search.c 90.3 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, 2001-2012
  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
#include <setjmp.h>
Jim Blandy's avatar
Jim Blandy committed
24 25
#include "lisp.h"
#include "syntax.h"
Karl Heuer's avatar
Karl Heuer committed
26
#include "category.h"
27
#include "character.h"
28
#include "buffer.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
  /* Nonzero means regexp was compiled to do full POSIX backtracking.  */
  char 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, int);
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
   POSIX is nonzero if we want full backtracking (POSIX style)
Karl Heuer's avatar
Karl Heuer committed
116
   for this pattern.  0 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
compile_pattern_1 (struct regexp_cache *cp, Lisp_Object pattern, Lisp_Object translate, int posix)
Jim Blandy's avatar
Jim Blandy committed
122
{
123
  char *val;
124
  reg_syntax_t old;
Jim Blandy's avatar
Jim Blandy committed
125

126
  cp->regexp = Qnil;
127
  cp->buf.translate = (! NILP (translate) ? translate : make_number (0));
128
  cp->posix = posix;
129
  cp->buf.multibyte = STRING_MULTIBYTE (pattern);
Kenichi Handa's avatar
Kenichi Handa committed
130
  cp->buf.charset_unibyte = charset_unibyte;
131 132 133 134 135
  if (STRINGP (Vsearch_spaces_regexp))
    cp->whitespace_regexp = Vsearch_spaces_regexp;
  else
    cp->whitespace_regexp = Qnil;

136 137 138 139 140
  /* 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;  */
141
  old = re_set_syntax (RE_SYNTAX_EMACS
142
		       | (posix ? 0 : RE_NO_POSIX_BACKTRACKING));
143 144

  if (STRINGP (Vsearch_spaces_regexp))
145
    re_set_whitespace_regexp (SSDATA (Vsearch_spaces_regexp));
146 147
  else
    re_set_whitespace_regexp (NULL);
148

149
  val = (char *) re_compile_pattern (SSDATA (pattern),
Kenichi Handa's avatar
Kenichi Handa committed
150
				     SBYTES (pattern), &cp->buf);
151

152 153
  /* 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
154
  cp->syntax_table = cp->buf.used_syntax ? BVAR (current_buffer, syntax_table) : Qt;
155

156 157
  re_set_whitespace_regexp (NULL);

158
  re_set_syntax (old);
159
  /* UNBLOCK_INPUT;  */
Jim Blandy's avatar
Jim Blandy committed
160
  if (val)
161
    xsignal1 (Qinvalid_regexp, build_string (val));
Jim Blandy's avatar
Jim Blandy committed
162

163 164 165
  cp->regexp = Fcopy_sequence (pattern);
}

166 167 168 169 170
/* Shrink each compiled regexp buffer in the cache
   to the size actually used right now.
   This is called from garbage collection.  */

void
171
shrink_regexp_cache (void)
172
{
173
  struct regexp_cache *cp;
174 175 176 177 178

  for (cp = searchbuf_head; cp != 0; cp = cp->next)
    {
      cp->buf.allocated = cp->buf.used;
      cp->buf.buffer
179
	= (unsigned char *) 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.
205 206 207 208 209 210
   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.  */
211 212

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

  for (cpp = &searchbuf_head; ; cpp = &cp->next)
    {
      cp = *cpp;
220 221
      /* 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
222 223 224
	 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.  */
225
      if (NILP (cp->regexp))
226
	goto compile_it;
227
      if (SCHARS (cp->regexp) == SCHARS (pattern)
228
	  && STRING_MULTIBYTE (cp->regexp) == STRING_MULTIBYTE (pattern)
229
	  && !NILP (Fstring_equal (cp->regexp, pattern))
230
	  && EQ (cp->buf.translate, (! NILP (translate) ? translate : make_number (0)))
Karl Heuer's avatar
Karl Heuer committed
231
	  && cp->posix == posix
232
	  && (EQ (cp->syntax_table, Qt)
Tom Tromey's avatar
Tom Tromey committed
233
	      || EQ (cp->syntax_table, BVAR (current_buffer, syntax_table)))
Kenichi Handa's avatar
Kenichi Handa committed
234 235
	  && !NILP (Fequal (cp->whitespace_regexp, Vsearch_spaces_regexp))
	  && cp->buf.charset_unibyte == charset_unibyte)
236 237
	break;

238 239 240
      /* 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.  */
241 242
      if (cp->next == 0)
	{
243
	compile_it:
244
	  compile_pattern_1 (cp, pattern, translate, posix);
245 246 247 248 249 250 251 252 253 254
	  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
255

256 257 258 259 260
  /* 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
261
  /* The compiled pattern can be used both for multibyte and unibyte
Kenichi Handa's avatar
Kenichi Handa committed
262 263 264
     target.  But, we have to tell which the pattern is used for. */
  cp->buf.target_multibyte = multibyte;

265
  return &cp->buf;
Jim Blandy's avatar
Jim Blandy committed
266 267 268
}


269
static Lisp_Object
270
looking_at_1 (Lisp_Object string, int posix)
Jim Blandy's avatar
Jim Blandy committed
271 272 273
{
  Lisp_Object val;
  unsigned char *p1, *p2;
274 275
  ptrdiff_t s1, s2;
  register ptrdiff_t i;
276
  struct re_pattern_buffer *bufp;
Jim Blandy's avatar
Jim Blandy committed
277

278 279 280
  if (running_asynch_code)
    save_search_regs ();

281
  /* This is so set_image_of_range_1 in regex.c can find the EQV table.  */
Tom Tromey's avatar
Tom Tromey committed
282 283
  XCHAR_TABLE (BVAR (current_buffer, case_canon_table))->extras[2]
    = BVAR (current_buffer, case_eqv_table);
284

285
  CHECK_STRING (string);
286 287 288
  bufp = compile_pattern (string,
			  (NILP (Vinhibit_changing_match_data)
			   ? &search_regs : NULL),
Tom Tromey's avatar
Tom Tromey committed
289 290
			  (!NILP (BVAR (current_buffer, case_fold_search))
			   ? BVAR (current_buffer, case_canon_table) : Qnil),
291
			  posix,
Tom Tromey's avatar
Tom Tromey committed
292
			  !NILP (BVAR (current_buffer, enable_multibyte_characters)));
Jim Blandy's avatar
Jim Blandy committed
293 294 295 296 297 298 299 300

  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;
301
  s1 = GPT_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
302
  p2 = GAP_END_ADDR;
303
  s2 = ZV_BYTE - GPT_BYTE;
Jim Blandy's avatar
Jim Blandy committed
304 305 306
  if (s1 < 0)
    {
      p2 = p1;
307
      s2 = ZV_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
308 309 310 311
      s1 = 0;
    }
  if (s2 < 0)
    {
312
      s1 = ZV_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
313 314
      s2 = 0;
    }
315 316

  re_match_object = Qnil;
317

318
  i = re_match_2 (bufp, (char *) p1, s1, (char *) p2, s2,
319 320 321
		  PT_BYTE - BEGV_BYTE,
		  (NILP (Vinhibit_changing_match_data)
		   ? &search_regs : NULL),
322
		  ZV_BYTE - BEGV_BYTE);
323
  immediate_quit = 0;
324

Jim Blandy's avatar
Jim Blandy committed
325 326 327 328
  if (i == -2)
    matcher_overflow ();

  val = (0 <= i ? Qt : Qnil);
329
  if (NILP (Vinhibit_changing_match_data) && i >= 0)
330 331 332 333 334 335 336 337
    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);
	}
338 339 340 341 342

  /* 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
343 344 345
  return val;
}

346
DEFUN ("looking-at", Flooking_at, Slooking_at, 1, 1, 0,
347 348 349 350
       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
351
  (Lisp_Object regexp)
352
{
353
  return looking_at_1 (regexp, 0);
354 355 356
}

DEFUN ("posix-looking-at", Fposix_looking_at, Sposix_looking_at, 1, 1, 0,
357 358 359 360 361
       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
362
  (Lisp_Object regexp)
363
{
364
  return looking_at_1 (regexp, 1);
365 366 367
}

static Lisp_Object
368
string_match_1 (Lisp_Object regexp, Lisp_Object string, Lisp_Object start, int posix)
Jim Blandy's avatar
Jim Blandy committed
369
{
370
  ptrdiff_t val;
371
  struct re_pattern_buffer *bufp;
372 373
  EMACS_INT pos;
  ptrdiff_t pos_byte, i;
Jim Blandy's avatar
Jim Blandy committed
374

375 376 377
  if (running_asynch_code)
    save_search_regs ();

378 379
  CHECK_STRING (regexp);
  CHECK_STRING (string);
Jim Blandy's avatar
Jim Blandy committed
380 381

  if (NILP (start))
382
    pos = 0, pos_byte = 0;
Jim Blandy's avatar
Jim Blandy committed
383 384
  else
    {
385
      ptrdiff_t len = SCHARS (string);
Jim Blandy's avatar
Jim Blandy committed
386

387
      CHECK_NUMBER (start);
388 389 390 391
      pos = XINT (start);
      if (pos < 0 && -pos <= len)
	pos = len + pos;
      else if (0 > pos || pos > len)
Jim Blandy's avatar
Jim Blandy committed
392
	args_out_of_range (string, start);
393
      pos_byte = string_char_to_byte (string, pos);
Jim Blandy's avatar
Jim Blandy committed
394 395
    }

396
  /* This is so set_image_of_range_1 in regex.c can find the EQV table.  */
Tom Tromey's avatar
Tom Tromey committed
397 398
  XCHAR_TABLE (BVAR (current_buffer, case_canon_table))->extras[2]
    = BVAR (current_buffer, case_eqv_table);
399

400 401 402
  bufp = compile_pattern (regexp,
			  (NILP (Vinhibit_changing_match_data)
			   ? &search_regs : NULL),
Tom Tromey's avatar
Tom Tromey committed
403 404
			  (!NILP (BVAR (current_buffer, case_fold_search))
			   ? BVAR (current_buffer, case_canon_table) : Qnil),
405 406
			  posix,
			  STRING_MULTIBYTE (string));
Jim Blandy's avatar
Jim Blandy committed
407
  immediate_quit = 1;
408
  re_match_object = string;
409

410
  val = re_search (bufp, SSDATA (string),
411 412
		   SBYTES (string), pos_byte,
		   SBYTES (string) - pos_byte,
413 414
		   (NILP (Vinhibit_changing_match_data)
		    ? &search_regs : NULL));
Jim Blandy's avatar
Jim Blandy committed
415
  immediate_quit = 0;
416 417 418 419 420

  /* 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
421 422 423
  if (val == -2)
    matcher_overflow ();
  if (val < 0) return Qnil;
424

425 426 427 428 429 430 431 432 433
  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]);
	}
434 435

  return make_number (string_byte_to_char (string, val));
Jim Blandy's avatar
Jim Blandy committed
436
}
Richard M. Stallman's avatar
Richard M. Stallman committed
437

Paul Eggert's avatar
Paul Eggert committed
438
DEFUN ("string-match", Fstring_match, Sstring_match, 2, 3, 0,
439
       doc: /* Return index of start of first match for REGEXP in STRING, or nil.
Richard M. Stallman's avatar
Richard M. Stallman committed
440
Matching ignores case if `case-fold-search' is non-nil.
441 442 443
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
444 445 446 447
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
448
  (Lisp_Object regexp, Lisp_Object string, Lisp_Object start)
449 450 451 452 453
{
  return string_match_1 (regexp, string, start, 0);
}

DEFUN ("posix-string-match", Fposix_string_match, Sposix_string_match, 2, 3, 0,
454 455 456 457 458 459 460
       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
461
  (Lisp_Object regexp, Lisp_Object string, Lisp_Object start)
462 463 464 465
{
  return string_match_1 (regexp, string, start, 1);
}

Richard M. Stallman's avatar
Richard M. Stallman committed
466 467 468 469
/* 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.  */

470
ptrdiff_t
471
fast_string_match (Lisp_Object regexp, Lisp_Object string)
Richard M. Stallman's avatar
Richard M. Stallman committed
472
{
473
  ptrdiff_t val;
474
  struct re_pattern_buffer *bufp;
Richard M. Stallman's avatar
Richard M. Stallman committed
475

476 477
  bufp = compile_pattern (regexp, 0, Qnil,
			  0, STRING_MULTIBYTE (string));
Richard M. Stallman's avatar
Richard M. Stallman committed
478
  immediate_quit = 1;
479
  re_match_object = string;
480

481
  val = re_search (bufp, SSDATA (string),
482 483
		   SBYTES (string), 0,
		   SBYTES (string), 0);
Richard M. Stallman's avatar
Richard M. Stallman committed
484 485 486
  immediate_quit = 0;
  return val;
}
Karl Heuer's avatar
Karl Heuer committed
487 488 489

/* Match REGEXP against STRING, searching all of STRING ignoring case,
   and return the index of the match, or negative on failure.
490 491
   This does not clobber the match data.
   We assume that STRING contains single-byte characters.  */
Karl Heuer's avatar
Karl Heuer committed
492

493
ptrdiff_t
494
fast_c_string_match_ignore_case (Lisp_Object regexp, const char *string)
Karl Heuer's avatar
Karl Heuer committed
495
{
496
  ptrdiff_t val;
Karl Heuer's avatar
Karl Heuer committed
497
  struct re_pattern_buffer *bufp;
498
  size_t len = strlen (string);
Karl Heuer's avatar
Karl Heuer committed
499

500
  regexp = string_make_unibyte (regexp);
501
  re_match_object = Qt;
Karl Heuer's avatar
Karl Heuer committed
502
  bufp = compile_pattern (regexp, 0,
503
			  Vascii_canon_table, 0,
504
			  0);
Karl Heuer's avatar
Karl Heuer committed
505 506 507 508 509
  immediate_quit = 1;
  val = re_search (bufp, string, len, 0, len, 0);
  immediate_quit = 0;
  return val;
}
510 511 512

/* Like fast_string_match but ignore case.  */

513
ptrdiff_t
514
fast_string_match_ignore_case (Lisp_Object regexp, Lisp_Object string)
515
{
516
  ptrdiff_t val;
517 518
  struct re_pattern_buffer *bufp;

519
  bufp = compile_pattern (regexp, 0, Vascii_canon_table,
520 521 522 523
			  0, STRING_MULTIBYTE (string));
  immediate_quit = 1;
  re_match_object = string;

524
  val = re_search (bufp, SSDATA (string),
525 526 527 528 529
		   SBYTES (string), 0,
		   SBYTES (string), 0);
  immediate_quit = 0;
  return val;
}
530

531
/* Match REGEXP against the characters after POS to LIMIT, and return
532 533 534 535 536
   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.  */

537 538
ptrdiff_t
fast_looking_at (Lisp_Object regexp, ptrdiff_t pos, ptrdiff_t pos_byte, ptrdiff_t limit, ptrdiff_t limit_byte, Lisp_Object string)
539 540 541 542
{
  int multibyte;
  struct re_pattern_buffer *buf;
  unsigned char *p1, *p2;
543 544
  ptrdiff_t s1, s2;
  ptrdiff_t len;
Juanma Barranquero's avatar
Juanma Barranquero committed
545

546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582
  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
583
      multibyte = ! NILP (BVAR (current_buffer, enable_multibyte_characters));
584 585 586 587 588 589 590 591 592 593 594
    }

  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;
}

595 596 597 598 599 600 601 602

/* 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
603
newline_cache_on_off (struct buffer *buf)
604
{
Tom Tromey's avatar
Tom Tromey committed
605
  if (NILP (BVAR (buf, cache_long_line_scans)))
606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631
    {
      /* 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
632 633

   If we find COUNT instances, set *SHORTAGE to zero, and return the
Richard M. Stallman's avatar
Richard M. Stallman committed
634
   position past the COUNTth match.  Note that for reverse motion
Richard M. Stallman's avatar
Richard M. Stallman committed
635
   this is not the same as the usual convention for Emacs motion commands.
Jim Blandy's avatar
Jim Blandy committed
636

637 638
   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
639

640 641 642
   If ALLOW_QUIT is non-zero, set immediate_quit.  That's good to do
   except when inside redisplay.  */

643 644 645
ptrdiff_t
scan_buffer (register int target, ptrdiff_t start, ptrdiff_t end,
	     ptrdiff_t count, ptrdiff_t *shortage, int allow_quit)
Jim Blandy's avatar
Jim Blandy committed
646
{
647
  struct region_cache *newline_cache;
648
  int direction;
Jim Blandy's avatar
Jim Blandy committed
649

650 651 652 653 654 655 656 657 658 659
  if (count > 0)
    {
      direction = 1;
      if (! end) end = ZV;
    }
  else
    {
      direction = -1;
      if (! end) end = BEGV;
    }
Jim Blandy's avatar
Jim Blandy committed
660

661 662
  newline_cache_on_off (current_buffer);
  newline_cache = current_buffer->newline_cache;
Jim Blandy's avatar
Jim Blandy committed
663 664 665 666

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

667
  immediate_quit = allow_quit;
Jim Blandy's avatar
Jim Blandy committed
668

Jim Blandy's avatar
Jim Blandy committed
669
  if (count > 0)
670
    while (start != end)
Jim Blandy's avatar
Jim Blandy committed
671
      {
672 673 674 675 676
        /* 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.  */
677 678 679
	ptrdiff_t ceiling_byte = CHAR_TO_BYTE (end) - 1;
	ptrdiff_t start_byte = CHAR_TO_BYTE (start);
	ptrdiff_t tem;
680 681 682 683 684

        /* If we're looking for a newline, consult the newline cache
           to see where we can avoid some scanning.  */
        if (target == '\n' && newline_cache)
          {
685
            ptrdiff_t next_change;
686 687
            immediate_quit = 0;
            while (region_cache_forward
688 689
                   (current_buffer, newline_cache, start_byte, &next_change))
              start_byte = next_change;
690
            immediate_quit = allow_quit;
691

692 693 694
            /* START should never be after END.  */
            if (start_byte > ceiling_byte)
              start_byte = ceiling_byte;
695 696 697

            /* Now the text after start is an unknown region, and
               next_change is the position of the next known region. */
698
            ceiling_byte = min (next_change - 1, ceiling_byte);
699 700 701 702 703 704
          }

        /* 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.  */
705 706
	tem = BUFFER_CEILING_OF (start_byte);
	ceiling_byte = min (tem, ceiling_byte);
707 708

        {
709
          /* The termination address of the dumb loop.  */
710 711 712 713
          register unsigned char *ceiling_addr
	    = BYTE_POS_ADDR (ceiling_byte) + 1;
          register unsigned char *cursor
	    = BYTE_POS_ADDR (start_byte);
714 715 716 717 718 719 720 721 722 723 724 725 726 727
          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,
728 729
                                   BYTE_TO_CHAR (start_byte + scan_start - base),
                                   BYTE_TO_CHAR (start_byte + cursor - base));
730 731 732 733 734 735 736

              /* Did we find the target character?  */
              if (cursor < ceiling_addr)
                {
                  if (--count == 0)
                    {
                      immediate_quit = 0;
737
                      return BYTE_TO_CHAR (start_byte + cursor - base + 1);
738 739 740 741 742
                    }
                  cursor++;
                }
            }

743
          start = BYTE_TO_CHAR (start_byte + cursor - base);
744
        }
Jim Blandy's avatar
Jim Blandy committed
745 746
      }
  else
747 748 749
    while (start > end)
      {
        /* The last character to check before the next obstacle.  */
750 751 752
	ptrdiff_t ceiling_byte = CHAR_TO_BYTE (end);
	ptrdiff_t start_byte = CHAR_TO_BYTE (start);
	ptrdiff_t tem;
753 754 755 756

        /* Consult the newline cache, if appropriate.  */
        if (target == '\n' && newline_cache)
          {
757
            ptrdiff_t next_change;
758 759
            immediate_quit = 0;
            while (region_cache_backward
760 761
                   (current_buffer, newline_cache, start_byte, &next_change))
              start_byte = next_change;
762
            immediate_quit = allow_quit;
763 764

            /* Start should never be at or before end.  */
765 766
            if (start_byte <= ceiling_byte)
              start_byte = ceiling_byte + 1;
767 768 769

            /* Now the text before start is an unknown region, and
               next_change is the position of the next known region. */
770
            ceiling_byte = max (next_change, ceiling_byte);
771 772 773
          }

        /* Stop scanning before the gap.  */
774 775
	tem = BUFFER_FLOOR_OF (start_byte - 1);
	ceiling_byte = max (tem, ceiling_byte);
776 777 778

        {
          /* The termination address of the dumb loop.  */
779 780
          register unsigned char *ceiling_addr = BYTE_POS_ADDR (ceiling_byte);
          register unsigned char *cursor = BYTE_POS_ADDR (start_byte - 1);
781 782 783 784 785 786 787 788 789 790 791 792 793
          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,
794 795
                                   BYTE_TO_CHAR (start_byte + cursor - base),
                                   BYTE_TO_CHAR (start_byte + scan_start - base));
796 797 798 799 800 801 802

              /* Did we find the target character?  */
              if (cursor >= ceiling_addr)
                {
                  if (++count >= 0)
                    {
                      immediate_quit = 0;
803
                      return BYTE_TO_CHAR (start_byte + cursor - base);
804 805 806 807 808
                    }
                  cursor--;
                }
            }

809
	  start = BYTE_TO_CHAR (start_byte + cursor - base);
810 811 812
        }
      }

Jim Blandy's avatar
Jim Blandy committed
813 814
  immediate_quit = 0;
  if (shortage != 0)
Jim Blandy's avatar
Jim Blandy committed
815
    *shortage = count * direction;
816
  return start;
Jim Blandy's avatar
Jim Blandy committed
817
}
818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833

/* 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
834
   except in special cases.  */
Jim Blandy's avatar
Jim Blandy committed
835

836
EMACS_INT
837 838
scan_newline (ptrdiff_t start, ptrdiff_t start_byte,
	      ptrdiff_t limit, ptrdiff_t limit_byte,
839
	      register EMACS_INT count, int allow_quit)
840
{
841 842 843 844 845
  int direction = ((count > 0) ? 1 : -1);

  register unsigned char *cursor;
  unsigned char *base;

846
  ptrdiff_t ceiling;
847 848
  register unsigned char *ceiling_addr;

849 850
  int old_immediate_quit = immediate_quit;

851 852 853
  /* The code that follows is like scan_buffer
     but checks for either newline or carriage return.  */

854 855
  if (allow_quit)
    immediate_quit++;
856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875

  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)
		    {
876
		      immediate_quit = old_immediate_quit;
877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908
		      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)
		    {
909
		      immediate_quit = old_immediate_quit;
910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926
		      /* 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);
927
  immediate_quit = old_immediate_quit;
928 929

  return count * direction;
930 931
}

932 933
ptrdiff_t
find_next_newline_no_quit (ptrdiff_t from, ptrdiff_t cnt)
Jim Blandy's avatar
Jim Blandy committed
934
{
935
  return scan_buffer ('\n', from, 0, cnt, (ptrdiff_t *) 0, 0);
936 937 938 939 940
}

/* 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.  */
941

942 943
ptrdiff_t
find_before_next_newline (ptrdiff_t from, ptrdiff_t to, ptrdiff_t cnt)
944
{
945 946
  ptrdiff_t shortage;
  ptrdiff_t pos = scan_buffer ('\n', from, to, cnt, &shortage, 1);
947 948 949

  if (shortage == 0)
    pos--;
950

951
  return pos;
Jim Blandy's avatar
Jim Blandy committed
952 953 954 955 956
}

/* Subroutines of Lisp buffer search functions. */

static Lisp_Object
957 958
search_command (Lisp_Object string, Lisp_Object bound, Lisp_Object noerror,
		Lisp_Object count, int direction, int RE, int posix)
Jim Blandy's avatar
Jim Blandy committed
959
{
960
  register EMACS_INT np;
961 962
  EMACS_INT lim;
  ptrdiff_t lim_byte;
963
  EMACS_INT n = direction;
Jim Blandy's avatar
Jim Blandy committed
964 965 966

  if (!NILP (count))
    {
967
      CHECK_NUMBER (count);
Jim Blandy's avatar
Jim Blandy committed
968 969 970
      n *= XINT (count);
    }

971
  CHECK_STRING (string);
Jim Blandy's avatar
Jim Blandy committed
972
  if (NILP (bound))
973 974 975 976 977 978
    {
      if (n > 0)
	lim = ZV, lim_byte = ZV_BYTE;
      else
	lim = BEGV, lim_byte = BEGV_BYTE;
    }
Jim Blandy's avatar
Jim Blandy committed
979 980
  else
    {
981
      CHECK_NUMBER_COERCE_MARKER (bound);
Jim Blandy's avatar
Jim Blandy committed
982
      lim = XINT (bound);
983
      if (n > 0 ? lim < PT : lim > PT)
Jim Blandy's avatar
Jim Blandy committed
984 985
	error ("Invalid search bound (wrong side of point)");
      if (lim > ZV)
986
	lim = ZV, lim_byte = ZV_BYTE;
987
      else if (lim < BEGV)
988
	lim = BEGV, lim_byte = BEGV_BYTE;
989 990
      else
	lim_byte = CHAR_TO_BYTE (lim);
Jim Blandy's avatar
Jim Blandy committed
991 992
    }