1 /* String search routines for GNU Emacs.
3 Copyright (C) 1985-1987, 1993-1994, 1997-1999, 2001-2012
4 Free Software Foundation, Inc.
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/>. */
28 #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 /* Nonzero 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
, int);
104 static void matcher_overflow (void) NO_RETURN
;
107 matcher_overflow (void)
109 error ("Stack overflow in regexp matcher");
112 /* Compile a regexp and signal a Lisp error if anything goes wrong.
113 PATTERN is the pattern to compile.
114 CP is the place to put the result.
115 TRANSLATE is a translation table for ignoring case, or nil for none.
116 POSIX is nonzero if we want full backtracking (POSIX style)
117 for this pattern. 0 means backtrack only enough to get a valid match.
119 The behavior also depends on Vsearch_spaces_regexp. */
122 compile_pattern_1 (struct regexp_cache
*cp
, Lisp_Object pattern
, Lisp_Object translate
, int 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
);
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
;
180 = (unsigned char *) xrealloc (cp
->buf
.buffer
, cp
->buf
.used
);
184 /* Clear the regexp cache w.r.t. a particular syntax table,
185 because it was changed.
186 There is no danger of memory leak here because re_compile_pattern
187 automagically manages the memory in each re_pattern_buffer struct,
188 based on its `allocated' and `buffer' values. */
190 clear_regexp_cache (void)
194 for (i
= 0; i
< REGEXP_CACHE_SIZE
; ++i
)
195 /* It's tempting to compare with the syntax-table we've actually changed,
196 but it's not sufficient because char-table inheritance means that
197 modifying one syntax-table can change others at the same time. */
198 if (!EQ (searchbufs
[i
].syntax_table
, Qt
))
199 searchbufs
[i
].regexp
= Qnil
;
202 /* Compile a regexp if necessary, but first check to see if there's one in
204 PATTERN is the pattern to compile.
205 TRANSLATE is a translation table for ignoring case, or nil for none.
206 REGP is the structure that says where to store the "register"
207 values that will result from matching this pattern.
208 If it is 0, we should compile the pattern not to record any
209 subexpression bounds.
210 POSIX is nonzero if we want full backtracking (POSIX style)
211 for this pattern. 0 means backtrack only enough to get a valid match. */
213 struct re_pattern_buffer
*
214 compile_pattern (Lisp_Object pattern
, struct re_registers
*regp
, Lisp_Object translate
, int posix
, int 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
, int 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 XCHAR_TABLE (BVAR (current_buffer
, case_canon_table
))->extras
[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
= (0 <= i
? Qt
: Qnil
);
330 if (NILP (Vinhibit_changing_match_data
) && i
>= 0)
331 for (i
= 0; i
< search_regs
.num_regs
; i
++)
332 if (search_regs
.start
[i
] >= 0)
335 = BYTE_TO_CHAR (search_regs
.start
[i
] + BEGV_BYTE
);
337 = BYTE_TO_CHAR (search_regs
.end
[i
] + BEGV_BYTE
);
340 /* Set last_thing_searched only when match data is changed. */
341 if (NILP (Vinhibit_changing_match_data
))
342 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
, int posix
)
372 struct re_pattern_buffer
*bufp
;
374 ptrdiff_t pos_byte
, i
;
376 if (running_asynch_code
)
379 CHECK_STRING (regexp
);
380 CHECK_STRING (string
);
383 pos
= 0, pos_byte
= 0;
386 ptrdiff_t len
= SCHARS (string
);
388 CHECK_NUMBER (start
);
390 if (pos
< 0 && -pos
<= len
)
392 else if (0 > pos
|| pos
> len
)
393 args_out_of_range (string
, start
);
394 pos_byte
= string_char_to_byte (string
, pos
);
397 /* This is so set_image_of_range_1 in regex.c can find the EQV table. */
398 XCHAR_TABLE (BVAR (current_buffer
, case_canon_table
))->extras
[2]
399 = BVAR (current_buffer
, case_eqv_table
);
401 bufp
= compile_pattern (regexp
,
402 (NILP (Vinhibit_changing_match_data
)
403 ? &search_regs
: NULL
),
404 (!NILP (BVAR (current_buffer
, case_fold_search
))
405 ? BVAR (current_buffer
, case_canon_table
) : Qnil
),
407 STRING_MULTIBYTE (string
));
409 re_match_object
= string
;
411 val
= re_search (bufp
, SSDATA (string
),
412 SBYTES (string
), pos_byte
,
413 SBYTES (string
) - pos_byte
,
414 (NILP (Vinhibit_changing_match_data
)
415 ? &search_regs
: NULL
));
418 /* Set last_thing_searched only when match data is changed. */
419 if (NILP (Vinhibit_changing_match_data
))
420 last_thing_searched
= Qt
;
424 if (val
< 0) return Qnil
;
426 if (NILP (Vinhibit_changing_match_data
))
427 for (i
= 0; i
< search_regs
.num_regs
; i
++)
428 if (search_regs
.start
[i
] >= 0)
431 = string_byte_to_char (string
, search_regs
.start
[i
]);
433 = string_byte_to_char (string
, search_regs
.end
[i
]);
436 return make_number (string_byte_to_char (string
, val
));
439 DEFUN ("string-match", Fstring_match
, Sstring_match
, 2, 3, 0,
440 doc
: /* Return index of start of first match for REGEXP in STRING, or nil.
441 Matching ignores case if `case-fold-search' is non-nil.
442 If third arg START is non-nil, start search at that index in STRING.
443 For index of first char beyond the match, do (match-end 0).
444 `match-end' and `match-beginning' also give indices of substrings
445 matched by parenthesis constructs in the pattern.
447 You can use the function `match-string' to extract the substrings
448 matched by the parenthesis constructions in REGEXP. */)
449 (Lisp_Object regexp
, Lisp_Object string
, Lisp_Object start
)
451 return string_match_1 (regexp
, string
, start
, 0);
454 DEFUN ("posix-string-match", Fposix_string_match
, Sposix_string_match
, 2, 3, 0,
455 doc
: /* Return index of start of first match for REGEXP in STRING, or nil.
456 Find the longest match, in accord with Posix regular expression rules.
457 Case is ignored if `case-fold-search' is non-nil in the current buffer.
458 If third arg START is non-nil, start search at that index in STRING.
459 For index of first char beyond the match, do (match-end 0).
460 `match-end' and `match-beginning' also give indices of substrings
461 matched by parenthesis constructs in the pattern. */)
462 (Lisp_Object regexp
, Lisp_Object string
, Lisp_Object start
)
464 return string_match_1 (regexp
, string
, start
, 1);
467 /* Match REGEXP against STRING, searching all of STRING,
468 and return the index of the match, or negative on failure.
469 This does not clobber the match data. */
472 fast_string_match (Lisp_Object regexp
, Lisp_Object string
)
475 struct re_pattern_buffer
*bufp
;
477 bufp
= compile_pattern (regexp
, 0, Qnil
,
478 0, STRING_MULTIBYTE (string
));
480 re_match_object
= string
;
482 val
= re_search (bufp
, SSDATA (string
),
489 /* Match REGEXP against STRING, searching all of STRING ignoring case,
490 and return the index of the match, or negative on failure.
491 This does not clobber the match data.
492 We assume that STRING contains single-byte characters. */
495 fast_c_string_match_ignore_case (Lisp_Object regexp
, const char *string
)
498 struct re_pattern_buffer
*bufp
;
499 size_t len
= strlen (string
);
501 regexp
= string_make_unibyte (regexp
);
502 re_match_object
= Qt
;
503 bufp
= compile_pattern (regexp
, 0,
504 Vascii_canon_table
, 0,
507 val
= re_search (bufp
, string
, len
, 0, len
, 0);
512 /* Like fast_string_match but ignore case. */
515 fast_string_match_ignore_case (Lisp_Object regexp
, Lisp_Object string
)
518 struct re_pattern_buffer
*bufp
;
520 bufp
= compile_pattern (regexp
, 0, Vascii_canon_table
,
521 0, STRING_MULTIBYTE (string
));
523 re_match_object
= string
;
525 val
= re_search (bufp
, SSDATA (string
),
532 /* Match REGEXP against the characters after POS to LIMIT, and return
533 the number of matched characters. If STRING is non-nil, match
534 against the characters in it. In that case, POS and LIMIT are
535 indices into the string. This function doesn't modify the match
539 fast_looking_at (Lisp_Object regexp
, ptrdiff_t pos
, ptrdiff_t pos_byte
, ptrdiff_t limit
, ptrdiff_t limit_byte
, Lisp_Object string
)
542 struct re_pattern_buffer
*buf
;
543 unsigned char *p1
, *p2
;
547 if (STRINGP (string
))
550 pos_byte
= string_char_to_byte (string
, pos
);
552 limit_byte
= string_char_to_byte (string
, limit
);
556 s2
= SBYTES (string
);
557 re_match_object
= string
;
558 multibyte
= STRING_MULTIBYTE (string
);
563 pos_byte
= CHAR_TO_BYTE (pos
);
565 limit_byte
= CHAR_TO_BYTE (limit
);
566 pos_byte
-= BEGV_BYTE
;
567 limit_byte
-= BEGV_BYTE
;
569 s1
= GPT_BYTE
- BEGV_BYTE
;
571 s2
= ZV_BYTE
- GPT_BYTE
;
575 s2
= ZV_BYTE
- BEGV_BYTE
;
580 s1
= ZV_BYTE
- BEGV_BYTE
;
583 re_match_object
= Qnil
;
584 multibyte
= ! NILP (BVAR (current_buffer
, enable_multibyte_characters
));
587 buf
= compile_pattern (regexp
, 0, Qnil
, 0, multibyte
);
589 len
= re_match_2 (buf
, (char *) p1
, s1
, (char *) p2
, s2
,
590 pos_byte
, NULL
, limit_byte
);
597 /* The newline cache: remembering which sections of text have no newlines. */
599 /* If the user has requested newline caching, make sure it's on.
600 Otherwise, make sure it's off.
601 This is our cheezy way of associating an action with the change of
602 state of a buffer-local variable. */
604 newline_cache_on_off (struct buffer
*buf
)
606 if (NILP (BVAR (buf
, cache_long_line_scans
)))
608 /* It should be off. */
609 if (buf
->newline_cache
)
611 free_region_cache (buf
->newline_cache
);
612 buf
->newline_cache
= 0;
617 /* It should be on. */
618 if (buf
->newline_cache
== 0)
619 buf
->newline_cache
= new_region_cache ();
624 /* Search for COUNT instances of the character TARGET between START and END.
626 If COUNT is positive, search forwards; END must be >= START.
627 If COUNT is negative, search backwards for the -COUNTth instance;
628 END must be <= START.
629 If COUNT is zero, do anything you please; run rogue, for all I care.
631 If END is zero, use BEGV or ZV instead, as appropriate for the
632 direction indicated by COUNT.
634 If we find COUNT instances, set *SHORTAGE to zero, and return the
635 position past the COUNTth match. Note that for reverse motion
636 this is not the same as the usual convention for Emacs motion commands.
638 If we don't find COUNT instances before reaching END, set *SHORTAGE
639 to the number of TARGETs left unfound, and return END.
641 If ALLOW_QUIT is non-zero, set immediate_quit. That's good to do
642 except when inside redisplay. */
645 scan_buffer (register int target
, ptrdiff_t start
, ptrdiff_t end
,
646 ptrdiff_t count
, ptrdiff_t *shortage
, int allow_quit
)
648 struct region_cache
*newline_cache
;
659 if (! end
) end
= BEGV
;
662 newline_cache_on_off (current_buffer
);
663 newline_cache
= current_buffer
->newline_cache
;
668 immediate_quit
= allow_quit
;
673 /* Our innermost scanning loop is very simple; it doesn't know
674 about gaps, buffer ends, or the newline cache. ceiling is
675 the position of the last character before the next such
676 obstacle --- the last character the dumb search loop should
678 ptrdiff_t ceiling_byte
= CHAR_TO_BYTE (end
) - 1;
679 ptrdiff_t start_byte
= CHAR_TO_BYTE (start
);
682 /* If we're looking for a newline, consult the newline cache
683 to see where we can avoid some scanning. */
684 if (target
== '\n' && newline_cache
)
686 ptrdiff_t next_change
;
688 while (region_cache_forward
689 (current_buffer
, newline_cache
, start_byte
, &next_change
))
690 start_byte
= next_change
;
691 immediate_quit
= allow_quit
;
693 /* START should never be after END. */
694 if (start_byte
> ceiling_byte
)
695 start_byte
= ceiling_byte
;
697 /* Now the text after start is an unknown region, and
698 next_change is the position of the next known region. */
699 ceiling_byte
= min (next_change
- 1, ceiling_byte
);
702 /* The dumb loop can only scan text stored in contiguous
703 bytes. BUFFER_CEILING_OF returns the last character
704 position that is contiguous, so the ceiling is the
705 position after that. */
706 tem
= BUFFER_CEILING_OF (start_byte
);
707 ceiling_byte
= min (tem
, ceiling_byte
);
710 /* The termination address of the dumb loop. */
711 register unsigned char *ceiling_addr
712 = BYTE_POS_ADDR (ceiling_byte
) + 1;
713 register unsigned char *cursor
714 = BYTE_POS_ADDR (start_byte
);
715 unsigned char *base
= cursor
;
717 while (cursor
< ceiling_addr
)
719 unsigned char *scan_start
= cursor
;
722 while (*cursor
!= target
&& ++cursor
< ceiling_addr
)
725 /* If we're looking for newlines, cache the fact that
726 the region from start to cursor is free of them. */
727 if (target
== '\n' && newline_cache
)
728 know_region_cache (current_buffer
, newline_cache
,
729 BYTE_TO_CHAR (start_byte
+ scan_start
- base
),
730 BYTE_TO_CHAR (start_byte
+ cursor
- base
));
732 /* Did we find the target character? */
733 if (cursor
< ceiling_addr
)
738 return BYTE_TO_CHAR (start_byte
+ cursor
- base
+ 1);
744 start
= BYTE_TO_CHAR (start_byte
+ cursor
- base
);
750 /* The last character to check before the next obstacle. */
751 ptrdiff_t ceiling_byte
= CHAR_TO_BYTE (end
);
752 ptrdiff_t start_byte
= CHAR_TO_BYTE (start
);
755 /* Consult the newline cache, if appropriate. */
756 if (target
== '\n' && newline_cache
)
758 ptrdiff_t next_change
;
760 while (region_cache_backward
761 (current_buffer
, newline_cache
, start_byte
, &next_change
))
762 start_byte
= next_change
;
763 immediate_quit
= allow_quit
;
765 /* Start should never be at or before end. */
766 if (start_byte
<= ceiling_byte
)
767 start_byte
= ceiling_byte
+ 1;
769 /* Now the text before start is an unknown region, and
770 next_change is the position of the next known region. */
771 ceiling_byte
= max (next_change
, ceiling_byte
);
774 /* Stop scanning before the gap. */
775 tem
= BUFFER_FLOOR_OF (start_byte
- 1);
776 ceiling_byte
= max (tem
, ceiling_byte
);
779 /* The termination address of the dumb loop. */
780 register unsigned char *ceiling_addr
= BYTE_POS_ADDR (ceiling_byte
);
781 register unsigned char *cursor
= BYTE_POS_ADDR (start_byte
- 1);
782 unsigned char *base
= cursor
;
784 while (cursor
>= ceiling_addr
)
786 unsigned char *scan_start
= cursor
;
788 while (*cursor
!= target
&& --cursor
>= ceiling_addr
)
791 /* If we're looking for newlines, cache the fact that
792 the region from after the cursor to start is free of them. */
793 if (target
== '\n' && newline_cache
)
794 know_region_cache (current_buffer
, newline_cache
,
795 BYTE_TO_CHAR (start_byte
+ cursor
- base
),
796 BYTE_TO_CHAR (start_byte
+ scan_start
- base
));
798 /* Did we find the target character? */
799 if (cursor
>= ceiling_addr
)
804 return BYTE_TO_CHAR (start_byte
+ cursor
- base
);
810 start
= BYTE_TO_CHAR (start_byte
+ cursor
- base
);
816 *shortage
= count
* direction
;
820 /* Search for COUNT instances of a line boundary, which means either a
821 newline or (if selective display enabled) a carriage return.
822 Start at START. If COUNT is negative, search backwards.
824 We report the resulting position by calling TEMP_SET_PT_BOTH.
826 If we find COUNT instances. we position after (always after,
827 even if scanning backwards) the COUNTth match, and return 0.
829 If we don't find COUNT instances before reaching the end of the
830 buffer (or the beginning, if scanning backwards), we return
831 the number of line boundaries left unfound, and position at
832 the limit we bumped up against.
834 If ALLOW_QUIT is non-zero, set immediate_quit. That's good to do
835 except in special cases. */
838 scan_newline (ptrdiff_t start
, ptrdiff_t start_byte
,
839 ptrdiff_t limit
, ptrdiff_t limit_byte
,
840 register EMACS_INT count
, int allow_quit
)
842 int direction
= ((count
> 0) ? 1 : -1);
844 register unsigned char *cursor
;
848 register unsigned char *ceiling_addr
;
850 int old_immediate_quit
= immediate_quit
;
852 /* The code that follows is like scan_buffer
853 but checks for either newline or carriage return. */
858 start_byte
= CHAR_TO_BYTE (start
);
862 while (start_byte
< limit_byte
)
864 ceiling
= BUFFER_CEILING_OF (start_byte
);
865 ceiling
= min (limit_byte
- 1, ceiling
);
866 ceiling_addr
= BYTE_POS_ADDR (ceiling
) + 1;
867 base
= (cursor
= BYTE_POS_ADDR (start_byte
));
870 while (*cursor
!= '\n' && ++cursor
!= ceiling_addr
)
873 if (cursor
!= ceiling_addr
)
877 immediate_quit
= old_immediate_quit
;
878 start_byte
= start_byte
+ cursor
- base
+ 1;
879 start
= BYTE_TO_CHAR (start_byte
);
880 TEMP_SET_PT_BOTH (start
, start_byte
);
884 if (++cursor
== ceiling_addr
)
890 start_byte
+= cursor
- base
;
895 while (start_byte
> limit_byte
)
897 ceiling
= BUFFER_FLOOR_OF (start_byte
- 1);
898 ceiling
= max (limit_byte
, ceiling
);
899 ceiling_addr
= BYTE_POS_ADDR (ceiling
) - 1;
900 base
= (cursor
= BYTE_POS_ADDR (start_byte
- 1) + 1);
903 while (--cursor
!= ceiling_addr
&& *cursor
!= '\n')
906 if (cursor
!= ceiling_addr
)
910 immediate_quit
= old_immediate_quit
;
911 /* Return the position AFTER the match we found. */
912 start_byte
= start_byte
+ cursor
- base
+ 1;
913 start
= BYTE_TO_CHAR (start_byte
);
914 TEMP_SET_PT_BOTH (start
, start_byte
);
921 /* Here we add 1 to compensate for the last decrement
922 of CURSOR, which took it past the valid range. */
923 start_byte
+= cursor
- base
+ 1;
927 TEMP_SET_PT_BOTH (limit
, limit_byte
);
928 immediate_quit
= old_immediate_quit
;
930 return count
* direction
;
934 find_next_newline_no_quit (ptrdiff_t from
, ptrdiff_t cnt
)
936 return scan_buffer ('\n', from
, 0, cnt
, (ptrdiff_t *) 0, 0);
939 /* Like find_next_newline, but returns position before the newline,
940 not after, and only search up to TO. This isn't just
941 find_next_newline (...)-1, because you might hit TO. */
944 find_before_next_newline (ptrdiff_t from
, ptrdiff_t to
, ptrdiff_t cnt
)
947 ptrdiff_t pos
= scan_buffer ('\n', from
, to
, cnt
, &shortage
, 1);
955 /* Subroutines of Lisp buffer search functions. */
958 search_command (Lisp_Object string
, Lisp_Object bound
, Lisp_Object noerror
,
959 Lisp_Object count
, int direction
, int RE
, int posix
)
961 register EMACS_INT np
;
964 EMACS_INT n
= direction
;
968 CHECK_NUMBER (count
);
972 CHECK_STRING (string
);
976 lim
= ZV
, lim_byte
= ZV_BYTE
;
978 lim
= BEGV
, lim_byte
= BEGV_BYTE
;
982 CHECK_NUMBER_COERCE_MARKER (bound
);
984 if (n
> 0 ? lim
< PT
: lim
> PT
)
985 error ("Invalid search bound (wrong side of point)");
987 lim
= ZV
, lim_byte
= ZV_BYTE
;
989 lim
= BEGV
, lim_byte
= BEGV_BYTE
;
991 lim_byte
= CHAR_TO_BYTE (lim
);
994 /* This is so set_image_of_range_1 in regex.c can find the EQV table. */
995 XCHAR_TABLE (BVAR (current_buffer
, case_canon_table
))->extras
[2]
996 = BVAR (current_buffer
, case_eqv_table
);
998 np
= search_buffer (string
, PT
, PT_BYTE
, lim
, lim_byte
, n
, RE
,
999 (!NILP (BVAR (current_buffer
, case_fold_search
))
1000 ? BVAR (current_buffer
, case_canon_table
)
1002 (!NILP (BVAR (current_buffer
, case_fold_search
))
1003 ? BVAR (current_buffer
, case_eqv_table
)
1009 xsignal1 (Qsearch_failed
, string
);
1011 if (!EQ (noerror
, Qt
))
1013 if (lim
< BEGV
|| lim
> ZV
)
1015 SET_PT_BOTH (lim
, lim_byte
);
1017 #if 0 /* This would be clean, but maybe programs depend on
1018 a value of nil here. */
1026 if (np
< BEGV
|| np
> ZV
)
1031 return make_number (np
);
1034 /* Return 1 if REGEXP it matches just one constant string. */
1037 trivial_regexp_p (Lisp_Object regexp
)
1039 ptrdiff_t len
= SBYTES (regexp
);
1040 unsigned char *s
= SDATA (regexp
);
1045 case '.': case '*': case '+': case '?': case '[': case '^': case '$':
1052 case '|': case '(': case ')': case '`': case '\'': case 'b':
1053 case 'B': case '<': case '>': case 'w': case 'W': case 's':
1054 case 'S': case '=': case '{': case '}': case '_':
1055 case 'c': case 'C': /* for categoryspec and notcategoryspec */
1056 case '1': case '2': case '3': case '4': case '5':
1057 case '6': case '7': case '8': case '9':
1065 /* Search for the n'th occurrence of STRING in the current buffer,
1066 starting at position POS and stopping at position LIM,
1067 treating STRING as a literal string if RE is false or as
1068 a regular expression if RE is true.
1070 If N is positive, searching is forward and LIM must be greater than POS.
1071 If N is negative, searching is backward and LIM must be less than POS.
1073 Returns -x if x occurrences remain to be found (x > 0),
1074 or else the position at the beginning of the Nth occurrence
1075 (if searching backward) or the end (if searching forward).
1077 POSIX is nonzero if we want full backtracking (POSIX style)
1078 for this pattern. 0 means backtrack only enough to get a valid match. */
1080 #define TRANSLATE(out, trt, d) \
1086 temp = Faref (trt, make_number (d)); \
1087 if (INTEGERP (temp)) \
1088 out = XINT (temp); \
1097 /* Only used in search_buffer, to record the end position of the match
1098 when searching regexps and SEARCH_REGS should not be changed
1099 (i.e. Vinhibit_changing_match_data is non-nil). */
1100 static struct re_registers search_regs_1
;
1103 search_buffer (Lisp_Object string
, ptrdiff_t pos
, ptrdiff_t pos_byte
,
1104 ptrdiff_t lim
, ptrdiff_t lim_byte
, EMACS_INT n
,
1105 int RE
, Lisp_Object trt
, Lisp_Object inverse_trt
, int posix
)
1107 ptrdiff_t len
= SCHARS (string
);
1108 ptrdiff_t len_byte
= SBYTES (string
);
1109 register ptrdiff_t i
;
1111 if (running_asynch_code
)
1112 save_search_regs ();
1114 /* Searching 0 times means don't move. */
1115 /* Null string is found at starting position. */
1116 if (len
== 0 || n
== 0)
1118 set_search_regs (pos_byte
, 0);
1122 if (RE
&& !(trivial_regexp_p (string
) && NILP (Vsearch_spaces_regexp
)))
1124 unsigned char *p1
, *p2
;
1126 struct re_pattern_buffer
*bufp
;
1128 bufp
= compile_pattern (string
,
1129 (NILP (Vinhibit_changing_match_data
)
1130 ? &search_regs
: &search_regs_1
),
1132 !NILP (BVAR (current_buffer
, enable_multibyte_characters
)));
1134 immediate_quit
= 1; /* Quit immediately if user types ^G,
1135 because letting this function finish
1136 can take too long. */
1137 QUIT
; /* Do a pending quit right away,
1138 to avoid paradoxical behavior */
1139 /* Get pointers and sizes of the two strings
1140 that make up the visible portion of the buffer. */
1143 s1
= GPT_BYTE
- BEGV_BYTE
;
1145 s2
= ZV_BYTE
- GPT_BYTE
;
1149 s2
= ZV_BYTE
- BEGV_BYTE
;
1154 s1
= ZV_BYTE
- BEGV_BYTE
;
1157 re_match_object
= Qnil
;
1162 val
= re_search_2 (bufp
, (char *) p1
, s1
, (char *) p2
, s2
,
1163 pos_byte
- BEGV_BYTE
, lim_byte
- pos_byte
,
1164 (NILP (Vinhibit_changing_match_data
)
1165 ? &search_regs
: &search_regs_1
),
1166 /* Don't allow match past current point */
1167 pos_byte
- BEGV_BYTE
);
1170 matcher_overflow ();
1174 if (NILP (Vinhibit_changing_match_data
))
1176 pos_byte
= search_regs
.start
[0] + BEGV_BYTE
;
1177 for (i
= 0; i
< search_regs
.num_regs
; i
++)
1178 if (search_regs
.start
[i
] >= 0)
1180 search_regs
.start
[i
]
1181 = BYTE_TO_CHAR (search_regs
.start
[i
] + BEGV_BYTE
);
1183 = BYTE_TO_CHAR (search_regs
.end
[i
] + BEGV_BYTE
);
1185 XSETBUFFER (last_thing_searched
, current_buffer
);
1186 /* Set pos to the new position. */
1187 pos
= search_regs
.start
[0];
1191 pos_byte
= search_regs_1
.start
[0] + BEGV_BYTE
;
1192 /* Set pos to the new position. */
1193 pos
= BYTE_TO_CHAR (search_regs_1
.start
[0] + BEGV_BYTE
);
1206 val
= re_search_2 (bufp
, (char *) p1
, s1
, (char *) p2
, s2
,
1207 pos_byte
- BEGV_BYTE
, lim_byte
- pos_byte
,
1208 (NILP (Vinhibit_changing_match_data
)
1209 ? &search_regs
: &search_regs_1
),
1210 lim_byte
- BEGV_BYTE
);
1213 matcher_overflow ();
1217 if (NILP (Vinhibit_changing_match_data
))
1219 pos_byte
= search_regs
.end
[0] + BEGV_BYTE
;
1220 for (i
= 0; i
< search_regs
.num_regs
; i
++)
1221 if (search_regs
.start
[i
] >= 0)
1223 search_regs
.start
[i
]
1224 = BYTE_TO_CHAR (search_regs
.start
[i
] + BEGV_BYTE
);
1226 = BYTE_TO_CHAR (search_regs
.end
[i
] + BEGV_BYTE
);
1228 XSETBUFFER (last_thing_searched
, current_buffer
);
1229 pos
= search_regs
.end
[0];
1233 pos_byte
= search_regs_1
.end
[0] + BEGV_BYTE
;
1234 pos
= BYTE_TO_CHAR (search_regs_1
.end
[0] + BEGV_BYTE
);
1247 else /* non-RE case */
1249 unsigned char *raw_pattern
, *pat
;
1250 ptrdiff_t raw_pattern_size
;
1251 ptrdiff_t raw_pattern_size_byte
;
1252 unsigned char *patbuf
;
1253 int multibyte
= !NILP (BVAR (current_buffer
, enable_multibyte_characters
));
1254 unsigned char *base_pat
;
1255 /* Set to positive if we find a non-ASCII char that need
1256 translation. Otherwise set to zero later. */
1258 int boyer_moore_ok
= 1;
1260 /* MULTIBYTE says whether the text to be searched is multibyte.
1261 We must convert PATTERN to match that, or we will not really
1262 find things right. */
1264 if (multibyte
== STRING_MULTIBYTE (string
))
1266 raw_pattern
= SDATA (string
);
1267 raw_pattern_size
= SCHARS (string
);
1268 raw_pattern_size_byte
= SBYTES (string
);
1272 raw_pattern_size
= SCHARS (string
);
1273 raw_pattern_size_byte
1274 = count_size_as_multibyte (SDATA (string
),
1276 raw_pattern
= (unsigned char *) alloca (raw_pattern_size_byte
+ 1);
1277 copy_text (SDATA (string
), raw_pattern
,
1278 SCHARS (string
), 0, 1);
1282 /* Converting multibyte to single-byte.
1284 ??? Perhaps this conversion should be done in a special way
1285 by subtracting nonascii-insert-offset from each non-ASCII char,
1286 so that only the multibyte chars which really correspond to
1287 the chosen single-byte character set can possibly match. */
1288 raw_pattern_size
= SCHARS (string
);
1289 raw_pattern_size_byte
= SCHARS (string
);
1290 raw_pattern
= (unsigned char *) alloca (raw_pattern_size
+ 1);
1291 copy_text (SDATA (string
), raw_pattern
,
1292 SBYTES (string
), 1, 0);
1295 /* Copy and optionally translate the pattern. */
1296 len
= raw_pattern_size
;
1297 len_byte
= raw_pattern_size_byte
;
1298 patbuf
= (unsigned char *) alloca (len
* MAX_MULTIBYTE_LENGTH
);
1300 base_pat
= raw_pattern
;
1303 /* Fill patbuf by translated characters in STRING while
1304 checking if we can use boyer-moore search. If TRT is
1305 non-nil, we can use boyer-moore search only if TRT can be
1306 represented by the byte array of 256 elements. For that,
1307 all non-ASCII case-equivalents of all case-sensitive
1308 characters in STRING must belong to the same charset and
1313 unsigned char str_base
[MAX_MULTIBYTE_LENGTH
], *str
;
1314 int c
, translated
, inverse
;
1315 int in_charlen
, charlen
;
1317 /* If we got here and the RE flag is set, it's because we're
1318 dealing with a regexp known to be trivial, so the backslash
1319 just quotes the next character. */
1320 if (RE
&& *base_pat
== '\\')
1328 c
= STRING_CHAR_AND_LENGTH (base_pat
, in_charlen
);
1333 charlen
= in_charlen
;
1337 /* Translate the character. */
1338 TRANSLATE (translated
, trt
, c
);
1339 charlen
= CHAR_STRING (translated
, str_base
);
1342 /* Check if C has any other case-equivalents. */
1343 TRANSLATE (inverse
, inverse_trt
, c
);
1344 /* If so, check if we can use boyer-moore. */
1345 if (c
!= inverse
&& boyer_moore_ok
)
1347 /* Check if all equivalents belong to the same
1348 group of characters. Note that the check of C
1349 itself is done by the last iteration. */
1350 int this_char_base
= -1;
1352 while (boyer_moore_ok
)
1354 if (ASCII_BYTE_P (inverse
))
1356 if (this_char_base
> 0)
1361 else if (CHAR_BYTE8_P (inverse
))
1362 /* Boyer-moore search can't handle a
1363 translation of an eight-bit
1366 else if (this_char_base
< 0)
1368 this_char_base
= inverse
& ~0x3F;
1370 char_base
= this_char_base
;
1371 else if (this_char_base
!= char_base
)
1374 else if ((inverse
& ~0x3F) != this_char_base
)
1378 TRANSLATE (inverse
, inverse_trt
, inverse
);
1383 /* Store this character into the translated pattern. */
1384 memcpy (pat
, str
, charlen
);
1386 base_pat
+= in_charlen
;
1387 len_byte
-= in_charlen
;
1390 /* If char_base is still negative we didn't find any translated
1391 non-ASCII characters. */
1397 /* Unibyte buffer. */
1403 /* If we got here and the RE flag is set, it's because we're
1404 dealing with a regexp known to be trivial, so the backslash
1405 just quotes the next character. */
1406 if (RE
&& *base_pat
== '\\')
1413 TRANSLATE (translated
, trt
, c
);
1414 *pat
++ = translated
;
1418 len_byte
= pat
- patbuf
;
1419 pat
= base_pat
= patbuf
;
1422 return boyer_moore (n
, pat
, len_byte
, trt
, inverse_trt
,
1426 return simple_search (n
, pat
, raw_pattern_size
, len_byte
, trt
,
1427 pos
, pos_byte
, lim
, lim_byte
);
1431 /* Do a simple string search N times for the string PAT,
1432 whose length is LEN/LEN_BYTE,
1433 from buffer position POS/POS_BYTE until LIM/LIM_BYTE.
1434 TRT is the translation table.
1436 Return the character position where the match is found.
1437 Otherwise, if M matches remained to be found, return -M.
1439 This kind of search works regardless of what is in PAT and
1440 regardless of what is in TRT. It is used in cases where
1441 boyer_moore cannot work. */
1444 simple_search (EMACS_INT n
, unsigned char *pat
,
1445 ptrdiff_t len
, ptrdiff_t len_byte
, Lisp_Object trt
,
1446 ptrdiff_t pos
, ptrdiff_t pos_byte
,
1447 ptrdiff_t lim
, ptrdiff_t lim_byte
)
1449 int multibyte
= ! NILP (BVAR (current_buffer
, enable_multibyte_characters
));
1450 int forward
= n
> 0;
1451 /* Number of buffer bytes matched. Note that this may be different
1452 from len_byte in a multibyte buffer. */
1453 ptrdiff_t match_byte
;
1455 if (lim
> pos
&& multibyte
)
1460 /* Try matching at position POS. */
1461 ptrdiff_t this_pos
= pos
;
1462 ptrdiff_t this_pos_byte
= pos_byte
;
1463 ptrdiff_t this_len
= len
;
1464 unsigned char *p
= pat
;
1465 if (pos
+ len
> lim
|| pos_byte
+ len_byte
> lim_byte
)
1468 while (this_len
> 0)
1470 int charlen
, buf_charlen
;
1473 pat_ch
= STRING_CHAR_AND_LENGTH (p
, charlen
);
1474 buf_ch
= STRING_CHAR_AND_LENGTH (BYTE_POS_ADDR (this_pos_byte
),
1476 TRANSLATE (buf_ch
, trt
, buf_ch
);
1478 if (buf_ch
!= pat_ch
)
1484 this_pos_byte
+= buf_charlen
;
1490 match_byte
= this_pos_byte
- pos_byte
;
1492 pos_byte
+= match_byte
;
1496 INC_BOTH (pos
, pos_byte
);
1506 /* Try matching at position POS. */
1507 ptrdiff_t this_pos
= pos
;
1508 ptrdiff_t this_len
= len
;
1509 unsigned char *p
= pat
;
1511 if (pos
+ len
> lim
)
1514 while (this_len
> 0)
1517 int buf_ch
= FETCH_BYTE (this_pos
);
1518 TRANSLATE (buf_ch
, trt
, buf_ch
);
1520 if (buf_ch
!= pat_ch
)
1539 /* Backwards search. */
1540 else if (lim
< pos
&& multibyte
)
1545 /* Try matching at position POS. */
1546 ptrdiff_t this_pos
= pos
;
1547 ptrdiff_t this_pos_byte
= pos_byte
;
1548 ptrdiff_t this_len
= len
;
1549 const unsigned char *p
= pat
+ len_byte
;
1551 if (this_pos
- len
< lim
|| (pos_byte
- len_byte
) < lim_byte
)
1554 while (this_len
> 0)
1558 DEC_BOTH (this_pos
, this_pos_byte
);
1559 PREV_CHAR_BOUNDARY (p
, pat
);
1560 pat_ch
= STRING_CHAR (p
);
1561 buf_ch
= STRING_CHAR (BYTE_POS_ADDR (this_pos_byte
));
1562 TRANSLATE (buf_ch
, trt
, buf_ch
);
1564 if (buf_ch
!= pat_ch
)
1572 match_byte
= pos_byte
- this_pos_byte
;
1574 pos_byte
= this_pos_byte
;
1578 DEC_BOTH (pos
, pos_byte
);
1588 /* Try matching at position POS. */
1589 ptrdiff_t this_pos
= pos
- len
;
1590 ptrdiff_t this_len
= len
;
1591 unsigned char *p
= pat
;
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
)
1625 set_search_regs ((multibyte
? pos_byte
: pos
) - match_byte
, match_byte
);
1627 set_search_regs (multibyte
? pos_byte
: pos
, match_byte
);
1637 /* Do Boyer-Moore search N times for the string BASE_PAT,
1638 whose length is LEN_BYTE,
1639 from buffer position POS_BYTE until LIM_BYTE.
1640 DIRECTION says which direction we search in.
1641 TRT and INVERSE_TRT are translation tables.
1642 Characters in PAT are already translated by TRT.
1644 This kind of search works if all the characters in BASE_PAT that
1645 have nontrivial translation are the same aside from the last byte.
1646 This makes it possible to translate just the last byte of a
1647 character, and do so after just a simple test of the context.
1648 CHAR_BASE is nonzero if there is such a non-ASCII character.
1650 If that criterion is not satisfied, do not call this function. */
1653 boyer_moore (EMACS_INT n
, unsigned char *base_pat
,
1655 Lisp_Object trt
, Lisp_Object inverse_trt
,
1656 ptrdiff_t pos_byte
, ptrdiff_t lim_byte
,
1659 int direction
= ((n
> 0) ? 1 : -1);
1660 register ptrdiff_t dirlen
;
1662 int stride_for_teases
= 0;
1664 register unsigned char *cursor
, *p_limit
;
1665 register ptrdiff_t i
;
1667 unsigned char *pat
, *pat_end
;
1668 int multibyte
= ! NILP (BVAR (current_buffer
, enable_multibyte_characters
));
1670 unsigned char simple_translate
[0400];
1671 /* These are set to the preceding bytes of a byte to be translated
1672 if char_base is nonzero. As the maximum byte length of a
1673 multibyte character is 5, we have to check at most four previous
1675 int translate_prev_byte1
= 0;
1676 int translate_prev_byte2
= 0;
1677 int translate_prev_byte3
= 0;
1679 /* The general approach is that we are going to maintain that we know
1680 the first (closest to the present position, in whatever direction
1681 we're searching) character that could possibly be the last
1682 (furthest from present position) character of a valid match. We
1683 advance the state of our knowledge by looking at that character
1684 and seeing whether it indeed matches the last character of the
1685 pattern. If it does, we take a closer look. If it does not, we
1686 move our pointer (to putative last characters) as far as is
1687 logically possible. This amount of movement, which I call a
1688 stride, will be the length of the pattern if the actual character
1689 appears nowhere in the pattern, otherwise it will be the distance
1690 from the last occurrence of that character to the end of the
1691 pattern. If the amount is zero we have a possible match. */
1693 /* Here we make a "mickey mouse" BM table. The stride of the search
1694 is determined only by the last character of the putative match.
1695 If that character does not match, we will stride the proper
1696 distance to propose a match that superimposes it on the last
1697 instance of a character that matches it (per trt), or misses
1698 it entirely if there is none. */
1700 dirlen
= len_byte
* direction
;
1702 /* Record position after the end of the pattern. */
1703 pat_end
= base_pat
+ len_byte
;
1704 /* BASE_PAT points to a character that we start scanning from.
1705 It is the first character in a forward search,
1706 the last character in a backward search. */
1708 base_pat
= pat_end
- 1;
1710 /* A character that does not appear in the pattern induces a
1711 stride equal to the pattern length. */
1712 for (i
= 0; i
< 0400; i
++)
1715 /* We use this for translation, instead of TRT itself.
1716 We fill this in to handle the characters that actually
1717 occur in the pattern. Others don't matter anyway! */
1718 for (i
= 0; i
< 0400; i
++)
1719 simple_translate
[i
] = i
;
1723 /* Setup translate_prev_byte1/2/3/4 from CHAR_BASE. Only a
1724 byte following them are the target of translation. */
1725 unsigned char str
[MAX_MULTIBYTE_LENGTH
];
1726 int cblen
= CHAR_STRING (char_base
, str
);
1728 translate_prev_byte1
= str
[cblen
- 2];
1731 translate_prev_byte2
= str
[cblen
- 3];
1733 translate_prev_byte3
= str
[cblen
- 4];
1740 unsigned char *ptr
= base_pat
+ i
;
1744 /* If the byte currently looking at is the last of a
1745 character to check case-equivalents, set CH to that
1746 character. An ASCII character and a non-ASCII character
1747 matching with CHAR_BASE are to be checked. */
1750 if (ASCII_BYTE_P (*ptr
) || ! multibyte
)
1753 && ((pat_end
- ptr
) == 1 || CHAR_HEAD_P (ptr
[1])))
1755 unsigned char *charstart
= ptr
- 1;
1757 while (! (CHAR_HEAD_P (*charstart
)))
1759 ch
= STRING_CHAR (charstart
);
1760 if (char_base
!= (ch
& ~0x3F))
1764 if (ch
>= 0200 && multibyte
)
1765 j
= (ch
& 0x3F) | 0200;
1770 stride_for_teases
= BM_tab
[j
];
1772 BM_tab
[j
] = dirlen
- i
;
1773 /* A translation table is accompanied by its inverse -- see
1774 comment following downcase_table for details. */
1777 int starting_ch
= ch
;
1782 TRANSLATE (ch
, inverse_trt
, ch
);
1783 if (ch
>= 0200 && multibyte
)
1784 j
= (ch
& 0x3F) | 0200;
1788 /* For all the characters that map into CH,
1789 set up simple_translate to map the last byte
1791 simple_translate
[j
] = starting_j
;
1792 if (ch
== starting_ch
)
1794 BM_tab
[j
] = dirlen
- i
;
1803 stride_for_teases
= BM_tab
[j
];
1804 BM_tab
[j
] = dirlen
- i
;
1806 /* stride_for_teases tells how much to stride if we get a
1807 match on the far character but are subsequently
1808 disappointed, by recording what the stride would have been
1809 for that character if the last character had been
1812 pos_byte
+= dirlen
- ((direction
> 0) ? direction
: 0);
1813 /* loop invariant - POS_BYTE points at where last char (first
1814 char if reverse) of pattern would align in a possible match. */
1818 unsigned char *tail_end_ptr
;
1820 /* It's been reported that some (broken) compiler thinks that
1821 Boolean expressions in an arithmetic context are unsigned.
1822 Using an explicit ?1:0 prevents this. */
1823 if ((lim_byte
- pos_byte
- ((direction
> 0) ? 1 : 0)) * direction
1825 return (n
* (0 - direction
));
1826 /* First we do the part we can by pointers (maybe nothing) */
1829 limit
= pos_byte
- dirlen
+ direction
;
1832 limit
= BUFFER_CEILING_OF (limit
);
1833 /* LIMIT is now the last (not beyond-last!) value POS_BYTE
1834 can take on without hitting edge of buffer or the gap. */
1835 limit
= min (limit
, pos_byte
+ 20000);
1836 limit
= min (limit
, lim_byte
- 1);
1840 limit
= BUFFER_FLOOR_OF (limit
);
1841 /* LIMIT is now the last (not beyond-last!) value POS_BYTE
1842 can take on without hitting edge of buffer or the gap. */
1843 limit
= max (limit
, pos_byte
- 20000);
1844 limit
= max (limit
, lim_byte
);
1846 tail_end
= BUFFER_CEILING_OF (pos_byte
) + 1;
1847 tail_end_ptr
= BYTE_POS_ADDR (tail_end
);
1849 if ((limit
- pos_byte
) * direction
> 20)
1853 p_limit
= BYTE_POS_ADDR (limit
);
1854 p2
= (cursor
= BYTE_POS_ADDR (pos_byte
));
1855 /* In this loop, pos + cursor - p2 is the surrogate for pos. */
1856 while (1) /* use one cursor setting as long as i can */
1858 if (direction
> 0) /* worth duplicating */
1860 while (cursor
<= p_limit
)
1862 if (BM_tab
[*cursor
] == 0)
1864 cursor
+= BM_tab
[*cursor
];
1869 while (cursor
>= p_limit
)
1871 if (BM_tab
[*cursor
] == 0)
1873 cursor
+= BM_tab
[*cursor
];
1876 /* If you are here, cursor is beyond the end of the
1877 searched region. You fail to match within the
1878 permitted region and would otherwise try a character
1879 beyond that region. */
1883 i
= dirlen
- direction
;
1886 while ((i
-= direction
) + direction
!= 0)
1889 cursor
-= direction
;
1890 /* Translate only the last byte of a character. */
1892 || ((cursor
== tail_end_ptr
1893 || CHAR_HEAD_P (cursor
[1]))
1894 && (CHAR_HEAD_P (cursor
[0])
1895 /* Check if this is the last byte of
1896 a translatable character. */
1897 || (translate_prev_byte1
== cursor
[-1]
1898 && (CHAR_HEAD_P (translate_prev_byte1
)
1899 || (translate_prev_byte2
== cursor
[-2]
1900 && (CHAR_HEAD_P (translate_prev_byte2
)
1901 || (translate_prev_byte3
== cursor
[-3]))))))))
1902 ch
= simple_translate
[*cursor
];
1911 while ((i
-= direction
) + direction
!= 0)
1913 cursor
-= direction
;
1914 if (pat
[i
] != *cursor
)
1918 cursor
+= dirlen
- i
- direction
; /* fix cursor */
1919 if (i
+ direction
== 0)
1921 ptrdiff_t position
, start
, end
;
1923 cursor
-= direction
;
1925 position
= pos_byte
+ cursor
- p2
+ ((direction
> 0)
1926 ? 1 - len_byte
: 0);
1927 set_search_regs (position
, len_byte
);
1929 if (NILP (Vinhibit_changing_match_data
))
1931 start
= search_regs
.start
[0];
1932 end
= search_regs
.end
[0];
1935 /* If Vinhibit_changing_match_data is non-nil,
1936 search_regs will not be changed. So let's
1937 compute start and end here. */
1939 start
= BYTE_TO_CHAR (position
);
1940 end
= BYTE_TO_CHAR (position
+ len_byte
);
1943 if ((n
-= direction
) != 0)
1944 cursor
+= dirlen
; /* to resume search */
1946 return direction
> 0 ? end
: start
;
1949 cursor
+= stride_for_teases
; /* <sigh> we lose - */
1951 pos_byte
+= cursor
- p2
;
1954 /* Now we'll pick up a clump that has to be done the hard
1955 way because it covers a discontinuity. */
1957 limit
= ((direction
> 0)
1958 ? BUFFER_CEILING_OF (pos_byte
- dirlen
+ 1)
1959 : BUFFER_FLOOR_OF (pos_byte
- dirlen
- 1));
1960 limit
= ((direction
> 0)
1961 ? min (limit
+ len_byte
, lim_byte
- 1)
1962 : max (limit
- len_byte
, lim_byte
));
1963 /* LIMIT is now the last value POS_BYTE can have
1964 and still be valid for a possible match. */
1967 /* This loop can be coded for space rather than
1968 speed because it will usually run only once.
1969 (the reach is at most len + 21, and typically
1970 does not exceed len). */
1971 while ((limit
- pos_byte
) * direction
>= 0)
1973 int ch
= FETCH_BYTE (pos_byte
);
1974 if (BM_tab
[ch
] == 0)
1976 pos_byte
+= BM_tab
[ch
];
1978 break; /* ran off the end */
1981 /* Found what might be a match. */
1982 i
= dirlen
- direction
;
1983 while ((i
-= direction
) + direction
!= 0)
1987 pos_byte
-= direction
;
1988 ptr
= BYTE_POS_ADDR (pos_byte
);
1989 /* Translate only the last byte of a character. */
1991 || ((ptr
== tail_end_ptr
1992 || CHAR_HEAD_P (ptr
[1]))
1993 && (CHAR_HEAD_P (ptr
[0])
1994 /* Check if this is the last byte of a
1995 translatable character. */
1996 || (translate_prev_byte1
== ptr
[-1]
1997 && (CHAR_HEAD_P (translate_prev_byte1
)
1998 || (translate_prev_byte2
== ptr
[-2]
1999 && (CHAR_HEAD_P (translate_prev_byte2
)
2000 || translate_prev_byte3
== ptr
[-3])))))))
2001 ch
= simple_translate
[*ptr
];
2007 /* Above loop has moved POS_BYTE part or all the way
2008 back to the first pos (last pos if reverse).
2009 Set it once again at the last (first if reverse) char. */
2010 pos_byte
+= dirlen
- i
- direction
;
2011 if (i
+ direction
== 0)
2013 ptrdiff_t position
, start
, end
;
2014 pos_byte
-= direction
;
2016 position
= pos_byte
+ ((direction
> 0) ? 1 - len_byte
: 0);
2017 set_search_regs (position
, len_byte
);
2019 if (NILP (Vinhibit_changing_match_data
))
2021 start
= search_regs
.start
[0];
2022 end
= search_regs
.end
[0];
2025 /* If Vinhibit_changing_match_data is non-nil,
2026 search_regs will not be changed. So let's
2027 compute start and end here. */
2029 start
= BYTE_TO_CHAR (position
);
2030 end
= BYTE_TO_CHAR (position
+ len_byte
);
2033 if ((n
-= direction
) != 0)
2034 pos_byte
+= dirlen
; /* to resume search */
2036 return direction
> 0 ? end
: start
;
2039 pos_byte
+= stride_for_teases
;
2042 /* We have done one clump. Can we continue? */
2043 if ((lim_byte
- pos_byte
) * direction
< 0)
2044 return ((0 - n
) * direction
);
2046 return BYTE_TO_CHAR (pos_byte
);
2049 /* Record beginning BEG_BYTE and end BEG_BYTE + NBYTES
2050 for the overall match just found in the current buffer.
2051 Also clear out the match data for registers 1 and up. */
2054 set_search_regs (ptrdiff_t beg_byte
, ptrdiff_t nbytes
)
2058 if (!NILP (Vinhibit_changing_match_data
))
2061 /* Make sure we have registers in which to store
2062 the match position. */
2063 if (search_regs
.num_regs
== 0)
2065 search_regs
.start
= (regoff_t
*) xmalloc (2 * sizeof (regoff_t
));
2066 search_regs
.end
= (regoff_t
*) xmalloc (2 * sizeof (regoff_t
));
2067 search_regs
.num_regs
= 2;
2070 /* Clear out the other registers. */
2071 for (i
= 1; i
< search_regs
.num_regs
; i
++)
2073 search_regs
.start
[i
] = -1;
2074 search_regs
.end
[i
] = -1;
2077 search_regs
.start
[0] = BYTE_TO_CHAR (beg_byte
);
2078 search_regs
.end
[0] = BYTE_TO_CHAR (beg_byte
+ nbytes
);
2079 XSETBUFFER (last_thing_searched
, current_buffer
);
2082 DEFUN ("search-backward", Fsearch_backward
, Ssearch_backward
, 1, 4,
2083 "MSearch backward: ",
2084 doc
: /* Search backward from point for STRING.
2085 Set point to the beginning of the occurrence found, and return point.
2086 An optional second argument bounds the search; it is a buffer position.
2087 The match found must not extend before that position.
2088 Optional third argument, if t, means if fail just return nil (no error).
2089 If not nil and not t, position at limit of search and return nil.
2090 Optional fourth argument COUNT, if non-nil, means to search for COUNT
2091 successive occurrences. If COUNT is negative, search forward,
2092 instead of backward, for -COUNT 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 (Lisp_Object string
, Lisp_Object bound
, Lisp_Object noerror
, Lisp_Object count
)
2100 return search_command (string
, bound
, noerror
, count
, -1, 0, 0);
2103 DEFUN ("search-forward", Fsearch_forward
, Ssearch_forward
, 1, 4, "MSearch: ",
2104 doc
: /* Search forward from point for STRING.
2105 Set point to the end of the occurrence found, and return point.
2106 An optional second argument bounds the search; it is a buffer position.
2107 The match found must not extend after that position. A value of nil is
2108 equivalent to (point-max).
2109 Optional third argument, if t, means if fail just return nil (no error).
2110 If not nil and not t, move to limit of search and return nil.
2111 Optional fourth argument COUNT, if non-nil, means to search for COUNT
2112 successive occurrences. If COUNT is negative, search backward,
2113 instead of forward, for -COUNT occurrences.
2115 Search case-sensitivity is determined by the value of the variable
2116 `case-fold-search', which see.
2118 See also the functions `match-beginning', `match-end' and `replace-match'. */)
2119 (Lisp_Object string
, Lisp_Object bound
, Lisp_Object noerror
, Lisp_Object count
)
2121 return search_command (string
, bound
, noerror
, count
, 1, 0, 0);
2124 DEFUN ("re-search-backward", Fre_search_backward
, Sre_search_backward
, 1, 4,
2125 "sRE search backward: ",
2126 doc
: /* Search backward from point for match for regular expression REGEXP.
2127 Set point to the beginning of the match, and return point.
2128 The match found is the one starting last in the buffer
2129 and yet ending before the origin of the search.
2130 An optional second argument bounds the search; it is a buffer position.
2131 The match found must start at or after that position.
2132 Optional third argument, if t, means if fail just return nil (no error).
2133 If not nil and not t, move to limit of search and return nil.
2134 Optional fourth argument is repeat count--search for successive occurrences.
2136 Search case-sensitivity is determined by the value of the variable
2137 `case-fold-search', which see.
2139 See also the functions `match-beginning', `match-end', `match-string',
2140 and `replace-match'. */)
2141 (Lisp_Object regexp
, Lisp_Object bound
, Lisp_Object noerror
, Lisp_Object count
)
2143 return search_command (regexp
, bound
, noerror
, count
, -1, 1, 0);
2146 DEFUN ("re-search-forward", Fre_search_forward
, Sre_search_forward
, 1, 4,
2148 doc
: /* Search forward from point for regular expression REGEXP.
2149 Set point to the end of the occurrence found, and return point.
2150 An optional second argument bounds the search; it is a buffer position.
2151 The match found must not extend after that position.
2152 Optional third argument, if t, means if fail just return nil (no error).
2153 If not nil and not t, move to limit of search and return nil.
2154 Optional fourth argument is repeat count--search for successive occurrences.
2156 Search case-sensitivity is determined by the value of the variable
2157 `case-fold-search', which see.
2159 See also the functions `match-beginning', `match-end', `match-string',
2160 and `replace-match'. */)
2161 (Lisp_Object regexp
, Lisp_Object bound
, Lisp_Object noerror
, Lisp_Object count
)
2163 return search_command (regexp
, bound
, noerror
, count
, 1, 1, 0);
2166 DEFUN ("posix-search-backward", Fposix_search_backward
, Sposix_search_backward
, 1, 4,
2167 "sPosix search backward: ",
2168 doc
: /* Search backward from point for match for regular expression REGEXP.
2169 Find the longest match in accord with Posix regular expression rules.
2170 Set point to the beginning of the match, and return point.
2171 The match found is the one starting last in the buffer
2172 and yet ending before the origin of the search.
2173 An optional second argument bounds the search; it is a buffer position.
2174 The match found must start at or after that position.
2175 Optional third argument, if t, means if fail just return nil (no error).
2176 If not nil and not t, move to limit of search and return nil.
2177 Optional fourth argument is repeat count--search for successive occurrences.
2179 Search case-sensitivity is determined by the value of the variable
2180 `case-fold-search', which see.
2182 See also the functions `match-beginning', `match-end', `match-string',
2183 and `replace-match'. */)
2184 (Lisp_Object regexp
, Lisp_Object bound
, Lisp_Object noerror
, Lisp_Object count
)
2186 return search_command (regexp
, bound
, noerror
, count
, -1, 1, 1);
2189 DEFUN ("posix-search-forward", Fposix_search_forward
, Sposix_search_forward
, 1, 4,
2191 doc
: /* Search forward from point for regular expression REGEXP.
2192 Find the longest match in accord with Posix regular expression rules.
2193 Set point to the end of the occurrence found, and return point.
2194 An optional second argument bounds the search; it is a buffer position.
2195 The match found must not extend after that position.
2196 Optional third argument, if t, means if fail just return nil (no error).
2197 If not nil and not t, move to limit of search and return nil.
2198 Optional fourth argument is repeat count--search for successive occurrences.
2200 Search case-sensitivity is determined by the value of the variable
2201 `case-fold-search', which see.
2203 See also the functions `match-beginning', `match-end', `match-string',
2204 and `replace-match'. */)
2205 (Lisp_Object regexp
, Lisp_Object bound
, Lisp_Object noerror
, Lisp_Object count
)
2207 return search_command (regexp
, bound
, noerror
, count
, 1, 1, 1);
2210 DEFUN ("replace-match", Freplace_match
, Sreplace_match
, 1, 5, 0,
2211 doc
: /* Replace text matched by last search with NEWTEXT.
2212 Leave point at the end of the replacement text.
2214 If second arg FIXEDCASE is non-nil, do not alter case of replacement text.
2215 Otherwise maybe capitalize the whole text, or maybe just word initials,
2216 based on the replaced text.
2217 If the replaced text has only capital letters
2218 and has at least one multiletter word, convert NEWTEXT to all caps.
2219 Otherwise if all words are capitalized in the replaced text,
2220 capitalize each word in NEWTEXT.
2222 If third arg LITERAL is non-nil, insert NEWTEXT literally.
2223 Otherwise treat `\\' as special:
2224 `\\&' in NEWTEXT means substitute original matched text.
2225 `\\N' means substitute what matched the Nth `\\(...\\)'.
2226 If Nth parens didn't match, substitute nothing.
2227 `\\\\' means insert one `\\'.
2228 Case conversion does not apply to these substitutions.
2230 FIXEDCASE and LITERAL are optional arguments.
2232 The optional fourth argument STRING can be a string to modify.
2233 This is meaningful when the previous match was done against STRING,
2234 using `string-match'. When used this way, `replace-match'
2235 creates and returns a new string made by copying STRING and replacing
2236 the part of STRING that was matched.
2238 The optional fifth argument SUBEXP specifies a subexpression;
2239 it says to replace just that subexpression with NEWTEXT,
2240 rather than replacing the entire matched text.
2241 This is, in a vague sense, the inverse of using `\\N' in NEWTEXT;
2242 `\\N' copies subexp N into NEWTEXT, but using N as SUBEXP puts
2243 NEWTEXT in place of subexp N.
2244 This is useful only after a regular expression search or match,
2245 since only regular expressions have distinguished subexpressions. */)
2246 (Lisp_Object newtext
, Lisp_Object fixedcase
, Lisp_Object literal
, Lisp_Object string
, Lisp_Object subexp
)
2248 enum { nochange
, all_caps
, cap_initial
} case_action
;
2249 register ptrdiff_t pos
, pos_byte
;
2250 int some_multiletter_word
;
2253 int some_nonuppercase_initial
;
2254 register int c
, prevc
;
2256 ptrdiff_t opoint
, newpoint
;
2258 CHECK_STRING (newtext
);
2260 if (! NILP (string
))
2261 CHECK_STRING (string
);
2263 case_action
= nochange
; /* We tried an initialization */
2264 /* but some C compilers blew it */
2266 if (search_regs
.num_regs
<= 0)
2267 error ("`replace-match' called before any match found");
2273 CHECK_NUMBER (subexp
);
2274 if (! (0 <= XINT (subexp
) && XINT (subexp
) < search_regs
.num_regs
))
2275 args_out_of_range (subexp
, make_number (search_regs
.num_regs
));
2276 sub
= XINT (subexp
);
2281 if (search_regs
.start
[sub
] < BEGV
2282 || search_regs
.start
[sub
] > search_regs
.end
[sub
]
2283 || search_regs
.end
[sub
] > ZV
)
2284 args_out_of_range (make_number (search_regs
.start
[sub
]),
2285 make_number (search_regs
.end
[sub
]));
2289 if (search_regs
.start
[sub
] < 0
2290 || search_regs
.start
[sub
] > search_regs
.end
[sub
]
2291 || search_regs
.end
[sub
] > SCHARS (string
))
2292 args_out_of_range (make_number (search_regs
.start
[sub
]),
2293 make_number (search_regs
.end
[sub
]));
2296 if (NILP (fixedcase
))
2298 /* Decide how to casify by examining the matched text. */
2301 pos
= search_regs
.start
[sub
];
2302 last
= search_regs
.end
[sub
];
2305 pos_byte
= CHAR_TO_BYTE (pos
);
2307 pos_byte
= string_char_to_byte (string
, pos
);
2310 case_action
= all_caps
;
2312 /* some_multiletter_word is set nonzero if any original word
2313 is more than one letter long. */
2314 some_multiletter_word
= 0;
2316 some_nonuppercase_initial
= 0;
2323 c
= FETCH_CHAR_AS_MULTIBYTE (pos_byte
);
2324 INC_BOTH (pos
, pos_byte
);
2327 FETCH_STRING_CHAR_AS_MULTIBYTE_ADVANCE (c
, string
, pos
, pos_byte
);
2331 /* Cannot be all caps if any original char is lower case */
2334 if (SYNTAX (prevc
) != Sword
)
2335 some_nonuppercase_initial
= 1;
2337 some_multiletter_word
= 1;
2339 else if (uppercasep (c
))
2342 if (SYNTAX (prevc
) != Sword
)
2345 some_multiletter_word
= 1;
2349 /* If the initial is a caseless word constituent,
2350 treat that like a lowercase initial. */
2351 if (SYNTAX (prevc
) != Sword
)
2352 some_nonuppercase_initial
= 1;
2358 /* Convert to all caps if the old text is all caps
2359 and has at least one multiletter word. */
2360 if (! some_lowercase
&& some_multiletter_word
)
2361 case_action
= all_caps
;
2362 /* Capitalize each word, if the old text has all capitalized words. */
2363 else if (!some_nonuppercase_initial
&& some_multiletter_word
)
2364 case_action
= cap_initial
;
2365 else if (!some_nonuppercase_initial
&& some_uppercase
)
2366 /* Should x -> yz, operating on X, give Yz or YZ?
2367 We'll assume the latter. */
2368 case_action
= all_caps
;
2370 case_action
= nochange
;
2373 /* Do replacement in a string. */
2376 Lisp_Object before
, after
;
2378 before
= Fsubstring (string
, make_number (0),
2379 make_number (search_regs
.start
[sub
]));
2380 after
= Fsubstring (string
, make_number (search_regs
.end
[sub
]), Qnil
);
2382 /* Substitute parts of the match into NEWTEXT
2386 ptrdiff_t lastpos
= 0;
2387 ptrdiff_t lastpos_byte
= 0;
2388 /* We build up the substituted string in ACCUM. */
2391 ptrdiff_t length
= SBYTES (newtext
);
2395 for (pos_byte
= 0, pos
= 0; pos_byte
< length
;)
2397 ptrdiff_t substart
= -1;
2398 ptrdiff_t subend
= 0;
2399 int delbackslash
= 0;
2401 FETCH_STRING_CHAR_ADVANCE (c
, newtext
, pos
, pos_byte
);
2405 FETCH_STRING_CHAR_ADVANCE (c
, newtext
, pos
, pos_byte
);
2409 substart
= search_regs
.start
[sub
];
2410 subend
= search_regs
.end
[sub
];
2412 else if (c
>= '1' && c
<= '9')
2414 if (c
- '0' < search_regs
.num_regs
2415 && 0 <= search_regs
.start
[c
- '0'])
2417 substart
= search_regs
.start
[c
- '0'];
2418 subend
= search_regs
.end
[c
- '0'];
2422 /* If that subexp did not match,
2423 replace \\N with nothing. */
2431 error ("Invalid use of `\\' in replacement text");
2435 if (pos
- 2 != lastpos
)
2436 middle
= substring_both (newtext
, lastpos
,
2438 pos
- 2, pos_byte
- 2);
2441 accum
= concat3 (accum
, middle
,
2443 make_number (substart
),
2444 make_number (subend
)));
2446 lastpos_byte
= pos_byte
;
2448 else if (delbackslash
)
2450 middle
= substring_both (newtext
, lastpos
,
2452 pos
- 1, pos_byte
- 1);
2454 accum
= concat2 (accum
, middle
);
2456 lastpos_byte
= pos_byte
;
2461 middle
= substring_both (newtext
, lastpos
,
2467 newtext
= concat2 (accum
, middle
);
2470 /* Do case substitution in NEWTEXT if desired. */
2471 if (case_action
== all_caps
)
2472 newtext
= Fupcase (newtext
);
2473 else if (case_action
== cap_initial
)
2474 newtext
= Fupcase_initials (newtext
);
2476 return concat3 (before
, newtext
, after
);
2479 /* Record point, then move (quietly) to the start of the match. */
2480 if (PT
>= search_regs
.end
[sub
])
2482 else if (PT
> search_regs
.start
[sub
])
2483 opoint
= search_regs
.end
[sub
] - ZV
;
2487 /* If we want non-literal replacement,
2488 perform substitution on the replacement string. */
2491 ptrdiff_t length
= SBYTES (newtext
);
2492 unsigned char *substed
;
2493 ptrdiff_t substed_alloc_size
, substed_len
;
2494 int buf_multibyte
= !NILP (BVAR (current_buffer
, enable_multibyte_characters
));
2495 int str_multibyte
= STRING_MULTIBYTE (newtext
);
2496 int really_changed
= 0;
2498 substed_alloc_size
= ((STRING_BYTES_BOUND
- 100) / 2 < length
2499 ? STRING_BYTES_BOUND
2500 : length
* 2 + 100);
2501 substed
= (unsigned char *) xmalloc (substed_alloc_size
);
2504 /* Go thru NEWTEXT, producing the actual text to insert in
2505 SUBSTED while adjusting multibyteness to that of the current
2508 for (pos_byte
= 0, pos
= 0; pos_byte
< length
;)
2510 unsigned char str
[MAX_MULTIBYTE_LENGTH
];
2511 const unsigned char *add_stuff
= NULL
;
2512 ptrdiff_t add_len
= 0;
2517 FETCH_STRING_CHAR_ADVANCE_NO_CHECK (c
, newtext
, pos
, pos_byte
);
2519 c
= multibyte_char_to_unibyte (c
);
2523 /* Note that we don't have to increment POS. */
2524 c
= SREF (newtext
, pos_byte
++);
2526 MAKE_CHAR_MULTIBYTE (c
);
2529 /* Either set ADD_STUFF and ADD_LEN to the text to put in SUBSTED,
2530 or set IDX to a match index, which means put that part
2531 of the buffer text into SUBSTED. */
2539 FETCH_STRING_CHAR_ADVANCE_NO_CHECK (c
, newtext
,
2541 if (!buf_multibyte
&& !ASCII_CHAR_P (c
))
2542 c
= multibyte_char_to_unibyte (c
);
2546 c
= SREF (newtext
, pos_byte
++);
2548 MAKE_CHAR_MULTIBYTE (c
);
2553 else if (c
>= '1' && c
<= '9' && c
- '0' < search_regs
.num_regs
)
2555 if (search_regs
.start
[c
- '0'] >= 1)
2559 add_len
= 1, add_stuff
= (unsigned char *) "\\";
2563 error ("Invalid use of `\\' in replacement text");
2568 add_len
= CHAR_STRING (c
, str
);
2572 /* If we want to copy part of a previous match,
2573 set up ADD_STUFF and ADD_LEN to point to it. */
2576 ptrdiff_t begbyte
= CHAR_TO_BYTE (search_regs
.start
[idx
]);
2577 add_len
= CHAR_TO_BYTE (search_regs
.end
[idx
]) - begbyte
;
2578 if (search_regs
.start
[idx
] < GPT
&& GPT
< search_regs
.end
[idx
])
2579 move_gap (search_regs
.start
[idx
]);
2580 add_stuff
= BYTE_POS_ADDR (begbyte
);
2583 /* Now the stuff we want to add to SUBSTED
2584 is invariably ADD_LEN bytes starting at ADD_STUFF. */
2586 /* Make sure SUBSTED is big enough. */
2587 if (substed_alloc_size
- substed_len
< add_len
)
2589 xpalloc (substed
, &substed_alloc_size
,
2590 add_len
- (substed_alloc_size
- substed_len
),
2591 STRING_BYTES_BOUND
, 1);
2593 /* Now add to the end of SUBSTED. */
2596 memcpy (substed
+ substed_len
, add_stuff
, add_len
);
2597 substed_len
+= add_len
;
2606 multibyte_chars_in_text (substed
, substed_len
);
2608 newtext
= make_multibyte_string ((char *) substed
, nchars
,
2612 newtext
= make_unibyte_string ((char *) substed
, substed_len
);
2617 /* Replace the old text with the new in the cleanest possible way. */
2618 replace_range (search_regs
.start
[sub
], search_regs
.end
[sub
],
2620 newpoint
= search_regs
.start
[sub
] + SCHARS (newtext
);
2622 if (case_action
== all_caps
)
2623 Fupcase_region (make_number (search_regs
.start
[sub
]),
2624 make_number (newpoint
));
2625 else if (case_action
== cap_initial
)
2626 Fupcase_initials_region (make_number (search_regs
.start
[sub
]),
2627 make_number (newpoint
));
2629 /* Adjust search data for this change. */
2631 ptrdiff_t oldend
= search_regs
.end
[sub
];
2632 ptrdiff_t oldstart
= search_regs
.start
[sub
];
2633 ptrdiff_t change
= newpoint
- search_regs
.end
[sub
];
2636 for (i
= 0; i
< search_regs
.num_regs
; i
++)
2638 if (search_regs
.start
[i
] >= oldend
)
2639 search_regs
.start
[i
] += change
;
2640 else if (search_regs
.start
[i
] > oldstart
)
2641 search_regs
.start
[i
] = oldstart
;
2642 if (search_regs
.end
[i
] >= oldend
)
2643 search_regs
.end
[i
] += change
;
2644 else if (search_regs
.end
[i
] > oldstart
)
2645 search_regs
.end
[i
] = oldstart
;
2649 /* Put point back where it was in the text. */
2651 TEMP_SET_PT (opoint
+ ZV
);
2653 TEMP_SET_PT (opoint
);
2655 /* Now move point "officially" to the start of the inserted replacement. */
2656 move_if_not_intangible (newpoint
);
2662 match_limit (Lisp_Object num
, int beginningp
)
2669 args_out_of_range (num
, make_number (0));
2670 if (search_regs
.num_regs
<= 0)
2671 error ("No match data, because no search succeeded");
2672 if (n
>= search_regs
.num_regs
2673 || search_regs
.start
[n
] < 0)
2675 return (make_number ((beginningp
) ? search_regs
.start
[n
]
2676 : search_regs
.end
[n
]));
2679 DEFUN ("match-beginning", Fmatch_beginning
, Smatch_beginning
, 1, 1, 0,
2680 doc
: /* Return position of start of text matched by last search.
2681 SUBEXP, a number, specifies which parenthesized expression in the last
2683 Value is nil if SUBEXPth pair didn't match, or there were less than
2685 Zero means the entire text matched by the whole regexp or whole string. */)
2686 (Lisp_Object subexp
)
2688 return match_limit (subexp
, 1);
2691 DEFUN ("match-end", Fmatch_end
, Smatch_end
, 1, 1, 0,
2692 doc
: /* Return position of end of text matched by last search.
2693 SUBEXP, a number, specifies which parenthesized expression in the last
2695 Value is nil if SUBEXPth pair didn't match, or there were less than
2697 Zero means the entire text matched by the whole regexp or whole string. */)
2698 (Lisp_Object subexp
)
2700 return match_limit (subexp
, 0);
2703 DEFUN ("match-data", Fmatch_data
, Smatch_data
, 0, 3, 0,
2704 doc
: /* Return a list containing all info on what the last search matched.
2705 Element 2N is `(match-beginning N)'; element 2N + 1 is `(match-end N)'.
2706 All the elements are markers or nil (nil if the Nth pair didn't match)
2707 if the last match was on a buffer; integers or nil if a string was matched.
2708 Use `set-match-data' to reinstate the data in this list.
2710 If INTEGERS (the optional first argument) is non-nil, always use
2711 integers \(rather than markers) to represent buffer positions. In
2712 this case, and if the last match was in a buffer, the buffer will get
2713 stored as one additional element at the end of the list.
2715 If REUSE is a list, reuse it as part of the value. If REUSE is long
2716 enough to hold all the values, and if INTEGERS is non-nil, no consing
2719 If optional third arg RESEAT is non-nil, any previous markers on the
2720 REUSE list will be modified to point to nowhere.
2722 Return value is undefined if the last search failed. */)
2723 (Lisp_Object integers
, Lisp_Object reuse
, Lisp_Object reseat
)
2725 Lisp_Object tail
, prev
;
2730 for (tail
= reuse
; CONSP (tail
); tail
= XCDR (tail
))
2731 if (MARKERP (XCAR (tail
)))
2733 unchain_marker (XMARKER (XCAR (tail
)));
2734 XSETCAR (tail
, Qnil
);
2737 if (NILP (last_thing_searched
))
2742 data
= (Lisp_Object
*) alloca ((2 * search_regs
.num_regs
+ 1)
2743 * sizeof (Lisp_Object
));
2746 for (i
= 0; i
< search_regs
.num_regs
; i
++)
2748 ptrdiff_t start
= search_regs
.start
[i
];
2751 if (EQ (last_thing_searched
, Qt
)
2752 || ! NILP (integers
))
2754 XSETFASTINT (data
[2 * i
], start
);
2755 XSETFASTINT (data
[2 * i
+ 1], search_regs
.end
[i
]);
2757 else if (BUFFERP (last_thing_searched
))
2759 data
[2 * i
] = Fmake_marker ();
2760 Fset_marker (data
[2 * i
],
2761 make_number (start
),
2762 last_thing_searched
);
2763 data
[2 * i
+ 1] = Fmake_marker ();
2764 Fset_marker (data
[2 * i
+ 1],
2765 make_number (search_regs
.end
[i
]),
2766 last_thing_searched
);
2769 /* last_thing_searched must always be Qt, a buffer, or Qnil. */
2775 data
[2 * i
] = data
[2 * i
+ 1] = Qnil
;
2778 if (BUFFERP (last_thing_searched
) && !NILP (integers
))
2780 data
[len
] = last_thing_searched
;
2784 /* If REUSE is not usable, cons up the values and return them. */
2785 if (! CONSP (reuse
))
2786 return Flist (len
, data
);
2788 /* If REUSE is a list, store as many value elements as will fit
2789 into the elements of REUSE. */
2790 for (i
= 0, tail
= reuse
; CONSP (tail
);
2791 i
++, tail
= XCDR (tail
))
2794 XSETCAR (tail
, data
[i
]);
2796 XSETCAR (tail
, Qnil
);
2800 /* If we couldn't fit all value elements into REUSE,
2801 cons up the rest of them and add them to the end of REUSE. */
2803 XSETCDR (prev
, Flist (len
- i
, data
+ i
));
2808 /* We used to have an internal use variant of `reseat' described as:
2810 If RESEAT is `evaporate', put the markers back on the free list
2811 immediately. No other references to the markers must exist in this
2812 case, so it is used only internally on the unwind stack and
2813 save-match-data from Lisp.
2815 But it was ill-conceived: those supposedly-internal markers get exposed via
2816 the undo-list, so freeing them here is unsafe. */
2818 DEFUN ("set-match-data", Fset_match_data
, Sset_match_data
, 1, 2, 0,
2819 doc
: /* Set internal data on last search match from elements of LIST.
2820 LIST should have been created by calling `match-data' previously.
2822 If optional arg RESEAT is non-nil, make markers on LIST point nowhere. */)
2823 (register Lisp_Object list
, Lisp_Object reseat
)
2826 register Lisp_Object marker
;
2828 if (running_asynch_code
)
2829 save_search_regs ();
2833 /* Unless we find a marker with a buffer or an explicit buffer
2834 in LIST, assume that this match data came from a string. */
2835 last_thing_searched
= Qt
;
2837 /* Allocate registers if they don't already exist. */
2839 EMACS_INT length
= XFASTINT (Flength (list
)) / 2;
2841 if (length
> search_regs
.num_regs
)
2843 ptrdiff_t num_regs
= search_regs
.num_regs
;
2844 if (PTRDIFF_MAX
< length
)
2845 memory_full (SIZE_MAX
);
2847 xpalloc (search_regs
.start
, &num_regs
, length
- num_regs
,
2848 min (PTRDIFF_MAX
, UINT_MAX
), sizeof (regoff_t
));
2850 xrealloc (search_regs
.end
, num_regs
* sizeof (regoff_t
));
2852 for (i
= search_regs
.num_regs
; i
< num_regs
; i
++)
2853 search_regs
.start
[i
] = -1;
2855 search_regs
.num_regs
= num_regs
;
2858 for (i
= 0; CONSP (list
); i
++)
2860 marker
= XCAR (list
);
2861 if (BUFFERP (marker
))
2863 last_thing_searched
= marker
;
2870 search_regs
.start
[i
] = -1;
2879 if (MARKERP (marker
))
2881 if (XMARKER (marker
)->buffer
== 0)
2882 XSETFASTINT (marker
, 0);
2884 XSETBUFFER (last_thing_searched
, XMARKER (marker
)->buffer
);
2887 CHECK_NUMBER_COERCE_MARKER (marker
);
2890 if (!NILP (reseat
) && MARKERP (m
))
2892 unchain_marker (XMARKER (m
));
2893 XSETCAR (list
, Qnil
);
2896 if ((list
= XCDR (list
), !CONSP (list
)))
2899 m
= marker
= XCAR (list
);
2901 if (MARKERP (marker
) && XMARKER (marker
)->buffer
== 0)
2902 XSETFASTINT (marker
, 0);
2904 CHECK_NUMBER_COERCE_MARKER (marker
);
2905 if ((XINT (from
) < 0
2906 ? TYPE_MINIMUM (regoff_t
) <= XINT (from
)
2907 : XINT (from
) <= TYPE_MAXIMUM (regoff_t
))
2908 && (XINT (marker
) < 0
2909 ? TYPE_MINIMUM (regoff_t
) <= XINT (marker
)
2910 : XINT (marker
) <= TYPE_MAXIMUM (regoff_t
)))
2912 search_regs
.start
[i
] = XINT (from
);
2913 search_regs
.end
[i
] = XINT (marker
);
2917 search_regs
.start
[i
] = -1;
2920 if (!NILP (reseat
) && MARKERP (m
))
2922 unchain_marker (XMARKER (m
));
2923 XSETCAR (list
, Qnil
);
2929 for (; i
< search_regs
.num_regs
; i
++)
2930 search_regs
.start
[i
] = -1;
2936 /* If non-zero the match data have been saved in saved_search_regs
2937 during the execution of a sentinel or filter. */
2938 static int search_regs_saved
;
2939 static struct re_registers saved_search_regs
;
2940 static Lisp_Object saved_last_thing_searched
;
2942 /* Called from Flooking_at, Fstring_match, search_buffer, Fstore_match_data
2943 if asynchronous code (filter or sentinel) is running. */
2945 save_search_regs (void)
2947 if (!search_regs_saved
)
2949 saved_search_regs
.num_regs
= search_regs
.num_regs
;
2950 saved_search_regs
.start
= search_regs
.start
;
2951 saved_search_regs
.end
= search_regs
.end
;
2952 saved_last_thing_searched
= last_thing_searched
;
2953 last_thing_searched
= Qnil
;
2954 search_regs
.num_regs
= 0;
2955 search_regs
.start
= 0;
2956 search_regs
.end
= 0;
2958 search_regs_saved
= 1;
2962 /* Called upon exit from filters and sentinels. */
2964 restore_search_regs (void)
2966 if (search_regs_saved
)
2968 if (search_regs
.num_regs
> 0)
2970 xfree (search_regs
.start
);
2971 xfree (search_regs
.end
);
2973 search_regs
.num_regs
= saved_search_regs
.num_regs
;
2974 search_regs
.start
= saved_search_regs
.start
;
2975 search_regs
.end
= saved_search_regs
.end
;
2976 last_thing_searched
= saved_last_thing_searched
;
2977 saved_last_thing_searched
= Qnil
;
2978 search_regs_saved
= 0;
2983 unwind_set_match_data (Lisp_Object list
)
2985 /* It is NOT ALWAYS safe to free (evaporate) the markers immediately. */
2986 return Fset_match_data (list
, Qt
);
2989 /* Called to unwind protect the match data. */
2991 record_unwind_save_match_data (void)
2993 record_unwind_protect (unwind_set_match_data
,
2994 Fmatch_data (Qnil
, Qnil
, Qnil
));
2997 /* Quote a string to deactivate reg-expr chars */
2999 DEFUN ("regexp-quote", Fregexp_quote
, Sregexp_quote
, 1, 1, 0,
3000 doc
: /* Return a regexp string which matches exactly STRING and nothing else. */)
3001 (Lisp_Object string
)
3003 register char *in
, *out
, *end
;
3004 register char *temp
;
3005 int backslashes_added
= 0;
3007 CHECK_STRING (string
);
3009 temp
= (char *) alloca (SBYTES (string
) * 2);
3011 /* Now copy the data into the new string, inserting escapes. */
3013 in
= SSDATA (string
);
3014 end
= in
+ SBYTES (string
);
3017 for (; in
!= end
; in
++)
3020 || *in
== '*' || *in
== '.' || *in
== '\\'
3021 || *in
== '?' || *in
== '+'
3022 || *in
== '^' || *in
== '$')
3023 *out
++ = '\\', backslashes_added
++;
3027 return make_specified_string (temp
,
3028 SCHARS (string
) + backslashes_added
,
3030 STRING_MULTIBYTE (string
));
3034 syms_of_search (void)
3038 for (i
= 0; i
< REGEXP_CACHE_SIZE
; ++i
)
3040 searchbufs
[i
].buf
.allocated
= 100;
3041 searchbufs
[i
].buf
.buffer
= (unsigned char *) xmalloc (100);
3042 searchbufs
[i
].buf
.fastmap
= searchbufs
[i
].fastmap
;
3043 searchbufs
[i
].regexp
= Qnil
;
3044 searchbufs
[i
].whitespace_regexp
= Qnil
;
3045 searchbufs
[i
].syntax_table
= Qnil
;
3046 staticpro (&searchbufs
[i
].regexp
);
3047 staticpro (&searchbufs
[i
].whitespace_regexp
);
3048 staticpro (&searchbufs
[i
].syntax_table
);
3049 searchbufs
[i
].next
= (i
== REGEXP_CACHE_SIZE
-1 ? 0 : &searchbufs
[i
+1]);
3051 searchbuf_head
= &searchbufs
[0];
3053 DEFSYM (Qsearch_failed
, "search-failed");
3054 DEFSYM (Qinvalid_regexp
, "invalid-regexp");
3056 Fput (Qsearch_failed
, Qerror_conditions
,
3057 pure_cons (Qsearch_failed
, pure_cons (Qerror
, Qnil
)));
3058 Fput (Qsearch_failed
, Qerror_message
,
3059 make_pure_c_string ("Search failed"));
3061 Fput (Qinvalid_regexp
, Qerror_conditions
,
3062 pure_cons (Qinvalid_regexp
, pure_cons (Qerror
, Qnil
)));
3063 Fput (Qinvalid_regexp
, Qerror_message
,
3064 make_pure_c_string ("Invalid regexp"));
3066 last_thing_searched
= Qnil
;
3067 staticpro (&last_thing_searched
);
3069 saved_last_thing_searched
= Qnil
;
3070 staticpro (&saved_last_thing_searched
);
3072 DEFVAR_LISP ("search-spaces-regexp", Vsearch_spaces_regexp
,
3073 doc
: /* Regexp to substitute for bunches of spaces in regexp search.
3074 Some commands use this for user-specified regexps.
3075 Spaces that occur inside character classes or repetition operators
3076 or other such regexp constructs are not replaced with this.
3077 A value of nil (which is the normal value) means treat spaces literally. */);
3078 Vsearch_spaces_regexp
= Qnil
;
3080 DEFVAR_LISP ("inhibit-changing-match-data", Vinhibit_changing_match_data
,
3081 doc
: /* Internal use only.
3082 If non-nil, the primitive searching and matching functions
3083 such as `looking-at', `string-match', `re-search-forward', etc.,
3084 do not set the match data. The proper way to use this variable
3085 is to bind it with `let' around a small expression. */);
3086 Vinhibit_changing_match_data
= Qnil
;
3088 defsubr (&Slooking_at
);
3089 defsubr (&Sposix_looking_at
);
3090 defsubr (&Sstring_match
);
3091 defsubr (&Sposix_string_match
);
3092 defsubr (&Ssearch_forward
);
3093 defsubr (&Ssearch_backward
);
3094 defsubr (&Sre_search_forward
);
3095 defsubr (&Sre_search_backward
);
3096 defsubr (&Sposix_search_forward
);
3097 defsubr (&Sposix_search_backward
);
3098 defsubr (&Sreplace_match
);
3099 defsubr (&Smatch_beginning
);
3100 defsubr (&Smatch_end
);
3101 defsubr (&Smatch_data
);
3102 defsubr (&Sset_match_data
);
3103 defsubr (&Sregexp_quote
);