bidi.c 75.9 KB
Newer Older
1
/* Low-level bidirectional buffer/string-scanning functions for GNU Emacs.
2
   Copyright (C) 2000-2001, 2004-2005, 2009-2011
3
   Free Software Foundation, Inc.
Eli Zaretskii's avatar
Eli Zaretskii committed
4 5 6

This file is part of GNU Emacs.

Eli Zaretskii's avatar
Eli Zaretskii committed
7
GNU Emacs is free software: you can redistribute it and/or modify
Eli Zaretskii's avatar
Eli Zaretskii committed
8
it under the terms of the GNU General Public License as published by
Eli Zaretskii's avatar
Eli Zaretskii committed
9 10
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
Eli Zaretskii's avatar
Eli Zaretskii 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
Eli Zaretskii's avatar
Eli Zaretskii committed
18
along with GNU Emacs.  If not, see <http://www.gnu.org/licenses/>.  */
Eli Zaretskii's avatar
Eli Zaretskii committed
19

20 21 22
/* Written by Eli Zaretskii <eliz@gnu.org>.

   A sequential implementation of the Unicode Bidirectional algorithm,
23
   (UBA) as per UAX#9, a part of the Unicode Standard.
Eli Zaretskii's avatar
Eli Zaretskii committed
24 25

   Unlike the reference and most other implementations, this one is
26 27
   designed to be called once for every character in the buffer or
   string.
Eli Zaretskii's avatar
Eli Zaretskii committed
28

29
   The main entry point is bidi_move_to_visually_next.  Each time it
Eli Zaretskii's avatar
Eli Zaretskii committed
30 31 32
   is called, it finds the next character in the visual order, and
   returns its information in a special structure.  The caller is then
   expected to process this character for display or any other
33 34 35
   purposes, and call bidi_move_to_visually_next for the next
   character.  See the comments in bidi_move_to_visually_next for more
   details about its algorithm that finds the next visual-order
Eli Zaretskii's avatar
Eli Zaretskii committed
36 37
   character by resolving their levels on the fly.

Eli Zaretskii's avatar
Eli Zaretskii committed
38
   Two other entry points are bidi_paragraph_init and
39 40 41 42
   bidi_mirror_char.  The first determines the base direction of a
   paragraph, while the second returns the mirrored version of its
   argument character.

Eli Zaretskii's avatar
Eli Zaretskii committed
43 44 45 46 47
   A few auxiliary entry points are used to initialize the bidi
   iterator for iterating an object (buffer or string), push and pop
   the bidi iterator state, and save and restore the state of the bidi
   cache.

48 49 50 51
   If you want to understand the code, you will have to read it
   together with the relevant portions of UAX#9.  The comments include
   references to UAX#9 rules, for that very reason.

Eli Zaretskii's avatar
Eli Zaretskii committed
52 53 54 55 56 57 58
   A note about references to UAX#9 rules: if the reference says
   something like "X9/Retaining", it means that you need to refer to
   rule X9 and to its modifications decribed in the "Implementation
   Notes" section of UAX#9, under "Retaining Format Codes".  */

#include <config.h>
#include <stdio.h>
59 60
#include <setjmp.h>

Eli Zaretskii's avatar
Eli Zaretskii committed
61 62 63 64 65 66 67
#include "lisp.h"
#include "buffer.h"
#include "character.h"
#include "dispextern.h"

static int bidi_initialized = 0;

68
static Lisp_Object bidi_type_table, bidi_mirror_table;
Eli Zaretskii's avatar
Eli Zaretskii committed
69 70 71 72 73 74 75 76 77 78 79 80 81

#define LRM_CHAR   0x200E
#define RLM_CHAR   0x200F
#define BIDI_EOB   -1

/* Data type for describing the bidirectional character categories.  */
typedef enum {
  UNKNOWN_BC,
  NEUTRAL,
  WEAK,
  STRONG
} bidi_category_t;

82
extern int bidi_ignore_explicit_marks_for_paragraph_level EXTERNALLY_VISIBLE;
Eli Zaretskii's avatar
Eli Zaretskii committed
83 84
int bidi_ignore_explicit_marks_for_paragraph_level = 1;

Eli Zaretskii's avatar
Eli Zaretskii committed
85
static Lisp_Object paragraph_start_re, paragraph_separate_re;
86
static Lisp_Object Qparagraph_start, Qparagraph_separate;
Eli Zaretskii's avatar
Eli Zaretskii committed
87

88 89 90 91

/***********************************************************************
			Utilities
 ***********************************************************************/
Eli Zaretskii's avatar
Eli Zaretskii committed
92

93 94
/* Return the bidi type of a character CH, subject to the current
   directional OVERRIDE.  */
Paul Eggert's avatar
Paul Eggert committed
95
static inline bidi_type_t
96
bidi_get_type (int ch, bidi_dir_t override)
Eli Zaretskii's avatar
Eli Zaretskii committed
97
{
98 99
  bidi_type_t default_type;

100 101
  if (ch == BIDI_EOB)
    return NEUTRAL_B;
102 103
  if (ch < 0 || ch > MAX_CHAR)
    abort ();
104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135

  default_type = (bidi_type_t) XINT (CHAR_TABLE_REF (bidi_type_table, ch));

  if (override == NEUTRAL_DIR)
    return default_type;

  switch (default_type)
    {
      /* Although UAX#9 does not tell, it doesn't make sense to
	 override NEUTRAL_B and LRM/RLM characters.  */
      case NEUTRAL_B:
      case LRE:
      case LRO:
      case RLE:
      case RLO:
      case PDF:
	return default_type;
      default:
	switch (ch)
	  {
	    case LRM_CHAR:
	    case RLM_CHAR:
	      return default_type;
	    default:
	      if (override == L2R) /* X6 */
		return STRONG_L;
	      else if (override == R2L)
		return STRONG_R;
	      else
		abort ();	/* can't happen: handled above */
	  }
    }
Eli Zaretskii's avatar
Eli Zaretskii committed
136 137
}

138
static void
139 140 141 142 143 144
bidi_check_type (bidi_type_t type)
{
  if (type < UNKNOWN_BT || type > NEUTRAL_ON)
    abort ();
}

Eli Zaretskii's avatar
Eli Zaretskii committed
145
/* Given a bidi TYPE of a character, return its category.  */
Paul Eggert's avatar
Paul Eggert committed
146
static inline bidi_category_t
Eli Zaretskii's avatar
Eli Zaretskii committed
147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179
bidi_get_category (bidi_type_t type)
{
  switch (type)
    {
      case UNKNOWN_BT:
	return UNKNOWN_BC;
      case STRONG_L:
      case STRONG_R:
      case STRONG_AL:
      case LRE:
      case LRO:
      case RLE:
      case RLO:
	return STRONG;
      case PDF:		/* ??? really?? */
      case WEAK_EN:
      case WEAK_ES:
      case WEAK_ET:
      case WEAK_AN:
      case WEAK_CS:
      case WEAK_NSM:
      case WEAK_BN:
	return WEAK;
      case NEUTRAL_B:
      case NEUTRAL_S:
      case NEUTRAL_WS:
      case NEUTRAL_ON:
	return NEUTRAL;
      default:
	abort ();
    }
}

180 181 182 183
/* Return the mirrored character of C, if it has one.  If C has no
   mirrored counterpart, return C.
   Note: The conditions in UAX#9 clause L4 regarding the surrounding
   context must be tested by the caller.  */
Eli Zaretskii's avatar
Eli Zaretskii committed
184 185 186
int
bidi_mirror_char (int c)
{
187 188 189 190 191 192
  Lisp_Object val;

  if (c == BIDI_EOB)
    return c;
  if (c < 0 || c > MAX_CHAR)
    abort ();
Eli Zaretskii's avatar
Eli Zaretskii committed
193

194 195
  val = CHAR_TABLE_REF (bidi_mirror_table, c);
  if (INTEGERP (val))
Eli Zaretskii's avatar
Eli Zaretskii committed
196
    {
197
      int v = XINT (val);
Eli Zaretskii's avatar
Eli Zaretskii committed
198

199 200 201 202
      if (v < 0 || v > MAX_CHAR)
	abort ();

      return v;
Eli Zaretskii's avatar
Eli Zaretskii committed
203
    }
204

Eli Zaretskii's avatar
Eli Zaretskii committed
205 206 207
  return c;
}

208 209 210 211
/* Determine the start-of-run (sor) directional type given the two
   embedding levels on either side of the run boundary.  Also, update
   the saved info about previously seen characters, since that info is
   generally valid for a single level run.  */
212
static inline void
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
bidi_set_sor_type (struct bidi_it *bidi_it, int level_before, int level_after)
{
  int higher_level = level_before > level_after ? level_before : level_after;

  /* The prev_was_pdf gork is required for when we have several PDFs
     in a row.  In that case, we want to compute the sor type for the
     next level run only once: when we see the first PDF.  That's
     because the sor type depends only on the higher of the two levels
     that we find on the two sides of the level boundary (see UAX#9,
     clause X10), and so we don't need to know the final embedding
     level to which we descend after processing all the PDFs.  */
  if (!bidi_it->prev_was_pdf || level_before < level_after)
    /* FIXME: should the default sor direction be user selectable?  */
    bidi_it->sor = (higher_level & 1) != 0 ? R2L : L2R;
  if (level_before > level_after)
    bidi_it->prev_was_pdf = 1;

  bidi_it->prev.type = UNKNOWN_BT;
  bidi_it->last_strong.type = bidi_it->last_strong.type_after_w1 =
    bidi_it->last_strong.orig_type = UNKNOWN_BT;
  bidi_it->prev_for_neutral.type = bidi_it->sor == R2L ? STRONG_R : STRONG_L;
  bidi_it->prev_for_neutral.charpos = bidi_it->charpos;
  bidi_it->prev_for_neutral.bytepos = bidi_it->bytepos;
  bidi_it->next_for_neutral.type = bidi_it->next_for_neutral.type_after_w1 =
    bidi_it->next_for_neutral.orig_type = UNKNOWN_BT;
  bidi_it->ignore_bn_limit = -1; /* meaning it's unknown */
}

/* Push the current embedding level and override status; reset the
   current level to LEVEL and the current override status to OVERRIDE.  */
243
static inline void
244 245 246 247
bidi_push_embedding_level (struct bidi_it *bidi_it,
			   int level, bidi_dir_t override)
{
  bidi_it->stack_idx++;
248
  xassert (bidi_it->stack_idx < BIDI_MAXLEVEL);
249 250 251 252 253 254
  bidi_it->level_stack[bidi_it->stack_idx].level = level;
  bidi_it->level_stack[bidi_it->stack_idx].override = override;
}

/* Pop the embedding level and directional override status from the
   stack, and return the new level.  */
255
static inline int
256 257 258 259 260 261 262 263 264
bidi_pop_embedding_level (struct bidi_it *bidi_it)
{
  /* UAX#9 says to ignore invalid PDFs.  */
  if (bidi_it->stack_idx > 0)
    bidi_it->stack_idx--;
  return bidi_it->level_stack[bidi_it->stack_idx].level;
}

/* Record in SAVED_INFO the information about the current character.  */
265
static inline void
266 267 268 269 270 271 272 273 274 275 276 277 278
bidi_remember_char (struct bidi_saved_info *saved_info,
		    struct bidi_it *bidi_it)
{
  saved_info->charpos = bidi_it->charpos;
  saved_info->bytepos = bidi_it->bytepos;
  saved_info->type = bidi_it->type;
  bidi_check_type (bidi_it->type);
  saved_info->type_after_w1 = bidi_it->type_after_w1;
  bidi_check_type (bidi_it->type_after_w1);
  saved_info->orig_type = bidi_it->orig_type;
  bidi_check_type (bidi_it->orig_type);
}

Eli Zaretskii's avatar
Eli Zaretskii committed
279 280
/* Copy the bidi iterator from FROM to TO.  To save cycles, this only
   copies the part of the level stack that is actually in use.  */
Paul Eggert's avatar
Paul Eggert committed
281
static inline void
Eli Zaretskii's avatar
Eli Zaretskii committed
282 283 284 285
bidi_copy_it (struct bidi_it *to, struct bidi_it *from)
{
  int i;

286
  /* Copy everything except the level stack and beyond.  */
287
  memcpy (to, from, offsetof (struct bidi_it, level_stack[0]));
Eli Zaretskii's avatar
Eli Zaretskii committed
288 289 290 291 292 293 294

  /* Copy the active part of the level stack.  */
  to->level_stack[0] = from->level_stack[0]; /* level zero is always in use */
  for (i = 1; i <= from->stack_idx; i++)
    to->level_stack[i] = from->level_stack[i];
}

295 296 297 298

/***********************************************************************
			Caching the bidi iterator states
 ***********************************************************************/
Eli Zaretskii's avatar
Eli Zaretskii committed
299

300 301
#define BIDI_CACHE_CHUNK 200
static struct bidi_it *bidi_cache;
302
static ptrdiff_t bidi_cache_size = 0;
303
enum { elsz = sizeof (struct bidi_it) };
304 305 306
static ptrdiff_t bidi_cache_idx;	/* next unused cache slot */
static ptrdiff_t bidi_cache_last_idx;	/* slot of last cache hit */
static ptrdiff_t bidi_cache_start = 0;	/* start of cache for this
307 308 309 310 311 312 313 314
					   "stack" level */

/* Reset the cache state to the empty state.  We only reset the part
   of the cache relevant to iteration of the current object.  Previous
   objects, which are pushed on the display iterator's stack, are left
   intact.  This is called when the cached information is no more
   useful for the current iteration, e.g. when we were reseated to a
   new position on the same object.  */
Paul Eggert's avatar
Paul Eggert committed
315
static inline void
Eli Zaretskii's avatar
Eli Zaretskii committed
316 317
bidi_cache_reset (void)
{
318
  bidi_cache_idx = bidi_cache_start;
Eli Zaretskii's avatar
Eli Zaretskii committed
319 320 321
  bidi_cache_last_idx = -1;
}

322 323 324 325
/* Shrink the cache to its minimal size.  Called when we init the bidi
   iterator for reordering a buffer or a string that does not come
   from display properties, because that means all the previously
   cached info is of no further use.  */
Paul Eggert's avatar
Paul Eggert committed
326
static inline void
327 328 329 330
bidi_cache_shrink (void)
{
  if (bidi_cache_size > BIDI_CACHE_CHUNK)
    {
331 332 333
      bidi_cache_size = BIDI_CACHE_CHUNK;
      bidi_cache =
	(struct bidi_it *) xrealloc (bidi_cache, bidi_cache_size * elsz);
334 335 336 337
    }
  bidi_cache_reset ();
}

Paul Eggert's avatar
Paul Eggert committed
338
static inline void
339
bidi_cache_fetch_state (ptrdiff_t idx, struct bidi_it *bidi_it)
Eli Zaretskii's avatar
Eli Zaretskii committed
340 341 342
{
  int current_scan_dir = bidi_it->scan_dir;

343
  if (idx < bidi_cache_start || idx >= bidi_cache_idx)
Eli Zaretskii's avatar
Eli Zaretskii committed
344 345 346 347 348 349 350 351 352 353 354
    abort ();

  bidi_copy_it (bidi_it, &bidi_cache[idx]);
  bidi_it->scan_dir = current_scan_dir;
  bidi_cache_last_idx = idx;
}

/* Find a cached state with a given CHARPOS and resolved embedding
   level less or equal to LEVEL.  if LEVEL is -1, disregard the
   resolved levels in cached states.  DIR, if non-zero, means search
   in that direction from the last cache hit.  */
355
static inline ptrdiff_t
356
bidi_cache_search (EMACS_INT charpos, int level, int dir)
Eli Zaretskii's avatar
Eli Zaretskii committed
357
{
358
  ptrdiff_t i, i_start;
Eli Zaretskii's avatar
Eli Zaretskii committed
359

360
  if (bidi_cache_idx > bidi_cache_start)
Eli Zaretskii's avatar
Eli Zaretskii committed
361
    {
362 363
      if (bidi_cache_last_idx == -1)
	bidi_cache_last_idx = bidi_cache_idx - 1;
Eli Zaretskii's avatar
Eli Zaretskii committed
364
      if (charpos < bidi_cache[bidi_cache_last_idx].charpos)
365 366 367 368 369 370 371 372 373 374 375
	{
	  dir = -1;
	  i_start = bidi_cache_last_idx - 1;
	}
      else if (charpos > (bidi_cache[bidi_cache_last_idx].charpos
			  + bidi_cache[bidi_cache_last_idx].nchars - 1))
	{
	  dir = 1;
	  i_start = bidi_cache_last_idx + 1;
	}
      else if (dir)
Eli Zaretskii's avatar
Eli Zaretskii committed
376 377 378 379 380 381 382 383 384 385
	i_start = bidi_cache_last_idx;
      else
	{
	  dir = -1;
	  i_start = bidi_cache_idx - 1;
	}

      if (dir < 0)
	{
	  /* Linear search for now; FIXME!  */
386
	  for (i = i_start; i >= bidi_cache_start; i--)
387 388
	    if (bidi_cache[i].charpos <= charpos
		&& charpos < bidi_cache[i].charpos + bidi_cache[i].nchars
Eli Zaretskii's avatar
Eli Zaretskii committed
389 390 391 392 393 394
		&& (level == -1 || bidi_cache[i].resolved_level <= level))
	      return i;
	}
      else
	{
	  for (i = i_start; i < bidi_cache_idx; i++)
395 396
	    if (bidi_cache[i].charpos <= charpos
		&& charpos < bidi_cache[i].charpos + bidi_cache[i].nchars
Eli Zaretskii's avatar
Eli Zaretskii committed
397 398 399 400 401 402 403 404 405 406 407
		&& (level == -1 || bidi_cache[i].resolved_level <= level))
	      return i;
	}
    }

  return -1;
}

/* Find a cached state where the resolved level changes to a value
   that is lower than LEVEL, and return its cache slot index.  DIR is
   the direction to search, starting with the last used cache slot.
408 409 410
   If DIR is zero, we search backwards from the last occupied cache
   slot.  BEFORE, if non-zero, means return the index of the slot that
   is ``before'' the level change in the search direction.  That is,
Eli Zaretskii's avatar
Eli Zaretskii committed
411 412 413 414 415 416 417 418 419
   given the cached levels like this:

	 1122333442211
	  AB        C

   and assuming we are at the position cached at the slot marked with
   C, searching backwards (DIR = -1) for LEVEL = 2 will return the
   index of slot B or A, depending whether BEFORE is, respectively,
   non-zero or zero.  */
420
static ptrdiff_t
Eli Zaretskii's avatar
Eli Zaretskii committed
421 422 423 424
bidi_cache_find_level_change (int level, int dir, int before)
{
  if (bidi_cache_idx)
    {
425
      ptrdiff_t i = dir ? bidi_cache_last_idx : bidi_cache_idx - 1;
Eli Zaretskii's avatar
Eli Zaretskii committed
426 427
      int incr = before ? 1 : 0;

428 429
      xassert (!dir || bidi_cache_last_idx >= 0);

Eli Zaretskii's avatar
Eli Zaretskii committed
430 431 432 433 434 435 436
      if (!dir)
	dir = -1;
      else if (!incr)
	i += dir;

      if (dir < 0)
	{
437
	  while (i >= bidi_cache_start + incr)
Eli Zaretskii's avatar
Eli Zaretskii committed
438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459
	    {
	      if (bidi_cache[i - incr].resolved_level >= 0
		  && bidi_cache[i - incr].resolved_level < level)
		return i;
	      i--;
	    }
	}
      else
	{
	  while (i < bidi_cache_idx - incr)
	    {
	      if (bidi_cache[i + incr].resolved_level >= 0
		  && bidi_cache[i + incr].resolved_level < level)
		return i;
	      i++;
	    }
	}
    }

  return -1;
}

460
static inline void
461
bidi_cache_ensure_space (ptrdiff_t idx)
462 463 464 465
{
  /* Enlarge the cache as needed.  */
  if (idx >= bidi_cache_size)
    {
466
      ptrdiff_t new_size;
467 468 469 470 471 472 473 474 475 476

      /* The bidi cache cannot be larger than the largest Lisp string
	 or buffer.  */
      ptrdiff_t string_or_buffer_bound =
	max (BUF_BYTES_MAX, STRING_BYTES_BOUND);

      /* Also, it cannot be larger than what C can represent.  */
      ptrdiff_t c_bound = min (PTRDIFF_MAX, SIZE_MAX) / elsz;

      if (min (string_or_buffer_bound, c_bound) <= idx)
477 478 479 480
	memory_full (SIZE_MAX);
      new_size = idx - idx % BIDI_CACHE_CHUNK + BIDI_CACHE_CHUNK;
      bidi_cache = (struct bidi_it *) xrealloc (bidi_cache, new_size * elsz);
      bidi_cache_size = new_size;
481 482 483
    }
}

Paul Eggert's avatar
Paul Eggert committed
484
static inline void
Eli Zaretskii's avatar
Eli Zaretskii committed
485 486
bidi_cache_iterator_state (struct bidi_it *bidi_it, int resolved)
{
487
  ptrdiff_t idx;
Eli Zaretskii's avatar
Eli Zaretskii committed
488 489 490 491 492 493 494 495 496

  /* We should never cache on backward scans.  */
  if (bidi_it->scan_dir == -1)
    abort ();
  idx = bidi_cache_search (bidi_it->charpos, -1, 1);

  if (idx < 0)
    {
      idx = bidi_cache_idx;
497
      bidi_cache_ensure_space (idx);
498 499 500
      /* Character positions should correspond to cache positions 1:1.
	 If we are outside the range of cached positions, the cache is
	 useless and must be reset.  */
501
      if (idx > bidi_cache_start &&
502 503
	  (bidi_it->charpos > (bidi_cache[idx - 1].charpos
			       + bidi_cache[idx - 1].nchars)
504
	   || bidi_it->charpos < bidi_cache[bidi_cache_start].charpos))
505 506
	{
	  bidi_cache_reset ();
507
	  idx = bidi_cache_start;
508
	}
509 510
      if (bidi_it->nchars <= 0)
	abort ();
Eli Zaretskii's avatar
Eli Zaretskii committed
511 512 513 514 515 516 517 518 519
      bidi_copy_it (&bidi_cache[idx], bidi_it);
      if (!resolved)
	bidi_cache[idx].resolved_level = -1;
    }
  else
    {
      /* Copy only the members which could have changed, to avoid
	 costly copying of the entire struct.  */
      bidi_cache[idx].type = bidi_it->type;
520
      bidi_check_type (bidi_it->type);
521 522
      bidi_cache[idx].type_after_w1 = bidi_it->type_after_w1;
      bidi_check_type (bidi_it->type_after_w1);
Eli Zaretskii's avatar
Eli Zaretskii committed
523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538
      if (resolved)
	bidi_cache[idx].resolved_level = bidi_it->resolved_level;
      else
	bidi_cache[idx].resolved_level = -1;
      bidi_cache[idx].invalid_levels = bidi_it->invalid_levels;
      bidi_cache[idx].invalid_rl_levels = bidi_it->invalid_rl_levels;
      bidi_cache[idx].next_for_neutral = bidi_it->next_for_neutral;
      bidi_cache[idx].next_for_ws = bidi_it->next_for_ws;
      bidi_cache[idx].ignore_bn_limit = bidi_it->ignore_bn_limit;
    }

  bidi_cache_last_idx = idx;
  if (idx >= bidi_cache_idx)
    bidi_cache_idx = idx + 1;
}

Paul Eggert's avatar
Paul Eggert committed
539
static inline bidi_type_t
540
bidi_cache_find (EMACS_INT charpos, int level, struct bidi_it *bidi_it)
Eli Zaretskii's avatar
Eli Zaretskii committed
541
{
542
  ptrdiff_t i = bidi_cache_search (charpos, level, bidi_it->scan_dir);
Eli Zaretskii's avatar
Eli Zaretskii committed
543

544
  if (i >= bidi_cache_start)
Eli Zaretskii's avatar
Eli Zaretskii committed
545 546 547
    {
      bidi_dir_t current_scan_dir = bidi_it->scan_dir;

548
      bidi_copy_it (bidi_it, &bidi_cache[i]);
Eli Zaretskii's avatar
Eli Zaretskii committed
549 550 551 552 553 554 555 556 557 558
      bidi_cache_last_idx = i;
      /* Don't let scan direction from from the cached state override
	 the current scan direction.  */
      bidi_it->scan_dir = current_scan_dir;
      return bidi_it->type;
    }

  return UNKNOWN_BT;
}

Paul Eggert's avatar
Paul Eggert committed
559
static inline int
Eli Zaretskii's avatar
Eli Zaretskii committed
560 561
bidi_peek_at_next_level (struct bidi_it *bidi_it)
{
562
  if (bidi_cache_idx == bidi_cache_start || bidi_cache_last_idx == -1)
Eli Zaretskii's avatar
Eli Zaretskii committed
563 564 565 566
    abort ();
  return bidi_cache[bidi_cache_last_idx + bidi_it->scan_dir].resolved_level;
}

567 568 569 570

/***********************************************************************
	     Pushing and popping the bidi iterator state
 ***********************************************************************/
571 572 573
/* 5-slot stack for saving the start of the previous level of the
   cache.  xdisp.c maintains a 5-slot stack for its iterator state,
   and we need the same size of our stack.  */
574
static ptrdiff_t bidi_cache_start_stack[IT_STACK_SIZE];
575 576 577 578
static int bidi_cache_sp;

/* Push the bidi iterator state in preparation for reordering a
   different object, e.g. display string found at certain buffer
579 580 581
   position.  Pushing the bidi iterator boils down to saving its
   entire state on the cache and starting a new cache "stacked" on top
   of the current cache.  */
582 583
void
bidi_push_it (struct bidi_it *bidi_it)
Eli Zaretskii's avatar
Eli Zaretskii committed
584
{
585 586 587 588
  /* Save the current iterator state in its entirety after the last
     used cache slot.  */
  bidi_cache_ensure_space (bidi_cache_idx);
  memcpy (&bidi_cache[bidi_cache_idx++], bidi_it, sizeof (struct bidi_it));
589

590
  /* Push the current cache start onto the stack.  */
591
  xassert (bidi_cache_sp < IT_STACK_SIZE);
592
  bidi_cache_start_stack[bidi_cache_sp++] = bidi_cache_start;
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
  /* Start a new level of cache, and make it empty.  */
  bidi_cache_start = bidi_cache_idx;
  bidi_cache_last_idx = -1;
}

/* Restore the iterator state saved by bidi_push_it and return the
   cache to the corresponding state.  */
void
bidi_pop_it (struct bidi_it *bidi_it)
{
  if (bidi_cache_start <= 0)
    abort ();

  /* Reset the next free cache slot index to what it was before the
     call to bidi_push_it.  */
  bidi_cache_idx = bidi_cache_start - 1;

  /* Restore the bidi iterator state saved in the cache.  */
  memcpy (bidi_it, &bidi_cache[bidi_cache_idx], sizeof (struct bidi_it));

  /* Pop the previous cache start from the stack.  */
  if (bidi_cache_sp <= 0)
    abort ();
  bidi_cache_start = bidi_cache_start_stack[--bidi_cache_sp];

  /* Invalidate the last-used cache slot data.  */
  bidi_cache_last_idx = -1;
}

623 624
ptrdiff_t bidi_cache_total_alloc;

625 626 627 628 629 630
/* Stash away a copy of the cache and its control variables.  */
void *
bidi_shelve_cache (void)
{
  unsigned char *databuf;

631
  /* Empty cache.  */
632 633 634 635 636 637 638 639
  if (bidi_cache_idx == 0)
    return NULL;

  databuf = xmalloc (sizeof (bidi_cache_idx)
		     + bidi_cache_idx * sizeof (struct bidi_it)
		     + sizeof (bidi_cache_start_stack)
		     + sizeof (bidi_cache_sp) + sizeof (bidi_cache_start)
		     + sizeof (bidi_cache_last_idx));
640 641 642 643 644 645
  bidi_cache_total_alloc +=
    sizeof (bidi_cache_idx) + bidi_cache_idx * sizeof (struct bidi_it)
    + sizeof (bidi_cache_start_stack)
    + sizeof (bidi_cache_sp) + sizeof (bidi_cache_start)
    + sizeof (bidi_cache_last_idx);

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
  memcpy (databuf, &bidi_cache_idx, sizeof (bidi_cache_idx));
  memcpy (databuf + sizeof (bidi_cache_idx),
	  bidi_cache, bidi_cache_idx * sizeof (struct bidi_it));
  memcpy (databuf + sizeof (bidi_cache_idx)
	  + bidi_cache_idx * sizeof (struct bidi_it),
	  bidi_cache_start_stack, sizeof (bidi_cache_start_stack));
  memcpy (databuf + sizeof (bidi_cache_idx)
	  + bidi_cache_idx * sizeof (struct bidi_it)
	  + sizeof (bidi_cache_start_stack),
	  &bidi_cache_sp, sizeof (bidi_cache_sp));
  memcpy (databuf + sizeof (bidi_cache_idx)
	  + bidi_cache_idx * sizeof (struct bidi_it)
	  + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp),
	  &bidi_cache_start, sizeof (bidi_cache_start));
  memcpy (databuf + sizeof (bidi_cache_idx)
	  + bidi_cache_idx * sizeof (struct bidi_it)
	  + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp)
	  + sizeof (bidi_cache_start),
	  &bidi_cache_last_idx, sizeof (bidi_cache_last_idx));

  return databuf;
}

/* Restore the cache state from a copy stashed away by bidi_shelve_cache.  */
void
671
bidi_unshelve_cache (void *databuf, int just_free)
672 673 674 675
{
  unsigned char *p = databuf;

  if (!p)
676
    {
677 678 679 680
      /* A NULL pointer means an empty cache.  */
      bidi_cache_start = 0;
      bidi_cache_sp = 0;
      bidi_cache_reset ();
681
    }
682 683
  else
    {
684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724
      if (just_free)
	{
	  ptrdiff_t idx;

	  memcpy (&idx, p, sizeof (bidi_cache_idx));
	  bidi_cache_total_alloc -=
	    sizeof (bidi_cache_idx) + idx * sizeof (struct bidi_it)
	    + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp)
	    + sizeof (bidi_cache_start) + sizeof (bidi_cache_last_idx);
	}
      else
	{
	  memcpy (&bidi_cache_idx, p, sizeof (bidi_cache_idx));
	  bidi_cache_ensure_space (bidi_cache_idx);
	  memcpy (bidi_cache, p + sizeof (bidi_cache_idx),
		  bidi_cache_idx * sizeof (struct bidi_it));
	  memcpy (bidi_cache_start_stack,
		  p + sizeof (bidi_cache_idx)
		  + bidi_cache_idx * sizeof (struct bidi_it),
		  sizeof (bidi_cache_start_stack));
	  memcpy (&bidi_cache_sp,
		  p + sizeof (bidi_cache_idx)
		  + bidi_cache_idx * sizeof (struct bidi_it)
		  + sizeof (bidi_cache_start_stack),
		  sizeof (bidi_cache_sp));
	  memcpy (&bidi_cache_start,
		  p + sizeof (bidi_cache_idx)
		  + bidi_cache_idx * sizeof (struct bidi_it)
		  + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp),
		  sizeof (bidi_cache_start));
	  memcpy (&bidi_cache_last_idx,
		  p + sizeof (bidi_cache_idx)
		  + bidi_cache_idx * sizeof (struct bidi_it)
		  + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp)
		  + sizeof (bidi_cache_start),
		  sizeof (bidi_cache_last_idx));
	  bidi_cache_total_alloc -=
	    sizeof (bidi_cache_idx) + bidi_cache_idx * sizeof (struct bidi_it)
	    + sizeof (bidi_cache_start_stack) + sizeof (bidi_cache_sp)
	    + sizeof (bidi_cache_start) + sizeof (bidi_cache_last_idx);
	}
725 726 727 728

      xfree (p);
    }
}
Eli Zaretskii's avatar
Eli Zaretskii committed
729

730 731 732 733 734 735

/***********************************************************************
			Initialization
 ***********************************************************************/
static void
bidi_initialize (void)
Eli Zaretskii's avatar
Eli Zaretskii committed
736
{
737

738 739
#include "biditype.h"
#include "bidimirror.h"
740

741
  int i;
Eli Zaretskii's avatar
Eli Zaretskii committed
742

743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768
  bidi_type_table = Fmake_char_table (Qnil, make_number (STRONG_L));
  staticpro (&bidi_type_table);

  for (i = 0; i < sizeof bidi_type / sizeof bidi_type[0]; i++)
    char_table_set_range (bidi_type_table, bidi_type[i].from, bidi_type[i].to,
			  make_number (bidi_type[i].type));

  bidi_mirror_table = Fmake_char_table (Qnil, Qnil);
  staticpro (&bidi_mirror_table);

  for (i = 0; i < sizeof bidi_mirror / sizeof bidi_mirror[0]; i++)
    char_table_set (bidi_mirror_table, bidi_mirror[i].from,
		    make_number (bidi_mirror[i].to));

  Qparagraph_start = intern ("paragraph-start");
  staticpro (&Qparagraph_start);
  paragraph_start_re = Fsymbol_value (Qparagraph_start);
  if (!STRINGP (paragraph_start_re))
    paragraph_start_re = build_string ("\f\\|[ \t]*$");
  staticpro (&paragraph_start_re);
  Qparagraph_separate = intern ("paragraph-separate");
  staticpro (&Qparagraph_separate);
  paragraph_separate_re = Fsymbol_value (Qparagraph_separate);
  if (!STRINGP (paragraph_separate_re))
    paragraph_separate_re = build_string ("[ \t\f]*$");
  staticpro (&paragraph_separate_re);
769 770 771

  bidi_cache_sp = 0;

772
  bidi_initialized = 1;
Eli Zaretskii's avatar
Eli Zaretskii committed
773 774
}

775 776
/* Do whatever UAX#9 clause X8 says should be done at paragraph's
   end.  */
Paul Eggert's avatar
Paul Eggert committed
777
static inline void
778
bidi_set_paragraph_end (struct bidi_it *bidi_it)
Eli Zaretskii's avatar
Eli Zaretskii committed
779
{
780 781 782 783 784
  bidi_it->invalid_levels = 0;
  bidi_it->invalid_rl_levels = -1;
  bidi_it->stack_idx = 0;
  bidi_it->resolved_level = bidi_it->level_stack[0].level;
}
Eli Zaretskii's avatar
Eli Zaretskii committed
785

786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808
/* Initialize the bidi iterator from buffer/string position CHARPOS.  */
void
bidi_init_it (EMACS_INT charpos, EMACS_INT bytepos, int frame_window_p,
	      struct bidi_it *bidi_it)
{
  if (! bidi_initialized)
    bidi_initialize ();
  if (charpos >= 0)
    bidi_it->charpos = charpos;
  if (bytepos >= 0)
    bidi_it->bytepos = bytepos;
  bidi_it->frame_window_p = frame_window_p;
  bidi_it->nchars = -1;	/* to be computed in bidi_resolve_explicit_1 */
  bidi_it->first_elt = 1;
  bidi_set_paragraph_end (bidi_it);
  bidi_it->new_paragraph = 1;
  bidi_it->separator_limit = -1;
  bidi_it->type = NEUTRAL_B;
  bidi_it->type_after_w1 = NEUTRAL_B;
  bidi_it->orig_type = NEUTRAL_B;
  bidi_it->prev_was_pdf = 0;
  bidi_it->prev.type = bidi_it->prev.type_after_w1 =
    bidi_it->prev.orig_type = UNKNOWN_BT;
809 810
  bidi_it->last_strong.type = bidi_it->last_strong.type_after_w1 =
    bidi_it->last_strong.orig_type = UNKNOWN_BT;
811 812 813
  bidi_it->next_for_neutral.charpos = -1;
  bidi_it->next_for_neutral.type =
    bidi_it->next_for_neutral.type_after_w1 =
814
    bidi_it->next_for_neutral.orig_type = UNKNOWN_BT;
815 816 817 818 819 820
  bidi_it->prev_for_neutral.charpos = -1;
  bidi_it->prev_for_neutral.type =
    bidi_it->prev_for_neutral.type_after_w1 =
    bidi_it->prev_for_neutral.orig_type = UNKNOWN_BT;
  bidi_it->sor = L2R;	 /* FIXME: should it be user-selectable? */
  bidi_it->disp_pos = -1;	/* invalid/unknown */
821
  bidi_it->disp_prop_p = 0;
822 823 824 825
  /* We can only shrink the cache if we are at the bottom level of its
     "stack".  */
  if (bidi_cache_start == 0)
    bidi_cache_shrink ();
826 827
  else
    bidi_cache_reset ();
Eli Zaretskii's avatar
Eli Zaretskii committed
828 829
}

830
/* Perform initializations for reordering a new line of bidi text.  */
831
static void
832 833 834 835 836 837 838 839 840
bidi_line_init (struct bidi_it *bidi_it)
{
  bidi_it->scan_dir = 1; /* FIXME: do we need to have control on this? */
  bidi_it->resolved_level = bidi_it->level_stack[0].level;
  bidi_it->level_stack[0].override = NEUTRAL_DIR; /* X1 */
  bidi_it->invalid_levels = 0;
  bidi_it->invalid_rl_levels = -1;
  bidi_it->next_en_pos = -1;
  bidi_it->next_for_ws.type = UNKNOWN_BT;
841 842
  bidi_set_sor_type (bidi_it,
		     bidi_it->paragraph_dir == R2L ? 1 : 0,
843 844 845 846 847
		     bidi_it->level_stack[0].level); /* X10 */

  bidi_cache_reset ();
}

848 849 850 851 852

/***********************************************************************
			Fetching characters
 ***********************************************************************/

853 854 855 856
/* Count bytes in string S between BEG/BEGBYTE and END.  BEG and END
   are zero-based character positions in S, BEGBYTE is byte position
   corresponding to BEG.  UNIBYTE, if non-zero, means S is a unibyte
   string.  */
857 858
static inline EMACS_INT
bidi_count_bytes (const unsigned char *s, const EMACS_INT beg,
859
		  const EMACS_INT begbyte, const EMACS_INT end, int unibyte)
860 861 862 863
{
  EMACS_INT pos = beg;
  const unsigned char *p = s + begbyte, *start = p;

864 865 866
  if (unibyte)
    p = s + end;
  else
867
    {
868 869 870 871 872 873 874 875
      if (!CHAR_HEAD_P (*p))
	abort ();

      while (pos < end)
	{
	  p += BYTES_BY_CHAR_HEAD (*p);
	  pos++;
	}
876 877 878 879 880 881 882
    }

  return p - start;
}

/* Fetch and returns the character at byte position BYTEPOS.  If S is
   non-NULL, fetch the character from string S; otherwise fetch the
883 884
   character from the current buffer.  UNIBYTE non-zero means S is a
   unibyte string.  */
885
static inline int
886
bidi_char_at_pos (EMACS_INT bytepos, const unsigned char *s, int unibyte)
887 888
{
  if (s)
889 890 891 892 893 894
    {
      if (unibyte)
	return s[bytepos];
      else
	return STRING_CHAR (s + bytepos);
    }
895 896 897 898
  else
    return FETCH_MULTIBYTE_CHAR (bytepos);
}

899 900 901 902 903
/* Fetch and return the character at BYTEPOS/CHARPOS.  If that
   character is covered by a display string, treat the entire run of
   covered characters as a single character u+FFFC, and return their
   combined length in CH_LEN and NCHARS.  DISP_POS specifies the
   character position of the next display string, or -1 if not yet
904 905 906 907 908 909 910
   computed.  DISP_PROP_P non-zero means that there's really a display
   string at DISP_POS, as opposed to when we searched till DISP_POS
   without findingone.  When the next character is at or beyond that
   position, the function updates DISP_POS with the position of the
   next display string.  STRING->s is the C string to iterate, or NULL
   if iterating over a buffer or a Lisp string; in the latter case,
   STRING->lstring is the Lisp string.  */
911
static inline int
912
bidi_fetch_char (EMACS_INT bytepos, EMACS_INT charpos, EMACS_INT *disp_pos,
913
		 int *disp_prop_p, struct bidi_string_data *string,
914
		 int frame_window_p, EMACS_INT *ch_len, EMACS_INT *nchars)
915 916
{
  int ch;
917 918
  EMACS_INT endpos =
    (string->s || STRINGP (string->lstring)) ? string->schars : ZV;
919
  struct text_pos pos;
920 921

  /* If we got past the last known position of display string, compute
922 923
     the position of the next one.  That position could be at CHARPOS.  */
  if (charpos < endpos && charpos > *disp_pos)
924 925
    {
      SET_TEXT_POS (pos, charpos, bytepos);
926 927
      *disp_pos = compute_display_string_pos (&pos, string, frame_window_p,
					      disp_prop_p);
928
    }
929 930

  /* Fetch the character at BYTEPOS.  */
931
  if (charpos >= endpos)
932 933 934 935
    {
      ch = BIDI_EOB;
      *ch_len = 1;
      *nchars = 1;
936
      *disp_pos = endpos;
937
      *disp_prop_p = 0;
938
    }
939
  else if (charpos >= *disp_pos && *disp_prop_p)
940
    {
941 942 943 944 945 946 947
      EMACS_INT disp_end_pos;

      /* We don't expect to find ourselves in the middle of a display
	 property.  Hopefully, it will never be needed.  */
      if (charpos > *disp_pos)
	abort ();
      /* Return the Unicode Object Replacement Character to represent
948
	 the entire run of characters covered by the display string.  */
949
      ch = 0xFFFC;
950
      disp_end_pos = compute_display_string_end (*disp_pos, string);
951
      *nchars = disp_end_pos - *disp_pos;
952 953
      if (*nchars <= 0)
	abort ();
954 955
      if (string->s)
	*ch_len = bidi_count_bytes (string->s, *disp_pos, bytepos,
956
				    disp_end_pos, string->unibyte);
957 958
      else if (STRINGP (string->lstring))
	*ch_len = bidi_count_bytes (SDATA (string->lstring), *disp_pos,
959
				    bytepos, disp_end_pos, string->unibyte);
960 961
      else
	*ch_len = CHAR_TO_BYTE (disp_end_pos) - bytepos;
962 963 964
    }
  else
    {
965 966
      if (string->s)
	{
967
	  int len;
968

969 970 971 972 973 974 975 976 977 978
	  if (!string->unibyte)
	    {
	      ch = STRING_CHAR_AND_LENGTH (string->s + bytepos, len);
	      *ch_len = len;
	    }
	  else
	    {
	      ch = UNIBYTE_TO_CHAR (string->s[bytepos]);
	      *ch_len = 1;
	    }
979
	}
980 981
      else if (STRINGP (string->lstring))
	{
982
	  int len;
983

984 985 986 987 988 989 990 991 992 993 994
	  if (!string->unibyte)
	    {
	      ch = STRING_CHAR_AND_LENGTH (SDATA (string->lstring) + bytepos,
					   len);
	      *ch_len = len;
	    }
	  else
	    {
	      ch = UNIBYTE_TO_CHAR (SREF (string->lstring, bytepos));
	      *ch_len = 1;
	    }
995
	}
996 997 998 999 1000
      else
	{
	  ch = FETCH_MULTIBYTE_CHAR (bytepos);
	  *ch_len = CHAR_BYTES (ch);
	}
1001 1002 1003 1004 1005
      *nchars = 1;
    }

  /* If we just entered a run of characters covered by a display
     string, compute the position of the next display string.  */
1006 1007
  if (charpos + *nchars <= endpos && charpos + *nchars > *disp_pos
      && *disp_prop_p)
1008 1009
    {
      SET_TEXT_POS (pos, charpos + *nchars, bytepos + *ch_len);
1010 1011
      *disp_pos = compute_display_string_pos (&pos, string, frame_window_p,
					      disp_prop_p);
1012
    }
1013 1014 1015 1016

  return ch;
}

1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048

/***********************************************************************
			Determining paragraph direction
 ***********************************************************************/

/* Check if buffer position CHARPOS/BYTEPOS is the end of a paragraph.
   Value is the non-negative length of the paragraph separator
   following the buffer position, -1 if position is at the beginning
   of a new paragraph, or -2 if position is neither at beginning nor
   at end of a paragraph.  */
static EMACS_INT
bidi_at_paragraph_end (EMACS_INT charpos, EMACS_INT bytepos)
{
  Lisp_Object sep_re;
  Lisp_Object start_re;
  EMACS_INT val;

  sep_re = paragraph_separate_re;
  start_re = paragraph_start_re;

  val = fast_looking_at (sep_re, charpos, bytepos, ZV, ZV_BYTE, Qnil);
  if (val < 0)
    {
      if (fast_looking_at (start_re, charpos, bytepos, ZV, ZV_BYTE, Qnil) >= 0)
	val = -1;
      else
	val = -2;
    }

  return val;
}

1049 1050 1051
/* Find the beginning of this paragraph by looking back in the buffer.
   Value is the byte position of the paragraph's beginning.  */
static EMACS_INT
1052
bidi_find_paragraph_start (EMACS_INT pos, EMACS_INT pos_byte)
1053
{
Eli Zaretskii's avatar
Eli Zaretskii committed
1054
  Lisp_Object re = paragraph_start_re;
1055 1056 1057 1058 1059
  EMACS_INT limit = ZV, limit_byte = ZV_BYTE;

  while (pos_byte > BEGV_BYTE
	 && fast_looking_at (re, pos, pos_byte, limit, limit_byte, Qnil) < 0)
    {
1060 1061 1062 1063
      /* FIXME: What if the paragraph beginning is covered by a
	 display string?  And what if a display string covering some
	 of the text over which we scan back includes
	 paragraph_start_re?  */
1064 1065
      pos = find_next_newline_no_quit (pos - 1, -1);
      pos_byte = CHAR_TO_BYTE (pos);
1066
    }
1067
  return pos_byte;
1068 1069
}

1070
/* Determine the base direction, a.k.a. base embedding level, of the
1071 1072
   paragraph we are about to iterate through.  If DIR is either L2R or
   R2L, just use that.  Otherwise, determine the paragraph direction
1073 1074
   from the first strong directional character of the paragraph.

1075
   NO_DEFAULT_P non-zero means don't default to L2R if the paragraph
1076 1077 1078 1079 1080 1081 1082 1083 1084
   has no strong directional characters and both DIR and
   bidi_it->paragraph_dir are NEUTRAL_DIR.  In that case, search back
   in the buffer until a paragraph is found with a strong character,
   or until hitting BEGV.  In the latter case, fall back to L2R.  This
   flag is used in current-bidi-paragraph-direction.

   Note that this function gives the paragraph separator the same
   direction as the preceding paragraph, even though Emacs generally
   views the separartor as not belonging to any paragraph.  */
Eli Zaretskii's avatar
Eli Zaretskii committed
1085
void
1086
bidi_paragraph_init (bidi_dir_t dir, struct bidi_it *bidi_it, int no_default_p)
Eli Zaretskii's avatar
Eli Zaretskii committed
1087
{
1088
  EMACS_INT bytepos = bidi_it->bytepos;
1089
  int string_p = bidi_it->string.s != NULL || STRINGP (bidi_it->string.lstring);
1090
  EMACS_INT pstartbyte;
1091 1092 1093 1094 1095
  /* Note that begbyte is a byte position, while end is a character
     position.  Yes, this is ugly, but we are trying to avoid costly
     calls to BYTE_TO_CHAR and its ilk.  */
  EMACS_INT begbyte = string_p ? 0 : BEGV_BYTE;
  EMACS_INT end = string_p ? bidi_it->string.schars : ZV;
1096

1097
  /* Special case for an empty buffer. */
1098
  if (bytepos == begbyte && bidi_it->charpos == end)
1099
    dir = L2R;
1100
  /* We should never be called at EOB or before BEGV.  */
1101
  else if (bidi_it->charpos >= end || bytepos < begbyte)
1102 1103
    abort ();

1104 1105 1106 1107 1108 1109 1110 1111 1112 1113
  if (dir == L2R)
    {
      bidi_it->paragraph_dir = L2R;
      bidi_it->new_paragraph = 0;
    }
  else if (dir == R2L)
    {
      bidi_it->paragraph_dir = R2L;
      bidi_it->new_paragraph = 0;
    }
Eli Zaretskii's avatar
Eli Zaretskii committed
1114 1115
  else if (dir == NEUTRAL_DIR)	/* P2 */
    {
1116 1117 1118
      int ch;
      EMACS_INT ch_len, nchars;
      EMACS_INT pos, disp_pos = -1;
1119
      int disp_prop_p = 0;
1120
      bidi_type_t type;
1121
      const unsigned char *s;
1122

1123 1124 1125
      if (!bidi_initialized)
	bidi_initialize ();

1126 1127
      /* If we are inside a paragraph separator, we are just waiting
	 for the separator to be exhausted; use the previous paragraph
1128 1129 1130 1131
	 direction.  But don't do that if we have been just reseated,
	 because we need to reinitialize below in that case.  */
      if (!bidi_it->first_elt
	  && bidi_it->charpos < bidi_it->separator_limit)
1132 1133
	return;

1134
      /* If we are on a newline, get past it to where the next
1135 1136 1137
	 paragraph might start.  But don't do that at BEGV since then
	 we are potentially in a new paragraph that doesn't yet
	 exist.  */
1138
      pos = bidi_it->charpos;
1139 1140
      s = STRINGP (bidi_it->string.lstring) ?
	SDATA (bidi_it->string.lstring) : bidi_it->string.s;
1141 1142
      if (bytepos > begbyte
	  && bidi_char_at_pos (bytepos, s, bidi_it->string.unibyte) == '\n')
1143
	{
1144
	  bytepos++;
1145
	  pos++;
1146
	}
1147 1148 1149

      /* We are either at the beginning of a paragraph or in the
	 middle of it.  Find where this paragraph starts.  */
1150 1151 1152 1153 1154 1155 1156 1157
      if (string_p)
	{
	  /* We don't support changes of paragraph direction inside a
	     string.  It is treated as a single paragraph.  */
	  pstartbyte = 0;
	}
      else
	pstartbyte = bidi_find_paragraph_start (pos, bytepos);
1158 1159
      bidi_it->separator_limit = -1;
      bidi_it->new_paragraph = 0;
1160 1161

      /* The following loop is run more than once only if NO_DEFAULT_P
1162
	 is non-zero, and only if we are iterating on a buffer.  */
1163 1164
      do {
	bytepos = pstartbyte;
1165 1166
	if (!string_p)
	  pos = BYTE_TO_CHAR (bytepos);
1167 1168
	ch = bidi_fetch_char (bytepos, pos, &disp_pos, &disp_prop_p,
			      &bidi_it->string,
1169
			      bidi_it->frame_window_p, &ch_len, &nchars);
1170 1171
	type = bidi_get_type (ch, NEUTRAL_DIR);

1172
	for (pos += nchars, bytepos += ch_len;
1173 1174 1175 1176 1177 1178 1179 1180 1181 1182
	     /* NOTE: UAX#9 says to search only for L, AL, or R types
		of characters, and ignore RLE, RLO, LRE, and LRO.
		However, I'm not sure it makes sense to omit those 4;
		should try with and without that to see the effect.  */
	     (bidi_get_category (type) != STRONG)
	       || (bidi_ignore_explicit_marks_for_paragraph_level
		   && (type == RLE || type == RLO
		       || type == LRE || type == LRO));
	     type = bidi_get_type (ch, NEUTRAL_DIR))
	  {
1183
	    if (pos >= end)
1184 1185
	      {
		/* Pretend there's a paragraph separator at end of
1186
		   buffer/string.  */
1187 1188 1189
		type = NEUTRAL_B;
		break;
	      }
1190 1191 1192
	    if (!string_p
		&& type == NEUTRAL_B
		&& bidi_at_paragraph_end (pos, bytepos) >= -1)
1193
	      break;
1194
	    /* Fetch next character and advance to get past it.  */
1195 1196
	    ch = bidi_fetch_char (bytepos, pos, &disp_pos,
				  &disp_prop_p, &bidi_it->string,