syntax.c 103 KB
Newer Older
Richard M. Stallman's avatar
Richard M. Stallman committed
1
/* GNU Emacs routines to deal with syntax tables; also word and list parsing.
2
   Copyright (C) 1985, 1987, 1993, 1994, 1995, 1997, 1998, 1999, 2001,
Glenn Morris's avatar
Glenn Morris committed
3
                 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010
4
                 Free Software Foundation, Inc.
Richard M. Stallman's avatar
Richard M. Stallman committed
5 6 7

This file is part of GNU Emacs.

8
GNU Emacs is free software: you can redistribute it and/or modify
Richard M. Stallman's avatar
Richard M. Stallman 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.
Richard M. Stallman's avatar
Richard M. Stallman 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/>.  */
Richard M. Stallman's avatar
Richard M. Stallman committed
20 21


22
#include <config.h>
Richard M. Stallman's avatar
Richard M. Stallman committed
23
#include <ctype.h>
24
#include <setjmp.h>
Richard M. Stallman's avatar
Richard M. Stallman committed
25 26 27
#include "lisp.h"
#include "commands.h"
#include "buffer.h"
28
#include "character.h"
Stefan Monnier's avatar
Stefan Monnier committed
29
#include "keymap.h"
Richard M. Stallman's avatar
Richard M. Stallman committed
30
#include "regex.h"
31 32 33 34

/* Make syntax table lookup grant data in gl_state.  */
#define SYNTAX_ENTRY_VIA_PROPERTY

Richard M. Stallman's avatar
Richard M. Stallman committed
35
#include "syntax.h"
36
#include "intervals.h"
37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90
#include "category.h"

/* Then there are seven single-bit flags that have the following meanings:
  1. This character is the first of a two-character comment-start sequence.
  2. This character is the second of a two-character comment-start sequence.
  3. This character is the first of a two-character comment-end sequence.
  4. This character is the second of a two-character comment-end sequence.
  5. This character is a prefix, for backward-prefix-chars.
  6. The char is part of a delimiter for comments of style "b".
  7. This character is part of a nestable comment sequence.
  8. The char is part of a delimiter for comments of style "c".
  Note that any two-character sequence whose first character has flag 1
  and whose second character has flag 2 will be interpreted as a comment start.

  bit 6 and 8 are used to discriminate between different comment styles.
  Languages such as C++ allow two orthogonal syntax start/end pairs
  and bit 6 is used to determine whether a comment-end or Scommentend
  ends style a or b.  Comment markers can start style a, b, c, or bc.
  Style a is always the default.
  For 2-char comment markers, the style b flag is only looked up on the second
  char of the comment marker and on the first char of the comment ender.
  For style c (like to for the nested flag), the flag can be placed on any
  one of the chars.
  */

/* These macros extract specific flags from an integer
   that holds the syntax code and the flags.  */

#define SYNTAX_FLAGS_COMSTART_FIRST(flags) (((flags) >> 16) & 1)

#define SYNTAX_FLAGS_COMSTART_SECOND(flags) (((flags) >> 17) & 1)

#define SYNTAX_FLAGS_COMEND_FIRST(flags) (((flags) >> 18) & 1)

#define SYNTAX_FLAGS_COMEND_SECOND(flags) (((flags) >> 19) & 1)

#define SYNTAX_FLAGS_PREFIX(flags) (((flags) >> 20) & 1)

#define SYNTAX_FLAGS_COMMENT_STYLEB(flags) (((flags) >> 21) & 1)
#define SYNTAX_FLAGS_COMMENT_STYLEC(flags) (((flags) >> 22) & 2)
/* FLAGS should be the flags of the main char of the comment marker, e.g.
   the second for comstart and the first for comend.  */
#define SYNTAX_FLAGS_COMMENT_STYLE(flags, other_flags) \
  (SYNTAX_FLAGS_COMMENT_STYLEB (flags) \
   | SYNTAX_FLAGS_COMMENT_STYLEC (flags) \
   | SYNTAX_FLAGS_COMMENT_STYLEC (other_flags))

#define SYNTAX_FLAGS_COMMENT_NESTED(flags) (((flags) >> 22) & 1)

/* These macros extract a particular flag for a given character.  */

#define SYNTAX_COMEND_FIRST(c) \
  (SYNTAX_FLAGS_COMEND_FIRST (SYNTAX_WITH_FLAGS (c)))
#define SYNTAX_PREFIX(c) (SYNTAX_FLAGS_PREFIX (SYNTAX_WITH_FLAGS (c)))
91 92 93 94 95 96 97

/* We use these constants in place for comment-style and
   string-ender-char to distinguish  comments/strings started by
   comment_fence and string_fence codes.  */

#define ST_COMMENT_STYLE (256 + 1)
#define ST_STRING_STYLE (256 + 2)
Richard M. Stallman's avatar
Richard M. Stallman committed
98

99
Lisp_Object Qsyntax_table_p, Qsyntax_table, Qscan_error;
Richard M. Stallman's avatar
Richard M. Stallman committed
100 101

int words_include_escapes;
102
int parse_sexp_lookup_properties;
Richard M. Stallman's avatar
Richard M. Stallman committed
103

104 105 106
/* Nonzero means `scan-sexps' treat all multibyte characters as symbol.  */
int multibyte_syntax_as_symbol;

107 108 109 110 111
/* Used as a temporary in SYNTAX_ENTRY and other macros in syntax.h,
   if not compiled with GCC.  No need to mark it, since it is used
   only very temporarily.  */
Lisp_Object syntax_temp;

112 113 114 115 116 117
/* Non-zero means an open parenthesis in column 0 is always considered
   to be the start of a defun.  Zero means an open parenthesis in
   column 0 has no special meaning.  */

int open_paren_in_column_0_is_defun_start;

118 119 120 121
/* This is the internal form of the parse state used in parse-partial-sexp.  */

struct lisp_parse_state
  {
122 123 124 125 126 127 128 129 130 131 132 133 134 135
    int depth;	   /* Depth at end of parsing.  */
    int instring;  /* -1 if not within string, else desired terminator.  */
    int incomment; /* -1 if in unnestable comment else comment nesting */
    int comstyle;  /* comment style a=0, or b=1, or ST_COMMENT_STYLE.  */
    int quoted;	   /* Nonzero if just after an escape char at end of parsing */
    int mindepth;  /* Minimum depth seen while scanning.  */
    /* Char number of most recent start-of-expression at current level */
    EMACS_INT thislevelstart;
    /* Char number of start of containing expression */
    EMACS_INT prevlevelstart;
    EMACS_INT location;	     /* Char number at which parsing stopped.  */
    EMACS_INT comstr_start;  /* Position of last comment/string starter.  */
    Lisp_Object levelstarts; /* Char numbers of starts-of-expression
				of levels (starting from outermost).  */
136 137
  };

138 139 140
/* These variables are a cache for finding the start of a defun.
   find_start_pos is the place for which the defun start was found.
   find_start_value is the defun start position found for it.
141
   find_start_value_byte is the corresponding byte position.
142 143 144 145
   find_start_buffer is the buffer it was found in.
   find_start_begv is the BEGV value when it was found.
   find_start_modiff is the value of MODIFF when it was found.  */

146 147 148
static EMACS_INT find_start_pos;
static EMACS_INT find_start_value;
static EMACS_INT find_start_value_byte;
149
static struct buffer *find_start_buffer;
150
static EMACS_INT find_start_begv;
151
static int find_start_modiff;
152 153


154 155 156 157 158 159 160
static Lisp_Object skip_chars (int, Lisp_Object, Lisp_Object, int);
static Lisp_Object skip_syntaxes (int, Lisp_Object, Lisp_Object);
static Lisp_Object scan_lists (EMACS_INT, EMACS_INT, EMACS_INT, int);
static void scan_sexps_forward (struct lisp_parse_state *,
                                EMACS_INT, EMACS_INT, EMACS_INT, int,
                                int, Lisp_Object, int);
static int in_classes (int, Lisp_Object);
161

162 163 164 165 166
/* Whether the syntax of the character C has the prefix flag set.  */
int syntax_prefix_flag_p (int c)
{
  return SYNTAX_PREFIX (c);
}
167 168 169

struct gl_state_s gl_state;		/* Global state of syntax parser.  */

170
INTERVAL interval_of (int, Lisp_Object);
171 172 173
#define INTERVALS_AT_ONCE 10		/* 1 + max-number of intervals
					   to scan to property-change.  */

174 175
/* Update gl_state to an appropriate interval which contains CHARPOS.  The
   sign of COUNT give the relative position of CHARPOS wrt the previously
176
   valid interval.  If INIT, only [be]_property fields of gl_state are
177
   valid at start, the rest is filled basing on OBJECT.
178

179
   `gl_state.*_i' are the intervals, and CHARPOS is further in the search
180 181
   direction than the intervals - or in an interval.  We update the
   current syntax-table basing on the property of this interval, and
182
   update the interval to start further than CHARPOS - or be
183
   NULL_INTERVAL.  We also update lim_property to be the next value of
184
   charpos to call this subroutine again - or be before/after the
185 186 187
   start/end of OBJECT.  */

void
188
update_syntax_table (int charpos, int count, int init, Lisp_Object object)
189 190
{
  Lisp_Object tmp_table;
191
  int cnt = 0, invalidate = 1;
192
  INTERVAL i;
193 194 195

  if (init)
    {
196
      gl_state.old_prop = Qnil;
197 198
      gl_state.start = gl_state.b_property;
      gl_state.stop = gl_state.e_property;
199 200
      i = interval_of (charpos, object);
      gl_state.backward_i = gl_state.forward_i = i;
201 202 203
      invalidate = 0;
      if (NULL_INTERVAL_P (i))
	return;
204
      /* interval_of updates only ->position of the return value, so
205
	 update the parents manually to speed up update_interval.  */
206
      while (!NULL_PARENT (i))
207 208
	{
	  if (AM_RIGHT_CHILD (i))
209
	    INTERVAL_PARENT (i)->position = i->position
210
	      - LEFT_TOTAL_LENGTH (i) + TOTAL_LENGTH (i) /* right end */
211 212
	      - TOTAL_LENGTH (INTERVAL_PARENT (i))
	      + LEFT_TOTAL_LENGTH (INTERVAL_PARENT (i));
213
	  else
214
	    INTERVAL_PARENT (i)->position = i->position - LEFT_TOTAL_LENGTH (i)
215
	      + TOTAL_LENGTH (i);
216
	  i = INTERVAL_PARENT (i);
217 218
	}
      i = gl_state.forward_i;
219
      gl_state.b_property = i->position - gl_state.offset;
220
      gl_state.e_property = INTERVAL_LAST_POS (i) - gl_state.offset;
221 222
      goto update;
    }
223
  i = count > 0 ? gl_state.forward_i : gl_state.backward_i;
224

225
  /* We are guaranteed to be called with CHARPOS either in i,
226
     or further off.  */
227 228
  if (NULL_INTERVAL_P (i))
    error ("Error in syntax_table logic for to-the-end intervals");
229
  else if (charpos < i->position)		/* Move left.  */
230 231
    {
      if (count > 0)
232
	error ("Error in syntax_table logic for intervals <-");
233
      /* Update the interval.  */
234
      i = update_interval (i, charpos);
235
      if (INTERVAL_LAST_POS (i) != gl_state.b_property)
236 237 238
	{
	  invalidate = 0;
	  gl_state.forward_i = i;
239
	  gl_state.e_property = INTERVAL_LAST_POS (i) - gl_state.offset;
240
	}
241
    }
242
  else if (charpos >= INTERVAL_LAST_POS (i)) /* Move right.  */
243 244
    {
      if (count < 0)
245
	error ("Error in syntax_table logic for intervals ->");
246
      /* Update the interval.  */
247
      i = update_interval (i, charpos);
248
      if (i->position != gl_state.e_property)
249 250 251
	{
	  invalidate = 0;
	  gl_state.backward_i = i;
252
	  gl_state.b_property = i->position - gl_state.offset;
253 254 255 256 257 258 259 260
	}
    }

  update:
  tmp_table = textget (i->plist, Qsyntax_table);

  if (invalidate)
    invalidate = !EQ (tmp_table, gl_state.old_prop); /* Need to invalidate? */
Juanma Barranquero's avatar
Juanma Barranquero committed
261

262 263 264
  if (invalidate)		/* Did not get to adjacent interval.  */
    {				/* with the same table => */
				/* invalidate the old range.  */
265 266 267
      if (count > 0)
	{
	  gl_state.backward_i = i;
268 269 270
	  gl_state.b_property = i->position - gl_state.offset;
	}
      else
271
	{
272
	  gl_state.forward_i = i;
273
	  gl_state.e_property = INTERVAL_LAST_POS (i) - gl_state.offset;
274 275
	}
    }
276

277
  if (!EQ (tmp_table, gl_state.old_prop))
278
    {
279 280 281 282 283
      gl_state.current_syntax_table = tmp_table;
      gl_state.old_prop = tmp_table;
      if (EQ (Fsyntax_table_p (tmp_table), Qt))
	{
	  gl_state.use_global = 0;
Juanma Barranquero's avatar
Juanma Barranquero committed
284
	}
285 286 287 288 289
      else if (CONSP (tmp_table))
	{
	  gl_state.use_global = 1;
	  gl_state.global_code = tmp_table;
	}
Juanma Barranquero's avatar
Juanma Barranquero committed
290
      else
291 292 293 294
	{
	  gl_state.use_global = 0;
	  gl_state.current_syntax_table = current_buffer->syntax_table;
	}
295 296 297 298 299 300 301
    }

  while (!NULL_INTERVAL_P (i))
    {
      if (cnt && !EQ (tmp_table, textget (i->plist, Qsyntax_table)))
	{
	  if (count > 0)
302 303 304 305 306 307
	    {
	      gl_state.e_property = i->position - gl_state.offset;
	      gl_state.forward_i = i;
	    }
	  else
	    {
308 309
	      gl_state.b_property
		= i->position + LENGTH (i) - gl_state.offset;
310 311 312
	      gl_state.backward_i = i;
	    }
	  return;
313
	}
Juanma Barranquero's avatar
Juanma Barranquero committed
314
      else if (cnt == INTERVALS_AT_ONCE)
315 316
	{
	  if (count > 0)
317
	    {
318 319 320 321 322 323
	      gl_state.e_property
		= i->position + LENGTH (i) - gl_state.offset
		/* e_property at EOB is not set to ZV but to ZV+1, so that
		   we can do INC(from);UPDATE_SYNTAX_TABLE_FORWARD without
		   having to check eob between the two.  */
		+ (NULL_INTERVAL_P (next_interval (i)) ? 1 : 0);
324 325 326 327 328 329 330 331
	      gl_state.forward_i = i;
	    }
	  else
	    {
	      gl_state.b_property = i->position - gl_state.offset;
	      gl_state.backward_i = i;
	    }
	  return;
332 333 334 335
	}
      cnt++;
      i = count > 0 ? next_interval (i) : previous_interval (i);
    }
336 337 338 339 340
  eassert (NULL_INTERVAL_P (i)); /* This property goes to the end.  */
  if (count > 0)
    gl_state.e_property = gl_state.stop;
  else
    gl_state.b_property = gl_state.start;
341 342
}

343 344 345
/* Returns TRUE if char at CHARPOS is quoted.
   Global syntax-table data should be set up already to be good at CHARPOS
   or after.  On return global syntax data is good for lookup at CHARPOS. */
346 347

static int
348
char_quoted (EMACS_INT charpos, EMACS_INT bytepos)
349 350
{
  register enum syntaxcode code;
351
  register EMACS_INT beg = BEGV;
352
  register int quoted = 0;
353
  EMACS_INT orig = charpos;
354

355
  while (charpos > beg)
356
    {
357
      int c;
358
      DEC_BOTH (charpos, bytepos);
359

360
      UPDATE_SYNTAX_TABLE_BACKWARD (charpos);
361
      c = FETCH_CHAR_AS_MULTIBYTE (bytepos);
362
      code = SYNTAX (c);
363 364 365 366
      if (! (code == Scharquote || code == Sescape))
	break;

      quoted = !quoted;
367
    }
368 369

  UPDATE_SYNTAX_TABLE (orig);
370 371
  return quoted;
}
372 373 374 375

/* Return the bytepos one character after BYTEPOS.
   We assume that BYTEPOS is not at the end of the buffer.  */

376
INLINE EMACS_INT
377
inc_bytepos (EMACS_INT bytepos)
378
{
379 380 381
  if (NILP (current_buffer->enable_multibyte_characters))
    return bytepos + 1;

382 383 384 385 386 387 388
  INC_POS (bytepos);
  return bytepos;
}

/* Return the bytepos one character before BYTEPOS.
   We assume that BYTEPOS is not at the start of the buffer.  */

389
INLINE EMACS_INT
390
dec_bytepos (EMACS_INT bytepos)
391
{
392 393 394
  if (NILP (current_buffer->enable_multibyte_characters))
    return bytepos - 1;

395 396 397
  DEC_POS (bytepos);
  return bytepos;
}
398

Juanma Barranquero's avatar
Juanma Barranquero committed
399
/* Return a defun-start position before POS and not too far before.
Richard M. Stallman's avatar
Richard M. Stallman committed
400 401 402
   It should be the last one before POS, or nearly the last.

   When open_paren_in_column_0_is_defun_start is nonzero,
403
   only the beginning of the buffer is treated as a defun-start.
Richard M. Stallman's avatar
Richard M. Stallman committed
404 405 406 407

   We record the information about where the scan started
   and what its result was, so that another call in the same area
   can return the same value very quickly.
408 409 410 411

   There is no promise at which position the global syntax data is
   valid on return from the subroutine, so the caller should explicitly
   update the global data.  */
412

413
static EMACS_INT
414
find_defun_start (EMACS_INT pos, EMACS_INT pos_byte)
415
{
416
  EMACS_INT opoint = PT, opoint_byte = PT_BYTE;
417

418 419 420 421 422 423
  if (!open_paren_in_column_0_is_defun_start)
    {
      find_start_value_byte = BEGV_BYTE;
      return BEGV;
    }

424 425 426 427 428 429 430 431 432 433 434 435
  /* Use previous finding, if it's valid and applies to this inquiry.  */
  if (current_buffer == find_start_buffer
      /* Reuse the defun-start even if POS is a little farther on.
	 POS might be in the next defun, but that's ok.
	 Our value may not be the best possible, but will still be usable.  */
      && pos <= find_start_pos + 1000
      && pos >= find_start_value
      && BEGV == find_start_begv
      && MODIFF == find_start_modiff)
    return find_start_value;

  /* Back up to start of line.  */
436
  scan_newline (pos, pos_byte, BEGV, BEGV_BYTE, -1, 1);
437

438 439 440
  /* We optimize syntax-table lookup for rare updates.  Thus we accept
     only those `^\s(' which are good in global _and_ text-property
     syntax-tables.  */
441
  SETUP_BUFFER_SYNTAX_TABLE ();
442
  while (PT > BEGV)
443
    {
444 445
      int c;

446 447
      /* Open-paren at start of line means we may have found our
	 defun-start.  */
448
      c = FETCH_CHAR_AS_MULTIBYTE (PT_BYTE);
449
      if (SYNTAX (c) == Sopen)
450
	{
451
	  SETUP_SYNTAX_TABLE (PT + 1, -1);	/* Try again... */
452
	  c = FETCH_CHAR_AS_MULTIBYTE (PT_BYTE);
453
	  if (SYNTAX (c) == Sopen)
454 455
	    break;
	  /* Now fallback to the default value.  */
456
	  SETUP_BUFFER_SYNTAX_TABLE ();
457
	}
458 459
      /* Move to beg of previous line.  */
      scan_newline (PT, PT_BYTE, BEGV, BEGV_BYTE, -2, 1);
460 461 462
    }

  /* Record what we found, for the next try.  */
463 464
  find_start_value = PT;
  find_start_value_byte = PT_BYTE;
465 466 467 468 469
  find_start_buffer = current_buffer;
  find_start_modiff = MODIFF;
  find_start_begv = BEGV;
  find_start_pos = pos;

470 471
  TEMP_SET_PT_BOTH (opoint, opoint_byte);

472 473 474
  return find_start_value;
}

475 476 477
/* Return the SYNTAX_COMEND_FIRST of the character before POS, POS_BYTE.  */

static int
478
prev_char_comend_first (int pos, int pos_byte)
479 480 481 482 483 484 485 486 487 488 489 490 491
{
  int c, val;

  DEC_BOTH (pos, pos_byte);
  UPDATE_SYNTAX_TABLE_BACKWARD (pos);
  c = FETCH_CHAR (pos_byte);
  val = SYNTAX_COMEND_FIRST (c);
  UPDATE_SYNTAX_TABLE_FORWARD (pos + 1);
  return val;
}

/* Return the SYNTAX_COMSTART_FIRST of the character before POS, POS_BYTE.  */

492 493 494 495 496
/* static int
 * prev_char_comstart_first (pos, pos_byte)
 *      int pos, pos_byte;
 * {
 *   int c, val;
Juanma Barranquero's avatar
Juanma Barranquero committed
497
 *
498 499 500 501 502 503 504
 *   DEC_BOTH (pos, pos_byte);
 *   UPDATE_SYNTAX_TABLE_BACKWARD (pos);
 *   c = FETCH_CHAR (pos_byte);
 *   val = SYNTAX_COMSTART_FIRST (c);
 *   UPDATE_SYNTAX_TABLE_FORWARD (pos + 1);
 *   return val;
 * } */
505

506 507 508 509 510 511 512 513 514
/* Checks whether charpos FROM is at the end of a comment.
   FROM_BYTE is the bytepos corresponding to FROM.
   Do not move back before STOP.

   Return a positive value if we find a comment ending at FROM/FROM_BYTE;
   return -1 otherwise.

   If successful, store the charpos of the comment's beginning
   into *CHARPOS_PTR, and the bytepos into *BYTEPOS_PTR.
515 516 517

   Global syntax data remains valid for backward search starting at
   the returned value (or at FROM, if the search was not successful).  */
518 519

static int
520
back_comment (EMACS_INT from, EMACS_INT from_byte, EMACS_INT stop, int comnested, int comstyle, EMACS_INT *charpos_ptr, EMACS_INT *bytepos_ptr)
521 522 523 524 525 526 527 528 529 530
{
  /* Look back, counting the parity of string-quotes,
     and recording the comment-starters seen.
     When we reach a safe place, assume that's not in a string;
     then step the main scan to the earliest comment-starter seen
     an even number of string quotes away from the safe place.

     OFROM[I] is position of the earliest comment-starter seen
     which is I+2X quotes from the comment-end.
     PARITY is current parity of quotes from the comment end.  */
531
  int string_style = -1;	/* Presumed outside of any string. */
532
  int string_lossage = 0;
533
  /* Not a real lossage: indicates that we have passed a matching comment
Juanma Barranquero's avatar
Juanma Barranquero committed
534
     starter plus a non-matching comment-ender, meaning that any matching
535 536 537 538
     comment-starter we might see later could be a false positive (hidden
     inside another comment).
     Test case:  { a (* b } c (* d *) */
  int comment_lossage = 0;
539 540 541 542
  EMACS_INT comment_end = from;
  EMACS_INT comment_end_byte = from_byte;
  EMACS_INT comstart_pos = 0;
  EMACS_INT comstart_byte;
543 544
  /* Place where the containing defun starts,
     or 0 if we didn't come across it yet.  */
545 546
  EMACS_INT defun_start = 0;
  EMACS_INT defun_start_byte = 0;
547
  register enum syntaxcode code;
548
  int nesting = 1;		/* current comment nesting */
549
  int c;
550 551 552 553 554
  int syntax = 0;

  /* FIXME: A }} comment-ender style leads to incorrect behavior
     in the case of {{ c }}} because we ignore the last two chars which are
     assumed to be comment-enders although they aren't.  */
555 556 557 558 559

  /* At beginning of range to scan, we're outside of strings;
     that determines quote parity to the comment-end.  */
  while (from != stop)
    {
560 561
      int temp_byte, prev_syntax;
      int com2start, com2end;
562
      int comstart;
563

564
      /* Move back and examine a character.  */
565
      DEC_BOTH (from, from_byte);
566 567
      UPDATE_SYNTAX_TABLE_BACKWARD (from);

568
      prev_syntax = syntax;
569
      c = FETCH_CHAR_AS_MULTIBYTE (from_byte);
570
      syntax = SYNTAX_WITH_FLAGS (c);
571 572
      code = SYNTAX (c);

573 574 575
      /* Check for 2-char comment markers.  */
      com2start = (SYNTAX_FLAGS_COMSTART_FIRST (syntax)
		   && SYNTAX_FLAGS_COMSTART_SECOND (prev_syntax)
576 577
		   && (comstyle
		       == SYNTAX_FLAGS_COMMENT_STYLE (prev_syntax, syntax))
578 579 580 581
		   && (SYNTAX_FLAGS_COMMENT_NESTED (prev_syntax)
		       || SYNTAX_FLAGS_COMMENT_NESTED (syntax)) == comnested);
      com2end = (SYNTAX_FLAGS_COMEND_FIRST (syntax)
		 && SYNTAX_FLAGS_COMEND_SECOND (prev_syntax));
582 583
      comstart = (com2start || code == Scomment);
      
584 585 586 587 588 589 590 591 592 593
      /* Nasty cases with overlapping 2-char comment markers:
	 - snmp-mode: -- c -- foo -- c --
	              --- c --
		      ------ c --
	 - c-mode:    *||*
		      |* *|* *|
		      |*| |* |*|
		      ///   */

      /* If a 2-char comment sequence partly overlaps with another,
594 595 596
	 we don't try to be clever.  E.g. |*| in C, or }% in modes that
	 have %..\n and %{..}%.  */
      if (from > stop && (com2end || comstart))
597
	{
598 599 600
	  int next = from, next_byte = from_byte, next_c, next_syntax;
	  DEC_BOTH (next, next_byte);
	  UPDATE_SYNTAX_TABLE_BACKWARD (next);
601
	  next_c = FETCH_CHAR_AS_MULTIBYTE (next_byte);
602
	  next_syntax = SYNTAX_WITH_FLAGS (next_c);
603
	  if (((comstart || comnested)
604 605 606 607
	       && SYNTAX_FLAGS_COMEND_SECOND (syntax)
	       && SYNTAX_FLAGS_COMEND_FIRST (next_syntax))
	      || ((com2end || comnested)
		  && SYNTAX_FLAGS_COMSTART_SECOND (syntax)
608 609
		  && (comstyle
		      == SYNTAX_FLAGS_COMMENT_STYLE (syntax, prev_syntax))
610 611 612
		  && SYNTAX_FLAGS_COMSTART_FIRST (next_syntax)))
	    goto lossage;
	  /* UPDATE_SYNTAX_TABLE_FORWARD (next + 1); */
613
	}
614 615 616 617 618 619 620 621 622 623 624 625 626 627 628

      if (com2start && comstart_pos == 0)
	/* We're looking at a comment starter.  But it might be a comment
	   ender as well (see snmp-mode).  The first time we see one, we
	   need to consider it as a comment starter,
	   and the subsequent times as a comment ender.  */
	com2end = 0;

      /* Turn a 2-char comment sequences into the appropriate syntax.  */
      if (com2end)
	code = Sendcomment;
      else if (com2start)
	code = Scomment;
      /* Ignore comment starters of a different style.  */
      else if (code == Scomment
629
	       && (comstyle != SYNTAX_FLAGS_COMMENT_STYLE (syntax, 0)
630
		   || SYNTAX_FLAGS_COMMENT_NESTED (syntax) != comnested))
631
	continue;
632

633 634
      /* Ignore escaped characters, except comment-enders.  */
      if (code != Sendcomment && char_quoted (from, from_byte))
635 636
	continue;

637
      switch (code)
638
	{
639 640 641 642 643 644 645 646 647 648 649 650 651 652
	case Sstring_fence:
	case Scomment_fence:
	  c = (code == Sstring_fence ? ST_STRING_STYLE : ST_COMMENT_STYLE);
	case Sstring:
	  /* Track parity of quotes.  */
	  if (string_style == -1)
	    /* Entering a string.  */
	    string_style = c;
	  else if (string_style == c)
	    /* Leaving the string.  */
	    string_style = -1;
	  else
	    /* If we have two kinds of string delimiters.
	       There's no way to grok this scanning backwards.  */
653
	    string_lossage = 1;
654
	  break;
Juanma Barranquero's avatar
Juanma Barranquero committed
655

656 657
	case Scomment:
	  /* We've already checked that it is the relevant comstyle.  */
658
	  if (string_style != -1 || comment_lossage || string_lossage)
659 660 661 662
	    /* There are odd string quotes involved, so let's be careful.
	       Test case in Pascal: " { " a { " } */
	    goto lossage;

663 664 665 666 667 668 669
	  if (!comnested)
	    {
	      /* Record best comment-starter so far.  */
	      comstart_pos = from;
	      comstart_byte = from_byte;
	    }
	  else if (--nesting <= 0)
670 671 672 673 674 675
	    /* nested comments have to be balanced, so we don't need to
	       keep looking for earlier ones.  We use here the same (slightly
	       incorrect) reasoning as below:  since it is followed by uniform
	       paired string quotes, this comment-start has to be outside of
	       strings, else the comment-end itself would be inside a string. */
	    goto done;
676 677
	  break;

678
	case Sendcomment:
679
	  if (SYNTAX_FLAGS_COMMENT_STYLE (syntax, 0) == comstyle
680
	      && ((com2end && SYNTAX_FLAGS_COMMENT_NESTED (prev_syntax))
681
		  || SYNTAX_FLAGS_COMMENT_NESTED (syntax)) == comnested)
682 683 684 685 686 687 688 689 690
	    /* This is the same style of comment ender as ours. */
	    {
	      if (comnested)
		nesting++;
	      else
		/* Anything before that can't count because it would match
		   this comment-ender rather than ours.  */
		from = stop;	/* Break out of the loop.  */
	    }
691 692 693 694 695 696 697 698 699
	  else if (comstart_pos != 0 || c != '\n')
	    /* We're mixing comment styles here, so we'd better be careful.
	       The (comstart_pos != 0 || c != '\n') check is not quite correct
	       (we should just always set comment_lossage), but removing it
	       would imply that any multiline comment in C would go through
	       lossage, which seems overkill.
	       The failure should only happen in the rare cases such as
	         { (* } *)   */
	    comment_lossage = 1;
700
	  break;
701

702 703 704 705 706 707 708 709 710 711 712
	case Sopen:
	  /* Assume a defun-start point is outside of strings.  */
	  if (open_paren_in_column_0_is_defun_start
	      && (from == stop
		  || (temp_byte = dec_bytepos (from_byte),
		      FETCH_CHAR (temp_byte) == '\n')))
	    {
	      defun_start = from;
	      defun_start_byte = from_byte;
	      from = stop;	/* Break out of the loop.  */
	    }
713
	  break;
714 715

	default:
716
	  break;
717
	}
718 719 720 721 722
    }

  if (comstart_pos == 0)
    {
      from = comment_end;
723
      from_byte = comment_end_byte;
724 725
      UPDATE_SYNTAX_TABLE_FORWARD (comment_end - 1);
    }
726 727 728
  /* If comstart_pos is set and we get here (ie. didn't jump to `lossage'
     or `done'), then we've found the beginning of the non-nested comment.  */
  else if (1)	/* !comnested */
729 730
    {
      from = comstart_pos;
731
      from_byte = comstart_byte;
732
      UPDATE_SYNTAX_TABLE_FORWARD (from - 1);
733 734 735
    }
  else
    {
736 737
      struct lisp_parse_state state;
    lossage:
738 739
      /* We had two kinds of string delimiters mixed up
	 together.  Decode this going forwards.
740
	 Scan fwd from a known safe place (beginning-of-defun)
741 742
	 to the one in question; this records where we
	 last passed a comment starter.  */
743 744 745 746 747 748
      /* If we did not already find the defun start, find it now.  */
      if (defun_start == 0)
	{
	  defun_start = find_defun_start (comment_end, comment_end_byte);
	  defun_start_byte = find_start_value_byte;
	}
749
      do
750
	{
751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772
	  scan_sexps_forward (&state,
			      defun_start, defun_start_byte,
			      comment_end, -10000, 0, Qnil, 0);
	  defun_start = comment_end;
	  if (state.incomment == (comnested ? 1 : -1)
	      && state.comstyle == comstyle)
	    from = state.comstr_start;
	  else
	    {
	      from = comment_end;
	      if (state.incomment)
		/* If comment_end is inside some other comment, maybe ours
		   is nested, so we need to try again from within the
		   surrounding comment.  Example: { a (* " *)  */
		{
		  /* FIXME: We should advance by one or two chars. */
		  defun_start = state.comstr_start + 2;
		  defun_start_byte = CHAR_TO_BYTE (defun_start);
		}
	    }
	} while (defun_start < comment_end);

773
      from_byte = CHAR_TO_BYTE (from);
774 775
      UPDATE_SYNTAX_TABLE_FORWARD (from - 1);
    }
Juanma Barranquero's avatar
Juanma Barranquero committed
776

777
 done:
778 779 780
  *charpos_ptr = from;
  *bytepos_ptr = from_byte;

781
  return (from == comment_end) ? -1 : from;
782 783
}

Richard M. Stallman's avatar
Richard M. Stallman committed
784
DEFUN ("syntax-table-p", Fsyntax_table_p, Ssyntax_table_p, 1, 1, 0,
785 786
       doc: /* Return t if OBJECT is a syntax table.
Currently, any char-table counts as a syntax table.  */)
Dan Nicolaescu's avatar
Dan Nicolaescu committed
787
  (Lisp_Object object)
Richard M. Stallman's avatar
Richard M. Stallman committed
788
{
789
  if (CHAR_TABLE_P (object)
790
      && EQ (XCHAR_TABLE (object)->purpose, Qsyntax_table))
Richard M. Stallman's avatar
Richard M. Stallman committed
791 792 793 794
    return Qt;
  return Qnil;
}

795
static void
796
check_syntax_table (Lisp_Object obj)
Richard M. Stallman's avatar
Richard M. Stallman committed
797
{
798 799
  CHECK_TYPE (CHAR_TABLE_P (obj) && EQ (XCHAR_TABLE (obj)->purpose, Qsyntax_table),
	      Qsyntax_table_p, obj);
Juanma Barranquero's avatar
Juanma Barranquero committed
800
}
Richard M. Stallman's avatar
Richard M. Stallman committed
801 802

DEFUN ("syntax-table", Fsyntax_table, Ssyntax_table, 0, 0, 0,
803 804
       doc: /* Return the current syntax table.
This is the one specified by the current buffer.  */)
Dan Nicolaescu's avatar
Dan Nicolaescu committed
805
  (void)
Richard M. Stallman's avatar
Richard M. Stallman committed
806 807 808 809 810 811
{
  return current_buffer->syntax_table;
}

DEFUN ("standard-syntax-table", Fstandard_syntax_table,
   Sstandard_syntax_table, 0, 0, 0,
812 813
       doc: /* Return the standard syntax table.
This is the one used for new buffers.  */)
Dan Nicolaescu's avatar
Dan Nicolaescu committed
814
  (void)
Richard M. Stallman's avatar
Richard M. Stallman committed
815 816 817 818 819
{
  return Vstandard_syntax_table;
}

DEFUN ("copy-syntax-table", Fcopy_syntax_table, Scopy_syntax_table, 0, 1, 0,
820 821
       doc: /* Construct a new syntax table and return it.
It is a copy of the TABLE, which defaults to the standard syntax table.  */)
Dan Nicolaescu's avatar
Dan Nicolaescu committed
822
  (Lisp_Object table)
Richard M. Stallman's avatar
Richard M. Stallman committed
823
{
824 825
  Lisp_Object copy;

Jim Blandy's avatar
Jim Blandy committed
826
  if (!NILP (table))
827 828 829 830 831
    check_syntax_table (table);
  else
    table = Vstandard_syntax_table;

  copy = Fcopy_sequence (table);
832 833 834 835 836 837 838 839 840 841

  /* Only the standard syntax table should have a default element.
     Other syntax tables should inherit from parents instead.  */
  XCHAR_TABLE (copy)->defalt = Qnil;

  /* Copied syntax tables should all have parents.
     If we copied one with no parent, such as the standard syntax table,
     use the standard syntax table as the copy's parent.  */
  if (NILP (XCHAR_TABLE (copy)->parent))
    Fset_char_table_parent (copy, Vstandard_syntax_table);
842
  return copy;
Richard M. Stallman's avatar
Richard M. Stallman committed
843 844 845
}

DEFUN ("set-syntax-table", Fset_syntax_table, Sset_syntax_table, 1, 1, 0,
846 847
       doc: /* Select a new syntax table for the current buffer.
One argument, a syntax table.  */)
Dan Nicolaescu's avatar
Dan Nicolaescu committed
848
  (Lisp_Object table)
Richard M. Stallman's avatar
Richard M. Stallman committed
849
{
850
  int idx;
851
  check_syntax_table (table);
Richard M. Stallman's avatar
Richard M. Stallman committed
852 853
  current_buffer->syntax_table = table;
  /* Indicate that this buffer now has a specified syntax table.  */
Gerd Moellmann's avatar
Gerd Moellmann committed
854 855
  idx = PER_BUFFER_VAR_IDX (syntax_table);
  SET_PER_BUFFER_VALUE_P (current_buffer, idx, 1);
Richard M. Stallman's avatar
Richard M. Stallman committed
856 857 858 859 860
  return table;
}

/* Convert a letter which signifies a syntax code
 into the code it signifies.
861
 This is used by modify-syntax-entry, and other things.  */
Richard M. Stallman's avatar
Richard M. Stallman committed
862 863 864 865 866 867

unsigned char syntax_spec_code[0400] =
  { 0377, 0377, 0377, 0377, 0377, 0377, 0377, 0377,
    0377, 0377, 0377, 0377, 0377, 0377, 0377, 0377,
    0377, 0377, 0377, 0377, 0377, 0377, 0377, 0377,
    0377, 0377, 0377, 0377, 0377, 0377, 0377, 0377,
868
    (char) Swhitespace, (char) Scomment_fence, (char) Sstring, 0377,
Richard M. Stallman's avatar
Richard M. Stallman committed
869 870 871 872 873 874
        (char) Smath, 0377, 0377, (char) Squote,
    (char) Sopen, (char) Sclose, 0377, 0377,
	0377, (char) Swhitespace, (char) Spunct, (char) Scharquote,
    0377, 0377, 0377, 0377, 0377, 0377, 0377, 0377,
    0377, 0377, 0377, 0377,
	(char) Scomment, 0377, (char) Sendcomment, 0377,
875
    (char) Sinherit, 0377, 0377, 0377, 0377, 0377, 0377, 0377,   /* @, A ... */
Richard M. Stallman's avatar
Richard M. Stallman committed
876 877 878 879 880 881
    0377, 0377, 0377, 0377, 0377, 0377, 0377, 0377,
    0377, 0377, 0377, 0377, 0377, 0377, 0377, (char) Sword,
    0377, 0377, 0377, 0377, (char) Sescape, 0377, 0377, (char) Ssymbol,
    0377, 0377, 0377, 0377, 0377, 0377, 0377, 0377,   /* `, a, ... */
    0377, 0377, 0377, 0377, 0377, 0377, 0377, 0377,
    0377, 0377, 0377, 0377, 0377, 0377, 0377, (char) Sword,
882
    0377, 0377, 0377, 0377, (char) Sstring_fence, 0377, 0377, 0377
Richard M. Stallman's avatar
Richard M. Stallman committed
883 884
  };

885
/* Indexed by syntax code, give the letter that describes it.  */
Richard M. Stallman's avatar
Richard M. Stallman committed
886

887
char syntax_code_spec[16] =
Richard M. Stallman's avatar
Richard M. Stallman committed
888
  {
889 890
    ' ', '.', 'w', '_', '(', ')', '\'', '\"', '$', '\\', '/', '<', '>', '@',
    '!', '|'
Richard M. Stallman's avatar
Richard M. Stallman committed
891
  };
Karl Heuer's avatar
Karl Heuer committed
892 893 894 895 896 897 898 899

/* Indexed by syntax code, give the object (cons of syntax code and
   nil) to be stored in syntax table.  Since these objects can be
   shared among syntax tables, we generate them in advance.  By
   sharing objects, the function `describe-syntax' can give a more
   compact listing.  */
static Lisp_Object Vsyntax_code_object;

Richard M. Stallman's avatar
Richard M. Stallman committed
900 901

DEFUN ("char-syntax", Fchar_syntax, Schar_syntax, 1, 1, 0,
902
       doc: /* Return the syntax code of CHARACTER, described by a character.
903 904
For example, if CHARACTER is a word constituent, the
character `w' (119) is returned.
905 906
The characters that correspond to various syntax codes
are listed in the documentation of `modify-syntax-entry'.  */)
Dan Nicolaescu's avatar
Dan Nicolaescu committed
907
  (Lisp_Object character)
Richard M. Stallman's avatar
Richard M. Stallman committed
908
{
909
  int char_int;
910
  CHECK_CHARACTER (character);
911
  char_int = XINT (character);
912
  SETUP_BUFFER_SYNTAX_TABLE ();
913
  return make_number (syntax_code_spec[(int) SYNTAX (char_int)]);
914 915 916
}

DEFUN ("matching-paren", Fmatching_paren, Smatching_paren, 1, 1, 0,
917
       doc: /* Return the matching parenthesis of CHARACTER, or nil if none.  */)
Dan Nicolaescu's avatar
Dan Nicolaescu committed
918
  (Lisp_Object character)
919
{
920
  int char_int, code;
921
  CHECK_NUMBER (character);
922
  char_int = XINT (character);
923
  SETUP_BUFFER_SYNTAX_TABLE ();
924
  code = SYNTAX (char_int);
925
  if (code == Sopen || code == Sclose)
926
    return SYNTAX_MATCH (char_int);
927
  return Qnil;
Richard M. Stallman's avatar
Richard M. Stallman committed
928 929
}

930
DEFUN ("string-to-syntax", Fstring_to_syntax, Sstring_to_syntax, 1, 1, 0,
931 932 933
       doc: /* Convert a syntax specification STRING into syntax cell form.
STRING should be a string as it is allowed as argument of
`modify-syntax-entry'.  Value is the equivalent cons cell
934
\(CODE . MATCHING-CHAR) that can be used as value of a `syntax-table'
935
text property.  */)
Dan Nicolaescu's avatar
Dan Nicolaescu committed
936
  (Lisp_Object string)
Richard M. Stallman's avatar
Richard M. Stallman committed
937
{
938
  register const unsigned char *p;
Richard M. Stallman's avatar
Richard M. Stallman committed
939
  register enum syntaxcode code;
940
  int val;
941
  Lisp_Object match;
Richard M. Stallman's avatar
Richard M. Stallman committed
942

943
  CHECK_STRING (string);
Richard M. Stallman's avatar
Richard M. Stallman committed
944

945
  p = SDATA (string);
Richard M. Stallman's avatar
Richard M. Stallman committed
946 947
  code = (enum syntaxcode) syntax_spec_code[*p++];
  if (((int) code & 0377) == 0377)
948
    error ("Invalid syntax description letter: %c", p[-1]);
Richard M. Stallman's avatar
Richard M. Stallman committed
949

950
  if (code == Sinherit)
951
    return Qnil;
952 953

  if (*p)
954
    {
Karl Heuer's avatar
Karl Heuer committed
955
      int len;
956
      int character = STRING_CHAR_AND_LENGTH (p, len);
Karl Heuer's avatar
Karl Heuer committed
957
      XSETINT (match, character);
958 959
      if (XFASTINT (match) == ' ')
	match = Qnil;
Karl Heuer's avatar
Karl Heuer committed
960
      p += len;
961 962
    }
  else
963
    match = Qnil;
Richard M. Stallman's avatar
Richard M. Stallman committed
964

965
  val = (int) code;
Richard M. Stallman's avatar
Richard M. Stallman committed
966 967 968 969
  while (*p)
    switch (*p++)
      {
      case '1':
970
	val |= 1 << 16;
Richard M. Stallman's avatar
Richard M. Stallman committed
971 972 973
	break;

      case '2':
974
	val |= 1 << 17;
Richard M. Stallman's avatar
Richard M. Stallman committed
975 976 977
	break;

      case '3':
978
	val |= 1 << 18;
Richard M. Stallman's avatar
Richard M. Stallman committed
979 980 981
	break;

      case '4':
982
	val |= 1 << 19;
Richard M. Stallman's avatar
Richard M. Stallman committed
983 984 985
	break;

      case 'p':
986
	val |= 1 << 20;
Richard M. Stallman's avatar
Richard M. Stallman committed
987
	break;
988 989

      case 'b':
990
	val |= 1 << 21;
991
	break;
992 993 994 995

      case 'n':
	val |= 1 << 22;
	break;
996 997 998 999

      case 'c':
	val |= 1 << 23;
	break;
Richard M. Stallman's avatar
Richard M. Stallman committed
1000
      }
Juanma Barranquero's avatar
Juanma Barranquero committed
1001

Karl Heuer's avatar
Karl Heuer committed
1002
  if (val < XVECTOR (Vsyntax_code_object)->size && NILP (match))
1003
    return XVECTOR (Vsyntax_code_object)->contents[val];
Karl Heuer's avatar
Karl Heuer committed
1004 1005
  else
    /* Since we can't use a shared object, let's make a new one.  */
1006 1007 1008
    return Fcons (make_number (val), match);
}

1009
/* I really don't know why this is interactive
Stefan Monnier's avatar
Stefan Monnier committed
1010
   help-form should at least be made useful whilst reading the second arg.  */
Juanma Barranquero's avatar
Juanma Barranquero committed
1011
DEFUN ("modify-syntax-entry", Fmodify_syntax_entry, Smodify_syntax_entry, 2, 3,
1012
  "cSet syntax for character: \nsSet syntax for %s to: ",
1013
       doc: /* Set syntax for character CHAR according to string NEWENTRY.
1014
The syntax is changed only for table SYNTAX-TABLE, which defaults to
1015
 the current buffer's syntax table.
1016
CHAR may be a cons (MIN . MAX), in which case, syntaxes of all characters
1017
in the range MIN to MAX are changed.
1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033
The first character of NEWENTRY should be one of the following:
  Space or -  whitespace syntax.    w   word constituent.
  _           symbol constituent.   .   punctuation.
  (           open-parenthesis.     )   close-parenthesis.
  "           string quote.         \\   escape.
  $           paired delimiter.     '   expression quote or prefix operator.
  <           comment starter.      >   comment ender.
  /           character-quote.      @   inherit from `standard-syntax-table'.
  |           generic string fence. !   generic comment fence.

Only single-character comment start and end sequences are represented thus.
Two-character sequences are represented as described below.
The second character of NEWENTRY is the matching parenthesis,
 used only if the first character is `(' or `)'.
Any additional characters are flags.
Defined flags are the characters 1, 2, 3, 4, b, p, and n.
1034 1035 1036 1037
 1 means CHAR is the start of a two-char comment start sequence.
 2 means CHAR is the second character of such a sequence.
 3 means CHAR is the start of a two-char comment end sequence.
 4 means CHAR is the second character of such a sequence.
1038

1039
There can be several orthogonal comment sequences.  This is to support
1040 1041
language modes such as C++.  By default, all comment sequences are of style
a, but you can set the comment sequence style to b (on the second character
1042 1043
of a comment-start, and the first character of a comment-end sequence) and/or
c (on any of its chars) using this flag:
1044
 b means CHAR is part of comment sequence b.
1045
 c means CHAR is part of comment sequence c.
1046
 n means CHAR is part of a nestable comment sequence.
1047

1048
 p means CHAR is a prefix character for `backward-prefix-chars';
1049
   such characters are treated as whitespace when they occur