search.c 95.1 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,
3
                 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011
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>
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"
Jim Blandy's avatar
Jim Blandy committed
27
#include "buffer.h"
28
#include "character.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 57
/* The instances of that struct.  */
struct regexp_cache searchbufs[REGEXP_CACHE_SIZE];
Jim Blandy's avatar
Jim Blandy committed
58

59 60
/* 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
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

Karl Heuer's avatar
Karl Heuer committed
87
/* error condition signaled when regexp compile_pattern fails */
Jim Blandy's avatar
Jim Blandy committed
88 89 90

Lisp_Object Qinvalid_regexp;

91 92 93
/* Error condition used for failing searches */
Lisp_Object Qsearch_failed;

94 95
static void set_search_regs (EMACS_INT, EMACS_INT);
static void save_search_regs (void);
96 97
static EMACS_INT simple_search (EMACS_INT, unsigned char *, EMACS_INT,
				EMACS_INT, Lisp_Object, EMACS_INT, EMACS_INT,
98
                                EMACS_INT, EMACS_INT);
99
static EMACS_INT boyer_moore (EMACS_INT, unsigned char *, EMACS_INT, EMACS_INT,
100 101 102 103
                              Lisp_Object, Lisp_Object,
                              EMACS_INT, EMACS_INT,
                              EMACS_INT, EMACS_INT, int);
static EMACS_INT search_buffer (Lisp_Object, EMACS_INT, EMACS_INT,
104
                                EMACS_INT, EMACS_INT, EMACS_INT, int,
105
                                Lisp_Object, Lisp_Object, int);
106
static void matcher_overflow (void) NO_RETURN;
107

Jim Blandy's avatar
Jim Blandy committed
108
static void
109
matcher_overflow (void)
Jim Blandy's avatar
Jim Blandy committed
110 111 112 113
{
  error ("Stack overflow in regexp matcher");
}

114 115 116
/* 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.
117
   TRANSLATE is a translation table for ignoring case, or nil for none.
118 119 120 121 122
   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
123
   for this pattern.  0 means backtrack only enough to get a valid match.
124 125

   The behavior also depends on Vsearch_spaces_regexp.  */
Jim Blandy's avatar
Jim Blandy committed
126

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

133
  cp->regexp = Qnil;
134
  cp->buf.translate = (! NILP (translate) ? translate : make_number (0));
135
  cp->posix = posix;
136
  cp->buf.multibyte = STRING_MULTIBYTE (pattern);
Kenichi Handa's avatar
Kenichi Handa committed
137
  cp->buf.charset_unibyte = charset_unibyte;
138 139 140 141 142
  if (STRINGP (Vsearch_spaces_regexp))
    cp->whitespace_regexp = Vsearch_spaces_regexp;
  else
    cp->whitespace_regexp = Qnil;

143 144 145 146 147
  /* 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;  */
148
  old = re_set_syntax (RE_SYNTAX_EMACS
149
		       | (posix ? 0 : RE_NO_POSIX_BACKTRACKING));
150 151 152 153 154

  if (STRINGP (Vsearch_spaces_regexp))
    re_set_whitespace_regexp (SDATA (Vsearch_spaces_regexp));
  else
    re_set_whitespace_regexp (NULL);
155

156
  val = (char *) re_compile_pattern (SSDATA (pattern),
Kenichi Handa's avatar
Kenichi Handa committed
157
				     SBYTES (pattern), &cp->buf);
158

159 160 161 162
  /* 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;

163 164
  re_set_whitespace_regexp (NULL);

165
  re_set_syntax (old);
166
  /* UNBLOCK_INPUT;  */
Jim Blandy's avatar
Jim Blandy committed
167
  if (val)
168
    xsignal1 (Qinvalid_regexp, build_string (val));
Jim Blandy's avatar
Jim Blandy committed
169

170 171 172
  cp->regexp = Fcopy_sequence (pattern);
}

173 174 175 176 177
/* Shrink each compiled regexp buffer in the cache
   to the size actually used right now.
   This is called from garbage collection.  */

void
178
shrink_regexp_cache (void)
179
{
180
  struct regexp_cache *cp;
181 182 183 184 185

  for (cp = searchbuf_head; cp != 0; cp = cp->next)
    {
      cp->buf.allocated = cp->buf.used;
      cp->buf.buffer
186
	= (unsigned char *) xrealloc (cp->buf.buffer, cp->buf.used);
187 188 189
    }
}

190 191
/* Clear the regexp cache w.r.t. a particular syntax table,
   because it was changed.
192 193 194 195
   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
196
clear_regexp_cache (void)
197 198 199 200
{
  int i;

  for (i = 0; i < REGEXP_CACHE_SIZE; ++i)
201 202
    /* It's tempting to compare with the syntax-table we've actually changed,
       but it's not sufficient because char-table inheritance means that
203 204 205
       modifying one syntax-table can change others at the same time.  */
    if (!EQ (searchbufs[i].syntax_table, Qt))
      searchbufs[i].regexp = Qnil;
206 207
}

208
/* Compile a regexp if necessary, but first check to see if there's one in
209 210
   the cache.
   PATTERN is the pattern to compile.
211
   TRANSLATE is a translation table for ignoring case, or nil for none.
212 213 214 215 216 217
   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.  */
218 219

struct re_pattern_buffer *
220
compile_pattern (Lisp_Object pattern, struct re_registers *regp, Lisp_Object translate, int posix, int multibyte)
221 222 223 224 225 226
{
  struct regexp_cache *cp, **cpp;

  for (cpp = &searchbuf_head; ; cpp = &cp->next)
    {
      cp = *cpp;
227 228
      /* 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
229 230 231
	 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.  */
232
      if (NILP (cp->regexp))
233
	goto compile_it;
234
      if (SCHARS (cp->regexp) == SCHARS (pattern)
235
	  && STRING_MULTIBYTE (cp->regexp) == STRING_MULTIBYTE (pattern)
236
	  && !NILP (Fstring_equal (cp->regexp, pattern))
237
	  && EQ (cp->buf.translate, (! NILP (translate) ? translate : make_number (0)))
Karl Heuer's avatar
Karl Heuer committed
238
	  && cp->posix == posix
239 240
	  && (EQ (cp->syntax_table, Qt)
	      || EQ (cp->syntax_table, current_buffer->syntax_table))
Kenichi Handa's avatar
Kenichi Handa committed
241 242
	  && !NILP (Fequal (cp->whitespace_regexp, Vsearch_spaces_regexp))
	  && cp->buf.charset_unibyte == charset_unibyte)
243 244
	break;

245 246 247
      /* 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.  */
248 249
      if (cp->next == 0)
	{
250
	compile_it:
Kenichi Handa's avatar
Kenichi Handa committed
251
	  compile_pattern_1 (cp, pattern, translate, regp, posix);
252 253 254 255 256 257 258 259 260 261
	  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
262

263 264 265 266 267
  /* 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
268
  /* The compiled pattern can be used both for multibyte and unibyte
Kenichi Handa's avatar
Kenichi Handa committed
269 270 271
     target.  But, we have to tell which the pattern is used for. */
  cp->buf.target_multibyte = multibyte;

272
  return &cp->buf;
Jim Blandy's avatar
Jim Blandy committed
273 274 275
}


276
static Lisp_Object
277
looking_at_1 (Lisp_Object string, int posix)
Jim Blandy's avatar
Jim Blandy committed
278 279 280
{
  Lisp_Object val;
  unsigned char *p1, *p2;
281
  EMACS_INT s1, s2;
282
  register EMACS_INT i;
283
  struct re_pattern_buffer *bufp;
Jim Blandy's avatar
Jim Blandy committed
284

285 286 287
  if (running_asynch_code)
    save_search_regs ();

288 289 290 291
  /* 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;

292
  CHECK_STRING (string);
293 294 295
  bufp = compile_pattern (string,
			  (NILP (Vinhibit_changing_match_data)
			   ? &search_regs : NULL),
296
			  (!NILP (current_buffer->case_fold_search)
297
			   ? current_buffer->case_canon_table : Qnil),
298 299
			  posix,
			  !NILP (current_buffer->enable_multibyte_characters));
Jim Blandy's avatar
Jim Blandy committed
300 301 302 303 304 305 306 307

  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;
308
  s1 = GPT_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
309
  p2 = GAP_END_ADDR;
310
  s2 = ZV_BYTE - GPT_BYTE;
Jim Blandy's avatar
Jim Blandy committed
311 312 313
  if (s1 < 0)
    {
      p2 = p1;
314
      s2 = ZV_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
315 316 317 318
      s1 = 0;
    }
  if (s2 < 0)
    {
319
      s1 = ZV_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
320 321
      s2 = 0;
    }
322 323

  re_match_object = Qnil;
324

325
  i = re_match_2 (bufp, (char *) p1, s1, (char *) p2, s2,
326 327 328
		  PT_BYTE - BEGV_BYTE,
		  (NILP (Vinhibit_changing_match_data)
		   ? &search_regs : NULL),
329
		  ZV_BYTE - BEGV_BYTE);
330
  immediate_quit = 0;
331

Jim Blandy's avatar
Jim Blandy committed
332 333 334 335
  if (i == -2)
    matcher_overflow ();

  val = (0 <= i ? Qt : Qnil);
336
  if (NILP (Vinhibit_changing_match_data) && i >= 0)
337 338 339 340 341 342 343 344
    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);
	}
345 346 347 348 349

  /* 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
350 351 352
  return val;
}

353
DEFUN ("looking-at", Flooking_at, Slooking_at, 1, 1, 0,
354 355 356 357
       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
358
  (Lisp_Object regexp)
359
{
360
  return looking_at_1 (regexp, 0);
361 362 363
}

DEFUN ("posix-looking-at", Fposix_looking_at, Sposix_looking_at, 1, 1, 0,
364 365 366 367 368
       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
369
  (Lisp_Object regexp)
370
{
371
  return looking_at_1 (regexp, 1);
372 373 374
}

static Lisp_Object
375
string_match_1 (Lisp_Object regexp, Lisp_Object string, Lisp_Object start, int posix)
Jim Blandy's avatar
Jim Blandy committed
376 377
{
  int val;
378
  struct re_pattern_buffer *bufp;
379
  EMACS_INT pos, pos_byte;
380
  int i;
Jim Blandy's avatar
Jim Blandy committed
381

382 383 384
  if (running_asynch_code)
    save_search_regs ();

385 386
  CHECK_STRING (regexp);
  CHECK_STRING (string);
Jim Blandy's avatar
Jim Blandy committed
387 388

  if (NILP (start))
389
    pos = 0, pos_byte = 0;
Jim Blandy's avatar
Jim Blandy committed
390 391
  else
    {
392
      EMACS_INT len = SCHARS (string);
Jim Blandy's avatar
Jim Blandy committed
393

394
      CHECK_NUMBER (start);
395 396 397 398
      pos = XINT (start);
      if (pos < 0 && -pos <= len)
	pos = len + pos;
      else if (0 > pos || pos > len)
Jim Blandy's avatar
Jim Blandy committed
399
	args_out_of_range (string, start);
400
      pos_byte = string_char_to_byte (string, pos);
Jim Blandy's avatar
Jim Blandy committed
401 402
    }

403 404 405 406
  /* 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;

407 408 409
  bufp = compile_pattern (regexp,
			  (NILP (Vinhibit_changing_match_data)
			   ? &search_regs : NULL),
410
			  (!NILP (current_buffer->case_fold_search)
411
			   ? current_buffer->case_canon_table : Qnil),
412 413
			  posix,
			  STRING_MULTIBYTE (string));
Jim Blandy's avatar
Jim Blandy committed
414
  immediate_quit = 1;
415
  re_match_object = string;
416

417
  val = re_search (bufp, SSDATA (string),
418 419
		   SBYTES (string), pos_byte,
		   SBYTES (string) - pos_byte,
420 421
		   (NILP (Vinhibit_changing_match_data)
		    ? &search_regs : NULL));
Jim Blandy's avatar
Jim Blandy committed
422
  immediate_quit = 0;
423 424 425 426 427

  /* 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
428 429 430
  if (val == -2)
    matcher_overflow ();
  if (val < 0) return Qnil;
431

432 433 434 435 436 437 438 439 440
  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]);
	}
441 442

  return make_number (string_byte_to_char (string, val));
Jim Blandy's avatar
Jim Blandy committed
443
}
Richard M. Stallman's avatar
Richard M. Stallman committed
444

445
DEFUN ("string-match", Fstring_match, Sstring_match, 2, 3, 0,
446
       doc: /* Return index of start of first match for REGEXP in STRING, or nil.
Richard M. Stallman's avatar
Richard M. Stallman committed
447
Matching ignores case if `case-fold-search' is non-nil.
448 449 450
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
451 452 453 454
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
455
  (Lisp_Object regexp, Lisp_Object string, Lisp_Object start)
456 457 458 459 460
{
  return string_match_1 (regexp, string, start, 0);
}

DEFUN ("posix-string-match", Fposix_string_match, Sposix_string_match, 2, 3, 0,
461 462 463 464 465 466 467
       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
468
  (Lisp_Object regexp, Lisp_Object string, Lisp_Object start)
469 470 471 472
{
  return string_match_1 (regexp, string, start, 1);
}

Richard M. Stallman's avatar
Richard M. Stallman committed
473 474 475 476 477
/* 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
478
fast_string_match (Lisp_Object regexp, Lisp_Object string)
Richard M. Stallman's avatar
Richard M. Stallman committed
479 480
{
  int val;
481
  struct re_pattern_buffer *bufp;
Richard M. Stallman's avatar
Richard M. Stallman committed
482

483 484
  bufp = compile_pattern (regexp, 0, Qnil,
			  0, STRING_MULTIBYTE (string));
Richard M. Stallman's avatar
Richard M. Stallman committed
485
  immediate_quit = 1;
486
  re_match_object = string;
487

488
  val = re_search (bufp, SSDATA (string),
489 490
		   SBYTES (string), 0,
		   SBYTES (string), 0);
Richard M. Stallman's avatar
Richard M. Stallman committed
491 492 493
  immediate_quit = 0;
  return val;
}
Karl Heuer's avatar
Karl Heuer committed
494 495 496

/* Match REGEXP against STRING, searching all of STRING ignoring case,
   and return the index of the match, or negative on failure.
497 498
   This does not clobber the match data.
   We assume that STRING contains single-byte characters.  */
Karl Heuer's avatar
Karl Heuer committed
499 500

int
501
fast_c_string_match_ignore_case (Lisp_Object regexp, const char *string)
Karl Heuer's avatar
Karl Heuer committed
502 503 504
{
  int val;
  struct re_pattern_buffer *bufp;
505
  size_t len = strlen (string);
Karl Heuer's avatar
Karl Heuer committed
506

507
  regexp = string_make_unibyte (regexp);
508
  re_match_object = Qt;
Karl Heuer's avatar
Karl Heuer committed
509
  bufp = compile_pattern (regexp, 0,
510
			  Vascii_canon_table, 0,
511
			  0);
Karl Heuer's avatar
Karl Heuer committed
512 513 514 515 516
  immediate_quit = 1;
  val = re_search (bufp, string, len, 0, len, 0);
  immediate_quit = 0;
  return val;
}
517 518 519 520

/* Like fast_string_match but ignore case.  */

int
521
fast_string_match_ignore_case (Lisp_Object regexp, Lisp_Object string)
522 523 524 525
{
  int val;
  struct re_pattern_buffer *bufp;

526
  bufp = compile_pattern (regexp, 0, Vascii_canon_table,
527 528 529 530
			  0, STRING_MULTIBYTE (string));
  immediate_quit = 1;
  re_match_object = string;

531
  val = re_search (bufp, SSDATA (string),
532 533 534 535 536
		   SBYTES (string), 0,
		   SBYTES (string), 0);
  immediate_quit = 0;
  return val;
}
537

538
/* Match REGEXP against the characters after POS to LIMIT, and return
539 540 541 542 543 544
   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
545
fast_looking_at (Lisp_Object regexp, EMACS_INT pos, EMACS_INT pos_byte, EMACS_INT limit, EMACS_INT limit_byte, Lisp_Object string)
546 547 548 549
{
  int multibyte;
  struct re_pattern_buffer *buf;
  unsigned char *p1, *p2;
550
  EMACS_INT s1, s2;
551
  EMACS_INT len;
Juanma Barranquero's avatar
Juanma Barranquero committed
552

553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601
  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;
}

602 603 604 605 606 607 608 609

/* 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
610
newline_cache_on_off (struct buffer *buf)
611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638
{
  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
639 640

   If we find COUNT instances, set *SHORTAGE to zero, and return the
Richard M. Stallman's avatar
Richard M. Stallman committed
641
   position past the COUNTth match.  Note that for reverse motion
Richard M. Stallman's avatar
Richard M. Stallman committed
642
   this is not the same as the usual convention for Emacs motion commands.
Jim Blandy's avatar
Jim Blandy committed
643

644 645
   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
646

647 648 649
   If ALLOW_QUIT is non-zero, set immediate_quit.  That's good to do
   except when inside redisplay.  */

650 651 652
EMACS_INT
scan_buffer (register int target, EMACS_INT start, EMACS_INT end,
	     EMACS_INT count, int *shortage, int allow_quit)
Jim Blandy's avatar
Jim Blandy committed
653
{
654
  struct region_cache *newline_cache;
655
  int direction;
Jim Blandy's avatar
Jim Blandy committed
656

657 658 659 660 661 662 663 664 665 666
  if (count > 0)
    {
      direction = 1;
      if (! end) end = ZV;
    }
  else
    {
      direction = -1;
      if (! end) end = BEGV;
    }
Jim Blandy's avatar
Jim Blandy committed
667

668 669
  newline_cache_on_off (current_buffer);
  newline_cache = current_buffer->newline_cache;
Jim Blandy's avatar
Jim Blandy committed
670 671 672 673

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

674
  immediate_quit = allow_quit;
Jim Blandy's avatar
Jim Blandy committed
675

Jim Blandy's avatar
Jim Blandy committed
676
  if (count > 0)
677
    while (start != end)
Jim Blandy's avatar
Jim Blandy committed
678
      {
679 680 681 682 683
        /* 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.  */
684 685 686
	EMACS_INT ceiling_byte = CHAR_TO_BYTE (end) - 1;
	EMACS_INT start_byte = CHAR_TO_BYTE (start);
	EMACS_INT tem;
687 688 689 690 691

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

699 700 701
            /* START should never be after END.  */
            if (start_byte > ceiling_byte)
              start_byte = ceiling_byte;
702 703 704

            /* Now the text after start is an unknown region, and
               next_change is the position of the next known region. */
705
            ceiling_byte = min (next_change - 1, ceiling_byte);
706 707 708 709 710 711
          }

        /* 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.  */
712 713
	tem = BUFFER_CEILING_OF (start_byte);
	ceiling_byte = min (tem, ceiling_byte);
714 715

        {
716
          /* The termination address of the dumb loop.  */
717 718 719 720
          register unsigned char *ceiling_addr
	    = BYTE_POS_ADDR (ceiling_byte) + 1;
          register unsigned char *cursor
	    = BYTE_POS_ADDR (start_byte);
721 722 723 724 725 726 727 728 729 730 731 732 733 734
          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,
735 736
                                   start_byte + scan_start - base,
                                   start_byte + cursor - base);
737 738 739 740 741 742 743

              /* Did we find the target character?  */
              if (cursor < ceiling_addr)
                {
                  if (--count == 0)
                    {
                      immediate_quit = 0;
744
                      return BYTE_TO_CHAR (start_byte + cursor - base + 1);
745 746 747 748 749
                    }
                  cursor++;
                }
            }

750
          start = BYTE_TO_CHAR (start_byte + cursor - base);
751
        }
Jim Blandy's avatar
Jim Blandy committed
752 753
      }
  else
754 755 756
    while (start > end)
      {
        /* The last character to check before the next obstacle.  */
757 758 759
	EMACS_INT ceiling_byte = CHAR_TO_BYTE (end);
	EMACS_INT start_byte = CHAR_TO_BYTE (start);
	EMACS_INT tem;
760 761 762 763

        /* Consult the newline cache, if appropriate.  */
        if (target == '\n' && newline_cache)
          {
764
            EMACS_INT next_change;
765 766
            immediate_quit = 0;
            while (region_cache_backward
767 768
                   (current_buffer, newline_cache, start_byte, &next_change))
              start_byte = next_change;
769
            immediate_quit = allow_quit;
770 771

            /* Start should never be at or before end.  */
772 773
            if (start_byte <= ceiling_byte)
              start_byte = ceiling_byte + 1;
774 775 776

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

        /* 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
                                   start_byte + cursor - base,
                                   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

843 844 845 846
EMACS_INT
scan_newline (EMACS_INT start, EMACS_INT start_byte,
	      EMACS_INT limit, EMACS_INT limit_byte,
	      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
  EMACS_INT 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
EMACS_INT
find_next_newline_no_quit (EMACS_INT from, EMACS_INT cnt)
Jim Blandy's avatar
Jim Blandy committed
941
{
942
  return scan_buffer ('\n', from, 0, cnt, (int *) 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
EMACS_INT
find_before_next_newline (EMACS_INT from, EMACS_INT to, EMACS_INT cnt)
951 952
{
  int shortage;
953
  EMACS_INT pos = scan_buffer ('\n', from, to, cnt, &shortage, 1);
954 955 956

  if (shortage == 0)
    pos--;
957

958
  return pos;
Jim Blandy's avatar
Jim Blandy committed
959 960 961 962 963
}

/* Subroutines of Lisp buffer search functions. */

static Lisp_Object
964 965
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
966 967
{
  register int np;
968
  EMACS_INT lim, lim_byte;
Jim Blandy's avatar
Jim Blandy committed
969 970 971 972
  int n = direction;

  if (!NILP (count))
    {
973
      CHECK_NUMBER (count);
Jim Blandy's avatar
Jim Blandy committed
974 975 976
      n *= XINT (count);
    }

977
  CHECK_STRING (string);
Jim Blandy's avatar
Jim Blandy committed
978
  if (NILP (bound))
979 980 981 982 983 984
    {
      if (n > 0)
	lim = ZV, lim_byte = ZV_BYTE;
      else
	lim = BEGV, lim_byte = BEGV_BYTE;
    }
Jim Blandy's avatar
Jim Blandy committed
985 986
  else
    {
987
      CHECK_NUMBER_COERCE_MARKER (bound);
Jim Blandy's avatar
Jim Blandy committed
988
      lim = XINT (bound);
989
      if (n > 0 ? lim < PT : lim > PT)
Jim Blandy's avatar
Jim Blandy committed
990 991
	error ("Invalid search bound (wrong side of point)");
      if (lim > ZV)
992
	lim = ZV, lim_byte = ZV_BYTE;
993
      else if (lim < BEGV)
994
	lim = BEGV, lim_byte