search.c 95 KB
Newer Older
Jim Blandy's avatar
Jim Blandy committed
1
/* String search routines for GNU Emacs.
2
   Copyright (C) 1985-1987, 1993-1994, 1997-1999, 2001-2011
Glenn Morris's avatar
Glenn Morris committed
3
                 Free Software Foundation, Inc.
Jim Blandy's avatar
Jim Blandy committed
4 5 6

This file is part of GNU Emacs.

7
GNU Emacs is free software: you can redistribute it and/or modify
Jim Blandy's avatar
Jim Blandy committed
8
it under the terms of the GNU General Public License as published by
9 10
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
Jim Blandy's avatar
Jim Blandy committed
11 12 13 14 15 16 17

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
18
along with GNU Emacs.  If not, see <http://www.gnu.org/licenses/>.  */
Jim Blandy's avatar
Jim Blandy committed
19 20


21
#include <config.h>
22
#include <setjmp.h>
Jim Blandy's avatar
Jim Blandy committed
23 24
#include "lisp.h"
#include "syntax.h"
Karl Heuer's avatar
Karl Heuer committed
25
#include "category.h"
Jim Blandy's avatar
Jim Blandy committed
26
#include "buffer.h"
27
#include "character.h"
Kenichi Handa's avatar
Kenichi Handa committed
28
#include "charset.h"
29
#include "region-cache.h"
Jim Blandy's avatar
Jim Blandy committed
30
#include "commands.h"
31
#include "blockinput.h"
32
#include "intervals.h"
Jim Blandy's avatar
Jim Blandy committed
33

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

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

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

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

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


Jim Blandy's avatar
Jim Blandy committed
62 63 64 65 66 67 68
/* Every call to re_match, etc., must pass &search_regs as the regs
   argument unless you can show it is unnecessary (i.e., if re_match
   is certainly going to be called again before region-around-match
   can be called).

   Since the registers are now dynamically allocated, we need to make
   sure not to refer to the Nth register before checking that it has
Jim Blandy's avatar
Jim Blandy committed
69 70 71
   been allocated by checking search_regs.num_regs.

   The regex code keeps track of whether it has allocated the search
72 73
   buffer using bits in the re_pattern_buffer.  This means that whenever
   you compile a new pattern, it completely forgets whether it has
Jim Blandy's avatar
Jim Blandy committed
74 75 76 77 78
   allocated any registers, and will allocate new registers the next
   time you call a searching or matching function.  Therefore, we need
   to call re_set_registers after compiling a new pattern or after
   setting the match registers, so that the regex functions will be
   able to free or re-allocate it properly.  */
Jim Blandy's avatar
Jim Blandy committed
79 80
static struct re_registers search_regs;

Jim Blandy's avatar
Jim Blandy committed
81 82 83 84
/* The buffer in which the last search was performed, or
   Qt if the last search was done in a string;
   Qnil if no searching has been done yet.  */
static Lisp_Object last_thing_searched;
Jim Blandy's avatar
Jim Blandy committed
85

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

Lisp_Object Qinvalid_regexp;

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

93 94
static void set_search_regs (EMACS_INT, EMACS_INT);
static void save_search_regs (void);
95 96
static EMACS_INT simple_search (EMACS_INT, unsigned char *, EMACS_INT,
				EMACS_INT, Lisp_Object, EMACS_INT, EMACS_INT,
97
                                EMACS_INT, EMACS_INT);
98
static EMACS_INT boyer_moore (EMACS_INT, unsigned char *, EMACS_INT, EMACS_INT,
99 100 101 102
                              Lisp_Object, Lisp_Object,
                              EMACS_INT, EMACS_INT,
                              EMACS_INT, EMACS_INT, int);
static EMACS_INT search_buffer (Lisp_Object, EMACS_INT, EMACS_INT,
103
                                EMACS_INT, EMACS_INT, EMACS_INT, int,
104
                                Lisp_Object, Lisp_Object, int);
105
static void matcher_overflow (void) NO_RETURN;
106

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

113 114 115
/* Compile a regexp and signal a Lisp error if anything goes wrong.
   PATTERN is the pattern to compile.
   CP is the place to put the result.
116
   TRANSLATE is a translation table for ignoring case, or nil for none.
117
   POSIX is nonzero if we want full backtracking (POSIX style)
Karl Heuer's avatar
Karl Heuer committed
118
   for this pattern.  0 means backtrack only enough to get a valid match.
119 120

   The behavior also depends on Vsearch_spaces_regexp.  */
Jim Blandy's avatar
Jim Blandy committed
121

122
static void
123
compile_pattern_1 (struct regexp_cache *cp, Lisp_Object pattern, Lisp_Object translate, int posix)
Jim Blandy's avatar
Jim Blandy committed
124
{
125
  char *val;
126
  reg_syntax_t old;
Jim Blandy's avatar
Jim Blandy committed
127

128
  cp->regexp = Qnil;
129
  cp->buf.translate = (! NILP (translate) ? translate : make_number (0));
130
  cp->posix = posix;
131
  cp->buf.multibyte = STRING_MULTIBYTE (pattern);
Kenichi Handa's avatar
Kenichi Handa committed
132
  cp->buf.charset_unibyte = charset_unibyte;
133 134 135 136 137
  if (STRINGP (Vsearch_spaces_regexp))
    cp->whitespace_regexp = Vsearch_spaces_regexp;
  else
    cp->whitespace_regexp = Qnil;

138 139 140 141 142
  /* 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;  */
143
  old = re_set_syntax (RE_SYNTAX_EMACS
144
		       | (posix ? 0 : RE_NO_POSIX_BACKTRACKING));
145 146

  if (STRINGP (Vsearch_spaces_regexp))
147
    re_set_whitespace_regexp (SSDATA (Vsearch_spaces_regexp));
148 149
  else
    re_set_whitespace_regexp (NULL);
150

151
  val = (char *) re_compile_pattern (SSDATA (pattern),
Kenichi Handa's avatar
Kenichi Handa committed
152
				     SBYTES (pattern), &cp->buf);
153

154 155
  /* 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
156
  cp->syntax_table = cp->buf.used_syntax ? BVAR (current_buffer, syntax_table) : Qt;
157

158 159
  re_set_whitespace_regexp (NULL);

160
  re_set_syntax (old);
161
  /* UNBLOCK_INPUT;  */
Jim Blandy's avatar
Jim Blandy committed
162
  if (val)
163
    xsignal1 (Qinvalid_regexp, build_string (val));
Jim Blandy's avatar
Jim Blandy committed
164

165 166 167
  cp->regexp = Fcopy_sequence (pattern);
}

168 169 170 171 172
/* Shrink each compiled regexp buffer in the cache
   to the size actually used right now.
   This is called from garbage collection.  */

void
173
shrink_regexp_cache (void)
174
{
175
  struct regexp_cache *cp;
176 177 178 179 180

  for (cp = searchbuf_head; cp != 0; cp = cp->next)
    {
      cp->buf.allocated = cp->buf.used;
      cp->buf.buffer
181
	= (unsigned char *) xrealloc (cp->buf.buffer, cp->buf.used);
182 183 184
    }
}

185 186
/* Clear the regexp cache w.r.t. a particular syntax table,
   because it was changed.
187 188 189 190
   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
191
clear_regexp_cache (void)
192 193 194 195
{
  int i;

  for (i = 0; i < REGEXP_CACHE_SIZE; ++i)
196 197
    /* It's tempting to compare with the syntax-table we've actually changed,
       but it's not sufficient because char-table inheritance means that
198 199 200
       modifying one syntax-table can change others at the same time.  */
    if (!EQ (searchbufs[i].syntax_table, Qt))
      searchbufs[i].regexp = Qnil;
201 202
}

203
/* Compile a regexp if necessary, but first check to see if there's one in
204 205
   the cache.
   PATTERN is the pattern to compile.
206
   TRANSLATE is a translation table for ignoring case, or nil for none.
207 208 209 210 211 212
   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.  */
213 214

struct re_pattern_buffer *
215
compile_pattern (Lisp_Object pattern, struct re_registers *regp, Lisp_Object translate, int posix, int multibyte)
216 217 218 219 220 221
{
  struct regexp_cache *cp, **cpp;

  for (cpp = &searchbuf_head; ; cpp = &cp->next)
    {
      cp = *cpp;
222 223
      /* 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
224 225 226
	 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.  */
227
      if (NILP (cp->regexp))
228
	goto compile_it;
229
      if (SCHARS (cp->regexp) == SCHARS (pattern)
230
	  && STRING_MULTIBYTE (cp->regexp) == STRING_MULTIBYTE (pattern)
231
	  && !NILP (Fstring_equal (cp->regexp, pattern))
232
	  && EQ (cp->buf.translate, (! NILP (translate) ? translate : make_number (0)))
Karl Heuer's avatar
Karl Heuer committed
233
	  && cp->posix == posix
234
	  && (EQ (cp->syntax_table, Qt)
Tom Tromey's avatar
Tom Tromey committed
235
	      || EQ (cp->syntax_table, BVAR (current_buffer, syntax_table)))
Kenichi Handa's avatar
Kenichi Handa committed
236 237
	  && !NILP (Fequal (cp->whitespace_regexp, Vsearch_spaces_regexp))
	  && cp->buf.charset_unibyte == charset_unibyte)
238 239
	break;

240 241 242
      /* 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.  */
243 244
      if (cp->next == 0)
	{
245
	compile_it:
246
	  compile_pattern_1 (cp, pattern, translate, posix);
247 248 249 250 251 252 253 254 255 256
	  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
257

258 259 260 261 262
  /* 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
263
  /* The compiled pattern can be used both for multibyte and unibyte
Kenichi Handa's avatar
Kenichi Handa committed
264 265 266
     target.  But, we have to tell which the pattern is used for. */
  cp->buf.target_multibyte = multibyte;

267
  return &cp->buf;
Jim Blandy's avatar
Jim Blandy committed
268 269 270
}


271
static Lisp_Object
272
looking_at_1 (Lisp_Object string, int posix)
Jim Blandy's avatar
Jim Blandy committed
273 274 275
{
  Lisp_Object val;
  unsigned char *p1, *p2;
276
  EMACS_INT s1, s2;
277
  register EMACS_INT i;
278
  struct re_pattern_buffer *bufp;
Jim Blandy's avatar
Jim Blandy committed
279

280 281 282
  if (running_asynch_code)
    save_search_regs ();

283
  /* This is so set_image_of_range_1 in regex.c can find the EQV table.  */
Tom Tromey's avatar
Tom Tromey committed
284 285
  XCHAR_TABLE (BVAR (current_buffer, case_canon_table))->extras[2]
    = BVAR (current_buffer, case_eqv_table);
286

287
  CHECK_STRING (string);
288 289 290
  bufp = compile_pattern (string,
			  (NILP (Vinhibit_changing_match_data)
			   ? &search_regs : NULL),
Tom Tromey's avatar
Tom Tromey committed
291 292
			  (!NILP (BVAR (current_buffer, case_fold_search))
			   ? BVAR (current_buffer, case_canon_table) : Qnil),
293
			  posix,
Tom Tromey's avatar
Tom Tromey committed
294
			  !NILP (BVAR (current_buffer, enable_multibyte_characters)));
Jim Blandy's avatar
Jim Blandy committed
295 296 297 298 299 300 301 302

  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;
303
  s1 = GPT_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
304
  p2 = GAP_END_ADDR;
305
  s2 = ZV_BYTE - GPT_BYTE;
Jim Blandy's avatar
Jim Blandy committed
306 307 308
  if (s1 < 0)
    {
      p2 = p1;
309
      s2 = ZV_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
310 311 312 313
      s1 = 0;
    }
  if (s2 < 0)
    {
314
      s1 = ZV_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
315 316
      s2 = 0;
    }
317 318

  re_match_object = Qnil;
319

320
  i = re_match_2 (bufp, (char *) p1, s1, (char *) p2, s2,
321 322 323
		  PT_BYTE - BEGV_BYTE,
		  (NILP (Vinhibit_changing_match_data)
		   ? &search_regs : NULL),
324
		  ZV_BYTE - BEGV_BYTE);
325
  immediate_quit = 0;
326

Jim Blandy's avatar
Jim Blandy committed
327 328 329 330
  if (i == -2)
    matcher_overflow ();

  val = (0 <= i ? Qt : Qnil);
331
  if (NILP (Vinhibit_changing_match_data) && i >= 0)
332 333 334 335 336 337 338 339
    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);
	}
340 341 342 343 344

  /* 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
345 346 347
  return val;
}

348
DEFUN ("looking-at", Flooking_at, Slooking_at, 1, 1, 0,
349 350 351 352
       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
353
  (Lisp_Object regexp)
354
{
355
  return looking_at_1 (regexp, 0);
356 357 358
}

DEFUN ("posix-looking-at", Fposix_looking_at, Sposix_looking_at, 1, 1, 0,
359 360 361 362 363
       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
364
  (Lisp_Object regexp)
365
{
366
  return looking_at_1 (regexp, 1);
367 368 369
}

static Lisp_Object
370
string_match_1 (Lisp_Object regexp, Lisp_Object string, Lisp_Object start, int posix)
Jim Blandy's avatar
Jim Blandy committed
371 372
{
  int val;
373
  struct re_pattern_buffer *bufp;
374
  EMACS_INT pos, pos_byte;
375
  int i;
Jim Blandy's avatar
Jim Blandy committed
376

377 378 379
  if (running_asynch_code)
    save_search_regs ();

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

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

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

398
  /* This is so set_image_of_range_1 in regex.c can find the EQV table.  */
Tom Tromey's avatar
Tom Tromey committed
399 400
  XCHAR_TABLE (BVAR (current_buffer, case_canon_table))->extras[2]
    = BVAR (current_buffer, case_eqv_table);
401

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

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

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

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

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

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

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

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

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

Richard M. Stallman's avatar
Richard M. Stallman committed
468 469 470 471 472
/* 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
473
fast_string_match (Lisp_Object regexp, Lisp_Object string)
Richard M. Stallman's avatar
Richard M. Stallman committed
474 475
{
  int val;
476
  struct re_pattern_buffer *bufp;
Richard M. Stallman's avatar
Richard M. Stallman committed
477

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

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

/* Match REGEXP against STRING, searching all of STRING ignoring case,
   and return the index of the match, or negative on failure.
492 493
   This does not clobber the match data.
   We assume that STRING contains single-byte characters.  */
Karl Heuer's avatar
Karl Heuer committed
494 495

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

502
  regexp = string_make_unibyte (regexp);
503
  re_match_object = Qt;
Karl Heuer's avatar
Karl Heuer committed
504
  bufp = compile_pattern (regexp, 0,
505
			  Vascii_canon_table, 0,
506
			  0);
Karl Heuer's avatar
Karl Heuer committed
507 508 509 510 511
  immediate_quit = 1;
  val = re_search (bufp, string, len, 0, len, 0);
  immediate_quit = 0;
  return val;
}
512 513 514 515

/* Like fast_string_match but ignore case.  */

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

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

526
  val = re_search (bufp, SSDATA (string),
527 528 529 530 531
		   SBYTES (string), 0,
		   SBYTES (string), 0);
  immediate_quit = 0;
  return val;
}
532

533
/* Match REGEXP against the characters after POS to LIMIT, and return
534 535 536 537 538 539
   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
540
fast_looking_at (Lisp_Object regexp, EMACS_INT pos, EMACS_INT pos_byte, EMACS_INT limit, EMACS_INT limit_byte, Lisp_Object string)
541 542 543 544
{
  int multibyte;
  struct re_pattern_buffer *buf;
  unsigned char *p1, *p2;
545
  EMACS_INT s1, s2;
546
  EMACS_INT len;
Juanma Barranquero's avatar
Juanma Barranquero committed
547

548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584
  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
585
      multibyte = ! NILP (BVAR (current_buffer, enable_multibyte_characters));
586 587 588 589 590 591 592 593 594 595 596
    }

  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;
}

597 598 599 600 601 602 603 604

/* 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
605
newline_cache_on_off (struct buffer *buf)
606
{
Tom Tromey's avatar
Tom Tromey committed
607
  if (NILP (BVAR (buf, cache_long_line_scans)))
608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633
    {
      /* 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
634 635

   If we find COUNT instances, set *SHORTAGE to zero, and return the
Richard M. Stallman's avatar
Richard M. Stallman committed
636
   position past the COUNTth match.  Note that for reverse motion
Richard M. Stallman's avatar
Richard M. Stallman committed
637
   this is not the same as the usual convention for Emacs motion commands.
Jim Blandy's avatar
Jim Blandy committed
638

639 640
   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
641

642 643 644
   If ALLOW_QUIT is non-zero, set immediate_quit.  That's good to do
   except when inside redisplay.  */

645 646 647
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
648
{
649
  struct region_cache *newline_cache;
650
  int direction;
Jim Blandy's avatar
Jim Blandy committed
651

652 653 654 655 656 657 658 659 660 661
  if (count > 0)
    {
      direction = 1;
      if (! end) end = ZV;
    }
  else
    {
      direction = -1;
      if (! end) end = BEGV;
    }
Jim Blandy's avatar
Jim Blandy committed
662

663 664
  newline_cache_on_off (current_buffer);
  newline_cache = current_buffer->newline_cache;
Jim Blandy's avatar
Jim Blandy committed
665 666 667 668

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

669
  immediate_quit = allow_quit;
Jim Blandy's avatar
Jim Blandy committed
670

Jim Blandy's avatar
Jim Blandy committed
671
  if (count > 0)
672
    while (start != end)
Jim Blandy's avatar
Jim Blandy committed
673
      {
674 675 676 677 678
        /* 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.  */
679 680 681
	EMACS_INT ceiling_byte = CHAR_TO_BYTE (end) - 1;
	EMACS_INT start_byte = CHAR_TO_BYTE (start);
	EMACS_INT tem;
682 683 684 685 686

        /* If we're looking for a newline, consult the newline cache
           to see where we can avoid some scanning.  */
        if (target == '\n' && newline_cache)
          {
687
            EMACS_INT next_change;
688 689
            immediate_quit = 0;
            while (region_cache_forward
690 691
                   (current_buffer, newline_cache, start_byte, &next_change))
              start_byte = next_change;
692
            immediate_quit = allow_quit;
693

694 695 696
            /* START should never be after END.  */
            if (start_byte > ceiling_byte)
              start_byte = ceiling_byte;
697 698 699

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

        /* 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.  */
707 708
	tem = BUFFER_CEILING_OF (start_byte);
	ceiling_byte = min (tem, ceiling_byte);
709 710

        {
711
          /* The termination address of the dumb loop.  */
712 713 714 715
          register unsigned char *ceiling_addr
	    = BYTE_POS_ADDR (ceiling_byte) + 1;
          register unsigned char *cursor
	    = BYTE_POS_ADDR (start_byte);
716 717 718 719 720 721 722 723 724 725 726 727 728 729
          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,
730 731
                                   start_byte + scan_start - base,
                                   start_byte + cursor - base);
732 733 734 735 736 737 738

              /* Did we find the target character?  */
              if (cursor < ceiling_addr)
                {
                  if (--count == 0)
                    {
                      immediate_quit = 0;
739
                      return BYTE_TO_CHAR (start_byte + cursor - base + 1);
740 741 742 743 744
                    }
                  cursor++;
                }
            }

745
          start = BYTE_TO_CHAR (start_byte + cursor - base);
746
        }
Jim Blandy's avatar
Jim Blandy committed
747 748
      }
  else
749 750 751
    while (start > end)
      {
        /* The last character to check before the next obstacle.  */
752 753 754
	EMACS_INT ceiling_byte = CHAR_TO_BYTE (end);
	EMACS_INT start_byte = CHAR_TO_BYTE (start);
	EMACS_INT tem;
755 756 757 758

        /* Consult the newline cache, if appropriate.  */
        if (target == '\n' && newline_cache)
          {
759
            EMACS_INT next_change;
760 761
            immediate_quit = 0;
            while (region_cache_backward
762 763
                   (current_buffer, newline_cache, start_byte, &next_change))
              start_byte = next_change;
764
            immediate_quit = allow_quit;
765 766

            /* Start should never be at or before end.  */
767 768
            if (start_byte <= ceiling_byte)
              start_byte = ceiling_byte + 1;
769 770 771

            /* Now the text before start is an unknown region, and
               next_change is the position of the next known region. */
772
            ceiling_byte = max (next_change, ceiling_byte);
773 774 775
          }

        /* Stop scanning before the gap.  */
776 777
	tem = BUFFER_FLOOR_OF (start_byte - 1);
	ceiling_byte = max (tem, ceiling_byte);
778 779 780

        {
          /* The termination address of the dumb loop.  */
781 782
          register unsigned char *ceiling_addr = BYTE_POS_ADDR (ceiling_byte);
          register unsigned char *cursor = BYTE_POS_ADDR (start_byte - 1);
783 784 785 786 787 788 789 790 791 792 793 794 795
          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,
796 797
                                   start_byte + cursor - base,
                                   start_byte + scan_start - base);
798 799 800 801 802 803 804

              /* Did we find the target character?  */
              if (cursor >= ceiling_addr)
                {
                  if (++count >= 0)
                    {
                      immediate_quit = 0;
805
                      return BYTE_TO_CHAR (start_byte + cursor - base);
806 807 808 809 810
                    }
                  cursor--;
                }
            }

811
	  start = BYTE_TO_CHAR (start_byte + cursor - base);
812 813 814
        }
      }

Jim Blandy's avatar
Jim Blandy committed
815 816
  immediate_quit = 0;
  if (shortage != 0)
Jim Blandy's avatar
Jim Blandy committed
817
    *shortage = count * direction;
818
  return start;
Jim Blandy's avatar
Jim Blandy committed
819
}
820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835

/* 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
836
   except in special cases.  */
Jim Blandy's avatar
Jim Blandy committed
837

838 839 840 841
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)
842
{
843 844 845 846 847
  int direction = ((count > 0) ? 1 : -1);

  register unsigned char *cursor;
  unsigned char *base;

848
  EMACS_INT ceiling;
849 850
  register unsigned char *ceiling_addr;

851 852
  int old_immediate_quit = immediate_quit;

853 854 855
  /* The code that follows is like scan_buffer
     but checks for either newline or carriage return.  */

856 857
  if (allow_quit)
    immediate_quit++;
858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877

  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)
		    {
878
		      immediate_quit = old_immediate_quit;
879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910
		      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)
		    {
911
		      immediate_quit = old_immediate_quit;
912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928
		      /* 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);
929
  immediate_quit = old_immediate_quit;
930 931

  return count * direction;
932 933
}

934 935
EMACS_INT
find_next_newline_no_quit (EMACS_INT from, EMACS_INT cnt)
Jim Blandy's avatar
Jim Blandy committed
936
{
937
  return scan_buffer ('\n', from, 0, cnt, (int *) 0, 0);
938 939 940 941 942
}

/* 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.  */
943

944 945
EMACS_INT
find_before_next_newline (EMACS_INT from, EMACS_INT to, EMACS_INT cnt)
946 947
{
  int shortage;
948
  EMACS_INT pos = scan_buffer ('\n', from, to, cnt, &shortage, 1);
949 950 951

  if (shortage == 0)
    pos--;
952

953
  return pos;
Jim Blandy's avatar
Jim Blandy committed
954 955 956 957 958
}

/* Subroutines of Lisp buffer search functions. */

static Lisp_Object
959 960
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
961 962
{
  register int np;
963
  EMACS_INT lim, lim_byte;
Jim Blandy's avatar
Jim Blandy committed
964 965 966 967
  int n = direction;

  if (!NILP (count))
    {
968
      CHECK_NUMBER (count);
Jim Blandy's avatar
Jim Blandy committed
969 970 971
      n *= XINT (count);
    }

972
  CHECK_STRING (string);
Jim Blandy's avatar
Jim Blandy committed
973
  if (NILP (bound))
974 975 976 977 978 979
    {
      if (n > 0)
	lim = ZV, lim_byte = ZV_BYTE;
      else
	lim = BEGV, lim_byte = BEGV_BYTE;
    }
Jim Blandy's avatar
Jim Blandy committed
980