1 /* String search routines for GNU Emacs.
2 Copyright (C) 1985, 1986, 1987, 1993, 1994, 1997, 1998, 1999, 2002, 2003,
3 2004, 2005, 2006 Free Software Foundation, Inc.
5 This file is part of GNU Emacs.
7 GNU Emacs is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 2, or (at your option)
12 GNU Emacs is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with GNU Emacs; see the file COPYING. If not, write to
19 the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor,
20 Boston, MA 02110-1301, USA. */
28 #include "character.h"
29 #include "region-cache.h"
31 #include "blockinput.h"
32 #include "intervals.h"
34 #include <sys/types.h>
37 #define REGEXP_CACHE_SIZE 20
39 /* If the regexp is non-nil, then the buffer contains the compiled form
40 of that regexp, suitable for searching. */
43 struct regexp_cache
*next
;
44 Lisp_Object regexp
, whitespace_regexp
;
45 struct re_pattern_buffer buf
;
47 /* Nonzero means regexp was compiled to do full POSIX backtracking. */
51 /* The instances of that struct. */
52 struct regexp_cache searchbufs
[REGEXP_CACHE_SIZE
];
54 /* The head of the linked list; points to the most recently used buffer. */
55 struct regexp_cache
*searchbuf_head
;
58 /* Every call to re_match, etc., must pass &search_regs as the regs
59 argument unless you can show it is unnecessary (i.e., if re_match
60 is certainly going to be called again before region-around-match
63 Since the registers are now dynamically allocated, we need to make
64 sure not to refer to the Nth register before checking that it has
65 been allocated by checking search_regs.num_regs.
67 The regex code keeps track of whether it has allocated the search
68 buffer using bits in the re_pattern_buffer. This means that whenever
69 you compile a new pattern, it completely forgets whether it has
70 allocated any registers, and will allocate new registers the next
71 time you call a searching or matching function. Therefore, we need
72 to call re_set_registers after compiling a new pattern or after
73 setting the match registers, so that the regex functions will be
74 able to free or re-allocate it properly. */
75 static struct re_registers search_regs
;
77 /* The buffer in which the last search was performed, or
78 Qt if the last search was done in a string;
79 Qnil if no searching has been done yet. */
80 static Lisp_Object last_thing_searched
;
82 /* error condition signaled when regexp compile_pattern fails */
84 Lisp_Object Qinvalid_regexp
;
86 Lisp_Object Vsearch_spaces_regexp
;
88 static void set_search_regs ();
89 static void save_search_regs ();
90 static int simple_search ();
91 static int boyer_moore ();
92 static int search_buffer ();
97 error ("Stack overflow in regexp matcher");
100 /* Compile a regexp and signal a Lisp error if anything goes wrong.
101 PATTERN is the pattern to compile.
102 CP is the place to put the result.
103 TRANSLATE is a translation table for ignoring case, or nil for none.
104 REGP is the structure that says where to store the "register"
105 values that will result from matching this pattern.
106 If it is 0, we should compile the pattern not to record any
107 subexpression bounds.
108 POSIX is nonzero if we want full backtracking (POSIX style)
109 for this pattern. 0 means backtrack only enough to get a valid match.
110 MULTIBYTE is nonzero iff a target of match is a multibyte buffer or
113 The behavior also depends on Vsearch_spaces_regexp. */
116 compile_pattern_1 (cp
, pattern
, translate
, regp
, posix
, multibyte
)
117 struct regexp_cache
*cp
;
119 Lisp_Object translate
;
120 struct re_registers
*regp
;
128 cp
->buf
.translate
= (! NILP (translate
) ? translate
: make_number (0));
130 cp
->buf
.multibyte
= STRING_MULTIBYTE (pattern
);
131 cp
->buf
.target_multibyte
= multibyte
;
132 cp
->whitespace_regexp
= Vsearch_spaces_regexp
;
134 old
= re_set_syntax (RE_SYNTAX_EMACS
135 | (posix
? 0 : RE_NO_POSIX_BACKTRACKING
));
136 re_set_whitespace_regexp (NILP (Vsearch_spaces_regexp
) ? NULL
137 : SDATA (Vsearch_spaces_regexp
));
139 val
= (char *) re_compile_pattern ((char *) SDATA (pattern
),
140 SBYTES (pattern
), &cp
->buf
);
142 re_set_whitespace_regexp (NULL
);
147 Fsignal (Qinvalid_regexp
, Fcons (build_string (val
), Qnil
));
149 cp
->regexp
= Fcopy_sequence (pattern
);
152 /* Shrink each compiled regexp buffer in the cache
153 to the size actually used right now.
154 This is called from garbage collection. */
157 shrink_regexp_cache ()
159 struct regexp_cache
*cp
;
161 for (cp
= searchbuf_head
; cp
!= 0; cp
= cp
->next
)
163 cp
->buf
.allocated
= cp
->buf
.used
;
165 = (unsigned char *) xrealloc (cp
->buf
.buffer
, cp
->buf
.used
);
169 /* Compile a regexp if necessary, but first check to see if there's one in
171 PATTERN is the pattern to compile.
172 TRANSLATE is a translation table for ignoring case, or nil for none.
173 REGP is the structure that says where to store the "register"
174 values that will result from matching this pattern.
175 If it is 0, we should compile the pattern not to record any
176 subexpression bounds.
177 POSIX is nonzero if we want full backtracking (POSIX style)
178 for this pattern. 0 means backtrack only enough to get a valid match. */
180 struct re_pattern_buffer
*
181 compile_pattern (pattern
, regp
, translate
, posix
, multibyte
)
183 struct re_registers
*regp
;
184 Lisp_Object translate
;
185 int posix
, multibyte
;
187 struct regexp_cache
*cp
, **cpp
;
189 for (cpp
= &searchbuf_head
; ; cpp
= &cp
->next
)
192 /* Entries are initialized to nil, and may be set to nil by
193 compile_pattern_1 if the pattern isn't valid. Don't apply
194 string accessors in those cases. However, compile_pattern_1
195 is only applied to the cache entry we pick here to reuse. So
196 nil should never appear before a non-nil entry. */
197 if (NILP (cp
->regexp
))
199 if (SCHARS (cp
->regexp
) == SCHARS (pattern
)
200 && STRING_MULTIBYTE (cp
->regexp
) == STRING_MULTIBYTE (pattern
)
201 && !NILP (Fstring_equal (cp
->regexp
, pattern
))
202 && EQ (cp
->buf
.translate
, (! NILP (translate
) ? translate
: make_number (0)))
203 && cp
->posix
== posix
204 && cp
->buf
.target_multibyte
== multibyte
205 && !NILP (Fequal (cp
->whitespace_regexp
, Vsearch_spaces_regexp
)))
208 /* If we're at the end of the cache, compile into the nil cell
209 we found, or the last (least recently used) cell with a
214 compile_pattern_1 (cp
, pattern
, translate
, regp
, posix
, multibyte
);
219 /* When we get here, cp (aka *cpp) contains the compiled pattern,
220 either because we found it in the cache or because we just compiled it.
221 Move it to the front of the queue to mark it as most recently used. */
223 cp
->next
= searchbuf_head
;
226 /* Advise the searching functions about the space we have allocated
227 for register data. */
229 re_set_registers (&cp
->buf
, regp
, regp
->num_regs
, regp
->start
, regp
->end
);
234 /* Error condition used for failing searches */
235 Lisp_Object Qsearch_failed
;
241 Fsignal (Qsearch_failed
, Fcons (arg
, Qnil
));
246 looking_at_1 (string
, posix
)
251 unsigned char *p1
, *p2
;
254 struct re_pattern_buffer
*bufp
;
256 if (running_asynch_code
)
259 CHECK_STRING (string
);
260 bufp
= compile_pattern (string
, &search_regs
,
261 (!NILP (current_buffer
->case_fold_search
)
262 ? current_buffer
->case_canon_table
: Qnil
),
264 !NILP (current_buffer
->enable_multibyte_characters
));
267 QUIT
; /* Do a pending quit right away, to avoid paradoxical behavior */
269 /* Get pointers and sizes of the two strings
270 that make up the visible portion of the buffer. */
273 s1
= GPT_BYTE
- BEGV_BYTE
;
275 s2
= ZV_BYTE
- GPT_BYTE
;
279 s2
= ZV_BYTE
- BEGV_BYTE
;
284 s1
= ZV_BYTE
- BEGV_BYTE
;
288 re_match_object
= Qnil
;
290 i
= re_match_2 (bufp
, (char *) p1
, s1
, (char *) p2
, s2
,
291 PT_BYTE
- BEGV_BYTE
, &search_regs
,
292 ZV_BYTE
- BEGV_BYTE
);
298 val
= (0 <= i
? Qt
: Qnil
);
300 for (i
= 0; i
< search_regs
.num_regs
; i
++)
301 if (search_regs
.start
[i
] >= 0)
304 = BYTE_TO_CHAR (search_regs
.start
[i
] + BEGV_BYTE
);
306 = BYTE_TO_CHAR (search_regs
.end
[i
] + BEGV_BYTE
);
308 XSETBUFFER (last_thing_searched
, current_buffer
);
312 DEFUN ("looking-at", Flooking_at
, Slooking_at
, 1, 1, 0,
313 doc
: /* Return t if text after point matches regular expression REGEXP.
314 This function modifies the match data that `match-beginning',
315 `match-end' and `match-data' access; save and restore the match
316 data if you want to preserve them. */)
320 return looking_at_1 (regexp
, 0);
323 DEFUN ("posix-looking-at", Fposix_looking_at
, Sposix_looking_at
, 1, 1, 0,
324 doc
: /* Return t if text after point matches regular expression REGEXP.
325 Find the longest match, in accord with Posix regular expression rules.
326 This function modifies the match data that `match-beginning',
327 `match-end' and `match-data' access; save and restore the match
328 data if you want to preserve them. */)
332 return looking_at_1 (regexp
, 1);
336 string_match_1 (regexp
, string
, start
, posix
)
337 Lisp_Object regexp
, string
, start
;
341 struct re_pattern_buffer
*bufp
;
345 if (running_asynch_code
)
348 CHECK_STRING (regexp
);
349 CHECK_STRING (string
);
352 pos
= 0, pos_byte
= 0;
355 int len
= SCHARS (string
);
357 CHECK_NUMBER (start
);
359 if (pos
< 0 && -pos
<= len
)
361 else if (0 > pos
|| pos
> len
)
362 args_out_of_range (string
, start
);
363 pos_byte
= string_char_to_byte (string
, pos
);
366 bufp
= compile_pattern (regexp
, &search_regs
,
367 (!NILP (current_buffer
->case_fold_search
)
368 ? current_buffer
->case_canon_table
: Qnil
),
370 STRING_MULTIBYTE (string
));
372 re_match_object
= string
;
374 val
= re_search (bufp
, (char *) SDATA (string
),
375 SBYTES (string
), pos_byte
,
376 SBYTES (string
) - pos_byte
,
379 last_thing_searched
= Qt
;
382 if (val
< 0) return Qnil
;
384 for (i
= 0; i
< search_regs
.num_regs
; i
++)
385 if (search_regs
.start
[i
] >= 0)
388 = string_byte_to_char (string
, search_regs
.start
[i
]);
390 = string_byte_to_char (string
, search_regs
.end
[i
]);
393 return make_number (string_byte_to_char (string
, val
));
396 DEFUN ("string-match", Fstring_match
, Sstring_match
, 2, 3, 0,
397 doc
: /* Return index of start of first match for REGEXP in STRING, or nil.
398 Matching ignores case if `case-fold-search' is non-nil.
399 If third arg START is non-nil, start search at that index in STRING.
400 For index of first char beyond the match, do (match-end 0).
401 `match-end' and `match-beginning' also give indices of substrings
402 matched by parenthesis constructs in the pattern.
404 You can use the function `match-string' to extract the substrings
405 matched by the parenthesis constructions in REGEXP. */)
406 (regexp
, string
, start
)
407 Lisp_Object regexp
, string
, start
;
409 return string_match_1 (regexp
, string
, start
, 0);
412 DEFUN ("posix-string-match", Fposix_string_match
, Sposix_string_match
, 2, 3, 0,
413 doc
: /* Return index of start of first match for REGEXP in STRING, or nil.
414 Find the longest match, in accord with Posix regular expression rules.
415 Case is ignored if `case-fold-search' is non-nil in the current buffer.
416 If third arg START is non-nil, start search at that index in STRING.
417 For index of first char beyond the match, do (match-end 0).
418 `match-end' and `match-beginning' also give indices of substrings
419 matched by parenthesis constructs in the pattern. */)
420 (regexp
, string
, start
)
421 Lisp_Object regexp
, string
, start
;
423 return string_match_1 (regexp
, string
, start
, 1);
426 /* Match REGEXP against STRING, searching all of STRING,
427 and return the index of the match, or negative on failure.
428 This does not clobber the match data. */
431 fast_string_match (regexp
, string
)
432 Lisp_Object regexp
, string
;
435 struct re_pattern_buffer
*bufp
;
437 bufp
= compile_pattern (regexp
, 0, Qnil
,
438 0, STRING_MULTIBYTE (string
));
440 re_match_object
= string
;
442 val
= re_search (bufp
, (char *) SDATA (string
),
449 /* Match REGEXP against STRING, searching all of STRING ignoring case,
450 and return the index of the match, or negative on failure.
451 This does not clobber the match data.
452 We assume that STRING contains single-byte characters. */
454 extern Lisp_Object Vascii_downcase_table
;
457 fast_c_string_match_ignore_case (regexp
, string
)
462 struct re_pattern_buffer
*bufp
;
463 int len
= strlen (string
);
465 regexp
= string_make_unibyte (regexp
);
466 re_match_object
= Qt
;
467 bufp
= compile_pattern (regexp
, 0,
468 Vascii_canon_table
, 0,
471 val
= re_search (bufp
, string
, len
, 0, len
, 0);
476 /* Like fast_string_match but ignore case. */
479 fast_string_match_ignore_case (regexp
, string
)
480 Lisp_Object regexp
, string
;
483 struct re_pattern_buffer
*bufp
;
485 bufp
= compile_pattern (regexp
, 0, Vascii_canon_table
,
486 0, STRING_MULTIBYTE (string
));
488 re_match_object
= string
;
490 val
= re_search (bufp
, (char *) SDATA (string
),
497 /* The newline cache: remembering which sections of text have no newlines. */
499 /* If the user has requested newline caching, make sure it's on.
500 Otherwise, make sure it's off.
501 This is our cheezy way of associating an action with the change of
502 state of a buffer-local variable. */
504 newline_cache_on_off (buf
)
507 if (NILP (buf
->cache_long_line_scans
))
509 /* It should be off. */
510 if (buf
->newline_cache
)
512 free_region_cache (buf
->newline_cache
);
513 buf
->newline_cache
= 0;
518 /* It should be on. */
519 if (buf
->newline_cache
== 0)
520 buf
->newline_cache
= new_region_cache ();
525 /* Search for COUNT instances of the character TARGET between START and END.
527 If COUNT is positive, search forwards; END must be >= START.
528 If COUNT is negative, search backwards for the -COUNTth instance;
529 END must be <= START.
530 If COUNT is zero, do anything you please; run rogue, for all I care.
532 If END is zero, use BEGV or ZV instead, as appropriate for the
533 direction indicated by COUNT.
535 If we find COUNT instances, set *SHORTAGE to zero, and return the
536 position past the COUNTth match. Note that for reverse motion
537 this is not the same as the usual convention for Emacs motion commands.
539 If we don't find COUNT instances before reaching END, set *SHORTAGE
540 to the number of TARGETs left unfound, and return END.
542 If ALLOW_QUIT is non-zero, set immediate_quit. That's good to do
543 except when inside redisplay. */
546 scan_buffer (target
, start
, end
, count
, shortage
, allow_quit
)
553 struct region_cache
*newline_cache
;
564 if (! end
) end
= BEGV
;
567 newline_cache_on_off (current_buffer
);
568 newline_cache
= current_buffer
->newline_cache
;
573 immediate_quit
= allow_quit
;
578 /* Our innermost scanning loop is very simple; it doesn't know
579 about gaps, buffer ends, or the newline cache. ceiling is
580 the position of the last character before the next such
581 obstacle --- the last character the dumb search loop should
583 int ceiling_byte
= CHAR_TO_BYTE (end
) - 1;
584 int start_byte
= CHAR_TO_BYTE (start
);
587 /* If we're looking for a newline, consult the newline cache
588 to see where we can avoid some scanning. */
589 if (target
== '\n' && newline_cache
)
593 while (region_cache_forward
594 (current_buffer
, newline_cache
, start_byte
, &next_change
))
595 start_byte
= next_change
;
596 immediate_quit
= allow_quit
;
598 /* START should never be after END. */
599 if (start_byte
> ceiling_byte
)
600 start_byte
= ceiling_byte
;
602 /* Now the text after start is an unknown region, and
603 next_change is the position of the next known region. */
604 ceiling_byte
= min (next_change
- 1, ceiling_byte
);
607 /* The dumb loop can only scan text stored in contiguous
608 bytes. BUFFER_CEILING_OF returns the last character
609 position that is contiguous, so the ceiling is the
610 position after that. */
611 tem
= BUFFER_CEILING_OF (start_byte
);
612 ceiling_byte
= min (tem
, ceiling_byte
);
615 /* The termination address of the dumb loop. */
616 register unsigned char *ceiling_addr
617 = BYTE_POS_ADDR (ceiling_byte
) + 1;
618 register unsigned char *cursor
619 = BYTE_POS_ADDR (start_byte
);
620 unsigned char *base
= cursor
;
622 while (cursor
< ceiling_addr
)
624 unsigned char *scan_start
= cursor
;
627 while (*cursor
!= target
&& ++cursor
< ceiling_addr
)
630 /* If we're looking for newlines, cache the fact that
631 the region from start to cursor is free of them. */
632 if (target
== '\n' && newline_cache
)
633 know_region_cache (current_buffer
, newline_cache
,
634 start_byte
+ scan_start
- base
,
635 start_byte
+ cursor
- base
);
637 /* Did we find the target character? */
638 if (cursor
< ceiling_addr
)
643 return BYTE_TO_CHAR (start_byte
+ cursor
- base
+ 1);
649 start
= BYTE_TO_CHAR (start_byte
+ cursor
- base
);
655 /* The last character to check before the next obstacle. */
656 int ceiling_byte
= CHAR_TO_BYTE (end
);
657 int start_byte
= CHAR_TO_BYTE (start
);
660 /* Consult the newline cache, if appropriate. */
661 if (target
== '\n' && newline_cache
)
665 while (region_cache_backward
666 (current_buffer
, newline_cache
, start_byte
, &next_change
))
667 start_byte
= next_change
;
668 immediate_quit
= allow_quit
;
670 /* Start should never be at or before end. */
671 if (start_byte
<= ceiling_byte
)
672 start_byte
= ceiling_byte
+ 1;
674 /* Now the text before start is an unknown region, and
675 next_change is the position of the next known region. */
676 ceiling_byte
= max (next_change
, ceiling_byte
);
679 /* Stop scanning before the gap. */
680 tem
= BUFFER_FLOOR_OF (start_byte
- 1);
681 ceiling_byte
= max (tem
, ceiling_byte
);
684 /* The termination address of the dumb loop. */
685 register unsigned char *ceiling_addr
= BYTE_POS_ADDR (ceiling_byte
);
686 register unsigned char *cursor
= BYTE_POS_ADDR (start_byte
- 1);
687 unsigned char *base
= cursor
;
689 while (cursor
>= ceiling_addr
)
691 unsigned char *scan_start
= cursor
;
693 while (*cursor
!= target
&& --cursor
>= ceiling_addr
)
696 /* If we're looking for newlines, cache the fact that
697 the region from after the cursor to start is free of them. */
698 if (target
== '\n' && newline_cache
)
699 know_region_cache (current_buffer
, newline_cache
,
700 start_byte
+ cursor
- base
,
701 start_byte
+ scan_start
- base
);
703 /* Did we find the target character? */
704 if (cursor
>= ceiling_addr
)
709 return BYTE_TO_CHAR (start_byte
+ cursor
- base
);
715 start
= BYTE_TO_CHAR (start_byte
+ cursor
- base
);
721 *shortage
= count
* direction
;
725 /* Search for COUNT instances of a line boundary, which means either a
726 newline or (if selective display enabled) a carriage return.
727 Start at START. If COUNT is negative, search backwards.
729 We report the resulting position by calling TEMP_SET_PT_BOTH.
731 If we find COUNT instances. we position after (always after,
732 even if scanning backwards) the COUNTth match, and return 0.
734 If we don't find COUNT instances before reaching the end of the
735 buffer (or the beginning, if scanning backwards), we return
736 the number of line boundaries left unfound, and position at
737 the limit we bumped up against.
739 If ALLOW_QUIT is non-zero, set immediate_quit. That's good to do
740 except in special cases. */
743 scan_newline (start
, start_byte
, limit
, limit_byte
, count
, allow_quit
)
744 int start
, start_byte
;
745 int limit
, limit_byte
;
749 int direction
= ((count
> 0) ? 1 : -1);
751 register unsigned char *cursor
;
754 register int ceiling
;
755 register unsigned char *ceiling_addr
;
757 int old_immediate_quit
= immediate_quit
;
759 /* The code that follows is like scan_buffer
760 but checks for either newline or carriage return. */
765 start_byte
= CHAR_TO_BYTE (start
);
769 while (start_byte
< limit_byte
)
771 ceiling
= BUFFER_CEILING_OF (start_byte
);
772 ceiling
= min (limit_byte
- 1, ceiling
);
773 ceiling_addr
= BYTE_POS_ADDR (ceiling
) + 1;
774 base
= (cursor
= BYTE_POS_ADDR (start_byte
));
777 while (*cursor
!= '\n' && ++cursor
!= ceiling_addr
)
780 if (cursor
!= ceiling_addr
)
784 immediate_quit
= old_immediate_quit
;
785 start_byte
= start_byte
+ cursor
- base
+ 1;
786 start
= BYTE_TO_CHAR (start_byte
);
787 TEMP_SET_PT_BOTH (start
, start_byte
);
791 if (++cursor
== ceiling_addr
)
797 start_byte
+= cursor
- base
;
802 while (start_byte
> limit_byte
)
804 ceiling
= BUFFER_FLOOR_OF (start_byte
- 1);
805 ceiling
= max (limit_byte
, ceiling
);
806 ceiling_addr
= BYTE_POS_ADDR (ceiling
) - 1;
807 base
= (cursor
= BYTE_POS_ADDR (start_byte
- 1) + 1);
810 while (--cursor
!= ceiling_addr
&& *cursor
!= '\n')
813 if (cursor
!= ceiling_addr
)
817 immediate_quit
= old_immediate_quit
;
818 /* Return the position AFTER the match we found. */
819 start_byte
= start_byte
+ cursor
- base
+ 1;
820 start
= BYTE_TO_CHAR (start_byte
);
821 TEMP_SET_PT_BOTH (start
, start_byte
);
828 /* Here we add 1 to compensate for the last decrement
829 of CURSOR, which took it past the valid range. */
830 start_byte
+= cursor
- base
+ 1;
834 TEMP_SET_PT_BOTH (limit
, limit_byte
);
835 immediate_quit
= old_immediate_quit
;
837 return count
* direction
;
841 find_next_newline_no_quit (from
, cnt
)
842 register int from
, cnt
;
844 return scan_buffer ('\n', from
, 0, cnt
, (int *) 0, 0);
847 /* Like find_next_newline, but returns position before the newline,
848 not after, and only search up to TO. This isn't just
849 find_next_newline (...)-1, because you might hit TO. */
852 find_before_next_newline (from
, to
, cnt
)
856 int pos
= scan_buffer ('\n', from
, to
, cnt
, &shortage
, 1);
864 /* Subroutines of Lisp buffer search functions. */
867 search_command (string
, bound
, noerror
, count
, direction
, RE
, posix
)
868 Lisp_Object string
, bound
, noerror
, count
;
879 CHECK_NUMBER (count
);
883 CHECK_STRING (string
);
887 lim
= ZV
, lim_byte
= ZV_BYTE
;
889 lim
= BEGV
, lim_byte
= BEGV_BYTE
;
893 CHECK_NUMBER_COERCE_MARKER (bound
);
895 if (n
> 0 ? lim
< PT
: lim
> PT
)
896 error ("Invalid search bound (wrong side of point)");
898 lim
= ZV
, lim_byte
= ZV_BYTE
;
900 lim
= BEGV
, lim_byte
= BEGV_BYTE
;
902 lim_byte
= CHAR_TO_BYTE (lim
);
905 np
= search_buffer (string
, PT
, PT_BYTE
, lim
, lim_byte
, n
, RE
,
906 (!NILP (current_buffer
->case_fold_search
)
907 ? current_buffer
->case_canon_table
909 (!NILP (current_buffer
->case_fold_search
)
910 ? current_buffer
->case_eqv_table
916 return signal_failure (string
);
917 if (!EQ (noerror
, Qt
))
919 if (lim
< BEGV
|| lim
> ZV
)
921 SET_PT_BOTH (lim
, lim_byte
);
923 #if 0 /* This would be clean, but maybe programs depend on
924 a value of nil here. */
932 if (np
< BEGV
|| np
> ZV
)
937 return make_number (np
);
940 /* Return 1 if REGEXP it matches just one constant string. */
943 trivial_regexp_p (regexp
)
946 int len
= SBYTES (regexp
);
947 unsigned char *s
= SDATA (regexp
);
952 case '.': case '*': case '+': case '?': case '[': case '^': case '$':
959 case '|': case '(': case ')': case '`': case '\'': case 'b':
960 case 'B': case '<': case '>': case 'w': case 'W': case 's':
961 case 'S': case '=': case '{': case '}': case '_':
962 case 'c': case 'C': /* for categoryspec and notcategoryspec */
963 case '1': case '2': case '3': case '4': case '5':
964 case '6': case '7': case '8': case '9':
972 /* Search for the n'th occurrence of STRING in the current buffer,
973 starting at position POS and stopping at position LIM,
974 treating STRING as a literal string if RE is false or as
975 a regular expression if RE is true.
977 If N is positive, searching is forward and LIM must be greater than POS.
978 If N is negative, searching is backward and LIM must be less than POS.
980 Returns -x if x occurrences remain to be found (x > 0),
981 or else the position at the beginning of the Nth occurrence
982 (if searching backward) or the end (if searching forward).
984 POSIX is nonzero if we want full backtracking (POSIX style)
985 for this pattern. 0 means backtrack only enough to get a valid match. */
987 #define TRANSLATE(out, trt, d) \
993 temp = Faref (trt, make_number (d)); \
994 if (INTEGERP (temp)) \
1005 search_buffer (string
, pos
, pos_byte
, lim
, lim_byte
, n
,
1006 RE
, trt
, inverse_trt
, posix
)
1015 Lisp_Object inverse_trt
;
1018 int len
= SCHARS (string
);
1019 int len_byte
= SBYTES (string
);
1022 if (running_asynch_code
)
1023 save_search_regs ();
1025 /* Searching 0 times means don't move. */
1026 /* Null string is found at starting position. */
1027 if (len
== 0 || n
== 0)
1029 set_search_regs (pos_byte
, 0);
1033 if (RE
&& !(trivial_regexp_p (string
) && NILP (Vsearch_spaces_regexp
)))
1035 unsigned char *p1
, *p2
;
1037 struct re_pattern_buffer
*bufp
;
1039 bufp
= compile_pattern (string
, &search_regs
, trt
, posix
,
1040 !NILP (current_buffer
->enable_multibyte_characters
));
1042 immediate_quit
= 1; /* Quit immediately if user types ^G,
1043 because letting this function finish
1044 can take too long. */
1045 QUIT
; /* Do a pending quit right away,
1046 to avoid paradoxical behavior */
1047 /* Get pointers and sizes of the two strings
1048 that make up the visible portion of the buffer. */
1051 s1
= GPT_BYTE
- BEGV_BYTE
;
1053 s2
= ZV_BYTE
- GPT_BYTE
;
1057 s2
= ZV_BYTE
- BEGV_BYTE
;
1062 s1
= ZV_BYTE
- BEGV_BYTE
;
1065 re_match_object
= Qnil
;
1070 val
= re_search_2 (bufp
, (char *) p1
, s1
, (char *) p2
, s2
,
1071 pos_byte
- BEGV_BYTE
, lim_byte
- pos_byte
,
1073 /* Don't allow match past current point */
1074 pos_byte
- BEGV_BYTE
);
1077 matcher_overflow ();
1081 pos_byte
= search_regs
.start
[0] + BEGV_BYTE
;
1082 for (i
= 0; i
< search_regs
.num_regs
; i
++)
1083 if (search_regs
.start
[i
] >= 0)
1085 search_regs
.start
[i
]
1086 = BYTE_TO_CHAR (search_regs
.start
[i
] + BEGV_BYTE
);
1088 = BYTE_TO_CHAR (search_regs
.end
[i
] + BEGV_BYTE
);
1090 XSETBUFFER (last_thing_searched
, current_buffer
);
1091 /* Set pos to the new position. */
1092 pos
= search_regs
.start
[0];
1104 val
= re_search_2 (bufp
, (char *) p1
, s1
, (char *) p2
, s2
,
1105 pos_byte
- BEGV_BYTE
, lim_byte
- pos_byte
,
1107 lim_byte
- BEGV_BYTE
);
1110 matcher_overflow ();
1114 pos_byte
= search_regs
.end
[0] + BEGV_BYTE
;
1115 for (i
= 0; i
< search_regs
.num_regs
; i
++)
1116 if (search_regs
.start
[i
] >= 0)
1118 search_regs
.start
[i
]
1119 = BYTE_TO_CHAR (search_regs
.start
[i
] + BEGV_BYTE
);
1121 = BYTE_TO_CHAR (search_regs
.end
[i
] + BEGV_BYTE
);
1123 XSETBUFFER (last_thing_searched
, current_buffer
);
1124 pos
= search_regs
.end
[0];
1136 else /* non-RE case */
1138 unsigned char *raw_pattern
, *pat
;
1139 int raw_pattern_size
;
1140 int raw_pattern_size_byte
;
1141 unsigned char *patbuf
;
1142 int multibyte
= !NILP (current_buffer
->enable_multibyte_characters
);
1143 unsigned char *base_pat
;
1144 /* Set to positive if we find a non-ASCII char that need
1145 translation. Otherwise set to zero later. */
1147 int boyer_moore_ok
= 1;
1149 /* MULTIBYTE says whether the text to be searched is multibyte.
1150 We must convert PATTERN to match that, or we will not really
1151 find things right. */
1153 if (multibyte
== STRING_MULTIBYTE (string
))
1155 raw_pattern
= (unsigned char *) SDATA (string
);
1156 raw_pattern_size
= SCHARS (string
);
1157 raw_pattern_size_byte
= SBYTES (string
);
1161 raw_pattern_size
= SCHARS (string
);
1162 raw_pattern_size_byte
1163 = count_size_as_multibyte (SDATA (string
),
1165 raw_pattern
= (unsigned char *) alloca (raw_pattern_size_byte
+ 1);
1166 copy_text (SDATA (string
), raw_pattern
,
1167 SCHARS (string
), 0, 1);
1171 /* Converting multibyte to single-byte.
1173 ??? Perhaps this conversion should be done in a special way
1174 by subtracting nonascii-insert-offset from each non-ASCII char,
1175 so that only the multibyte chars which really correspond to
1176 the chosen single-byte character set can possibly match. */
1177 raw_pattern_size
= SCHARS (string
);
1178 raw_pattern_size_byte
= SCHARS (string
);
1179 raw_pattern
= (unsigned char *) alloca (raw_pattern_size
+ 1);
1180 copy_text (SDATA (string
), raw_pattern
,
1181 SBYTES (string
), 1, 0);
1184 /* Copy and optionally translate the pattern. */
1185 len
= raw_pattern_size
;
1186 len_byte
= raw_pattern_size_byte
;
1187 patbuf
= (unsigned char *) alloca (len
* MAX_MULTIBYTE_LENGTH
);
1189 base_pat
= raw_pattern
;
1192 /* Fill patbuf by translated characters in STRING while
1193 checking if we can use boyer-moore search. If TRT is
1194 non-nil, we can use boyer-moore search only if TRT can be
1195 represented by the byte array of 256 elements. For that,
1196 all non-ASCII case-equivalents of all case-senstive
1197 characters in STRING must belong to the same charset and
1202 unsigned char str_base
[MAX_MULTIBYTE_LENGTH
], *str
;
1203 int c
, translated
, inverse
;
1204 int in_charlen
, charlen
;
1206 /* If we got here and the RE flag is set, it's because we're
1207 dealing with a regexp known to be trivial, so the backslash
1208 just quotes the next character. */
1209 if (RE
&& *base_pat
== '\\')
1217 c
= STRING_CHAR_AND_LENGTH (base_pat
, len_byte
, in_charlen
);
1222 charlen
= in_charlen
;
1226 /* Translate the character. */
1227 TRANSLATE (translated
, trt
, c
);
1228 charlen
= CHAR_STRING (translated
, str_base
);
1231 /* Check if C has any other case-equivalents. */
1232 TRANSLATE (inverse
, inverse_trt
, c
);
1233 /* If so, check if we can use boyer-moore. */
1234 if (c
!= inverse
&& boyer_moore_ok
)
1236 /* Check if all equivalents belong to the same
1237 group of characters. Note that the check of C
1238 itself is done by the last iteration. */
1239 int this_char_base
= -1;
1241 while (boyer_moore_ok
)
1243 if (ASCII_BYTE_P (inverse
))
1245 if (this_char_base
> 0)
1251 char_base
= this_char_base
;
1254 else if (CHAR_BYTE8_P (inverse
))
1255 /* Boyer-moore search can't handle a
1256 translation of an eight-bit
1259 else if (this_char_base
< 0)
1261 this_char_base
= inverse
& ~0x3F;
1263 char_base
= this_char_base
;
1264 else if (char_base
> 0
1265 && this_char_base
!= char_base
)
1268 else if ((inverse
& ~0x3F) != this_char_base
)
1272 TRANSLATE (inverse
, inverse_trt
, inverse
);
1279 /* Store this character into the translated pattern. */
1280 bcopy (str
, pat
, charlen
);
1282 base_pat
+= in_charlen
;
1283 len_byte
-= in_charlen
;
1288 /* Unibyte buffer. */
1294 /* If we got here and the RE flag is set, it's because we're
1295 dealing with a regexp known to be trivial, so the backslash
1296 just quotes the next character. */
1297 if (RE
&& *base_pat
== '\\')
1304 TRANSLATE (translated
, trt
, c
);
1305 *pat
++ = translated
;
1309 len_byte
= pat
- patbuf
;
1310 len
= raw_pattern_size
;
1311 pat
= base_pat
= patbuf
;
1314 return boyer_moore (n
, pat
, len
, len_byte
, trt
, inverse_trt
,
1315 pos
, pos_byte
, lim
, lim_byte
,
1318 return simple_search (n
, pat
, len
, len_byte
, trt
,
1319 pos
, pos_byte
, lim
, lim_byte
);
1323 /* Do a simple string search N times for the string PAT,
1324 whose length is LEN/LEN_BYTE,
1325 from buffer position POS/POS_BYTE until LIM/LIM_BYTE.
1326 TRT is the translation table.
1328 Return the character position where the match is found.
1329 Otherwise, if M matches remained to be found, return -M.
1331 This kind of search works regardless of what is in PAT and
1332 regardless of what is in TRT. It is used in cases where
1333 boyer_moore cannot work. */
1336 simple_search (n
, pat
, len
, len_byte
, trt
, pos
, pos_byte
, lim
, lim_byte
)
1344 int multibyte
= ! NILP (current_buffer
->enable_multibyte_characters
);
1345 int forward
= n
> 0;
1346 /* Number of buffer bytes matched. Note that this may be different
1347 from len_byte in a multibyte buffer. */
1350 if (lim
> pos
&& multibyte
)
1355 /* Try matching at position POS. */
1357 int this_pos_byte
= pos_byte
;
1359 int this_len_byte
= len_byte
;
1360 unsigned char *p
= pat
;
1361 if (pos
+ len
> lim
)
1364 while (this_len
> 0)
1366 int charlen
, buf_charlen
;
1369 pat_ch
= STRING_CHAR_AND_LENGTH (p
, this_len_byte
, charlen
);
1370 buf_ch
= STRING_CHAR_AND_LENGTH (BYTE_POS_ADDR (this_pos_byte
),
1371 ZV_BYTE
- this_pos_byte
,
1373 TRANSLATE (buf_ch
, trt
, buf_ch
);
1375 if (buf_ch
!= pat_ch
)
1378 this_len_byte
-= charlen
;
1382 this_pos_byte
+= buf_charlen
;
1388 match_byte
= this_pos_byte
- pos_byte
;
1390 pos_byte
+= match_byte
;
1394 INC_BOTH (pos
, pos_byte
);
1404 /* Try matching at position POS. */
1407 unsigned char *p
= pat
;
1409 if (pos
+ len
> lim
)
1412 while (this_len
> 0)
1415 int buf_ch
= FETCH_BYTE (this_pos
);
1416 TRANSLATE (buf_ch
, trt
, buf_ch
);
1418 if (buf_ch
!= pat_ch
)
1437 /* Backwards search. */
1438 else if (lim
< pos
&& multibyte
)
1443 /* Try matching at position POS. */
1444 int this_pos
= pos
- len
;
1447 int this_len_byte
= len_byte
;
1448 unsigned char *p
= pat
;
1450 if (pos
- len
< lim
)
1452 this_pos_byte
= CHAR_TO_BYTE (this_pos
);
1453 match_byte
= pos_byte
- this_pos_byte
;
1455 while (this_len
> 0)
1457 int charlen
, buf_charlen
;
1460 pat_ch
= STRING_CHAR_AND_LENGTH (p
, this_len_byte
, charlen
);
1461 buf_ch
= STRING_CHAR_AND_LENGTH (BYTE_POS_ADDR (this_pos_byte
),
1462 ZV_BYTE
- this_pos_byte
,
1464 TRANSLATE (buf_ch
, trt
, buf_ch
);
1466 if (buf_ch
!= pat_ch
)
1469 this_len_byte
-= charlen
;
1472 this_pos_byte
+= buf_charlen
;
1479 pos_byte
-= match_byte
;
1483 DEC_BOTH (pos
, pos_byte
);
1493 /* Try matching at position POS. */
1494 int this_pos
= pos
- len
;
1496 unsigned char *p
= pat
;
1498 if (pos
- len
< lim
)
1501 while (this_len
> 0)
1504 int buf_ch
= FETCH_BYTE (this_pos
);
1505 TRANSLATE (buf_ch
, trt
, buf_ch
);
1507 if (buf_ch
!= pat_ch
)
1530 set_search_regs ((multibyte
? pos_byte
: pos
) - match_byte
, match_byte
);
1532 set_search_regs (multibyte
? pos_byte
: pos
, match_byte
);
1542 /* Do Boyer-Moore search N times for the string BASE_PAT,
1543 whose length is LEN/LEN_BYTE,
1544 from buffer position POS/POS_BYTE until LIM/LIM_BYTE.
1545 DIRECTION says which direction we search in.
1546 TRT and INVERSE_TRT are translation tables.
1547 Characters in PAT are already translated by TRT.
1549 This kind of search works if all the characters in BASE_PAT that
1550 have nontrivial translation are the same aside from the last byte.
1551 This makes it possible to translate just the last byte of a
1552 character, and do so after just a simple test of the context.
1553 CHAR_BASE is nonzero iff there is such a non-ASCII character.
1555 If that criterion is not satisfied, do not call this function. */
1558 boyer_moore (n
, base_pat
, len
, len_byte
, trt
, inverse_trt
,
1559 pos
, pos_byte
, lim
, lim_byte
, char_base
)
1561 unsigned char *base_pat
;
1564 Lisp_Object inverse_trt
;
1569 int direction
= ((n
> 0) ? 1 : -1);
1570 register int dirlen
;
1571 int infinity
, limit
, stride_for_teases
= 0;
1572 register int *BM_tab
;
1574 register unsigned char *cursor
, *p_limit
;
1576 unsigned char *pat
, *pat_end
;
1577 int multibyte
= ! NILP (current_buffer
->enable_multibyte_characters
);
1579 unsigned char simple_translate
[0400];
1580 /* These are set to the preceding bytes of a byte to be translated
1581 if char_base is nonzero. As the maximum byte length of a
1582 multibyte character is 5, we have to check at most four previous
1584 int translate_prev_byte1
= 0;
1585 int translate_prev_byte2
= 0;
1586 int translate_prev_byte3
= 0;
1587 int translate_prev_byte4
= 0;
1590 int BM_tab_space
[0400];
1591 BM_tab
= &BM_tab_space
[0];
1593 BM_tab
= (int *) alloca (0400 * sizeof (int));
1595 /* The general approach is that we are going to maintain that we know */
1596 /* the first (closest to the present position, in whatever direction */
1597 /* we're searching) character that could possibly be the last */
1598 /* (furthest from present position) character of a valid match. We */
1599 /* advance the state of our knowledge by looking at that character */
1600 /* and seeing whether it indeed matches the last character of the */
1601 /* pattern. If it does, we take a closer look. If it does not, we */
1602 /* move our pointer (to putative last characters) as far as is */
1603 /* logically possible. This amount of movement, which I call a */
1604 /* stride, will be the length of the pattern if the actual character */
1605 /* appears nowhere in the pattern, otherwise it will be the distance */
1606 /* from the last occurrence of that character to the end of the */
1608 /* As a coding trick, an enormous stride is coded into the table for */
1609 /* characters that match the last character. This allows use of only */
1610 /* a single test, a test for having gone past the end of the */
1611 /* permissible match region, to test for both possible matches (when */
1612 /* the stride goes past the end immediately) and failure to */
1613 /* match (where you get nudged past the end one stride at a time). */
1615 /* Here we make a "mickey mouse" BM table. The stride of the search */
1616 /* is determined only by the last character of the putative match. */
1617 /* If that character does not match, we will stride the proper */
1618 /* distance to propose a match that superimposes it on the last */
1619 /* instance of a character that matches it (per trt), or misses */
1620 /* it entirely if there is none. */
1622 dirlen
= len_byte
* direction
;
1623 infinity
= dirlen
- (lim_byte
+ pos_byte
+ len_byte
+ len_byte
) * direction
;
1625 /* Record position after the end of the pattern. */
1626 pat_end
= base_pat
+ len_byte
;
1627 /* BASE_PAT points to a character that we start scanning from.
1628 It is the first character in a forward search,
1629 the last character in a backward search. */
1631 base_pat
= pat_end
- 1;
1633 BM_tab_base
= BM_tab
;
1635 j
= dirlen
; /* to get it in a register */
1636 /* A character that does not appear in the pattern induces a */
1637 /* stride equal to the pattern length. */
1638 while (BM_tab_base
!= BM_tab
)
1646 /* We use this for translation, instead of TRT itself.
1647 We fill this in to handle the characters that actually
1648 occur in the pattern. Others don't matter anyway! */
1649 bzero (simple_translate
, sizeof simple_translate
);
1650 for (i
= 0; i
< 0400; i
++)
1651 simple_translate
[i
] = i
;
1655 /* Setup translate_prev_byte1/2/3/4 from CHAR_BASE. Only a
1656 byte following them are the target of translation. */
1657 unsigned char str
[MAX_MULTIBYTE_LENGTH
];
1658 int len
= CHAR_STRING (char_base
, str
);
1660 translate_prev_byte1
= str
[len
- 2];
1663 translate_prev_byte2
= str
[len
- 3];
1666 translate_prev_byte3
= str
[len
- 4];
1668 translate_prev_byte4
= str
[len
- 5];
1674 while (i
!= infinity
)
1676 unsigned char *ptr
= base_pat
+ i
;
1682 /* If the byte currently looking at is the last of a
1683 character to check case-equivalents, set CH to that
1684 character. An ASCII character and a non-ASCII character
1685 matching with CHAR_BASE are to be checked. */
1688 if (ASCII_BYTE_P (*ptr
) || ! multibyte
)
1691 && ((pat_end
- ptr
) == 1 || CHAR_HEAD_P (ptr
[1])))
1693 unsigned char *charstart
= ptr
- 1;
1695 while (! (CHAR_HEAD_P (*charstart
)))
1697 ch
= STRING_CHAR (charstart
, ptr
- charstart
+ 1);
1698 if (char_base
!= (ch
& ~0x3F))
1703 j
= (ch
& 0x3F) | 0200;
1708 stride_for_teases
= BM_tab
[j
];
1710 BM_tab
[j
] = dirlen
- i
;
1711 /* A translation table is accompanied by its inverse -- see */
1712 /* comment following downcase_table for details */
1715 int starting_ch
= ch
;
1720 TRANSLATE (ch
, inverse_trt
, ch
);
1722 j
= (ch
& 0x3F) | 0200;
1726 /* For all the characters that map into CH,
1727 set up simple_translate to map the last byte
1729 simple_translate
[j
] = starting_j
;
1730 if (ch
== starting_ch
)
1732 BM_tab
[j
] = dirlen
- i
;
1741 stride_for_teases
= BM_tab
[j
];
1742 BM_tab
[j
] = dirlen
- i
;
1744 /* stride_for_teases tells how much to stride if we get a */
1745 /* match on the far character but are subsequently */
1746 /* disappointed, by recording what the stride would have been */
1747 /* for that character if the last character had been */
1750 infinity
= dirlen
- infinity
;
1751 pos_byte
+= dirlen
- ((direction
> 0) ? direction
: 0);
1752 /* loop invariant - POS_BYTE points at where last char (first
1753 char if reverse) of pattern would align in a possible match. */
1757 unsigned char *tail_end_ptr
;
1759 /* It's been reported that some (broken) compiler thinks that
1760 Boolean expressions in an arithmetic context are unsigned.
1761 Using an explicit ?1:0 prevents this. */
1762 if ((lim_byte
- pos_byte
- ((direction
> 0) ? 1 : 0)) * direction
1764 return (n
* (0 - direction
));
1765 /* First we do the part we can by pointers (maybe nothing) */
1768 limit
= pos_byte
- dirlen
+ direction
;
1771 limit
= BUFFER_CEILING_OF (limit
);
1772 /* LIMIT is now the last (not beyond-last!) value POS_BYTE
1773 can take on without hitting edge of buffer or the gap. */
1774 limit
= min (limit
, pos_byte
+ 20000);
1775 limit
= min (limit
, lim_byte
- 1);
1779 limit
= BUFFER_FLOOR_OF (limit
);
1780 /* LIMIT is now the last (not beyond-last!) value POS_BYTE
1781 can take on without hitting edge of buffer or the gap. */
1782 limit
= max (limit
, pos_byte
- 20000);
1783 limit
= max (limit
, lim_byte
);
1785 tail_end
= BUFFER_CEILING_OF (pos_byte
) + 1;
1786 tail_end_ptr
= BYTE_POS_ADDR (tail_end
);
1788 if ((limit
- pos_byte
) * direction
> 20)
1792 p_limit
= BYTE_POS_ADDR (limit
);
1793 p2
= (cursor
= BYTE_POS_ADDR (pos_byte
));
1794 /* In this loop, pos + cursor - p2 is the surrogate for pos */
1795 while (1) /* use one cursor setting as long as i can */
1797 if (direction
> 0) /* worth duplicating */
1799 /* Use signed comparison if appropriate
1800 to make cursor+infinity sure to be > p_limit.
1801 Assuming that the buffer lies in a range of addresses
1802 that are all "positive" (as ints) or all "negative",
1803 either kind of comparison will work as long
1804 as we don't step by infinity. So pick the kind
1805 that works when we do step by infinity. */
1806 if ((EMACS_INT
) (p_limit
+ infinity
) > (EMACS_INT
) p_limit
)
1807 while ((EMACS_INT
) cursor
<= (EMACS_INT
) p_limit
)
1808 cursor
+= BM_tab
[*cursor
];
1810 while ((EMACS_UINT
) cursor
<= (EMACS_UINT
) p_limit
)
1811 cursor
+= BM_tab
[*cursor
];
1815 if ((EMACS_INT
) (p_limit
+ infinity
) < (EMACS_INT
) p_limit
)
1816 while ((EMACS_INT
) cursor
>= (EMACS_INT
) p_limit
)
1817 cursor
+= BM_tab
[*cursor
];
1819 while ((EMACS_UINT
) cursor
>= (EMACS_UINT
) p_limit
)
1820 cursor
+= BM_tab
[*cursor
];
1822 /* If you are here, cursor is beyond the end of the searched region. */
1823 /* This can happen if you match on the far character of the pattern, */
1824 /* because the "stride" of that character is infinity, a number able */
1825 /* to throw you well beyond the end of the search. It can also */
1826 /* happen if you fail to match within the permitted region and would */
1827 /* otherwise try a character beyond that region */
1828 if ((cursor
- p_limit
) * direction
<= len_byte
)
1829 break; /* a small overrun is genuine */
1830 cursor
-= infinity
; /* large overrun = hit */
1831 i
= dirlen
- direction
;
1834 while ((i
-= direction
) + direction
!= 0)
1837 cursor
-= direction
;
1838 /* Translate only the last byte of a character. */
1840 || ((cursor
== tail_end_ptr
1841 || CHAR_HEAD_P (cursor
[1]))
1842 && (CHAR_HEAD_P (cursor
[0])
1843 /* Check if this is the last byte of
1844 a translable character. */
1845 || (translate_prev_byte1
== cursor
[-1]
1846 && (CHAR_HEAD_P (translate_prev_byte1
)
1847 || (translate_prev_byte2
== cursor
[-2]
1848 && (CHAR_HEAD_P (translate_prev_byte2
)
1849 || (translate_prev_byte3
== cursor
[-3]))))))))
1850 ch
= simple_translate
[*cursor
];
1859 while ((i
-= direction
) + direction
!= 0)
1861 cursor
-= direction
;
1862 if (pat
[i
] != *cursor
)
1866 cursor
+= dirlen
- i
- direction
; /* fix cursor */
1867 if (i
+ direction
== 0)
1871 cursor
-= direction
;
1873 position
= pos_byte
+ cursor
- p2
+ ((direction
> 0)
1874 ? 1 - len_byte
: 0);
1875 set_search_regs (position
, len_byte
);
1877 if ((n
-= direction
) != 0)
1878 cursor
+= dirlen
; /* to resume search */
1880 return ((direction
> 0)
1881 ? search_regs
.end
[0] : search_regs
.start
[0]);
1884 cursor
+= stride_for_teases
; /* <sigh> we lose - */
1886 pos_byte
+= cursor
- p2
;
1889 /* Now we'll pick up a clump that has to be done the hard */
1890 /* way because it covers a discontinuity */
1892 limit
= ((direction
> 0)
1893 ? BUFFER_CEILING_OF (pos_byte
- dirlen
+ 1)
1894 : BUFFER_FLOOR_OF (pos_byte
- dirlen
- 1));
1895 limit
= ((direction
> 0)
1896 ? min (limit
+ len_byte
, lim_byte
- 1)
1897 : max (limit
- len_byte
, lim_byte
));
1898 /* LIMIT is now the last value POS_BYTE can have
1899 and still be valid for a possible match. */
1902 /* This loop can be coded for space rather than */
1903 /* speed because it will usually run only once. */
1904 /* (the reach is at most len + 21, and typically */
1905 /* does not exceed len) */
1906 while ((limit
- pos_byte
) * direction
>= 0)
1907 pos_byte
+= BM_tab
[FETCH_BYTE (pos_byte
)];
1908 /* now run the same tests to distinguish going off the */
1909 /* end, a match or a phony match. */
1910 if ((pos_byte
- limit
) * direction
<= len_byte
)
1911 break; /* ran off the end */
1912 /* Found what might be a match.
1913 Set POS_BYTE back to last (first if reverse) pos. */
1914 pos_byte
-= infinity
;
1915 i
= dirlen
- direction
;
1916 while ((i
-= direction
) + direction
!= 0)
1920 pos_byte
-= direction
;
1921 ptr
= BYTE_POS_ADDR (pos_byte
);
1922 /* Translate only the last byte of a character. */
1924 || ((ptr
== tail_end_ptr
1925 || CHAR_HEAD_P (ptr
[1]))
1926 && (CHAR_HEAD_P (ptr
[0])
1927 /* Check if this is the last byte of a
1928 translable character. */
1929 || (translate_prev_byte1
== ptr
[-1]
1930 && (CHAR_HEAD_P (translate_prev_byte1
)
1931 || (translate_prev_byte2
== ptr
[-2]
1932 && (CHAR_HEAD_P (translate_prev_byte2
)
1933 || translate_prev_byte3
== ptr
[-3])))))))
1934 ch
= simple_translate
[*ptr
];
1940 /* Above loop has moved POS_BYTE part or all the way
1941 back to the first pos (last pos if reverse).
1942 Set it once again at the last (first if reverse) char. */
1943 pos_byte
+= dirlen
- i
- direction
;
1944 if (i
+ direction
== 0)
1947 pos_byte
-= direction
;
1949 position
= pos_byte
+ ((direction
> 0) ? 1 - len_byte
: 0);
1951 set_search_regs (position
, len_byte
);
1953 if ((n
-= direction
) != 0)
1954 pos_byte
+= dirlen
; /* to resume search */
1956 return ((direction
> 0)
1957 ? search_regs
.end
[0] : search_regs
.start
[0]);
1960 pos_byte
+= stride_for_teases
;
1963 /* We have done one clump. Can we continue? */
1964 if ((lim_byte
- pos_byte
) * direction
< 0)
1965 return ((0 - n
) * direction
);
1967 return BYTE_TO_CHAR (pos_byte
);
1970 /* Record beginning BEG_BYTE and end BEG_BYTE + NBYTES
1971 for the overall match just found in the current buffer.
1972 Also clear out the match data for registers 1 and up. */
1975 set_search_regs (beg_byte
, nbytes
)
1976 int beg_byte
, nbytes
;
1980 /* Make sure we have registers in which to store
1981 the match position. */
1982 if (search_regs
.num_regs
== 0)
1984 search_regs
.start
= (regoff_t
*) xmalloc (2 * sizeof (regoff_t
));
1985 search_regs
.end
= (regoff_t
*) xmalloc (2 * sizeof (regoff_t
));
1986 search_regs
.num_regs
= 2;
1989 /* Clear out the other registers. */
1990 for (i
= 1; i
< search_regs
.num_regs
; i
++)
1992 search_regs
.start
[i
] = -1;
1993 search_regs
.end
[i
] = -1;
1996 search_regs
.start
[0] = BYTE_TO_CHAR (beg_byte
);
1997 search_regs
.end
[0] = BYTE_TO_CHAR (beg_byte
+ nbytes
);
1998 XSETBUFFER (last_thing_searched
, current_buffer
);
2001 /* Given a string of words separated by word delimiters,
2002 compute a regexp that matches those exact words
2003 separated by arbitrary punctuation. */
2009 register unsigned char *p
, *o
;
2010 register int i
, i_byte
, len
, punct_count
= 0, word_count
= 0;
2015 CHECK_STRING (string
);
2017 len
= SCHARS (string
);
2019 for (i
= 0, i_byte
= 0; i
< len
; )
2023 FETCH_STRING_CHAR_AS_MULTIBYTE_ADVANCE (c
, string
, i
, i_byte
);
2025 if (SYNTAX (c
) != Sword
)
2028 if (i
> 0 && SYNTAX (prev_c
) == Sword
)
2035 if (SYNTAX (prev_c
) == Sword
)
2038 return empty_string
;
2040 adjust
= - punct_count
+ 5 * (word_count
- 1) + 4;
2041 if (STRING_MULTIBYTE (string
))
2042 val
= make_uninit_multibyte_string (len
+ adjust
,
2046 val
= make_uninit_string (len
+ adjust
);
2053 for (i
= 0, i_byte
= 0; i
< len
; )
2056 int i_byte_orig
= i_byte
;
2058 FETCH_STRING_CHAR_AS_MULTIBYTE_ADVANCE (c
, string
, i
, i_byte
);
2060 if (SYNTAX (c
) == Sword
)
2062 bcopy (SDATA (string
) + i_byte_orig
, o
,
2063 i_byte
- i_byte_orig
);
2064 o
+= i_byte
- i_byte_orig
;
2066 else if (i
> 0 && SYNTAX (prev_c
) == Sword
&& --word_count
)
2084 DEFUN ("search-backward", Fsearch_backward
, Ssearch_backward
, 1, 4,
2085 "MSearch backward: ",
2086 doc
: /* Search backward from point for STRING.
2087 Set point to the beginning of the occurrence found, and return point.
2088 An optional second argument bounds the search; it is a buffer position.
2089 The match found must not extend before that position.
2090 Optional third argument, if t, means if fail just return nil (no error).
2091 If not nil and not t, position at limit of search and return nil.
2092 Optional fourth argument is repeat count--search for successive occurrences.
2094 Search case-sensitivity is determined by the value of the variable
2095 `case-fold-search', which see.
2097 See also the functions `match-beginning', `match-end' and `replace-match'. */)
2098 (string
, bound
, noerror
, count
)
2099 Lisp_Object string
, bound
, noerror
, count
;
2101 return search_command (string
, bound
, noerror
, count
, -1, 0, 0);
2104 DEFUN ("search-forward", Fsearch_forward
, Ssearch_forward
, 1, 4, "MSearch: ",
2105 doc
: /* Search forward from point for STRING.
2106 Set point to the end of the occurrence found, and return point.
2107 An optional second argument bounds the search; it is a buffer position.
2108 The match found must not extend after that position. nil is equivalent
2110 Optional third argument, if t, means if fail just return nil (no error).
2111 If not nil and not t, move to limit of search and return nil.
2112 Optional fourth argument is repeat count--search for successive occurrences.
2114 Search case-sensitivity is determined by the value of the variable
2115 `case-fold-search', which see.
2117 See also the functions `match-beginning', `match-end' and `replace-match'. */)
2118 (string
, bound
, noerror
, count
)
2119 Lisp_Object string
, bound
, noerror
, count
;
2121 return search_command (string
, bound
, noerror
, count
, 1, 0, 0);
2124 DEFUN ("word-search-backward", Fword_search_backward
, Sword_search_backward
, 1, 4,
2125 "sWord search backward: ",
2126 doc
: /* Search backward from point for STRING, ignoring differences in punctuation.
2127 Set point to the beginning of the occurrence found, and return point.
2128 An optional second argument bounds the search; it is a buffer position.
2129 The match found must not extend before that position.
2130 Optional third argument, if t, means if fail just return nil (no error).
2131 If not nil and not t, move to limit of search and return nil.
2132 Optional fourth argument is repeat count--search for successive occurrences. */)
2133 (string
, bound
, noerror
, count
)
2134 Lisp_Object string
, bound
, noerror
, count
;
2136 return search_command (wordify (string
), bound
, noerror
, count
, -1, 1, 0);
2139 DEFUN ("word-search-forward", Fword_search_forward
, Sword_search_forward
, 1, 4,
2141 doc
: /* Search forward from point for STRING, ignoring differences in punctuation.
2142 Set point to the end of the occurrence found, and return point.
2143 An optional second argument bounds the search; it is a buffer position.
2144 The match found must not extend after that position.
2145 Optional third argument, if t, means if fail just return nil (no error).
2146 If not nil and not t, move to limit of search and return nil.
2147 Optional fourth argument is repeat count--search for successive occurrences. */)
2148 (string
, bound
, noerror
, count
)
2149 Lisp_Object string
, bound
, noerror
, count
;
2151 return search_command (wordify (string
), bound
, noerror
, count
, 1, 1, 0);
2154 DEFUN ("re-search-backward", Fre_search_backward
, Sre_search_backward
, 1, 4,
2155 "sRE search backward: ",
2156 doc
: /* Search backward from point for match for regular expression REGEXP.
2157 Set point to the beginning of the match, and return point.
2158 The match found is the one starting last in the buffer
2159 and yet ending before the origin of the search.
2160 An optional second argument bounds the search; it is a buffer position.
2161 The match found must start at or after that position.
2162 Optional third argument, if t, means if fail just return nil (no error).
2163 If not nil and not t, move to limit of search and return nil.
2164 Optional fourth argument is repeat count--search for successive occurrences.
2165 See also the functions `match-beginning', `match-end', `match-string',
2166 and `replace-match'. */)
2167 (regexp
, bound
, noerror
, count
)
2168 Lisp_Object regexp
, bound
, noerror
, count
;
2170 return search_command (regexp
, bound
, noerror
, count
, -1, 1, 0);
2173 DEFUN ("re-search-forward", Fre_search_forward
, Sre_search_forward
, 1, 4,
2175 doc
: /* Search forward from point for regular expression REGEXP.
2176 Set point to the end of the occurrence found, and return point.
2177 An optional second argument bounds the search; it is a buffer position.
2178 The match found must not extend after that position.
2179 Optional third argument, if t, means if fail just return nil (no error).
2180 If not nil and not t, move to limit of search and return nil.
2181 Optional fourth argument is repeat count--search for successive occurrences.
2182 See also the functions `match-beginning', `match-end', `match-string',
2183 and `replace-match'. */)
2184 (regexp
, bound
, noerror
, count
)
2185 Lisp_Object regexp
, bound
, noerror
, count
;
2187 return search_command (regexp
, bound
, noerror
, count
, 1, 1, 0);
2190 DEFUN ("posix-search-backward", Fposix_search_backward
, Sposix_search_backward
, 1, 4,
2191 "sPosix search backward: ",
2192 doc
: /* Search backward from point for match for regular expression REGEXP.
2193 Find the longest match in accord with Posix regular expression rules.
2194 Set point to the beginning of the match, and return point.
2195 The match found is the one starting last in the buffer
2196 and yet ending before the origin of the search.
2197 An optional second argument bounds the search; it is a buffer position.
2198 The match found must start at or after that position.
2199 Optional third argument, if t, means if fail just return nil (no error).
2200 If not nil and not t, move to limit of search and return nil.
2201 Optional fourth argument is repeat count--search for successive occurrences.
2202 See also the functions `match-beginning', `match-end', `match-string',
2203 and `replace-match'. */)
2204 (regexp
, bound
, noerror
, count
)
2205 Lisp_Object regexp
, bound
, noerror
, count
;
2207 return search_command (regexp
, bound
, noerror
, count
, -1, 1, 1);
2210 DEFUN ("posix-search-forward", Fposix_search_forward
, Sposix_search_forward
, 1, 4,
2212 doc
: /* Search forward from point for regular expression REGEXP.
2213 Find the longest match in accord with Posix regular expression rules.
2214 Set point to the end of the occurrence found, and return point.
2215 An optional second argument bounds the search; it is a buffer position.
2216 The match found must not extend after that position.
2217 Optional third argument, if t, means if fail just return nil (no error).
2218 If not nil and not t, move to limit of search and return nil.
2219 Optional fourth argument is repeat count--search for successive occurrences.
2220 See also the functions `match-beginning', `match-end', `match-string',
2221 and `replace-match'. */)
2222 (regexp
, bound
, noerror
, count
)
2223 Lisp_Object regexp
, bound
, noerror
, count
;
2225 return search_command (regexp
, bound
, noerror
, count
, 1, 1, 1);
2228 DEFUN ("replace-match", Freplace_match
, Sreplace_match
, 1, 5, 0,
2229 doc
: /* Replace text matched by last search with NEWTEXT.
2230 Leave point at the end of the replacement text.
2232 If second arg FIXEDCASE is non-nil, do not alter case of replacement text.
2233 Otherwise maybe capitalize the whole text, or maybe just word initials,
2234 based on the replaced text.
2235 If the replaced text has only capital letters
2236 and has at least one multiletter word, convert NEWTEXT to all caps.
2237 Otherwise if all words are capitalized in the replaced text,
2238 capitalize each word in NEWTEXT.
2240 If third arg LITERAL is non-nil, insert NEWTEXT literally.
2241 Otherwise treat `\\' as special:
2242 `\\&' in NEWTEXT means substitute original matched text.
2243 `\\N' means substitute what matched the Nth `\\(...\\)'.
2244 If Nth parens didn't match, substitute nothing.
2245 `\\\\' means insert one `\\'.
2246 Case conversion does not apply to these substitutions.
2248 FIXEDCASE and LITERAL are optional arguments.
2250 The optional fourth argument STRING can be a string to modify.
2251 This is meaningful when the previous match was done against STRING,
2252 using `string-match'. When used this way, `replace-match'
2253 creates and returns a new string made by copying STRING and replacing
2254 the part of STRING that was matched.
2256 The optional fifth argument SUBEXP specifies a subexpression;
2257 it says to replace just that subexpression with NEWTEXT,
2258 rather than replacing the entire matched text.
2259 This is, in a vague sense, the inverse of using `\\N' in NEWTEXT;
2260 `\\N' copies subexp N into NEWTEXT, but using N as SUBEXP puts
2261 NEWTEXT in place of subexp N.
2262 This is useful only after a regular expression search or match,
2263 since only regular expressions have distinguished subexpressions. */)
2264 (newtext
, fixedcase
, literal
, string
, subexp
)
2265 Lisp_Object newtext
, fixedcase
, literal
, string
, subexp
;
2267 enum { nochange
, all_caps
, cap_initial
} case_action
;
2268 register int pos
, pos_byte
;
2269 int some_multiletter_word
;
2272 int some_nonuppercase_initial
;
2273 register int c
, prevc
;
2275 int opoint
, newpoint
;
2277 CHECK_STRING (newtext
);
2279 if (! NILP (string
))
2280 CHECK_STRING (string
);
2282 case_action
= nochange
; /* We tried an initialization */
2283 /* but some C compilers blew it */
2285 if (search_regs
.num_regs
<= 0)
2286 error ("`replace-match' called before any match found");
2292 CHECK_NUMBER (subexp
);
2293 sub
= XINT (subexp
);
2294 if (sub
< 0 || sub
>= search_regs
.num_regs
)
2295 args_out_of_range (subexp
, make_number (search_regs
.num_regs
));
2300 if (search_regs
.start
[sub
] < BEGV
2301 || search_regs
.start
[sub
] > search_regs
.end
[sub
]
2302 || search_regs
.end
[sub
] > ZV
)
2303 args_out_of_range (make_number (search_regs
.start
[sub
]),
2304 make_number (search_regs
.end
[sub
]));
2308 if (search_regs
.start
[sub
] < 0
2309 || search_regs
.start
[sub
] > search_regs
.end
[sub
]
2310 || search_regs
.end
[sub
] > SCHARS (string
))
2311 args_out_of_range (make_number (search_regs
.start
[sub
]),
2312 make_number (search_regs
.end
[sub
]));
2315 if (NILP (fixedcase
))
2317 /* Decide how to casify by examining the matched text. */
2320 pos
= search_regs
.start
[sub
];
2321 last
= search_regs
.end
[sub
];
2324 pos_byte
= CHAR_TO_BYTE (pos
);
2326 pos_byte
= string_char_to_byte (string
, pos
);
2329 case_action
= all_caps
;
2331 /* some_multiletter_word is set nonzero if any original word
2332 is more than one letter long. */
2333 some_multiletter_word
= 0;
2335 some_nonuppercase_initial
= 0;
2342 c
= FETCH_CHAR_AS_MULTIBYTE (pos_byte
);
2343 INC_BOTH (pos
, pos_byte
);
2346 FETCH_STRING_CHAR_AS_MULTIBYTE_ADVANCE (c
, string
, pos
, pos_byte
);
2350 /* Cannot be all caps if any original char is lower case */
2353 if (SYNTAX (prevc
) != Sword
)
2354 some_nonuppercase_initial
= 1;
2356 some_multiletter_word
= 1;
2358 else if (UPPERCASEP (c
))
2361 if (SYNTAX (prevc
) != Sword
)
2364 some_multiletter_word
= 1;
2368 /* If the initial is a caseless word constituent,
2369 treat that like a lowercase initial. */
2370 if (SYNTAX (prevc
) != Sword
)
2371 some_nonuppercase_initial
= 1;
2377 /* Convert to all caps if the old text is all caps
2378 and has at least one multiletter word. */
2379 if (! some_lowercase
&& some_multiletter_word
)
2380 case_action
= all_caps
;
2381 /* Capitalize each word, if the old text has all capitalized words. */
2382 else if (!some_nonuppercase_initial
&& some_multiletter_word
)
2383 case_action
= cap_initial
;
2384 else if (!some_nonuppercase_initial
&& some_uppercase
)
2385 /* Should x -> yz, operating on X, give Yz or YZ?
2386 We'll assume the latter. */
2387 case_action
= all_caps
;
2389 case_action
= nochange
;
2392 /* Do replacement in a string. */
2395 Lisp_Object before
, after
;
2397 before
= Fsubstring (string
, make_number (0),
2398 make_number (search_regs
.start
[sub
]));
2399 after
= Fsubstring (string
, make_number (search_regs
.end
[sub
]), Qnil
);
2401 /* Substitute parts of the match into NEWTEXT
2406 int lastpos_byte
= 0;
2407 /* We build up the substituted string in ACCUM. */
2410 int length
= SBYTES (newtext
);
2414 for (pos_byte
= 0, pos
= 0; pos_byte
< length
;)
2418 int delbackslash
= 0;
2420 FETCH_STRING_CHAR_ADVANCE (c
, newtext
, pos
, pos_byte
);
2424 FETCH_STRING_CHAR_ADVANCE (c
, newtext
, pos
, pos_byte
);
2428 substart
= search_regs
.start
[sub
];
2429 subend
= search_regs
.end
[sub
];
2431 else if (c
>= '1' && c
<= '9')
2433 if (search_regs
.start
[c
- '0'] >= 0
2434 && c
<= search_regs
.num_regs
+ '0')
2436 substart
= search_regs
.start
[c
- '0'];
2437 subend
= search_regs
.end
[c
- '0'];
2441 /* If that subexp did not match,
2442 replace \\N with nothing. */
2450 error ("Invalid use of `\\' in replacement text");
2454 if (pos
- 2 != lastpos
)
2455 middle
= substring_both (newtext
, lastpos
,
2457 pos
- 2, pos_byte
- 2);
2460 accum
= concat3 (accum
, middle
,
2462 make_number (substart
),
2463 make_number (subend
)));
2465 lastpos_byte
= pos_byte
;
2467 else if (delbackslash
)
2469 middle
= substring_both (newtext
, lastpos
,
2471 pos
- 1, pos_byte
- 1);
2473 accum
= concat2 (accum
, middle
);
2475 lastpos_byte
= pos_byte
;
2480 middle
= substring_both (newtext
, lastpos
,
2486 newtext
= concat2 (accum
, middle
);
2489 /* Do case substitution in NEWTEXT if desired. */
2490 if (case_action
== all_caps
)
2491 newtext
= Fupcase (newtext
);
2492 else if (case_action
== cap_initial
)
2493 newtext
= Fupcase_initials (newtext
);
2495 return concat3 (before
, newtext
, after
);
2498 /* Record point, then move (quietly) to the start of the match. */
2499 if (PT
>= search_regs
.end
[sub
])
2501 else if (PT
> search_regs
.start
[sub
])
2502 opoint
= search_regs
.end
[sub
] - ZV
;
2506 /* If we want non-literal replacement,
2507 perform substitution on the replacement string. */
2510 int length
= SBYTES (newtext
);
2511 unsigned char *substed
;
2512 int substed_alloc_size
, substed_len
;
2513 int buf_multibyte
= !NILP (current_buffer
->enable_multibyte_characters
);
2514 int str_multibyte
= STRING_MULTIBYTE (newtext
);
2515 Lisp_Object rev_tbl
;
2516 int really_changed
= 0;
2520 substed_alloc_size
= length
* 2 + 100;
2521 substed
= (unsigned char *) xmalloc (substed_alloc_size
+ 1);
2524 /* Go thru NEWTEXT, producing the actual text to insert in
2525 SUBSTED while adjusting multibyteness to that of the current
2528 for (pos_byte
= 0, pos
= 0; pos_byte
< length
;)
2530 unsigned char str
[MAX_MULTIBYTE_LENGTH
];
2531 unsigned char *add_stuff
= NULL
;
2537 FETCH_STRING_CHAR_ADVANCE_NO_CHECK (c
, newtext
, pos
, pos_byte
);
2539 c
= multibyte_char_to_unibyte (c
, rev_tbl
);
2543 /* Note that we don't have to increment POS. */
2544 c
= SREF (newtext
, pos_byte
++);
2546 c
= unibyte_char_to_multibyte (c
);
2549 /* Either set ADD_STUFF and ADD_LEN to the text to put in SUBSTED,
2550 or set IDX to a match index, which means put that part
2551 of the buffer text into SUBSTED. */
2559 FETCH_STRING_CHAR_ADVANCE_NO_CHECK (c
, newtext
,
2561 if (!buf_multibyte
&& !ASCII_CHAR_P (c
))
2562 c
= multibyte_char_to_unibyte (c
, rev_tbl
);
2566 c
= SREF (newtext
, pos_byte
++);
2568 c
= unibyte_char_to_multibyte (c
);
2573 else if (c
>= '1' && c
<= '9' && c
<= search_regs
.num_regs
+ '0')
2575 if (search_regs
.start
[c
- '0'] >= 1)
2579 add_len
= 1, add_stuff
= "\\";
2583 error ("Invalid use of `\\' in replacement text");
2588 add_len
= CHAR_STRING (c
, str
);
2592 /* If we want to copy part of a previous match,
2593 set up ADD_STUFF and ADD_LEN to point to it. */
2596 int begbyte
= CHAR_TO_BYTE (search_regs
.start
[idx
]);
2597 add_len
= CHAR_TO_BYTE (search_regs
.end
[idx
]) - begbyte
;
2598 if (search_regs
.start
[idx
] < GPT
&& GPT
< search_regs
.end
[idx
])
2599 move_gap (search_regs
.start
[idx
]);
2600 add_stuff
= BYTE_POS_ADDR (begbyte
);
2603 /* Now the stuff we want to add to SUBSTED
2604 is invariably ADD_LEN bytes starting at ADD_STUFF. */
2606 /* Make sure SUBSTED is big enough. */
2607 if (substed_len
+ add_len
>= substed_alloc_size
)
2609 substed_alloc_size
= substed_len
+ add_len
+ 500;
2610 substed
= (unsigned char *) xrealloc (substed
,
2611 substed_alloc_size
+ 1);
2614 /* Now add to the end of SUBSTED. */
2617 bcopy (add_stuff
, substed
+ substed_len
, add_len
);
2618 substed_len
+= add_len
;
2626 int nchars
= multibyte_chars_in_text (substed
, substed_len
);
2628 newtext
= make_multibyte_string (substed
, nchars
, substed_len
);
2631 newtext
= make_unibyte_string (substed
, substed_len
);
2636 /* Replace the old text with the new in the cleanest possible way. */
2637 replace_range (search_regs
.start
[sub
], search_regs
.end
[sub
],
2639 newpoint
= search_regs
.start
[sub
] + SCHARS (newtext
);
2641 if (case_action
== all_caps
)
2642 Fupcase_region (make_number (search_regs
.start
[sub
]),
2643 make_number (newpoint
));
2644 else if (case_action
== cap_initial
)
2645 Fupcase_initials_region (make_number (search_regs
.start
[sub
]),
2646 make_number (newpoint
));
2648 /* Adjust search data for this change. */
2650 int oldend
= search_regs
.end
[sub
];
2651 int oldstart
= search_regs
.start
[sub
];
2652 int change
= newpoint
- search_regs
.end
[sub
];
2655 for (i
= 0; i
< search_regs
.num_regs
; i
++)
2657 if (search_regs
.start
[i
] >= oldend
)
2658 search_regs
.start
[i
] += change
;
2659 else if (search_regs
.start
[i
] > oldstart
)
2660 search_regs
.start
[i
] = oldstart
;
2661 if (search_regs
.end
[i
] >= oldend
)
2662 search_regs
.end
[i
] += change
;
2663 else if (search_regs
.end
[i
] > oldstart
)
2664 search_regs
.end
[i
] = oldstart
;
2668 /* Put point back where it was in the text. */
2670 TEMP_SET_PT (opoint
+ ZV
);
2672 TEMP_SET_PT (opoint
);
2674 /* Now move point "officially" to the start of the inserted replacement. */
2675 move_if_not_intangible (newpoint
);
2681 match_limit (num
, beginningp
)
2690 args_out_of_range (num
, make_number (0));
2691 if (search_regs
.num_regs
<= 0)
2692 error ("No match data, because no search succeeded");
2693 if (n
>= search_regs
.num_regs
2694 || search_regs
.start
[n
] < 0)
2696 return (make_number ((beginningp
) ? search_regs
.start
[n
]
2697 : search_regs
.end
[n
]));
2700 DEFUN ("match-beginning", Fmatch_beginning
, Smatch_beginning
, 1, 1, 0,
2701 doc
: /* Return position of start of text matched by last search.
2702 SUBEXP, a number, specifies which parenthesized expression in the last
2704 Value is nil if SUBEXPth pair didn't match, or there were less than
2706 Zero means the entire text matched by the whole regexp or whole string. */)
2710 return match_limit (subexp
, 1);
2713 DEFUN ("match-end", Fmatch_end
, Smatch_end
, 1, 1, 0,
2714 doc
: /* Return position of end of text matched by last search.
2715 SUBEXP, a number, specifies which parenthesized expression in the last
2717 Value is nil if SUBEXPth pair didn't match, or there were less than
2719 Zero means the entire text matched by the whole regexp or whole string. */)
2723 return match_limit (subexp
, 0);
2726 DEFUN ("match-data", Fmatch_data
, Smatch_data
, 0, 3, 0,
2727 doc
: /* Return a list containing all info on what the last search matched.
2728 Element 2N is `(match-beginning N)'; element 2N + 1 is `(match-end N)'.
2729 All the elements are markers or nil (nil if the Nth pair didn't match)
2730 if the last match was on a buffer; integers or nil if a string was matched.
2731 Use `store-match-data' to reinstate the data in this list.
2733 If INTEGERS (the optional first argument) is non-nil, always use
2734 integers \(rather than markers) to represent buffer positions. In
2735 this case, and if the last match was in a buffer, the buffer will get
2736 stored as one additional element at the end of the list.
2738 If REUSE is a list, reuse it as part of the value. If REUSE is long
2739 enough to hold all the values, and if INTEGERS is non-nil, no consing
2742 If optional third arg RESEAT is non-nil, any previous markers on the
2743 REUSE list will be modified to point to nowhere.
2745 Return value is undefined if the last search failed. */)
2746 (integers
, reuse
, reseat
)
2747 Lisp_Object integers
, reuse
, reseat
;
2749 Lisp_Object tail
, prev
;
2754 for (tail
= reuse
; CONSP (tail
); tail
= XCDR (tail
))
2755 if (MARKERP (XCAR (tail
)))
2757 unchain_marker (XMARKER (XCAR (tail
)));
2758 XSETCAR (tail
, Qnil
);
2761 if (NILP (last_thing_searched
))
2766 data
= (Lisp_Object
*) alloca ((2 * search_regs
.num_regs
+ 1)
2767 * sizeof (Lisp_Object
));
2770 for (i
= 0; i
< search_regs
.num_regs
; i
++)
2772 int start
= search_regs
.start
[i
];
2775 if (EQ (last_thing_searched
, Qt
)
2776 || ! NILP (integers
))
2778 XSETFASTINT (data
[2 * i
], start
);
2779 XSETFASTINT (data
[2 * i
+ 1], search_regs
.end
[i
]);
2781 else if (BUFFERP (last_thing_searched
))
2783 data
[2 * i
] = Fmake_marker ();
2784 Fset_marker (data
[2 * i
],
2785 make_number (start
),
2786 last_thing_searched
);
2787 data
[2 * i
+ 1] = Fmake_marker ();
2788 Fset_marker (data
[2 * i
+ 1],
2789 make_number (search_regs
.end
[i
]),
2790 last_thing_searched
);
2793 /* last_thing_searched must always be Qt, a buffer, or Qnil. */
2799 data
[2 * i
] = data
[2 * i
+ 1] = Qnil
;
2802 if (BUFFERP (last_thing_searched
) && !NILP (integers
))
2804 data
[len
] = last_thing_searched
;
2808 /* If REUSE is not usable, cons up the values and return them. */
2809 if (! CONSP (reuse
))
2810 return Flist (len
, data
);
2812 /* If REUSE is a list, store as many value elements as will fit
2813 into the elements of REUSE. */
2814 for (i
= 0, tail
= reuse
; CONSP (tail
);
2815 i
++, tail
= XCDR (tail
))
2818 XSETCAR (tail
, data
[i
]);
2820 XSETCAR (tail
, Qnil
);
2824 /* If we couldn't fit all value elements into REUSE,
2825 cons up the rest of them and add them to the end of REUSE. */
2827 XSETCDR (prev
, Flist (len
- i
, data
+ i
));
2832 /* Internal usage only:
2833 If RESEAT is `evaporate', put the markers back on the free list
2834 immediately. No other references to the markers must exist in this case,
2835 so it is used only internally on the unwind stack and save-match-data from
2838 DEFUN ("set-match-data", Fset_match_data
, Sset_match_data
, 1, 2, 0,
2839 doc
: /* Set internal data on last search match from elements of LIST.
2840 LIST should have been created by calling `match-data' previously.
2842 If optional arg RESEAT is non-nil, make markers on LIST point nowhere. */)
2844 register Lisp_Object list
, reseat
;
2847 register Lisp_Object marker
;
2849 if (running_asynch_code
)
2850 save_search_regs ();
2852 if (!CONSP (list
) && !NILP (list
))
2853 list
= wrong_type_argument (Qconsp
, list
);
2855 /* Unless we find a marker with a buffer or an explicit buffer
2856 in LIST, assume that this match data came from a string. */
2857 last_thing_searched
= Qt
;
2859 /* Allocate registers if they don't already exist. */
2861 int length
= XFASTINT (Flength (list
)) / 2;
2863 if (length
> search_regs
.num_regs
)
2865 if (search_regs
.num_regs
== 0)
2868 = (regoff_t
*) xmalloc (length
* sizeof (regoff_t
));
2870 = (regoff_t
*) xmalloc (length
* sizeof (regoff_t
));
2875 = (regoff_t
*) xrealloc (search_regs
.start
,
2876 length
* sizeof (regoff_t
));
2878 = (regoff_t
*) xrealloc (search_regs
.end
,
2879 length
* sizeof (regoff_t
));
2882 for (i
= search_regs
.num_regs
; i
< length
; i
++)
2883 search_regs
.start
[i
] = -1;
2885 search_regs
.num_regs
= length
;
2888 for (i
= 0; CONSP (list
); i
++)
2890 marker
= XCAR (list
);
2891 if (BUFFERP (marker
))
2893 last_thing_searched
= marker
;
2900 search_regs
.start
[i
] = -1;
2909 if (MARKERP (marker
))
2911 if (XMARKER (marker
)->buffer
== 0)
2912 XSETFASTINT (marker
, 0);
2914 XSETBUFFER (last_thing_searched
, XMARKER (marker
)->buffer
);
2917 CHECK_NUMBER_COERCE_MARKER (marker
);
2918 from
= XINT (marker
);
2920 if (!NILP (reseat
) && MARKERP (m
))
2922 if (EQ (reseat
, Qevaporate
))
2925 unchain_marker (XMARKER (m
));
2926 XSETCAR (list
, Qnil
);
2929 if ((list
= XCDR (list
), !CONSP (list
)))
2932 m
= marker
= XCAR (list
);
2934 if (MARKERP (marker
) && XMARKER (marker
)->buffer
== 0)
2935 XSETFASTINT (marker
, 0);
2937 CHECK_NUMBER_COERCE_MARKER (marker
);
2938 search_regs
.start
[i
] = from
;
2939 search_regs
.end
[i
] = XINT (marker
);
2941 if (!NILP (reseat
) && MARKERP (m
))
2943 if (EQ (reseat
, Qevaporate
))
2946 unchain_marker (XMARKER (m
));
2947 XSETCAR (list
, Qnil
);
2953 for (; i
< search_regs
.num_regs
; i
++)
2954 search_regs
.start
[i
] = -1;
2960 /* If non-zero the match data have been saved in saved_search_regs
2961 during the execution of a sentinel or filter. */
2962 static int search_regs_saved
;
2963 static struct re_registers saved_search_regs
;
2964 static Lisp_Object saved_last_thing_searched
;
2966 /* Called from Flooking_at, Fstring_match, search_buffer, Fstore_match_data
2967 if asynchronous code (filter or sentinel) is running. */
2971 if (!search_regs_saved
)
2973 saved_search_regs
.num_regs
= search_regs
.num_regs
;
2974 saved_search_regs
.start
= search_regs
.start
;
2975 saved_search_regs
.end
= search_regs
.end
;
2976 saved_last_thing_searched
= last_thing_searched
;
2977 last_thing_searched
= Qnil
;
2978 search_regs
.num_regs
= 0;
2979 search_regs
.start
= 0;
2980 search_regs
.end
= 0;
2982 search_regs_saved
= 1;
2986 /* Called upon exit from filters and sentinels. */
2988 restore_search_regs ()
2990 if (search_regs_saved
)
2992 if (search_regs
.num_regs
> 0)
2994 xfree (search_regs
.start
);
2995 xfree (search_regs
.end
);
2997 search_regs
.num_regs
= saved_search_regs
.num_regs
;
2998 search_regs
.start
= saved_search_regs
.start
;
2999 search_regs
.end
= saved_search_regs
.end
;
3000 last_thing_searched
= saved_last_thing_searched
;
3001 saved_last_thing_searched
= Qnil
;
3002 search_regs_saved
= 0;
3007 unwind_set_match_data (list
)
3010 /* It is safe to free (evaporate) the markers immediately. */
3011 return Fset_match_data (list
, Qevaporate
);
3014 /* Called to unwind protect the match data. */
3016 record_unwind_save_match_data ()
3018 record_unwind_protect (unwind_set_match_data
,
3019 Fmatch_data (Qnil
, Qnil
, Qnil
));
3022 /* Quote a string to inactivate reg-expr chars */
3024 DEFUN ("regexp-quote", Fregexp_quote
, Sregexp_quote
, 1, 1, 0,
3025 doc
: /* Return a regexp string which matches exactly STRING and nothing else. */)
3029 register unsigned char *in
, *out
, *end
;
3030 register unsigned char *temp
;
3031 int backslashes_added
= 0;
3033 CHECK_STRING (string
);
3035 temp
= (unsigned char *) alloca (SBYTES (string
) * 2);
3037 /* Now copy the data into the new string, inserting escapes. */
3039 in
= SDATA (string
);
3040 end
= in
+ SBYTES (string
);
3043 for (; in
!= end
; in
++)
3046 || *in
== '*' || *in
== '.' || *in
== '\\'
3047 || *in
== '?' || *in
== '+'
3048 || *in
== '^' || *in
== '$')
3049 *out
++ = '\\', backslashes_added
++;
3053 return make_specified_string (temp
,
3054 SCHARS (string
) + backslashes_added
,
3056 STRING_MULTIBYTE (string
));
3064 for (i
= 0; i
< REGEXP_CACHE_SIZE
; ++i
)
3066 searchbufs
[i
].buf
.allocated
= 100;
3067 searchbufs
[i
].buf
.buffer
= (unsigned char *) xmalloc (100);
3068 searchbufs
[i
].buf
.fastmap
= searchbufs
[i
].fastmap
;
3069 searchbufs
[i
].regexp
= Qnil
;
3070 searchbufs
[i
].whitespace_regexp
= Qnil
;
3071 staticpro (&searchbufs
[i
].regexp
);
3072 staticpro (&searchbufs
[i
].whitespace_regexp
);
3073 searchbufs
[i
].next
= (i
== REGEXP_CACHE_SIZE
-1 ? 0 : &searchbufs
[i
+1]);
3075 searchbuf_head
= &searchbufs
[0];
3077 Qsearch_failed
= intern ("search-failed");
3078 staticpro (&Qsearch_failed
);
3079 Qinvalid_regexp
= intern ("invalid-regexp");
3080 staticpro (&Qinvalid_regexp
);
3082 Fput (Qsearch_failed
, Qerror_conditions
,
3083 Fcons (Qsearch_failed
, Fcons (Qerror
, Qnil
)));
3084 Fput (Qsearch_failed
, Qerror_message
,
3085 build_string ("Search failed"));
3087 Fput (Qinvalid_regexp
, Qerror_conditions
,
3088 Fcons (Qinvalid_regexp
, Fcons (Qerror
, Qnil
)));
3089 Fput (Qinvalid_regexp
, Qerror_message
,
3090 build_string ("Invalid regexp"));
3092 last_thing_searched
= Qnil
;
3093 staticpro (&last_thing_searched
);
3095 saved_last_thing_searched
= Qnil
;
3096 staticpro (&saved_last_thing_searched
);
3098 DEFVAR_LISP ("search-spaces-regexp", &Vsearch_spaces_regexp
,
3099 doc
: /* Regexp to substitute for bunches of spaces in regexp search.
3100 Some commands use this for user-specified regexps.
3101 Spaces that occur inside character classes or repetition operators
3102 or other such regexp constructs are not replaced with this.
3103 A value of nil (which is the normal value) means treat spaces literally. */);
3104 Vsearch_spaces_regexp
= Qnil
;
3106 defsubr (&Slooking_at
);
3107 defsubr (&Sposix_looking_at
);
3108 defsubr (&Sstring_match
);
3109 defsubr (&Sposix_string_match
);
3110 defsubr (&Ssearch_forward
);
3111 defsubr (&Ssearch_backward
);
3112 defsubr (&Sword_search_forward
);
3113 defsubr (&Sword_search_backward
);
3114 defsubr (&Sre_search_forward
);
3115 defsubr (&Sre_search_backward
);
3116 defsubr (&Sposix_search_forward
);
3117 defsubr (&Sposix_search_backward
);
3118 defsubr (&Sreplace_match
);
3119 defsubr (&Smatch_beginning
);
3120 defsubr (&Smatch_end
);
3121 defsubr (&Smatch_data
);
3122 defsubr (&Sset_match_data
);
3123 defsubr (&Sregexp_quote
);
3126 /* arch-tag: a6059d79-0552-4f14-a2cb-d379a4e3c78f
3127 (do not change this comment) */