search.c 96.2 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
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>
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;

Richard M. Stallman's avatar
Richard M. Stallman committed
93
Lisp_Object Vsearch_spaces_regexp;
94

95 96 97 98 99
/* 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;

100 101 102 103 104 105 106 107 108 109 110 111
static void set_search_regs P_ ((EMACS_INT, EMACS_INT));
static void save_search_regs P_ ((void));
static EMACS_INT simple_search P_ ((int, unsigned char *, int, int, 
				    Lisp_Object, EMACS_INT, EMACS_INT,
				    EMACS_INT, EMACS_INT));
static EMACS_INT boyer_moore P_ ((int, unsigned char *, int, int, 
				  Lisp_Object, Lisp_Object,
				  EMACS_INT, EMACS_INT,
				  EMACS_INT, EMACS_INT, int));
static EMACS_INT search_buffer P_ ((Lisp_Object, EMACS_INT, EMACS_INT,
				    EMACS_INT, EMACS_INT, int, int,
				    Lisp_Object, Lisp_Object, int));
112
static void matcher_overflow () NO_RETURN;
113

Jim Blandy's avatar
Jim Blandy committed
114 115 116 117 118 119
static void
matcher_overflow ()
{
  error ("Stack overflow in regexp matcher");
}

120 121 122
/* 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.
123
   TRANSLATE is a translation table for ignoring case, or nil for none.
124 125 126 127 128
   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
129
   for this pattern.  0 means backtrack only enough to get a valid match.
130 131

   The behavior also depends on Vsearch_spaces_regexp.  */
Jim Blandy's avatar
Jim Blandy committed
132

133
static void
Kenichi Handa's avatar
Kenichi Handa committed
134
compile_pattern_1 (cp, pattern, translate, regp, posix)
135
     struct regexp_cache *cp;
Jim Blandy's avatar
Jim Blandy committed
136
     Lisp_Object pattern;
137
     Lisp_Object translate;
138
     struct re_registers *regp;
139
     int posix;
Jim Blandy's avatar
Jim Blandy committed
140
{
141
  char *val;
142
  reg_syntax_t old;
Jim Blandy's avatar
Jim Blandy committed
143

144
  cp->regexp = Qnil;
145
  cp->buf.translate = (! NILP (translate) ? translate : make_number (0));
146
  cp->posix = posix;
147
  cp->buf.multibyte = STRING_MULTIBYTE (pattern);
Kenichi Handa's avatar
Kenichi Handa committed
148
  cp->buf.charset_unibyte = charset_unibyte;
149 150 151 152 153
  if (STRINGP (Vsearch_spaces_regexp))
    cp->whitespace_regexp = Vsearch_spaces_regexp;
  else
    cp->whitespace_regexp = Qnil;

154 155 156 157 158
  /* 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;  */
159
  old = re_set_syntax (RE_SYNTAX_EMACS
160
		       | (posix ? 0 : RE_NO_POSIX_BACKTRACKING));
161 162 163 164 165

  if (STRINGP (Vsearch_spaces_regexp))
    re_set_whitespace_regexp (SDATA (Vsearch_spaces_regexp));
  else
    re_set_whitespace_regexp (NULL);
166

Kenichi Handa's avatar
Kenichi Handa committed
167 168
  val = (char *) re_compile_pattern ((char *) SDATA (pattern),
				     SBYTES (pattern), &cp->buf);
169

170 171 172 173
  /* 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;

174 175
  re_set_whitespace_regexp (NULL);

176
  re_set_syntax (old);
177
  /* UNBLOCK_INPUT;  */
Jim Blandy's avatar
Jim Blandy committed
178
  if (val)
179
    xsignal1 (Qinvalid_regexp, build_string (val));
Jim Blandy's avatar
Jim Blandy committed
180

181 182 183
  cp->regexp = Fcopy_sequence (pattern);
}

184 185 186 187 188 189 190
/* Shrink each compiled regexp buffer in the cache
   to the size actually used right now.
   This is called from garbage collection.  */

void
shrink_regexp_cache ()
{
191
  struct regexp_cache *cp;
192 193 194 195 196

  for (cp = searchbuf_head; cp != 0; cp = cp->next)
    {
      cp->buf.allocated = cp->buf.used;
      cp->buf.buffer
197
	= (unsigned char *) xrealloc (cp->buf.buffer, cp->buf.used);
198 199 200
    }
}

201 202
/* Clear the regexp cache w.r.t. a particular syntax table,
   because it was changed.
203 204 205 206 207 208 209 210 211
   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
clear_regexp_cache ()
{
  int i;

  for (i = 0; i < REGEXP_CACHE_SIZE; ++i)
212 213 214 215 216
    /* It's tempting to compare with the syntax-table we've actually changd,
       but it's not sufficient because char-table inheritance mewans that
       modifying one syntax-table can change others at the same time.  */
    if (!EQ (searchbufs[i].syntax_table, Qt))
      searchbufs[i].regexp = Qnil;
217 218
}

219
/* Compile a regexp if necessary, but first check to see if there's one in
220 221
   the cache.
   PATTERN is the pattern to compile.
222
   TRANSLATE is a translation table for ignoring case, or nil for none.
223 224 225 226 227 228
   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.  */
229 230

struct re_pattern_buffer *
231
compile_pattern (pattern, regp, translate, posix, multibyte)
232 233
     Lisp_Object pattern;
     struct re_registers *regp;
234
     Lisp_Object translate;
235
     int posix, multibyte;
236 237 238 239 240 241
{
  struct regexp_cache *cp, **cpp;

  for (cpp = &searchbuf_head; ; cpp = &cp->next)
    {
      cp = *cpp;
242 243
      /* 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
244 245 246
	 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.  */
247
      if (NILP (cp->regexp))
248
	goto compile_it;
249
      if (SCHARS (cp->regexp) == SCHARS (pattern)
250
	  && STRING_MULTIBYTE (cp->regexp) == STRING_MULTIBYTE (pattern)
251
	  && !NILP (Fstring_equal (cp->regexp, pattern))
252
	  && EQ (cp->buf.translate, (! NILP (translate) ? translate : make_number (0)))
Karl Heuer's avatar
Karl Heuer committed
253
	  && cp->posix == posix
254 255
	  && (EQ (cp->syntax_table, Qt)
	      || EQ (cp->syntax_table, current_buffer->syntax_table))
Kenichi Handa's avatar
Kenichi Handa committed
256 257
	  && !NILP (Fequal (cp->whitespace_regexp, Vsearch_spaces_regexp))
	  && cp->buf.charset_unibyte == charset_unibyte)
258 259
	break;

260 261 262
      /* 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.  */
263 264
      if (cp->next == 0)
	{
265
	compile_it:
Kenichi Handa's avatar
Kenichi Handa committed
266
	  compile_pattern_1 (cp, pattern, translate, regp, posix);
267 268 269 270 271 272 273 274 275 276
	  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
277

278 279 280 281 282
  /* 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);

Kenichi Handa's avatar
Kenichi Handa committed
283 284 285 286
  /* The compiled pattern can be used both for mulitbyte and unibyte
     target.  But, we have to tell which the pattern is used for. */
  cp->buf.target_multibyte = multibyte;

287
  return &cp->buf;
Jim Blandy's avatar
Jim Blandy committed
288 289 290
}


291 292
static Lisp_Object
looking_at_1 (string, posix)
Jim Blandy's avatar
Jim Blandy committed
293
     Lisp_Object string;
294
     int posix;
Jim Blandy's avatar
Jim Blandy committed
295 296 297
{
  Lisp_Object val;
  unsigned char *p1, *p2;
298
  EMACS_INT s1, s2;
Jim Blandy's avatar
Jim Blandy committed
299
  register int i;
300
  struct re_pattern_buffer *bufp;
Jim Blandy's avatar
Jim Blandy committed
301

302 303 304
  if (running_asynch_code)
    save_search_regs ();

305 306 307 308
  /* 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;

309
  CHECK_STRING (string);
310 311 312
  bufp = compile_pattern (string,
			  (NILP (Vinhibit_changing_match_data)
			   ? &search_regs : NULL),
313
			  (!NILP (current_buffer->case_fold_search)
314
			   ? current_buffer->case_canon_table : Qnil),
315 316
			  posix,
			  !NILP (current_buffer->enable_multibyte_characters));
Jim Blandy's avatar
Jim Blandy committed
317 318 319 320 321 322 323 324

  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;
325
  s1 = GPT_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
326
  p2 = GAP_END_ADDR;
327
  s2 = ZV_BYTE - GPT_BYTE;
Jim Blandy's avatar
Jim Blandy committed
328 329 330
  if (s1 < 0)
    {
      p2 = p1;
331
      s2 = ZV_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
332 333 334 335
      s1 = 0;
    }
  if (s2 < 0)
    {
336
      s1 = ZV_BYTE - BEGV_BYTE;
Jim Blandy's avatar
Jim Blandy committed
337 338
      s2 = 0;
    }
339 340

  re_match_object = Qnil;
341

342
  i = re_match_2 (bufp, (char *) p1, s1, (char *) p2, s2,
343 344 345
		  PT_BYTE - BEGV_BYTE,
		  (NILP (Vinhibit_changing_match_data)
		   ? &search_regs : NULL),
346
		  ZV_BYTE - BEGV_BYTE);
347
  immediate_quit = 0;
348

Jim Blandy's avatar
Jim Blandy committed
349 350 351 352
  if (i == -2)
    matcher_overflow ();

  val = (0 <= i ? Qt : Qnil);
353
  if (NILP (Vinhibit_changing_match_data) && i >= 0)
354 355 356 357 358 359 360 361
    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);
	}
362 363 364 365 366

  /* 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
367 368 369
  return val;
}

370
DEFUN ("looking-at", Flooking_at, Slooking_at, 1, 1, 0,
371 372 373 374 375
       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.  */)
     (regexp)
376
     Lisp_Object regexp;
377
{
378
  return looking_at_1 (regexp, 0);
379 380 381
}

DEFUN ("posix-looking-at", Fposix_looking_at, Sposix_looking_at, 1, 1, 0,
382 383 384 385 386 387
       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.  */)
     (regexp)
388
     Lisp_Object regexp;
389
{
390
  return looking_at_1 (regexp, 1);
391 392 393 394
}

static Lisp_Object
string_match_1 (regexp, string, start, posix)
Jim Blandy's avatar
Jim Blandy committed
395
     Lisp_Object regexp, string, start;
396
     int posix;
Jim Blandy's avatar
Jim Blandy committed
397 398
{
  int val;
399
  struct re_pattern_buffer *bufp;
400
  EMACS_INT pos, pos_byte;
401
  int i;
Jim Blandy's avatar
Jim Blandy committed
402

403 404 405
  if (running_asynch_code)
    save_search_regs ();

406 407
  CHECK_STRING (regexp);
  CHECK_STRING (string);
Jim Blandy's avatar
Jim Blandy committed
408 409

  if (NILP (start))
410
    pos = 0, pos_byte = 0;
Jim Blandy's avatar
Jim Blandy committed
411 412
  else
    {
413
      int len = SCHARS (string);
Jim Blandy's avatar
Jim Blandy committed
414

415
      CHECK_NUMBER (start);
416 417 418 419
      pos = XINT (start);
      if (pos < 0 && -pos <= len)
	pos = len + pos;
      else if (0 > pos || pos > len)
Jim Blandy's avatar
Jim Blandy committed
420
	args_out_of_range (string, start);
421
      pos_byte = string_char_to_byte (string, pos);
Jim Blandy's avatar
Jim Blandy committed
422 423
    }

424 425 426 427
  /* 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;

428 429 430
  bufp = compile_pattern (regexp,
			  (NILP (Vinhibit_changing_match_data)
			   ? &search_regs : NULL),
431
			  (!NILP (current_buffer->case_fold_search)
432
			   ? current_buffer->case_canon_table : Qnil),
433 434
			  posix,
			  STRING_MULTIBYTE (string));
Jim Blandy's avatar
Jim Blandy committed
435
  immediate_quit = 1;
436
  re_match_object = string;
437

438 439 440
  val = re_search (bufp, (char *) SDATA (string),
		   SBYTES (string), pos_byte,
		   SBYTES (string) - pos_byte,
441 442
		   (NILP (Vinhibit_changing_match_data)
		    ? &search_regs : NULL));
Jim Blandy's avatar
Jim Blandy committed
443
  immediate_quit = 0;
444 445 446 447 448

  /* 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
449 450 451
  if (val == -2)
    matcher_overflow ();
  if (val < 0) return Qnil;
452

453 454 455 456 457 458 459 460 461
  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]);
	}
462 463

  return make_number (string_byte_to_char (string, val));
Jim Blandy's avatar
Jim Blandy committed
464
}
Richard M. Stallman's avatar
Richard M. Stallman committed
465

466
DEFUN ("string-match", Fstring_match, Sstring_match, 2, 3, 0,
467
       doc: /* Return index of start of first match for REGEXP in STRING, or nil.
Richard M. Stallman's avatar
Richard M. Stallman committed
468
Matching ignores case if `case-fold-search' is non-nil.
469 470 471
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
472 473 474 475
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. */)
476
     (regexp, string, start)
477 478 479 480 481 482
     Lisp_Object regexp, string, start;
{
  return string_match_1 (regexp, string, start, 0);
}

DEFUN ("posix-string-match", Fposix_string_match, Sposix_string_match, 2, 3, 0,
483 484 485 486 487 488 489 490
       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.  */)
     (regexp, string, start)
491 492 493 494 495
     Lisp_Object regexp, string, start;
{
  return string_match_1 (regexp, string, start, 1);
}

Richard M. Stallman's avatar
Richard M. Stallman committed
496 497 498 499 500 501 502 503 504
/* 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
fast_string_match (regexp, string)
     Lisp_Object regexp, string;
{
  int val;
505
  struct re_pattern_buffer *bufp;
Richard M. Stallman's avatar
Richard M. Stallman committed
506

507 508
  bufp = compile_pattern (regexp, 0, Qnil,
			  0, STRING_MULTIBYTE (string));
Richard M. Stallman's avatar
Richard M. Stallman committed
509
  immediate_quit = 1;
510
  re_match_object = string;
511

512 513 514
  val = re_search (bufp, (char *) SDATA (string),
		   SBYTES (string), 0,
		   SBYTES (string), 0);
Richard M. Stallman's avatar
Richard M. Stallman committed
515 516 517
  immediate_quit = 0;
  return val;
}
Karl Heuer's avatar
Karl Heuer committed
518 519 520

/* Match REGEXP against STRING, searching all of STRING ignoring case,
   and return the index of the match, or negative on failure.
521 522
   This does not clobber the match data.
   We assume that STRING contains single-byte characters.  */
Karl Heuer's avatar
Karl Heuer committed
523 524 525 526

extern Lisp_Object Vascii_downcase_table;

int
527
fast_c_string_match_ignore_case (regexp, string)
Karl Heuer's avatar
Karl Heuer committed
528
     Lisp_Object regexp;
529
     const char *string;
Karl Heuer's avatar
Karl Heuer committed
530 531 532 533 534
{
  int val;
  struct re_pattern_buffer *bufp;
  int len = strlen (string);

535
  regexp = string_make_unibyte (regexp);
536
  re_match_object = Qt;
Karl Heuer's avatar
Karl Heuer committed
537
  bufp = compile_pattern (regexp, 0,
538
			  Vascii_canon_table, 0,
539
			  0);
Karl Heuer's avatar
Karl Heuer committed
540 541 542 543 544
  immediate_quit = 1;
  val = re_search (bufp, string, len, 0, len, 0);
  immediate_quit = 0;
  return val;
}
545 546 547 548 549 550 551 552 553 554

/* Like fast_string_match but ignore case.  */

int
fast_string_match_ignore_case (regexp, string)
     Lisp_Object regexp, string;
{
  int val;
  struct re_pattern_buffer *bufp;

555
  bufp = compile_pattern (regexp, 0, Vascii_canon_table,
556 557 558 559 560 561 562 563 564 565
			  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;
}
566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581

/* Match REGEXP atainst the characters after POS to LIMIT, and return
   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
fast_looking_at (regexp, pos, pos_byte, limit, limit_byte, string)
     Lisp_Object regexp;
     EMACS_INT pos, pos_byte, limit, limit_byte;
     Lisp_Object string;
{
  int multibyte;
  struct re_pattern_buffer *buf;
  unsigned char *p1, *p2;
582
  EMACS_INT s1, s2;
583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633
  EMACS_INT len;
  
  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;
}

634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671

/* 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
newline_cache_on_off (buf)
     struct buffer *buf;
{
  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
672 673

   If we find COUNT instances, set *SHORTAGE to zero, and return the
Richard M. Stallman's avatar
Richard M. Stallman committed
674
   position past the COUNTth match.  Note that for reverse motion
Richard M. Stallman's avatar
Richard M. Stallman committed
675
   this is not the same as the usual convention for Emacs motion commands.
Jim Blandy's avatar
Jim Blandy committed
676

677 678
   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
679

680 681 682
   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
683
int
684 685
scan_buffer (target, start, end, count, shortage, allow_quit)
     register int target;
686
     EMACS_INT start, end;
687 688
     int count;
     int *shortage;
689
     int allow_quit;
Jim Blandy's avatar
Jim Blandy committed
690
{
691
  struct region_cache *newline_cache;
692
  int direction;
Jim Blandy's avatar
Jim Blandy committed
693

694 695 696 697 698 699 700 701 702 703
  if (count > 0)
    {
      direction = 1;
      if (! end) end = ZV;
    }
  else
    {
      direction = -1;
      if (! end) end = BEGV;
    }
Jim Blandy's avatar
Jim Blandy committed
704

705 706
  newline_cache_on_off (current_buffer);
  newline_cache = current_buffer->newline_cache;
Jim Blandy's avatar
Jim Blandy committed
707 708 709 710

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

711
  immediate_quit = allow_quit;
Jim Blandy's avatar
Jim Blandy committed
712

Jim Blandy's avatar
Jim Blandy committed
713
  if (count > 0)
714
    while (start != end)
Jim Blandy's avatar
Jim Blandy committed
715
      {
716 717 718 719 720
        /* 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.  */
721 722 723
	EMACS_INT ceiling_byte = CHAR_TO_BYTE (end) - 1;
	EMACS_INT start_byte = CHAR_TO_BYTE (start);
	EMACS_INT tem;
724 725 726 727 728 729 730 731

        /* 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
732 733
                   (current_buffer, newline_cache, start_byte, &next_change))
              start_byte = next_change;
734
            immediate_quit = allow_quit;
735

736 737 738
            /* START should never be after END.  */
            if (start_byte > ceiling_byte)
              start_byte = ceiling_byte;
739 740 741

            /* Now the text after start is an unknown region, and
               next_change is the position of the next known region. */
742
            ceiling_byte = min (next_change - 1, ceiling_byte);
743 744 745 746 747 748
          }

        /* 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.  */
749 750
	tem = BUFFER_CEILING_OF (start_byte);
	ceiling_byte = min (tem, ceiling_byte);
751 752

        {
753
          /* The termination address of the dumb loop.  */
754 755 756 757
          register unsigned char *ceiling_addr
	    = BYTE_POS_ADDR (ceiling_byte) + 1;
          register unsigned char *cursor
	    = BYTE_POS_ADDR (start_byte);
758 759 760 761 762 763 764 765 766 767 768 769 770 771
          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,
772 773
                                   start_byte + scan_start - base,
                                   start_byte + cursor - base);
774 775 776 777 778 779 780

              /* Did we find the target character?  */
              if (cursor < ceiling_addr)
                {
                  if (--count == 0)
                    {
                      immediate_quit = 0;
781
                      return BYTE_TO_CHAR (start_byte + cursor - base + 1);
782 783 784 785 786
                    }
                  cursor++;
                }
            }

787
          start = BYTE_TO_CHAR (start_byte + cursor - base);
788
        }
Jim Blandy's avatar
Jim Blandy committed
789 790
      }
  else
791 792 793
    while (start > end)
      {
        /* The last character to check before the next obstacle.  */
794 795 796
	EMACS_INT ceiling_byte = CHAR_TO_BYTE (end);
	EMACS_INT start_byte = CHAR_TO_BYTE (start);
	EMACS_INT tem;
797 798 799 800 801 802 803

        /* Consult the newline cache, if appropriate.  */
        if (target == '\n' && newline_cache)
          {
            int next_change;
            immediate_quit = 0;
            while (region_cache_backward
804 805
                   (current_buffer, newline_cache, start_byte, &next_change))
              start_byte = next_change;
806
            immediate_quit = allow_quit;
807 808

            /* Start should never be at or before end.  */
809 810
            if (start_byte <= ceiling_byte)
              start_byte = ceiling_byte + 1;
811 812 813

            /* Now the text before start is an unknown region, and
               next_change is the position of the next known region. */
814
            ceiling_byte = max (next_change, ceiling_byte);
815 816 817
          }

        /* Stop scanning before the gap.  */
818 819
	tem = BUFFER_FLOOR_OF (start_byte - 1);
	ceiling_byte = max (tem, ceiling_byte);
820 821 822

        {
          /* The termination address of the dumb loop.  */
823 824
          register unsigned char *ceiling_addr = BYTE_POS_ADDR (ceiling_byte);
          register unsigned char *cursor = BYTE_POS_ADDR (start_byte - 1);
825 826 827 828 829 830 831 832 833 834 835 836 837
          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,
838 839
                                   start_byte + cursor - base,
                                   start_byte + scan_start - base);
840 841 842 843 844 845 846

              /* Did we find the target character?  */
              if (cursor >= ceiling_addr)
                {
                  if (++count >= 0)
                    {
                      immediate_quit = 0;
847
                      return BYTE_TO_CHAR (start_byte + cursor - base);
848 849 850 851 852
                    }
                  cursor--;
                }
            }

853
	  start = BYTE_TO_CHAR (start_byte + cursor - base);
854 855 856
        }
      }

Jim Blandy's avatar
Jim Blandy committed
857 858
  immediate_quit = 0;
  if (shortage != 0)
Jim Blandy's avatar
Jim Blandy committed
859
    *shortage = count * direction;
860
  return start;
Jim Blandy's avatar
Jim Blandy committed
861
}
862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877

/* 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
878
   except in special cases.  */
Jim Blandy's avatar
Jim Blandy committed
879

880
int
881
scan_newline (start, start_byte, limit, limit_byte, count, allow_quit)
882 883
     EMACS_INT start, start_byte;
     EMACS_INT limit, limit_byte;
884 885
     register int count;
     int allow_quit;
886
{
887 888 889 890 891
  int direction = ((count > 0) ? 1 : -1);

  register unsigned char *cursor;
  unsigned char *base;

892
  EMACS_INT ceiling;
893 894
  register unsigned char *ceiling_addr;

895 896
  int old_immediate_quit = immediate_quit;

897 898 899
  /* The code that follows is like scan_buffer
     but checks for either newline or carriage return.  */

900 901
  if (allow_quit)
    immediate_quit++;
902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921

  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)
		    {
922
		      immediate_quit = old_immediate_quit;
923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954
		      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)
		    {
955
		      immediate_quit = old_immediate_quit;
956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972
		      /* 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);
973
  immediate_quit = old_immediate_quit;
974 975

  return count * direction;
976 977
}

Jim Blandy's avatar
Jim Blandy committed
978
int
979
find_next_newline_no_quit (from, cnt)
980 981
     EMACS_INT from;
     int cnt;
Jim Blandy's avatar
Jim Blandy committed
982
{
983
  return scan_buffer ('\n', from, 0, cnt, (int *) 0, 0);
984 985 986 987 988
}

/* 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.  */
989

990 991
int
find_before_next_newline (from, to, cnt)
992 993
     EMACS_INT from, to;
     int cnt;
994 995 996 997 998 999
{
  int shortage;
  int pos = scan_buffer ('\n', from, to, cnt, &shortage, 1);

  if (shortage == 0)
    pos--;
1000

1001
  return pos;
Jim Blandy's avatar
Jim Blandy committed
1002 1003 1004 1005 1006
}

/* Subroutines of Lisp buffer search functions. */

static Lisp_Object
1007
search_command (string, bound, noerror, count, direction, RE, posix)
Jim Blandy's avatar
Jim Blandy committed
1008 1009 1010
     Lisp_Object string, bound, noerror, count;
     int direction;
     int RE;
1011
     int posix;
Jim Blandy's avatar
Jim Blandy committed
1012 1013
{
  register int np;
1014
  int lim, lim_byte;
Jim Blandy's avatar
Jim Blandy committed
1015 1016 1017 1018
  int n = direction;

  if (!NILP (count))
    {
1019
      CHECK_NUMBER (count);
Jim Blandy's avatar
Jim Blandy committed
1020 1021 1022
      n *= XINT (count);
    }

1023
  CHECK_STRING (string);
Jim Blandy's avatar
Jim Blandy committed
1024
  if (NILP (bound))
1025 1026 1027 1028 1029 1030
    {
      if (n > 0)
	lim = ZV, lim_byte = ZV_BYTE;
      else
	lim = BEGV, lim_byte = BEGV_BYTE;
    }
Jim Blandy's avatar
Jim Blandy committed
1031 1032
  else
    {
1033
      CHECK_NUMBER_COERCE_MARKER (bound);
Jim Blandy's avatar
Jim Blandy committed
1034
      lim = XINT (bound);
1035
      if (n > 0 ? lim < PT : lim > PT)
Jim Blandy's avatar
Jim Blandy committed
1036 1037
	error ("Invalid search bound (wrong side of point)");
      if (lim > ZV)
1038
	lim = ZV, lim_byte = ZV_BYTE;
1039
      else if (lim < BEGV)
1040
	lim = BEGV, lim_byte = BEGV_BYTE;
1041 1042
      else
	lim_byte = CHAR_TO_BYTE (lim);
Jim Blandy's avatar
Jim Blandy committed
1043 1044
    }

1045 1046 1047 1048
  /* 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;

1049
  np = search_buffer (string, PT, PT_BYTE, lim, lim_byte, n, RE,
Jim Blandy's avatar
Jim Blandy committed
1050
		      (!NILP (current_buffer->case_fold_search)
1051
		       ? current_buffer->case_canon_table
1052
		       : Qnil),
Jim Blandy's avatar
Jim Blandy committed
1053
		      (!NILP (current_buffer->case_fold_search)
1054
		       ? current_buffer->case_eqv_table
1055
		       : Qnil),
1056
		      posix);
Jim Blandy's avatar
Jim Blandy committed
1057 1058 1059
  if (np <= 0)
    {
      if (NILP (noerror))
1060 1061
	xsignal1 (Qsearch_failed, string);

Jim Blandy's avatar
Jim Blandy committed
1062 1063 1064 1065
      if (!EQ (noerror, Qt))
	{
	  if (lim < BEGV || lim > ZV)
	    abort ();
1066
	  SET_PT_BOTH (lim, lim_byte);
1067 1068 1069
	  return Qnil;
#if 0 /* This would be clean, but maybe programs depend on
	 a value of nil here.  */
1070
	  np = lim;