search.c 95.5 KB
Newer Older
Jim Blandy's avatar
Jim Blandy committed
1
/* String search routines for GNU Emacs.
2
   Copyright (C) 1985, 1986, 1987, 1993, 1994, 1997, 1998, 1999, 2001, 2002,
Glenn Morris's avatar
Glenn Morris committed
3
                 2003, 2004, 2005, 2006, 2007, 2008, 2009
Glenn Morris's avatar
Glenn Morris committed
4
                 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>
Jim Blandy's avatar
Jim Blandy committed
23 24
#include "lisp.h"
#include "syntax.h"
Karl Heuer's avatar
Karl Heuer committed
25
#include "category.h"
Jim Blandy's avatar
Jim Blandy committed
26
#include "buffer.h"
27
#include "character.h"
Kenichi Handa's avatar
Kenichi Handa committed
28
#include "charset.h"
29
#include "region-cache.h"
Jim Blandy's avatar
Jim Blandy committed
30
#include "commands.h"
31
#include "blockinput.h"
32
#include "intervals.h"
Jim Blandy's avatar
Jim Blandy committed
33

Jim Blandy's avatar
Jim Blandy committed
34 35 36
#include <sys/types.h>
#include "regex.h"

37
#define REGEXP_CACHE_SIZE 20
Jim Blandy's avatar
Jim Blandy committed
38

39 40
/* If the regexp is non-nil, then the buffer contains the compiled form
   of that regexp, suitable for searching.  */
41 42
struct regexp_cache
{
43
  struct regexp_cache *next;
44
  Lisp_Object regexp, whitespace_regexp;
45
  /* Syntax table for which the regexp applies.  We need this because
46 47
     of character classes.  If this is t, then the compiled pattern is valid
     for any syntax-table.  */
48
  Lisp_Object syntax_table;
49 50
  struct re_pattern_buffer buf;
  char fastmap[0400];
51 52
  /* Nonzero means regexp was compiled to do full POSIX backtracking.  */
  char posix;
53
};
Jim Blandy's avatar
Jim Blandy committed
54

55 56
/* The instances of that struct.  */
struct regexp_cache searchbufs[REGEXP_CACHE_SIZE];
Jim Blandy's avatar
Jim Blandy committed
57

58 59
/* The head of the linked list; points to the most recently used buffer.  */
struct regexp_cache *searchbuf_head;
Jim Blandy's avatar
Jim Blandy committed
60 61


Jim Blandy's avatar
Jim Blandy committed
62 63 64 65 66 67 68
/* 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
69 70 71
   been allocated by checking search_regs.num_regs.

   The regex code keeps track of whether it has allocated the search
72 73
   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
74 75 76 77 78
   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
79 80
static struct re_registers search_regs;

Jim Blandy's avatar
Jim Blandy committed
81 82 83 84
/* 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
85

Karl Heuer's avatar
Karl Heuer committed
86
/* error condition signaled when regexp compile_pattern fails */
Jim Blandy's avatar
Jim Blandy committed
87 88 89

Lisp_Object Qinvalid_regexp;

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

Richard M. Stallman's avatar
Richard M. Stallman committed
93
Lisp_Object Vsearch_spaces_regexp;
94

95 96 97 98 99
/* If non-nil, the match data will not be changed during call to
   searching or matching functions.  This variable is for internal use
   only.  */
Lisp_Object Vinhibit_changing_match_data;

100
static void set_search_regs ();
101
static void save_search_regs ();
102 103
static int simple_search ();
static int boyer_moore ();
104
static int search_buffer ();
105
static void matcher_overflow () NO_RETURN;
106

Jim Blandy's avatar
Jim Blandy committed
107 108 109 110 111 112
static void
matcher_overflow ()
{
  error ("Stack overflow in regexp matcher");
}

113 114 115
/* 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.
116
   TRANSLATE is a translation table for ignoring case, or nil for none.
117 118 119 120 121
   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.
   POSIX is nonzero if we want full backtracking (POSIX style)
Karl Heuer's avatar
Karl Heuer committed
122
   for this pattern.  0 means backtrack only enough to get a valid match.
123 124

   The behavior also depends on Vsearch_spaces_regexp.  */
Jim Blandy's avatar
Jim Blandy committed
125

126
static void
Kenichi Handa's avatar
Kenichi Handa committed
127
compile_pattern_1 (cp, pattern, translate, regp, posix)
128
     struct regexp_cache *cp;
Jim Blandy's avatar
Jim Blandy committed
129
     Lisp_Object pattern;
130
     Lisp_Object translate;
131
     struct re_registers *regp;
132
     int posix;
Jim Blandy's avatar
Jim Blandy committed
133
{
134
  char *val;
135
  reg_syntax_t old;
Jim Blandy's avatar
Jim Blandy committed
136

137
  cp->regexp = Qnil;
138
  cp->buf.translate = (! NILP (translate) ? translate : make_number (0));
139
  cp->posix = posix;
140
  cp->buf.multibyte = STRING_MULTIBYTE (pattern);
Kenichi Handa's avatar
Kenichi Handa committed
141
  cp->buf.charset_unibyte = charset_unibyte;
142 143 144 145 146
  if (STRINGP (Vsearch_spaces_regexp))
    cp->whitespace_regexp = Vsearch_spaces_regexp;
  else
    cp->whitespace_regexp = Qnil;

147 148 149 150 151
  /* 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;  */
152
  old = re_set_syntax (RE_SYNTAX_EMACS
153
		       | (posix ? 0 : RE_NO_POSIX_BACKTRACKING));
154 155 156 157 158

  if (STRINGP (Vsearch_spaces_regexp))
    re_set_whitespace_regexp (SDATA (Vsearch_spaces_regexp));
  else
    re_set_whitespace_regexp (NULL);
159

Kenichi Handa's avatar
Kenichi Handa committed
160 161
  val = (char *) re_compile_pattern ((char *) SDATA (pattern),
				     SBYTES (pattern), &cp->buf);
162

163 164 165 166
  /* If the compiled pattern hard codes some of the contents of the
     syntax-table, it can only be reused with *this* syntax table.  */
  cp->syntax_table = cp->buf.used_syntax ? current_buffer->syntax_table : Qt;

167 168
  re_set_whitespace_regexp (NULL);

169
  re_set_syntax (old);
170
  /* UNBLOCK_INPUT;  */
Jim Blandy's avatar
Jim Blandy committed
171
  if (val)
172
    xsignal1 (Qinvalid_regexp, build_string (val));
Jim Blandy's avatar
Jim Blandy committed
173

174 175 176
  cp->regexp = Fcopy_sequence (pattern);
}

177 178 179 180 181 182 183
/* Shrink each compiled regexp buffer in the cache
   to the size actually used right now.
   This is called from garbage collection.  */

void
shrink_regexp_cache ()
{
184
  struct regexp_cache *cp;
185 186 187 188 189

  for (cp = searchbuf_head; cp != 0; cp = cp->next)
    {
      cp->buf.allocated = cp->buf.used;
      cp->buf.buffer
190
	= (unsigned char *) xrealloc (cp->buf.buffer, cp->buf.used);
191 192 193
    }
}

194 195
/* Clear the regexp cache w.r.t. a particular syntax table,
   because it was changed.
196 197 198 199 200 201 202 203 204
   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
clear_regexp_cache ()
{
  int i;

  for (i = 0; i < REGEXP_CACHE_SIZE; ++i)
205 206 207 208 209
    /* It's tempting to compare with the syntax-table we've actually changd,
       but it's not sufficient because char-table inheritance mewans that
       modifying one syntax-table can change others at the same time.  */
    if (!EQ (searchbufs[i].syntax_table, Qt))
      searchbufs[i].regexp = Qnil;
210 211
}

212
/* Compile a regexp if necessary, but first check to see if there's one in
213 214
   the cache.
   PATTERN is the pattern to compile.
215
   TRANSLATE is a translation table for ignoring case, or nil for none.
216 217 218 219 220 221
   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.
   POSIX is nonzero if we want full backtracking (POSIX style)
   for this pattern.  0 means backtrack only enough to get a valid match.  */
222 223

struct re_pattern_buffer *
224
compile_pattern (pattern, regp, translate, posix, multibyte)
225 226
     Lisp_Object pattern;
     struct re_registers *regp;
227
     Lisp_Object translate;
228
     int posix, multibyte;
229 230 231 232 233 234
{
  struct regexp_cache *cp, **cpp;

  for (cpp = &searchbuf_head; ; cpp = &cp->next)
    {
      cp = *cpp;
235 236
      /* 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
237 238 239
	 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.  */
240
      if (NILP (cp->regexp))
241
	goto compile_it;
242
      if (SCHARS (cp->regexp) == SCHARS (pattern)
243
	  && STRING_MULTIBYTE (cp->regexp) == STRING_MULTIBYTE (pattern)
244
	  && !NILP (Fstring_equal (cp->regexp, pattern))
245
	  && EQ (cp->buf.translate, (! NILP (translate) ? translate : make_number (0)))
Karl Heuer's avatar
Karl Heuer committed
246
	  && cp->posix == posix
247 248
	  && (EQ (cp->syntax_table, Qt)
	      || EQ (cp->syntax_table, current_buffer->syntax_table))
Kenichi Handa's avatar
Kenichi Handa committed
249 250
	  && !NILP (Fequal (cp->whitespace_regexp, Vsearch_spaces_regexp))
	  && cp->buf.charset_unibyte == charset_unibyte)
251 252
	break;

253 254 255
      /* 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.  */
256 257
      if (cp->next == 0)
	{
258
	compile_it:
Kenichi Handa's avatar
Kenichi Handa committed
259
	  compile_pattern_1 (cp, pattern, translate, regp, posix);
260 261 262 263 264 265 266 267 268 269
	  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
270

271 272 273 274 275
  /* 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);

Kenichi Handa's avatar
Kenichi Handa committed
276 277 278 279
  /* The compiled pattern can be used both for mulitbyte and unibyte
     target.  But, we have to tell which the pattern is used for. */
  cp->buf.target_multibyte = multibyte;

280
  return &cp->buf;
Jim Blandy's avatar
Jim Blandy committed
281 282 283
}


284 285
static Lisp_Object
looking_at_1 (string, posix)
Jim Blandy's avatar
Jim Blandy committed
286
     Lisp_Object string;
287
     int posix;
Jim Blandy's avatar
Jim Blandy committed
288 289 290 291 292
{
  Lisp_Object val;
  unsigned char *p1, *p2;
  int s1, s2;
  register int i;
293
  struct re_pattern_buffer *bufp;
Jim Blandy's avatar
Jim Blandy committed
294

295 296 297
  if (running_asynch_code)
    save_search_regs ();

298 299 300 301
  /* This is so set_image_of_range_1 in regex.c can find the EQV table.  */
  XCHAR_TABLE (current_buffer->case_canon_table)->extras[2]
    = current_buffer->case_eqv_table;

302
  CHECK_STRING (string);
303 304 305
  bufp = compile_pattern (string,
			  (NILP (Vinhibit_changing_match_data)
			   ? &search_regs : NULL),
306
			  (!NILP (current_buffer->case_fold_search)
307
			   ? current_buffer->case_canon_table : Qnil),
308 309
			  posix,
			  !NILP (current_buffer->enable_multibyte_characters));
Jim Blandy's avatar
Jim Blandy committed
310 311 312 313 314 315 316 317

  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;
318
  s1 = GPT_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
319
  p2 = GAP_END_ADDR;
320
  s2 = ZV_BYTE - GPT_BYTE;
Jim Blandy's avatar
Jim Blandy committed
321 322 323
  if (s1 < 0)
    {
      p2 = p1;
324
      s2 = ZV_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
325 326 327 328
      s1 = 0;
    }
  if (s2 < 0)
    {
329
      s1 = ZV_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
330 331
      s2 = 0;
    }
332 333

  re_match_object = Qnil;
334

335
  i = re_match_2 (bufp, (char *) p1, s1, (char *) p2, s2,
336 337 338
		  PT_BYTE - BEGV_BYTE,
		  (NILP (Vinhibit_changing_match_data)
		   ? &search_regs : NULL),
339
		  ZV_BYTE - BEGV_BYTE);
340
  immediate_quit = 0;
341

Jim Blandy's avatar
Jim Blandy committed
342 343 344 345
  if (i == -2)
    matcher_overflow ();

  val = (0 <= i ? Qt : Qnil);
346
  if (NILP (Vinhibit_changing_match_data) && i >= 0)
347 348 349 350 351 352 353 354
    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);
	  search_regs.end[i]
	    = BYTE_TO_CHAR (search_regs.end[i] + BEGV_BYTE);
	}
355 356 357 358 359

  /* Set last_thing_searched only when match data is changed.  */
  if (NILP (Vinhibit_changing_match_data))
    XSETBUFFER (last_thing_searched, current_buffer);

Jim Blandy's avatar
Jim Blandy committed
360 361 362
  return val;
}

363
DEFUN ("looking-at", Flooking_at, Slooking_at, 1, 1, 0,
364 365 366 367 368
       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.  */)
     (regexp)
369
     Lisp_Object regexp;
370
{
371
  return looking_at_1 (regexp, 0);
372 373 374
}

DEFUN ("posix-looking-at", Fposix_looking_at, Sposix_looking_at, 1, 1, 0,
375 376 377 378 379 380
       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.  */)
     (regexp)
381
     Lisp_Object regexp;
382
{
383
  return looking_at_1 (regexp, 1);
384 385 386 387
}

static Lisp_Object
string_match_1 (regexp, string, start, posix)
Jim Blandy's avatar
Jim Blandy committed
388
     Lisp_Object regexp, string, start;
389
     int posix;
Jim Blandy's avatar
Jim Blandy committed
390 391
{
  int val;
392
  struct re_pattern_buffer *bufp;
393 394
  int pos, pos_byte;
  int i;
Jim Blandy's avatar
Jim Blandy committed
395

396 397 398
  if (running_asynch_code)
    save_search_regs ();

399 400
  CHECK_STRING (regexp);
  CHECK_STRING (string);
Jim Blandy's avatar
Jim Blandy committed
401 402

  if (NILP (start))
403
    pos = 0, pos_byte = 0;
Jim Blandy's avatar
Jim Blandy committed
404 405
  else
    {
406
      int len = SCHARS (string);
Jim Blandy's avatar
Jim Blandy committed
407

408
      CHECK_NUMBER (start);
409 410 411 412
      pos = XINT (start);
      if (pos < 0 && -pos <= len)
	pos = len + pos;
      else if (0 > pos || pos > len)
Jim Blandy's avatar
Jim Blandy committed
413
	args_out_of_range (string, start);
414
      pos_byte = string_char_to_byte (string, pos);
Jim Blandy's avatar
Jim Blandy committed
415 416
    }

417 418 419 420
  /* This is so set_image_of_range_1 in regex.c can find the EQV table.  */
  XCHAR_TABLE (current_buffer->case_canon_table)->extras[2]
    = current_buffer->case_eqv_table;

421 422 423
  bufp = compile_pattern (regexp,
			  (NILP (Vinhibit_changing_match_data)
			   ? &search_regs : NULL),
424
			  (!NILP (current_buffer->case_fold_search)
425
			   ? current_buffer->case_canon_table : Qnil),
426 427
			  posix,
			  STRING_MULTIBYTE (string));
Jim Blandy's avatar
Jim Blandy committed
428
  immediate_quit = 1;
429
  re_match_object = string;
430

431 432 433
  val = re_search (bufp, (char *) SDATA (string),
		   SBYTES (string), pos_byte,
		   SBYTES (string) - pos_byte,
434 435
		   (NILP (Vinhibit_changing_match_data)
		    ? &search_regs : NULL));
Jim Blandy's avatar
Jim Blandy committed
436
  immediate_quit = 0;
437 438 439 440 441

  /* 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
442 443 444
  if (val == -2)
    matcher_overflow ();
  if (val < 0) return Qnil;
445

446 447 448 449 450 451 452 453 454
  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]);
	}
455 456

  return make_number (string_byte_to_char (string, val));
Jim Blandy's avatar
Jim Blandy committed
457
}
Richard M. Stallman's avatar
Richard M. Stallman committed
458

459
DEFUN ("string-match", Fstring_match, Sstring_match, 2, 3, 0,
460
       doc: /* Return index of start of first match for REGEXP in STRING, or nil.
Richard M. Stallman's avatar
Richard M. Stallman committed
461
Matching ignores case if `case-fold-search' is non-nil.
462 463 464
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
465 466 467 468
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. */)
469
     (regexp, string, start)
470 471 472 473 474 475
     Lisp_Object regexp, string, start;
{
  return string_match_1 (regexp, string, start, 0);
}

DEFUN ("posix-string-match", Fposix_string_match, Sposix_string_match, 2, 3, 0,
476 477 478 479 480 481 482 483
       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.  */)
     (regexp, string, start)
484 485 486 487 488
     Lisp_Object regexp, string, start;
{
  return string_match_1 (regexp, string, start, 1);
}

Richard M. Stallman's avatar
Richard M. Stallman committed
489 490 491 492 493 494 495 496 497
/* 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.  */

int
fast_string_match (regexp, string)
     Lisp_Object regexp, string;
{
  int val;
498
  struct re_pattern_buffer *bufp;
Richard M. Stallman's avatar
Richard M. Stallman committed
499

500 501
  bufp = compile_pattern (regexp, 0, Qnil,
			  0, STRING_MULTIBYTE (string));
Richard M. Stallman's avatar
Richard M. Stallman committed
502
  immediate_quit = 1;
503
  re_match_object = string;
504

505 506 507
  val = re_search (bufp, (char *) SDATA (string),
		   SBYTES (string), 0,
		   SBYTES (string), 0);
Richard M. Stallman's avatar
Richard M. Stallman committed
508 509 510
  immediate_quit = 0;
  return val;
}
Karl Heuer's avatar
Karl Heuer committed
511 512 513

/* Match REGEXP against STRING, searching all of STRING ignoring case,
   and return the index of the match, or negative on failure.
514 515
   This does not clobber the match data.
   We assume that STRING contains single-byte characters.  */
Karl Heuer's avatar
Karl Heuer committed
516 517 518 519

extern Lisp_Object Vascii_downcase_table;

int
520
fast_c_string_match_ignore_case (regexp, string)
Karl Heuer's avatar
Karl Heuer committed
521
     Lisp_Object regexp;
522
     const char *string;
Karl Heuer's avatar
Karl Heuer committed
523 524 525 526 527
{
  int val;
  struct re_pattern_buffer *bufp;
  int len = strlen (string);

528
  regexp = string_make_unibyte (regexp);
529
  re_match_object = Qt;
Karl Heuer's avatar
Karl Heuer committed
530
  bufp = compile_pattern (regexp, 0,
531
			  Vascii_canon_table, 0,
532
			  0);
Karl Heuer's avatar
Karl Heuer committed
533 534 535 536 537
  immediate_quit = 1;
  val = re_search (bufp, string, len, 0, len, 0);
  immediate_quit = 0;
  return val;
}
538 539 540 541 542 543 544 545 546 547

/* Like fast_string_match but ignore case.  */

int
fast_string_match_ignore_case (regexp, string)
     Lisp_Object regexp, string;
{
  int val;
  struct re_pattern_buffer *bufp;

548
  bufp = compile_pattern (regexp, 0, Vascii_canon_table,
549 550 551 552 553 554 555 556 557 558
			  0, STRING_MULTIBYTE (string));
  immediate_quit = 1;
  re_match_object = string;

  val = re_search (bufp, (char *) SDATA (string),
		   SBYTES (string), 0,
		   SBYTES (string), 0);
  immediate_quit = 0;
  return val;
}
559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626

/* Match REGEXP atainst the characters after POS to LIMIT, and return
   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.  */

EMACS_INT
fast_looking_at (regexp, pos, pos_byte, limit, limit_byte, string)
     Lisp_Object regexp;
     EMACS_INT pos, pos_byte, limit, limit_byte;
     Lisp_Object string;
{
  int multibyte;
  struct re_pattern_buffer *buf;
  unsigned char *p1, *p2;
  int s1, s2;
  EMACS_INT len;
  
  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;
      multibyte = ! NILP (current_buffer->enable_multibyte_characters);
    }

  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;
}

627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664

/* 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
newline_cache_on_off (buf)
     struct buffer *buf;
{
  if (NILP (buf->cache_long_line_scans))
    {
      /* 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 ();
    }
}


/* Search for COUNT instances of the character TARGET between START and END.

   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
665 666

   If we find COUNT instances, set *SHORTAGE to zero, and return the
Richard M. Stallman's avatar
Richard M. Stallman committed
667
   position past the COUNTth match.  Note that for reverse motion
Richard M. Stallman's avatar
Richard M. Stallman committed
668
   this is not the same as the usual convention for Emacs motion commands.
Jim Blandy's avatar
Jim Blandy committed
669

670 671
   If we don't find COUNT instances before reaching END, set *SHORTAGE
   to the number of TARGETs left unfound, and return END.
Jim Blandy's avatar
Jim Blandy committed
672

673 674 675
   If ALLOW_QUIT is non-zero, set immediate_quit.  That's good to do
   except when inside redisplay.  */

Andreas Schwab's avatar
Andreas Schwab committed
676
int
677 678 679 680 681
scan_buffer (target, start, end, count, shortage, allow_quit)
     register int target;
     int start, end;
     int count;
     int *shortage;
682
     int allow_quit;
Jim Blandy's avatar
Jim Blandy committed
683
{
684
  struct region_cache *newline_cache;
685
  int direction;
Jim Blandy's avatar
Jim Blandy committed
686

687 688 689 690 691 692 693 694 695 696
  if (count > 0)
    {
      direction = 1;
      if (! end) end = ZV;
    }
  else
    {
      direction = -1;
      if (! end) end = BEGV;
    }
Jim Blandy's avatar
Jim Blandy committed
697

698 699
  newline_cache_on_off (current_buffer);
  newline_cache = current_buffer->newline_cache;
Jim Blandy's avatar
Jim Blandy committed
700 701 702 703

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

704
  immediate_quit = allow_quit;
Jim Blandy's avatar
Jim Blandy committed
705

Jim Blandy's avatar
Jim Blandy committed
706
  if (count > 0)
707
    while (start != end)
Jim Blandy's avatar
Jim Blandy committed
708
      {
709 710 711 712 713
        /* 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.  */
714 715
	int ceiling_byte = CHAR_TO_BYTE (end) - 1;
	int start_byte = CHAR_TO_BYTE (start);
716
	int tem;
717 718 719 720 721 722 723 724

        /* If we're looking for a newline, consult the newline cache
           to see where we can avoid some scanning.  */
        if (target == '\n' && newline_cache)
          {
            int next_change;
            immediate_quit = 0;
            while (region_cache_forward
725 726
                   (current_buffer, newline_cache, start_byte, &next_change))
              start_byte = next_change;
727
            immediate_quit = allow_quit;
728

729 730 731
            /* START should never be after END.  */
            if (start_byte > ceiling_byte)
              start_byte = ceiling_byte;
732 733 734

            /* Now the text after start is an unknown region, and
               next_change is the position of the next known region. */
735
            ceiling_byte = min (next_change - 1, ceiling_byte);
736 737 738 739 740 741
          }

        /* 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.  */
742 743
	tem = BUFFER_CEILING_OF (start_byte);
	ceiling_byte = min (tem, ceiling_byte);
744 745

        {
746
          /* The termination address of the dumb loop.  */
747 748 749 750
          register unsigned char *ceiling_addr
	    = BYTE_POS_ADDR (ceiling_byte) + 1;
          register unsigned char *cursor
	    = BYTE_POS_ADDR (start_byte);
751 752 753 754 755 756 757 758 759 760 761 762 763 764
          unsigned char *base = cursor;

          while (cursor < ceiling_addr)
            {
              unsigned char *scan_start = cursor;

              /* The dumb loop.  */
              while (*cursor != target && ++cursor < ceiling_addr)
                ;

              /* If we're looking for newlines, cache the fact that
                 the region from start to cursor is free of them. */
              if (target == '\n' && newline_cache)
                know_region_cache (current_buffer, newline_cache,
765 766
                                   start_byte + scan_start - base,
                                   start_byte + cursor - base);
767 768 769 770 771 772 773

              /* Did we find the target character?  */
              if (cursor < ceiling_addr)
                {
                  if (--count == 0)
                    {
                      immediate_quit = 0;
774
                      return BYTE_TO_CHAR (start_byte + cursor - base + 1);
775 776 777 778 779
                    }
                  cursor++;
                }
            }

780
          start = BYTE_TO_CHAR (start_byte + cursor - base);
781
        }
Jim Blandy's avatar
Jim Blandy committed
782 783
      }
  else
784 785 786
    while (start > end)
      {
        /* The last character to check before the next obstacle.  */
787 788
	int ceiling_byte = CHAR_TO_BYTE (end);
	int start_byte = CHAR_TO_BYTE (start);
789
	int tem;
790 791 792 793 794 795 796

        /* Consult the newline cache, if appropriate.  */
        if (target == '\n' && newline_cache)
          {
            int next_change;
            immediate_quit = 0;
            while (region_cache_backward
797 798
                   (current_buffer, newline_cache, start_byte, &next_change))
              start_byte = next_change;
799
            immediate_quit = allow_quit;
800 801

            /* Start should never be at or before end.  */
802 803
            if (start_byte <= ceiling_byte)
              start_byte = ceiling_byte + 1;
804 805 806

            /* Now the text before start is an unknown region, and
               next_change is the position of the next known region. */
807
            ceiling_byte = max (next_change, ceiling_byte);
808 809 810
          }

        /* Stop scanning before the gap.  */
811 812
	tem = BUFFER_FLOOR_OF (start_byte - 1);
	ceiling_byte = max (tem, ceiling_byte);
813 814 815

        {
          /* The termination address of the dumb loop.  */
816 817
          register unsigned char *ceiling_addr = BYTE_POS_ADDR (ceiling_byte);
          register unsigned char *cursor = BYTE_POS_ADDR (start_byte - 1);
818 819 820 821 822 823 824 825 826 827 828 829 830
          unsigned char *base = cursor;

          while (cursor >= ceiling_addr)
            {
              unsigned char *scan_start = cursor;

              while (*cursor != target && --cursor >= ceiling_addr)
                ;

              /* If we're looking for newlines, cache the fact that
                 the region from after the cursor to start is free of them.  */
              if (target == '\n' && newline_cache)
                know_region_cache (current_buffer, newline_cache,
831 832
                                   start_byte + cursor - base,
                                   start_byte + scan_start - base);
833 834 835 836 837 838 839

              /* Did we find the target character?  */
              if (cursor >= ceiling_addr)
                {
                  if (++count >= 0)
                    {
                      immediate_quit = 0;
840
                      return BYTE_TO_CHAR (start_byte + cursor - base);
841 842 843 844 845
                    }
                  cursor--;
                }
            }

846
	  start = BYTE_TO_CHAR (start_byte + cursor - base);
847 848 849
        }
      }

Jim Blandy's avatar
Jim Blandy committed
850 851
  immediate_quit = 0;
  if (shortage != 0)
Jim Blandy's avatar
Jim Blandy committed
852
    *shortage = count * direction;
853
  return start;
Jim Blandy's avatar
Jim Blandy committed
854
}
855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870

/* Search for COUNT instances of a line boundary, which means either a
   newline or (if selective display enabled) a carriage return.
   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.

   If ALLOW_QUIT is non-zero, set immediate_quit.  That's good to do
871
   except in special cases.  */
Jim Blandy's avatar
Jim Blandy committed
872

873
int
874 875 876 877 878
scan_newline (start, start_byte, limit, limit_byte, count, allow_quit)
     int start, start_byte;
     int limit, limit_byte;
     register int count;
     int allow_quit;
879
{
880 881 882 883 884 885 886 887
  int direction = ((count > 0) ? 1 : -1);

  register unsigned char *cursor;
  unsigned char *base;

  register int ceiling;
  register unsigned char *ceiling_addr;

888 889
  int old_immediate_quit = immediate_quit;

890 891 892
  /* The code that follows is like scan_buffer
     but checks for either newline or carriage return.  */

893 894
  if (allow_quit)
    immediate_quit++;
895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914

  start_byte = CHAR_TO_BYTE (start);

  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));
	  while (1)
	    {
	      while (*cursor != '\n' && ++cursor != ceiling_addr)
		;

	      if (cursor != ceiling_addr)
		{
		  if (--count == 0)
		    {
915
		      immediate_quit = old_immediate_quit;
916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947
		      start_byte = start_byte + cursor - base + 1;
		      start = BYTE_TO_CHAR (start_byte);
		      TEMP_SET_PT_BOTH (start, start_byte);
		      return 0;
		    }
		  else
		    if (++cursor == ceiling_addr)
		      break;
		}
	      else
		break;
	    }
	  start_byte += cursor - base;
	}
    }
  else
    {
      while (start_byte > limit_byte)
	{
	  ceiling = BUFFER_FLOOR_OF (start_byte - 1);
	  ceiling = max (limit_byte, ceiling);
	  ceiling_addr = BYTE_POS_ADDR (ceiling) - 1;
	  base = (cursor = BYTE_POS_ADDR (start_byte - 1) + 1);
	  while (1)
	    {
	      while (--cursor != ceiling_addr && *cursor != '\n')
		;

	      if (cursor != ceiling_addr)
		{
		  if (++count == 0)
		    {
948
		      immediate_quit = old_immediate_quit;
949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965
		      /* Return the position AFTER the match we found.  */
		      start_byte = start_byte + cursor - base + 1;
		      start = BYTE_TO_CHAR (start_byte);
		      TEMP_SET_PT_BOTH (start, start_byte);
		      return 0;
		    }
		}
	      else
		break;
	    }
	  /* Here we add 1 to compensate for the last decrement
	     of CURSOR, which took it past the valid range.  */
	  start_byte += cursor - base + 1;
	}
    }

  TEMP_SET_PT_BOTH (limit, limit_byte);
966
  immediate_quit = old_immediate_quit;
967 968

  return count * direction;
969 970
}

Jim Blandy's avatar
Jim Blandy committed
971
int
972
find_next_newline_no_quit (from, cnt)
Jim Blandy's avatar
Jim Blandy committed
973 974
     register int from, cnt;
{
975
  return scan_buffer ('\n', from, 0, cnt, (int *) 0, 0);
976 977 978 979 980
}

/* Like find_next_newline, but returns position before the newline,
   not after, and only search up to TO.  This isn't just
   find_next_newline (...)-1, because you might hit TO.  */
981

982 983
int
find_before_next_newline (from, to, cnt)
984
     int from, to, cnt;
985 986 987 988 989 990
{
  int shortage;
  int pos = scan_buffer ('\n', from, to, cnt, &shortage, 1);

  if (shortage == 0)
    pos--;
991

992
  return pos;
Jim Blandy's avatar
Jim Blandy committed
993 994 995 996 997
}

/* Subroutines of Lisp buffer search functions. */

static Lisp_Object
998
search_command (string, bound, noerror, count, direction, RE, posix)
Jim Blandy's avatar
Jim Blandy committed
999 1000 1001
     Lisp_Object string, bound, noerror, count;
     int direction;
     int RE;
1002
     int posix;
Jim Blandy's avatar
Jim Blandy committed
1003 1004
{
  register int np;
1005
  int lim, lim_byte;
Jim Blandy's avatar
Jim Blandy committed
1006 1007 1008 1009
  int n = direction;

  if (!NILP (count))
    {
1010
      CHECK_NUMBER (count);
Jim Blandy's avatar
Jim Blandy committed
1011 1012 1013
      n *= XINT (count);
    }

1014
  CHECK_STRING (string);
Jim Blandy's avatar
Jim Blandy committed
1015
  if (NILP (bound))
1016 1017 1018 1019 1020 1021
    {
      if (n > 0)
	lim = ZV, lim_byte = ZV_BYTE;
      else
	lim = BEGV, lim_byte = BEGV_BYTE;
    }
Jim Blandy's avatar
Jim Blandy committed
1022 1023
  else
    {
1024
      CHECK_NUMBER_COERCE_MARKER (bound);
Jim Blandy's avatar
Jim Blandy committed
1025
      lim = XINT (bound);
1026
      if (n > 0 ? lim < PT : lim > PT)
Jim Blandy's avatar
Jim Blandy committed
1027 1028
	error ("Invalid search bound (wrong side of point)");
      if (lim > ZV)
1029
	lim = ZV, lim_byte = ZV_BYTE;
1030
      else if (lim < BEGV)
1031
	lim = BEGV, lim_byte = BEGV_BYTE;
1032 1033
      else
	lim_byte = CHAR_TO_BYTE (lim);
Jim Blandy's avatar
Jim Blandy committed
1034 1035
    }

1036 1037 1038 1039
  /* This is so set_image_of_range_1 in regex.c can find the EQV table.  */
  XCHAR_TABLE (current_buffer->case_canon_table)->extras[2]
    = current_buffer->case_eqv_table;

1040
  np = search_buffer (string, PT, PT_BYTE, lim, lim_byte, n, RE,
Jim Blandy's avatar
Jim Blandy committed
1041
		      (!NILP (current_buffer->case_fold_search)
1042
		       ? current_buffer->case_canon_table
1043
		       : Qnil),
Jim Blandy's avatar
Jim Blandy committed
1044
		      (!NILP (current_buffer->case_fold_search)
1045
		       ? current_buffer->case_eqv_table