search.c 90.8 KB
Newer Older
Jim Blandy's avatar
Jim Blandy committed
1
/* String search routines for GNU Emacs.
2

3 4
Copyright (C) 1985-1987, 1993-1994, 1997-1999, 2001-2013 Free Software
Foundation, Inc.
Jim Blandy's avatar
Jim Blandy committed
5 6 7

This file is part of GNU Emacs.

8
GNU Emacs is free software: you can redistribute it and/or modify
Jim Blandy's avatar
Jim Blandy committed
9
it under the terms of the GNU General Public License as published by
10 11
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
Jim Blandy's avatar
Jim Blandy committed
12 13 14 15 16 17 18

GNU Emacs is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

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/>.  */
Jim Blandy's avatar
Jim Blandy committed
20 21


22
#include <config.h>
23

Jim Blandy's avatar
Jim Blandy committed
24
#include "lisp.h"
Karl Heuer's avatar
Karl Heuer committed
25
#include "category.h"
26
#include "character.h"
27
#include "buffer.h"
28
#include "syntax.h"
Kenichi Handa's avatar
Kenichi Handa committed
29
#include "charset.h"
30
#include "region-cache.h"
Jim Blandy's avatar
Jim Blandy committed
31
#include "commands.h"
32
#include "blockinput.h"
33
#include "intervals.h"
Jim Blandy's avatar
Jim Blandy committed
34

Jim Blandy's avatar
Jim Blandy committed
35 36 37
#include <sys/types.h>
#include "regex.h"

38
#define REGEXP_CACHE_SIZE 20
Jim Blandy's avatar
Jim Blandy committed
39

40 41
/* If the regexp is non-nil, then the buffer contains the compiled form
   of that regexp, suitable for searching.  */
42 43
struct regexp_cache
{
44
  struct regexp_cache *next;
45
  Lisp_Object regexp, whitespace_regexp;
46
  /* Syntax table for which the regexp applies.  We need this because
47 48
     of character classes.  If this is t, then the compiled pattern is valid
     for any syntax-table.  */
49
  Lisp_Object syntax_table;
50 51
  struct re_pattern_buffer buf;
  char fastmap[0400];
52 53
  /* True means regexp was compiled to do full POSIX backtracking.  */
  bool posix;
54
};
Jim Blandy's avatar
Jim Blandy committed
55

56
/* The instances of that struct.  */
57
static struct regexp_cache searchbufs[REGEXP_CACHE_SIZE];
Jim Blandy's avatar
Jim Blandy committed
58

59
/* The head of the linked list; points to the most recently used buffer.  */
60
static struct regexp_cache *searchbuf_head;
Jim Blandy's avatar
Jim Blandy committed
61 62


Jim Blandy's avatar
Jim Blandy committed
63 64 65 66 67 68 69
/* Every call to re_match, etc., must pass &search_regs as the regs
   argument unless you can show it is unnecessary (i.e., if re_match
   is certainly going to be called again before region-around-match
   can be called).

   Since the registers are now dynamically allocated, we need to make
   sure not to refer to the Nth register before checking that it has
Jim Blandy's avatar
Jim Blandy committed
70 71 72
   been allocated by checking search_regs.num_regs.

   The regex code keeps track of whether it has allocated the search
73 74
   buffer using bits in the re_pattern_buffer.  This means that whenever
   you compile a new pattern, it completely forgets whether it has
Jim Blandy's avatar
Jim Blandy committed
75 76 77 78 79
   allocated any registers, and will allocate new registers the next
   time you call a searching or matching function.  Therefore, we need
   to call re_set_registers after compiling a new pattern or after
   setting the match registers, so that the regex functions will be
   able to free or re-allocate it properly.  */
Jim Blandy's avatar
Jim Blandy committed
80 81
static struct re_registers search_regs;

Jim Blandy's avatar
Jim Blandy committed
82 83 84 85
/* The buffer in which the last search was performed, or
   Qt if the last search was done in a string;
   Qnil if no searching has been done yet.  */
static Lisp_Object last_thing_searched;
Jim Blandy's avatar
Jim Blandy committed
86

87
/* Error condition signaled when regexp compile_pattern fails.  */
88
static Lisp_Object Qinvalid_regexp;
Jim Blandy's avatar
Jim Blandy committed
89

90
/* Error condition used for failing searches.  */
91
static Lisp_Object Qsearch_failed;
92

93
static void set_search_regs (ptrdiff_t, ptrdiff_t);
94
static void save_search_regs (void);
95 96 97 98 99 100 101 102
static EMACS_INT simple_search (EMACS_INT, unsigned char *, ptrdiff_t,
				ptrdiff_t, Lisp_Object, ptrdiff_t, ptrdiff_t,
                                ptrdiff_t, ptrdiff_t);
static EMACS_INT boyer_moore (EMACS_INT, unsigned char *, ptrdiff_t,
                              Lisp_Object, Lisp_Object, ptrdiff_t,
                              ptrdiff_t, int);
static EMACS_INT search_buffer (Lisp_Object, ptrdiff_t, ptrdiff_t,
                                ptrdiff_t, ptrdiff_t, EMACS_INT, int,
103
                                Lisp_Object, Lisp_Object, bool);
104

105
static _Noreturn void
106
matcher_overflow (void)
Jim Blandy's avatar
Jim Blandy committed
107 108 109 110
{
  error ("Stack overflow in regexp matcher");
}

111 112 113
/* Compile a regexp and signal a Lisp error if anything goes wrong.
   PATTERN is the pattern to compile.
   CP is the place to put the result.
114
   TRANSLATE is a translation table for ignoring case, or nil for none.
115 116
   POSIX is true if we want full backtracking (POSIX style) for this pattern.
   False means backtrack only enough to get a valid match.
117 118

   The behavior also depends on Vsearch_spaces_regexp.  */
Jim Blandy's avatar
Jim Blandy committed
119

120
static void
121 122
compile_pattern_1 (struct regexp_cache *cp, Lisp_Object pattern,
		   Lisp_Object translate, bool posix)
Jim Blandy's avatar
Jim Blandy committed
123
{
124
  char *val;
125
  reg_syntax_t old;
Jim Blandy's avatar
Jim Blandy committed
126

127
  cp->regexp = Qnil;
128
  cp->buf.translate = (! NILP (translate) ? translate : make_number (0));
129
  cp->posix = posix;
130
  cp->buf.multibyte = STRING_MULTIBYTE (pattern);
Kenichi Handa's avatar
Kenichi Handa committed
131
  cp->buf.charset_unibyte = charset_unibyte;
132 133 134 135 136
  if (STRINGP (Vsearch_spaces_regexp))
    cp->whitespace_regexp = Vsearch_spaces_regexp;
  else
    cp->whitespace_regexp = Qnil;

137 138 139 140 141
  /* rms: I think BLOCK_INPUT is not needed here any more,
     because regex.c defines malloc to call xmalloc.
     Using BLOCK_INPUT here means the debugger won't run if an error occurs.
     So let's turn it off.  */
  /*  BLOCK_INPUT;  */
142
  old = re_set_syntax (RE_SYNTAX_EMACS
143
		       | (posix ? 0 : RE_NO_POSIX_BACKTRACKING));
144 145

  if (STRINGP (Vsearch_spaces_regexp))
146
    re_set_whitespace_regexp (SSDATA (Vsearch_spaces_regexp));
147 148
  else
    re_set_whitespace_regexp (NULL);
149

150
  val = (char *) re_compile_pattern (SSDATA (pattern),
Kenichi Handa's avatar
Kenichi Handa committed
151
				     SBYTES (pattern), &cp->buf);
152

153 154
  /* If the compiled pattern hard codes some of the contents of the
     syntax-table, it can only be reused with *this* syntax table.  */
Tom Tromey's avatar
Tom Tromey committed
155
  cp->syntax_table = cp->buf.used_syntax ? BVAR (current_buffer, syntax_table) : Qt;
156

157 158
  re_set_whitespace_regexp (NULL);

159
  re_set_syntax (old);
160
  /* unblock_input ();  */
Jim Blandy's avatar
Jim Blandy committed
161
  if (val)
162
    xsignal1 (Qinvalid_regexp, build_string (val));
Jim Blandy's avatar
Jim Blandy committed
163

164 165 166
  cp->regexp = Fcopy_sequence (pattern);
}

167 168 169 170 171
/* Shrink each compiled regexp buffer in the cache
   to the size actually used right now.
   This is called from garbage collection.  */

void
172
shrink_regexp_cache (void)
173
{
174
  struct regexp_cache *cp;
175 176 177 178

  for (cp = searchbuf_head; cp != 0; cp = cp->next)
    {
      cp->buf.allocated = cp->buf.used;
179
      cp->buf.buffer = xrealloc (cp->buf.buffer, cp->buf.used);
180 181 182
    }
}

183 184
/* Clear the regexp cache w.r.t. a particular syntax table,
   because it was changed.
185 186 187 188
   There is no danger of memory leak here because re_compile_pattern
   automagically manages the memory in each re_pattern_buffer struct,
   based on its `allocated' and `buffer' values.  */
void
189
clear_regexp_cache (void)
190 191 192 193
{
  int i;

  for (i = 0; i < REGEXP_CACHE_SIZE; ++i)
194 195
    /* It's tempting to compare with the syntax-table we've actually changed,
       but it's not sufficient because char-table inheritance means that
196 197 198
       modifying one syntax-table can change others at the same time.  */
    if (!EQ (searchbufs[i].syntax_table, Qt))
      searchbufs[i].regexp = Qnil;
199 200
}

201
/* Compile a regexp if necessary, but first check to see if there's one in
202 203
   the cache.
   PATTERN is the pattern to compile.
204
   TRANSLATE is a translation table for ignoring case, or nil for none.
205 206 207 208
   REGP is the structure that says where to store the "register"
   values that will result from matching this pattern.
   If it is 0, we should compile the pattern not to record any
   subexpression bounds.
209 210
   POSIX is true if we want full backtracking (POSIX style) for this pattern.
   False means backtrack only enough to get a valid match.  */
211 212

struct re_pattern_buffer *
213
compile_pattern (Lisp_Object pattern, struct re_registers *regp,
214
		 Lisp_Object translate, bool posix, bool multibyte)
215 216 217 218 219 220
{
  struct regexp_cache *cp, **cpp;

  for (cpp = &searchbuf_head; ; cpp = &cp->next)
    {
      cp = *cpp;
221 222
      /* Entries are initialized to nil, and may be set to nil by
	 compile_pattern_1 if the pattern isn't valid.  Don't apply
Ken Raeburn's avatar
Ken Raeburn committed
223 224 225
	 string accessors in those cases.  However, compile_pattern_1
	 is only applied to the cache entry we pick here to reuse.  So
	 nil should never appear before a non-nil entry.  */
226
      if (NILP (cp->regexp))
227
	goto compile_it;
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)))
Karl Heuer's avatar
Karl Heuer committed
232
	  && cp->posix == posix
233
	  && (EQ (cp->syntax_table, Qt)
Tom Tromey's avatar
Tom Tromey committed
234
	      || EQ (cp->syntax_table, BVAR (current_buffer, syntax_table)))
Kenichi Handa's avatar
Kenichi Handa committed
235 236
	  && !NILP (Fequal (cp->whitespace_regexp, Vsearch_spaces_regexp))
	  && cp->buf.charset_unibyte == charset_unibyte)
237 238
	break;

239 240 241
      /* If we're at the end of the cache, compile into the nil cell
	 we found, or the last (least recently used) cell with a
	 string value.  */
242 243
      if (cp->next == 0)
	{
244
	compile_it:
245
	  compile_pattern_1 (cp, pattern, translate, posix);
246 247 248 249 250 251 252 253 254 255
	  break;
	}
    }

  /* When we get here, cp (aka *cpp) contains the compiled pattern,
     either because we found it in the cache or because we just compiled it.
     Move it to the front of the queue to mark it as most recently used.  */
  *cpp = cp->next;
  cp->next = searchbuf_head;
  searchbuf_head = cp;
Jim Blandy's avatar
Jim Blandy committed
256

257 258 259 260 261
  /* Advise the searching functions about the space we have allocated
     for register data.  */
  if (regp)
    re_set_registers (&cp->buf, regp, regp->num_regs, regp->start, regp->end);

Juanma Barranquero's avatar
Juanma Barranquero committed
262
  /* The compiled pattern can be used both for multibyte and unibyte
Kenichi Handa's avatar
Kenichi Handa committed
263 264 265
     target.  But, we have to tell which the pattern is used for. */
  cp->buf.target_multibyte = multibyte;

266
  return &cp->buf;
Jim Blandy's avatar
Jim Blandy committed
267 268 269
}


270
static Lisp_Object
271
looking_at_1 (Lisp_Object string, bool posix)
Jim Blandy's avatar
Jim Blandy committed
272 273 274
{
  Lisp_Object val;
  unsigned char *p1, *p2;
275 276
  ptrdiff_t s1, s2;
  register ptrdiff_t i;
277
  struct re_pattern_buffer *bufp;
Jim Blandy's avatar
Jim Blandy committed
278

279 280 281
  if (running_asynch_code)
    save_search_regs ();

282
  /* This is so set_image_of_range_1 in regex.c can find the EQV table.  */
283
  set_char_table_extras (BVAR (current_buffer, case_canon_table), 2,
284
			 BVAR (current_buffer, case_eqv_table));
285

286
  CHECK_STRING (string);
287 288 289
  bufp = compile_pattern (string,
			  (NILP (Vinhibit_changing_match_data)
			   ? &search_regs : NULL),
Tom Tromey's avatar
Tom Tromey committed
290 291
			  (!NILP (BVAR (current_buffer, case_fold_search))
			   ? BVAR (current_buffer, case_canon_table) : Qnil),
292
			  posix,
Tom Tromey's avatar
Tom Tromey committed
293
			  !NILP (BVAR (current_buffer, enable_multibyte_characters)));
Jim Blandy's avatar
Jim Blandy committed
294 295 296 297 298 299 300 301

  immediate_quit = 1;
  QUIT;			/* Do a pending quit right away, to avoid paradoxical behavior */

  /* Get pointers and sizes of the two strings
     that make up the visible portion of the buffer. */

  p1 = BEGV_ADDR;
302
  s1 = GPT_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
303
  p2 = GAP_END_ADDR;
304
  s2 = ZV_BYTE - GPT_BYTE;
Jim Blandy's avatar
Jim Blandy committed
305 306 307
  if (s1 < 0)
    {
      p2 = p1;
308
      s2 = ZV_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
309 310 311 312
      s1 = 0;
    }
  if (s2 < 0)
    {
313
      s1 = ZV_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
314 315
      s2 = 0;
    }
316 317

  re_match_object = Qnil;
318

319
  i = re_match_2 (bufp, (char *) p1, s1, (char *) p2, s2,
320 321 322
		  PT_BYTE - BEGV_BYTE,
		  (NILP (Vinhibit_changing_match_data)
		   ? &search_regs : NULL),
323
		  ZV_BYTE - BEGV_BYTE);
324
  immediate_quit = 0;
325

Jim Blandy's avatar
Jim Blandy committed
326 327 328
  if (i == -2)
    matcher_overflow ();

329
  val = (i >= 0 ? Qt : Qnil);
330
  if (NILP (Vinhibit_changing_match_data) && i >= 0)
331
  {
332 333 334 335 336
    for (i = 0; i < search_regs.num_regs; i++)
      if (search_regs.start[i] >= 0)
	{
	  search_regs.start[i]
	    = BYTE_TO_CHAR (search_regs.start[i] + BEGV_BYTE);
337 338 339 340
         search_regs.end[i]
           = BYTE_TO_CHAR (search_regs.end[i] + BEGV_BYTE);
       }
    /* Set last_thing_searched only when match data is changed.  */
341
    XSETBUFFER (last_thing_searched, current_buffer);
342
  }
343

Jim Blandy's avatar
Jim Blandy committed
344 345 346
  return val;
}

347
DEFUN ("looking-at", Flooking_at, Slooking_at, 1, 1, 0,
348 349 350 351
       doc: /* Return t if text after point matches regular expression REGEXP.
This function modifies the match data that `match-beginning',
`match-end' and `match-data' access; save and restore the match
data if you want to preserve them.  */)
Dan Nicolaescu's avatar
Dan Nicolaescu committed
352
  (Lisp_Object regexp)
353
{
354
  return looking_at_1 (regexp, 0);
355 356 357
}

DEFUN ("posix-looking-at", Fposix_looking_at, Sposix_looking_at, 1, 1, 0,
358 359 360 361 362
       doc: /* Return t if text after point matches regular expression REGEXP.
Find the longest match, in accord with Posix regular expression rules.
This function modifies the match data that `match-beginning',
`match-end' and `match-data' access; save and restore the match
data if you want to preserve them.  */)
Dan Nicolaescu's avatar
Dan Nicolaescu committed
363
  (Lisp_Object regexp)
364
{
365
  return looking_at_1 (regexp, 1);
366 367 368
}

static Lisp_Object
369 370
string_match_1 (Lisp_Object regexp, Lisp_Object string, Lisp_Object start,
		bool posix)
Jim Blandy's avatar
Jim Blandy committed
371
{
372
  ptrdiff_t val;
373
  struct re_pattern_buffer *bufp;
374 375
  EMACS_INT pos;
  ptrdiff_t pos_byte, i;
Jim Blandy's avatar
Jim Blandy committed
376

377 378 379
  if (running_asynch_code)
    save_search_regs ();

380 381
  CHECK_STRING (regexp);
  CHECK_STRING (string);
Jim Blandy's avatar
Jim Blandy committed
382 383

  if (NILP (start))
384
    pos = 0, pos_byte = 0;
Jim Blandy's avatar
Jim Blandy committed
385 386
  else
    {
387
      ptrdiff_t len = SCHARS (string);
Jim Blandy's avatar
Jim Blandy committed
388

389
      CHECK_NUMBER (start);
390 391 392 393
      pos = XINT (start);
      if (pos < 0 && -pos <= len)
	pos = len + pos;
      else if (0 > pos || pos > len)
Jim Blandy's avatar
Jim Blandy committed
394
	args_out_of_range (string, start);
395
      pos_byte = string_char_to_byte (string, pos);
Jim Blandy's avatar
Jim Blandy committed
396 397
    }

398
  /* This is so set_image_of_range_1 in regex.c can find the EQV table.  */
399
  set_char_table_extras (BVAR (current_buffer, case_canon_table), 2,
400
			 BVAR (current_buffer, case_eqv_table));
401

402 403 404
  bufp = compile_pattern (regexp,
			  (NILP (Vinhibit_changing_match_data)
			   ? &search_regs : NULL),
Tom Tromey's avatar
Tom Tromey committed
405 406
			  (!NILP (BVAR (current_buffer, case_fold_search))
			   ? BVAR (current_buffer, case_canon_table) : Qnil),
407 408
			  posix,
			  STRING_MULTIBYTE (string));
Jim Blandy's avatar
Jim Blandy committed
409
  immediate_quit = 1;
410
  re_match_object = string;
411

412
  val = re_search (bufp, SSDATA (string),
413 414
		   SBYTES (string), pos_byte,
		   SBYTES (string) - pos_byte,
415 416
		   (NILP (Vinhibit_changing_match_data)
		    ? &search_regs : NULL));
Jim Blandy's avatar
Jim Blandy committed
417
  immediate_quit = 0;
418 419 420 421 422

  /* Set last_thing_searched only when match data is changed.  */
  if (NILP (Vinhibit_changing_match_data))
    last_thing_searched = Qt;

Jim Blandy's avatar
Jim Blandy committed
423 424 425
  if (val == -2)
    matcher_overflow ();
  if (val < 0) return Qnil;
426

427 428 429 430 431 432 433 434 435
  if (NILP (Vinhibit_changing_match_data))
    for (i = 0; i < search_regs.num_regs; i++)
      if (search_regs.start[i] >= 0)
	{
	  search_regs.start[i]
	    = string_byte_to_char (string, search_regs.start[i]);
	  search_regs.end[i]
	    = string_byte_to_char (string, search_regs.end[i]);
	}
436 437

  return make_number (string_byte_to_char (string, val));
Jim Blandy's avatar
Jim Blandy committed
438
}
Richard M. Stallman's avatar
Richard M. Stallman committed
439

Paul Eggert's avatar
Paul Eggert committed
440
DEFUN ("string-match", Fstring_match, Sstring_match, 2, 3, 0,
441
       doc: /* Return index of start of first match for REGEXP in STRING, or nil.
Richard M. Stallman's avatar
Richard M. Stallman committed
442
Matching ignores case if `case-fold-search' is non-nil.
443 444 445
If third arg START is non-nil, start search at that index in STRING.
For index of first char beyond the match, do (match-end 0).
`match-end' and `match-beginning' also give indices of substrings
446 447 448 449
matched by parenthesis constructs in the pattern.

You can use the function `match-string' to extract the substrings
matched by the parenthesis constructions in REGEXP. */)
Dan Nicolaescu's avatar
Dan Nicolaescu committed
450
  (Lisp_Object regexp, Lisp_Object string, Lisp_Object start)
451 452 453 454 455
{
  return string_match_1 (regexp, string, start, 0);
}

DEFUN ("posix-string-match", Fposix_string_match, Sposix_string_match, 2, 3, 0,
456 457 458 459 460 461 462
       doc: /* Return index of start of first match for REGEXP in STRING, or nil.
Find the longest match, in accord with Posix regular expression rules.
Case is ignored if `case-fold-search' is non-nil in the current buffer.
If third arg START is non-nil, start search at that index in STRING.
For index of first char beyond the match, do (match-end 0).
`match-end' and `match-beginning' also give indices of substrings
matched by parenthesis constructs in the pattern.  */)
Dan Nicolaescu's avatar
Dan Nicolaescu committed
463
  (Lisp_Object regexp, Lisp_Object string, Lisp_Object start)
464 465 466 467
{
  return string_match_1 (regexp, string, start, 1);
}

Richard M. Stallman's avatar
Richard M. Stallman committed
468 469 470 471
/* Match REGEXP against STRING, searching all of STRING,
   and return the index of the match, or negative on failure.
   This does not clobber the match data.  */

472
ptrdiff_t
473
fast_string_match (Lisp_Object regexp, Lisp_Object string)
Richard M. Stallman's avatar
Richard M. Stallman committed
474
{
475
  ptrdiff_t val;
476
  struct re_pattern_buffer *bufp;
Richard M. Stallman's avatar
Richard M. Stallman committed
477

478 479
  bufp = compile_pattern (regexp, 0, Qnil,
			  0, STRING_MULTIBYTE (string));
Richard M. Stallman's avatar
Richard M. Stallman committed
480
  immediate_quit = 1;
481
  re_match_object = string;
482

483
  val = re_search (bufp, SSDATA (string),
484 485
		   SBYTES (string), 0,
		   SBYTES (string), 0);
Richard M. Stallman's avatar
Richard M. Stallman committed
486 487 488
  immediate_quit = 0;
  return val;
}
Karl Heuer's avatar
Karl Heuer committed
489 490 491

/* Match REGEXP against STRING, searching all of STRING ignoring case,
   and return the index of the match, or negative on failure.
492 493
   This does not clobber the match data.
   We assume that STRING contains single-byte characters.  */
Karl Heuer's avatar
Karl Heuer committed
494

495
ptrdiff_t
496 497
fast_c_string_match_ignore_case (Lisp_Object regexp,
				 const char *string, ptrdiff_t len)
Karl Heuer's avatar
Karl Heuer committed
498
{
499
  ptrdiff_t val;
Karl Heuer's avatar
Karl Heuer committed
500 501
  struct re_pattern_buffer *bufp;

502
  regexp = string_make_unibyte (regexp);
503
  re_match_object = Qt;
Karl Heuer's avatar
Karl Heuer committed
504
  bufp = compile_pattern (regexp, 0,
505
			  Vascii_canon_table, 0,
506
			  0);
Karl Heuer's avatar
Karl Heuer committed
507 508 509 510 511
  immediate_quit = 1;
  val = re_search (bufp, string, len, 0, len, 0);
  immediate_quit = 0;
  return val;
}
512 513 514

/* Like fast_string_match but ignore case.  */

515
ptrdiff_t
516
fast_string_match_ignore_case (Lisp_Object regexp, Lisp_Object string)
517
{
518
  ptrdiff_t val;
519 520
  struct re_pattern_buffer *bufp;

521
  bufp = compile_pattern (regexp, 0, Vascii_canon_table,
522 523 524 525
			  0, STRING_MULTIBYTE (string));
  immediate_quit = 1;
  re_match_object = string;

526
  val = re_search (bufp, SSDATA (string),
527 528 529 530 531
		   SBYTES (string), 0,
		   SBYTES (string), 0);
  immediate_quit = 0;
  return val;
}
532

533
/* Match REGEXP against the characters after POS to LIMIT, and return
534 535 536 537 538
   the number of matched characters.  If STRING is non-nil, match
   against the characters in it.  In that case, POS and LIMIT are
   indices into the string.  This function doesn't modify the match
   data.  */

539
ptrdiff_t
540 541
fast_looking_at (Lisp_Object regexp, ptrdiff_t pos, ptrdiff_t pos_byte,
		 ptrdiff_t limit, ptrdiff_t limit_byte, Lisp_Object string)
542
{
543
  bool multibyte;
544 545
  struct re_pattern_buffer *buf;
  unsigned char *p1, *p2;
546 547
  ptrdiff_t s1, s2;
  ptrdiff_t len;
Juanma Barranquero's avatar
Juanma Barranquero committed
548

549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585
  if (STRINGP (string))
    {
      if (pos_byte < 0)
	pos_byte = string_char_to_byte (string, pos);
      if (limit_byte < 0)
	limit_byte = string_char_to_byte (string, limit);
      p1 = NULL;
      s1 = 0;
      p2 = SDATA (string);
      s2 = SBYTES (string);
      re_match_object = string;
      multibyte = STRING_MULTIBYTE (string);
    }
  else
    {
      if (pos_byte < 0)
	pos_byte = CHAR_TO_BYTE (pos);
      if (limit_byte < 0)
	limit_byte = CHAR_TO_BYTE (limit);
      pos_byte -= BEGV_BYTE;
      limit_byte -= BEGV_BYTE;
      p1 = BEGV_ADDR;
      s1 = GPT_BYTE - BEGV_BYTE;
      p2 = GAP_END_ADDR;
      s2 = ZV_BYTE - GPT_BYTE;
      if (s1 < 0)
	{
	  p2 = p1;
	  s2 = ZV_BYTE - BEGV_BYTE;
	  s1 = 0;
	}
      if (s2 < 0)
	{
	  s1 = ZV_BYTE - BEGV_BYTE;
	  s2 = 0;
	}
      re_match_object = Qnil;
Tom Tromey's avatar
Tom Tromey committed
586
      multibyte = ! NILP (BVAR (current_buffer, enable_multibyte_characters));
587 588 589 590 591 592 593 594 595 596 597
    }

  buf = compile_pattern (regexp, 0, Qnil, 0, multibyte);
  immediate_quit = 1;
  len = re_match_2 (buf, (char *) p1, s1, (char *) p2, s2,
		    pos_byte, NULL, limit_byte);
  immediate_quit = 0;

  return len;
}

598 599 600 601 602 603 604 605

/* The newline cache: remembering which sections of text have no newlines.  */

/* If the user has requested newline caching, make sure it's on.
   Otherwise, make sure it's off.
   This is our cheezy way of associating an action with the change of
   state of a buffer-local variable.  */
static void
606
newline_cache_on_off (struct buffer *buf)
607
{
Tom Tromey's avatar
Tom Tromey committed
608
  if (NILP (BVAR (buf, cache_long_line_scans)))
609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625
    {
      /* It should be off.  */
      if (buf->newline_cache)
        {
          free_region_cache (buf->newline_cache);
          buf->newline_cache = 0;
        }
    }
  else
    {
      /* It should be on.  */
      if (buf->newline_cache == 0)
        buf->newline_cache = new_region_cache ();
    }
}


626
/* Search for COUNT newlines between START/START_BYTE and END/END_BYTE.
627 628 629 630 631 632 633 634

   If COUNT is positive, search forwards; END must be >= START.
   If COUNT is negative, search backwards for the -COUNTth instance;
      END must be <= START.
   If COUNT is zero, do anything you please; run rogue, for all I care.

   If END is zero, use BEGV or ZV instead, as appropriate for the
   direction indicated by COUNT.
Jim Blandy's avatar
Jim Blandy committed
635 636

   If we find COUNT instances, set *SHORTAGE to zero, and return the
Richard M. Stallman's avatar
Richard M. Stallman committed
637
   position past the COUNTth match.  Note that for reverse motion
Richard M. Stallman's avatar
Richard M. Stallman committed
638
   this is not the same as the usual convention for Emacs motion commands.
Jim Blandy's avatar
Jim Blandy committed
639

640
   If we don't find COUNT instances before reaching END, set *SHORTAGE
641
   to the number of newlines left unfound, and return END.
Jim Blandy's avatar
Jim Blandy committed
642

643 644 645
   If BYTEPOS is not NULL, set *BYTEPOS to the byte position corresponding
   to the returned character position.

646
   If ALLOW_QUIT, set immediate_quit.  That's good to do
647 648
   except when inside redisplay.  */

649
ptrdiff_t
650 651 652
find_newline (ptrdiff_t start, ptrdiff_t start_byte, ptrdiff_t end,
	      ptrdiff_t end_byte, ptrdiff_t count, ptrdiff_t *shortage,
	      ptrdiff_t *bytepos, bool allow_quit)
Jim Blandy's avatar
Jim Blandy committed
653
{
654
  struct region_cache *newline_cache;
655
  int direction;
Jim Blandy's avatar
Jim Blandy committed
656

657 658 659
  if (count > 0)
    {
      direction = 1;
660 661
      if (!end)
	end = ZV, end_byte = ZV_BYTE;
662 663 664 665
    }
  else
    {
      direction = -1;
666
      if (!end)
667
	end = BEGV, end_byte = BEGV_BYTE;
668
    }
669 670
  if (end_byte == -1)
    end_byte = CHAR_TO_BYTE (end);
Jim Blandy's avatar
Jim Blandy committed
671

672 673
  newline_cache_on_off (current_buffer);
  newline_cache = current_buffer->newline_cache;
Jim Blandy's avatar
Jim Blandy committed
674 675 676 677

  if (shortage != 0)
    *shortage = 0;

678
  immediate_quit = allow_quit;
Jim Blandy's avatar
Jim Blandy committed
679

Jim Blandy's avatar
Jim Blandy committed
680
  if (count > 0)
681
    while (start != end)
Jim Blandy's avatar
Jim Blandy committed
682
      {
683 684 685 686 687
        /* Our innermost scanning loop is very simple; it doesn't know
           about gaps, buffer ends, or the newline cache.  ceiling is
           the position of the last character before the next such
           obstacle --- the last character the dumb search loop should
           examine.  */
688
	ptrdiff_t tem, ceiling_byte = end_byte - 1;
689 690 691

        /* If we're looking for a newline, consult the newline cache
           to see where we can avoid some scanning.  */
692
        if (newline_cache)
693
          {
694
            ptrdiff_t next_change;
695 696
            immediate_quit = 0;
            while (region_cache_forward
697 698
                   (current_buffer, newline_cache, start, &next_change))
              start = next_change;
699
            immediate_quit = allow_quit;
700

701 702
	    start_byte = CHAR_TO_BYTE (start);

703 704 705
            /* START should never be after END.  */
            if (start_byte > ceiling_byte)
              start_byte = ceiling_byte;
706 707 708

            /* Now the text after start is an unknown region, and
               next_change is the position of the next known region. */
709
            ceiling_byte = min (CHAR_TO_BYTE (next_change) - 1, ceiling_byte);
710
          }
711
	else if (start_byte == -1)
712
	  start_byte = CHAR_TO_BYTE (start);
713 714 715 716 717

        /* The dumb loop can only scan text stored in contiguous
           bytes. BUFFER_CEILING_OF returns the last character
           position that is contiguous, so the ceiling is the
           position after that.  */
718 719
	tem = BUFFER_CEILING_OF (start_byte);
	ceiling_byte = min (tem, ceiling_byte);
720 721

        {
722
          /* The termination address of the dumb loop.  */
723 724 725 726
          register unsigned char *ceiling_addr
	    = BYTE_POS_ADDR (ceiling_byte) + 1;
          register unsigned char *cursor
	    = BYTE_POS_ADDR (start_byte);
727 728 729 730 731
          unsigned char *base = cursor;

          while (cursor < ceiling_addr)
            {
              /* The dumb loop.  */
732
	      unsigned char *nl = memchr (cursor, '\n', ceiling_addr - cursor);
733 734 735

              /* If we're looking for newlines, cache the fact that
                 the region from start to cursor is free of them. */
736 737 738 739 740 741 742 743 744 745 746 747 748 749 750
              if (newline_cache)
		{
		  unsigned char *low = cursor;
		  unsigned char *lim = nl ? nl : ceiling_addr;
		  know_region_cache (current_buffer, newline_cache,
				     BYTE_TO_CHAR (low - base + start_byte),
				     BYTE_TO_CHAR (lim - base + start_byte));
		}

              if (! nl)
		break;

	      if (--count == 0)
		{
		  immediate_quit = 0;
751 752
		  if (bytepos)
		    *bytepos = nl + 1 - base + start_byte;
753 754 755
		  return BYTE_TO_CHAR (nl + 1 - base + start_byte);
		}
	      cursor = nl + 1;
756 757
            }

758 759
	  start_byte += ceiling_addr - base;
	  start = BYTE_TO_CHAR (start_byte);
760
        }
Jim Blandy's avatar
Jim Blandy committed
761 762
      }
  else
763 764 765
    while (start > end)
      {
        /* The last character to check before the next obstacle.  */
766
	ptrdiff_t tem, ceiling_byte = end_byte;
767 768

        /* Consult the newline cache, if appropriate.  */
769
        if (newline_cache)
770
          {
771
            ptrdiff_t next_change;
772 773
            immediate_quit = 0;
            while (region_cache_backward
774 775
                   (current_buffer, newline_cache, start, &next_change))
              start = next_change;
776
            immediate_quit = allow_quit;
777

778 779
	    start_byte = CHAR_TO_BYTE (start);

780
            /* Start should never be at or before end.  */
781 782
            if (start_byte <= ceiling_byte)
              start_byte = ceiling_byte + 1;
783 784 785

            /* Now the text before start is an unknown region, and
               next_change is the position of the next known region. */
786
            ceiling_byte = max (CHAR_TO_BYTE (next_change), ceiling_byte);
787
          }
788
	else if (start_byte == -1)
789
	  start_byte = CHAR_TO_BYTE (start);
790 791

        /* Stop scanning before the gap.  */
792 793
	tem = BUFFER_FLOOR_OF (start_byte - 1);
	ceiling_byte = max (tem, ceiling_byte);
794 795 796

        {
          /* The termination address of the dumb loop.  */
797 798
          register unsigned char *ceiling_addr = BYTE_POS_ADDR (ceiling_byte);
          register unsigned char *cursor = BYTE_POS_ADDR (start_byte - 1);
799 800 801 802
          unsigned char *base = cursor;

          while (cursor >= ceiling_addr)
            {
803 804
	      unsigned char *nl = memrchr (ceiling_addr, '\n',
					   cursor + 1 - ceiling_addr);
805 806 807

              /* If we're looking for newlines, cache the fact that
                 the region from after the cursor to start is free of them.  */
808 809 810 811 812 813 814 815 816 817 818 819 820 821 822
              if (newline_cache)
		{
		  unsigned char *low = nl ? nl : ceiling_addr - 1;
		  unsigned char *lim = cursor;
		  know_region_cache (current_buffer, newline_cache,
				     BYTE_TO_CHAR (low - base + start_byte),
				     BYTE_TO_CHAR (lim - base + start_byte));
		}

              if (! nl)
		break;

	      if (++count >= 0)
		{
		  immediate_quit = 0;
823 824
		  if (bytepos)
		    *bytepos = nl - base + start_byte;
825 826 827
		  return BYTE_TO_CHAR (nl - base + start_byte);
		}
	      cursor = nl - 1;
828 829
            }

830 831
	  start_byte += ceiling_addr - 1 - base;
	  start = BYTE_TO_CHAR (start_byte);
832 833 834
        }
      }

Jim Blandy's avatar
Jim Blandy committed
835
  immediate_quit = 0;
836
  if (shortage)
Jim Blandy's avatar
Jim Blandy committed
837
    *shortage = count * direction;
838 839 840 841 842
  if (bytepos)
    {
      *bytepos = start_byte == -1 ? CHAR_TO_BYTE (start) : start_byte;
      eassert (*bytepos == CHAR_TO_BYTE (start));
    }
843
  return start;
Jim Blandy's avatar
Jim Blandy committed
844
}
845

846
/* Search for COUNT instances of a line boundary.
847 848 849 850 851 852 853 854 855 856 857 858
   Start at START.  If COUNT is negative, search backwards.

   We report the resulting position by calling TEMP_SET_PT_BOTH.

   If we find COUNT instances. we position after (always after,
   even if scanning backwards) the COUNTth match, and return 0.

   If we don't find COUNT instances before reaching the end of the
   buffer (or the beginning, if scanning backwards), we return
   the number of line boundaries left unfound, and position at
   the limit we bumped up against.

859
   If ALLOW_QUIT, set immediate_quit.  That's good to do
860
   except in special cases.  */
Jim Blandy's avatar
Jim Blandy committed
861

862
EMACS_INT
863 864
scan_newline (ptrdiff_t start, ptrdiff_t start_byte,
	      ptrdiff_t limit, ptrdiff_t limit_byte,
865
	      EMACS_INT count, bool allow_quit)
866
{
867 868
  int direction = ((count > 0) ? 1 : -1);

869
  unsigned char *cursor;
870 871
  unsigned char *base;

872
  ptrdiff_t ceiling;
873
  unsigned char *ceiling_addr;
874

875
  bool old_immediate_quit = immediate_quit;
876 877 878

  if (allow_quit)
    immediate_quit++;
879 880 881 882 883 884 885 886 887 888

  if (count > 0)
    {
      while (start_byte < limit_byte)
	{
	  ceiling =  BUFFER_CEILING_OF (start_byte);
	  ceiling = min (limit_byte - 1, ceiling);
	  ceiling_addr = BYTE_POS_ADDR (ceiling) + 1;
	  base = (cursor = BYTE_POS_ADDR (start_byte));

889 890 891 892 893 894
	  do
	    {
	      unsigned char *nl = memchr (cursor, '\n', ceiling_addr - cursor);
	      if (! nl)
		break;
	      if (--count == 0)
895
		{
896 897 898 899 900
		  immediate_quit = old_immediate_quit;
		  start_byte += nl - base + 1;
		  start = BYTE_TO_CHAR (start_byte);
		  TEMP_SET_PT_BOTH (start, start_byte);
		  return 0;
901
		}
902
	      cursor = nl + 1;
903
	    }
904 905 906
	  while (cursor < ceiling_addr);

	  start_byte += ceiling_addr - base;
907 908 909 910 911 912 913 914
	}
    }
  else
    {
      while (start_byte > limit_byte)
	{
	  ceiling = BUFFER_FLOOR_OF (start_byte - 1);
	  ceiling = max (limit_byte, ceiling);
915
	  ceiling_addr = BYTE_POS_ADDR (ceiling);
916 917 918
	  base = (cursor = BYTE_POS_ADDR (start_byte - 1) + 1);
	  while (1)
	    {
919 920 921 922
	      unsigned char *nl = memrchr (ceiling_addr, '\n',
					   cursor - ceiling_addr);
	      if (! nl)
		break;