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,
Glenn Morris's avatar
Glenn Morris committed
3
                 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010
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;

Richard M. Stallman's avatar
Richard M. Stallman committed
94
Lisp_Object Vsearch_spaces_regexp;
95

96 97 98 99 100
/* If non-nil, the match data will not be changed during call to
   searching or matching functions.  This variable is for internal use
   only.  */
Lisp_Object Vinhibit_changing_match_data;

101 102 103 104 105 106 107 108 109 110 111 112
static void set_search_regs (EMACS_INT, EMACS_INT);
static void save_search_regs (void);
static EMACS_INT simple_search (int, unsigned char *, int, int,
                                Lisp_Object, EMACS_INT, EMACS_INT,
                                EMACS_INT, EMACS_INT);
static EMACS_INT boyer_moore (int, unsigned char *, int, int,
                              Lisp_Object, Lisp_Object,
                              EMACS_INT, EMACS_INT,
                              EMACS_INT, EMACS_INT, int);
static EMACS_INT search_buffer (Lisp_Object, EMACS_INT, EMACS_INT,
                                EMACS_INT, EMACS_INT, int, int,
                                Lisp_Object, Lisp_Object, int);
113
static void matcher_overflow (void) NO_RETURN;
114

Jim Blandy's avatar
Jim Blandy committed
115
static void
116
matcher_overflow (void)
Jim Blandy's avatar
Jim Blandy committed
117 118 119 120
{
  error ("Stack overflow in regexp matcher");
}

121 122 123
/* 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.
124
   TRANSLATE is a translation table for ignoring case, or nil for none.
125 126 127 128 129
   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
130
   for this pattern.  0 means backtrack only enough to get a valid match.
131 132

   The behavior also depends on Vsearch_spaces_regexp.  */
Jim Blandy's avatar
Jim Blandy committed
133

134
static void
135
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
136
{
137
  char *val;
138
  reg_syntax_t old;
Jim Blandy's avatar
Jim Blandy committed
139

140
  cp->regexp = Qnil;
141
  cp->buf.translate = (! NILP (translate) ? translate : make_number (0));
142
  cp->posix = posix;
143
  cp->buf.multibyte = STRING_MULTIBYTE (pattern);
Kenichi Handa's avatar
Kenichi Handa committed
144
  cp->buf.charset_unibyte = charset_unibyte;
145 146 147 148 149
  if (STRINGP (Vsearch_spaces_regexp))
    cp->whitespace_regexp = Vsearch_spaces_regexp;
  else
    cp->whitespace_regexp = Qnil;

150 151 152 153 154
  /* 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;  */
155
  old = re_set_syntax (RE_SYNTAX_EMACS
156
		       | (posix ? 0 : RE_NO_POSIX_BACKTRACKING));
157 158 159 160 161

  if (STRINGP (Vsearch_spaces_regexp))
    re_set_whitespace_regexp (SDATA (Vsearch_spaces_regexp));
  else
    re_set_whitespace_regexp (NULL);
162

Kenichi Handa's avatar
Kenichi Handa committed
163 164
  val = (char *) re_compile_pattern ((char *) SDATA (pattern),
				     SBYTES (pattern), &cp->buf);
165

166 167 168 169
  /* 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;

170 171
  re_set_whitespace_regexp (NULL);

172
  re_set_syntax (old);
173
  /* UNBLOCK_INPUT;  */
Jim Blandy's avatar
Jim Blandy committed
174
  if (val)
175
    xsignal1 (Qinvalid_regexp, build_string (val));
Jim Blandy's avatar
Jim Blandy committed
176

177 178 179
  cp->regexp = Fcopy_sequence (pattern);
}

180 181 182 183 184
/* Shrink each compiled regexp buffer in the cache
   to the size actually used right now.
   This is called from garbage collection.  */

void
185
shrink_regexp_cache (void)
186
{
187
  struct regexp_cache *cp;
188 189 190 191 192

  for (cp = searchbuf_head; cp != 0; cp = cp->next)
    {
      cp->buf.allocated = cp->buf.used;
      cp->buf.buffer
193
	= (unsigned char *) xrealloc (cp->buf.buffer, cp->buf.used);
194 195 196
    }
}

197 198
/* Clear the regexp cache w.r.t. a particular syntax table,
   because it was changed.
199 200 201 202
   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
203
clear_regexp_cache (void)
204 205 206 207
{
  int i;

  for (i = 0; i < REGEXP_CACHE_SIZE; ++i)
208 209
    /* It's tempting to compare with the syntax-table we've actually changed,
       but it's not sufficient because char-table inheritance means that
210 211 212
       modifying one syntax-table can change others at the same time.  */
    if (!EQ (searchbufs[i].syntax_table, Qt))
      searchbufs[i].regexp = Qnil;
213 214
}

215
/* Compile a regexp if necessary, but first check to see if there's one in
216 217
   the cache.
   PATTERN is the pattern to compile.
218
   TRANSLATE is a translation table for ignoring case, or nil for none.
219 220 221 222 223 224
   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.  */
225 226

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

  for (cpp = &searchbuf_head; ; cpp = &cp->next)
    {
      cp = *cpp;
234 235
      /* 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
236 237 238
	 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.  */
239
      if (NILP (cp->regexp))
240
	goto compile_it;
241
      if (SCHARS (cp->regexp) == SCHARS (pattern)
242
	  && STRING_MULTIBYTE (cp->regexp) == STRING_MULTIBYTE (pattern)
243
	  && !NILP (Fstring_equal (cp->regexp, pattern))
244
	  && EQ (cp->buf.translate, (! NILP (translate) ? translate : make_number (0)))
Karl Heuer's avatar
Karl Heuer committed
245
	  && cp->posix == posix
246 247
	  && (EQ (cp->syntax_table, Qt)
	      || EQ (cp->syntax_table, current_buffer->syntax_table))
Kenichi Handa's avatar
Kenichi Handa committed
248 249
	  && !NILP (Fequal (cp->whitespace_regexp, Vsearch_spaces_regexp))
	  && cp->buf.charset_unibyte == charset_unibyte)
250 251
	break;

252 253 254
      /* 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.  */
255 256
      if (cp->next == 0)
	{
257
	compile_it:
Kenichi Handa's avatar
Kenichi Handa committed
258
	  compile_pattern_1 (cp, pattern, translate, regp, posix);
259 260 261 262 263 264 265 266 267 268
	  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
269

270 271 272 273 274
  /* 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
275
  /* The compiled pattern can be used both for multibyte and unibyte
Kenichi Handa's avatar
Kenichi Handa committed
276 277 278
     target.  But, we have to tell which the pattern is used for. */
  cp->buf.target_multibyte = multibyte;

279
  return &cp->buf;
Jim Blandy's avatar
Jim Blandy committed
280 281 282
}


283
static Lisp_Object
284
looking_at_1 (Lisp_Object string, int posix)
Jim Blandy's avatar
Jim Blandy committed
285 286 287
{
  Lisp_Object val;
  unsigned char *p1, *p2;
288
  EMACS_INT s1, s2;
Jim Blandy's avatar
Jim Blandy committed
289
  register int i;
290
  struct re_pattern_buffer *bufp;
Jim Blandy's avatar
Jim Blandy committed
291

292 293 294
  if (running_asynch_code)
    save_search_regs ();

295 296 297 298
  /* 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;

299
  CHECK_STRING (string);
300 301 302
  bufp = compile_pattern (string,
			  (NILP (Vinhibit_changing_match_data)
			   ? &search_regs : NULL),
303
			  (!NILP (current_buffer->case_fold_search)
304
			   ? current_buffer->case_canon_table : Qnil),
305 306
			  posix,
			  !NILP (current_buffer->enable_multibyte_characters));
Jim Blandy's avatar
Jim Blandy committed
307 308 309 310 311 312 313 314

  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;
315
  s1 = GPT_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
316
  p2 = GAP_END_ADDR;
317
  s2 = ZV_BYTE - GPT_BYTE;
Jim Blandy's avatar
Jim Blandy committed
318 319 320
  if (s1 < 0)
    {
      p2 = p1;
321
      s2 = ZV_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
322 323 324 325
      s1 = 0;
    }
  if (s2 < 0)
    {
326
      s1 = ZV_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
327 328
      s2 = 0;
    }
329 330

  re_match_object = Qnil;
331

332
  i = re_match_2 (bufp, (char *) p1, s1, (char *) p2, s2,
333 334 335
		  PT_BYTE - BEGV_BYTE,
		  (NILP (Vinhibit_changing_match_data)
		   ? &search_regs : NULL),
336
		  ZV_BYTE - BEGV_BYTE);
337
  immediate_quit = 0;
338

Jim Blandy's avatar
Jim Blandy committed
339 340 341 342
  if (i == -2)
    matcher_overflow ();

  val = (0 <= i ? Qt : Qnil);
343
  if (NILP (Vinhibit_changing_match_data) && i >= 0)
344 345 346 347 348 349 350 351
    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);
	}
352 353 354 355 356

  /* 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
357 358 359
  return val;
}

360
DEFUN ("looking-at", Flooking_at, Slooking_at, 1, 1, 0,
361 362 363 364
       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
365
  (Lisp_Object regexp)
366
{
367
  return looking_at_1 (regexp, 0);
368 369 370
}

DEFUN ("posix-looking-at", Fposix_looking_at, Sposix_looking_at, 1, 1, 0,
371 372 373 374 375
       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
376
  (Lisp_Object regexp)
377
{
378
  return looking_at_1 (regexp, 1);
379 380 381
}

static Lisp_Object
382
string_match_1 (Lisp_Object regexp, Lisp_Object string, Lisp_Object start, int posix)
Jim Blandy's avatar
Jim Blandy committed
383 384
{
  int val;
385
  struct re_pattern_buffer *bufp;
386
  EMACS_INT pos, pos_byte;
387
  int i;
Jim Blandy's avatar
Jim Blandy committed
388

389 390 391
  if (running_asynch_code)
    save_search_regs ();

392 393
  CHECK_STRING (regexp);
  CHECK_STRING (string);
Jim Blandy's avatar
Jim Blandy committed
394 395

  if (NILP (start))
396
    pos = 0, pos_byte = 0;
Jim Blandy's avatar
Jim Blandy committed
397 398
  else
    {
399
      int len = SCHARS (string);
Jim Blandy's avatar
Jim Blandy committed
400

401
      CHECK_NUMBER (start);
402 403 404 405
      pos = XINT (start);
      if (pos < 0 && -pos <= len)
	pos = len + pos;
      else if (0 > pos || pos > len)
Jim Blandy's avatar
Jim Blandy committed
406
	args_out_of_range (string, start);
407
      pos_byte = string_char_to_byte (string, pos);
Jim Blandy's avatar
Jim Blandy committed
408 409
    }

410 411 412 413
  /* 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;

414 415 416
  bufp = compile_pattern (regexp,
			  (NILP (Vinhibit_changing_match_data)
			   ? &search_regs : NULL),
417
			  (!NILP (current_buffer->case_fold_search)
418
			   ? current_buffer->case_canon_table : Qnil),
419 420
			  posix,
			  STRING_MULTIBYTE (string));
Jim Blandy's avatar
Jim Blandy committed
421
  immediate_quit = 1;
422
  re_match_object = string;
423

424 425 426
  val = re_search (bufp, (char *) SDATA (string),
		   SBYTES (string), pos_byte,
		   SBYTES (string) - pos_byte,
427 428
		   (NILP (Vinhibit_changing_match_data)
		    ? &search_regs : NULL));
Jim Blandy's avatar
Jim Blandy committed
429
  immediate_quit = 0;
430 431 432 433 434

  /* 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
435 436 437
  if (val == -2)
    matcher_overflow ();
  if (val < 0) return Qnil;
438

439 440 441 442 443 444 445 446 447
  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]);
	}
448 449

  return make_number (string_byte_to_char (string, val));
Jim Blandy's avatar
Jim Blandy committed
450
}
Richard M. Stallman's avatar
Richard M. Stallman committed
451

452
DEFUN ("string-match", Fstring_match, Sstring_match, 2, 3, 0,
453
       doc: /* Return index of start of first match for REGEXP in STRING, or nil.
Richard M. Stallman's avatar
Richard M. Stallman committed
454
Matching ignores case if `case-fold-search' is non-nil.
455 456 457
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
458 459 460 461
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
462
  (Lisp_Object regexp, Lisp_Object string, Lisp_Object start)
463 464 465 466 467
{
  return string_match_1 (regexp, string, start, 0);
}

DEFUN ("posix-string-match", Fposix_string_match, Sposix_string_match, 2, 3, 0,
468 469 470 471 472 473 474
       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
475
  (Lisp_Object regexp, Lisp_Object string, Lisp_Object start)
476 477 478 479
{
  return string_match_1 (regexp, string, start, 1);
}

Richard M. Stallman's avatar
Richard M. Stallman committed
480 481 482 483 484
/* 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
485
fast_string_match (Lisp_Object regexp, Lisp_Object string)
Richard M. Stallman's avatar
Richard M. Stallman committed
486 487
{
  int val;
488
  struct re_pattern_buffer *bufp;
Richard M. Stallman's avatar
Richard M. Stallman committed
489

490 491
  bufp = compile_pattern (regexp, 0, Qnil,
			  0, STRING_MULTIBYTE (string));
Richard M. Stallman's avatar
Richard M. Stallman committed
492
  immediate_quit = 1;
493
  re_match_object = string;
494

495 496 497
  val = re_search (bufp, (char *) SDATA (string),
		   SBYTES (string), 0,
		   SBYTES (string), 0);
Richard M. Stallman's avatar
Richard M. Stallman committed
498 499 500
  immediate_quit = 0;
  return val;
}
Karl Heuer's avatar
Karl Heuer committed
501 502 503

/* Match REGEXP against STRING, searching all of STRING ignoring case,
   and return the index of the match, or negative on failure.
504 505
   This does not clobber the match data.
   We assume that STRING contains single-byte characters.  */
Karl Heuer's avatar
Karl Heuer committed
506 507

int
508
fast_c_string_match_ignore_case (Lisp_Object regexp, const char *string)
Karl Heuer's avatar
Karl Heuer committed
509 510 511 512 513
{
  int val;
  struct re_pattern_buffer *bufp;
  int len = strlen (string);

514
  regexp = string_make_unibyte (regexp);
515
  re_match_object = Qt;
Karl Heuer's avatar
Karl Heuer committed
516
  bufp = compile_pattern (regexp, 0,
517
			  Vascii_canon_table, 0,
518
			  0);
Karl Heuer's avatar
Karl Heuer committed
519 520 521 522 523
  immediate_quit = 1;
  val = re_search (bufp, string, len, 0, len, 0);
  immediate_quit = 0;
  return val;
}
524 525 526 527

/* Like fast_string_match but ignore case.  */

int
528
fast_string_match_ignore_case (Lisp_Object regexp, Lisp_Object string)
529 530 531 532
{
  int val;
  struct re_pattern_buffer *bufp;

533
  bufp = compile_pattern (regexp, 0, Vascii_canon_table,
534 535 536 537 538 539 540 541 542 543
			  0, STRING_MULTIBYTE (string));
  immediate_quit = 1;
  re_match_object = string;

  val = re_search (bufp, (char *) SDATA (string),
		   SBYTES (string), 0,
		   SBYTES (string), 0);
  immediate_quit = 0;
  return val;
}
544

545
/* Match REGEXP against the characters after POS to LIMIT, and return
546 547 548 549 550 551
   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
552
fast_looking_at (Lisp_Object regexp, EMACS_INT pos, EMACS_INT pos_byte, EMACS_INT limit, EMACS_INT limit_byte, Lisp_Object string)
553 554 555 556
{
  int multibyte;
  struct re_pattern_buffer *buf;
  unsigned char *p1, *p2;
557
  EMACS_INT s1, s2;
558
  EMACS_INT len;
Juanma Barranquero's avatar
Juanma Barranquero committed
559

560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608
  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;
}

609 610 611 612 613 614 615 616

/* 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
617
newline_cache_on_off (struct buffer *buf)
618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645
{
  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
646 647

   If we find COUNT instances, set *SHORTAGE to zero, and return the
Richard M. Stallman's avatar
Richard M. Stallman committed
648
   position past the COUNTth match.  Note that for reverse motion
Richard M. Stallman's avatar
Richard M. Stallman committed
649
   this is not the same as the usual convention for Emacs motion commands.
Jim Blandy's avatar
Jim Blandy committed
650

651 652
   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
653

654 655 656
   If ALLOW_QUIT is non-zero, set immediate_quit.  That's good to do
   except when inside redisplay.  */

Andreas Schwab's avatar
Andreas Schwab committed
657
int
658
scan_buffer (register int target, EMACS_INT start, EMACS_INT end, int count, int *shortage, int allow_quit)
Jim Blandy's avatar
Jim Blandy committed
659
{
660
  struct region_cache *newline_cache;
661
  int direction;
Jim Blandy's avatar
Jim Blandy committed
662

663 664 665 666 667 668 669 670 671 672
  if (count > 0)
    {
      direction = 1;
      if (! end) end = ZV;
    }
  else
    {
      direction = -1;
      if (! end) end = BEGV;
    }
Jim Blandy's avatar
Jim Blandy committed
673

674 675
  newline_cache_on_off (current_buffer);
  newline_cache = current_buffer->newline_cache;
Jim Blandy's avatar
Jim Blandy committed
676 677 678 679

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

680
  immediate_quit = allow_quit;
Jim Blandy's avatar
Jim Blandy committed
681

Jim Blandy's avatar
Jim Blandy committed
682
  if (count > 0)
683
    while (start != end)
Jim Blandy's avatar
Jim Blandy committed
684
      {
685 686 687 688 689
        /* 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.  */
690 691 692
	EMACS_INT ceiling_byte = CHAR_TO_BYTE (end) - 1;
	EMACS_INT start_byte = CHAR_TO_BYTE (start);
	EMACS_INT tem;
693 694 695 696 697 698 699 700

        /* If we're looking for a newline, consult the newline cache
           to see where we can avoid some scanning.  */
        if (target == '\n' && newline_cache)
          {
            int next_change;
            immediate_quit = 0;
            while (region_cache_forward
701 702
                   (current_buffer, newline_cache, start_byte, &next_change))
              start_byte = next_change;
703
            immediate_quit = allow_quit;
704

705 706 707
            /* START should never be after END.  */
            if (start_byte > ceiling_byte)
              start_byte = ceiling_byte;
708 709 710

            /* Now the text after start is an unknown region, and
               next_change is the position of the next known region. */
711
            ceiling_byte = min (next_change - 1, ceiling_byte);
712 713 714 715 716 717
          }

        /* 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.  */
718 719
	tem = BUFFER_CEILING_OF (start_byte);
	ceiling_byte = min (tem, ceiling_byte);
720 721

        {
722
          /* The termination address of the dumb loop.  */
723 724 725 726
          register unsigned char *ceiling_addr
	    = BYTE_POS_ADDR (ceiling_byte) + 1;
          register unsigned char *cursor
	    = BYTE_POS_ADDR (start_byte);
727 728 729 730 731 732 733 734 735 736 737 738 739 740
          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,
741 742
                                   start_byte + scan_start - base,
                                   start_byte + cursor - base);
743 744 745 746 747 748 749

              /* Did we find the target character?  */
              if (cursor < ceiling_addr)
                {
                  if (--count == 0)
                    {
                      immediate_quit = 0;
750
                      return BYTE_TO_CHAR (start_byte + cursor - base + 1);
751 752 753 754 755
                    }
                  cursor++;
                }
            }

756
          start = BYTE_TO_CHAR (start_byte + cursor - base);
757
        }
Jim Blandy's avatar
Jim Blandy committed
758 759
      }
  else
760 761 762
    while (start > end)
      {
        /* The last character to check before the next obstacle.  */
763 764 765
	EMACS_INT ceiling_byte = CHAR_TO_BYTE (end);
	EMACS_INT start_byte = CHAR_TO_BYTE (start);
	EMACS_INT tem;
766 767 768 769 770 771 772

        /* Consult the newline cache, if appropriate.  */
        if (target == '\n' && newline_cache)
          {
            int next_change;
            immediate_quit = 0;
            while (region_cache_backward
773 774
                   (current_buffer, newline_cache, start_byte, &next_change))
              start_byte = next_change;
775
            immediate_quit = allow_quit;
776 777

            /* Start should never be at or before end.  */
778 779
            if (start_byte <= ceiling_byte)
              start_byte = ceiling_byte + 1;
780 781 782

            /* Now the text before start is an unknown region, and
               next_change is the position of the next known region. */
783
            ceiling_byte = max (next_change, ceiling_byte);
784 785 786
          }

        /* Stop scanning before the gap.  */
787 788
	tem = BUFFER_FLOOR_OF (start_byte - 1);
	ceiling_byte = max (tem, ceiling_byte);
789 790 791

        {
          /* The termination address of the dumb loop.  */
792 793
          register unsigned char *ceiling_addr = BYTE_POS_ADDR (ceiling_byte);
          register unsigned char *cursor = BYTE_POS_ADDR (start_byte - 1);
794 795 796 797 798 799 800 801 802 803 804 805 806
          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,
807 808
                                   start_byte + cursor - base,
                                   start_byte + scan_start - base);
809 810 811 812 813 814 815

              /* Did we find the target character?  */
              if (cursor >= ceiling_addr)
                {
                  if (++count >= 0)
                    {
                      immediate_quit = 0;
816
                      return BYTE_TO_CHAR (start_byte + cursor - base);
817 818 819 820 821
                    }
                  cursor--;
                }
            }

822
	  start = BYTE_TO_CHAR (start_byte + cursor - base);
823 824 825
        }
      }

Jim Blandy's avatar
Jim Blandy committed
826 827
  immediate_quit = 0;
  if (shortage != 0)
Jim Blandy's avatar
Jim Blandy committed
828
    *shortage = count * direction;
829
  return start;
Jim Blandy's avatar
Jim Blandy committed
830
}
831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846

/* 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
847
   except in special cases.  */
Jim Blandy's avatar
Jim Blandy committed
848

849
int
850
scan_newline (EMACS_INT start, EMACS_INT start_byte, EMACS_INT limit, EMACS_INT limit_byte, register int count, int allow_quit)
851
{
852 853 854 855 856
  int direction = ((count > 0) ? 1 : -1);

  register unsigned char *cursor;
  unsigned char *base;

857
  EMACS_INT ceiling;
858 859
  register unsigned char *ceiling_addr;

860 861
  int old_immediate_quit = immediate_quit;

862 863 864
  /* The code that follows is like scan_buffer
     but checks for either newline or carriage return.  */

865 866
  if (allow_quit)
    immediate_quit++;
867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886

  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)
		    {
887
		      immediate_quit = old_immediate_quit;
888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919
		      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)
		    {
920
		      immediate_quit = old_immediate_quit;
921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937
		      /* 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);
938
  immediate_quit = old_immediate_quit;
939 940

  return count * direction;
941 942
}

Jim Blandy's avatar
Jim Blandy committed
943
int
944
find_next_newline_no_quit (EMACS_INT from, int cnt)
Jim Blandy's avatar
Jim Blandy committed
945
{
946
  return scan_buffer ('\n', from, 0, cnt, (int *) 0, 0);
947 948 949 950 951
}

/* 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.  */
952

953
int
954
find_before_next_newline (EMACS_INT from, EMACS_INT to, int cnt)
955 956 957 958 959 960
{
  int shortage;
  int pos = scan_buffer ('\n', from, to, cnt, &shortage, 1);

  if (shortage == 0)
    pos--;
961

962
  return pos;
Jim Blandy's avatar
Jim Blandy committed
963 964 965 966 967
}

/* Subroutines of Lisp buffer search functions. */

static Lisp_Object
968
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
969 970
{
  register int np;
971
  int lim, lim_byte;
Jim Blandy's avatar
Jim Blandy committed
972 973 974 975
  int n = direction;

  if (!NILP (count))
    {
976
      CHECK_NUMBER (count);
Jim Blandy's avatar
Jim Blandy committed
977 978 979
      n *= XINT (count);
    }

980
  CHECK_STRING (string);
Jim Blandy's avatar
Jim Blandy committed
981
  if (NILP (bound))
982 983 984 985 986 987
    {
      if (n > 0)
	lim = ZV, lim_byte = ZV_BYTE;
      else
	lim = BEGV, lim_byte = BEGV_BYTE;
    }
Jim Blandy's avatar
Jim Blandy committed
988 989
  else
    {
990
      CHECK_NUMBER_COERCE_MARKER (bound);
Jim Blandy's avatar
Jim Blandy committed
991
      lim = XINT (bound);
992
      if (n > 0 ? lim < PT : lim > PT)
Jim Blandy's avatar
Jim Blandy committed
993 994
	error ("Invalid search bound (wrong side of point)");
      if (lim > ZV)
995
	lim = ZV, lim_byte = ZV_BYTE;
996
      else if (lim < BEGV)
997
	lim = BEGV, lim_byte = BEGV_BYTE;
998 999
      else
	lim_byte = CHAR_TO_BYTE (lim);
Jim Blandy's avatar
Jim Blandy committed
1000 1001
    }

1002 1003 1004 1005
  /* 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;

1006
  np = search_buffer (string, PT, PT_BYTE, lim, lim_byte, n, RE,
Jim Blandy's avatar
Jim Blandy committed
1007
		      (!NILP (current_buffer->case_fold_search)
1008
		       ? current_buffer->case_canon_table
1009
		       : Qnil),
Jim Blandy's avatar
Jim Blandy committed
1010
		      (!NILP (current_buffer->case_fold_search)
1011
		       ? current_buffer->case_eqv_table
1012
		       : Qnil),
1013
		      posix);
Jim Blandy's avatar
Jim Blandy committed
1014 1015 1016
  if (np <= 0)
    {
      if (NILP (noerror))
1017 1018
	xsignal1 (Qsearch_failed, string);

Jim Blandy's avatar
Jim Blandy committed
1019 1020 1021 1022
      if (!EQ (noerror, Qt))
	{
	  if (lim < BEGV || lim > ZV)
	    abort ();
1023
	  SET_PT_BOTH (lim, lim_byte);
1024 1025 1026
	  return Qnil;
#if 0 /* This would be clean, but maybe programs depend on
	 a value of nil here.  */
1027
	  np = lim;
1028
#endif
Jim Blandy's avatar
Jim Blandy committed
1029
	}
1030 1031
      else
	return Qnil;
Jim Blandy's avatar
Jim Blandy committed
1032 1033 1034 1035 1036 1037 1038 1039 1040 1041
    }

  if (np < BEGV || np > ZV)
    abort ();

  SET_PT (np);

  return make_number (np);
}

1042 1043
/* Return 1 if REGEXP it matches just one constant string.  */