1 /*
2 * Copyright © 2015 Mozilla Foundation.
3 * Copyright © 2015 Google, Inc.
4 *
5 * This is part of HarfBuzz, a text shaping library.
6 *
7 * Permission is hereby granted, without written agreement and without
8 * license or royalty fees, to use, copy, modify, and distribute this
9 * software and its documentation for any purpose, provided that the
10 * above copyright notice and the following two paragraphs appear in
11 * all copies of this software.
12 *
13 * IN NO EVENT SHALL THE COPYRIGHT HOLDER BE LIABLE TO ANY PARTY FOR
14 * DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES
15 * ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS DOCUMENTATION, EVEN
16 * IF THE COPYRIGHT HOLDER HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH
17 * DAMAGE.
18 *
19 * THE COPYRIGHT HOLDER SPECIFICALLY DISCLAIMS ANY WARRANTIES, INCLUDING,
20 * BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
21 * FITNESS FOR A PARTICULAR PURPOSE. THE SOFTWARE PROVIDED HEREUNDER IS
22 * ON AN "AS IS" BASIS, AND THE COPYRIGHT HOLDER HAS NO OBLIGATION TO
23 * PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
24 *
25 * Mozilla Author(s): Jonathan Kew
26 * Google Author(s): Behdad Esfahbod
27 */
28
29 #include "hb-ot-shape-complex-use.hh"
30 #include "hb-ot-shape-complex-arabic.hh"
31 #include "hb-ot-shape-complex-vowel-constraints.hh"
32
33 /* buffer var allocations */
34 #define use_category() complex_var_u8_0()
35
36
37 /*
38 * Universal Shaping Engine.
39 * https://docs.microsoft.com/en-us/typography/script-development/use
40 */
41
42 static const hb_tag_t
43 basic_features[] =
44 {
45 /*
46 * Basic features.
47 * These features are applied all at once, before reordering.
48 */
49 HB_TAG('r','k','r','f'),
50 HB_TAG('a','b','v','f'),
51 HB_TAG('b','l','w','f'),
52 HB_TAG('h','a','l','f'),
53 HB_TAG('p','s','t','f'),
54 HB_TAG('v','a','t','u'),
55 HB_TAG('c','j','c','t'),
56 };
57 static const hb_tag_t
58 arabic_features[] =
59 {
60 HB_TAG('i','s','o','l'),
61 HB_TAG('i','n','i','t'),
62 HB_TAG('m','e','d','i'),
63 HB_TAG('f','i','n','a'),
64 /* The spec doesn't specify these but we apply anyway, since our Arabic shaper
65 * does. These are only used in Syriac spec. */
66 HB_TAG('m','e','d','2'),
67 HB_TAG('f','i','n','2'),
68 HB_TAG('f','i','n','3'),
69 };
70 /* Same order as arabic_features. Don't need Syriac stuff.*/
71 enum joining_form_t {
72 ISOL,
73 INIT,
74 MEDI,
75 FINA,
76 _NONE
77 };
78 static const hb_tag_t
79 other_features[] =
80 {
81 /*
82 * Other features.
83 * These features are applied all at once, after reordering and
84 * clearing syllables.
85 */
86 HB_TAG('a','b','v','s'),
87 HB_TAG('b','l','w','s'),
88 HB_TAG('h','a','l','n'),
89 HB_TAG('p','r','e','s'),
90 HB_TAG('p','s','t','s'),
91 };
92 static const hb_tag_t
93 positioning_features[] =
94 {
95 /*
96 * Positioning features.
97 * We don't care about the types.
98 */
99 HB_TAG('d','i','s','t'),
100 HB_TAG('a','b','v','m'),
101 HB_TAG('b','l','w','m'),
102 };
103
104 static void
105 setup_syllables (const hb_ot_shape_plan_t *plan,
106 hb_font_t *font,
107 hb_buffer_t *buffer);
108 static void
109 clear_substitution_flags (const hb_ot_shape_plan_t *plan,
110 hb_font_t *font,
111 hb_buffer_t *buffer);
112 static void
113 record_rphf (const hb_ot_shape_plan_t *plan,
114 hb_font_t *font,
115 hb_buffer_t *buffer);
116 static void
117 record_pref (const hb_ot_shape_plan_t *plan,
118 hb_font_t *font,
119 hb_buffer_t *buffer);
120 static void
121 reorder (const hb_ot_shape_plan_t *plan,
122 hb_font_t *font,
123 hb_buffer_t *buffer);
124 static void
125 clear_syllables (const hb_ot_shape_plan_t *plan,
126 hb_font_t *font,
127 hb_buffer_t *buffer);
128
129 static void
collect_features_use(hb_ot_shape_planner_t * plan)130 collect_features_use (hb_ot_shape_planner_t *plan)
131 {
132 hb_ot_map_builder_t *map = &plan->map;
133
134 /* Do this before any lookups have been applied. */
135 map->add_gsub_pause (setup_syllables);
136
137 /* "Default glyph pre-processing group" */
138 map->enable_feature (HB_TAG('l','o','c','l'));
139 map->enable_feature (HB_TAG('c','c','m','p'));
140 map->enable_feature (HB_TAG('n','u','k','t'));
141 map->enable_feature (HB_TAG('a','k','h','n'), F_MANUAL_ZWJ);
142
143 /* "Reordering group" */
144 map->add_gsub_pause (clear_substitution_flags);
145 map->add_feature (HB_TAG('r','p','h','f'), F_MANUAL_ZWJ);
146 map->add_gsub_pause (record_rphf);
147 map->add_gsub_pause (clear_substitution_flags);
148 map->enable_feature (HB_TAG('p','r','e','f'), F_MANUAL_ZWJ);
149 map->add_gsub_pause (record_pref);
150
151 /* "Orthographic unit shaping group" */
152 for (unsigned int i = 0; i < ARRAY_LENGTH (basic_features); i++)
153 map->enable_feature (basic_features[i], F_MANUAL_ZWJ);
154
155 map->add_gsub_pause (reorder);
156 map->add_gsub_pause (clear_syllables);
157
158 /* "Topographical features" */
159 for (unsigned int i = 0; i < ARRAY_LENGTH (arabic_features); i++)
160 map->add_feature (arabic_features[i]);
161 map->add_gsub_pause (nullptr);
162
163 /* "Standard typographic presentation" */
164 for (unsigned int i = 0; i < ARRAY_LENGTH (other_features); i++)
165 map->enable_feature (other_features[i], F_MANUAL_ZWJ);
166
167 /* "Positional feature application" */
168 for (unsigned int i = 0; i < ARRAY_LENGTH (positioning_features); i++)
169 map->enable_feature (positioning_features[i]);
170 }
171
172 struct use_shape_plan_t
173 {
174 hb_mask_t rphf_mask;
175
176 arabic_shape_plan_t *arabic_plan;
177 };
178
179 static bool
has_arabic_joining(hb_script_t script)180 has_arabic_joining (hb_script_t script)
181 {
182 /* List of scripts that have data in arabic-table. */
183 switch ((int) script)
184 {
185 /* Unicode-1.1 additions */
186 case HB_SCRIPT_ARABIC:
187
188 /* Unicode-3.0 additions */
189 case HB_SCRIPT_MONGOLIAN:
190 case HB_SCRIPT_SYRIAC:
191
192 /* Unicode-5.0 additions */
193 case HB_SCRIPT_NKO:
194 case HB_SCRIPT_PHAGS_PA:
195
196 /* Unicode-6.0 additions */
197 case HB_SCRIPT_MANDAIC:
198
199 /* Unicode-7.0 additions */
200 case HB_SCRIPT_MANICHAEAN:
201 case HB_SCRIPT_PSALTER_PAHLAVI:
202
203 /* Unicode-9.0 additions */
204 case HB_SCRIPT_ADLAM:
205
206 return true;
207
208 default:
209 return false;
210 }
211 }
212
213 static void *
data_create_use(const hb_ot_shape_plan_t * plan)214 data_create_use (const hb_ot_shape_plan_t *plan)
215 {
216 use_shape_plan_t *use_plan = (use_shape_plan_t *) calloc (1, sizeof (use_shape_plan_t));
217 if (unlikely (!use_plan))
218 return nullptr;
219
220 use_plan->rphf_mask = plan->map.get_1_mask (HB_TAG('r','p','h','f'));
221
222 if (has_arabic_joining (plan->props.script))
223 {
224 use_plan->arabic_plan = (arabic_shape_plan_t *) data_create_arabic (plan);
225 if (unlikely (!use_plan->arabic_plan))
226 {
227 free (use_plan);
228 return nullptr;
229 }
230 }
231
232 return use_plan;
233 }
234
235 static void
data_destroy_use(void * data)236 data_destroy_use (void *data)
237 {
238 use_shape_plan_t *use_plan = (use_shape_plan_t *) data;
239
240 if (use_plan->arabic_plan)
241 data_destroy_arabic (use_plan->arabic_plan);
242
243 free (data);
244 }
245
246 enum syllable_type_t {
247 independent_cluster,
248 virama_terminated_cluster,
249 standard_cluster,
250 number_joiner_terminated_cluster,
251 numeral_cluster,
252 symbol_cluster,
253 broken_cluster,
254 non_cluster,
255 };
256
257 #include "hb-ot-shape-complex-use-machine.hh"
258
259
260 static void
setup_masks_use(const hb_ot_shape_plan_t * plan,hb_buffer_t * buffer,hb_font_t * font HB_UNUSED)261 setup_masks_use (const hb_ot_shape_plan_t *plan,
262 hb_buffer_t *buffer,
263 hb_font_t *font HB_UNUSED)
264 {
265 const use_shape_plan_t *use_plan = (const use_shape_plan_t *) plan->data;
266
267 /* Do this before allocating use_category(). */
268 if (use_plan->arabic_plan)
269 {
270 setup_masks_arabic_plan (use_plan->arabic_plan, buffer, plan->props.script);
271 }
272
273 HB_BUFFER_ALLOCATE_VAR (buffer, use_category);
274
275 /* We cannot setup masks here. We save information about characters
276 * and setup masks later on in a pause-callback. */
277
278 unsigned int count = buffer->len;
279 hb_glyph_info_t *info = buffer->info;
280 for (unsigned int i = 0; i < count; i++)
281 info[i].use_category() = hb_use_get_category (info[i].codepoint);
282 }
283
284 static void
setup_rphf_mask(const hb_ot_shape_plan_t * plan,hb_buffer_t * buffer)285 setup_rphf_mask (const hb_ot_shape_plan_t *plan,
286 hb_buffer_t *buffer)
287 {
288 const use_shape_plan_t *use_plan = (const use_shape_plan_t *) plan->data;
289
290 hb_mask_t mask = use_plan->rphf_mask;
291 if (!mask) return;
292
293 hb_glyph_info_t *info = buffer->info;
294
295 foreach_syllable (buffer, start, end)
296 {
297 unsigned int limit = info[start].use_category() == USE_R ? 1 : MIN (3u, end - start);
298 for (unsigned int i = start; i < start + limit; i++)
299 info[i].mask |= mask;
300 }
301 }
302
303 static void
setup_topographical_masks(const hb_ot_shape_plan_t * plan,hb_buffer_t * buffer)304 setup_topographical_masks (const hb_ot_shape_plan_t *plan,
305 hb_buffer_t *buffer)
306 {
307 const use_shape_plan_t *use_plan = (const use_shape_plan_t *) plan->data;
308 if (use_plan->arabic_plan)
309 return;
310
311 static_assert ((INIT < 4 && ISOL < 4 && MEDI < 4 && FINA < 4), "");
312 hb_mask_t masks[4], all_masks = 0;
313 for (unsigned int i = 0; i < 4; i++)
314 {
315 masks[i] = plan->map.get_1_mask (arabic_features[i]);
316 if (masks[i] == plan->map.get_global_mask ())
317 masks[i] = 0;
318 all_masks |= masks[i];
319 }
320 if (!all_masks)
321 return;
322 hb_mask_t other_masks = ~all_masks;
323
324 unsigned int last_start = 0;
325 joining_form_t last_form = _NONE;
326 hb_glyph_info_t *info = buffer->info;
327 foreach_syllable (buffer, start, end)
328 {
329 syllable_type_t syllable_type = (syllable_type_t) (info[start].syllable() & 0x0F);
330 switch (syllable_type)
331 {
332 case independent_cluster:
333 case symbol_cluster:
334 case non_cluster:
335 /* These don't join. Nothing to do. */
336 last_form = _NONE;
337 break;
338
339 case virama_terminated_cluster:
340 case standard_cluster:
341 case number_joiner_terminated_cluster:
342 case numeral_cluster:
343 case broken_cluster:
344
345 bool join = last_form == FINA || last_form == ISOL;
346
347 if (join)
348 {
349 /* Fixup previous syllable's form. */
350 last_form = last_form == FINA ? MEDI : INIT;
351 for (unsigned int i = last_start; i < start; i++)
352 info[i].mask = (info[i].mask & other_masks) | masks[last_form];
353 }
354
355 /* Form for this syllable. */
356 last_form = join ? FINA : ISOL;
357 for (unsigned int i = start; i < end; i++)
358 info[i].mask = (info[i].mask & other_masks) | masks[last_form];
359
360 break;
361 }
362
363 last_start = start;
364 }
365 }
366
367 static void
setup_syllables(const hb_ot_shape_plan_t * plan,hb_font_t * font HB_UNUSED,hb_buffer_t * buffer)368 setup_syllables (const hb_ot_shape_plan_t *plan,
369 hb_font_t *font HB_UNUSED,
370 hb_buffer_t *buffer)
371 {
372 find_syllables (buffer);
373 foreach_syllable (buffer, start, end)
374 buffer->unsafe_to_break (start, end);
375 setup_rphf_mask (plan, buffer);
376 setup_topographical_masks (plan, buffer);
377 }
378
379 static void
clear_substitution_flags(const hb_ot_shape_plan_t * plan HB_UNUSED,hb_font_t * font HB_UNUSED,hb_buffer_t * buffer)380 clear_substitution_flags (const hb_ot_shape_plan_t *plan HB_UNUSED,
381 hb_font_t *font HB_UNUSED,
382 hb_buffer_t *buffer)
383 {
384 hb_glyph_info_t *info = buffer->info;
385 unsigned int count = buffer->len;
386 for (unsigned int i = 0; i < count; i++)
387 _hb_glyph_info_clear_substituted (&info[i]);
388 }
389
390 static void
record_rphf(const hb_ot_shape_plan_t * plan,hb_font_t * font HB_UNUSED,hb_buffer_t * buffer)391 record_rphf (const hb_ot_shape_plan_t *plan,
392 hb_font_t *font HB_UNUSED,
393 hb_buffer_t *buffer)
394 {
395 const use_shape_plan_t *use_plan = (const use_shape_plan_t *) plan->data;
396
397 hb_mask_t mask = use_plan->rphf_mask;
398 if (!mask) return;
399 hb_glyph_info_t *info = buffer->info;
400
401 foreach_syllable (buffer, start, end)
402 {
403 /* Mark a substituted repha as USE_R. */
404 for (unsigned int i = start; i < end && (info[i].mask & mask); i++)
405 if (_hb_glyph_info_substituted (&info[i]))
406 {
407 info[i].use_category() = USE_R;
408 break;
409 }
410 }
411 }
412
413 static void
record_pref(const hb_ot_shape_plan_t * plan HB_UNUSED,hb_font_t * font HB_UNUSED,hb_buffer_t * buffer)414 record_pref (const hb_ot_shape_plan_t *plan HB_UNUSED,
415 hb_font_t *font HB_UNUSED,
416 hb_buffer_t *buffer)
417 {
418 hb_glyph_info_t *info = buffer->info;
419
420 foreach_syllable (buffer, start, end)
421 {
422 /* Mark a substituted pref as VPre, as they behave the same way. */
423 for (unsigned int i = start; i < end; i++)
424 if (_hb_glyph_info_substituted (&info[i]))
425 {
426 info[i].use_category() = USE_VPre;
427 break;
428 }
429 }
430 }
431
432 static inline bool
is_halant(const hb_glyph_info_t & info)433 is_halant (const hb_glyph_info_t &info)
434 {
435 return (info.use_category() == USE_H || info.use_category() == USE_HVM) &&
436 !_hb_glyph_info_ligated (&info);
437 }
438
439 static void
reorder_syllable(hb_buffer_t * buffer,unsigned int start,unsigned int end)440 reorder_syllable (hb_buffer_t *buffer, unsigned int start, unsigned int end)
441 {
442 syllable_type_t syllable_type = (syllable_type_t) (buffer->info[start].syllable() & 0x0F);
443 /* Only a few syllable types need reordering. */
444 if (unlikely (!(FLAG_UNSAFE (syllable_type) &
445 (FLAG (virama_terminated_cluster) |
446 FLAG (standard_cluster) |
447 FLAG (broken_cluster) |
448 0))))
449 return;
450
451 hb_glyph_info_t *info = buffer->info;
452
453 #define POST_BASE_FLAGS64 (FLAG64 (USE_FM) | \
454 FLAG64 (USE_FAbv) | \
455 FLAG64 (USE_FBlw) | \
456 FLAG64 (USE_FPst) | \
457 FLAG64 (USE_MAbv) | \
458 FLAG64 (USE_MBlw) | \
459 FLAG64 (USE_MPst) | \
460 FLAG64 (USE_MPre) | \
461 FLAG64 (USE_VAbv) | \
462 FLAG64 (USE_VBlw) | \
463 FLAG64 (USE_VPst) | \
464 FLAG64 (USE_VPre) | \
465 FLAG64 (USE_VMAbv) | \
466 FLAG64 (USE_VMBlw) | \
467 FLAG64 (USE_VMPst) | \
468 FLAG64 (USE_VMPre))
469
470 /* Move things forward. */
471 if (info[start].use_category() == USE_R && end - start > 1)
472 {
473 /* Got a repha. Reorder it towards the end, but before the first post-base
474 * glyph. */
475 for (unsigned int i = start + 1; i < end; i++)
476 {
477 bool is_post_base_glyph = (FLAG64_UNSAFE (info[i].use_category()) & POST_BASE_FLAGS64) ||
478 is_halant (info[i]);
479 if (is_post_base_glyph || i == end - 1)
480 {
481 /* If we hit a post-base glyph, move before it; otherwise move to the
482 * end. Shift things in between backward. */
483
484 if (is_post_base_glyph)
485 i--;
486
487 buffer->merge_clusters (start, i + 1);
488 hb_glyph_info_t t = info[start];
489 memmove (&info[start], &info[start + 1], (i - start) * sizeof (info[0]));
490 info[i] = t;
491
492 break;
493 }
494 }
495 }
496
497 /* Move things back. */
498 unsigned int j = start;
499 for (unsigned int i = start; i < end; i++)
500 {
501 uint32_t flag = FLAG_UNSAFE (info[i].use_category());
502 if (is_halant (info[i]))
503 {
504 /* If we hit a halant, move after it; otherwise move to the beginning, and
505 * shift things in between forward. */
506 j = i + 1;
507 }
508 else if (((flag) & (FLAG (USE_VPre) | FLAG (USE_VMPre))) &&
509 /* Only move the first component of a MultipleSubst. */
510 0 == _hb_glyph_info_get_lig_comp (&info[i]) &&
511 j < i)
512 {
513 buffer->merge_clusters (j, i + 1);
514 hb_glyph_info_t t = info[i];
515 memmove (&info[j + 1], &info[j], (i - j) * sizeof (info[0]));
516 info[j] = t;
517 }
518 }
519 }
520
521 static inline void
insert_dotted_circles(const hb_ot_shape_plan_t * plan HB_UNUSED,hb_font_t * font,hb_buffer_t * buffer)522 insert_dotted_circles (const hb_ot_shape_plan_t *plan HB_UNUSED,
523 hb_font_t *font,
524 hb_buffer_t *buffer)
525 {
526 /* Note: This loop is extra overhead, but should not be measurable. */
527 bool has_broken_syllables = false;
528 unsigned int count = buffer->len;
529 hb_glyph_info_t *info = buffer->info;
530 for (unsigned int i = 0; i < count; i++)
531 if ((info[i].syllable() & 0x0F) == broken_cluster)
532 {
533 has_broken_syllables = true;
534 break;
535 }
536 if (likely (!has_broken_syllables))
537 return;
538
539 hb_glyph_info_t dottedcircle = {0};
540 if (!font->get_nominal_glyph (0x25CCu, &dottedcircle.codepoint))
541 return;
542 dottedcircle.use_category() = hb_use_get_category (0x25CC);
543
544 buffer->clear_output ();
545
546 buffer->idx = 0;
547 unsigned int last_syllable = 0;
548 while (buffer->idx < buffer->len && buffer->successful)
549 {
550 unsigned int syllable = buffer->cur().syllable();
551 syllable_type_t syllable_type = (syllable_type_t) (syllable & 0x0F);
552 if (unlikely (last_syllable != syllable && syllable_type == broken_cluster))
553 {
554 last_syllable = syllable;
555
556 hb_glyph_info_t ginfo = dottedcircle;
557 ginfo.cluster = buffer->cur().cluster;
558 ginfo.mask = buffer->cur().mask;
559 ginfo.syllable() = buffer->cur().syllable();
560 /* TODO Set glyph_props? */
561
562 /* Insert dottedcircle after possible Repha. */
563 while (buffer->idx < buffer->len && buffer->successful &&
564 last_syllable == buffer->cur().syllable() &&
565 buffer->cur().use_category() == USE_R)
566 buffer->next_glyph ();
567
568 buffer->output_info (ginfo);
569 }
570 else
571 buffer->next_glyph ();
572 }
573 buffer->swap_buffers ();
574 }
575
576 static void
reorder(const hb_ot_shape_plan_t * plan,hb_font_t * font,hb_buffer_t * buffer)577 reorder (const hb_ot_shape_plan_t *plan,
578 hb_font_t *font,
579 hb_buffer_t *buffer)
580 {
581 insert_dotted_circles (plan, font, buffer);
582
583 foreach_syllable (buffer, start, end)
584 reorder_syllable (buffer, start, end);
585
586 HB_BUFFER_DEALLOCATE_VAR (buffer, use_category);
587 }
588
589 static void
clear_syllables(const hb_ot_shape_plan_t * plan HB_UNUSED,hb_font_t * font HB_UNUSED,hb_buffer_t * buffer)590 clear_syllables (const hb_ot_shape_plan_t *plan HB_UNUSED,
591 hb_font_t *font HB_UNUSED,
592 hb_buffer_t *buffer)
593 {
594 hb_glyph_info_t *info = buffer->info;
595 unsigned int count = buffer->len;
596 for (unsigned int i = 0; i < count; i++)
597 info[i].syllable() = 0;
598 }
599
600
601 static void
preprocess_text_use(const hb_ot_shape_plan_t * plan,hb_buffer_t * buffer,hb_font_t * font)602 preprocess_text_use (const hb_ot_shape_plan_t *plan,
603 hb_buffer_t *buffer,
604 hb_font_t *font)
605 {
606 _hb_preprocess_text_vowel_constraints (plan, buffer, font);
607 }
608
609 static bool
compose_use(const hb_ot_shape_normalize_context_t * c,hb_codepoint_t a,hb_codepoint_t b,hb_codepoint_t * ab)610 compose_use (const hb_ot_shape_normalize_context_t *c,
611 hb_codepoint_t a,
612 hb_codepoint_t b,
613 hb_codepoint_t *ab)
614 {
615 /* Avoid recomposing split matras. */
616 if (HB_UNICODE_GENERAL_CATEGORY_IS_MARK (c->unicode->general_category (a)))
617 return false;
618
619 return (bool)c->unicode->compose (a, b, ab);
620 }
621
622
623 const hb_ot_complex_shaper_t _hb_ot_complex_shaper_use =
624 {
625 collect_features_use,
626 nullptr, /* override_features */
627 data_create_use,
628 data_destroy_use,
629 preprocess_text_use,
630 nullptr, /* postprocess_glyphs */
631 HB_OT_SHAPE_NORMALIZATION_MODE_COMPOSED_DIACRITICS_NO_SHORT_CIRCUIT,
632 nullptr, /* decompose */
633 compose_use,
634 setup_masks_use,
635 HB_TAG_NONE, /* gpos_tag */
636 nullptr, /* reorder_marks */
637 HB_OT_SHAPE_ZERO_WIDTH_MARKS_BY_GDEF_EARLY,
638 false, /* fallback_position */
639 };
640