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.hh"
30
31#ifndef HB_NO_OT_SHAPE
32
33#include "hb-ot-shaper-use-machine.hh"
34#include "hb-ot-shaper-use-table.hh"
35#include "hb-ot-shaper-arabic.hh"
36#include "hb-ot-shaper-arabic-joining-list.hh"
37#include "hb-ot-shaper-vowel-constraints.hh"
38
39
40/*
41 * Universal Shaping Engine.
42 * https://docs.microsoft.com/en-us/typography/script-development/use
43 */
44
45static const hb_tag_t
46use_basic_features[] =
47{
48 /*
49 * Basic features.
50 * These features are applied all at once, before reordering, constrained
51 * to the syllable.
52 */
53 HB_TAG('r','k','r','f'),
54 HB_TAG('a','b','v','f'),
55 HB_TAG('b','l','w','f'),
56 HB_TAG('h','a','l','f'),
57 HB_TAG('p','s','t','f'),
58 HB_TAG('v','a','t','u'),
59 HB_TAG('c','j','c','t'),
60};
61static const hb_tag_t
62use_topographical_features[] =
63{
64 HB_TAG('i','s','o','l'),
65 HB_TAG('i','n','i','t'),
66 HB_TAG('m','e','d','i'),
67 HB_TAG('f','i','n','a'),
68};
69/* Same order as use_topographical_features. */
70enum joining_form_t {
71 JOINING_FORM_ISOL,
72 JOINING_FORM_INIT,
73 JOINING_FORM_MEDI,
74 JOINING_FORM_FINA,
75 _JOINING_FORM_NONE
76};
77static const hb_tag_t
78use_other_features[] =
79{
80 /*
81 * Other features.
82 * These features are applied all at once, after reordering and
83 * clearing syllables.
84 */
85 HB_TAG('a','b','v','s'),
86 HB_TAG('b','l','w','s'),
87 HB_TAG('h','a','l','n'),
88 HB_TAG('p','r','e','s'),
89 HB_TAG('p','s','t','s'),
90};
91
92static bool
93setup_syllables_use (const hb_ot_shape_plan_t *plan,
94 hb_font_t *font,
95 hb_buffer_t *buffer);
96static bool
97record_rphf_use (const hb_ot_shape_plan_t *plan,
98 hb_font_t *font,
99 hb_buffer_t *buffer);
100static bool
101record_pref_use (const hb_ot_shape_plan_t *plan,
102 hb_font_t *font,
103 hb_buffer_t *buffer);
104static bool
105reorder_use (const hb_ot_shape_plan_t *plan,
106 hb_font_t *font,
107 hb_buffer_t *buffer);
108
109static void
110collect_features_use (hb_ot_shape_planner_t *plan)
111{
112 hb_ot_map_builder_t *map = &plan->map;
113
114 /* Do this before any lookups have been applied. */
115 map->add_gsub_pause (setup_syllables_use);
116
117 /* "Default glyph pre-processing group" */
118 map->enable_feature (HB_TAG('l','o','c','l'), F_PER_SYLLABLE);
119 map->enable_feature (HB_TAG('c','c','m','p'), F_PER_SYLLABLE);
120 map->enable_feature (HB_TAG('n','u','k','t'), F_PER_SYLLABLE);
121 map->enable_feature (HB_TAG('a','k','h','n'), F_MANUAL_ZWJ | F_PER_SYLLABLE);
122
123 /* "Reordering group" */
124 map->add_gsub_pause (_hb_clear_substitution_flags);
125 map->add_feature (HB_TAG('r','p','h','f'), F_MANUAL_ZWJ | F_PER_SYLLABLE);
126 map->add_gsub_pause (record_rphf_use);
127 map->add_gsub_pause (_hb_clear_substitution_flags);
128 map->enable_feature (HB_TAG('p','r','e','f'), F_MANUAL_ZWJ | F_PER_SYLLABLE);
129 map->add_gsub_pause (record_pref_use);
130
131 /* "Orthographic unit shaping group" */
132 for (unsigned int i = 0; i < ARRAY_LENGTH (use_basic_features); i++)
133 map->enable_feature (use_basic_features[i], F_MANUAL_ZWJ | F_PER_SYLLABLE);
134
135 map->add_gsub_pause (reorder_use);
136 map->add_gsub_pause (hb_syllabic_clear_var); // Don't need syllables anymore, use stop to free buffer var
137
138 /* "Topographical features" */
139 for (unsigned int i = 0; i < ARRAY_LENGTH (use_topographical_features); i++)
140 map->add_feature (use_topographical_features[i]);
141 map->add_gsub_pause (nullptr);
142
143 /* "Standard typographic presentation" */
144 for (unsigned int i = 0; i < ARRAY_LENGTH (use_other_features); i++)
145 map->enable_feature (use_other_features[i], F_MANUAL_ZWJ);
146}
147
148struct use_shape_plan_t
149{
150 hb_mask_t rphf_mask;
151
152 arabic_shape_plan_t *arabic_plan;
153};
154
155static void *
156data_create_use (const hb_ot_shape_plan_t *plan)
157{
158 use_shape_plan_t *use_plan = (use_shape_plan_t *) hb_calloc (1, sizeof (use_shape_plan_t));
159 if (unlikely (!use_plan))
160 return nullptr;
161
162 use_plan->rphf_mask = plan->map.get_1_mask (HB_TAG('r','p','h','f'));
163
164 if (has_arabic_joining (plan->props.script))
165 {
166 use_plan->arabic_plan = (arabic_shape_plan_t *) data_create_arabic (plan);
167 if (unlikely (!use_plan->arabic_plan))
168 {
169 hb_free (use_plan);
170 return nullptr;
171 }
172 }
173
174 return use_plan;
175}
176
177static void
178data_destroy_use (void *data)
179{
180 use_shape_plan_t *use_plan = (use_shape_plan_t *) data;
181
182 if (use_plan->arabic_plan)
183 data_destroy_arabic (use_plan->arabic_plan);
184
185 hb_free (data);
186}
187
188static void
189setup_masks_use (const hb_ot_shape_plan_t *plan,
190 hb_buffer_t *buffer,
191 hb_font_t *font HB_UNUSED)
192{
193 const use_shape_plan_t *use_plan = (const use_shape_plan_t *) plan->data;
194
195 /* Do this before allocating use_category(). */
196 if (use_plan->arabic_plan)
197 {
198 setup_masks_arabic_plan (use_plan->arabic_plan, buffer, plan->props.script);
199 }
200
201 HB_BUFFER_ALLOCATE_VAR (buffer, use_category);
202
203 /* We cannot setup masks here. We save information about characters
204 * and setup masks later on in a pause-callback. */
205
206 unsigned int count = buffer->len;
207 hb_glyph_info_t *info = buffer->info;
208 for (unsigned int i = 0; i < count; i++)
209 info[i].use_category() = hb_use_get_category (info[i].codepoint);
210}
211
212static void
213setup_rphf_mask (const hb_ot_shape_plan_t *plan,
214 hb_buffer_t *buffer)
215{
216 const use_shape_plan_t *use_plan = (const use_shape_plan_t *) plan->data;
217
218 hb_mask_t mask = use_plan->rphf_mask;
219 if (!mask) return;
220
221 hb_glyph_info_t *info = buffer->info;
222
223 foreach_syllable (buffer, start, end)
224 {
225 unsigned int limit = info[start].use_category() == USE(R) ? 1 : hb_min (3u, end - start);
226 for (unsigned int i = start; i < start + limit; i++)
227 info[i].mask |= mask;
228 }
229}
230
231static void
232setup_topographical_masks (const hb_ot_shape_plan_t *plan,
233 hb_buffer_t *buffer)
234{
235 const use_shape_plan_t *use_plan = (const use_shape_plan_t *) plan->data;
236 if (use_plan->arabic_plan)
237 return;
238
239 static_assert ((JOINING_FORM_INIT < 4 && JOINING_FORM_ISOL < 4 && JOINING_FORM_MEDI < 4 && JOINING_FORM_FINA < 4), "");
240 hb_mask_t masks[4], all_masks = 0;
241 for (unsigned int i = 0; i < 4; i++)
242 {
243 masks[i] = plan->map.get_1_mask (use_topographical_features[i]);
244 if (masks[i] == plan->map.get_global_mask ())
245 masks[i] = 0;
246 all_masks |= masks[i];
247 }
248 if (!all_masks)
249 return;
250 hb_mask_t other_masks = ~all_masks;
251
252 unsigned int last_start = 0;
253 joining_form_t last_form = _JOINING_FORM_NONE;
254 hb_glyph_info_t *info = buffer->info;
255 foreach_syllable (buffer, start, end)
256 {
257 use_syllable_type_t syllable_type = (use_syllable_type_t) (info[start].syllable() & 0x0F);
258 switch (syllable_type)
259 {
260 case use_hieroglyph_cluster:
261 case use_non_cluster:
262 /* These don't join. Nothing to do. */
263 last_form = _JOINING_FORM_NONE;
264 break;
265
266 case use_virama_terminated_cluster:
267 case use_sakot_terminated_cluster:
268 case use_standard_cluster:
269 case use_number_joiner_terminated_cluster:
270 case use_numeral_cluster:
271 case use_symbol_cluster:
272 case use_broken_cluster:
273
274 bool join = last_form == JOINING_FORM_FINA || last_form == JOINING_FORM_ISOL;
275
276 if (join)
277 {
278 /* Fixup previous syllable's form. */
279 last_form = last_form == JOINING_FORM_FINA ? JOINING_FORM_MEDI : JOINING_FORM_INIT;
280 for (unsigned int i = last_start; i < start; i++)
281 info[i].mask = (info[i].mask & other_masks) | masks[last_form];
282 }
283
284 /* Form for this syllable. */
285 last_form = join ? JOINING_FORM_FINA : JOINING_FORM_ISOL;
286 for (unsigned int i = start; i < end; i++)
287 info[i].mask = (info[i].mask & other_masks) | masks[last_form];
288
289 break;
290 }
291
292 last_start = start;
293 }
294}
295
296static bool
297setup_syllables_use (const hb_ot_shape_plan_t *plan,
298 hb_font_t *font HB_UNUSED,
299 hb_buffer_t *buffer)
300{
301 HB_BUFFER_ALLOCATE_VAR (buffer, syllable);
302 find_syllables_use (buffer);
303 foreach_syllable (buffer, start, end)
304 buffer->unsafe_to_break (start, end);
305 setup_rphf_mask (plan, buffer);
306 setup_topographical_masks (plan, buffer);
307 return false;
308}
309
310static bool
311record_rphf_use (const hb_ot_shape_plan_t *plan,
312 hb_font_t *font HB_UNUSED,
313 hb_buffer_t *buffer)
314{
315 const use_shape_plan_t *use_plan = (const use_shape_plan_t *) plan->data;
316
317 hb_mask_t mask = use_plan->rphf_mask;
318 if (!mask) return false;
319 hb_glyph_info_t *info = buffer->info;
320
321 foreach_syllable (buffer, start, end)
322 {
323 /* Mark a substituted repha as USE(R). */
324 for (unsigned int i = start; i < end && (info[i].mask & mask); i++)
325 if (_hb_glyph_info_substituted (&info[i]))
326 {
327 info[i].use_category() = USE(R);
328 break;
329 }
330 }
331 return false;
332}
333
334static bool
335record_pref_use (const hb_ot_shape_plan_t *plan HB_UNUSED,
336 hb_font_t *font HB_UNUSED,
337 hb_buffer_t *buffer)
338{
339 hb_glyph_info_t *info = buffer->info;
340
341 foreach_syllable (buffer, start, end)
342 {
343 /* Mark a substituted pref as VPre, as they behave the same way. */
344 for (unsigned int i = start; i < end; i++)
345 if (_hb_glyph_info_substituted (&info[i]))
346 {
347 info[i].use_category() = USE(VPre);
348 break;
349 }
350 }
351 return false;
352}
353
354static inline bool
355is_halant_use (const hb_glyph_info_t &info)
356{
357 return (info.use_category() == USE(H) || info.use_category() == USE(HVM) || info.use_category() == USE(IS)) &&
358 !_hb_glyph_info_ligated (&info);
359}
360
361static void
362reorder_syllable_use (hb_buffer_t *buffer, unsigned int start, unsigned int end)
363{
364 use_syllable_type_t syllable_type = (use_syllable_type_t) (buffer->info[start].syllable() & 0x0F);
365 /* Only a few syllable types need reordering. */
366 if (unlikely (!(FLAG_UNSAFE (syllable_type) &
367 (FLAG (use_virama_terminated_cluster) |
368 FLAG (use_sakot_terminated_cluster) |
369 FLAG (use_standard_cluster) |
370 FLAG (use_symbol_cluster) |
371 FLAG (use_broken_cluster) |
372 0))))
373 return;
374
375 hb_glyph_info_t *info = buffer->info;
376
377#define POST_BASE_FLAGS64 (FLAG64 (USE(FAbv)) | \
378 FLAG64 (USE(FBlw)) | \
379 FLAG64 (USE(FPst)) | \
380 FLAG64 (USE(FMAbv)) | \
381 FLAG64 (USE(FMBlw)) | \
382 FLAG64 (USE(FMPst)) | \
383 FLAG64 (USE(MAbv)) | \
384 FLAG64 (USE(MBlw)) | \
385 FLAG64 (USE(MPst)) | \
386 FLAG64 (USE(MPre)) | \
387 FLAG64 (USE(VAbv)) | \
388 FLAG64 (USE(VBlw)) | \
389 FLAG64 (USE(VPst)) | \
390 FLAG64 (USE(VPre)) | \
391 FLAG64 (USE(VMAbv)) | \
392 FLAG64 (USE(VMBlw)) | \
393 FLAG64 (USE(VMPst)) | \
394 FLAG64 (USE(VMPre)))
395
396 /* Move things forward. */
397 if (info[start].use_category() == USE(R) && end - start > 1)
398 {
399 /* Got a repha. Reorder it towards the end, but before the first post-base
400 * glyph. */
401 for (unsigned int i = start + 1; i < end; i++)
402 {
403 bool is_post_base_glyph = (FLAG64_UNSAFE (info[i].use_category()) & POST_BASE_FLAGS64) ||
404 is_halant_use (info[i]);
405 if (is_post_base_glyph || i == end - 1)
406 {
407 /* If we hit a post-base glyph, move before it; otherwise move to the
408 * end. Shift things in between backward. */
409
410 if (is_post_base_glyph)
411 i--;
412
413 buffer->merge_clusters (start, i + 1);
414 hb_glyph_info_t t = info[start];
415 memmove (&info[start], &info[start + 1], (i - start) * sizeof (info[0]));
416 info[i] = t;
417
418 break;
419 }
420 }
421 }
422
423 /* Move things back. */
424 unsigned int j = start;
425 for (unsigned int i = start; i < end; i++)
426 {
427 uint32_t flag = FLAG_UNSAFE (info[i].use_category());
428 if (is_halant_use (info[i]))
429 {
430 /* If we hit a halant, move after it; otherwise move to the beginning, and
431 * shift things in between forward. */
432 j = i + 1;
433 }
434 else if (((flag) & (FLAG (USE(VPre)) | FLAG (USE(VMPre)))) &&
435 /* Only move the first component of a MultipleSubst. */
436 0 == _hb_glyph_info_get_lig_comp (&info[i]) &&
437 j < i)
438 {
439 buffer->merge_clusters (j, i + 1);
440 hb_glyph_info_t t = info[i];
441 memmove (&info[j + 1], &info[j], (i - j) * sizeof (info[0]));
442 info[j] = t;
443 }
444 }
445}
446
447static bool
448reorder_use (const hb_ot_shape_plan_t *plan,
449 hb_font_t *font,
450 hb_buffer_t *buffer)
451{
452 bool ret = false;
453 if (buffer->message (font, "start reordering USE"))
454 {
455 if (hb_syllabic_insert_dotted_circles (font, buffer,
456 use_broken_cluster,
457 USE(B),
458 USE(R)))
459 ret = true;
460
461 foreach_syllable (buffer, start, end)
462 reorder_syllable_use (buffer, start, end);
463
464 (void) buffer->message (font, "end reordering USE");
465 }
466
467 HB_BUFFER_DEALLOCATE_VAR (buffer, use_category);
468
469 return ret;
470}
471
472
473static void
474preprocess_text_use (const hb_ot_shape_plan_t *plan,
475 hb_buffer_t *buffer,
476 hb_font_t *font)
477{
478 _hb_preprocess_text_vowel_constraints (plan, buffer, font);
479}
480
481static bool
482compose_use (const hb_ot_shape_normalize_context_t *c,
483 hb_codepoint_t a,
484 hb_codepoint_t b,
485 hb_codepoint_t *ab)
486{
487 /* Avoid recomposing split matras. */
488 if (HB_UNICODE_GENERAL_CATEGORY_IS_MARK (c->unicode->general_category (a)))
489 return false;
490
491 return (bool)c->unicode->compose (a, b, ab);
492}
493
494
495const hb_ot_shaper_t _hb_ot_shaper_use =
496{
497 collect_features_use,
498 nullptr, /* override_features */
499 data_create_use,
500 data_destroy_use,
501 preprocess_text_use,
502 nullptr, /* postprocess_glyphs */
503 nullptr, /* decompose */
504 compose_use,
505 setup_masks_use,
506 nullptr, /* reorder_marks */
507 HB_TAG_NONE, /* gpos_tag */
508 HB_OT_SHAPE_NORMALIZATION_MODE_COMPOSED_DIACRITICS_NO_SHORT_CIRCUIT,
509 HB_OT_SHAPE_ZERO_WIDTH_MARKS_BY_GDEF_EARLY,
510 false, /* fallback_position */
511};
512
513
514#endif
515