composite.c 62.1 KB
Newer Older
Kenichi Handa's avatar
Kenichi Handa committed
1
/* Composite sequence support.
Paul Eggert's avatar
Paul Eggert committed
2
   Copyright (C) 2001-2020 Free Software Foundation, Inc.
3
   Copyright (C) 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011
Kenichi Handa's avatar
Kenichi Handa committed
4 5
     National Institute of Advanced Industrial Science and Technology (AIST)
     Registration Number H14PRO021
6
   Copyright (C) 2003, 2006
7 8
     National Institute of Advanced Industrial Science and Technology (AIST)
     Registration Number H13PRO009
Kenichi Handa's avatar
Kenichi Handa committed
9 10 11

This file is part of GNU Emacs.

12
GNU Emacs is free software: you can redistribute it and/or modify
Kenichi Handa's avatar
Kenichi Handa committed
13
it under the terms of the GNU General Public License as published by
14 15
the Free Software Foundation, either version 3 of the License, or (at
your option) any later version.
Kenichi Handa's avatar
Kenichi Handa committed
16 17 18 19 20 21 22

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
23
along with GNU Emacs.  If not, see <https://www.gnu.org/licenses/>.  */
Kenichi Handa's avatar
Kenichi Handa committed
24 25

#include <config.h>
26

Kenichi Handa's avatar
Kenichi Handa committed
27
#include "lisp.h"
28
#include "character.h"
29
#include "composite.h"
30
#include "buffer.h"
31
#include "coding.h"
Kenichi Handa's avatar
Kenichi Handa committed
32
#include "intervals.h"
33 34
#include "frame.h"
#include "dispextern.h"
35 36
#include "termhooks.h"

Kenichi Handa's avatar
Kenichi Handa committed
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

/* Emacs uses special text property `composition' to support character
   composition.  A sequence of characters that have the same (i.e. eq)
   `composition' property value is treated as a single composite
   sequence (we call it just `composition' here after).  Characters in
   a composition are all composed somehow on the screen.

   The property value has this form when the composition is made:
	((LENGTH . COMPONENTS) . MODIFICATION-FUNC)
   then turns to this form:
	(COMPOSITION-ID . (LENGTH COMPONENTS-VEC . MODIFICATION-FUNC))
   when the composition is registered in composition_hash_table and
   composition_table.  These rather peculiar structures were designed
   to make it easy to distinguish them quickly (we can do that by
   checking only the first element) and to extract LENGTH (from the
   former form) and COMPOSITION-ID (from the latter form).

   We register a composition when it is displayed, or when the width
   is required (for instance, to calculate columns).

   LENGTH -- Length of the composition.  This information is used to
	check the validity of the composition.

   COMPONENTS --  Character, string, vector, list, or nil.

	If it is nil, characters in the text are composed relatively
	according to their metrics in font glyphs.

	If it is a character or a string, the character or characters
	in the string are composed relatively.

	If it is a vector or list of integers, the element is a
	character or an encoded composition rule.  The characters are
	composed according to the rules.  (2N)th elements are
	characters to be composed and (2N+1)th elements are
	composition rules to tell how to compose (2N+2)th element with
	the previously composed 2N glyphs.

75 76
   COMPONENTS-VEC -- Vector of integers.  In a relative composition,
	the elements are the characters to be composed.  In a rule-base
Kenichi Handa's avatar
Kenichi Handa committed
77 78 79
	composition, the elements are characters or encoded
	composition rules.

80
   MODIFICATION-FUNC -- If non-nil, it is a function to call when the
Kenichi Handa's avatar
Kenichi Handa committed
81 82 83 84 85 86 87 88 89 90 91 92 93 94 95
	composition gets invalid after a modification in a buffer.  If
	it is nil, a function in `composition-function-table' of the
	first character in the sequence is called.

   COMPOSITION-ID --Identification number of the composition.  It is
	used as an index to composition_table for the composition.

   When Emacs has to display a composition or has to know its
   displaying width, the function get_composition_id is called.  It
   returns COMPOSITION-ID so that the caller can access the
   information about the composition through composition_table.  If a
   COMPOSITION-ID has not yet been assigned to the composition,
   get_composition_id checks the validity of `composition' property,
   and, if valid, assigns a new ID, registers the information in
   composition_hash_table and composition_table, and changes the form
96 97
   of the property value.  If the property is invalid,
   get_composition_id returns -1 without changing the property value.
Kenichi Handa's avatar
Kenichi Handa committed
98

99
   We use two tables to keep the information about composition;
Kenichi Handa's avatar
Kenichi Handa committed
100 101
   composition_hash_table and composition_table.

102
   The former is a hash table whose keys are COMPONENTS-VECs and
Kenichi Handa's avatar
Kenichi Handa committed
103
   values are the corresponding COMPOSITION-IDs.  This hash table is
104
   weak, but as each key (COMPONENTS-VEC) is also kept as a value of the
Kenichi Handa's avatar
Kenichi Handa committed
105
   `composition' property, it won't be collected as garbage until all
106
   bits of text that have the same COMPONENTS-VEC are deleted.
Kenichi Handa's avatar
Kenichi Handa committed
107 108

   The latter is a table of pointers to `struct composition' indexed
109
   by COMPOSITION-ID.  This structure keeps the other information (see
Kenichi Handa's avatar
Kenichi Handa committed
110 111 112 113
   composite.h).

   In general, a text property holds information about individual
   characters.  But, a `composition' property holds information about
114
   a sequence of characters (in this sense, it is like the `intangible'
Kenichi Handa's avatar
Kenichi Handa committed
115
   property).  That means that we should not share the property value
116
   in adjacent compositions -- we can't distinguish them if they have the
Kenichi Handa's avatar
Kenichi Handa committed
117 118 119 120 121 122
   same property.  So, after any changes, we call
   `update_compositions' and change a property of one of adjacent
   compositions to a copy of it.  This function also runs a proper
   composition modification function to make a composition that gets
   invalid by the change valid again.

123
   As the value of the `composition' property holds information about a
Kenichi Handa's avatar
Kenichi Handa committed
124
   specific range of text, the value gets invalid if we change the
125
   text in the range.  We treat the `composition' property as always
Kenichi Handa's avatar
Kenichi Handa committed
126 127 128
   rear-nonsticky (currently by setting default-text-properties to
   (rear-nonsticky (composition))) and we never make properties of
   adjacent compositions identical.  Thus, any such changes make the
129
   range just shorter.  So, we can check the validity of the `composition'
Kenichi Handa's avatar
Kenichi Handa committed
130 131 132 133 134 135 136 137 138 139 140 141
   property by comparing LENGTH information with the actual length of
   the composition.

*/


/* Table of pointers to the structure `composition' indexed by
   COMPOSITION-ID.  This structure is for storing information about
   each composition except for COMPONENTS-VEC.  */
struct composition **composition_table;

/* The current size of `composition_table'.  */
142
static ptrdiff_t composition_table_size;
Kenichi Handa's avatar
Kenichi Handa committed
143 144

/* Number of compositions currently made. */
145
ptrdiff_t n_compositions;
Kenichi Handa's avatar
Kenichi Handa committed
146 147 148 149 150 151

/* Hash table for compositions.  The key is COMPONENTS-VEC of
   `composition' property.  The value is the corresponding
   COMPOSITION-ID.  */
Lisp_Object composition_hash_table;

152 153
/* Maximum number of characters to look back for
   auto-compositions.  */
154 155
#define MAX_AUTO_COMPOSITION_LOOKBACK 3

Kenichi Handa's avatar
Kenichi Handa committed
156 157 158 159 160 161 162
/* Return COMPOSITION-ID of a composition at buffer position
   CHARPOS/BYTEPOS and length NCHARS.  The `composition' property of
   the sequence is PROP.  STRING, if non-nil, is a string that
   contains the composition instead of the current buffer.

   If the composition is invalid, return -1.  */

163
ptrdiff_t
164
get_composition_id (ptrdiff_t charpos, ptrdiff_t bytepos, ptrdiff_t nchars,
165
		    Lisp_Object prop, Lisp_Object string)
Kenichi Handa's avatar
Kenichi Handa committed
166
{
167
  Lisp_Object id, length, components, key, *key_contents, hash_code;
168
  ptrdiff_t glyph_len;
Kenichi Handa's avatar
Kenichi Handa committed
169
  struct Lisp_Hash_Table *hash_table = XHASH_TABLE (composition_hash_table);
170
  ptrdiff_t hash_index;
171
  enum composition_method method;
Kenichi Handa's avatar
Kenichi Handa committed
172
  struct composition *cmp;
173
  ptrdiff_t i;
174
  int ch;
Kenichi Handa's avatar
Kenichi Handa committed
175

176 177
  /* Maximum length of a string of glyphs.  XftGlyphExtents limits
     this to INT_MAX, and Emacs limits it further.  Divide INT_MAX - 1
178
     by 2 because gui_produce_glyphs computes glyph_len * 2 + 1.  Divide
179 180
     the size by MAX_MULTIBYTE_LENGTH because encode_terminal_code
     multiplies glyph_len by MAX_MULTIBYTE_LENGTH.  */
181
  enum {
182 183
    GLYPH_LEN_MAX = min ((INT_MAX - 1) / 2,
			 min (PTRDIFF_MAX, SIZE_MAX) / MAX_MULTIBYTE_LENGTH)
184 185
  };

Kenichi Handa's avatar
Kenichi Handa committed
186 187 188 189 190 191 192 193 194
  /* PROP should be
	Form-A: ((LENGTH . COMPONENTS) . MODIFICATION-FUNC)
     or
	Form-B: (COMPOSITION-ID . (LENGTH COMPONENTS-VEC . MODIFICATION-FUNC))
  */
  if (nchars == 0 || !CONSP (prop))
    goto invalid_composition;

  id = XCAR (prop);
195
  if (FIXNUMP (id))
Kenichi Handa's avatar
Kenichi Handa committed
196 197
    {
      /* PROP should be Form-B.  */
Tom Tromey's avatar
Tom Tromey committed
198
      if (XFIXNUM (id) < 0 || XFIXNUM (id) >= n_compositions)
Kenichi Handa's avatar
Kenichi Handa committed
199
	goto invalid_composition;
Tom Tromey's avatar
Tom Tromey committed
200
      return XFIXNUM (id);
Kenichi Handa's avatar
Kenichi Handa committed
201 202 203 204 205 206 207
    }

  /* PROP should be Form-A.
     Thus, ID should be (LENGTH . COMPONENTS).  */
  if (!CONSP (id))
    goto invalid_composition;
  length = XCAR (id);
Tom Tromey's avatar
Tom Tromey committed
208
  if (!FIXNUMP (length) || XFIXNUM (length) != nchars)
Kenichi Handa's avatar
Kenichi Handa committed
209 210 211 212 213 214 215 216
    goto invalid_composition;

  components = XCDR (id);

  /* Check if the same composition has already been registered or not
     by consulting composition_hash_table.  The key for this table is
     COMPONENTS (converted to a vector COMPONENTS-VEC) or, if it is
     nil, vector of characters in the composition range.  */
217
  if (FIXNUMP (components))
218
    key = make_vector (1, components);
Kenichi Handa's avatar
Kenichi Handa committed
219 220 221 222 223 224
  else if (STRINGP (components) || CONSP (components))
    key = Fvconcat (1, &components);
  else if (VECTORP (components))
    key = components;
  else if (NILP (components))
    {
225
      key = make_uninit_vector (nchars);
Kenichi Handa's avatar
Kenichi Handa committed
226 227 228 229
      if (STRINGP (string))
	for (i = 0; i < nchars; i++)
	  {
	    FETCH_STRING_CHAR_ADVANCE (ch, string, charpos, bytepos);
230
	    ASET (key, i, make_fixnum (ch));
Kenichi Handa's avatar
Kenichi Handa committed
231 232 233 234 235
	  }
      else
	for (i = 0; i < nchars; i++)
	  {
	    FETCH_CHAR_ADVANCE (ch, charpos, bytepos);
236
	    ASET (key, i, make_fixnum (ch));
Kenichi Handa's avatar
Kenichi Handa committed
237 238 239 240 241 242 243 244 245 246 247 248 249 250
	  }
    }
  else
    goto invalid_composition;

  hash_index = hash_lookup (hash_table, key, &hash_code);
  if (hash_index >= 0)
    {
      /* We have already registered the same composition.  Change PROP
	 from Form-A above to Form-B while replacing COMPONENTS with
	 COMPONENTS-VEC stored in the hash table.  We can directly
	 modify the cons cell of PROP because it is not shared.  */
      key = HASH_KEY (hash_table, hash_index);
      id = HASH_VALUE (hash_table, hash_index);
251
      XSETCAR (prop, id);
252
      XSETCDR (prop, Fcons (make_fixnum (nchars), Fcons (key, XCDR (prop))));
Tom Tromey's avatar
Tom Tromey committed
253
      return XFIXNUM (id);
Kenichi Handa's avatar
Kenichi Handa committed
254 255 256
    }

  /* This composition is a new one.  We must register it.  */
257

Kenichi Handa's avatar
Kenichi Handa committed
258
  /* Check if we have sufficient memory to store this information.  */
259 260 261
  if (composition_table_size <= n_compositions)
    composition_table = xpalloc (composition_table, &composition_table_size,
				 1, -1, sizeof *composition_table);
Kenichi Handa's avatar
Kenichi Handa committed
262

Paul Eggert's avatar
Paul Eggert committed
263
  key_contents = XVECTOR (key)->contents;
Kenichi Handa's avatar
Kenichi Handa committed
264 265 266 267

  /* Check if the contents of COMPONENTS are valid if COMPONENTS is a
     vector or a list.  It should be a sequence of:
	char1 rule1 char2 rule2 char3 ...    ruleN charN+1  */
268

269
  if (VECTORP (components)
270 271 272 273
      && ASIZE (components) >= 2
      && VECTORP (AREF (components, 0)))
    {
      /* COMPONENTS is a glyph-string.  */
274
      ptrdiff_t len = ASIZE (key);
275 276 277 278 279

      for (i = 1; i < len; i++)
	if (! VECTORP (AREF (key, i)))
	  goto invalid_composition;
    }
280
  else if (VECTORP (components) || CONSP (components))
Kenichi Handa's avatar
Kenichi Handa committed
281
    {
282
      ptrdiff_t len = ASIZE (key);
Kenichi Handa's avatar
Kenichi Handa committed
283 284 285 286 287 288 289 290

      /* The number of elements should be odd.  */
      if ((len % 2) == 0)
	goto invalid_composition;
      /* All elements should be integers (character or encoded
         composition rule).  */
      for (i = 0; i < len; i++)
	{
291
	  if (!FIXNUMP (key_contents[i]))
Kenichi Handa's avatar
Kenichi Handa committed
292 293 294 295 296 297 298
	    goto invalid_composition;
	}
    }

  /* Change PROP from Form-A above to Form-B.  We can directly modify
     the cons cell of PROP because it is not shared.  */
  XSETFASTINT (id, n_compositions);
299
  XSETCAR (prop, id);
300
  XSETCDR (prop, Fcons (make_fixnum (nchars), Fcons (key, XCDR (prop))));
Kenichi Handa's avatar
Kenichi Handa committed
301 302 303 304

  /* Register the composition in composition_hash_table.  */
  hash_index = hash_put (hash_table, key, id, hash_code);

305 306
  method = (NILP (components)
	    ? COMPOSITION_RELATIVE
307
	    : ((FIXNUMP (components) || STRINGP (components))
308 309 310 311 312 313 314
	       ? COMPOSITION_WITH_ALTCHARS
	       : COMPOSITION_WITH_RULE_ALTCHARS));

  glyph_len = (method == COMPOSITION_WITH_RULE_ALTCHARS
	       ? (ASIZE (key) + 1) / 2
	       : ASIZE (key));

315
  if (GLYPH_LEN_MAX < glyph_len)
316 317
    memory_full (SIZE_MAX);

Kenichi Handa's avatar
Kenichi Handa committed
318
  /* Register the composition in composition_table.  */
319
  cmp = xmalloc (sizeof *cmp);
Kenichi Handa's avatar
Kenichi Handa committed
320

321
  cmp->method = method;
Kenichi Handa's avatar
Kenichi Handa committed
322 323
  cmp->hash_index = hash_index;
  cmp->glyph_len = glyph_len;
324
  cmp->offsets = xnmalloc (glyph_len, 2 * sizeof *cmp->offsets);
Kenichi Handa's avatar
Kenichi Handa committed
325 326
  cmp->font = NULL;

327
  if (cmp->method != COMPOSITION_WITH_RULE_ALTCHARS)
Kenichi Handa's avatar
Kenichi Handa committed
328 329 330 331 332 333
    {
      /* Relative composition.  */
      cmp->width = 0;
      for (i = 0; i < glyph_len; i++)
	{
	  int this_width;
Tom Tromey's avatar
Tom Tromey committed
334
	  ch = XFIXNUM (key_contents[i]);
335 336
	  /* TAB in a composition means display glyphs with padding
	     space on the left or right.  */
Paul Eggert's avatar
Paul Eggert committed
337
	  this_width = (ch == '\t' ? 1 : CHARACTER_WIDTH (ch));
Kenichi Handa's avatar
Kenichi Handa committed
338 339 340 341 342 343 344
	  if (cmp->width < this_width)
	    cmp->width = this_width;
	}
    }
  else
    {
      /* Rule-base composition.  */
345
      double leftmost = 0.0, rightmost;
Kenichi Handa's avatar
Kenichi Handa committed
346

Tom Tromey's avatar
Tom Tromey committed
347
      ch = XFIXNUM (key_contents[0]);
Paul Eggert's avatar
Paul Eggert committed
348
      rightmost = ch != '\t' ? CHARACTER_WIDTH (ch) : 1;
Kenichi Handa's avatar
Kenichi Handa committed
349 350 351

      for (i = 1; i < glyph_len; i += 2)
	{
352
	  int rule, gref, nref;
Kenichi Handa's avatar
Kenichi Handa committed
353
	  int this_width;
354
	  double this_left;
Kenichi Handa's avatar
Kenichi Handa committed
355

Tom Tromey's avatar
Tom Tromey committed
356 357
	  rule = XFIXNUM (key_contents[i]);
	  ch = XFIXNUM (key_contents[i + 1]);
Paul Eggert's avatar
Paul Eggert committed
358
	  this_width = ch != '\t' ? CHARACTER_WIDTH (ch) : 1;
Kenichi Handa's avatar
Kenichi Handa committed
359 360 361 362 363 364 365 366 367 368 369 370 371 372 373

	  /* A composition rule is specified by an integer value
	     that encodes global and new reference points (GREF and
	     NREF).  GREF and NREF are specified by numbers as
	     below:
		0---1---2 -- ascent
		|       |
		|       |
		|       |
		9--10--11 -- center
		|       |
	     ---3---4---5--- baseline
		|       |
		6---7---8 -- descent
	  */
374
	  COMPOSITION_DECODE_REFS (rule, gref, nref);
Kenichi Handa's avatar
Kenichi Handa committed
375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400
	  this_left = (leftmost
		       + (gref % 3) * (rightmost - leftmost) / 2.0
		       - (nref % 3) * this_width / 2.0);

	  if (this_left < leftmost)
	    leftmost = this_left;
	  if (this_left + this_width > rightmost)
	    rightmost = this_left + this_width;
	}

      cmp->width = rightmost - leftmost;
      if (cmp->width < (rightmost - leftmost))
	/* To get a ceiling integer value.  */
	cmp->width++;
    }

  composition_table[n_compositions] = cmp;

  return n_compositions++;

 invalid_composition:
  /* Would it be better to remove this `composition' property?  */
  return -1;
}


401 402
/* Find a static composition at or nearest to position POS of OBJECT
   (buffer or string).
Kenichi Handa's avatar
Kenichi Handa committed
403 404 405 406 407 408 409 410 411 412 413 414 415 416 417

   OBJECT defaults to the current buffer.  If there's a composition at
   POS, set *START and *END to the start and end of the sequence,
   *PROP to the `composition' property, and return 1.

   If there's no composition at POS and LIMIT is negative, return 0.

   Otherwise, search for a composition forward (LIMIT > POS) or
   backward (LIMIT < POS).  In this case, LIMIT bounds the search.

   If a composition is found, set *START, *END, and *PROP as above,
   and return 1, else return 0.

   This doesn't check the validity of composition.  */

418
bool
419 420
find_composition (ptrdiff_t pos, ptrdiff_t limit,
		  ptrdiff_t *start, ptrdiff_t *end,
421
		  Lisp_Object *prop, Lisp_Object object)
Kenichi Handa's avatar
Kenichi Handa committed
422 423 424 425 426 427 428 429 430 431
{
  Lisp_Object val;

  if (get_property_and_range (pos, Qcomposition, prop, start, end, object))
    return 1;

  if (limit < 0 || limit == pos)
    return 0;

  if (limit > pos)		/* search forward */
432
    {
433 434
      val = Fnext_single_property_change (make_fixnum (pos), Qcomposition,
					  object, make_fixnum (limit));
Tom Tromey's avatar
Tom Tromey committed
435
      pos = XFIXNUM (val);
436 437 438
      if (pos == limit)
	return 0;
    }
Kenichi Handa's avatar
Kenichi Handa committed
439
  else				/* search backward */
440 441 442 443
    {
      if (get_property_and_range (pos - 1, Qcomposition, prop, start, end,
				  object))
	return 1;
444 445
      val = Fprevious_single_property_change (make_fixnum (pos), Qcomposition,
					      object, make_fixnum (limit));
Tom Tromey's avatar
Tom Tromey committed
446
      pos = XFIXNUM (val);
447 448 449 450
      if (pos == limit)
	return 0;
      pos--;
    }
Kenichi Handa's avatar
Kenichi Handa committed
451 452 453 454 455 456 457 458
  get_property_and_range (pos, Qcomposition, prop, start, end, object);
  return 1;
}

/* Run a proper function to adjust the composition sitting between
   FROM and TO with property PROP.  */

static void
459
run_composition_function (ptrdiff_t from, ptrdiff_t to, Lisp_Object prop)
Kenichi Handa's avatar
Kenichi Handa committed
460
{
461
  Lisp_Object func;
462
  ptrdiff_t start, end;
Kenichi Handa's avatar
Kenichi Handa committed
463 464 465 466 467 468

  func = COMPOSITION_MODIFICATION_FUNC (prop);
  /* If an invalid composition precedes or follows, try to make them
     valid too.  */
  if (from > BEGV
      && find_composition (from - 1, -1, &start, &end, &prop, Qnil)
469
      && !composition_valid_p (start, end, prop))
Kenichi Handa's avatar
Kenichi Handa committed
470 471 472
    from = start;
  if (to < ZV
      && find_composition (to, -1, &start, &end, &prop, Qnil)
473
      && !composition_valid_p (start, end, prop))
Kenichi Handa's avatar
Kenichi Handa committed
474
    to = end;
Kenichi Handa's avatar
Kenichi Handa committed
475
  if (!NILP (Ffboundp (func)))
476
    call2 (func, make_fixnum (from), make_fixnum (to));
Kenichi Handa's avatar
Kenichi Handa committed
477 478 479 480 481 482
}

/* Make invalid compositions adjacent to or inside FROM and TO valid.
   CHECK_MASK is bitwise `or' of mask bits defined by macros
   CHECK_XXX (see the comment in composite.h).

483
   It also resets the text-property `auto-composed' to a proper region
Kenichi Handa's avatar
Kenichi Handa committed
484 485
   so that automatic character composition works correctly later while
   displaying the region.
Kenichi Handa's avatar
Kenichi Handa committed
486

Kenichi Handa's avatar
Kenichi Handa committed
487 488 489 490
   This function is called when a buffer text is changed.  If the
   change is deletion, FROM == TO.  Otherwise, FROM < TO.  */

void
491
update_compositions (ptrdiff_t from, ptrdiff_t to, int check_mask)
Kenichi Handa's avatar
Kenichi Handa committed
492
{
493
  Lisp_Object prop;
494
  ptrdiff_t start, end;
Kenichi Handa's avatar
Kenichi Handa committed
495 496
  /* The beginning and end of the region to set the property
     `auto-composed' to nil.  */
497
  ptrdiff_t min_pos = from, max_pos = to;
Kenichi Handa's avatar
Kenichi Handa committed
498

499 500
  if (inhibit_modification_hooks)
    return;
501

502 503 504 505
  /* If FROM and TO are not in a valid range, do nothing.  */
  if (! (BEGV <= from && from <= to && to <= ZV))
    return;

Kenichi Handa's avatar
Kenichi Handa committed
506 507 508 509 510 511 512 513
  if (check_mask & CHECK_HEAD)
    {
      /* FROM should be at composition boundary.  But, insertion or
	 deletion will make two compositions adjacent and
	 indistinguishable when they have same (eq) property.  To
	 avoid it, in such a case, we change the property of the
	 latter to the copy of it.  */
      if (from > BEGV
514
	  && find_composition (from - 1, -1, &start, &end, &prop, Qnil)
515
	  && composition_valid_p (start, end, prop))
Kenichi Handa's avatar
Kenichi Handa committed
516
	{
Kenichi Handa's avatar
Kenichi Handa committed
517 518 519
	  min_pos = start;
	  if (end > to)
	    max_pos = end;
Kenichi Handa's avatar
Kenichi Handa committed
520
	  if (from < end)
521
	    Fput_text_property (make_fixnum (from), make_fixnum (end),
Kenichi Handa's avatar
Kenichi Handa committed
522 523 524 525 526
				Qcomposition,
				Fcons (XCAR (prop), XCDR (prop)), Qnil);
	  run_composition_function (start, end, prop);
	  from = end;
	}
527
      else if (from < ZV
528
	       && find_composition (from, -1, &start, &from, &prop, Qnil)
529
	       && composition_valid_p (start, from, prop))
Kenichi Handa's avatar
Kenichi Handa committed
530 531 532 533 534
	{
	  if (from > to)
	    max_pos = from;
	  run_composition_function (start, from, prop);
	}
Kenichi Handa's avatar
Kenichi Handa committed
535 536 537 538 539 540 541 542 543
    }

  if (check_mask & CHECK_INSIDE)
    {
      /* In this case, we are sure that (check & CHECK_TAIL) is also
         nonzero.  Thus, here we should check only compositions before
         (to - 1).  */
      while (from < to - 1
	     && find_composition (from, to, &start, &from, &prop, Qnil)
544
	     && composition_valid_p (start, from, prop)
Kenichi Handa's avatar
Kenichi Handa committed
545 546 547 548 549 550 551
	     && from < to - 1)
	run_composition_function (start, from, prop);
    }

  if (check_mask & CHECK_TAIL)
    {
      if (from < to
552
	  && find_composition (to - 1, -1, &start, &end, &prop, Qnil)
553
	  && composition_valid_p (start, end, prop))
Kenichi Handa's avatar
Kenichi Handa committed
554 555 556 557 558 559 560
	{
	  /* TO should be also at composition boundary.  But,
	     insertion or deletion will make two compositions adjacent
	     and indistinguishable when they have same (eq) property.
	     To avoid it, in such a case, we change the property of
	     the former to the copy of it.  */
	  if (to < end)
Kenichi Handa's avatar
Kenichi Handa committed
561
	    {
562
	      Fput_text_property (make_fixnum (start), make_fixnum (to),
Kenichi Handa's avatar
Kenichi Handa committed
563 564 565 566
				  Qcomposition,
				  Fcons (XCAR (prop), XCDR (prop)), Qnil);
	      max_pos = end;
	    }
Kenichi Handa's avatar
Kenichi Handa committed
567 568 569
	  run_composition_function (start, end, prop);
	}
      else if (to < ZV
570
	       && find_composition (to, -1, &start, &end, &prop, Qnil)
571
	       && composition_valid_p (start, end, prop))
Kenichi Handa's avatar
Kenichi Handa committed
572 573 574 575
	{
	  run_composition_function (start, end, prop);
	  max_pos = end;
	}
Kenichi Handa's avatar
Kenichi Handa committed
576
    }
Kenichi Handa's avatar
Kenichi Handa committed
577
  if (min_pos < max_pos)
578
    {
579
      ptrdiff_t count = SPECPDL_INDEX ();
580 581 582 583

      specbind (Qinhibit_read_only, Qt);
      specbind (Qinhibit_modification_hooks, Qt);
      specbind (Qinhibit_point_motion_hooks, Qt);
584 585
      Fremove_list_of_text_properties (make_fixnum (min_pos),
				       make_fixnum (max_pos),
586
				       list1 (Qauto_composed), Qnil);
587 588
      unbind_to (count, Qnil);
    }
Kenichi Handa's avatar
Kenichi Handa committed
589 590
}

591 592 593 594 595 596

/* Modify composition property values in LIST destructively.  LIST is
   a list as returned from text_property_list.  Change values to the
   top-level copies of them so that none of them are `eq'.  */

void
597
make_composition_value_copy (Lisp_Object list)
598 599 600 601 602 603 604 605 606 607
{
  Lisp_Object plist, val;

  for (; CONSP (list); list = XCDR (list))
    {
      plist = XCAR (XCDR (XCDR (XCAR (list))));
      while (CONSP (plist) && CONSP (XCDR (plist)))
	{
	  if (EQ (XCAR (plist), Qcomposition)
	      && (val = XCAR (XCDR (plist)), CONSP (val)))
608
	    XSETCAR (XCDR (plist), Fcons (XCAR (val), XCDR (val)));
609 610 611 612 613 614
	  plist = XCDR (XCDR (plist));
	}
    }
}


Kenichi Handa's avatar
Kenichi Handa committed
615 616 617 618 619 620 621
/* Make text in the region between START and END a composition that
   has COMPONENTS and MODIFICATION-FUNC.

   If STRING is non-nil, then operate on characters contained between
   indices START and END in STRING.  */

void
622
compose_text (ptrdiff_t start, ptrdiff_t end, Lisp_Object components,
623
	      Lisp_Object modification_func, Lisp_Object string)
Kenichi Handa's avatar
Kenichi Handa committed
624 625 626
{
  Lisp_Object prop;

627
  prop = Fcons (Fcons (make_fixnum (end - start), components),
Kenichi Handa's avatar
Kenichi Handa committed
628
		modification_func);
629
  Fput_text_property  (make_fixnum (start), make_fixnum (end),
Kenichi Handa's avatar
Kenichi Handa committed
630 631
		       Qcomposition, prop, string);
}
632

633
/* Lisp glyph-string handlers.  */
634 635 636 637 638 639 640

/* Hash table for automatic composition.  The key is a header of a
   lgstring (Lispy glyph-string), and the value is a body of a
   lgstring.  */

static Lisp_Object gstring_hash_table;

641
static Lisp_Object gstring_lookup_cache (Lisp_Object);
642 643

static Lisp_Object
644
gstring_lookup_cache (Lisp_Object header)
645 646
{
  struct Lisp_Hash_Table *h = XHASH_TABLE (gstring_hash_table);
647
  ptrdiff_t i = hash_lookup (h, header, NULL);
648 649 650 651 652

  return (i >= 0 ? HASH_VALUE (h, i) : Qnil);
}

Lisp_Object
653
composition_gstring_put_cache (Lisp_Object gstring, ptrdiff_t len)
654 655
{
  struct Lisp_Hash_Table *h = XHASH_TABLE (gstring_hash_table);
Daniel Colascione's avatar
Daniel Colascione committed
656
  hash_rehash_if_needed (h);
657
  Lisp_Object header = LGSTRING_HEADER (gstring);
658
  Lisp_Object hash = h->test.hashfn (header, h);
659 660
  if (len < 0)
    {
661 662 663
      ptrdiff_t glyph_len = LGSTRING_GLYPH_LEN (gstring);
      for (len = 0; len < glyph_len; len++)
	if (NILP (LGSTRING_GLYPH (gstring, len)))
664 665
	  break;
    }
Juanma Barranquero's avatar
Juanma Barranquero committed
666

667
  Lisp_Object copy = make_nil_vector (len + 2);
668
  LGSTRING_SET_HEADER (copy, Fcopy_sequence (header));
669
  for (ptrdiff_t i = 0; i < len; i++)
670
    LGSTRING_SET_GLYPH (copy, i, Fcopy_sequence (LGSTRING_GLYPH (gstring, i)));
671 672
  ptrdiff_t id = hash_put (h, LGSTRING_HEADER (copy), copy, hash);
  LGSTRING_SET_ID (copy, make_fixnum (id));
673 674 675 676
  return copy;
}

Lisp_Object
677
composition_gstring_from_id (ptrdiff_t id)
678 679 680 681 682 683
{
  struct Lisp_Hash_Table *h = XHASH_TABLE (gstring_hash_table);

  return HASH_VALUE (h, id);
}

684 685 686 687 688 689
DEFUN ("clear-composition-cache", Fclear_composition_cache,
       Sclear_composition_cache, 0, 0, 0,
       doc: /* Internal use only.
Clear composition cache.  */)
  (void)
{
690
  Lisp_Object args[] = {QCtest, Qequal, QCsize, make_fixnum (311)};
691 692 693 694
  gstring_hash_table = CALLMANY (Fmake_hash_table, args);
  /* Fixme: We call Fclear_face_cache to force complete re-building of
     display glyphs.  But, it may be better to call this function from
     Fclear_face_cache instead.  */
695
  return Fclear_face_cache (Qt);
696 697
}

698
bool
699
composition_gstring_p (Lisp_Object gstring)
700 701
{
  Lisp_Object header;
702
  ptrdiff_t i;
703 704 705 706 707 708 709

  if (! VECTORP (gstring) || ASIZE (gstring) < 2)
    return 0;
  header = LGSTRING_HEADER (gstring);
  if (! VECTORP (header) || ASIZE (header) < 2)
    return 0;
  if (! NILP (LGSTRING_FONT (gstring))
710 711
      && (! FONT_OBJECT_P (LGSTRING_FONT (gstring))
	  && ! CODING_SYSTEM_P (LGSTRING_FONT (gstring))))
712 713
    return 0;
  for (i = 1; i < ASIZE (LGSTRING_HEADER (gstring)); i++)
714
    if (! FIXNATP (AREF (LGSTRING_HEADER (gstring), i)))
715
      return 0;
716
  if (! NILP (LGSTRING_ID (gstring)) && ! FIXNATP (LGSTRING_ID (gstring)))
717 718 719 720 721 722 723 724 725 726 727 728 729
    return 0;
  for (i = 0; i < LGSTRING_GLYPH_LEN (gstring); i++)
    {
      Lisp_Object glyph = LGSTRING_GLYPH (gstring, i);
      if (NILP (glyph))
	break;
      if (! VECTORP (glyph) || ASIZE (glyph) != LGLYPH_SIZE)
	return 0;
    }
  return 1;
}

int
730
composition_gstring_width (Lisp_Object gstring, ptrdiff_t from, ptrdiff_t to,
731
			   struct font_metrics *metrics)
732 733 734 735 736 737 738 739
{
  Lisp_Object *glyph;
  int width = 0;

  if (metrics)
    {
      Lisp_Object font_object = LGSTRING_FONT (gstring);

740 741 742
      if (FONT_OBJECT_P (font_object))
	{
	  struct font *font = XFONT_OBJECT (font_object);
743
	  int font_ascent, font_descent;
744

745 746 747
	  get_font_ascent_descent (font, &font_ascent, &font_descent);
	  metrics->ascent = font_ascent;
	  metrics->descent = font_descent;
748 749 750 751 752 753
	}
      else
	{
	  metrics->ascent = 1;
	  metrics->descent = 0;
	}
754 755
      metrics->width = metrics->lbearing = metrics->rbearing = 0;
    }
756
  for (glyph = lgstring_glyph_addr (gstring, from); from < to; from++, glyph++)
757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775
    {
      int x;

      if (NILP (LGLYPH_ADJUSTMENT (*glyph)))
	width += LGLYPH_WIDTH (*glyph);
      else
	width += LGLYPH_WADJUST (*glyph);
      if (metrics)
	{
	  x = metrics->width + LGLYPH_LBEARING (*glyph) + LGLYPH_XOFF (*glyph);
	  if (metrics->lbearing > x)
	    metrics->lbearing = x;
	  x = metrics->width + LGLYPH_RBEARING (*glyph) + LGLYPH_XOFF (*glyph);
	  if (metrics->rbearing < x)
	    metrics->rbearing = x;
	  metrics->width = width;
	  x = LGLYPH_ASCENT (*glyph) - LGLYPH_YOFF (*glyph);
	  if (metrics->ascent < x)
	    metrics->ascent = x;
776
	  x = LGLYPH_DESCENT (*glyph) + LGLYPH_YOFF (*glyph);
777 778 779 780 781 782 783 784 785 786 787 788
	  if (metrics->descent < x)
	    metrics->descent = x;
	}
    }
  return width;
}


static Lisp_Object gstring_work;
static Lisp_Object gstring_work_headers;

static Lisp_Object
Paul Eggert's avatar
Paul Eggert committed
789
fill_gstring_header (ptrdiff_t from, ptrdiff_t from_byte,
790
		     ptrdiff_t to, Lisp_Object font_object, Lisp_Object string)
791
{
Paul Eggert's avatar
Paul Eggert committed
792
  ptrdiff_t len = to - from;
793 794
  if (len == 0)
    error ("Attempt to shape zero-length text");
Paul Eggert's avatar
Paul Eggert committed
795 796 797 798
  eassume (0 < len);
  Lisp_Object header = (len <= 8
			? AREF (gstring_work_headers, len - 1)
			: make_uninit_vector (len + 1));
799 800

  ASET (header, 0, font_object);
Paul Eggert's avatar
Paul Eggert committed
801
  for (ptrdiff_t i = 0; i < len; i++)
802 803 804 805 806 807 808
    {
      int c;

      if (NILP (string))
	FETCH_CHAR_ADVANCE_NO_CHECK (c, from, from_byte);
      else
	FETCH_STRING_CHAR_ADVANCE_NO_CHECK (c, string, from, from_byte);
809
      ASET (header, i + 1, make_fixnum (c));
810 811 812 813 814
    }
  return header;
}

static void
815
fill_gstring_body (Lisp_Object gstring)
816 817 818
{
  Lisp_Object font_object = LGSTRING_FONT (gstring);
  Lisp_Object header = AREF (gstring, 0);
819 820
  ptrdiff_t len = LGSTRING_CHAR_LEN (gstring);
  ptrdiff_t i;
821 822 823 824 825
  struct font *font = NULL;
  unsigned int code;

  if (FONT_OBJECT_P (font_object))
    font = XFONT_OBJECT (font_object);
826 827 828 829

  for (i = 0; i < len; i++)
    {
      Lisp_Object g = LGSTRING_GLYPH (gstring, i);
Tom Tromey's avatar
Tom Tromey committed
830
      int c = XFIXNAT (AREF (header, i + 1));
831 832 833 834 835 836 837 838 839

      if (NILP (g))
	{
	  g = LGLYPH_NEW ();
	  LGSTRING_SET_GLYPH (gstring, i, g);
	}
      LGLYPH_SET_FROM (g, i);
      LGLYPH_SET_TO (g, i);
      LGLYPH_SET_CHAR (g, c);
840 841 842 843 844 845 846 847 848

      if (font != NULL)
        code = font->driver->encode_char (font, LGLYPH_CHAR (g));
      else
        code = FONT_INVALID_CODE;
      if (code != FONT_INVALID_CODE)
        {
	  font_fill_lglyph_metrics (g, font, code);
        }
849 850
      else
	{
Tom Tromey's avatar
Tom Tromey committed
851
	  int width = XFIXNAT (CHAR_TABLE_REF (Vchar_width_table, c));
852 853 854 855 856 857 858 859 860 861 862 863 864 865 866

	  LGLYPH_SET_CODE (g, c);
	  LGLYPH_SET_LBEARING (g, 0);
	  LGLYPH_SET_RBEARING (g, width);
	  LGLYPH_SET_WIDTH (g, width);
	  LGLYPH_SET_ASCENT (g, 1);
	  LGLYPH_SET_DESCENT (g, 0);
	}
      LGLYPH_SET_ADJUSTMENT (g, Qnil);
    }
  if (i < LGSTRING_GLYPH_LEN (gstring))
    LGSTRING_SET_GLYPH (gstring, i, Qnil);
}


867 868 869 870 871 872
/* Try to compose the characters at CHARPOS according to composition
   rule RULE ([PATTERN PREV-CHARS FUNC]).  LIMIT limits the characters
   to compose.  STRING, if not nil, is a target string.  WIN is a
   window where the characters are being displayed.  If characters are
   successfully composed, return the composition as a glyph-string
   object.  Otherwise return nil.  */
873 874

static Lisp_Object
875 876
autocmp_chars (Lisp_Object rule, ptrdiff_t charpos, ptrdiff_t bytepos,
	       ptrdiff_t limit, struct window *win, struct face *face,
877
	       Lisp_Object string, Lisp_Object direction)
878
{
879
  ptrdiff_t count = SPECPDL_INDEX ();
880
  Lisp_Object pos = make_fixnum (charpos);
881 882
  ptrdiff_t to;
  ptrdiff_t pt = PT, pt_byte = PT_BYTE;
883
  Lisp_Object re, font_object, lgstring;
884
  ptrdiff_t len;
Juanma Barranquero's avatar
Juanma Barranquero committed
885

886
  record_unwind_save_match_data ();
887 888 889 890 891 892 893
  re = AREF (rule, 0);
  if (NILP (re))
    len = 1;
  else if (! STRINGP (re))
    return unbind_to (count, Qnil);
  else if ((len = fast_looking_at (re, charpos, bytepos, limit, -1, string))
	   > 0)
894
    {
895 896 897 898 899 900 901 902
      if (NILP (string))
	len = BYTE_TO_CHAR (bytepos + len) - charpos;
      else
	len = string_byte_to_char (string, bytepos + len) - charpos;
    }
  if (len <= 0)
    return unbind_to (count, Qnil);
  to = limit = charpos + len;
Eli Zaretskii's avatar
Eli Zaretskii committed
903 904
  font_object = win->frame;
#ifdef HAVE_WINDOW_SYSTEM
905
  struct frame *f = XFRAME (font_object);
906 907
  if (FRAME_WINDOW_P (f))
    {
908
      font_object = font_range (charpos, bytepos, &to, win, face, string);
909 910 911 912 913 914
      if (! FONT_OBJECT_P (font_object)
	  || (! NILP (re)
	      && to < limit
	      && (fast_looking_at (re, charpos, bytepos, to, -1, string) <= 0)))
	return unbind_to (count, Qnil);
    }
Eli Zaretskii's avatar
Eli Zaretskii committed
915
#endif
916
  lgstring = Fcomposition_get_gstring (pos, make_fixnum (to), font_object,
917 918 919 920 921
				       string);
  if (NILP (LGSTRING_ID (lgstring)))
    {
      /* Save point as marker before calling out to lisp.  */
      if (NILP (string))
922 923
	record_unwind_protect (restore_point_unwind,
			       build_marker (current_buffer, pt, pt_byte));
924 925 926
      lgstring = safe_call (7, Vauto_composition_function, AREF (rule, 2),
			    pos, make_fixnum (to), font_object, string,
			    direction);
927
    }
928
  return unbind_to (count, lgstring);
929 930
}

931
/* 1 iff the character C is composable.  Characters of general
932 933
   category Z? or C? are not composable except for ZWNJ and ZWJ,
   and characters of category Zs. */
934

935 936 937 938
static bool
char_composable_p (int c)
{
  Lisp_Object val;
939
  return (c >= ' '
940
	  && (c == ZERO_WIDTH_NON_JOINER || c == ZERO_WIDTH_JOINER
941
	      || (val = CHAR_TABLE_REF (Vunicode_category_table, c),
942
		  (FIXNUMP (val) && (XFIXNUM (val) <= UNICODE_CATEGORY_Zs)))));
943
}
944 945 946

/* Update cmp_it->stop_pos to the next position after CHARPOS (and
   BYTEPOS) where character composition may happen.  If BYTEPOS is
947 948 949 950 951 952 953 954 955 956
   negative, compute it.  ENDPOS is a limit of searching.  If it is
   less than CHARPOS, search backward to ENDPOS+1 assuming that
   set_iterator_to_next works in reverse order.  In this case, if a
   composition closest to CHARPOS is found, set cmp_it->stop_pos to
   the last character of the composition.

   If no composition is found, set cmp_it->ch to -2.  If a static
   composition is found, set cmp_it->ch to -1.  Otherwise, set
   cmp_it->ch to the character that triggers the automatic
   composition.  */
957 958

void
959
composition_compute_stop_pos (struct composition_it *cmp_it, ptrdiff_t charpos, ptrdiff_t bytepos, ptrdiff_t endpos, Lisp_Object string)
960
{
961
  ptrdiff_t start, end;
962
  int c;
963
  Lisp_Object prop, val;
964 965
  /* This is from forward_to_next_line_start in xdisp.c.  */
  const int MAX_NEWLINE_DISTANCE = 500;
966

967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982
  if (charpos < endpos)
    {
      if (endpos > charpos + MAX_NEWLINE_DISTANCE)
	endpos = charpos + MAX_NEWLINE_DISTANCE;
    }
  else if (endpos < charpos)
    {
      /* We search backward for a position to check composition.  */
      if (endpos < 0)
	{
	  /* But we don't know where to stop the searching.  */
	  endpos = NILP (string) ? BEGV - 1 : -1;
	  /* Usually we don't reach ENDPOS because we stop searching
	     at an uncomposable character (NL, LRE, etc).  */
	}
    }
983
  cmp_it->id = -1;
984
  cmp_it->ch = -2;
985 986 987 988 989 990 991
  cmp_it->reversed_p = 0;
  cmp_it->stop_pos = endpos;
  if (charpos == endpos)
    return;
  /* FIXME: Bidi is not yet handled well in static composition.  */
  if (charpos < endpos
      && find_composition (charpos, endpos, &start, &end, &prop, string)
992
      && start >= charpos
993
      && composition_valid_p (start, end, prop))
994 995 996 997
    {
      cmp_it->stop_pos = endpos = start;
      cmp_it->ch = -1;
    }
Tom Tromey's avatar
Tom Tromey committed
998
  if (NILP (BVAR (current_buffer, enable_multibyte_characters))
999
      || NILP (Vauto_composition_mode))
1000 1001 1002
    return;
  if (bytepos < 0)
    {
1003
      if (NILP (string))
1004
	bytepos = CHAR_TO_BYTE (charpos);
1005 1006
      else
	bytepos = string_char_to_byte (string, charpos);
1007 1008 1009
    }

  start = charpos;
1010
  if (charpos < endpos)
1011
    {
1012 1013
      /* Forward search.  */
      while (charpos < endpos)
1014
	{
1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026
	  if (STRINGP (string))
	    FETCH_STRING_CHAR_ADVANCE (c, string, charpos, bytepos);
	  else
	    FETCH_CHAR_ADVANCE (c, charpos, bytepos);
	  if (c == '\n')
	    {
	      cmp_it->ch = -2;
	      break;
	    }
	  val = CHAR_TABLE_REF (Vcomposition_function_table, c);
	  if (! NILP (val))
	    {
1027
	      for (EMACS_INT ridx = 0; CONSP (val); val = XCDR (val), ridx++)
1028
		{
1029
		  Lisp_Object elt = XCAR (val);
1030
		  if (VECTORP (elt) && ASIZE (elt) == 3
1031
		      && FIXNATP (AREF (elt, 1))
Tom Tromey's avatar
Tom Tromey committed
1032
		      && charpos - 1 - XFIXNAT (AREF (elt, 1)) >= start)
1033 1034
		    {
		      cmp_it->rule_idx = ridx;
Tom Tromey's avatar
Tom Tromey committed
1035
		      cmp_it->lookback = XFIXNAT (AREF (elt, 1));
1036 1037 1038 1039
		      cmp_it->stop_pos = charpos - 1 - cmp_it->lookback;
		      cmp_it->ch = c;
		      return;
		    }
1040 1041
		}
	    }
1042
	}
1043 1044
      if (charpos == endpos
	  && !(STRINGP (string) && endpos == SCHARS (string)))
1045 1046 1047 1048 1049 1050 1051 1052 1053
	{
	  /* We couldn't find a composition point before ENDPOS.  But,
	     some character after ENDPOS may be composed with
	     characters before ENDPOS.  So, we should stop at the safe
	     point.  */
	  charpos = endpos - MAX_AUTO_COMPOSITION_LOOKBACK;
	  if (charpos < start)
	    charpos = start;
	}
1054
    }
1055
  else if (charpos > endpos)
1056 1057 1058 1059
    {
      /* Search backward for a pattern that may be composed and the
	 position of (possibly) the last character of the match is
	 closest to (but not after) START.  The reason for the last