinhibit_changing_match_data
))
2144 /* Make sure we have registers in which to store
2145 the match position. */
2146 if (search_regs
.num_regs
== 0)
2148 search_regs
.start
= xmalloc (2 * sizeof (regoff_t
));
2149 search_regs
.end
= xmalloc (2 * sizeof (regoff_t
));
2150 search_regs
.num_regs
= 2;
2153 /* Clear out the other registers. */
2154 for (i
= 1; i
< search_regs
.num_regs
; i
++)
2156 search_regs
.start
[i
] = -1;
2157 search_regs
.end
[i
] = -1;
2160 search_regs
.start
[0] = BYTE_TO_CHAR (beg_byte
);
2161 search_regs
.end
[0] = BYTE_TO_CHAR (beg_byte
+ nbytes
);
2162 XSETBUFFER (last_thing_searched
, current_buffer
);
2165 DEFUN ("search-backward", Fsearch_backward
, Ssearch_backward
, 1, 4,
2166 "MSearch backward: ",
2167 doc
: /* Search backward from point for STRING.
2168 Set point to the beginning of the occurrence found, and return point.
2169 An optional second argument bounds the search; it is a buffer position.
2170 The match found must not extend before that position.
2171 Optional third argument, if t, means if fail just return nil (no error).
2172 If not nil and not t, position at limit of search and return nil.
2173 Optional fourth argument COUNT, if non-nil, means to search for COUNT
2174 successive occurrences. If COUNT is negative, search forward,
2175 instead of backward, for -COUNT occurrences.
2177 Search case-sensitivity is determined by the value of the variable
2178 `case-fold-search', which see.
2180 See also the functions `match-beginning', `match-end' and `replace-match'. */)
2181 (Lisp_Object string
, Lisp_Object bound
, Lisp_Object noerror
, Lisp_Object count
)
2183 return search_command (string
, bound
, noerror
, count
, -1, 0, 0);
2186 DEFUN ("search-forward", Fsearch_forward
, Ssearch_forward
, 1, 4, "MSearch: ",
2187 doc
: /* Search forward from point for STRING.
2188 Set point to the end of the occurrence found, and return point.
2189 An optional second argument bounds the search; it is a buffer position.
2190 The match found must not extend after that position. A value of nil is
2191 equivalent to (point-max).
2192 Optional third argument, if t, means if fail just return nil (no error).
2193 If not nil and not t, move to limit of search and return nil.
2194 Optional fourth argument COUNT, if non-nil, means to search for COUNT
2195 successive occurrences. If COUNT is negative, search backward,
2196 instead of forward, for -COUNT occurrences.
2198 Search case-sensitivity is determined by the value of the variable
2199 `case-fold-search', which see.
2201 See also the functions `match-beginning', `match-end' and `replace-match'. */)
2202 (Lisp_Object string
, Lisp_Object bound
, Lisp_Object noerror
, Lisp_Object count
)
2204 return search_command (string
, bound
, noerror
, count
, 1, 0, 0);
2207 DEFUN ("re-search-backward", Fre_search_backward
, Sre_search_backward
, 1, 4,
2208 "sRE search backward: ",
2209 doc
: /* Search backward from point for match for regular expression REGEXP.
2210 Set point to the beginning of the match, and return point.
2211 The match found is the one starting last in the buffer
2212 and yet ending before the origin of the search.
2213 An optional second argument bounds the search; it is a buffer position.
2214 The match found must start at or after that position.
2215 Optional third argument, if t, means if fail just return nil (no error).
2216 If not nil and not t, move to limit of search and return nil.
2217 Optional fourth argument is repeat count--search for successive occurrences.
2219 Search case-sensitivity is determined by the value of the variable
2220 `case-fold-search', which see.
2222 See also the functions `match-beginning', `match-end', `match-string',
2223 and `replace-match'. */)
2224 (Lisp_Object regexp
, Lisp_Object bound
, Lisp_Object noerror
, Lisp_Object count
)
2226 return search_command (regexp
, bound
, noerror
, count
, -1, 1, 0);
2229 DEFUN ("re-search-forward", Fre_search_forward
, Sre_search_forward
, 1, 4,
2231 doc
: /* Search forward from point for regular expression REGEXP.
2232 Set point to the end of the occurrence found, and return point.
2233 An optional second argument bounds the search; it is a buffer position.
2234 The match found must not extend after that position.
2235 Optional third argument, if t, means if fail just return nil (no error).
2236 If not nil and not t, move to limit of search and return nil.
2237 Optional fourth argument is repeat count--search for successive occurrences.
2239 Search case-sensitivity is determined by the value of the variable
2240 `case-fold-search', which see.
2242 See also the functions `match-beginning', `match-end', `match-string',
2243 and `replace-match'. */)
2244 (Lisp_Object regexp
, Lisp_Object bound
, Lisp_Object noerror
, Lisp_Object count
)
2246 return search_command (regexp
, bound
, noerror
, count
, 1, 1, 0);
2249 DEFUN ("posix-search-backward", Fposix_search_backward
, Sposix_search_backward
, 1, 4,
2250 "sPosix search backward: ",
2251 doc
: /* Search backward from point for match for regular expression REGEXP.
2252 Find the longest match in accord with Posix regular expression rules.
2253 Set point to the beginning of the match, and return point.
2254 The match found is the one starting last in the buffer
2255 and yet ending before the origin of the search.
2256 An optional second argument bounds the search; it is a buffer position.
2257 The match found must start at or after that position.
2258 Optional third argument, if t, means if fail just return nil (no error).
2259 If not nil and not t, move to limit of search and return nil.
2260 Optional fourth argument is repeat count--search for successive occurrences.
2262 Search case-sensitivity is determined by the value of the variable
2263 `case-fold-search', which see.
2265 See also the functions `match-beginning', `match-end', `match-string',
2266 and `replace-match'. */)
2267 (Lisp_Object regexp
, Lisp_Object bound
, Lisp_Object noerror
, Lisp_Object count
)
2269 return search_command (regexp
, bound
, noerror
, count
, -1, 1, 1);
2272 DEFUN ("posix-search-forward", Fposix_search_forward
, Sposix_search_forward
, 1, 4,
2274 doc
: /* Search forward from point for regular expression REGEXP.
2275 Find the longest match in accord with Posix regular expression rules.
2276 Set point to the end of the occurrence found, and return point.
2277 An optional second argument bounds the search; it is a buffer position.
2278 The match found must not extend after that position.
2279 Optional third argument, if t, means if fail just return nil (no error).
2280 If not nil and not t, move to limit of search and return nil.
2281 Optional fourth argument is repeat count--search for successive occurrences.
2283 Search case-sensitivity is determined by the value of the variable
2284 `case-fold-search', which see.
2286 See also the functions `match-beginning', `match-end', `match-string',
2287 and `replace-match'. */)
2288 (Lisp_Object regexp
, Lisp_Object bound
, Lisp_Object noerror
, Lisp_Object count
)
2290 return search_command (regexp
, bound
, noerror
, count
, 1, 1, 1);
2293 DEFUN ("replace-match", Freplace_match
, Sreplace_match
, 1, 5, 0,
2294 doc
: /* Replace text matched by last search with NEWTEXT.
2295 Leave point at the end of the replacement text.
2297 If optional second arg FIXEDCASE is non-nil, do not alter the case of
2298 the replacement text. Otherwise, maybe capitalize the whole text, or
2299 maybe just word initials, based on the replaced text. If the replaced
2300 text has only capital letters and has at least one multiletter word,
2301 convert NEWTEXT to all caps. Otherwise if all words are capitalized
2302 in the replaced text, capitalize each word in NEWTEXT.
2304 If optional third arg LITERAL is non-nil, insert NEWTEXT literally.
2305 Otherwise treat `\\' as special:
2306 `\\&' in NEWTEXT means substitute original matched text.
2307 `\\N' means substitute what matched the Nth `\\(...\\)'.
2308 If Nth parens didn't match, substitute nothing.
2309 `\\\\' means insert one `\\'.
2310 `\\?' is treated literally
2311 (for compatibility with `query-replace-regexp').
2312 Any other character following `\\' signals an error.
2313 Case conversion does not apply to these substitutions.
2315 If optional fourth argument STRING is non-nil, it should be a string
2316 to act on; this should be the string on which the previous match was
2317 done via `string-match'. In this case, `replace-match' creates and
2318 returns a new string, made by copying STRING and replacing the part of
2319 STRING that was matched (the original STRING itself is not altered).
2321 The optional fifth argument SUBEXP specifies a subexpression;
2322 it says to replace just that subexpression with NEWTEXT,
2323 rather than replacing the entire matched text.
2324 This is, in a vague sense, the inverse of using `\\N' in NEWTEXT;
2325 `\\N' copies subexp N into NEWTEXT, but using N as SUBEXP puts
2326 NEWTEXT in place of subexp N.
2327 This is useful only after a regular expression search or match,
2328 since only regular expressions have distinguished subexpressions. */)
2329 (Lisp_Object newtext
, Lisp_Object fixedcase
, Lisp_Object literal
, Lisp_Object string
, Lisp_Object subexp
)
2331 enum { nochange
, all_caps
, cap_initial
} case_action
;
2332 ptrdiff_t pos
, pos_byte
;
2333 bool some_multiletter_word
;
2334 bool some_lowercase
;
2335 bool some_uppercase
;
2336 bool some_nonuppercase_initial
;
2339 ptrdiff_t opoint
, newpoint
;
2341 CHECK_STRING (newtext
);
2343 if (! NILP (string
))
2344 CHECK_STRING (string
);
2346 case_action
= nochange
; /* We tried an initialization */
2347 /* but some C compilers blew it */
2349 if (search_regs
.num_regs
<= 0)
2350 error ("`replace-match' called before any match found");
2356 CHECK_NUMBER (subexp
);
2357 if (! (0 <= XINT (subexp
) && XINT (subexp
) < search_regs
.num_regs
))
2358 args_out_of_range (subexp
, make_number (search_regs
.num_regs
));
2359 sub
= XINT (subexp
);
2364 if (search_regs
.start
[sub
] < BEGV
2365 || search_regs
.start
[sub
] > search_regs
.end
[sub
]
2366 || search_regs
.end
[sub
] > ZV
)
2367 args_out_of_range (make_number (search_regs
.start
[sub
]),
2368 make_number (search_regs
.end
[sub
]));
2372 if (search_regs
.start
[sub
] < 0
2373 || search_regs
.start
[sub
] > search_regs
.end
[sub
]
2374 || search_regs
.end
[sub
] > SCHARS (string
))
2375 args_out_of_range (make_number (search_regs
.start
[sub
]),
2376 make_number (search_regs
.end
[sub
]));
2379 if (NILP (fixedcase
))
2381 /* Decide how to casify by examining the matched text. */
2384 pos
= search_regs
.start
[sub
];
2385 last
= search_regs
.end
[sub
];
2388 pos_byte
= CHAR_TO_BYTE (pos
);
2390 pos_byte
= string_char_to_byte (string
, pos
);
2393 case_action
= all_caps
;
2395 /* some_multiletter_word is set nonzero if any original word
2396 is more than one letter long. */
2397 some_multiletter_word
= 0;
2399 some_nonuppercase_initial
= 0;
2406 c
= FETCH_CHAR_AS_MULTIBYTE (pos_byte
);
2407 INC_BOTH (pos
, pos_byte
);
2410 FETCH_STRING_CHAR_AS_MULTIBYTE_ADVANCE (c
, string
, pos
, pos_byte
);
2414 /* Cannot be all caps if any original char is lower case */
2417 if (SYNTAX (prevc
) != Sword
)
2418 some_nonuppercase_initial
= 1;
2420 some_multiletter_word
= 1;
2422 else if (uppercasep (c
))
2425 if (SYNTAX (prevc
) != Sword
)
2428 some_multiletter_word
= 1;
2432 /* If the initial is a caseless word constituent,
2433 treat that like a lowercase initial. */
2434 if (SYNTAX (prevc
) != Sword
)
2435 some_nonuppercase_initial
= 1;
2441 /* Convert to all caps if the old text is all caps
2442 and has at least one multiletter word. */
2443 if (! some_lowercase
&& some_multiletter_word
)
2444 case_action
= all_caps
;
2445 /* Capitalize each word, if the old text has all capitalized words. */
2446 else if (!some_nonuppercase_initial
&& some_multiletter_word
)
2447 case_action
= cap_initial
;
2448 else if (!some_nonuppercase_initial
&& some_uppercase
)
2449 /* Should x -> yz, operating on X, give Yz or YZ?
2450 We'll assume the latter. */
2451 case_action
= all_caps
;
2453 case_action
= nochange
;
2456 /* Do replacement in a string. */
2459 Lisp_Object before
, after
;
2461 before
= Fsubstring (string
, make_number (0),
2462 make_number (search_regs
.start
[sub
]));
2463 after
= Fsubstring (string
, make_number (search_regs
.end
[sub
]), Qnil
);
2465 /* Substitute parts of the match into NEWTEXT
2469 ptrdiff_t lastpos
= 0;
2470 ptrdiff_t lastpos_byte
= 0;
2471 /* We build up the substituted string in ACCUM. */
2474 ptrdiff_t length
= SBYTES (newtext
);
2478 for (pos_byte
= 0, pos
= 0; pos_byte
< length
;)
2480 ptrdiff_t substart
= -1;
2481 ptrdiff_t subend
= 0;
2482 bool delbackslash
= 0;
2484 FETCH_STRING_CHAR_ADVANCE (c
, newtext
, pos
, pos_byte
);
2488 FETCH_STRING_CHAR_ADVANCE (c
, newtext
, pos
, pos_byte
);
2492 substart
= search_regs
.start
[sub
];
2493 subend
= search_regs
.end
[sub
];
2495 else if (c
>= '1' && c
<= '9')
2497 if (c
- '0' < search_regs
.num_regs
2498 && search_regs
.start
[c
- '0'] >= 0)
2500 substart
= search_regs
.start
[c
- '0'];
2501 subend
= search_regs
.end
[c
- '0'];
2505 /* If that subexp did not match,
2506 replace \\N with nothing. */
2514 error ("Invalid use of `\\' in replacement text");
2518 if (pos
- 2 != lastpos
)
2519 middle
= substring_both (newtext
, lastpos
,
2521 pos
- 2, pos_byte
- 2);
2524 accum
= concat3 (accum
, middle
,
2526 make_number (substart
),
2527 make_number (subend
)));
2529 lastpos_byte
= pos_byte
;
2531 else if (delbackslash
)
2533 middle
= substring_both (newtext
, lastpos
,
2535 pos
- 1, pos_byte
- 1);
2537 accum
= concat2 (accum
, middle
);
2539 lastpos_byte
= pos_byte
;
2544 middle
= substring_both (newtext
, lastpos
,
2550 newtext
= concat2 (accum
, middle
);
2553 /* Do case substitution in NEWTEXT if desired. */
2554 if (case_action
== all_caps
)
2555 newtext
= Fupcase (newtext
);
2556 else if (case_action
== cap_initial
)
2557 newtext
= Fupcase_initials (newtext
);
2559 return concat3 (before
, newtext
, after
);
2562 /* Record point, then move (quietly) to the start of the match. */
2563 if (PT
>= search_regs
.end
[sub
])
2565 else if (PT
> search_regs
.start
[sub
])
2566 opoint
= search_regs
.end
[sub
] - ZV
;
2570 /* If we want non-literal replacement,
2571 perform substitution on the replacement string. */
2574 ptrdiff_t length
= SBYTES (newtext
);
2575 unsigned char *substed
;
2576 ptrdiff_t substed_alloc_size
, substed_len
;
2577 bool buf_multibyte
= !NILP (BVAR (current_buffer
, enable_multibyte_characters
));
2578 bool str_multibyte
= STRING_MULTIBYTE (newtext
);
2579 bool really_changed
= 0;
2581 substed_alloc_size
= (length
<= (STRING_BYTES_BOUND
- 100) / 2
2583 : STRING_BYTES_BOUND
);
2584 substed
= xmalloc (substed_alloc_size
);
2587 /* Go thru NEWTEXT, producing the actual text to insert in
2588 SUBSTED while adjusting multibyteness to that of the current
2591 for (pos_byte
= 0, pos
= 0; pos_byte
< length
;)
2593 unsigned char str
[MAX_MULTIBYTE_LENGTH
];
2594 const unsigned char *add_stuff
= NULL
;
2595 ptrdiff_t add_len
= 0;
2600 FETCH_STRING_CHAR_ADVANCE_NO_CHECK (c
, newtext
, pos
, pos_byte
);
2602 c
= CHAR_TO_BYTE8 (c
);
2606 /* Note that we don't have to increment POS. */
2607 c
= SREF (newtext
, pos_byte
++);
2609 MAKE_CHAR_MULTIBYTE (c
);
2612 /* Either set ADD_STUFF and ADD_LEN to the text to put in SUBSTED,
2613 or set IDX to a match index, which means put that part
2614 of the buffer text into SUBSTED. */
2622 FETCH_STRING_CHAR_ADVANCE_NO_CHECK (c
, newtext
,
2624 if (!buf_multibyte
&& !ASCII_CHAR_P (c
))
2625 c
= CHAR_TO_BYTE8 (c
);
2629 c
= SREF (newtext
, pos_byte
++);
2631 MAKE_CHAR_MULTIBYTE (c
);
2636 else if (c
>= '1' && c
<= '9' && c
- '0' < search_regs
.num_regs
)
2638 if (search_regs
.start
[c
- '0'] >= 1)
2642 add_len
= 1, add_stuff
= (unsigned char *) "\\";
2646 error ("Invalid use of `\\' in replacement text");
2651 add_len
= CHAR_STRING (c
, str
);
2655 /* If we want to copy part of a previous match,
2656 set up ADD_STUFF and ADD_LEN to point to it. */
2659 ptrdiff_t begbyte
= CHAR_TO_BYTE (search_regs
.start
[idx
]);
2660 add_len
= CHAR_TO_BYTE (search_regs
.end
[idx
]) - begbyte
;
2661 if (search_regs
.start
[idx
] < GPT
&& GPT
< search_regs
.end
[idx
])
2662 move_gap_both (search_regs
.start
[idx
], begbyte
);
2663 add_stuff
= BYTE_POS_ADDR (begbyte
);
2666 /* Now the stuff we want to add to SUBSTED
2667 is invariably ADD_LEN bytes starting at ADD_STUFF. */
2669 /* Make sure SUBSTED is big enough. */
2670 if (substed_alloc_size
- substed_len
< add_len
)
2672 xpalloc (substed
, &substed_alloc_size
,
2673 add_len
- (substed_alloc_size
- substed_len
),
2674 STRING_BYTES_BOUND
, 1);
2676 /* Now add to the end of SUBSTED. */
2679 memcpy (substed
+ substed_len
, add_stuff
, add_len
);
2680 substed_len
+= add_len
;
2685 newtext
= make_specified_string ((const char *) substed
, -1,
2686 substed_len
, buf_multibyte
);
2690 /* Replace the old text with the new in the cleanest possible way. */
2691 replace_range (search_regs
.start
[sub
], search_regs
.end
[sub
],
2693 newpoint
= search_regs
.start
[sub
] + SCHARS (newtext
);
2695 if (case_action
== all_caps
)
2696 Fupcase_region (make_number (search_regs
.start
[sub
]),
2697 make_number (newpoint
));
2698 else if (case_action
== cap_initial
)
2699 Fupcase_initials_region (make_number (search_regs
.start
[sub
]),
2700 make_number (newpoint
));
2702 /* Adjust search data for this change. */
2704 ptrdiff_t oldend
= search_regs
.end
[sub
];
2705 ptrdiff_t oldstart
= search_regs
.start
[sub
];
2706 ptrdiff_t change
= newpoint
- search_regs
.end
[sub
];
2709 for (i
= 0; i
< search_regs
.num_regs
; i
++)
2711 if (search_regs
.start
[i
] >= oldend
)
2712 search_regs
.start
[i
] += change
;
2713 else if (search_regs
.start
[i
] > oldstart
)
2714 search_regs
.start
[i
] = oldstart
;
2715 if (search_regs
.end
[i
] >= oldend
)
2716 search_regs
.end
[i
] += change
;
2717 else if (search_regs
.end
[i
] > oldstart
)
2718 search_regs
.end
[i
] = oldstart
;
2722 /* Put point back where it was in the text. */
2724 TEMP_SET_PT (opoint
+ ZV
);
2726 TEMP_SET_PT (opoint
);
2728 /* Now move point "officially" to the start of the inserted replacement. */
2729 move_if_not_intangible (newpoint
);
2735 match_limit (Lisp_Object num
, bool beginningp
)
2742 args_out_of_range (num
, make_number (0));
2743 if (search_regs
.num_regs
<= 0)
2744 error ("No match data, because no search succeeded");
2745 if (n
>= search_regs
.num_regs
2746 || search_regs
.start
[n
] < 0)
2748 return (make_number ((beginningp
) ? search_regs
.start
[n
]
2749 : search_regs
.end
[n
]));
2752 DEFUN ("match-beginning", Fmatch_beginning
, Smatch_beginning
, 1, 1, 0,
2753 doc
: /* Return position of start of text matched by last search.
2754 SUBEXP, a number, specifies which parenthesized expression in the last
2756 Value is nil if SUBEXPth pair didn't match, or there were less than
2758 Zero means the entire text matched by the whole regexp or whole string. */)
2759 (Lisp_Object subexp
)
2761 return match_limit (subexp
, 1);
2764 DEFUN ("match-end", Fmatch_end
, Smatch_end
, 1, 1, 0,
2765 doc
: /* Return position of end of text matched by last search.
2766 SUBEXP, a number, specifies which parenthesized expression in the last
2768 Value is nil if SUBEXPth pair didn't match, or there were less than
2770 Zero means the entire text matched by the whole regexp or whole string. */)
2771 (Lisp_Object subexp
)
2773 return match_limit (subexp
, 0);
2776 DEFUN ("match-data", Fmatch_data
, Smatch_data
, 0, 3, 0,
2777 doc
: /* Return a list containing all info on what the last search matched.
2778 Element 2N is `(match-beginning N)'; element 2N + 1 is `(match-end N)'.
2779 All the elements are markers or nil (nil if the Nth pair didn't match)
2780 if the last match was on a buffer; integers or nil if a string was matched.
2781 Use `set-match-data' to reinstate the data in this list.
2783 If INTEGERS (the optional first argument) is non-nil, always use
2784 integers \(rather than markers) to represent buffer positions. In
2785 this case, and if the last match was in a buffer, the buffer will get
2786 stored as one additional element at the end of the list.
2788 If REUSE is a list, reuse it as part of the value. If REUSE is long
2789 enough to hold all the values, and if INTEGERS is non-nil, no consing
2792 If optional third arg RESEAT is non-nil, any previous markers on the
2793 REUSE list will be modified to point to nowhere.
2795 Return value is undefined if the last search failed. */)
2796 (Lisp_Object integers
, Lisp_Object reuse
, Lisp_Object reseat
)
2798 Lisp_Object tail
, prev
;
2803 for (tail
= reuse
; CONSP (tail
); tail
= XCDR (tail
))
2804 if (MARKERP (XCAR (tail
)))
2806 unchain_marker (XMARKER (XCAR (tail
)));
2807 XSETCAR (tail
, Qnil
);
2810 if (NILP (last_thing_searched
))
2816 SAFE_NALLOCA (data
, 1, 2 * search_regs
.num_regs
+ 1);
2819 for (i
= 0; i
< search_regs
.num_regs
; i
++)
2821 ptrdiff_t start
= search_regs
.start
[i
];
2824 if (EQ (last_thing_searched
, Qt
)
2825 || ! NILP (integers
))
2827 XSETFASTINT (data
[2 * i
], start
);
2828 XSETFASTINT (data
[2 * i
+ 1], search_regs
.end
[i
]);
2830 else if (BUFFERP (last_thing_searched
))
2832 data
[2 * i
] = Fmake_marker ();
2833 Fset_marker (data
[2 * i
],
2834 make_number (start
),
2835 last_thing_searched
);
2836 data
[2 * i
+ 1] = Fmake_marker ();
2837 Fset_marker (data
[2 * i
+ 1],
2838 make_number (search_regs
.end
[i
]),
2839 last_thing_searched
);
2842 /* last_thing_searched must always be Qt, a buffer, or Qnil. */
2848 data
[2 * i
] = data
[2 * i
+ 1] = Qnil
;
2851 if (BUFFERP (last_thing_searched
) && !NILP (integers
))
2853 data
[len
] = last_thing_searched
;
2857 /* If REUSE is not usable, cons up the values and return them. */
2858 if (! CONSP (reuse
))
2859 reuse
= Flist (len
, data
);
2862 /* If REUSE is a list, store as many value elements as will fit
2863 into the elements of REUSE. */
2864 for (i
= 0, tail
= reuse
; CONSP (tail
);
2865 i
++, tail
= XCDR (tail
))
2868 XSETCAR (tail
, data
[i
]);
2870 XSETCAR (tail
, Qnil
);
2874 /* If we couldn't fit all value elements into REUSE,
2875 cons up the rest of them and add them to the end of REUSE. */
2877 XSETCDR (prev
, Flist (len
- i
, data
+ i
));
2884 /* We used to have an internal use variant of `reseat' described as:
2886 If RESEAT is `evaporate', put the markers back on the free list
2887 immediately. No other references to the markers must exist in this
2888 case, so it is used only internally on the unwind stack and
2889 save-match-data from Lisp.
2891 But it was ill-conceived: those supposedly-internal markers get exposed via
2892 the undo-list, so freeing them here is unsafe. */
2894 DEFUN ("set-match-data", Fset_match_data
, Sset_match_data
, 1, 2, 0,
2895 doc
: /* Set internal data on last search match from elements of LIST.
2896 LIST should have been created by calling `match-data' previously.
2898 If optional arg RESEAT is non-nil, make markers on LIST point nowhere. */)
2899 (register Lisp_Object list
, Lisp_Object reseat
)
2902 register Lisp_Object marker
;
2904 if (running_asynch_code
)
2905 save_search_regs ();
2909 /* Unless we find a marker with a buffer or an explicit buffer
2910 in LIST, assume that this match data came from a string. */
2911 last_thing_searched
= Qt
;
2913 /* Allocate registers if they don't already exist. */
2915 EMACS_INT length
= XFASTINT (Flength (list
)) / 2;
2917 if (length
> search_regs
.num_regs
)
2919 ptrdiff_t num_regs
= search_regs
.num_regs
;
2920 if (PTRDIFF_MAX
< length
)
2921 memory_full (SIZE_MAX
);
2923 xpalloc (search_regs
.start
, &num_regs
, length
- num_regs
,
2924 min (PTRDIFF_MAX
, UINT_MAX
), sizeof (regoff_t
));
2926 xrealloc (search_regs
.end
, num_regs
* sizeof (regoff_t
));
2928 for (i
= search_regs
.num_regs
; i
< num_regs
; i
++)
2929 search_regs
.start
[i
] = -1;
2931 search_regs
.num_regs
= num_regs
;
2934 for (i
= 0; CONSP (list
); i
++)
2936 marker
= XCAR (list
);
2937 if (BUFFERP (marker
))
2939 last_thing_searched
= marker
;
2946 search_regs
.start
[i
] = -1;
2955 if (MARKERP (marker
))
2957 if (XMARKER (marker
)->buffer
== 0)
2958 XSETFASTINT (marker
, 0);
2960 XSETBUFFER (last_thing_searched
, XMARKER (marker
)->buffer
);
2963 CHECK_NUMBER_COERCE_MARKER (marker
);
2966 if (!NILP (reseat
) && MARKERP (m
))
2968 unchain_marker (XMARKER (m
));
2969 XSETCAR (list
, Qnil
);
2972 if ((list
= XCDR (list
), !CONSP (list
)))
2975 m
= marker
= XCAR (list
);
2977 if (MARKERP (marker
) && XMARKER (marker
)->buffer
== 0)
2978 XSETFASTINT (marker
, 0);
2980 CHECK_NUMBER_COERCE_MARKER (marker
);
2981 if ((XINT (from
) < 0
2982 ? TYPE_MINIMUM (regoff_t
) <= XINT (from
)
2983 : XINT (from
) <= TYPE_MAXIMUM (regoff_t
))
2984 && (XINT (marker
) < 0
2985 ? TYPE_MINIMUM (regoff_t
) <= XINT (marker
)
2986 : XINT (marker
) <= TYPE_MAXIMUM (regoff_t
)))
2988 search_regs
.start
[i
] = XINT (from
);
2989 search_regs
.end
[i
] = XINT (marker
);
2993 search_regs
.start
[i
] = -1;
2996 if (!NILP (reseat
) && MARKERP (m
))
2998 unchain_marker (XMARKER (m
));
2999 XSETCAR (list
, Qnil
);
3005 for (; i
< search_regs
.num_regs
; i
++)
3006 search_regs
.start
[i
] = -1;
3012 /* If true the match data have been saved in saved_search_regs
3013 during the execution of a sentinel or filter. */
3014 static bool search_regs_saved
;
3015 static struct re_registers saved_search_regs
;
3016 static Lisp_Object saved_last_thing_searched
;
3018 /* Called from Flooking_at, Fstring_match, search_buffer, Fstore_match_data
3019 if asynchronous code (filter or sentinel) is running. */
3021 save_search_regs (void)
3023 if (!search_regs_saved
)
3025 saved_search_regs
.num_regs
= search_regs
.num_regs
;
3026 saved_search_regs
.start
= search_regs
.start
;
3027 saved_search_regs
.end
= search_regs
.end
;
3028 saved_last_thing_searched
= last_thing_searched
;
3029 last_thing_searched
= Qnil
;
3030 search_regs
.num_regs
= 0;
3031 search_regs
.start
= 0;
3032 search_regs
.end
= 0;
3034 search_regs_saved
= 1;
3038 /* Called upon exit from filters and sentinels. */
3040 restore_search_regs (void)
3042 if (search_regs_saved
)
3044 if (search_regs
.num_regs
> 0)
3046 xfree (search_regs
.start
);
3047 xfree (search_regs
.end
);
3049 search_regs
.num_regs
= saved_search_regs
.num_regs
;
3050 search_regs
.start
= saved_search_regs
.start
;
3051 search_regs
.end
= saved_search_regs
.end
;
3052 last_thing_searched
= saved_last_thing_searched
;
3053 saved_last_thing_searched
= Qnil
;
3054 search_regs_saved
= 0;
3059 unwind_set_match_data (Lisp_Object list
)
3061 /* It is NOT ALWAYS safe to free (evaporate) the markers immediately. */
3062 Fset_match_data (list
, Qt
);
3065 /* Called to unwind protect the match data. */
3067 record_unwind_save_match_data (void)
3069 record_unwind_protect (unwind_set_match_data
,
3070 Fmatch_data (Qnil
, Qnil
, Qnil
));
3073 /* Quote a string to deactivate reg-expr chars */
3075 DEFUN ("regexp-quote", Fregexp_quote
, Sregexp_quote
, 1, 1, 0,
3076 doc
: /* Return a regexp string which matches exactly STRING and nothing else. */)
3077 (Lisp_Object string
)
3079 char *in
, *out
, *end
;
3081 ptrdiff_t backslashes_added
= 0;
3083 CHECK_STRING (string
);
3086 SAFE_NALLOCA (temp
, 2, SBYTES (string
));
3088 /* Now copy the data into the new string, inserting escapes. */
3090 in
= SSDATA (string
);
3091 end
= in
+ SBYTES (string
);
3094 for (; in
!= end
; in
++)
3097 || *in
== '*' || *in
== '.' || *in
== '\\'
3098 || *in
== '?' || *in
== '+'
3099 || *in
== '^' || *in
== '$')
3100 *out
++ = '\\', backslashes_added
++;
3105 = make_specified_string (temp
,
3106 SCHARS (string
) + backslashes_added
,
3108 STRING_MULTIBYTE (string
));
3113 /* Like find_newline, but doesn't use the cache, and only searches forward. */
3115 find_newline1 (ptrdiff_t start
, ptrdiff_t start_byte
, ptrdiff_t end
,
3116 ptrdiff_t end_byte
, ptrdiff_t count
, ptrdiff_t *shortage
,
3117 ptrdiff_t *bytepos
, bool allow_quit
)
3122 end
= ZV
, end_byte
= ZV_BYTE
;
3127 end
= BEGV
, end_byte
= BEGV_BYTE
;
3130 end_byte
= CHAR_TO_BYTE (end
);
3135 immediate_quit
= allow_quit
;
3138 while (start
!= end
)
3140 /* Our innermost scanning loop is very simple; it doesn't know
3141 about gaps, buffer ends, or the newline cache. ceiling is
3142 the position of the last character before the next such
3143 obstacle --- the last character the dumb search loop should
3145 ptrdiff_t tem
, ceiling_byte
= end_byte
- 1;
3147 if (start_byte
== -1)
3148 start_byte
= CHAR_TO_BYTE (start
);
3150 /* The dumb loop can only scan text stored in contiguous
3151 bytes. BUFFER_CEILING_OF returns the last character
3152 position that is contiguous, so the ceiling is the
3153 position after that. */
3154 tem
= BUFFER_CEILING_OF (start_byte
);
3155 ceiling_byte
= min (tem
, ceiling_byte
);
3158 /* The termination address of the dumb loop. */
3159 unsigned char *lim_addr
= BYTE_POS_ADDR (ceiling_byte
) + 1;
3160 ptrdiff_t lim_byte
= ceiling_byte
+ 1;
3162 /* Nonpositive offsets (relative to LIM_ADDR and LIM_BYTE)
3163 of the base, the cursor, and the next line. */
3164 ptrdiff_t base
= start_byte
- lim_byte
;
3165 ptrdiff_t cursor
, next
;
3167 for (cursor
= base
; cursor
< 0; cursor
= next
)
3169 /* The dumb loop. */
3170 unsigned char *nl
= memchr (lim_addr
+ cursor
, '\n', - cursor
);
3171 next
= nl
? nl
- lim_addr
: 0;
3181 *bytepos
= lim_byte
+ next
;
3182 return BYTE_TO_CHAR (lim_byte
+ next
);
3186 start_byte
= lim_byte
;
3187 start
= BYTE_TO_CHAR (start_byte
);
3196 *bytepos
= start_byte
== -1 ? CHAR_TO_BYTE (start
) : start_byte
;
3197 eassert (*bytepos
== CHAR_TO_BYTE (start
));
3202 DEFUN ("newline-cache-check", Fnewline_cache_check
, Snewline_cache_check
,
3204 doc
: /* Check the newline cache of BUFFER against buffer contents.
3206 BUFFER defaults to the current buffer.
3208 Value is an array of 2 sub-arrays of buffer positions for newlines,
3209 the first based on the cache, the second based on actually scanning
3210 the buffer. If the buffer doesn't have a cache, the value is nil. */)
3211 (Lisp_Object buffer
)
3213 struct buffer
*buf
, *old
= NULL
;
3214 ptrdiff_t shortage
, nl_count_cache
, nl_count_buf
;
3215 Lisp_Object cache_newlines
, buf_newlines
, val
;
3216 ptrdiff_t from
, found
, i
;
3219 buf
= current_buffer
;
3222 CHECK_BUFFER (buffer
);
3223 buf
= XBUFFER (buffer
);
3224 old
= current_buffer
;
3226 if (buf
->base_buffer
)
3227 buf
= buf
->base_buffer
;
3229 /* If the buffer doesn't have a newline cache, return nil. */
3230 if (NILP (BVAR (buf
, cache_long_scans
))
3231 || buf
->newline_cache
== NULL
)
3234 /* find_newline can only work on the current buffer. */
3236 set_buffer_internal_1 (buf
);
3238 /* How many newlines are there according to the cache? */
3239 find_newline (BEGV
, BEGV_BYTE
, ZV
, ZV_BYTE
,
3240 TYPE_MAXIMUM (ptrdiff_t), &shortage
, NULL
, true);
3241 nl_count_cache
= TYPE_MAXIMUM (ptrdiff_t) - shortage
;
3243 /* Create vector and populate it. */
3244 cache_newlines
= make_uninit_vector (nl_count_cache
);
3248 for (from
= BEGV
, found
= from
, i
= 0; from
< ZV
; from
= found
, i
++)
3250 ptrdiff_t from_byte
= CHAR_TO_BYTE (from
);
3252 found
= find_newline (from
, from_byte
, 0, -1, 1, &shortage
,
3254 if (shortage
!= 0 || i
>= nl_count_cache
)
3256 ASET (cache_newlines
, i
, make_number (found
- 1));
3258 /* Fill the rest of slots with an invalid position. */
3259 for ( ; i
< nl_count_cache
; i
++)
3260 ASET (cache_newlines
, i
, make_number (-1));
3263 /* Now do the same, but without using the cache. */
3264 find_newline1 (BEGV
, BEGV_BYTE
, ZV
, ZV_BYTE
,
3265 TYPE_MAXIMUM (ptrdiff_t), &shortage
, NULL
, true);
3266 nl_count_buf
= TYPE_MAXIMUM (ptrdiff_t) - shortage
;
3267 buf_newlines
= make_uninit_vector (nl_count_buf
);
3270 for (from
= BEGV
, found
= from
, i
= 0; from
< ZV
; from
= found
, i
++)
3272 ptrdiff_t from_byte
= CHAR_TO_BYTE (from
);
3274 found
= find_newline1 (from
, from_byte
, 0, -1, 1, &shortage
,
3276 if (shortage
!= 0 || i
>= nl_count_buf
)
3278 ASET (buf_newlines
, i
, make_number (found
- 1));
3280 for ( ; i
< nl_count_buf
; i
++)
3281 ASET (buf_newlines
, i
, make_number (-1));
3284 /* Construct the value and return it. */
3285 val
= make_uninit_vector (2);
3286 ASET (val
, 0, cache_newlines
);
3287 ASET (val
, 1, buf_newlines
);
3290 set_buffer_internal_1 (old
);
3295 syms_of_search (void)
3299 for (i
= 0; i
< REGEXP_CACHE_SIZE
; ++i
)
3301 searchbufs
[i
].buf
.allocated
= 100;
3302 searchbufs
[i
].buf
.buffer
= xmalloc (100);
3303 searchbufs
[i
].buf
.fastmap
= searchbufs
[i
].fastmap
;
3304 searchbufs
[i
].regexp
= Qnil
;
3305 searchbufs
[i
].whitespace_regexp
= Qnil
;
3306 searchbufs
[i
].syntax_table
= Qnil
;
3307 staticpro (&searchbufs
[i
].regexp
);
3308 staticpro (&searchbufs
[i
].whitespace_regexp
);
3309 staticpro (&searchbufs
[i
].syntax_table
);
3310 searchbufs
[i
].next
= (i
== REGEXP_CACHE_SIZE
-1 ? 0 : &searchbufs
[i
+1]);
3312 searchbuf_head
= &searchbufs
[0];
3314 DEFSYM (Qsearch_failed
, "search-failed");
3315 DEFSYM (Qinvalid_regexp
, "invalid-regexp");
3317 Fput (Qsearch_failed
, Qerror_conditions
,
3318 listn (CONSTYPE_PURE
, 2, Qsearch_failed
, Qerror
));
3319 Fput (Qsearch_failed
, Qerror_message
,
3320 build_pure_c_string ("Search failed"));
3322 Fput (Qinvalid_regexp
, Qerror_conditions
,
3323 listn (CONSTYPE_PURE
, 2, Qinvalid_regexp
, Qerror
));
3324 Fput (Qinvalid_regexp
, Qerror_message
,
3325 build_pure_c_string ("Invalid regexp"));
3327 last_thing_searched
= Qnil
;
3328 staticpro (&last_thing_searched
);
3330 saved_last_thing_searched
= Qnil
;
3331 staticpro (&saved_last_thing_searched
);
3333 DEFVAR_LISP ("search-spaces-regexp", Vsearch_spaces_regexp
,
3334 doc
: /* Regexp to substitute for bunches of spaces in regexp search.
3335 Some commands use this for user-specified regexps.
3336 Spaces that occur inside character classes or repetition operators
3337 or other such regexp constructs are not replaced with this.
3338 A value of nil (which is the normal value) means treat spaces literally. */);
3339 Vsearch_spaces_regexp
= Qnil
;
3341 DEFVAR_LISP ("inhibit-changing-match-data", Vinhibit_changing_match_data
,
3342 doc
: /* Internal use only.
3343 If non-nil, the primitive searching and matching functions
3344 such as `looking-at', `string-match', `re-search-forward', etc.,
3345 do not set the match data. The proper way to use this variable
3346 is to bind it with `let' around a small expression. */);
3347 Vinhibit_changing_match_data
= Qnil
;
3349 defsubr (&Slooking_at
);
3350 defsubr (&Sposix_looking_at
);
3351 defsubr (&Sstring_match
);
3352 defsubr (&Sposix_string_match
);
3353 defsubr (&Ssearch_forward
);
3354 defsubr (&Ssearch_backward
);
3355 defsubr (&Sre_search_forward
);
3356 defsubr (&Sre_search_backward
);
3357 defsubr (&Sposix_search_forward
);
3358 defsubr (&Sposix_search_backward
);
3359 defsubr (&Sreplace_match
);
3360 defsubr (&Smatch_beginning
);
3361 defsubr (&Smatch_end
);
3362 defsubr (&Smatch_data
);
3363 defsubr (&Sset_match_data
);
3364 defsubr (&Sregexp_quote
);
3365 defsubr (&Snewline_cache_check
);