search.c 90.5 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

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

  for (cp = searchbuf_head; cp != 0; cp = cp->next)
    {
      cp->buf.allocated = cp->buf.used;
178
      cp->buf.buffer = xrealloc (cp->buf.buffer, cp->buf.used);
179 180 181
    }
}

182 183
/* Clear the regexp cache w.r.t. a particular syntax table,
   because it was changed.
184 185 186 187
   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
188
clear_regexp_cache (void)
189 190 191 192
{
  int i;

  for (i = 0; i < REGEXP_CACHE_SIZE; ++i)
193 194
    /* It's tempting to compare with the syntax-table we've actually changed,
       but it's not sufficient because char-table inheritance means that
195 196 197
       modifying one syntax-table can change others at the same time.  */
    if (!EQ (searchbufs[i].syntax_table, Qt))
      searchbufs[i].regexp = Qnil;
198 199
}

200
/* Compile a regexp if necessary, but first check to see if there's one in
201 202
   the cache.
   PATTERN is the pattern to compile.
203
   TRANSLATE is a translation table for ignoring case, or nil for none.
204 205 206 207 208 209
   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.  */
210 211

struct re_pattern_buffer *
212
compile_pattern (Lisp_Object pattern, struct re_registers *regp, Lisp_Object translate, int posix, int multibyte)
213 214 215 216 217 218
{
  struct regexp_cache *cp, **cpp;

  for (cpp = &searchbuf_head; ; cpp = &cp->next)
    {
      cp = *cpp;
219 220
      /* 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
221 222 223
	 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.  */
224
      if (NILP (cp->regexp))
225
	goto compile_it;
226
      if (SCHARS (cp->regexp) == SCHARS (pattern)
227
	  && STRING_MULTIBYTE (cp->regexp) == STRING_MULTIBYTE (pattern)
228
	  && !NILP (Fstring_equal (cp->regexp, pattern))
229
	  && EQ (cp->buf.translate, (! NILP (translate) ? translate : make_number (0)))
Karl Heuer's avatar
Karl Heuer committed
230
	  && cp->posix == posix
231
	  && (EQ (cp->syntax_table, Qt)
Tom Tromey's avatar
Tom Tromey committed
232
	      || EQ (cp->syntax_table, BVAR (current_buffer, syntax_table)))
Kenichi Handa's avatar
Kenichi Handa committed
233 234
	  && !NILP (Fequal (cp->whitespace_regexp, Vsearch_spaces_regexp))
	  && cp->buf.charset_unibyte == charset_unibyte)
235 236
	break;

237 238 239
      /* 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.  */
240 241
      if (cp->next == 0)
	{
242
	compile_it:
243
	  compile_pattern_1 (cp, pattern, translate, posix);
244 245 246 247 248 249 250 251 252 253
	  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
254

255 256 257 258 259
  /* 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
260
  /* The compiled pattern can be used both for multibyte and unibyte
Kenichi Handa's avatar
Kenichi Handa committed
261 262 263
     target.  But, we have to tell which the pattern is used for. */
  cp->buf.target_multibyte = multibyte;

264
  return &cp->buf;
Jim Blandy's avatar
Jim Blandy committed
265 266 267
}


268
static Lisp_Object
269
looking_at_1 (Lisp_Object string, int posix)
Jim Blandy's avatar
Jim Blandy committed
270 271 272
{
  Lisp_Object val;
  unsigned char *p1, *p2;
273 274
  ptrdiff_t s1, s2;
  register ptrdiff_t i;
275
  struct re_pattern_buffer *bufp;
Jim Blandy's avatar
Jim Blandy committed
276

277 278 279
  if (running_asynch_code)
    save_search_regs ();

280
  /* This is so set_image_of_range_1 in regex.c can find the EQV table.  */
281
  set_char_table_extras (BVAR (current_buffer, case_canon_table), 2,
282
			 BVAR (current_buffer, case_eqv_table));
283

284
  CHECK_STRING (string);
285 286 287
  bufp = compile_pattern (string,
			  (NILP (Vinhibit_changing_match_data)
			   ? &search_regs : NULL),
Tom Tromey's avatar
Tom Tromey committed
288 289
			  (!NILP (BVAR (current_buffer, case_fold_search))
			   ? BVAR (current_buffer, case_canon_table) : Qnil),
290
			  posix,
Tom Tromey's avatar
Tom Tromey committed
291
			  !NILP (BVAR (current_buffer, enable_multibyte_characters)));
Jim Blandy's avatar
Jim Blandy committed
292 293 294 295 296 297 298 299

  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;
300
  s1 = GPT_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
301
  p2 = GAP_END_ADDR;
302
  s2 = ZV_BYTE - GPT_BYTE;
Jim Blandy's avatar
Jim Blandy committed
303 304 305
  if (s1 < 0)
    {
      p2 = p1;
306
      s2 = ZV_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
307 308 309 310
      s1 = 0;
    }
  if (s2 < 0)
    {
311
      s1 = ZV_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
312 313
      s2 = 0;
    }
314 315

  re_match_object = Qnil;
316

317
  i = re_match_2 (bufp, (char *) p1, s1, (char *) p2, s2,
318 319 320
		  PT_BYTE - BEGV_BYTE,
		  (NILP (Vinhibit_changing_match_data)
		   ? &search_regs : NULL),
321
		  ZV_BYTE - BEGV_BYTE);
322
  immediate_quit = 0;
323

Jim Blandy's avatar
Jim Blandy committed
324 325 326 327
  if (i == -2)
    matcher_overflow ();

  val = (0 <= i ? Qt : Qnil);
328
  if (NILP (Vinhibit_changing_match_data) && i >= 0)
329 330 331 332 333 334 335 336
    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);
	  search_regs.end[i]
	    = BYTE_TO_CHAR (search_regs.end[i] + BEGV_BYTE);
	}
337 338 339 340 341

  /* 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
342 343 344
  return val;
}

345
DEFUN ("looking-at", Flooking_at, Slooking_at, 1, 1, 0,
346 347 348 349
       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
350
  (Lisp_Object regexp)
351
{
352
  return looking_at_1 (regexp, 0);
353 354 355
}

DEFUN ("posix-looking-at", Fposix_looking_at, Sposix_looking_at, 1, 1, 0,
356 357 358 359 360
       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
361
  (Lisp_Object regexp)
362
{
363
  return looking_at_1 (regexp, 1);
364 365 366
}

static Lisp_Object
367
string_match_1 (Lisp_Object regexp, Lisp_Object string, Lisp_Object start, int posix)
Jim Blandy's avatar
Jim Blandy committed
368
{
369
  ptrdiff_t val;
370
  struct re_pattern_buffer *bufp;
371 372
  EMACS_INT pos;
  ptrdiff_t pos_byte, i;
Jim Blandy's avatar
Jim Blandy committed
373

374 375 376
  if (running_asynch_code)
    save_search_regs ();

377 378
  CHECK_STRING (regexp);
  CHECK_STRING (string);
Jim Blandy's avatar
Jim Blandy committed
379 380

  if (NILP (start))
381
    pos = 0, pos_byte = 0;
Jim Blandy's avatar
Jim Blandy committed
382 383
  else
    {
384
      ptrdiff_t len = SCHARS (string);
Jim Blandy's avatar
Jim Blandy committed
385

386
      CHECK_NUMBER (start);
387 388 389 390
      pos = XINT (start);
      if (pos < 0 && -pos <= len)
	pos = len + pos;
      else if (0 > pos || pos > len)
Jim Blandy's avatar
Jim Blandy committed
391
	args_out_of_range (string, start);
392
      pos_byte = string_char_to_byte (string, pos);
Jim Blandy's avatar
Jim Blandy committed
393 394
    }

395
  /* This is so set_image_of_range_1 in regex.c can find the EQV table.  */
396
  set_char_table_extras (BVAR (current_buffer, case_canon_table), 2,
397
			 BVAR (current_buffer, case_eqv_table));
398

399 400 401
  bufp = compile_pattern (regexp,
			  (NILP (Vinhibit_changing_match_data)
			   ? &search_regs : NULL),
Tom Tromey's avatar
Tom Tromey committed
402 403
			  (!NILP (BVAR (current_buffer, case_fold_search))
			   ? BVAR (current_buffer, case_canon_table) : Qnil),
404 405
			  posix,
			  STRING_MULTIBYTE (string));
Jim Blandy's avatar
Jim Blandy committed
406
  immediate_quit = 1;
407
  re_match_object = string;
408

409
  val = re_search (bufp, SSDATA (string),
410 411
		   SBYTES (string), pos_byte,
		   SBYTES (string) - pos_byte,
412 413
		   (NILP (Vinhibit_changing_match_data)
		    ? &search_regs : NULL));
Jim Blandy's avatar
Jim Blandy committed
414
  immediate_quit = 0;
415 416 417 418 419

  /* 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
420 421 422
  if (val == -2)
    matcher_overflow ();
  if (val < 0) return Qnil;
423

424 425 426 427 428 429 430 431 432
  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]);
	}
433 434

  return make_number (string_byte_to_char (string, val));
Jim Blandy's avatar
Jim Blandy committed
435
}
Richard M. Stallman's avatar
Richard M. Stallman committed
436

Paul Eggert's avatar
Paul Eggert committed
437
DEFUN ("string-match", Fstring_match, Sstring_match, 2, 3, 0,
438
       doc: /* Return index of start of first match for REGEXP in STRING, or nil.
Richard M. Stallman's avatar
Richard M. Stallman committed
439
Matching ignores case if `case-fold-search' is non-nil.
440 441 442
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
443 444 445 446
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
447
  (Lisp_Object regexp, Lisp_Object string, Lisp_Object start)
448 449 450 451 452
{
  return string_match_1 (regexp, string, start, 0);
}

DEFUN ("posix-string-match", Fposix_string_match, Sposix_string_match, 2, 3, 0,
453 454 455 456 457 458 459
       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
460
  (Lisp_Object regexp, Lisp_Object string, Lisp_Object start)
461 462 463 464
{
  return string_match_1 (regexp, string, start, 1);
}

Richard M. Stallman's avatar
Richard M. Stallman committed
465 466 467 468
/* 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.  */

469
ptrdiff_t
470
fast_string_match (Lisp_Object regexp, Lisp_Object string)
Richard M. Stallman's avatar
Richard M. Stallman committed
471
{
472
  ptrdiff_t val;
473
  struct re_pattern_buffer *bufp;
Richard M. Stallman's avatar
Richard M. Stallman committed
474

475 476
  bufp = compile_pattern (regexp, 0, Qnil,
			  0, STRING_MULTIBYTE (string));
Richard M. Stallman's avatar
Richard M. Stallman committed
477
  immediate_quit = 1;
478
  re_match_object = string;
479

480
  val = re_search (bufp, SSDATA (string),
481 482
		   SBYTES (string), 0,
		   SBYTES (string), 0);
Richard M. Stallman's avatar
Richard M. Stallman committed
483 484 485
  immediate_quit = 0;
  return val;
}
Karl Heuer's avatar
Karl Heuer committed
486 487 488

/* Match REGEXP against STRING, searching all of STRING ignoring case,
   and return the index of the match, or negative on failure.
489 490
   This does not clobber the match data.
   We assume that STRING contains single-byte characters.  */
Karl Heuer's avatar
Karl Heuer committed
491

492
ptrdiff_t
493 494
fast_c_string_match_ignore_case (Lisp_Object regexp,
				 const char *string, ptrdiff_t len)
Karl Heuer's avatar
Karl Heuer committed
495
{
496
  ptrdiff_t val;
Karl Heuer's avatar
Karl Heuer committed
497 498
  struct re_pattern_buffer *bufp;

499
  regexp = string_make_unibyte (regexp);
500
  re_match_object = Qt;
Karl Heuer's avatar
Karl Heuer committed
501
  bufp = compile_pattern (regexp, 0,
502
			  Vascii_canon_table, 0,
503
			  0);
Karl Heuer's avatar
Karl Heuer committed
504 505 506 507 508
  immediate_quit = 1;
  val = re_search (bufp, string, len, 0, len, 0);
  immediate_quit = 0;
  return val;
}
509 510 511

/* Like fast_string_match but ignore case.  */

512
ptrdiff_t
513
fast_string_match_ignore_case (Lisp_Object regexp, Lisp_Object string)
514
{
515
  ptrdiff_t val;
516 517
  struct re_pattern_buffer *bufp;

518
  bufp = compile_pattern (regexp, 0, Vascii_canon_table,
519 520 521 522
			  0, STRING_MULTIBYTE (string));
  immediate_quit = 1;
  re_match_object = string;

523
  val = re_search (bufp, SSDATA (string),
524 525 526 527 528
		   SBYTES (string), 0,
		   SBYTES (string), 0);
  immediate_quit = 0;
  return val;
}
529

530
/* Match REGEXP against the characters after POS to LIMIT, and return
531 532 533 534 535
   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.  */

536 537
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)
538 539 540 541
{
  int multibyte;
  struct re_pattern_buffer *buf;
  unsigned char *p1, *p2;
542 543
  ptrdiff_t s1, s2;
  ptrdiff_t len;
Juanma Barranquero's avatar
Juanma Barranquero committed
544

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
  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
582
      multibyte = ! NILP (BVAR (current_buffer, enable_multibyte_characters));
583 584 585 586 587 588 589 590 591 592 593
    }

  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;
}

594 595 596 597 598 599 600 601

/* 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
602
newline_cache_on_off (struct buffer *buf)
603
{
Tom Tromey's avatar
Tom Tromey committed
604
  if (NILP (BVAR (buf, cache_long_line_scans)))
605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630
    {
      /* 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
631 632

   If we find COUNT instances, set *SHORTAGE to zero, and return the
Richard M. Stallman's avatar
Richard M. Stallman committed
633
   position past the COUNTth match.  Note that for reverse motion
Richard M. Stallman's avatar
Richard M. Stallman committed
634
   this is not the same as the usual convention for Emacs motion commands.
Jim Blandy's avatar
Jim Blandy committed
635

636 637
   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
638

639 640 641
   If ALLOW_QUIT is non-zero, set immediate_quit.  That's good to do
   except when inside redisplay.  */

642 643 644
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
645
{
646
  struct region_cache *newline_cache;
647
  int direction;
Jim Blandy's avatar
Jim Blandy committed
648

649 650 651 652 653 654 655 656 657 658
  if (count > 0)
    {
      direction = 1;
      if (! end) end = ZV;
    }
  else
    {
      direction = -1;
      if (! end) end = BEGV;
    }
Jim Blandy's avatar
Jim Blandy committed
659

660 661
  newline_cache_on_off (current_buffer);
  newline_cache = current_buffer->newline_cache;
Jim Blandy's avatar
Jim Blandy committed
662 663 664 665

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

666
  immediate_quit = allow_quit;
Jim Blandy's avatar
Jim Blandy committed
667

Jim Blandy's avatar
Jim Blandy committed
668
  if (count > 0)
669
    while (start != end)
Jim Blandy's avatar
Jim Blandy committed
670
      {
671 672 673 674 675
        /* 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.  */
676
	ptrdiff_t ceiling_byte = CHAR_TO_BYTE (end) - 1;
677
	ptrdiff_t start_byte;
678
	ptrdiff_t tem;
679 680 681 682 683

        /* If we're looking for a newline, consult the newline cache
           to see where we can avoid some scanning.  */
        if (target == '\n' && newline_cache)
          {
684
            ptrdiff_t next_change;
685 686
            immediate_quit = 0;
            while (region_cache_forward
687 688
                   (current_buffer, newline_cache, start, &next_change))
              start = next_change;
689
            immediate_quit = allow_quit;
690

691 692
	    start_byte = CHAR_TO_BYTE (start);

693 694 695
            /* START should never be after END.  */
            if (start_byte > ceiling_byte)
              start_byte = ceiling_byte;
696 697 698

            /* Now the text after start is an unknown region, and
               next_change is the position of the next known region. */
699
            ceiling_byte = min (CHAR_TO_BYTE (next_change) - 1, ceiling_byte);
700
          }
701 702
	else
	  start_byte = CHAR_TO_BYTE (start);
703 704 705 706 707

        /* 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.  */
708 709
	tem = BUFFER_CEILING_OF (start_byte);
	ceiling_byte = min (tem, ceiling_byte);
710 711

        {
712
          /* The termination address of the dumb loop.  */
713 714 715 716
          register unsigned char *ceiling_addr
	    = BYTE_POS_ADDR (ceiling_byte) + 1;
          register unsigned char *cursor
	    = BYTE_POS_ADDR (start_byte);
717 718 719 720 721 722 723 724 725 726 727 728 729 730
          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,
731 732
                                   BYTE_TO_CHAR (start_byte + scan_start - base),
                                   BYTE_TO_CHAR (start_byte + cursor - base));
733 734 735 736 737 738 739

              /* Did we find the target character?  */
              if (cursor < ceiling_addr)
                {
                  if (--count == 0)
                    {
                      immediate_quit = 0;
740
                      return BYTE_TO_CHAR (start_byte + cursor - base + 1);
741 742 743 744 745
                    }
                  cursor++;
                }
            }

746
          start = BYTE_TO_CHAR (start_byte + cursor - base);
747
        }
Jim Blandy's avatar
Jim Blandy committed
748 749
      }
  else
750 751 752
    while (start > end)
      {
        /* The last character to check before the next obstacle.  */
753
	ptrdiff_t ceiling_byte = CHAR_TO_BYTE (end);
754
	ptrdiff_t start_byte;
755
	ptrdiff_t tem;
756 757 758 759

        /* Consult the newline cache, if appropriate.  */
        if (target == '\n' && newline_cache)
          {
760
            ptrdiff_t next_change;
761 762
            immediate_quit = 0;
            while (region_cache_backward
763 764
                   (current_buffer, newline_cache, start, &next_change))
              start = next_change;
765
            immediate_quit = allow_quit;
766

767 768
	    start_byte = CHAR_TO_BYTE (start);

769
            /* Start should never be at or before end.  */
770 771
            if (start_byte <= ceiling_byte)
              start_byte = ceiling_byte + 1;
772 773 774

            /* Now the text before start is an unknown region, and
               next_change is the position of the next known region. */
775
            ceiling_byte = max (CHAR_TO_BYTE (next_change), ceiling_byte);
776
          }
777 778
	else
	  start_byte = CHAR_TO_BYTE (start);
779 780

        /* Stop scanning before the gap.  */
781 782
	tem = BUFFER_FLOOR_OF (start_byte - 1);
	ceiling_byte = max (tem, ceiling_byte);
783 784 785

        {
          /* The termination address of the dumb loop.  */
786 787
          register unsigned char *ceiling_addr = BYTE_POS_ADDR (ceiling_byte);
          register unsigned char *cursor = BYTE_POS_ADDR (start_byte - 1);
788 789 790 791 792 793 794 795 796 797 798 799 800
          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,
801 802
                                   BYTE_TO_CHAR (start_byte + cursor - base),
                                   BYTE_TO_CHAR (start_byte + scan_start - base));
803 804 805 806 807 808 809

              /* Did we find the target character?  */
              if (cursor >= ceiling_addr)
                {
                  if (++count >= 0)
                    {
                      immediate_quit = 0;
810
                      return BYTE_TO_CHAR (start_byte + cursor - base);
811 812 813 814 815
                    }
                  cursor--;
                }
            }

816
	  start = BYTE_TO_CHAR (start_byte + cursor - base);
817 818 819
        }
      }

Jim Blandy's avatar
Jim Blandy committed
820 821
  immediate_quit = 0;
  if (shortage != 0)
Jim Blandy's avatar
Jim Blandy committed
822
    *shortage = count * direction;
823
  return start;
Jim Blandy's avatar
Jim Blandy committed
824
}
825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840

/* 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
841
   except in special cases.  */
Jim Blandy's avatar
Jim Blandy committed
842

843
EMACS_INT
844 845
scan_newline (ptrdiff_t start, ptrdiff_t start_byte,
	      ptrdiff_t limit, ptrdiff_t limit_byte,
846
	      register EMACS_INT count, int allow_quit)
847
{
848 849 850 851 852
  int direction = ((count > 0) ? 1 : -1);

  register unsigned char *cursor;
  unsigned char *base;

853
  ptrdiff_t ceiling;
854 855
  register unsigned char *ceiling_addr;

856 857
  int old_immediate_quit = immediate_quit;

858 859 860
  /* The code that follows is like scan_buffer
     but checks for either newline or carriage return.  */

861 862
  if (allow_quit)
    immediate_quit++;
863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882

  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)
		    {
883
		      immediate_quit = old_immediate_quit;
884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915
		      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)
		    {
916
		      immediate_quit = old_immediate_quit;
917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933
		      /* 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);
934
  immediate_quit = old_immediate_quit;
935 936

  return count * direction;
937 938
}

939 940
ptrdiff_t
find_next_newline_no_quit (ptrdiff_t from, ptrdiff_t cnt)
Jim Blandy's avatar
Jim Blandy committed
941
{
942
  return scan_buffer ('\n', from, 0, cnt, (ptrdiff_t *) 0, 0);
943 944 945 946 947
}

/* 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.  */
948

949 950
ptrdiff_t
find_before_next_newline (ptrdiff_t from, ptrdiff_t to, ptrdiff_t cnt)
951
{
952 953
  ptrdiff_t shortage;
  ptrdiff_t pos = scan_buffer ('\n', from, to, cnt, &shortage, 1);
954 955 956

  if (shortage == 0)
    pos--;
Juanma Barranquero's avatar
<