syntax.c 112 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.
Paul Eggert's avatar
Paul Eggert committed
2
   Copyright (C) 1985, 1987, 1993-1995, 1997-1999, 2001-2018 Free
3
   Software Foundation, Inc.
Richard M. Stallman's avatar
Richard M. Stallman committed
4 5 6

This file is part of GNU Emacs.

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

GNU Emacs is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License
18
along with GNU Emacs.  If not, see <https://www.gnu.org/licenses/>.  */
Richard M. Stallman's avatar
Richard M. Stallman committed
19 20


21
#include <config.h>
22

Richard M. Stallman's avatar
Richard M. Stallman committed
23
#include "lisp.h"
24
#include "character.h"
25
#include "buffer.h"
26
#include "regex-emacs.h"
Richard M. Stallman's avatar
Richard M. Stallman committed
27
#include "syntax.h"
28
#include "intervals.h"
29 30
#include "category.h"

31 32 33 34 35
/* Make syntax table lookup grant data in gl_state.  */
#define SYNTAX(c) syntax_property (c, 1)
#define SYNTAX_ENTRY(c) syntax_property_entry (c, 1)
#define SYNTAX_WITH_FLAGS(c) syntax_property_with_flags (c, 1)

Paul Eggert's avatar
Paul Eggert committed
36
/* Eight single-bit flags have the following meanings:
37 38 39 40 41 42 43 44 45 46 47
  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.

Paul Eggert's avatar
Paul Eggert committed
48
  Bits 6 and 8 discriminate among different comment styles.
49
  Languages such as C++ allow two orthogonal syntax start/end pairs
Paul Eggert's avatar
Paul Eggert committed
50
  and bit 6 determines whether a comment-end or Scommentend
51 52
  ends style a or b.  Comment markers can start style a, b, c, or bc.
  Style a is always the default.
Paul Eggert's avatar
Paul Eggert committed
53
  For 2-char comment markers, the style b flag is looked up only on the second
54
  char of the comment marker and on the first char of the comment ender.
Paul Eggert's avatar
Paul Eggert committed
55 56
  For style c (like the nested flag), the flag can be placed on any of
  the chars.  */
57

58
/* These functions extract specific flags from an integer
59 60
   that holds the syntax code and the flags.  */

61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81
static bool
SYNTAX_FLAGS_COMSTART_FIRST (int flags)
{
  return (flags >> 16) & 1;
}
static bool
SYNTAX_FLAGS_COMSTART_SECOND (int flags)
{
  return (flags >> 17) & 1;
}
static bool
SYNTAX_FLAGS_COMEND_FIRST (int flags)
{
  return (flags >> 18) & 1;
}
static bool
SYNTAX_FLAGS_COMEND_SECOND (int flags)
{
  return (flags >> 19) & 1;
}
static bool
82 83 84 85 86
SYNTAX_FLAGS_COMSTARTEND_FIRST (int flags)
{
  return (flags & 0x50000) != 0;
}
static bool
87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110
SYNTAX_FLAGS_PREFIX (int flags)
{
  return (flags >> 20) & 1;
}
static bool
SYNTAX_FLAGS_COMMENT_STYLEB (int flags)
{
  return (flags >> 21) & 1;
}
static bool
SYNTAX_FLAGS_COMMENT_STYLEC (int flags)
{
  return (flags >> 23) & 1;
}
static int
SYNTAX_FLAGS_COMMENT_STYLEC2 (int flags)
{
  return (flags >> 22) & 2; /* SYNTAX_FLAGS_COMMENT_STYLEC (flags) * 2 */
}
static bool
SYNTAX_FLAGS_COMMENT_NESTED (int flags)
{
  return (flags >> 22) & 1;
}
111 112 113

/* FLAGS should be the flags of the main char of the comment marker, e.g.
   the second for comstart and the first for comend.  */
114 115 116 117 118 119 120
static int
SYNTAX_FLAGS_COMMENT_STYLE (int flags, int other_flags)
{
  return (SYNTAX_FLAGS_COMMENT_STYLEB (flags)
	  | SYNTAX_FLAGS_COMMENT_STYLEC2 (flags)
	  | SYNTAX_FLAGS_COMMENT_STYLEC2 (other_flags));
}
121

122
/* Extract a particular flag for a given character.  */
123

124 125 126 127 128
static bool
SYNTAX_COMEND_FIRST (int c)
{
  return SYNTAX_FLAGS_COMEND_FIRST (SYNTAX_WITH_FLAGS (c));
}
129 130

/* We use these constants in place for comment-style and
131
   string-ender-char to distinguish comments/strings started by
132 133
   comment_fence and string_fence codes.  */

134 135 136 137 138
enum
  {
    ST_COMMENT_STYLE = 256 + 1,
    ST_STRING_STYLE = 256 + 2
  };
Richard M. Stallman's avatar
Richard M. Stallman committed
139

140 141 142 143
/* This is the internal form of the parse state used in parse-partial-sexp.  */

struct lisp_parse_state
  {
144
    EMACS_INT depth;	/* Depth at end of parsing.  */
145
    int instring;  /* -1 if not within string, else desired terminator.  */
146
    EMACS_INT incomment; /* -1 if in unnestable comment else comment nesting */
147
    int comstyle;  /* comment style a=0, or b=1, or ST_COMMENT_STYLE.  */
Paul Eggert's avatar
Paul Eggert committed
148
    bool quoted;   /* True if just after an escape char at end of parsing.  */
149
    EMACS_INT mindepth;	/* Minimum depth seen while scanning.  */
150
    /* Char number of most recent start-of-expression at current level */
151
    ptrdiff_t thislevelstart;
152
    /* Char number of start of containing expression */
153 154
    ptrdiff_t prevlevelstart;
    ptrdiff_t location;	     /* Char number at which parsing stopped.  */
155
    ptrdiff_t location_byte; /* Corresponding byte position.  */
156
    ptrdiff_t comstr_start;  /* Position of last comment/string starter.  */
157 158
    Lisp_Object levelstarts; /* Char numbers of starts-of-expression
				of levels (starting from outermost).  */
159 160 161 162
    int prev_syntax; /* Syntax of previous position scanned, when
                        that position (potentially) holds the first char
                        of a 2-char construct, i.e. comment delimiter
                        or Sescape, etc.  Smax otherwise. */
163 164
  };

165 166 167
/* 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.
168
   find_start_value_byte is the corresponding byte position.
169 170 171 172
   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.  */

173 174 175
static ptrdiff_t find_start_pos;
static ptrdiff_t find_start_value;
static ptrdiff_t find_start_value_byte;
176
static struct buffer *find_start_buffer;
177 178
static ptrdiff_t find_start_begv;
static EMACS_INT find_start_modiff;
179 180


Paul Eggert's avatar
Paul Eggert committed
181 182 183
static Lisp_Object skip_chars (bool, Lisp_Object, Lisp_Object, bool);
static Lisp_Object skip_syntaxes (bool, Lisp_Object, Lisp_Object);
static Lisp_Object scan_lists (EMACS_INT, EMACS_INT, EMACS_INT, bool);
184
static void scan_sexps_forward (struct lisp_parse_state *,
185
                                ptrdiff_t, ptrdiff_t, ptrdiff_t, EMACS_INT,
186 187
                                bool, int);
static void internalize_parse_state (Lisp_Object, struct lisp_parse_state *);
Paul Eggert's avatar
Paul Eggert committed
188
static bool in_classes (int, Lisp_Object);
189
static void parse_sexp_propertize (ptrdiff_t charpos);
Paul Eggert's avatar
Paul Eggert committed
190 191

/* This setter is used only in this file, so it can be private.  */
192
static void
Paul Eggert's avatar
Paul Eggert committed
193 194
bset_syntax_table (struct buffer *b, Lisp_Object val)
{
195
  b->syntax_table_ = val;
Paul Eggert's avatar
Paul Eggert committed
196
}
197

198
/* Whether the syntax of the character C has the prefix flag set.  */
Paul Eggert's avatar
Paul Eggert committed
199 200
bool
syntax_prefix_flag_p (int c)
201
{
202
  return SYNTAX_FLAGS_PREFIX (SYNTAX_WITH_FLAGS (c));
203
}
204 205 206

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

207
enum { INTERVALS_AT_ONCE = 10 };	/* 1 + max-number of intervals
208 209
					   to scan to property-change.  */

210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254
/* Set the syntax entry VAL for char C in table TABLE.  */

static void
SET_RAW_SYNTAX_ENTRY (Lisp_Object table, int c, Lisp_Object val)
{
  CHAR_TABLE_SET (table, c, val);
}

/* Set the syntax entry VAL for char-range RANGE in table TABLE.
   RANGE is a cons (FROM . TO) specifying the range of characters.  */

static void
SET_RAW_SYNTAX_ENTRY_RANGE (Lisp_Object table, Lisp_Object range,
			    Lisp_Object val)
{
  Fset_char_table_range (table, range, val);
}

/* Extract the information from the entry for character C
   in the current syntax table.  */

static Lisp_Object
SYNTAX_MATCH (int c)
{
  Lisp_Object ent = SYNTAX_ENTRY (c);
  return CONSP (ent) ? XCDR (ent) : Qnil;
}

/* This should be called with FROM at the start of forward
   search, or after the last position of the backward search.  It
   makes sure that the first char is picked up with correct table, so
   one does not need to call UPDATE_SYNTAX_TABLE immediately after the
   call.
   Sign of COUNT gives the direction of the search.
 */

static void
SETUP_SYNTAX_TABLE (ptrdiff_t from, ptrdiff_t count)
{
  SETUP_BUFFER_SYNTAX_TABLE ();
  gl_state.b_property = BEGV;
  gl_state.e_property = ZV + 1;
  gl_state.object = Qnil;
  gl_state.offset = 0;
  if (parse_sexp_lookup_properties)
255
    {
256
      if (count > 0)
257
	update_syntax_table_forward (from, true, Qnil);
258
      else if (from > BEGV)
259
	{
260 261
	  update_syntax_table (from - 1, count, true, Qnil);
	  parse_sexp_propertize (from - 1);
262 263
	}
    }
264 265 266 267 268 269
}

/* Same as above, but in OBJECT.  If OBJECT is nil, use current buffer.
   If it is t (which is only used in fast_c_string_match_ignore_case),
   ignore properties altogether.

270 271 272 273
   This is meant for regex-emacs.c to use.  For buffers, regex-emacs.c
   passes arguments to the UPDATE_SYNTAX_TABLE functions which are
   relative to BEGV.  So if it is a buffer, we set the offset field to
   BEGV.  */
274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310

void
SETUP_SYNTAX_TABLE_FOR_OBJECT (Lisp_Object object,
			       ptrdiff_t from, ptrdiff_t count)
{
  SETUP_BUFFER_SYNTAX_TABLE ();
  gl_state.object = object;
  if (BUFFERP (gl_state.object))
    {
      struct buffer *buf = XBUFFER (gl_state.object);
      gl_state.b_property = 1;
      gl_state.e_property = BUF_ZV (buf) - BUF_BEGV (buf) + 1;
      gl_state.offset = BUF_BEGV (buf) - 1;
    }
  else if (NILP (gl_state.object))
    {
      gl_state.b_property = 1;
      gl_state.e_property = ZV - BEGV + 1;
      gl_state.offset = BEGV - 1;
    }
  else if (EQ (gl_state.object, Qt))
    {
      gl_state.b_property = 0;
      gl_state.e_property = PTRDIFF_MAX;
      gl_state.offset = 0;
    }
  else
    {
      gl_state.b_property = 0;
      gl_state.e_property = 1 + SCHARS (gl_state.object);
      gl_state.offset = 0;
    }
  if (parse_sexp_lookup_properties)
    update_syntax_table (from + gl_state.offset - (count <= 0),
			 count, 1, gl_state.object);
}

311 312
/* Update gl_state to an appropriate interval which contains CHARPOS.  The
   sign of COUNT give the relative position of CHARPOS wrt the previously
313
   valid interval.  If INIT, only [be]_property fields of gl_state are
314
   valid at start, the rest is filled basing on OBJECT.
315

316
   `gl_state.*_i' are the intervals, and CHARPOS is further in the search
317 318
   direction than the intervals - or in an interval.  We update the
   current syntax-table basing on the property of this interval, and
319
   update the interval to start further than CHARPOS - or be
Dmitry Antipov's avatar
Dmitry Antipov committed
320
   NULL.  We also update lim_property to be the next value of
321
   charpos to call this subroutine again - or be before/after the
322 323 324
   start/end of OBJECT.  */

void
Paul Eggert's avatar
Paul Eggert committed
325
update_syntax_table (ptrdiff_t charpos, EMACS_INT count, bool init,
326
		     Lisp_Object object)
327 328
{
  Lisp_Object tmp_table;
Paul Eggert's avatar
Paul Eggert committed
329
  int cnt = 0;
330
  bool invalidate = true;
331
  INTERVAL i;
332 333 334

  if (init)
    {
335
      gl_state.old_prop = Qnil;
336 337
      gl_state.start = gl_state.b_property;
      gl_state.stop = gl_state.e_property;
338 339
      i = interval_of (charpos, object);
      gl_state.backward_i = gl_state.forward_i = i;
340
      invalidate = false;
Dmitry Antipov's avatar
Dmitry Antipov committed
341
      if (!i)
342
	return;
343
      /* interval_of updates only ->position of the return value, so
344
	 update the parents manually to speed up update_interval.  */
345
      while (!NULL_PARENT (i))
346 347
	{
	  if (AM_RIGHT_CHILD (i))
348
	    INTERVAL_PARENT (i)->position = i->position
349
	      - LEFT_TOTAL_LENGTH (i) + TOTAL_LENGTH (i) /* right end */
350 351
	      - TOTAL_LENGTH (INTERVAL_PARENT (i))
	      + LEFT_TOTAL_LENGTH (INTERVAL_PARENT (i));
352
	  else
353
	    INTERVAL_PARENT (i)->position = i->position - LEFT_TOTAL_LENGTH (i)
354
	      + TOTAL_LENGTH (i);
355
	  i = INTERVAL_PARENT (i);
356 357
	}
      i = gl_state.forward_i;
358
      gl_state.b_property = i->position - gl_state.offset;
359
      gl_state.e_property = INTERVAL_LAST_POS (i) - gl_state.offset;
360 361
      goto update;
    }
362
  i = count > 0 ? gl_state.forward_i : gl_state.backward_i;
363

364
  /* We are guaranteed to be called with CHARPOS either in i,
365
     or further off.  */
Dmitry Antipov's avatar
Dmitry Antipov committed
366
  if (!i)
367
    error ("Error in syntax_table logic for to-the-end intervals");
368
  else if (charpos < i->position)		/* Move left.  */
369 370
    {
      if (count > 0)
371
	error ("Error in syntax_table logic for intervals <-");
372
      /* Update the interval.  */
373
      i = update_interval (i, charpos);
374
      if (INTERVAL_LAST_POS (i) != gl_state.b_property)
375
	{
376
	  invalidate = false;
377
	  gl_state.forward_i = i;
378
	  gl_state.e_property = INTERVAL_LAST_POS (i) - gl_state.offset;
379
	}
380
    }
381
  else if (charpos >= INTERVAL_LAST_POS (i)) /* Move right.  */
382 383
    {
      if (count < 0)
384
	error ("Error in syntax_table logic for intervals ->");
385
      /* Update the interval.  */
386
      i = update_interval (i, charpos);
387
      if (i->position != gl_state.e_property)
388
	{
389
	  invalidate = false;
390
	  gl_state.backward_i = i;
391
	  gl_state.b_property = i->position - gl_state.offset;
392 393 394 395 396 397 398 399
	}
    }

  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
400

401 402 403
  if (invalidate)		/* Did not get to adjacent interval.  */
    {				/* with the same table => */
				/* invalidate the old range.  */
404 405 406
      if (count > 0)
	{
	  gl_state.backward_i = i;
407 408 409
	  gl_state.b_property = i->position - gl_state.offset;
	}
      else
410
	{
411
	  gl_state.forward_i = i;
412
	  gl_state.e_property = INTERVAL_LAST_POS (i) - gl_state.offset;
413 414
	}
    }
415

416
  if (!EQ (tmp_table, gl_state.old_prop))
417
    {
418 419 420 421 422
      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
423
	}
424 425 426 427 428
      else if (CONSP (tmp_table))
	{
	  gl_state.use_global = 1;
	  gl_state.global_code = tmp_table;
	}
Juanma Barranquero's avatar
Juanma Barranquero committed
429
      else
430 431
	{
	  gl_state.use_global = 0;
Tom Tromey's avatar
Tom Tromey committed
432
	  gl_state.current_syntax_table = BVAR (current_buffer, syntax_table);
433
	}
434 435
    }

Dmitry Antipov's avatar
Dmitry Antipov committed
436
  while (i)
437 438 439 440
    {
      if (cnt && !EQ (tmp_table, textget (i->plist, Qsyntax_table)))
	{
	  if (count > 0)
441 442 443 444 445 446
	    {
	      gl_state.e_property = i->position - gl_state.offset;
	      gl_state.forward_i = i;
	    }
	  else
	    {
447 448
	      gl_state.b_property
		= i->position + LENGTH (i) - gl_state.offset;
449 450 451
	      gl_state.backward_i = i;
	    }
	  return;
452
	}
Juanma Barranquero's avatar
Juanma Barranquero committed
453
      else if (cnt == INTERVALS_AT_ONCE)
454 455
	{
	  if (count > 0)
456
	    {
457 458 459 460 461
	      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.  */
Dmitry Antipov's avatar
Dmitry Antipov committed
462
		+ (next_interval (i) ? 0 : 1);
463 464 465 466 467 468 469 470
	      gl_state.forward_i = i;
	    }
	  else
	    {
	      gl_state.b_property = i->position - gl_state.offset;
	      gl_state.backward_i = i;
	    }
	  return;
471 472 473 474
	}
      cnt++;
      i = count > 0 ? next_interval (i) : previous_interval (i);
    }
Dmitry Antipov's avatar
Dmitry Antipov committed
475
  eassert (i == NULL); /* This property goes to the end.  */
476
  if (count > 0)
477 478 479 480
    {
      gl_state.e_property = gl_state.stop;
      gl_state.forward_i = i;
    }
481 482
  else
    gl_state.b_property = gl_state.start;
483
}
484 485 486 487

static void
parse_sexp_propertize (ptrdiff_t charpos)
{
488 489 490 491 492 493
  EMACS_INT zv = ZV;
  if (syntax_propertize__done <= charpos
      && syntax_propertize__done < zv)
    {
      EMACS_INT modiffs = CHARS_MODIFF;
      safe_call1 (Qinternal__syntax_propertize,
494
		  make_fixnum (min (zv, 1 + charpos)));
495 496 497 498 499 500 501 502 503 504 505 506 507
      if (modiffs != CHARS_MODIFF)
	error ("parse-sexp-propertize-function modified the buffer!");
      if (syntax_propertize__done <= charpos
	  && syntax_propertize__done < zv)
	error ("parse-sexp-propertize-function did not move"
	       " syntax-propertize--done");
      SETUP_SYNTAX_TABLE (charpos, 1);
    }
  else if (gl_state.e_property > syntax_propertize__done)
    {
      gl_state.e_property = syntax_propertize__done;
      gl_state.e_property_truncated = true;
    }
Stefan Monnier's avatar
Stefan Monnier committed
508 509
  else if (gl_state.e_property_truncated
	   && gl_state.e_property < syntax_propertize__done)
510 511 512 513
    { /* When moving backward, e_property might be set without resetting
	 e_property_truncated, so the e_property_truncated flag may
	 occasionally be left raised spuriously.  This should be rare.  */
      gl_state.e_property_truncated = false;
514
      update_syntax_table_forward (charpos, false, Qnil);
515
    }
516 517 518
}

void
519
update_syntax_table_forward (ptrdiff_t charpos, bool init,
520
			     Lisp_Object object)
521
{
522
  if (gl_state.e_property_truncated)
523
    {
524 525
      eassert (NILP (object));
      eassert (charpos >= gl_state.e_property);
526
      parse_sexp_propertize (charpos);
527 528 529 530
    }
  else
    {
      update_syntax_table (charpos, 1, init, object);
531 532
      if (NILP (object) && gl_state.e_property > syntax_propertize__done)
	parse_sexp_propertize (charpos);
533 534
    }
}
535

Paul Eggert's avatar
Paul Eggert committed
536
/* Returns true if char at CHARPOS is quoted.
537
   Global syntax-table data should be set up already to be good at CHARPOS
538
   or after.  On return global syntax data is good for lookup at CHARPOS.  */
539

Paul Eggert's avatar
Paul Eggert committed
540
static bool
541
char_quoted (ptrdiff_t charpos, ptrdiff_t bytepos)
542
{
Paul Eggert's avatar
Paul Eggert committed
543 544 545
  enum syntaxcode code;
  ptrdiff_t beg = BEGV;
  bool quoted = 0;
546
  ptrdiff_t orig = charpos;
547

548
  while (charpos > beg)
549
    {
550
      int c;
551
      DEC_BOTH (charpos, bytepos);
552

553
      UPDATE_SYNTAX_TABLE_BACKWARD (charpos);
554
      c = FETCH_CHAR_AS_MULTIBYTE (bytepos);
555
      code = SYNTAX (c);
556 557 558 559
      if (! (code == Scharquote || code == Sescape))
	break;

      quoted = !quoted;
560
    }
561 562

  UPDATE_SYNTAX_TABLE (orig);
563 564
  return quoted;
}
565 566 567 568

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

569
static ptrdiff_t
570
dec_bytepos (ptrdiff_t bytepos)
571
{
Tom Tromey's avatar
Tom Tromey committed
572
  if (NILP (BVAR (current_buffer, enable_multibyte_characters)))
573 574
    return bytepos - 1;

575 576 577
  DEC_POS (bytepos);
  return bytepos;
}
578

Juanma Barranquero's avatar
Juanma Barranquero committed
579
/* Return a defun-start position before POS and not too far before.
Richard M. Stallman's avatar
Richard M. Stallman committed
580 581 582
   It should be the last one before POS, or nearly the last.

   When open_paren_in_column_0_is_defun_start is nonzero,
583
   only the beginning of the buffer is treated as a defun-start.
Richard M. Stallman's avatar
Richard M. Stallman committed
584 585 586 587

   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.
588 589 590 591

   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.  */
592

593 594
static ptrdiff_t
find_defun_start (ptrdiff_t pos, ptrdiff_t pos_byte)
595
{
596
  ptrdiff_t opoint = PT, opoint_byte = PT_BYTE;
597 598 599 600 601 602 603 604 605 606 607 608

  /* 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;

609 610 611
  if (!NILP (Vcomment_use_syntax_ppss))
    {
      EMACS_INT modiffs = CHARS_MODIFF;
612
      Lisp_Object ppss = call1 (Qsyntax_ppss, make_fixnum (pos));
613 614 615
      if (modiffs != CHARS_MODIFF)
	error ("syntax-ppss modified the buffer!");
      TEMP_SET_PT_BOTH (opoint, opoint_byte);
616 617
      Lisp_Object boc = Fnth (make_fixnum (8), ppss);
      if (FIXED_OR_FLOATP (boc))
618
        {
Tom Tromey's avatar
Tom Tromey committed
619
          find_start_value = XFIXNUM (boc);
620 621 622 623 624 625 626 627 628
          find_start_value_byte = CHAR_TO_BYTE (find_start_value);
        }
      else
        {
          find_start_value = pos;
          find_start_value_byte = pos_byte;
        }
      goto found;
    }
629 630 631 632 633 634 635
  if (!open_paren_in_column_0_is_defun_start)
    {
      find_start_value = BEGV;
      find_start_value_byte = BEGV_BYTE;
      goto found;
    }

636
  /* Back up to start of line.  */
637
  scan_newline (pos, pos_byte, BEGV, BEGV_BYTE, -1, 1);
638

639 640 641
  /* We optimize syntax-table lookup for rare updates.  Thus we accept
     only those `^\s(' which are good in global _and_ text-property
     syntax-tables.  */
642
  SETUP_BUFFER_SYNTAX_TABLE ();
643
  while (PT > BEGV)
644
    {
645 646
      /* Open-paren at start of line means we may have found our
	 defun-start.  */
Paul Eggert's avatar
Paul Eggert committed
647
      int c = FETCH_CHAR_AS_MULTIBYTE (PT_BYTE);
648
      if (SYNTAX (c) == Sopen)
649
	{
650
	  SETUP_SYNTAX_TABLE (PT + 1, -1);	/* Try again... */
651
	  c = FETCH_CHAR_AS_MULTIBYTE (PT_BYTE);
652
	  if (SYNTAX (c) == Sopen)
653 654
	    break;
	  /* Now fallback to the default value.  */
655
	  SETUP_BUFFER_SYNTAX_TABLE ();
656
	}
657 658
      /* Move to beg of previous line.  */
      scan_newline (PT, PT_BYTE, BEGV, BEGV_BYTE, -2, 1);
659 660 661
    }

  /* Record what we found, for the next try.  */
662 663
  find_start_value = PT;
  find_start_value_byte = PT_BYTE;
664 665 666
  TEMP_SET_PT_BOTH (opoint, opoint_byte);

 found:
667 668 669 670 671 672 673 674
  find_start_buffer = current_buffer;
  find_start_modiff = MODIFF;
  find_start_begv = BEGV;
  find_start_pos = pos;

  return find_start_value;
}

675 676
/* Return the SYNTAX_COMEND_FIRST of the character before POS, POS_BYTE.  */

Paul Eggert's avatar
Paul Eggert committed
677
static bool
678
prev_char_comend_first (ptrdiff_t pos, ptrdiff_t pos_byte)
679
{
Paul Eggert's avatar
Paul Eggert committed
680 681
  int c;
  bool val;
682 683 684 685 686 687 688 689 690

  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;
}

Paul Eggert's avatar
Paul Eggert committed
691
/* Check whether charpos FROM is at the end of a comment.
692 693 694
   FROM_BYTE is the bytepos corresponding to FROM.
   Do not move back before STOP.

Paul Eggert's avatar
Paul Eggert committed
695
   Return true if we find a comment ending at FROM/FROM_BYTE.
696 697 698

   If successful, store the charpos of the comment's beginning
   into *CHARPOS_PTR, and the bytepos into *BYTEPOS_PTR.
699 700 701

   Global syntax data remains valid for backward search starting at
   the returned value (or at FROM, if the search was not successful).  */
702

Paul Eggert's avatar
Paul Eggert committed
703 704 705 706
static bool
back_comment (ptrdiff_t from, ptrdiff_t from_byte, ptrdiff_t stop,
	      bool comnested, int comstyle, ptrdiff_t *charpos_ptr,
	      ptrdiff_t *bytepos_ptr)
707 708 709 710 711 712 713 714 715 716
{
  /* 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.  */
717
  int string_style = -1;	/* Presumed outside of any string.  */
Paul Eggert's avatar
Paul Eggert committed
718
  bool string_lossage = 0;
719
  /* Not a real lossage: indicates that we have passed a matching comment
Juanma Barranquero's avatar
Juanma Barranquero committed
720
     starter plus a non-matching comment-ender, meaning that any matching
721 722 723
     comment-starter we might see later could be a false positive (hidden
     inside another comment).
     Test case:  { a (* b } c (* d *) */
Paul Eggert's avatar
Paul Eggert committed
724
  bool comment_lossage = 0;
725 726 727
  ptrdiff_t comment_end = from;
  ptrdiff_t comment_end_byte = from_byte;
  ptrdiff_t comstart_pos = 0;
728
  ptrdiff_t comstart_byte;
729 730
  /* Place where the containing defun starts,
     or 0 if we didn't come across it yet.  */
731 732
  ptrdiff_t defun_start = 0;
  ptrdiff_t defun_start_byte = 0;
Paul Eggert's avatar
Paul Eggert committed
733
  enum syntaxcode code;
734
  ptrdiff_t nesting = 1;		/* Current comment nesting.  */
735
  int c;
736
  int syntax = 0;
Paul Eggert's avatar
Paul Eggert committed
737
  unsigned short int quit_count = 0;
738 739 740 741

  /* 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.  */
742 743 744 745 746

  /* At beginning of range to scan, we're outside of strings;
     that determines quote parity to the comment-end.  */
  while (from != stop)
    {
747
      rarely_quit (++quit_count);
Paul Eggert's avatar
Paul Eggert committed
748

749
      ptrdiff_t temp_byte;
Paul Eggert's avatar
Paul Eggert committed
750 751
      int prev_syntax;
      bool com2start, com2end, comstart;
752

753
      /* Move back and examine a character.  */
754
      DEC_BOTH (from, from_byte);
755 756
      UPDATE_SYNTAX_TABLE_BACKWARD (from);

757
      prev_syntax = syntax;
758
      c = FETCH_CHAR_AS_MULTIBYTE (from_byte);
759
      syntax = SYNTAX_WITH_FLAGS (c);
760 761
      code = SYNTAX (c);

762 763 764
      /* Check for 2-char comment markers.  */
      com2start = (SYNTAX_FLAGS_COMSTART_FIRST (syntax)
		   && SYNTAX_FLAGS_COMSTART_SECOND (prev_syntax)
765 766
		   && (comstyle
		       == SYNTAX_FLAGS_COMMENT_STYLE (prev_syntax, syntax))
767 768 769 770
		   && (SYNTAX_FLAGS_COMMENT_NESTED (prev_syntax)
		       || SYNTAX_FLAGS_COMMENT_NESTED (syntax)) == comnested);
      com2end = (SYNTAX_FLAGS_COMEND_FIRST (syntax)
		 && SYNTAX_FLAGS_COMEND_SECOND (prev_syntax));
771
      comstart = (com2start || code == Scomment);
772

773 774 775 776 777 778 779 780 781 782
      /* 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,
783 784 785
	 we don't try to be clever.  E.g. |*| in C, or }% in modes that
	 have %..\n and %{..}%.  */
      if (from > stop && (com2end || comstart))
786
	{
787
	  ptrdiff_t next = from, next_byte = from_byte;
788
	  int next_c, next_syntax;
789 790
	  DEC_BOTH (next, next_byte);
	  UPDATE_SYNTAX_TABLE_BACKWARD (next);
791
	  next_c = FETCH_CHAR_AS_MULTIBYTE (next_byte);
792
	  next_syntax = SYNTAX_WITH_FLAGS (next_c);
793
	  if (((comstart || comnested)
794 795 796 797
	       && SYNTAX_FLAGS_COMEND_SECOND (syntax)
	       && SYNTAX_FLAGS_COMEND_FIRST (next_syntax))
	      || ((com2end || comnested)
		  && SYNTAX_FLAGS_COMSTART_SECOND (syntax)
798 799
		  && (comstyle
		      == SYNTAX_FLAGS_COMMENT_STYLE (syntax, prev_syntax))
800 801 802
		  && SYNTAX_FLAGS_COMSTART_FIRST (next_syntax)))
	    goto lossage;
	  /* UPDATE_SYNTAX_TABLE_FORWARD (next + 1); */
803
	}
804 805 806 807 808 809 810 811 812 813 814 815 816 817 818

      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
819
	       && (comstyle != SYNTAX_FLAGS_COMMENT_STYLE (syntax, 0)
820
		   || SYNTAX_FLAGS_COMMENT_NESTED (syntax) != comnested))
821
	continue;
822

823 824 825 826
      /* Ignore escaped characters, except comment-enders which cannot
         be escaped.  */
      if ((Vcomment_end_can_be_escaped || code != Sendcomment)
          && char_quoted (from, from_byte))
827 828
	continue;

829
      switch (code)
830
	{
831 832 833
	case Sstring_fence:
	case Scomment_fence:
	  c = (code == Sstring_fence ? ST_STRING_STYLE : ST_COMMENT_STYLE);
834
	  FALLTHROUGH;
835 836 837 838 839 840 841 842 843 844 845
	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.  */
846
	    string_lossage = 1;
847
	  break;
Juanma Barranquero's avatar
Juanma Barranquero committed
848

849 850
	case Scomment:
	  /* We've already checked that it is the relevant comstyle.  */
851
	  if (string_style != -1 || comment_lossage || string_lossage)
852 853 854 855
	    /* There are odd string quotes involved, so let's be careful.
	       Test case in Pascal: " { " a { " } */
	    goto lossage;

856 857 858 859 860 861 862
	  if (!comnested)
	    {
	      /* Record best comment-starter so far.  */
	      comstart_pos = from;
	      comstart_byte = from_byte;
	    }
	  else if (--nesting <= 0)
863 864 865 866 867 868
	    /* 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;
869 870
	  break;

871
	case Sendcomment:
872
	  if (SYNTAX_FLAGS_COMMENT_STYLE (syntax, 0) == comstyle
873
	      && ((com2end && SYNTAX_FLAGS_COMMENT_NESTED (prev_syntax))
874
		  || SYNTAX_FLAGS_COMMENT_NESTED (syntax)) == comnested)
875 876 877 878 879 880 881 882 883
	    /* 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.  */
	    }
884 885 886 887 888 889 890 891 892
	  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;
893
	  break;
894

895 896 897
	case Sopen:
	  /* Assume a defun-start point is outside of strings.  */
	  if (open_paren_in_column_0_is_defun_start
898
              && NILP (Vcomment_use_syntax_ppss)
899 900 901 902 903 904 905 906
	      && (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.  */
	    }
907
	  break;
908 909

	default:
910
	  break;
911
	}
912 913 914 915 916
    }

  if (comstart_pos == 0)
    {
      from = comment_end;
917
      from_byte = comment_end_byte;
918
      UPDATE_SYNTAX_TABLE_FORWARD (comment_end);
919
    }
920 921 922
  /* 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 */
923 924
    {
      from = comstart_pos;
925
      from_byte = comstart_byte;
926
      UPDATE_SYNTAX_TABLE_FORWARD (from - 1);
927
    }
928
  else lossage:
929
    {
930
      struct lisp_parse_state state;
931
      bool adjusted = true;
932 933
      /* We had two kinds of string delimiters mixed up
	 together.  Decode this going forwards.
934
	 Scan fwd from a known safe place (beginning-of-defun)
935 936
	 to the one in question; this records where we
	 last passed a comment starter.  */
937 938 939 940 941
      /* 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;
942
	  adjusted = (defun_start > BEGV);
943
	}
944
      do
945
	{
946
          internalize_parse_state (Qnil, &state);
947 948
	  scan_sexps_forward (&state,
			      defun_start, defun_start_byte,
949
			      comment_end, TYPE_MINIMUM (EMACS_INT),
950
			      0, 0);
951
	  defun_start = comment_end;
952 953 954 955
	  if (!adjusted)
	    {
	      adjusted = true;
	      find_start_value
Tom Tromey's avatar
Tom Tromey committed
956
		= CONSP (state.levelstarts) ? XFIXNUM (XCAR (state.levelstarts))
957 958 959 960 961
		: state.thislevelstart >= 0 ? state.thislevelstart
		: find_start_value;
	      find_start_value_byte = CHAR_TO_BYTE (find_start_value);
	    }

962 963 964 965 966 967 968 969 970 971 972
	  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 (* " *)  */
		{
973
		  /* FIXME: We should advance by one or two chars.  */