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/>. */
27 #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);
105 static _Noreturn
void
106 matcher_overflow (void)
108 error ("Stack overflow in regexp matcher");
111 /* Compile a regexp and signal a Lisp error if anything goes wrong.
112 PATTERN is the pattern to compile.
113 CP is the place to put the result.
114 TRANSLATE is a translation table for ignoring case, or nil for none.
115 POSIX is nonzero if we want full backtracking (POSIX style)
116 for this pattern. 0 means backtrack only enough to get a valid match.
118 The behavior also depends on Vsearch_spaces_regexp. */
121 compile_pattern_1 (struct regexp_cache
*cp
, Lisp_Object pattern
, Lisp_Object translate
, int posix
)
127 cp
->buf
.translate
= (! NILP (translate
) ? translate
: make_number (0));
129 cp
->buf
.multibyte
= STRING_MULTIBYTE (pattern
);
130 cp
->buf
.charset_unibyte
= charset_unibyte
;
131 if (STRINGP (Vsearch_spaces_regexp
))
132 cp
->whitespace_regexp
= Vsearch_spaces_regexp
;
134 cp
->whitespace_regexp
= Qnil
;
136 /* rms: I think BLOCK_INPUT is not needed here any more,
137 because regex.c defines malloc to call xmalloc.
138 Using BLOCK_INPUT here means the debugger won't run if an error occurs.
139 So let's turn it off. */
141 old
= re_set_syntax (RE_SYNTAX_EMACS
142 | (posix
? 0 : RE_NO_POSIX_BACKTRACKING
));
144 if (STRINGP (Vsearch_spaces_regexp
))
145 re_set_whitespace_regexp (SSDATA (Vsearch_spaces_regexp
));
147 re_set_whitespace_regexp (NULL
);
149 val
= (char *) re_compile_pattern (SSDATA (pattern
),
150 SBYTES (pattern
), &cp
->buf
);
152 /* If the compiled pattern hard codes some of the contents of the
153 syntax-table, it can only be reused with *this* syntax table. */
154 cp
->syntax_table
= cp
->buf
.used_syntax
? BVAR (current_buffer
, syntax_table
) : Qt
;
156 re_set_whitespace_regexp (NULL
);
159 /* unblock_input (); */
161 xsignal1 (Qinvalid_regexp
, build_string (val
));
163 cp
->regexp
= Fcopy_sequence (pattern
);
166 /* Shrink each compiled regexp buffer in the cache
167 to the size actually used right now.
168 This is called from garbage collection. */
171 shrink_regexp_cache (void)
173 struct regexp_cache
*cp
;
175 for (cp
= searchbuf_head
; cp
!= 0; cp
= cp
->next
)
177 cp
->buf
.allocated
= cp
->buf
.used
;
178 cp
->buf
.buffer
= xrealloc (cp
->buf
.buffer
, cp
->buf
.used
);
182 /* Clear the regexp cache w.r.t. a particular syntax table,
183 because it was changed.
184 There is no danger of memory leak here because re_compile_pattern
185 automagically manages the memory in each re_pattern_buffer struct,
186 based on its `allocated' and `buffer' values. */
188 clear_regexp_cache (void)
192 for (i
= 0; i
< REGEXP_CACHE_SIZE
; ++i
)
193 /* It's tempting to compare with the syntax-table we've actually changed,
194 but it's not sufficient because char-table inheritance means that
195 modifying one syntax-table can change others at the same time. */
196 if (!EQ (searchbufs
[i
].syntax_table
, Qt
))
197 searchbufs
[i
].regexp
= Qnil
;
200 /* Compile a regexp if necessary, but first check to see if there's one in
202 PATTERN is the pattern to compile.
203 TRANSLATE is a translation table for ignoring case, or nil for none.
204 REGP is the structure that says where to store the "register"
205 values that will result from matching this pattern.
206 If it is 0, we should compile the pattern not to record any
207 subexpression bounds.
208 POSIX is nonzero if we want full backtracking (POSIX style)
209 for this pattern. 0 means backtrack only enough to get a valid match. */
211 struct re_pattern_buffer
*
212 compile_pattern (Lisp_Object pattern
, struct re_registers
*regp
, Lisp_Object translate
, int posix
, int multibyte
)
214 struct regexp_cache
*cp
, **cpp
;
216 for (cpp
= &searchbuf_head
; ; cpp
= &cp
->next
)
219 /* Entries are initialized to nil, and may be set to nil by
220 compile_pattern_1 if the pattern isn't valid. Don't apply
221 string accessors in those cases. However, compile_pattern_1
222 is only applied to the cache entry we pick here to reuse. So
223 nil should never appear before a non-nil entry. */
224 if (NILP (cp
->regexp
))
226 if (SCHARS (cp
->regexp
) == SCHARS (pattern
)
227 && STRING_MULTIBYTE (cp
->regexp
) == STRING_MULTIBYTE (pattern
)
228 && !NILP (Fstring_equal (cp
->regexp
, pattern
))
229 && EQ (cp
->buf
.translate
, (! NILP (translate
) ? translate
: make_number (0)))
230 && cp
->posix
== posix
231 && (EQ (cp
->syntax_table
, Qt
)
232 || EQ (cp
->syntax_table
, BVAR (current_buffer
, syntax_table
)))
233 && !NILP (Fequal (cp
->whitespace_regexp
, Vsearch_spaces_regexp
))
234 && cp
->buf
.charset_unibyte
== charset_unibyte
)
237 /* If we're at the end of the cache, compile into the nil cell
238 we found, or the last (least recently used) cell with a
243 compile_pattern_1 (cp
, pattern
, translate
, posix
);
248 /* When we get here, cp (aka *cpp) contains the compiled pattern,
249 either because we found it in the cache or because we just compiled it.
250 Move it to the front of the queue to mark it as most recently used. */
252 cp
->next
= searchbuf_head
;
255 /* Advise the searching functions about the space we have allocated
256 for register data. */
258 re_set_registers (&cp
->buf
, regp
, regp
->num_regs
, regp
->start
, regp
->end
);
260 /* The compiled pattern can be used both for multibyte and unibyte
261 target. But, we have to tell which the pattern is used for. */
262 cp
->buf
.target_multibyte
= multibyte
;
269 looking_at_1 (Lisp_Object string
, int posix
)
272 unsigned char *p1
, *p2
;
274 register ptrdiff_t i
;
275 struct re_pattern_buffer
*bufp
;
277 if (running_asynch_code
)
280 /* This is so set_image_of_range_1 in regex.c can find the EQV table. */
281 set_char_table_extras (BVAR (current_buffer
, case_canon_table
), 2,
282 BVAR (current_buffer
, case_eqv_table
));
284 CHECK_STRING (string
);
285 bufp
= compile_pattern (string
,
286 (NILP (Vinhibit_changing_match_data
)
287 ? &search_regs
: NULL
),
288 (!NILP (BVAR (current_buffer
, case_fold_search
))
289 ? BVAR (current_buffer
, case_canon_table
) : Qnil
),
291 !NILP (BVAR (current_buffer
, enable_multibyte_characters
)));
294 QUIT
; /* Do a pending quit right away, to avoid paradoxical behavior */
296 /* Get pointers and sizes of the two strings
297 that make up the visible portion of the buffer. */
300 s1
= GPT_BYTE
- BEGV_BYTE
;
302 s2
= ZV_BYTE
- GPT_BYTE
;
306 s2
= ZV_BYTE
- BEGV_BYTE
;
311 s1
= ZV_BYTE
- BEGV_BYTE
;
315 re_match_object
= Qnil
;
317 i
= re_match_2 (bufp
, (char *) p1
, s1
, (char *) p2
, s2
,
319 (NILP (Vinhibit_changing_match_data
)
320 ? &search_regs
: NULL
),
321 ZV_BYTE
- BEGV_BYTE
);
327 val
= (0 <= i
? Qt
: Qnil
);
328 if (NILP (Vinhibit_changing_match_data
) && i
>= 0)
329 for (i
= 0; i
< search_regs
.num_regs
; i
++)
330 if (search_regs
.start
[i
] >= 0)
333 = BYTE_TO_CHAR (search_regs
.start
[i
] + BEGV_BYTE
);
335 = BYTE_TO_CHAR (search_regs
.end
[i
] + BEGV_BYTE
);
338 /* Set last_thing_searched only when match data is changed. */
339 if (NILP (Vinhibit_changing_match_data
))
340 XSETBUFFER (last_thing_searched
, current_buffer
);
345 DEFUN ("looking-at", Flooking_at
, Slooking_at
, 1, 1, 0,
346 doc
: /* Return t if text after point matches regular expression REGEXP.
347 This function modifies the match data that `match-beginning',
348 `match-end' and `match-data' access; save and restore the match
349 data if you want to preserve them. */)
352 return looking_at_1 (regexp
, 0);
355 DEFUN ("posix-looking-at", Fposix_looking_at
, Sposix_looking_at
, 1, 1, 0,
356 doc
: /* Return t if text after point matches regular expression REGEXP.
357 Find the longest match, in accord with Posix regular expression rules.
358 This function modifies the match data that `match-beginning',
359 `match-end' and `match-data' access; save and restore the match
360 data if you want to preserve them. */)
363 return looking_at_1 (regexp
, 1);
367 string_match_1 (Lisp_Object regexp
, Lisp_Object string
, Lisp_Object start
, int posix
)
370 struct re_pattern_buffer
*bufp
;
372 ptrdiff_t pos_byte
, i
;
374 if (running_asynch_code
)
377 CHECK_STRING (regexp
);
378 CHECK_STRING (string
);
381 pos
= 0, pos_byte
= 0;
384 ptrdiff_t len
= SCHARS (string
);
386 CHECK_NUMBER (start
);
388 if (pos
< 0 && -pos
<= len
)
390 else if (0 > pos
|| pos
> len
)
391 args_out_of_range (string
, start
);
392 pos_byte
= string_char_to_byte (string
, pos
);
395 /* This is so set_image_of_range_1 in regex.c can find the EQV table. */
396 set_char_table_extras (BVAR (current_buffer
, case_canon_table
), 2,
397 BVAR (current_buffer
, case_eqv_table
));
399 bufp
= compile_pattern (regexp
,
400 (NILP (Vinhibit_changing_match_data
)
401 ? &search_regs
: NULL
),
402 (!NILP (BVAR (current_buffer
, case_fold_search
))
403 ? BVAR (current_buffer
, case_canon_table
) : Qnil
),
405 STRING_MULTIBYTE (string
));
407 re_match_object
= string
;
409 val
= re_search (bufp
, SSDATA (string
),
410 SBYTES (string
), pos_byte
,
411 SBYTES (string
) - pos_byte
,
412 (NILP (Vinhibit_changing_match_data
)
413 ? &search_regs
: NULL
));
416 /* Set last_thing_searched only when match data is changed. */
417 if (NILP (Vinhibit_changing_match_data
))
418 last_thing_searched
= Qt
;
422 if (val
< 0) return Qnil
;
424 if (NILP (Vinhibit_changing_match_data
))
425 for (i
= 0; i
< search_regs
.num_regs
; i
++)
426 if (search_regs
.start
[i
] >= 0)
429 = string_byte_to_char (string
, search_regs
.start
[i
]);
431 = string_byte_to_char (string
, search_regs
.end
[i
]);
434 return make_number (string_byte_to_char (string
, val
));
437 DEFUN ("string-match", Fstring_match
, Sstring_match
, 2, 3, 0,
438 doc
: /* Return index of start of first match for REGEXP in STRING, or nil.
439 Matching ignores case if `case-fold-search' is non-nil.
440 If third arg START is non-nil, start search at that index in STRING.
441 For index of first char beyond the match, do (match-end 0).
442 `match-end' and `match-beginning' also give indices of substrings
443 matched by parenthesis constructs in the pattern.
445 You can use the function `match-string' to extract the substrings
446 matched by the parenthesis constructions in REGEXP. */)
447 (Lisp_Object regexp
, Lisp_Object string
, Lisp_Object start
)
449 return string_match_1 (regexp
, string
, start
, 0);
452 DEFUN ("posix-string-match", Fposix_string_match
, Sposix_string_match
, 2, 3, 0,
453 doc
: /* Return index of start of first match for REGEXP in STRING, or nil.
454 Find the longest match, in accord with Posix regular expression rules.
455 Case is ignored if `case-fold-search' is non-nil in the current buffer.
456 If third arg START is non-nil, start search at that index in STRING.
457 For index of first char beyond the match, do (match-end 0).
458 `match-end' and `match-beginning' also give indices of substrings
459 matched by parenthesis constructs in the pattern. */)
460 (Lisp_Object regexp
, Lisp_Object string
, Lisp_Object start
)
462 return string_match_1 (regexp
, string
, start
, 1);
465 /* Match REGEXP against STRING, searching all of STRING,
466 and return the index of the match, or negative on failure.
467 This does not clobber the match data. */
470 fast_string_match (Lisp_Object regexp
, Lisp_Object string
)
473 struct re_pattern_buffer
*bufp
;
475 bufp
= compile_pattern (regexp
, 0, Qnil
,
476 0, STRING_MULTIBYTE (string
));
478 re_match_object
= string
;
480 val
= re_search (bufp
, SSDATA (string
),
487 /* Match REGEXP against STRING, searching all of STRING ignoring case,
488 and return the index of the match, or negative on failure.
489 This does not clobber the match data.
490 We assume that STRING contains single-byte characters. */
493 fast_c_string_match_ignore_case (Lisp_Object regexp
,
494 const char *string
, ptrdiff_t len
)
497 struct re_pattern_buffer
*bufp
;
499 regexp
= string_make_unibyte (regexp
);
500 re_match_object
= Qt
;
501 bufp
= compile_pattern (regexp
, 0,
502 Vascii_canon_table
, 0,
505 val
= re_search (bufp
, string
, len
, 0, len
, 0);
510 /* Like fast_string_match but ignore case. */
513 fast_string_match_ignore_case (Lisp_Object regexp
, Lisp_Object string
)
516 struct re_pattern_buffer
*bufp
;
518 bufp
= compile_pattern (regexp
, 0, Vascii_canon_table
,
519 0, STRING_MULTIBYTE (string
));
521 re_match_object
= string
;
523 val
= re_search (bufp
, SSDATA (string
),
530 /* Match REGEXP against the characters after POS to LIMIT, and return
531 the number of matched characters. If STRING is non-nil, match
532 against the characters in it. In that case, POS and LIMIT are
533 indices into the string. This function doesn't modify the match
537 fast_looking_at (Lisp_Object regexp
, ptrdiff_t pos
, ptrdiff_t pos_byte
, ptrdiff_t limit
, ptrdiff_t limit_byte
, Lisp_Object string
)
540 struct re_pattern_buffer
*buf
;
541 unsigned char *p1
, *p2
;
545 if (STRINGP (string
))
548 pos_byte
= string_char_to_byte (string
, pos
);
550 limit_byte
= string_char_to_byte (string
, limit
);
554 s2
= SBYTES (string
);
555 re_match_object
= string
;
556 multibyte
= STRING_MULTIBYTE (string
);
561 pos_byte
= CHAR_TO_BYTE (pos
);
563 limit_byte
= CHAR_TO_BYTE (limit
);
564 pos_byte
-= BEGV_BYTE
;
565 limit_byte
-= BEGV_BYTE
;
567 s1
= GPT_BYTE
- BEGV_BYTE
;
569 s2
= ZV_BYTE
- GPT_BYTE
;
573 s2
= ZV_BYTE
- BEGV_BYTE
;
578 s1
= ZV_BYTE
- BEGV_BYTE
;
581 re_match_object
= Qnil
;
582 multibyte
= ! NILP (BVAR (current_buffer
, enable_multibyte_characters
));
585 buf
= compile_pattern (regexp
, 0, Qnil
, 0, multibyte
);
587 len
= re_match_2 (buf
, (char *) p1
, s1
, (char *) p2
, s2
,
588 pos_byte
, NULL
, limit_byte
);
595 /* The newline cache: remembering which sections of text have no newlines. */
597 /* If the user has requested newline caching, make sure it's on.
598 Otherwise, make sure it's off.
599 This is our cheezy way of associating an action with the change of
600 state of a buffer-local variable. */
602 newline_cache_on_off (struct buffer
*buf
)
604 if (NILP (BVAR (buf
, cache_long_line_scans
)))
606 /* It should be off. */
607 if (buf
->newline_cache
)
609 free_region_cache (buf
->newline_cache
);
610 buf
->newline_cache
= 0;
615 /* It should be on. */
616 if (buf
->newline_cache
== 0)
617 buf
->newline_cache
= new_region_cache ();
622 /* Search for COUNT instances of the character TARGET between START and END.
624 If COUNT is positive, search forwards; END must be >= START.
625 If COUNT is negative, search backwards for the -COUNTth instance;
626 END must be <= START.
627 If COUNT is zero, do anything you please; run rogue, for all I care.
629 If END is zero, use BEGV or ZV instead, as appropriate for the
630 direction indicated by COUNT.
632 If we find COUNT instances, set *SHORTAGE to zero, and return the
633 position past the COUNTth match. Note that for reverse motion
634 this is not the same as the usual convention for Emacs motion commands.
636 If we don't find COUNT instances before reaching END, set *SHORTAGE
637 to the number of TARGETs left unfound, and return END.
639 If ALLOW_QUIT, set immediate_quit. That's good to do
640 except when inside redisplay. */
643 scan_buffer (int target
, ptrdiff_t start
, ptrdiff_t end
,
644 ptrdiff_t count
, ptrdiff_t *shortage
, bool allow_quit
)
646 struct region_cache
*newline_cache
;
657 if (! end
) end
= BEGV
;
660 newline_cache_on_off (current_buffer
);
661 newline_cache
= current_buffer
->newline_cache
;
666 immediate_quit
= allow_quit
;
671 /* Our innermost scanning loop is very simple; it doesn't know
672 about gaps, buffer ends, or the newline cache. ceiling is
673 the position of the last character before the next such
674 obstacle --- the last character the dumb search loop should
676 ptrdiff_t ceiling_byte
= CHAR_TO_BYTE (end
) - 1;
677 ptrdiff_t start_byte
;
680 /* If we're looking for a newline, consult the newline cache
681 to see where we can avoid some scanning. */
682 if (target
== '\n' && newline_cache
)
684 ptrdiff_t next_change
;
686 while (region_cache_forward
687 (current_buffer
, newline_cache
, start
, &next_change
))
689 immediate_quit
= allow_quit
;
691 start_byte
= CHAR_TO_BYTE (start
);
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 (CHAR_TO_BYTE (next_change
) - 1, ceiling_byte
);
702 start_byte
= CHAR_TO_BYTE (start
);
704 /* The dumb loop can only scan text stored in contiguous
705 bytes. BUFFER_CEILING_OF returns the last character
706 position that is contiguous, so the ceiling is the
707 position after that. */
708 tem
= BUFFER_CEILING_OF (start_byte
);
709 ceiling_byte
= min (tem
, ceiling_byte
);
712 /* The termination address of the dumb loop. */
713 register unsigned char *ceiling_addr
714 = BYTE_POS_ADDR (ceiling_byte
) + 1;
715 register unsigned char *cursor
716 = BYTE_POS_ADDR (start_byte
);
717 unsigned char *base
= cursor
;
719 while (cursor
< ceiling_addr
)
721 unsigned char *scan_start
= cursor
;
724 while (*cursor
!= target
&& ++cursor
< ceiling_addr
)
727 /* If we're looking for newlines, cache the fact that
728 the region from start to cursor is free of them. */
729 if (target
== '\n' && newline_cache
)
730 know_region_cache (current_buffer
, newline_cache
,
731 BYTE_TO_CHAR (start_byte
+ scan_start
- base
),
732 BYTE_TO_CHAR (start_byte
+ cursor
- base
));
734 /* Did we find the target character? */
735 if (cursor
< ceiling_addr
)
740 return BYTE_TO_CHAR (start_byte
+ cursor
- base
+ 1);
746 start
= BYTE_TO_CHAR (start_byte
+ cursor
- base
);
752 /* The last character to check before the next obstacle. */
753 ptrdiff_t ceiling_byte
= CHAR_TO_BYTE (end
);
754 ptrdiff_t start_byte
;
757 /* Consult the newline cache, if appropriate. */
758 if (target
== '\n' && newline_cache
)
760 ptrdiff_t next_change
;
762 while (region_cache_backward
763 (current_buffer
, newline_cache
, start
, &next_change
))
765 immediate_quit
= allow_quit
;
767 start_byte
= CHAR_TO_BYTE (start
);
769 /* Start should never be at or before end. */
770 if (start_byte
<= ceiling_byte
)
771 start_byte
= ceiling_byte
+ 1;
773 /* Now the text before start is an unknown region, and
774 next_change is the position of the next known region. */
775 ceiling_byte
= max (CHAR_TO_BYTE (next_change
), ceiling_byte
);
778 start_byte
= CHAR_TO_BYTE (start
);
780 /* Stop scanning before the gap. */
781 tem
= BUFFER_FLOOR_OF (start_byte
- 1);
782 ceiling_byte
= max (tem
, ceiling_byte
);
785 /* The termination address of the dumb loop. */
786 register unsigned char *ceiling_addr
= BYTE_POS_ADDR (ceiling_byte
);
787 register unsigned char *cursor
= BYTE_POS_ADDR (start_byte
- 1);
788 unsigned char *base
= cursor
;
790 while (cursor
>= ceiling_addr
)
792 unsigned char *scan_start
= cursor
;
794 while (*cursor
!= target
&& --cursor
>= ceiling_addr
)
797 /* If we're looking for newlines, cache the fact that
798 the region from after the cursor to start is free of them. */
799 if (target
== '\n' && newline_cache
)
800 know_region_cache (current_buffer
, newline_cache
,
801 BYTE_TO_CHAR (start_byte
+ cursor
- base
),
802 BYTE_TO_CHAR (start_byte
+ scan_start
- base
));
804 /* Did we find the target character? */
805 if (cursor
>= ceiling_addr
)
810 return BYTE_TO_CHAR (start_byte
+ cursor
- base
);
816 start
= BYTE_TO_CHAR (start_byte
+ cursor
- base
);
822 *shortage
= count
* direction
;
826 /* Search for COUNT instances of a line boundary, which means either a
827 newline or (if selective display enabled) a carriage return.
828 Start at START. If COUNT is negative, search backwards.
830 We report the resulting position by calling TEMP_SET_PT_BOTH.
832 If we find COUNT instances. we position after (always after,
833 even if scanning backwards) the COUNTth match, and return 0.
835 If we don't find COUNT instances before reaching the end of the
836 buffer (or the beginning, if scanning backwards), we return
837 the number of line boundaries left unfound, and position at
838 the limit we bumped up against.
840 If ALLOW_QUIT, set immediate_quit. That's good to do
841 except in special cases. */
844 scan_newline (ptrdiff_t start
, ptrdiff_t start_byte
,
845 ptrdiff_t limit
, ptrdiff_t limit_byte
,
846 EMACS_INT count
, bool allow_quit
)
848 int direction
= ((count
> 0) ? 1 : -1);
850 unsigned char *cursor
;
854 unsigned char *ceiling_addr
;
856 bool old_immediate_quit
= immediate_quit
;
858 /* The code that follows is like scan_buffer
859 but checks for either newline or carriage return. */
864 start_byte
= CHAR_TO_BYTE (start
);
868 while (start_byte
< limit_byte
)
870 ceiling
= BUFFER_CEILING_OF (start_byte
);
871 ceiling
= min (limit_byte
- 1, ceiling
);
872 ceiling_addr
= BYTE_POS_ADDR (ceiling
) + 1;
873 base
= (cursor
= BYTE_POS_ADDR (start_byte
));
876 while (*cursor
!= '\n' && ++cursor
!= ceiling_addr
)
879 if (cursor
!= ceiling_addr
)
883 immediate_quit
= old_immediate_quit
;
884 start_byte
= start_byte
+ cursor
- base
+ 1;
885 start
= BYTE_TO_CHAR (start_byte
);
886 TEMP_SET_PT_BOTH (start
, start_byte
);
890 if (++cursor
== ceiling_addr
)
896 start_byte
+= cursor
- base
;
901 while (start_byte
> limit_byte
)
903 ceiling
= BUFFER_FLOOR_OF (start_byte
- 1);
904 ceiling
= max (limit_byte
, ceiling
);
905 ceiling_addr
= BYTE_POS_ADDR (ceiling
) - 1;
906 base
= (cursor
= BYTE_POS_ADDR (start_byte
- 1) + 1);
909 while (--cursor
!= ceiling_addr
&& *cursor
!= '\n')
912 if (cursor
!= ceiling_addr
)
916 immediate_quit
= old_immediate_quit
;
917 /* Return the position AFTER the match we found. */
918 start_byte
= start_byte
+ cursor
- base
+ 1;
919 start
= BYTE_TO_CHAR (start_byte
);
920 TEMP_SET_PT_BOTH (start
, start_byte
);
927 /* Here we add 1 to compensate for the last decrement
928 of CURSOR, which took it past the valid range. */
929 start_byte
+= cursor
- base
+ 1;
933 TEMP_SET_PT_BOTH (limit
, limit_byte
);
934 immediate_quit
= old_immediate_quit
;
936 return count
* direction
;
940 find_next_newline_no_quit (ptrdiff_t from
, ptrdiff_t cnt
)
942 return scan_buffer ('\n', from
, 0, cnt
, (ptrdiff_t *) 0, 0);
945 /* Like find_next_newline, but returns position before the newline,
946 not after, and only search up to TO. This isn't just
947 find_next_newline (...)-1, because you might hit TO. */
950 find_before_next_newline (ptrdiff_t from
, ptrdiff_t to
, ptrdiff_t cnt
)
953 ptrdiff_t pos
= scan_buffer ('\n', from
, to
, cnt
, &shortage
, 1);
961 /* Subroutines of Lisp buffer search functions. */
964 search_command (Lisp_Object string
, Lisp_Object bound
, Lisp_Object noerror
,
965 Lisp_Object count
, int direction
, int RE
, int posix
)
967 register EMACS_INT np
;
970 EMACS_INT n
= direction
;
974 CHECK_NUMBER (count
);
978 CHECK_STRING (string
);
982 lim
= ZV
, lim_byte
= ZV_BYTE
;
984 lim
= BEGV
, lim_byte
= BEGV_BYTE
;
988 CHECK_NUMBER_COERCE_MARKER (bound
);
990 if (n
> 0 ? lim
< PT
: lim
> PT
)
991 error ("Invalid search bound (wrong side of point)");
993 lim
= ZV
, lim_byte
= ZV_BYTE
;
995 lim
= BEGV
, lim_byte
= BEGV_BYTE
;
997 lim_byte
= CHAR_TO_BYTE (lim
);
1000 /* This is so set_image_of_range_1 in regex.c can find the EQV table. */
1001 set_char_table_extras (BVAR (current_buffer
, case_canon_table
), 2,
1002 BVAR (current_buffer
, case_eqv_table
));
1004 np
= search_buffer (string
, PT
, PT_BYTE
, lim
, lim_byte
, n
, RE
,
1005 (!NILP (BVAR (current_buffer
, case_fold_search
))
1006 ? BVAR (current_buffer
, case_canon_table
)
1008 (!NILP (BVAR (current_buffer
, case_fold_search
))
1009 ? BVAR (current_buffer
, case_eqv_table
)
1015 xsignal1 (Qsearch_failed
, string
);
1017 if (!EQ (noerror
, Qt
))
1019 if (lim
< BEGV
|| lim
> ZV
)
1021 SET_PT_BOTH (lim
, lim_byte
);
1023 #if 0 /* This would be clean, but maybe programs depend on
1024 a value of nil here. */
1032 if (np
< BEGV
|| np
> ZV
)
1037 return make_number (np
);
1040 /* Return 1 if REGEXP it matches just one constant string. */
1043 trivial_regexp_p (Lisp_Object regexp
)
1045 ptrdiff_t len
= SBYTES (regexp
);
1046 unsigned char *s
= SDATA (regexp
);
1051 case '.': case '*': case '+': case '?': case '[': case '^': case '$':
1058 case '|': case '(': case ')': case '`': case '\'': case 'b':
1059 case 'B': case '<': case '>': case 'w': case 'W': case 's':
1060 case 'S': case '=': case '{': case '}': case '_':
1061 case 'c': case 'C': /* for categoryspec and notcategoryspec */
1062 case '1': case '2': case '3': case '4': case '5':
1063 case '6': case '7': case '8': case '9':
1071 /* Search for the n'th occurrence of STRING in the current buffer,
1072 starting at position POS and stopping at position LIM,
1073 treating STRING as a literal string if RE is false or as
1074 a regular expression if RE is true.
1076 If N is positive, searching is forward and LIM must be greater than POS.
1077 If N is negative, searching is backward and LIM must be less than POS.
1079 Returns -x if x occurrences remain to be found (x > 0),
1080 or else the position at the beginning of the Nth occurrence
1081 (if searching backward) or the end (if searching forward).
1083 POSIX is nonzero if we want full backtracking (POSIX style)
1084 for this pattern. 0 means backtrack only enough to get a valid match. */
1086 #define TRANSLATE(out, trt, d) \
1092 temp = Faref (trt, make_number (d)); \
1093 if (INTEGERP (temp)) \
1094 out = XINT (temp); \
1103 /* Only used in search_buffer, to record the end position of the match
1104 when searching regexps and SEARCH_REGS should not be changed
1105 (i.e. Vinhibit_changing_match_data is non-nil). */
1106 static struct re_registers search_regs_1
;
1109 search_buffer (Lisp_Object string
, ptrdiff_t pos
, ptrdiff_t pos_byte
,
1110 ptrdiff_t lim
, ptrdiff_t lim_byte
, EMACS_INT n
,
1111 int RE
, Lisp_Object trt
, Lisp_Object inverse_trt
, int posix
)
1113 ptrdiff_t len
= SCHARS (string
);
1114 ptrdiff_t len_byte
= SBYTES (string
);
1115 register ptrdiff_t i
;
1117 if (running_asynch_code
)
1118 save_search_regs ();
1120 /* Searching 0 times means don't move. */
1121 /* Null string is found at starting position. */
1122 if (len
== 0 || n
== 0)
1124 set_search_regs (pos_byte
, 0);
1128 if (RE
&& !(trivial_regexp_p (string
) && NILP (Vsearch_spaces_regexp
)))
1130 unsigned char *p1
, *p2
;
1132 struct re_pattern_buffer
*bufp
;
1134 bufp
= compile_pattern (string
,
1135 (NILP (Vinhibit_changing_match_data
)
1136 ? &search_regs
: &search_regs_1
),
1138 !NILP (BVAR (current_buffer
, enable_multibyte_characters
)));
1140 immediate_quit
= 1; /* Quit immediately if user types ^G,
1141 because letting this function finish
1142 can take too long. */
1143 QUIT
; /* Do a pending quit right away,
1144 to avoid paradoxical behavior */
1145 /* Get pointers and sizes of the two strings
1146 that make up the visible portion of the buffer. */
1149 s1
= GPT_BYTE
- BEGV_BYTE
;
1151 s2
= ZV_BYTE
- GPT_BYTE
;
1155 s2
= ZV_BYTE
- BEGV_BYTE
;
1160 s1
= ZV_BYTE
- BEGV_BYTE
;
1163 re_match_object
= Qnil
;
1169 val
= re_search_2 (bufp
, (char *) p1
, s1
, (char *) p2
, s2
,
1170 pos_byte
- BEGV_BYTE
, lim_byte
- pos_byte
,
1171 (NILP (Vinhibit_changing_match_data
)
1172 ? &search_regs
: &search_regs_1
),
1173 /* Don't allow match past current point */
1174 pos_byte
- BEGV_BYTE
);
1177 matcher_overflow ();
1181 if (NILP (Vinhibit_changing_match_data
))
1183 pos_byte
= search_regs
.start
[0] + BEGV_BYTE
;
1184 for (i
= 0; i
< search_regs
.num_regs
; i
++)
1185 if (search_regs
.start
[i
] >= 0)
1187 search_regs
.start
[i
]
1188 = BYTE_TO_CHAR (search_regs
.start
[i
] + BEGV_BYTE
);
1190 = BYTE_TO_CHAR (search_regs
.end
[i
] + BEGV_BYTE
);
1192 XSETBUFFER (last_thing_searched
, current_buffer
);
1193 /* Set pos to the new position. */
1194 pos
= search_regs
.start
[0];
1198 pos_byte
= search_regs_1
.start
[0] + BEGV_BYTE
;
1199 /* Set pos to the new position. */
1200 pos
= BYTE_TO_CHAR (search_regs_1
.start
[0] + BEGV_BYTE
);
1214 val
= re_search_2 (bufp
, (char *) p1
, s1
, (char *) p2
, s2
,
1215 pos_byte
- BEGV_BYTE
, lim_byte
- pos_byte
,
1216 (NILP (Vinhibit_changing_match_data
)
1217 ? &search_regs
: &search_regs_1
),
1218 lim_byte
- BEGV_BYTE
);
1221 matcher_overflow ();
1225 if (NILP (Vinhibit_changing_match_data
))
1227 pos_byte
= search_regs
.end
[0] + BEGV_BYTE
;
1228 for (i
= 0; i
< search_regs
.num_regs
; i
++)
1229 if (search_regs
.start
[i
] >= 0)
1231 search_regs
.start
[i
]
1232 = BYTE_TO_CHAR (search_regs
.start
[i
] + BEGV_BYTE
);
1234 = BYTE_TO_CHAR (search_regs
.end
[i
] + BEGV_BYTE
);
1236 XSETBUFFER (last_thing_searched
, current_buffer
);
1237 pos
= search_regs
.end
[0];
1241 pos_byte
= search_regs_1
.end
[0] + BEGV_BYTE
;
1242 pos
= BYTE_TO_CHAR (search_regs_1
.end
[0] + BEGV_BYTE
);
1255 else /* non-RE case */
1257 unsigned char *raw_pattern
, *pat
;
1258 ptrdiff_t raw_pattern_size
;
1259 ptrdiff_t raw_pattern_size_byte
;
1260 unsigned char *patbuf
;
1261 int multibyte
= !NILP (BVAR (current_buffer
, enable_multibyte_characters
));
1262 unsigned char *base_pat
;
1263 /* Set to positive if we find a non-ASCII char that need
1264 translation. Otherwise set to zero later. */
1266 int boyer_moore_ok
= 1;
1268 /* MULTIBYTE says whether the text to be searched is multibyte.
1269 We must convert PATTERN to match that, or we will not really
1270 find things right. */
1272 if (multibyte
== STRING_MULTIBYTE (string
))
1274 raw_pattern
= SDATA (string
);
1275 raw_pattern_size
= SCHARS (string
);
1276 raw_pattern_size_byte
= SBYTES (string
);
1280 raw_pattern_size
= SCHARS (string
);
1281 raw_pattern_size_byte
1282 = count_size_as_multibyte (SDATA (string
),
1284 raw_pattern
= alloca (raw_pattern_size_byte
+ 1);
1285 copy_text (SDATA (string
), raw_pattern
,
1286 SCHARS (string
), 0, 1);
1290 /* Converting multibyte to single-byte.
1292 ??? Perhaps this conversion should be done in a special way
1293 by subtracting nonascii-insert-offset from each non-ASCII char,
1294 so that only the multibyte chars which really correspond to
1295 the chosen single-byte character set can possibly match. */
1296 raw_pattern_size
= SCHARS (string
);
1297 raw_pattern_size_byte
= SCHARS (string
);
1298 raw_pattern
= alloca (raw_pattern_size
+ 1);
1299 copy_text (SDATA (string
), raw_pattern
,
1300 SBYTES (string
), 1, 0);
1303 /* Copy and optionally translate the pattern. */
1304 len
= raw_pattern_size
;
1305 len_byte
= raw_pattern_size_byte
;
1306 patbuf
= alloca (len
* MAX_MULTIBYTE_LENGTH
);
1308 base_pat
= raw_pattern
;
1311 /* Fill patbuf by translated characters in STRING while
1312 checking if we can use boyer-moore search. If TRT is
1313 non-nil, we can use boyer-moore search only if TRT can be
1314 represented by the byte array of 256 elements. For that,
1315 all non-ASCII case-equivalents of all case-sensitive
1316 characters in STRING must belong to the same charset and
1321 unsigned char str_base
[MAX_MULTIBYTE_LENGTH
], *str
;
1322 int c
, translated
, inverse
;
1323 int in_charlen
, charlen
;
1325 /* If we got here and the RE flag is set, it's because we're
1326 dealing with a regexp known to be trivial, so the backslash
1327 just quotes the next character. */
1328 if (RE
&& *base_pat
== '\\')
1336 c
= STRING_CHAR_AND_LENGTH (base_pat
, in_charlen
);
1341 charlen
= in_charlen
;
1345 /* Translate the character. */
1346 TRANSLATE (translated
, trt
, c
);
1347 charlen
= CHAR_STRING (translated
, str_base
);
1350 /* Check if C has any other case-equivalents. */
1351 TRANSLATE (inverse
, inverse_trt
, c
);
1352 /* If so, check if we can use boyer-moore. */
1353 if (c
!= inverse
&& boyer_moore_ok
)
1355 /* Check if all equivalents belong to the same
1356 group of characters. Note that the check of C
1357 itself is done by the last iteration. */
1358 int this_char_base
= -1;
1360 while (boyer_moore_ok
)
1362 if (ASCII_BYTE_P (inverse
))
1364 if (this_char_base
> 0)
1369 else if (CHAR_BYTE8_P (inverse
))
1370 /* Boyer-moore search can't handle a
1371 translation of an eight-bit
1374 else if (this_char_base
< 0)
1376 this_char_base
= inverse
& ~0x3F;
1378 char_base
= this_char_base
;
1379 else if (this_char_base
!= char_base
)
1382 else if ((inverse
& ~0x3F) != this_char_base
)
1386 TRANSLATE (inverse
, inverse_trt
, inverse
);
1391 /* Store this character into the translated pattern. */
1392 memcpy (pat
, str
, charlen
);
1394 base_pat
+= in_charlen
;
1395 len_byte
-= in_charlen
;
1398 /* If char_base is still negative we didn't find any translated
1399 non-ASCII characters. */
1405 /* Unibyte buffer. */
1409 int c
, translated
, inverse
;
1411 /* If we got here and the RE flag is set, it's because we're
1412 dealing with a regexp known to be trivial, so the backslash
1413 just quotes the next character. */
1414 if (RE
&& *base_pat
== '\\')
1421 TRANSLATE (translated
, trt
, c
);
1422 *pat
++ = translated
;
1423 /* Check that none of C's equivalents violates the
1424 assumptions of boyer_moore. */
1425 TRANSLATE (inverse
, inverse_trt
, c
);
1428 if (inverse
>= 0200)
1435 TRANSLATE (inverse
, inverse_trt
, inverse
);
1440 len_byte
= pat
- patbuf
;
1441 pat
= base_pat
= patbuf
;
1444 return boyer_moore (n
, pat
, len_byte
, trt
, inverse_trt
,
1448 return simple_search (n
, pat
, raw_pattern_size
, len_byte
, trt
,
1449 pos
, pos_byte
, lim
, lim_byte
);
1453 /* Do a simple string search N times for the string PAT,
1454 whose length is LEN/LEN_BYTE,
1455 from buffer position POS/POS_BYTE until LIM/LIM_BYTE.
1456 TRT is the translation table.
1458 Return the character position where the match is found.
1459 Otherwise, if M matches remained to be found, return -M.
1461 This kind of search works regardless of what is in PAT and
1462 regardless of what is in TRT. It is used in cases where
1463 boyer_moore cannot work. */
1466 simple_search (EMACS_INT n
, unsigned char *pat
,
1467 ptrdiff_t len
, ptrdiff_t len_byte
, Lisp_Object trt
,
1468 ptrdiff_t pos
, ptrdiff_t pos_byte
,
1469 ptrdiff_t lim
, ptrdiff_t lim_byte
)
1471 int multibyte
= ! NILP (BVAR (current_buffer
, enable_multibyte_characters
));
1472 int forward
= n
> 0;
1473 /* Number of buffer bytes matched. Note that this may be different
1474 from len_byte in a multibyte buffer. */
1475 ptrdiff_t match_byte
= PTRDIFF_MIN
;
1477 if (lim
> pos
&& multibyte
)
1482 /* Try matching at position POS. */
1483 ptrdiff_t this_pos
= pos
;
1484 ptrdiff_t this_pos_byte
= pos_byte
;
1485 ptrdiff_t this_len
= len
;
1486 unsigned char *p
= pat
;
1487 if (pos
+ len
> lim
|| pos_byte
+ len_byte
> lim_byte
)
1490 while (this_len
> 0)
1492 int charlen
, buf_charlen
;
1495 pat_ch
= STRING_CHAR_AND_LENGTH (p
, charlen
);
1496 buf_ch
= STRING_CHAR_AND_LENGTH (BYTE_POS_ADDR (this_pos_byte
),
1498 TRANSLATE (buf_ch
, trt
, buf_ch
);
1500 if (buf_ch
!= pat_ch
)
1506 this_pos_byte
+= buf_charlen
;
1512 match_byte
= this_pos_byte
- pos_byte
;
1514 pos_byte
+= match_byte
;
1518 INC_BOTH (pos
, pos_byte
);
1528 /* Try matching at position POS. */
1529 ptrdiff_t this_pos
= pos
;
1530 ptrdiff_t this_len
= len
;
1531 unsigned char *p
= pat
;
1533 if (pos
+ len
> lim
)
1536 while (this_len
> 0)
1539 int buf_ch
= FETCH_BYTE (this_pos
);
1540 TRANSLATE (buf_ch
, trt
, buf_ch
);
1542 if (buf_ch
!= pat_ch
)
1561 /* Backwards search. */
1562 else if (lim
< pos
&& multibyte
)
1567 /* Try matching at position POS. */
1568 ptrdiff_t this_pos
= pos
;
1569 ptrdiff_t this_pos_byte
= pos_byte
;
1570 ptrdiff_t this_len
= len
;
1571 const unsigned char *p
= pat
+ len_byte
;
1573 if (this_pos
- len
< lim
|| (pos_byte
- len_byte
) < lim_byte
)
1576 while (this_len
> 0)
1580 DEC_BOTH (this_pos
, this_pos_byte
);
1581 PREV_CHAR_BOUNDARY (p
, pat
);
1582 pat_ch
= STRING_CHAR (p
);
1583 buf_ch
= STRING_CHAR (BYTE_POS_ADDR (this_pos_byte
));
1584 TRANSLATE (buf_ch
, trt
, buf_ch
);
1586 if (buf_ch
!= pat_ch
)
1594 match_byte
= pos_byte
- this_pos_byte
;
1596 pos_byte
= this_pos_byte
;
1600 DEC_BOTH (pos
, pos_byte
);
1610 /* Try matching at position POS. */
1611 ptrdiff_t this_pos
= pos
- len
;
1612 ptrdiff_t this_len
= len
;
1613 unsigned char *p
= pat
;
1618 while (this_len
> 0)
1621 int buf_ch
= FETCH_BYTE (this_pos
);
1622 TRANSLATE (buf_ch
, trt
, buf_ch
);
1624 if (buf_ch
!= pat_ch
)
1646 eassert (match_byte
!= PTRDIFF_MIN
);
1648 set_search_regs ((multibyte
? pos_byte
: pos
) - match_byte
, match_byte
);
1650 set_search_regs (multibyte
? pos_byte
: pos
, match_byte
);
1660 /* Do Boyer-Moore search N times for the string BASE_PAT,
1661 whose length is LEN_BYTE,
1662 from buffer position POS_BYTE until LIM_BYTE.
1663 DIRECTION says which direction we search in.
1664 TRT and INVERSE_TRT are translation tables.
1665 Characters in PAT are already translated by TRT.
1667 This kind of search works if all the characters in BASE_PAT that
1668 have nontrivial translation are the same aside from the last byte.
1669 This makes it possible to translate just the last byte of a
1670 character, and do so after just a simple test of the context.
1671 CHAR_BASE is nonzero if there is such a non-ASCII character.
1673 If that criterion is not satisfied, do not call this function. */
1676 boyer_moore (EMACS_INT n
, unsigned char *base_pat
,
1678 Lisp_Object trt
, Lisp_Object inverse_trt
,
1679 ptrdiff_t pos_byte
, ptrdiff_t lim_byte
,
1682 int direction
= ((n
> 0) ? 1 : -1);
1683 register ptrdiff_t dirlen
;
1685 int stride_for_teases
= 0;
1687 register unsigned char *cursor
, *p_limit
;
1688 register ptrdiff_t i
;
1690 unsigned char *pat
, *pat_end
;
1691 int multibyte
= ! NILP (BVAR (current_buffer
, enable_multibyte_characters
));
1693 unsigned char simple_translate
[0400];
1694 /* These are set to the preceding bytes of a byte to be translated
1695 if char_base is nonzero. As the maximum byte length of a
1696 multibyte character is 5, we have to check at most four previous
1698 int translate_prev_byte1
= 0;
1699 int translate_prev_byte2
= 0;
1700 int translate_prev_byte3
= 0;
1702 /* The general approach is that we are going to maintain that we know
1703 the first (closest to the present position, in whatever direction
1704 we're searching) character that could possibly be the last
1705 (furthest from present position) character of a valid match. We
1706 advance the state of our knowledge by looking at that character
1707 and seeing whether it indeed matches the last character of the
1708 pattern. If it does, we take a closer look. If it does not, we
1709 move our pointer (to putative last characters) as far as is
1710 logically possible. This amount of movement, which I call a
1711 stride, will be the length of the pattern if the actual character
1712 appears nowhere in the pattern, otherwise it will be the distance
1713 from the last occurrence of that character to the end of the
1714 pattern. If the amount is zero we have a possible match. */
1716 /* Here we make a "mickey mouse" BM table. The stride of the search
1717 is determined only by the last character of the putative match.
1718 If that character does not match, we will stride the proper
1719 distance to propose a match that superimposes it on the last
1720 instance of a character that matches it (per trt), or misses
1721 it entirely if there is none. */
1723 dirlen
= len_byte
* direction
;
1725 /* Record position after the end of the pattern. */
1726 pat_end
= base_pat
+ len_byte
;
1727 /* BASE_PAT points to a character that we start scanning from.
1728 It is the first character in a forward search,
1729 the last character in a backward search. */
1731 base_pat
= pat_end
- 1;
1733 /* A character that does not appear in the pattern induces a
1734 stride equal to the pattern length. */
1735 for (i
= 0; i
< 0400; i
++)
1738 /* We use this for translation, instead of TRT itself.
1739 We fill this in to handle the characters that actually
1740 occur in the pattern. Others don't matter anyway! */
1741 for (i
= 0; i
< 0400; i
++)
1742 simple_translate
[i
] = i
;
1746 /* Setup translate_prev_byte1/2/3/4 from CHAR_BASE. Only a
1747 byte following them are the target of translation. */
1748 unsigned char str
[MAX_MULTIBYTE_LENGTH
];
1749 int cblen
= CHAR_STRING (char_base
, str
);
1751 translate_prev_byte1
= str
[cblen
- 2];
1754 translate_prev_byte2
= str
[cblen
- 3];
1756 translate_prev_byte3
= str
[cblen
- 4];
1763 unsigned char *ptr
= base_pat
+ i
;
1767 /* If the byte currently looking at is the last of a
1768 character to check case-equivalents, set CH to that
1769 character. An ASCII character and a non-ASCII character
1770 matching with CHAR_BASE are to be checked. */
1773 if (ASCII_BYTE_P (*ptr
) || ! multibyte
)
1776 && ((pat_end
- ptr
) == 1 || CHAR_HEAD_P (ptr
[1])))
1778 unsigned char *charstart
= ptr
- 1;
1780 while (! (CHAR_HEAD_P (*charstart
)))
1782 ch
= STRING_CHAR (charstart
);
1783 if (char_base
!= (ch
& ~0x3F))
1787 if (ch
>= 0200 && multibyte
)
1788 j
= (ch
& 0x3F) | 0200;
1793 stride_for_teases
= BM_tab
[j
];
1795 BM_tab
[j
] = dirlen
- i
;
1796 /* A translation table is accompanied by its inverse -- see
1797 comment following downcase_table for details. */
1800 int starting_ch
= ch
;
1805 TRANSLATE (ch
, inverse_trt
, ch
);
1806 if (ch
>= 0200 && multibyte
)
1807 j
= (ch
& 0x3F) | 0200;
1811 /* For all the characters that map into CH,
1812 set up simple_translate to map the last byte
1814 simple_translate
[j
] = starting_j
;
1815 if (ch
== starting_ch
)
1817 BM_tab
[j
] = dirlen
- i
;
1826 stride_for_teases
= BM_tab
[j
];
1827 BM_tab
[j
] = dirlen
- i
;
1829 /* stride_for_teases tells how much to stride if we get a
1830 match on the far character but are subsequently
1831 disappointed, by recording what the stride would have been
1832 for that character if the last character had been
1835 pos_byte
+= dirlen
- ((direction
> 0) ? direction
: 0);
1836 /* loop invariant - POS_BYTE points at where last char (first
1837 char if reverse) of pattern would align in a possible match. */
1841 unsigned char *tail_end_ptr
;
1843 /* It's been reported that some (broken) compiler thinks that
1844 Boolean expressions in an arithmetic context are unsigned.
1845 Using an explicit ?1:0 prevents this. */
1846 if ((lim_byte
- pos_byte
- ((direction
> 0) ? 1 : 0)) * direction
1848 return (n
* (0 - direction
));
1849 /* First we do the part we can by pointers (maybe nothing) */
1852 limit
= pos_byte
- dirlen
+ direction
;
1855 limit
= BUFFER_CEILING_OF (limit
);
1856 /* LIMIT is now the last (not beyond-last!) value POS_BYTE
1857 can take on without hitting edge of buffer or the gap. */
1858 limit
= min (limit
, pos_byte
+ 20000);
1859 limit
= min (limit
, lim_byte
- 1);
1863 limit
= BUFFER_FLOOR_OF (limit
);
1864 /* LIMIT is now the last (not beyond-last!) value POS_BYTE
1865 can take on without hitting edge of buffer or the gap. */
1866 limit
= max (limit
, pos_byte
- 20000);
1867 limit
= max (limit
, lim_byte
);
1869 tail_end
= BUFFER_CEILING_OF (pos_byte
) + 1;
1870 tail_end_ptr
= BYTE_POS_ADDR (tail_end
);
1872 if ((limit
- pos_byte
) * direction
> 20)
1876 p_limit
= BYTE_POS_ADDR (limit
);
1877 p2
= (cursor
= BYTE_POS_ADDR (pos_byte
));
1878 /* In this loop, pos + cursor - p2 is the surrogate for pos. */
1879 while (1) /* use one cursor setting as long as i can */
1881 if (direction
> 0) /* worth duplicating */
1883 while (cursor
<= p_limit
)
1885 if (BM_tab
[*cursor
] == 0)
1887 cursor
+= BM_tab
[*cursor
];
1892 while (cursor
>= p_limit
)
1894 if (BM_tab
[*cursor
] == 0)
1896 cursor
+= BM_tab
[*cursor
];
1899 /* If you are here, cursor is beyond the end of the
1900 searched region. You fail to match within the
1901 permitted region and would otherwise try a character
1902 beyond that region. */
1906 i
= dirlen
- direction
;
1909 while ((i
-= direction
) + direction
!= 0)
1912 cursor
-= direction
;
1913 /* Translate only the last byte of a character. */
1915 || ((cursor
== tail_end_ptr
1916 || CHAR_HEAD_P (cursor
[1]))
1917 && (CHAR_HEAD_P (cursor
[0])
1918 /* Check if this is the last byte of
1919 a translatable character. */
1920 || (translate_prev_byte1
== cursor
[-1]
1921 && (CHAR_HEAD_P (translate_prev_byte1
)
1922 || (translate_prev_byte2
== cursor
[-2]
1923 && (CHAR_HEAD_P (translate_prev_byte2
)
1924 || (translate_prev_byte3
== cursor
[-3]))))))))
1925 ch
= simple_translate
[*cursor
];
1934 while ((i
-= direction
) + direction
!= 0)
1936 cursor
-= direction
;
1937 if (pat
[i
] != *cursor
)
1941 cursor
+= dirlen
- i
- direction
; /* fix cursor */
1942 if (i
+ direction
== 0)
1944 ptrdiff_t position
, start
, end
;
1946 cursor
-= direction
;
1948 position
= pos_byte
+ cursor
- p2
+ ((direction
> 0)
1949 ? 1 - len_byte
: 0);
1950 set_search_regs (position
, len_byte
);
1952 if (NILP (Vinhibit_changing_match_data
))
1954 start
= search_regs
.start
[0];
1955 end
= search_regs
.end
[0];
1958 /* If Vinhibit_changing_match_data is non-nil,
1959 search_regs will not be changed. So let's
1960 compute start and end here. */
1962 start
= BYTE_TO_CHAR (position
);
1963 end
= BYTE_TO_CHAR (position
+ len_byte
);
1966 if ((n
-= direction
) != 0)
1967 cursor
+= dirlen
; /* to resume search */
1969 return direction
> 0 ? end
: start
;
1972 cursor
+= stride_for_teases
; /* <sigh> we lose - */
1974 pos_byte
+= cursor
- p2
;
1977 /* Now we'll pick up a clump that has to be done the hard
1978 way because it covers a discontinuity. */
1980 limit
= ((direction
> 0)
1981 ? BUFFER_CEILING_OF (pos_byte
- dirlen
+ 1)
1982 : BUFFER_FLOOR_OF (pos_byte
- dirlen
- 1));
1983 limit
= ((direction
> 0)
1984 ? min (limit
+ len_byte
, lim_byte
- 1)
1985 : max (limit
- len_byte
, lim_byte
));
1986 /* LIMIT is now the last value POS_BYTE can have
1987 and still be valid for a possible match. */
1990 /* This loop can be coded for space rather than
1991 speed because it will usually run only once.
1992 (the reach is at most len + 21, and typically
1993 does not exceed len). */
1994 while ((limit
- pos_byte
) * direction
>= 0)
1996 int ch
= FETCH_BYTE (pos_byte
);
1997 if (BM_tab
[ch
] == 0)
1999 pos_byte
+= BM_tab
[ch
];
2001 break; /* ran off the end */
2004 /* Found what might be a match. */
2005 i
= dirlen
- direction
;
2006 while ((i
-= direction
) + direction
!= 0)
2010 pos_byte
-= direction
;
2011 ptr
= BYTE_POS_ADDR (pos_byte
);
2012 /* Translate only the last byte of a character. */
2014 || ((ptr
== tail_end_ptr
2015 || CHAR_HEAD_P (ptr
[1]))
2016 && (CHAR_HEAD_P (ptr
[0])
2017 /* Check if this is the last byte of a
2018 translatable character. */
2019 || (translate_prev_byte1
== ptr
[-1]
2020 && (CHAR_HEAD_P (translate_prev_byte1
)
2021 || (translate_prev_byte2
== ptr
[-2]
2022 && (CHAR_HEAD_P (translate_prev_byte2
)
2023 || translate_prev_byte3
== ptr
[-3])))))))
2024 ch
= simple_translate
[*ptr
];
2030 /* Above loop has moved POS_BYTE part or all the way
2031 back to the first pos (last pos if reverse).
2032 Set it once again at the last (first if reverse) char. */
2033 pos_byte
+= dirlen
- i
- direction
;
2034 if (i
+ direction
== 0)
2036 ptrdiff_t position
, start
, end
;
2037 pos_byte
-= direction
;
2039 position
= pos_byte
+ ((direction
> 0) ? 1 - len_byte
: 0);
2040 set_search_regs (position
, len_byte
);
2042 if (NILP (Vinhibit_changing_match_data
))
2044 start
= search_regs
.start
[0];
2045 end
= search_regs
.end
[0];
2048 /* If Vinhibit_changing_match_data is non-nil,
2049 search_regs will not be changed. So let's
2050 compute start and end here. */
2052 start
= BYTE_TO_CHAR (position
);
2053 end
= BYTE_TO_CHAR (position
+ len_byte
);
2056 if ((n
-= direction
) != 0)
2057 pos_byte
+= dirlen
; /* to resume search */
2059 return direction
> 0 ? end
: start
;
2062 pos_byte
+= stride_for_teases
;
2065 /* We have done one clump. Can we continue? */
2066 if ((lim_byte
- pos_byte
) * direction
< 0)
2067 return ((0 - n
) * direction
);
2069 return BYTE_TO_CHAR (pos_byte
);
2072 /* Record beginning BEG_BYTE and end BEG_BYTE + NBYTES
2073 for the overall match just found in the current buffer.
2074 Also clear out the match data for registers 1 and up. */
2077 set_search_regs (ptrdiff_t beg_byte
, ptrdiff_t nbytes
)
2081 if (!NILP (Vinhibit_changing_match_data
))
2084 /* Make sure we have registers in which to store
2085 the match position. */
2086 if (search_regs
.num_regs
== 0)
2088 search_regs
.start
= xmalloc (2 * sizeof (regoff_t
));
2089 search_regs
.end
= xmalloc (2 * sizeof (regoff_t
));
2090 search_regs
.num_regs
= 2;
2093 /* Clear out the other registers. */
2094 for (i
= 1; i
< search_regs
.num_regs
; i
++)
2096 search_regs
.start
[i
] = -1;
2097 search_regs
.end
[i
] = -1;
2100 search_regs
.start
[0] = BYTE_TO_CHAR (beg_byte
);
2101 search_regs
.end
[0] = BYTE_TO_CHAR (beg_byte
+ nbytes
);
2102 XSETBUFFER (last_thing_searched
, current_buffer
);
2105 DEFUN ("search-backward", Fsearch_backward
, Ssearch_backward
, 1, 4,
2106 "MSearch backward: ",
2107 doc
: /* Search backward from point for STRING.
2108 Set point to the beginning of the occurrence found, and return point.
2109 An optional second argument bounds the search; it is a buffer position.
2110 The match found must not extend before that position.
2111 Optional third argument, if t, means if fail just return nil (no error).
2112 If not nil and not t, position at limit of search and return nil.
2113 Optional fourth argument COUNT, if non-nil, means to search for COUNT
2114 successive occurrences. If COUNT is negative, search forward,
2115 instead of backward, for -COUNT occurrences.
2117 Search case-sensitivity is determined by the value of the variable
2118 `case-fold-search', which see.
2120 See also the functions `match-beginning', `match-end' and `replace-match'. */)
2121 (Lisp_Object string
, Lisp_Object bound
, Lisp_Object noerror
, Lisp_Object count
)
2123 return search_command (string
, bound
, noerror
, count
, -1, 0, 0);
2126 DEFUN ("search-forward", Fsearch_forward
, Ssearch_forward
, 1, 4, "MSearch: ",
2127 doc
: /* Search forward from point for STRING.
2128 Set point to the end of the occurrence found, and return point.
2129 An optional second argument bounds the search; it is a buffer position.
2130 The match found must not extend after that position. A value of nil is
2131 equivalent to (point-max).
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 COUNT, if non-nil, means to search for COUNT
2135 successive occurrences. If COUNT is negative, search backward,
2136 instead of forward, for -COUNT occurrences.
2138 Search case-sensitivity is determined by the value of the variable
2139 `case-fold-search', which see.
2141 See also the functions `match-beginning', `match-end' and `replace-match'. */)
2142 (Lisp_Object string
, Lisp_Object bound
, Lisp_Object noerror
, Lisp_Object count
)
2144 return search_command (string
, bound
, noerror
, count
, 1, 0, 0);
2147 DEFUN ("re-search-backward", Fre_search_backward
, Sre_search_backward
, 1, 4,
2148 "sRE search backward: ",
2149 doc
: /* Search backward from point for match for regular expression REGEXP.
2150 Set point to the beginning of the match, and return point.
2151 The match found is the one starting last in the buffer
2152 and yet ending before the origin of the search.
2153 An optional second argument bounds the search; it is a buffer position.
2154 The match found must start at or after that position.
2155 Optional third argument, if t, means if fail just return nil (no error).
2156 If not nil and not t, move to limit of search and return nil.
2157 Optional fourth argument is repeat count--search for successive occurrences.
2159 Search case-sensitivity is determined by the value of the variable
2160 `case-fold-search', which see.
2162 See also the functions `match-beginning', `match-end', `match-string',
2163 and `replace-match'. */)
2164 (Lisp_Object regexp
, Lisp_Object bound
, Lisp_Object noerror
, Lisp_Object count
)
2166 return search_command (regexp
, bound
, noerror
, count
, -1, 1, 0);
2169 DEFUN ("re-search-forward", Fre_search_forward
, Sre_search_forward
, 1, 4,
2171 doc
: /* Search forward from point for regular expression REGEXP.
2172 Set point to the end of the occurrence found, and return point.
2173 An optional second argument bounds the search; it is a buffer position.
2174 The match found must not extend 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, 0);
2189 DEFUN ("posix-search-backward", Fposix_search_backward
, Sposix_search_backward
, 1, 4,
2190 "sPosix search backward: ",
2191 doc
: /* Search backward from point for match for regular expression REGEXP.
2192 Find the longest match in accord with Posix regular expression rules.
2193 Set point to the beginning of the match, and return point.
2194 The match found is the one starting last in the buffer
2195 and yet ending before the origin of the search.
2196 An optional second argument bounds the search; it is a buffer position.
2197 The match found must start at or after that position.
2198 Optional third argument, if t, means if fail just return nil (no error).
2199 If not nil and not t, move to limit of search and return nil.
2200 Optional fourth argument is repeat count--search for successive occurrences.
2202 Search case-sensitivity is determined by the value of the variable
2203 `case-fold-search', which see.
2205 See also the functions `match-beginning', `match-end', `match-string',
2206 and `replace-match'. */)
2207 (Lisp_Object regexp
, Lisp_Object bound
, Lisp_Object noerror
, Lisp_Object count
)
2209 return search_command (regexp
, bound
, noerror
, count
, -1, 1, 1);
2212 DEFUN ("posix-search-forward", Fposix_search_forward
, Sposix_search_forward
, 1, 4,
2214 doc
: /* Search forward from point for regular expression REGEXP.
2215 Find the longest match in accord with Posix regular expression rules.
2216 Set point to the end of the occurrence found, and return point.
2217 An optional second argument bounds the search; it is a buffer position.
2218 The match found must not extend after that position.
2219 Optional third argument, if t, means if fail just return nil (no error).
2220 If not nil and not t, move to limit of search and return nil.
2221 Optional fourth argument is repeat count--search for successive occurrences.
2223 Search case-sensitivity is determined by the value of the variable
2224 `case-fold-search', which see.
2226 See also the functions `match-beginning', `match-end', `match-string',
2227 and `replace-match'. */)
2228 (Lisp_Object regexp
, Lisp_Object bound
, Lisp_Object noerror
, Lisp_Object count
)
2230 return search_command (regexp
, bound
, noerror
, count
, 1, 1, 1);
2233 DEFUN ("replace-match", Freplace_match
, Sreplace_match
, 1, 5, 0,
2234 doc
: /* Replace text matched by last search with NEWTEXT.
2235 Leave point at the end of the replacement text.
2237 If optional second arg FIXEDCASE is non-nil, do not alter the case of
2238 the replacement text. Otherwise, maybe capitalize the whole text, or
2239 maybe just word initials, based on the replaced text. If the replaced
2240 text has only capital letters and has at least one multiletter word,
2241 convert NEWTEXT to all caps. Otherwise if all words are capitalized
2242 in the replaced text, capitalize each word in NEWTEXT.
2244 If optional third arg LITERAL is non-nil, insert NEWTEXT literally.
2245 Otherwise treat `\\' as special:
2246 `\\&' in NEWTEXT means substitute original matched text.
2247 `\\N' means substitute what matched the Nth `\\(...\\)'.
2248 If Nth parens didn't match, substitute nothing.
2249 `\\\\' means insert one `\\'.
2250 `\\?' is treated literally
2251 (for compatibility with `query-replace-regexp').
2252 Any other character following `\\' signals an error.
2253 Case conversion does not apply to these substitutions.
2255 If optional fourth argument STRING is non-nil, it should be a string
2256 to act on; this should be the string on which the previous match was
2257 done via `string-match'. In this case, `replace-match' creates and
2258 returns a new string, made by copying STRING and replacing the part of
2259 STRING that was matched (the original STRING itself is not altered).
2261 The optional fifth argument SUBEXP specifies a subexpression;
2262 it says to replace just that subexpression with NEWTEXT,
2263 rather than replacing the entire matched text.
2264 This is, in a vague sense, the inverse of using `\\N' in NEWTEXT;
2265 `\\N' copies subexp N into NEWTEXT, but using N as SUBEXP puts
2266 NEWTEXT in place of subexp N.
2267 This is useful only after a regular expression search or match,
2268 since only regular expressions have distinguished subexpressions. */)
2269 (Lisp_Object newtext
, Lisp_Object fixedcase
, Lisp_Object literal
, Lisp_Object string
, Lisp_Object subexp
)
2271 enum { nochange
, all_caps
, cap_initial
} case_action
;
2272 register ptrdiff_t pos
, pos_byte
;
2273 int some_multiletter_word
;
2276 int some_nonuppercase_initial
;
2277 register int c
, prevc
;
2279 ptrdiff_t opoint
, newpoint
;
2281 CHECK_STRING (newtext
);
2283 if (! NILP (string
))
2284 CHECK_STRING (string
);
2286 case_action
= nochange
; /* We tried an initialization */
2287 /* but some C compilers blew it */
2289 if (search_regs
.num_regs
<= 0)
2290 error ("`replace-match' called before any match found");
2296 CHECK_NUMBER (subexp
);
2297 if (! (0 <= XINT (subexp
) && XINT (subexp
) < search_regs
.num_regs
))
2298 args_out_of_range (subexp
, make_number (search_regs
.num_regs
));
2299 sub
= XINT (subexp
);
2304 if (search_regs
.start
[sub
] < BEGV
2305 || search_regs
.start
[sub
] > search_regs
.end
[sub
]
2306 || search_regs
.end
[sub
] > ZV
)
2307 args_out_of_range (make_number (search_regs
.start
[sub
]),
2308 make_number (search_regs
.end
[sub
]));
2312 if (search_regs
.start
[sub
] < 0
2313 || search_regs
.start
[sub
] > search_regs
.end
[sub
]
2314 || search_regs
.end
[sub
] > SCHARS (string
))
2315 args_out_of_range (make_number (search_regs
.start
[sub
]),
2316 make_number (search_regs
.end
[sub
]));
2319 if (NILP (fixedcase
))
2321 /* Decide how to casify by examining the matched text. */
2324 pos
= search_regs
.start
[sub
];
2325 last
= search_regs
.end
[sub
];
2328 pos_byte
= CHAR_TO_BYTE (pos
);
2330 pos_byte
= string_char_to_byte (string
, pos
);
2333 case_action
= all_caps
;
2335 /* some_multiletter_word is set nonzero if any original word
2336 is more than one letter long. */
2337 some_multiletter_word
= 0;
2339 some_nonuppercase_initial
= 0;
2346 c
= FETCH_CHAR_AS_MULTIBYTE (pos_byte
);
2347 INC_BOTH (pos
, pos_byte
);
2350 FETCH_STRING_CHAR_AS_MULTIBYTE_ADVANCE (c
, string
, pos
, pos_byte
);
2354 /* Cannot be all caps if any original char is lower case */
2357 if (SYNTAX (prevc
) != Sword
)
2358 some_nonuppercase_initial
= 1;
2360 some_multiletter_word
= 1;
2362 else if (uppercasep (c
))
2365 if (SYNTAX (prevc
) != Sword
)
2368 some_multiletter_word
= 1;
2372 /* If the initial is a caseless word constituent,
2373 treat that like a lowercase initial. */
2374 if (SYNTAX (prevc
) != Sword
)
2375 some_nonuppercase_initial
= 1;
2381 /* Convert to all caps if the old text is all caps
2382 and has at least one multiletter word. */
2383 if (! some_lowercase
&& some_multiletter_word
)
2384 case_action
= all_caps
;
2385 /* Capitalize each word, if the old text has all capitalized words. */
2386 else if (!some_nonuppercase_initial
&& some_multiletter_word
)
2387 case_action
= cap_initial
;
2388 else if (!some_nonuppercase_initial
&& some_uppercase
)
2389 /* Should x -> yz, operating on X, give Yz or YZ?
2390 We'll assume the latter. */
2391 case_action
= all_caps
;
2393 case_action
= nochange
;
2396 /* Do replacement in a string. */
2399 Lisp_Object before
, after
;
2401 before
= Fsubstring (string
, make_number (0),
2402 make_number (search_regs
.start
[sub
]));
2403 after
= Fsubstring (string
, make_number (search_regs
.end
[sub
]), Qnil
);
2405 /* Substitute parts of the match into NEWTEXT
2409 ptrdiff_t lastpos
= 0;
2410 ptrdiff_t lastpos_byte
= 0;
2411 /* We build up the substituted string in ACCUM. */
2414 ptrdiff_t length
= SBYTES (newtext
);
2418 for (pos_byte
= 0, pos
= 0; pos_byte
< length
;)
2420 ptrdiff_t substart
= -1;
2421 ptrdiff_t subend
= 0;
2422 int delbackslash
= 0;
2424 FETCH_STRING_CHAR_ADVANCE (c
, newtext
, pos
, pos_byte
);
2428 FETCH_STRING_CHAR_ADVANCE (c
, newtext
, pos
, pos_byte
);
2432 substart
= search_regs
.start
[sub
];
2433 subend
= search_regs
.end
[sub
];
2435 else if (c
>= '1' && c
<= '9')
2437 if (c
- '0' < search_regs
.num_regs
2438 && 0 <= search_regs
.start
[c
- '0'])
2440 substart
= search_regs
.start
[c
- '0'];
2441 subend
= search_regs
.end
[c
- '0'];
2445 /* If that subexp did not match,
2446 replace \\N with nothing. */
2454 error ("Invalid use of `\\' in replacement text");
2458 if (pos
- 2 != lastpos
)
2459 middle
= substring_both (newtext
, lastpos
,
2461 pos
- 2, pos_byte
- 2);
2464 accum
= concat3 (accum
, middle
,
2466 make_number (substart
),
2467 make_number (subend
)));
2469 lastpos_byte
= pos_byte
;
2471 else if (delbackslash
)
2473 middle
= substring_both (newtext
, lastpos
,
2475 pos
- 1, pos_byte
- 1);
2477 accum
= concat2 (accum
, middle
);
2479 lastpos_byte
= pos_byte
;
2484 middle
= substring_both (newtext
, lastpos
,
2490 newtext
= concat2 (accum
, middle
);
2493 /* Do case substitution in NEWTEXT if desired. */
2494 if (case_action
== all_caps
)
2495 newtext
= Fupcase (newtext
);
2496 else if (case_action
== cap_initial
)
2497 newtext
= Fupcase_initials (newtext
);
2499 return concat3 (before
, newtext
, after
);
2502 /* Record point, then move (quietly) to the start of the match. */
2503 if (PT
>= search_regs
.end
[sub
])
2505 else if (PT
> search_regs
.start
[sub
])
2506 opoint
= search_regs
.end
[sub
] - ZV
;
2510 /* If we want non-literal replacement,
2511 perform substitution on the replacement string. */
2514 ptrdiff_t length
= SBYTES (newtext
);
2515 unsigned char *substed
;
2516 ptrdiff_t substed_alloc_size
, substed_len
;
2517 int buf_multibyte
= !NILP (BVAR (current_buffer
, enable_multibyte_characters
));
2518 int str_multibyte
= STRING_MULTIBYTE (newtext
);
2519 int really_changed
= 0;
2521 substed_alloc_size
= ((STRING_BYTES_BOUND
- 100) / 2 < length
2522 ? STRING_BYTES_BOUND
2523 : length
* 2 + 100);
2524 substed
= xmalloc (substed_alloc_size
);
2527 /* Go thru NEWTEXT, producing the actual text to insert in
2528 SUBSTED while adjusting multibyteness to that of the current
2531 for (pos_byte
= 0, pos
= 0; pos_byte
< length
;)
2533 unsigned char str
[MAX_MULTIBYTE_LENGTH
];
2534 const unsigned char *add_stuff
= NULL
;
2535 ptrdiff_t add_len
= 0;
2540 FETCH_STRING_CHAR_ADVANCE_NO_CHECK (c
, newtext
, pos
, pos_byte
);
2542 c
= multibyte_char_to_unibyte (c
);
2546 /* Note that we don't have to increment POS. */
2547 c
= SREF (newtext
, pos_byte
++);
2549 MAKE_CHAR_MULTIBYTE (c
);
2552 /* Either set ADD_STUFF and ADD_LEN to the text to put in SUBSTED,
2553 or set IDX to a match index, which means put that part
2554 of the buffer text into SUBSTED. */
2562 FETCH_STRING_CHAR_ADVANCE_NO_CHECK (c
, newtext
,
2564 if (!buf_multibyte
&& !ASCII_CHAR_P (c
))
2565 c
= multibyte_char_to_unibyte (c
);
2569 c
= SREF (newtext
, pos_byte
++);
2571 MAKE_CHAR_MULTIBYTE (c
);
2576 else if (c
>= '1' && c
<= '9' && c
- '0' < search_regs
.num_regs
)
2578 if (search_regs
.start
[c
- '0'] >= 1)
2582 add_len
= 1, add_stuff
= (unsigned char *) "\\";
2586 error ("Invalid use of `\\' in replacement text");
2591 add_len
= CHAR_STRING (c
, str
);
2595 /* If we want to copy part of a previous match,
2596 set up ADD_STUFF and ADD_LEN to point to it. */
2599 ptrdiff_t begbyte
= CHAR_TO_BYTE (search_regs
.start
[idx
]);
2600 add_len
= CHAR_TO_BYTE (search_regs
.end
[idx
]) - begbyte
;
2601 if (search_regs
.start
[idx
] < GPT
&& GPT
< search_regs
.end
[idx
])
2602 move_gap (search_regs
.start
[idx
]);
2603 add_stuff
= BYTE_POS_ADDR (begbyte
);
2606 /* Now the stuff we want to add to SUBSTED
2607 is invariably ADD_LEN bytes starting at ADD_STUFF. */
2609 /* Make sure SUBSTED is big enough. */
2610 if (substed_alloc_size
- substed_len
< add_len
)
2612 xpalloc (substed
, &substed_alloc_size
,
2613 add_len
- (substed_alloc_size
- substed_len
),
2614 STRING_BYTES_BOUND
, 1);
2616 /* Now add to the end of SUBSTED. */
2619 memcpy (substed
+ substed_len
, add_stuff
, add_len
);
2620 substed_len
+= add_len
;
2629 multibyte_chars_in_text (substed
, substed_len
);
2631 newtext
= make_multibyte_string ((char *) substed
, nchars
,
2635 newtext
= make_unibyte_string ((char *) substed
, substed_len
);
2640 /* Replace the old text with the new in the cleanest possible way. */
2641 replace_range (search_regs
.start
[sub
], search_regs
.end
[sub
],
2643 newpoint
= search_regs
.start
[sub
] + SCHARS (newtext
);
2645 if (case_action
== all_caps
)
2646 Fupcase_region (make_number (search_regs
.start
[sub
]),
2647 make_number (newpoint
));
2648 else if (case_action
== cap_initial
)
2649 Fupcase_initials_region (make_number (search_regs
.start
[sub
]),
2650 make_number (newpoint
));
2652 /* Adjust search data for this change. */
2654 ptrdiff_t oldend
= search_regs
.end
[sub
];
2655 ptrdiff_t oldstart
= search_regs
.start
[sub
];
2656 ptrdiff_t change
= newpoint
- search_regs
.end
[sub
];
2659 for (i
= 0; i
< search_regs
.num_regs
; i
++)
2661 if (search_regs
.start
[i
] >= oldend
)
2662 search_regs
.start
[i
] += change
;
2663 else if (search_regs
.start
[i
] > oldstart
)
2664 search_regs
.start
[i
] = oldstart
;
2665 if (search_regs
.end
[i
] >= oldend
)
2666 search_regs
.end
[i
] += change
;
2667 else if (search_regs
.end
[i
] > oldstart
)
2668 search_regs
.end
[i
] = oldstart
;
2672 /* Put point back where it was in the text. */
2674 TEMP_SET_PT (opoint
+ ZV
);
2676 TEMP_SET_PT (opoint
);
2678 /* Now move point "officially" to the start of the inserted replacement. */
2679 move_if_not_intangible (newpoint
);
2685 match_limit (Lisp_Object num
, int beginningp
)
2692 args_out_of_range (num
, make_number (0));
2693 if (search_regs
.num_regs
<= 0)
2694 error ("No match data, because no search succeeded");
2695 if (n
>= search_regs
.num_regs
2696 || search_regs
.start
[n
] < 0)
2698 return (make_number ((beginningp
) ? search_regs
.start
[n
]
2699 : search_regs
.end
[n
]));
2702 DEFUN ("match-beginning", Fmatch_beginning
, Smatch_beginning
, 1, 1, 0,
2703 doc
: /* Return position of start of text matched by last search.
2704 SUBEXP, a number, specifies which parenthesized expression in the last
2706 Value is nil if SUBEXPth pair didn't match, or there were less than
2708 Zero means the entire text matched by the whole regexp or whole string. */)
2709 (Lisp_Object subexp
)
2711 return match_limit (subexp
, 1);
2714 DEFUN ("match-end", Fmatch_end
, Smatch_end
, 1, 1, 0,
2715 doc
: /* Return position of end of text matched by last search.
2716 SUBEXP, a number, specifies which parenthesized expression in the last
2718 Value is nil if SUBEXPth pair didn't match, or there were less than
2720 Zero means the entire text matched by the whole regexp or whole string. */)
2721 (Lisp_Object subexp
)
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 `set-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 (Lisp_Object integers
, Lisp_Object reuse
, Lisp_Object reseat
)
2748 Lisp_Object tail
, prev
;
2753 for (tail
= reuse
; CONSP (tail
); tail
= XCDR (tail
))
2754 if (MARKERP (XCAR (tail
)))
2756 unchain_marker (XMARKER (XCAR (tail
)));
2757 XSETCAR (tail
, Qnil
);
2760 if (NILP (last_thing_searched
))
2765 data
= alloca ((2 * search_regs
.num_regs
+ 1) * sizeof *data
);
2768 for (i
= 0; i
< search_regs
.num_regs
; i
++)
2770 ptrdiff_t start
= search_regs
.start
[i
];
2773 if (EQ (last_thing_searched
, Qt
)
2774 || ! NILP (integers
))
2776 XSETFASTINT (data
[2 * i
], start
);
2777 XSETFASTINT (data
[2 * i
+ 1], search_regs
.end
[i
]);
2779 else if (BUFFERP (last_thing_searched
))
2781 data
[2 * i
] = Fmake_marker ();
2782 Fset_marker (data
[2 * i
],
2783 make_number (start
),
2784 last_thing_searched
);
2785 data
[2 * i
+ 1] = Fmake_marker ();
2786 Fset_marker (data
[2 * i
+ 1],
2787 make_number (search_regs
.end
[i
]),
2788 last_thing_searched
);
2791 /* last_thing_searched must always be Qt, a buffer, or Qnil. */
2797 data
[2 * i
] = data
[2 * i
+ 1] = Qnil
;
2800 if (BUFFERP (last_thing_searched
) && !NILP (integers
))
2802 data
[len
] = last_thing_searched
;
2806 /* If REUSE is not usable, cons up the values and return them. */
2807 if (! CONSP (reuse
))
2808 return Flist (len
, data
);
2810 /* If REUSE is a list, store as many value elements as will fit
2811 into the elements of REUSE. */
2812 for (i
= 0, tail
= reuse
; CONSP (tail
);
2813 i
++, tail
= XCDR (tail
))
2816 XSETCAR (tail
, data
[i
]);
2818 XSETCAR (tail
, Qnil
);
2822 /* If we couldn't fit all value elements into REUSE,
2823 cons up the rest of them and add them to the end of REUSE. */
2825 XSETCDR (prev
, Flist (len
- i
, data
+ i
));
2830 /* We used to have an internal use variant of `reseat' described as:
2832 If RESEAT is `evaporate', put the markers back on the free list
2833 immediately. No other references to the markers must exist in this
2834 case, so it is used only internally on the unwind stack and
2835 save-match-data from Lisp.
2837 But it was ill-conceived: those supposedly-internal markers get exposed via
2838 the undo-list, so freeing them here is unsafe. */
2840 DEFUN ("set-match-data", Fset_match_data
, Sset_match_data
, 1, 2, 0,
2841 doc
: /* Set internal data on last search match from elements of LIST.
2842 LIST should have been created by calling `match-data' previously.
2844 If optional arg RESEAT is non-nil, make markers on LIST point nowhere. */)
2845 (register Lisp_Object list
, Lisp_Object reseat
)
2848 register Lisp_Object marker
;
2850 if (running_asynch_code
)
2851 save_search_regs ();
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 EMACS_INT length
= XFASTINT (Flength (list
)) / 2;
2863 if (length
> search_regs
.num_regs
)
2865 ptrdiff_t num_regs
= search_regs
.num_regs
;
2866 if (PTRDIFF_MAX
< length
)
2867 memory_full (SIZE_MAX
);
2869 xpalloc (search_regs
.start
, &num_regs
, length
- num_regs
,
2870 min (PTRDIFF_MAX
, UINT_MAX
), sizeof (regoff_t
));
2872 xrealloc (search_regs
.end
, num_regs
* sizeof (regoff_t
));
2874 for (i
= search_regs
.num_regs
; i
< num_regs
; i
++)
2875 search_regs
.start
[i
] = -1;
2877 search_regs
.num_regs
= num_regs
;
2880 for (i
= 0; CONSP (list
); i
++)
2882 marker
= XCAR (list
);
2883 if (BUFFERP (marker
))
2885 last_thing_searched
= marker
;
2892 search_regs
.start
[i
] = -1;
2901 if (MARKERP (marker
))
2903 if (XMARKER (marker
)->buffer
== 0)
2904 XSETFASTINT (marker
, 0);
2906 XSETBUFFER (last_thing_searched
, XMARKER (marker
)->buffer
);
2909 CHECK_NUMBER_COERCE_MARKER (marker
);
2912 if (!NILP (reseat
) && MARKERP (m
))
2914 unchain_marker (XMARKER (m
));
2915 XSETCAR (list
, Qnil
);
2918 if ((list
= XCDR (list
), !CONSP (list
)))
2921 m
= marker
= XCAR (list
);
2923 if (MARKERP (marker
) && XMARKER (marker
)->buffer
== 0)
2924 XSETFASTINT (marker
, 0);
2926 CHECK_NUMBER_COERCE_MARKER (marker
);
2927 if ((XINT (from
) < 0
2928 ? TYPE_MINIMUM (regoff_t
) <= XINT (from
)
2929 : XINT (from
) <= TYPE_MAXIMUM (regoff_t
))
2930 && (XINT (marker
) < 0
2931 ? TYPE_MINIMUM (regoff_t
) <= XINT (marker
)
2932 : XINT (marker
) <= TYPE_MAXIMUM (regoff_t
)))
2934 search_regs
.start
[i
] = XINT (from
);
2935 search_regs
.end
[i
] = XINT (marker
);
2939 search_regs
.start
[i
] = -1;
2942 if (!NILP (reseat
) && MARKERP (m
))
2944 unchain_marker (XMARKER (m
));
2945 XSETCAR (list
, Qnil
);
2951 for (; i
< search_regs
.num_regs
; i
++)
2952 search_regs
.start
[i
] = -1;
2958 /* If non-zero the match data have been saved in saved_search_regs
2959 during the execution of a sentinel or filter. */
2960 static int search_regs_saved
;
2961 static struct re_registers saved_search_regs
;
2962 static Lisp_Object saved_last_thing_searched
;
2964 /* Called from Flooking_at, Fstring_match, search_buffer, Fstore_match_data
2965 if asynchronous code (filter or sentinel) is running. */
2967 save_search_regs (void)
2969 if (!search_regs_saved
)
2971 saved_search_regs
.num_regs
= search_regs
.num_regs
;
2972 saved_search_regs
.start
= search_regs
.start
;
2973 saved_search_regs
.end
= search_regs
.end
;
2974 saved_last_thing_searched
= last_thing_searched
;
2975 last_thing_searched
= Qnil
;
2976 search_regs
.num_regs
= 0;
2977 search_regs
.start
= 0;
2978 search_regs
.end
= 0;
2980 search_regs_saved
= 1;
2984 /* Called upon exit from filters and sentinels. */
2986 restore_search_regs (void)
2988 if (search_regs_saved
)
2990 if (search_regs
.num_regs
> 0)
2992 xfree (search_regs
.start
);
2993 xfree (search_regs
.end
);
2995 search_regs
.num_regs
= saved_search_regs
.num_regs
;
2996 search_regs
.start
= saved_search_regs
.start
;
2997 search_regs
.end
= saved_search_regs
.end
;
2998 last_thing_searched
= saved_last_thing_searched
;
2999 saved_last_thing_searched
= Qnil
;
3000 search_regs_saved
= 0;
3005 unwind_set_match_data (Lisp_Object list
)
3007 /* It is NOT ALWAYS safe to free (evaporate) the markers immediately. */
3008 return Fset_match_data (list
, Qt
);
3011 /* Called to unwind protect the match data. */
3013 record_unwind_save_match_data (void)
3015 record_unwind_protect (unwind_set_match_data
,
3016 Fmatch_data (Qnil
, Qnil
, Qnil
));
3019 /* Quote a string to deactivate reg-expr chars */
3021 DEFUN ("regexp-quote", Fregexp_quote
, Sregexp_quote
, 1, 1, 0,
3022 doc
: /* Return a regexp string which matches exactly STRING and nothing else. */)
3023 (Lisp_Object string
)
3025 register char *in
, *out
, *end
;
3026 register char *temp
;
3027 int backslashes_added
= 0;
3029 CHECK_STRING (string
);
3031 temp
= alloca (SBYTES (string
) * 2);
3033 /* Now copy the data into the new string, inserting escapes. */
3035 in
= SSDATA (string
);
3036 end
= in
+ SBYTES (string
);
3039 for (; in
!= end
; in
++)
3042 || *in
== '*' || *in
== '.' || *in
== '\\'
3043 || *in
== '?' || *in
== '+'
3044 || *in
== '^' || *in
== '$')
3045 *out
++ = '\\', backslashes_added
++;
3049 return make_specified_string (temp
,
3050 SCHARS (string
) + backslashes_added
,
3052 STRING_MULTIBYTE (string
));
3056 syms_of_search (void)
3060 for (i
= 0; i
< REGEXP_CACHE_SIZE
; ++i
)
3062 searchbufs
[i
].buf
.allocated
= 100;
3063 searchbufs
[i
].buf
.buffer
= xmalloc (100);
3064 searchbufs
[i
].buf
.fastmap
= searchbufs
[i
].fastmap
;
3065 searchbufs
[i
].regexp
= Qnil
;
3066 searchbufs
[i
].whitespace_regexp
= Qnil
;
3067 searchbufs
[i
].syntax_table
= Qnil
;
3068 staticpro (&searchbufs
[i
].regexp
);
3069 staticpro (&searchbufs
[i
].whitespace_regexp
);
3070 staticpro (&searchbufs
[i
].syntax_table
);
3071 searchbufs
[i
].next
= (i
== REGEXP_CACHE_SIZE
-1 ? 0 : &searchbufs
[i
+1]);
3073 searchbuf_head
= &searchbufs
[0];
3075 DEFSYM (Qsearch_failed
, "search-failed");
3076 DEFSYM (Qinvalid_regexp
, "invalid-regexp");
3078 Fput (Qsearch_failed
, Qerror_conditions
,
3079 listn (CONSTYPE_PURE
, 2, Qsearch_failed
, Qerror
));
3080 Fput (Qsearch_failed
, Qerror_message
,
3081 build_pure_c_string ("Search failed"));
3083 Fput (Qinvalid_regexp
, Qerror_conditions
,
3084 listn (CONSTYPE_PURE
, 2, Qinvalid_regexp
, Qerror
));
3085 Fput (Qinvalid_regexp
, Qerror_message
,
3086 build_pure_c_string ("Invalid regexp"));
3088 last_thing_searched
= Qnil
;
3089 staticpro (&last_thing_searched
);
3091 saved_last_thing_searched
= Qnil
;
3092 staticpro (&saved_last_thing_searched
);
3094 DEFVAR_LISP ("search-spaces-regexp", Vsearch_spaces_regexp
,
3095 doc
: /* Regexp to substitute for bunches of spaces in regexp search.
3096 Some commands use this for user-specified regexps.
3097 Spaces that occur inside character classes or repetition operators
3098 or other such regexp constructs are not replaced with this.
3099 A value of nil (which is the normal value) means treat spaces literally. */);
3100 Vsearch_spaces_regexp
= Qnil
;
3102 DEFVAR_LISP ("inhibit-changing-match-data", Vinhibit_changing_match_data
,
3103 doc
: /* Internal use only.
3104 If non-nil, the primitive searching and matching functions
3105 such as `looking-at', `string-match', `re-search-forward', etc.,
3106 do not set the match data. The proper way to use this variable
3107 is to bind it with `let' around a small expression. */);
3108 Vinhibit_changing_match_data
= Qnil
;
3110 defsubr (&Slooking_at
);
3111 defsubr (&Sposix_looking_at
);
3112 defsubr (&Sstring_match
);
3113 defsubr (&Sposix_string_match
);
3114 defsubr (&Ssearch_forward
);
3115 defsubr (&Ssearch_backward
);
3116 defsubr (&Sre_search_forward
);
3117 defsubr (&Sre_search_backward
);
3118 defsubr (&Sposix_search_forward
);
3119 defsubr (&Sposix_search_backward
);
3120 defsubr (&Sreplace_match
);
3121 defsubr (&Smatch_beginning
);
3122 defsubr (&Smatch_end
);
3123 defsubr (&Smatch_data
);
3124 defsubr (&Sset_match_data
);
3125 defsubr (&Sregexp_quote
);