1/*
2 * Copyright © 2011,2012 Google, Inc.
3 *
4 * This is part of HarfBuzz, a text shaping library.
5 *
6 * Permission is hereby granted, without written agreement and without
7 * license or royalty fees, to use, copy, modify, and distribute this
8 * software and its documentation for any purpose, provided that the
9 * above copyright notice and the following two paragraphs appear in
10 * all copies of this software.
11 *
12 * IN NO EVENT SHALL THE COPYRIGHT HOLDER BE LIABLE TO ANY PARTY FOR
13 * DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES
14 * ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS DOCUMENTATION, EVEN
15 * IF THE COPYRIGHT HOLDER HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH
16 * DAMAGE.
17 *
18 * THE COPYRIGHT HOLDER SPECIFICALLY DISCLAIMS ANY WARRANTIES, INCLUDING,
19 * BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
20 * FITNESS FOR A PARTICULAR PURPOSE. THE SOFTWARE PROVIDED HEREUNDER IS
21 * ON AN "AS IS" BASIS, AND THE COPYRIGHT HOLDER HAS NO OBLIGATION TO
22 * PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
23 *
24 * Google Author(s): Behdad Esfahbod
25 */
26
27#include "hb-ot-shape-fallback.hh"
28#include "hb-ot-layout-gsubgpos.hh"
29
30static unsigned int
31recategorize_combining_class (hb_codepoint_t u,
32 unsigned int klass)
33{
34 if (klass >= 200)
35 return klass;
36
37 /* Thai / Lao need some per-character work. */
38 if ((u & ~0xFF) == 0x0E00u)
39 {
40 if (unlikely (klass == 0))
41 {
42 switch (u)
43 {
44 case 0x0E31u:
45 case 0x0E34u:
46 case 0x0E35u:
47 case 0x0E36u:
48 case 0x0E37u:
49 case 0x0E47u:
50 case 0x0E4Cu:
51 case 0x0E4Du:
52 case 0x0E4Eu:
53 klass = HB_UNICODE_COMBINING_CLASS_ABOVE_RIGHT;
54 break;
55
56 case 0x0EB1u:
57 case 0x0EB4u:
58 case 0x0EB5u:
59 case 0x0EB6u:
60 case 0x0EB7u:
61 case 0x0EBBu:
62 case 0x0ECCu:
63 case 0x0ECDu:
64 klass = HB_UNICODE_COMBINING_CLASS_ABOVE;
65 break;
66
67 case 0x0EBCu:
68 klass = HB_UNICODE_COMBINING_CLASS_BELOW;
69 break;
70 }
71 } else {
72 /* Thai virama is below-right */
73 if (u == 0x0E3Au)
74 klass = HB_UNICODE_COMBINING_CLASS_BELOW_RIGHT;
75 }
76 }
77
78 switch (klass)
79 {
80
81 /* Hebrew */
82
83 case HB_MODIFIED_COMBINING_CLASS_CCC10: /* sheva */
84 case HB_MODIFIED_COMBINING_CLASS_CCC11: /* hataf segol */
85 case HB_MODIFIED_COMBINING_CLASS_CCC12: /* hataf patah */
86 case HB_MODIFIED_COMBINING_CLASS_CCC13: /* hataf qamats */
87 case HB_MODIFIED_COMBINING_CLASS_CCC14: /* hiriq */
88 case HB_MODIFIED_COMBINING_CLASS_CCC15: /* tsere */
89 case HB_MODIFIED_COMBINING_CLASS_CCC16: /* segol */
90 case HB_MODIFIED_COMBINING_CLASS_CCC17: /* patah */
91 case HB_MODIFIED_COMBINING_CLASS_CCC18: /* qamats */
92 case HB_MODIFIED_COMBINING_CLASS_CCC20: /* qubuts */
93 case HB_MODIFIED_COMBINING_CLASS_CCC22: /* meteg */
94 return HB_UNICODE_COMBINING_CLASS_BELOW;
95
96 case HB_MODIFIED_COMBINING_CLASS_CCC23: /* rafe */
97 return HB_UNICODE_COMBINING_CLASS_ATTACHED_ABOVE;
98
99 case HB_MODIFIED_COMBINING_CLASS_CCC24: /* shin dot */
100 return HB_UNICODE_COMBINING_CLASS_ABOVE_RIGHT;
101
102 case HB_MODIFIED_COMBINING_CLASS_CCC25: /* sin dot */
103 case HB_MODIFIED_COMBINING_CLASS_CCC19: /* holam */
104 return HB_UNICODE_COMBINING_CLASS_ABOVE_LEFT;
105
106 case HB_MODIFIED_COMBINING_CLASS_CCC26: /* point varika */
107 return HB_UNICODE_COMBINING_CLASS_ABOVE;
108
109 case HB_MODIFIED_COMBINING_CLASS_CCC21: /* dagesh */
110 break;
111
112
113 /* Arabic and Syriac */
114
115 case HB_MODIFIED_COMBINING_CLASS_CCC27: /* fathatan */
116 case HB_MODIFIED_COMBINING_CLASS_CCC28: /* dammatan */
117 case HB_MODIFIED_COMBINING_CLASS_CCC30: /* fatha */
118 case HB_MODIFIED_COMBINING_CLASS_CCC31: /* damma */
119 case HB_MODIFIED_COMBINING_CLASS_CCC33: /* shadda */
120 case HB_MODIFIED_COMBINING_CLASS_CCC34: /* sukun */
121 case HB_MODIFIED_COMBINING_CLASS_CCC35: /* superscript alef */
122 case HB_MODIFIED_COMBINING_CLASS_CCC36: /* superscript alaph */
123 return HB_UNICODE_COMBINING_CLASS_ABOVE;
124
125 case HB_MODIFIED_COMBINING_CLASS_CCC29: /* kasratan */
126 case HB_MODIFIED_COMBINING_CLASS_CCC32: /* kasra */
127 return HB_UNICODE_COMBINING_CLASS_BELOW;
128
129
130 /* Thai */
131
132 case HB_MODIFIED_COMBINING_CLASS_CCC103: /* sara u / sara uu */
133 return HB_UNICODE_COMBINING_CLASS_BELOW_RIGHT;
134
135 case HB_MODIFIED_COMBINING_CLASS_CCC107: /* mai */
136 return HB_UNICODE_COMBINING_CLASS_ABOVE_RIGHT;
137
138
139 /* Lao */
140
141 case HB_MODIFIED_COMBINING_CLASS_CCC118: /* sign u / sign uu */
142 return HB_UNICODE_COMBINING_CLASS_BELOW;
143
144 case HB_MODIFIED_COMBINING_CLASS_CCC122: /* mai */
145 return HB_UNICODE_COMBINING_CLASS_ABOVE;
146
147
148 /* Tibetan */
149
150 case HB_MODIFIED_COMBINING_CLASS_CCC129: /* sign aa */
151 return HB_UNICODE_COMBINING_CLASS_BELOW;
152
153 case HB_MODIFIED_COMBINING_CLASS_CCC130: /* sign i*/
154 return HB_UNICODE_COMBINING_CLASS_ABOVE;
155
156 case HB_MODIFIED_COMBINING_CLASS_CCC132: /* sign u */
157 return HB_UNICODE_COMBINING_CLASS_BELOW;
158
159 }
160
161 return klass;
162}
163
164void
165_hb_ot_shape_fallback_position_recategorize_marks (const hb_ot_shape_plan_t *plan HB_UNUSED,
166 hb_font_t *font HB_UNUSED,
167 hb_buffer_t *buffer)
168{
169 unsigned int count = buffer->len;
170 hb_glyph_info_t *info = buffer->info;
171 for (unsigned int i = 0; i < count; i++)
172 if (_hb_glyph_info_get_general_category (&info[i]) == HB_UNICODE_GENERAL_CATEGORY_NON_SPACING_MARK) {
173 unsigned int combining_class = _hb_glyph_info_get_modified_combining_class (&info[i]);
174 combining_class = recategorize_combining_class (info[i].codepoint, combining_class);
175 _hb_glyph_info_set_modified_combining_class (&info[i], combining_class);
176 }
177}
178
179
180static void
181zero_mark_advances (hb_buffer_t *buffer,
182 unsigned int start,
183 unsigned int end)
184{
185 hb_glyph_info_t *info = buffer->info;
186 for (unsigned int i = start; i < end; i++)
187 if (_hb_glyph_info_get_general_category (&info[i]) == HB_UNICODE_GENERAL_CATEGORY_NON_SPACING_MARK)
188 {
189 buffer->pos[i].x_advance = 0;
190 buffer->pos[i].y_advance = 0;
191 }
192}
193
194static inline void
195position_mark (const hb_ot_shape_plan_t *plan,
196 hb_font_t *font,
197 hb_buffer_t *buffer,
198 hb_glyph_extents_t &base_extents,
199 unsigned int i,
200 unsigned int combining_class)
201{
202 hb_glyph_extents_t mark_extents;
203 if (!font->get_glyph_extents (buffer->info[i].codepoint, &mark_extents))
204 return;
205
206 hb_position_t y_gap = font->y_scale / 16;
207
208 hb_glyph_position_t &pos = buffer->pos[i];
209 pos.x_offset = pos.y_offset = 0;
210
211
212 /* We don't position LEFT and RIGHT marks. */
213
214 /* X positioning */
215 switch (combining_class)
216 {
217 case HB_UNICODE_COMBINING_CLASS_DOUBLE_BELOW:
218 case HB_UNICODE_COMBINING_CLASS_DOUBLE_ABOVE:
219 if (buffer->props.direction == HB_DIRECTION_LTR) {
220 pos.x_offset += base_extents.x_bearing + base_extents.width - mark_extents.width / 2 - mark_extents.x_bearing;
221 break;
222 } else if (buffer->props.direction == HB_DIRECTION_RTL) {
223 pos.x_offset += base_extents.x_bearing - mark_extents.width / 2 - mark_extents.x_bearing;
224 break;
225 }
226 HB_FALLTHROUGH;
227
228 default:
229 case HB_UNICODE_COMBINING_CLASS_ATTACHED_BELOW:
230 case HB_UNICODE_COMBINING_CLASS_ATTACHED_ABOVE:
231 case HB_UNICODE_COMBINING_CLASS_BELOW:
232 case HB_UNICODE_COMBINING_CLASS_ABOVE:
233 /* Center align. */
234 pos.x_offset += base_extents.x_bearing + (base_extents.width - mark_extents.width) / 2 - mark_extents.x_bearing;
235 break;
236
237 case HB_UNICODE_COMBINING_CLASS_ATTACHED_BELOW_LEFT:
238 case HB_UNICODE_COMBINING_CLASS_BELOW_LEFT:
239 case HB_UNICODE_COMBINING_CLASS_ABOVE_LEFT:
240 /* Left align. */
241 pos.x_offset += base_extents.x_bearing - mark_extents.x_bearing;
242 break;
243
244 case HB_UNICODE_COMBINING_CLASS_ATTACHED_ABOVE_RIGHT:
245 case HB_UNICODE_COMBINING_CLASS_BELOW_RIGHT:
246 case HB_UNICODE_COMBINING_CLASS_ABOVE_RIGHT:
247 /* Right align. */
248 pos.x_offset += base_extents.x_bearing + base_extents.width - mark_extents.width - mark_extents.x_bearing;
249 break;
250 }
251
252 /* Y positioning */
253 switch (combining_class)
254 {
255 case HB_UNICODE_COMBINING_CLASS_DOUBLE_BELOW:
256 case HB_UNICODE_COMBINING_CLASS_BELOW_LEFT:
257 case HB_UNICODE_COMBINING_CLASS_BELOW:
258 case HB_UNICODE_COMBINING_CLASS_BELOW_RIGHT:
259 /* Add gap, fall-through. */
260 base_extents.height -= y_gap;
261 HB_FALLTHROUGH;
262
263 case HB_UNICODE_COMBINING_CLASS_ATTACHED_BELOW_LEFT:
264 case HB_UNICODE_COMBINING_CLASS_ATTACHED_BELOW:
265 pos.y_offset = base_extents.y_bearing + base_extents.height - mark_extents.y_bearing;
266 /* Never shift up "below" marks. */
267 if ((y_gap > 0) == (pos.y_offset > 0))
268 {
269 base_extents.height -= pos.y_offset;
270 pos.y_offset = 0;
271 }
272 base_extents.height += mark_extents.height;
273 break;
274
275 case HB_UNICODE_COMBINING_CLASS_DOUBLE_ABOVE:
276 case HB_UNICODE_COMBINING_CLASS_ABOVE_LEFT:
277 case HB_UNICODE_COMBINING_CLASS_ABOVE:
278 case HB_UNICODE_COMBINING_CLASS_ABOVE_RIGHT:
279 /* Add gap, fall-through. */
280 base_extents.y_bearing += y_gap;
281 base_extents.height -= y_gap;
282 HB_FALLTHROUGH;
283
284 case HB_UNICODE_COMBINING_CLASS_ATTACHED_ABOVE:
285 case HB_UNICODE_COMBINING_CLASS_ATTACHED_ABOVE_RIGHT:
286 pos.y_offset = base_extents.y_bearing - (mark_extents.y_bearing + mark_extents.height);
287 /* Don't shift down "above" marks too much. */
288 if ((y_gap > 0) != (pos.y_offset > 0))
289 {
290 unsigned int correction = -pos.y_offset / 2;
291 base_extents.y_bearing += correction;
292 base_extents.height -= correction;
293 pos.y_offset += correction;
294 }
295 base_extents.y_bearing -= mark_extents.height;
296 base_extents.height += mark_extents.height;
297 break;
298 }
299}
300
301static inline void
302position_around_base (const hb_ot_shape_plan_t *plan,
303 hb_font_t *font,
304 hb_buffer_t *buffer,
305 unsigned int base,
306 unsigned int end)
307{
308 hb_direction_t horiz_dir = HB_DIRECTION_INVALID;
309
310 buffer->unsafe_to_break (base, end);
311
312 hb_glyph_extents_t base_extents;
313 if (!font->get_glyph_extents (buffer->info[base].codepoint,
314 &base_extents))
315 {
316 /* If extents don't work, zero marks and go home. */
317 zero_mark_advances (buffer, base + 1, end);
318 return;
319 }
320 base_extents.x_bearing += buffer->pos[base].x_offset;
321 base_extents.y_bearing += buffer->pos[base].y_offset;
322
323 unsigned int lig_id = _hb_glyph_info_get_lig_id (&buffer->info[base]);
324 /* Use integer for num_lig_components such that it doesn't convert to unsigned
325 * when we divide or multiply by it. */
326 int num_lig_components = _hb_glyph_info_get_lig_num_comps (&buffer->info[base]);
327
328 hb_position_t x_offset = 0, y_offset = 0;
329 if (HB_DIRECTION_IS_FORWARD (buffer->props.direction)) {
330 x_offset -= buffer->pos[base].x_advance;
331 y_offset -= buffer->pos[base].y_advance;
332 }
333
334 hb_glyph_extents_t component_extents = base_extents;
335 int last_lig_component = -1;
336 unsigned int last_combining_class = 255;
337 hb_glyph_extents_t cluster_extents = base_extents; /* Initialization is just to shut gcc up. */
338 hb_glyph_info_t *info = buffer->info;
339 for (unsigned int i = base + 1; i < end; i++)
340 if (_hb_glyph_info_get_modified_combining_class (&info[i]))
341 {
342 if (num_lig_components > 1) {
343 unsigned int this_lig_id = _hb_glyph_info_get_lig_id (&info[i]);
344 int this_lig_component = _hb_glyph_info_get_lig_comp (&info[i]) - 1;
345 /* Conditions for attaching to the last component. */
346 if (!lig_id || lig_id != this_lig_id || this_lig_component >= num_lig_components)
347 this_lig_component = num_lig_components - 1;
348 if (last_lig_component != this_lig_component)
349 {
350 last_lig_component = this_lig_component;
351 last_combining_class = 255;
352 component_extents = base_extents;
353 if (unlikely (horiz_dir == HB_DIRECTION_INVALID)) {
354 if (HB_DIRECTION_IS_HORIZONTAL (plan->props.direction))
355 horiz_dir = plan->props.direction;
356 else
357 horiz_dir = hb_script_get_horizontal_direction (plan->props.script);
358 }
359 if (horiz_dir == HB_DIRECTION_LTR)
360 component_extents.x_bearing += (this_lig_component * component_extents.width) / num_lig_components;
361 else
362 component_extents.x_bearing += ((num_lig_components - 1 - this_lig_component) * component_extents.width) / num_lig_components;
363 component_extents.width /= num_lig_components;
364 }
365 }
366
367 unsigned int this_combining_class = _hb_glyph_info_get_modified_combining_class (&info[i]);
368 if (last_combining_class != this_combining_class)
369 {
370 last_combining_class = this_combining_class;
371 cluster_extents = component_extents;
372 }
373
374 position_mark (plan, font, buffer, cluster_extents, i, this_combining_class);
375
376 buffer->pos[i].x_advance = 0;
377 buffer->pos[i].y_advance = 0;
378 buffer->pos[i].x_offset += x_offset;
379 buffer->pos[i].y_offset += y_offset;
380
381 } else {
382 if (HB_DIRECTION_IS_FORWARD (buffer->props.direction)) {
383 x_offset -= buffer->pos[i].x_advance;
384 y_offset -= buffer->pos[i].y_advance;
385 } else {
386 x_offset += buffer->pos[i].x_advance;
387 y_offset += buffer->pos[i].y_advance;
388 }
389 }
390}
391
392static inline void
393position_cluster (const hb_ot_shape_plan_t *plan,
394 hb_font_t *font,
395 hb_buffer_t *buffer,
396 unsigned int start,
397 unsigned int end)
398{
399 if (end - start < 2)
400 return;
401
402 /* Find the base glyph */
403 hb_glyph_info_t *info = buffer->info;
404 for (unsigned int i = start; i < end; i++)
405 if (!HB_UNICODE_GENERAL_CATEGORY_IS_MARK (_hb_glyph_info_get_general_category (&info[i])))
406 {
407 /* Find mark glyphs */
408 unsigned int j;
409 for (j = i + 1; j < end; j++)
410 if (!HB_UNICODE_GENERAL_CATEGORY_IS_MARK (_hb_glyph_info_get_general_category (&info[j])))
411 break;
412
413 position_around_base (plan, font, buffer, i, j);
414
415 i = j - 1;
416 }
417}
418
419void
420_hb_ot_shape_fallback_position (const hb_ot_shape_plan_t *plan,
421 hb_font_t *font,
422 hb_buffer_t *buffer)
423{
424 _hb_buffer_assert_gsubgpos_vars (buffer);
425
426 unsigned int start = 0;
427 unsigned int count = buffer->len;
428 hb_glyph_info_t *info = buffer->info;
429 for (unsigned int i = 1; i < count; i++)
430 if (likely (!HB_UNICODE_GENERAL_CATEGORY_IS_MARK (_hb_glyph_info_get_general_category (&info[i])))) {
431 position_cluster (plan, font, buffer, start, i);
432 start = i;
433 }
434 position_cluster (plan, font, buffer, start, count);
435}
436
437
438/* Performs old-style TrueType kerning. */
439void
440_hb_ot_shape_fallback_kern (const hb_ot_shape_plan_t *plan,
441 hb_font_t *font,
442 hb_buffer_t *buffer)
443{
444 if (!plan->has_kern) return;
445
446 OT::hb_ot_apply_context_t c (1, font, buffer);
447 c.set_lookup_mask (plan->kern_mask);
448 c.set_lookup_props (OT::LookupFlag::IgnoreMarks);
449 OT::hb_ot_apply_context_t::skipping_iterator_t &skippy_iter = c.iter_input;
450 skippy_iter.init (&c);
451
452 unsigned int count = buffer->len;
453 hb_glyph_info_t *info = buffer->info;
454 hb_glyph_position_t *pos = buffer->pos;
455 for (unsigned int idx = 0; idx < count;)
456 {
457 skippy_iter.reset (idx, 1);
458 if (!skippy_iter.next ())
459 {
460 idx++;
461 continue;
462 }
463
464 hb_position_t x_kern, y_kern;
465 font->get_glyph_kerning_for_direction (info[idx].codepoint,
466 info[skippy_iter.idx].codepoint,
467 buffer->props.direction,
468 &x_kern, &y_kern);
469
470 if (x_kern)
471 {
472 hb_position_t kern1 = x_kern >> 1;
473 hb_position_t kern2 = x_kern - kern1;
474 pos[idx].x_advance += kern1;
475 pos[skippy_iter.idx].x_advance += kern2;
476 pos[skippy_iter.idx].x_offset += kern2;
477 buffer->unsafe_to_break (idx, skippy_iter.idx + 1);
478 }
479
480 if (y_kern)
481 {
482 hb_position_t kern1 = y_kern >> 1;
483 hb_position_t kern2 = y_kern - kern1;
484 pos[idx].y_advance += kern1;
485 pos[skippy_iter.idx].y_advance += kern2;
486 pos[skippy_iter.idx].y_offset += kern2;
487 buffer->unsafe_to_break (idx, skippy_iter.idx + 1);
488 }
489
490 idx = skippy_iter.idx;
491 }
492}
493
494
495/* Adjusts width of various spaces. */
496void
497_hb_ot_shape_fallback_spaces (const hb_ot_shape_plan_t *plan,
498 hb_font_t *font,
499 hb_buffer_t *buffer)
500{
501 if (!HB_DIRECTION_IS_HORIZONTAL (buffer->props.direction))
502 return;
503
504 hb_glyph_info_t *info = buffer->info;
505 hb_glyph_position_t *pos = buffer->pos;
506 unsigned int count = buffer->len;
507 for (unsigned int i = 0; i < count; i++)
508 if (_hb_glyph_info_is_unicode_space (&info[i]) && !_hb_glyph_info_ligated (&info[i]))
509 {
510 hb_unicode_funcs_t::space_t space_type = _hb_glyph_info_get_unicode_space_fallback_type (&info[i]);
511 hb_codepoint_t glyph;
512 typedef hb_unicode_funcs_t t;
513 switch (space_type)
514 {
515 case t::NOT_SPACE: /* Shouldn't happen. */
516 case t::SPACE:
517 break;
518
519 case t::SPACE_EM:
520 case t::SPACE_EM_2:
521 case t::SPACE_EM_3:
522 case t::SPACE_EM_4:
523 case t::SPACE_EM_5:
524 case t::SPACE_EM_6:
525 case t::SPACE_EM_16:
526 pos[i].x_advance = (font->x_scale + ((int) space_type)/2) / (int) space_type;
527 break;
528
529 case t::SPACE_4_EM_18:
530 pos[i].x_advance = (int64_t) font->x_scale * 4 / 18;
531 break;
532
533 case t::SPACE_FIGURE:
534 for (char u = '0'; u <= '9'; u++)
535 if (font->get_nominal_glyph (u, &glyph))
536 {
537 pos[i].x_advance = font->get_glyph_h_advance (glyph);
538 break;
539 }
540 break;
541
542 case t::SPACE_PUNCTUATION:
543 if (font->get_nominal_glyph ('.', &glyph))
544 pos[i].x_advance = font->get_glyph_h_advance (glyph);
545 else if (font->get_nominal_glyph (',', &glyph))
546 pos[i].x_advance = font->get_glyph_h_advance (glyph);
547 break;
548
549 case t::SPACE_NARROW:
550 /* Half-space?
551 * Unicode doc https://unicode.org/charts/PDF/U2000.pdf says ~1/4 or 1/5 of EM.
552 * However, in my testing, many fonts have their regular space being about that
553 * size. To me, a percentage of the space width makes more sense. Half is as
554 * good as any. */
555 pos[i].x_advance /= 2;
556 break;
557 }
558 }
559}
560