2 * Copyright © 2015 Mozilla Foundation.
3 * Copyright © 2015 Google, Inc.
5 * This is part of HarfBuzz, a text shaping library.
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.
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
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.
25 * Mozilla Author(s): Jonathan Kew
26 * Google Author(s): Behdad Esfahbod
29 #include "hb-ot-shape-complex-use-private.hh"
30 #include "hb-ot-shape-complex-arabic-private.hh"
32 /* buffer var allocations */
33 #define use_category() complex_var_u8_0()
37 * Universal Shaping Engine.
38 * https://www.microsoft.com/typography/OpenTypeDev/USE/intro.htm
46 * These features are applied all at once, before reordering.
48 HB_TAG('r','k','r','f'),
49 HB_TAG('a','b','v','f'),
50 HB_TAG('b','l','w','f'),
51 HB_TAG('h','a','l','f'),
52 HB_TAG('p','s','t','f'),
53 HB_TAG('v','a','t','u'),
54 HB_TAG('c','j','c','t'),
59 HB_TAG('i','s','o','l'),
60 HB_TAG('i','n','i','t'),
61 HB_TAG('m','e','d','i'),
62 HB_TAG('f','i','n','a'),
63 /* The spec doesn't specify these but we apply anyway, since our Arabic shaper
64 * does. These are only used in Syriac spec. */
65 HB_TAG('m','e','d','2'),
66 HB_TAG('f','i','n','2'),
67 HB_TAG('f','i','n','3'),
69 /* Same order as arabic_features. Don't need Syriac stuff.*/
82 * These features are applied all at once, after reordering.
84 HB_TAG('a','b','v','s'),
85 HB_TAG('b','l','w','s'),
86 HB_TAG('h','a','l','n'),
87 HB_TAG('p','r','e','s'),
88 HB_TAG('p','s','t','s'),
89 /* Positioning features, though we don't care about the types. */
90 HB_TAG('d','i','s','t'),
91 HB_TAG('a','b','v','m'),
92 HB_TAG('b','l','w','m'),
96 setup_syllables (const hb_ot_shape_plan_t *plan,
100 clear_substitution_flags (const hb_ot_shape_plan_t *plan,
102 hb_buffer_t *buffer);
104 record_rphf (const hb_ot_shape_plan_t *plan,
106 hb_buffer_t *buffer);
108 record_pref (const hb_ot_shape_plan_t *plan,
110 hb_buffer_t *buffer);
112 reorder (const hb_ot_shape_plan_t *plan,
114 hb_buffer_t *buffer);
117 collect_features_use (hb_ot_shape_planner_t *plan)
119 hb_ot_map_builder_t *map = &plan->map;
121 /* Do this before any lookups have been applied. */
122 map->add_gsub_pause (setup_syllables);
124 /* "Default glyph pre-processing group" */
125 map->add_global_bool_feature (HB_TAG('l','o','c','l'));
126 map->add_global_bool_feature (HB_TAG('c','c','m','p'));
127 map->add_global_bool_feature (HB_TAG('n','u','k','t'));
128 map->add_global_bool_feature (HB_TAG('a','k','h','n'));
130 /* "Reordering group" */
131 map->add_gsub_pause (clear_substitution_flags);
132 map->add_feature (HB_TAG('r','p','h','f'), 1, F_MANUAL_ZWJ);
133 map->add_gsub_pause (record_rphf);
134 map->add_gsub_pause (clear_substitution_flags);
135 map->add_feature (HB_TAG('p','r','e','f'), 1, F_GLOBAL | F_MANUAL_ZWJ);
136 map->add_gsub_pause (record_pref);
138 /* "Orthographic unit shaping group" */
139 for (unsigned int i = 0; i < ARRAY_LENGTH (basic_features); i++)
140 map->add_feature (basic_features[i], 1, F_GLOBAL | F_MANUAL_ZWJ);
142 map->add_gsub_pause (reorder);
144 /* "Topographical features" */
145 for (unsigned int i = 0; i < ARRAY_LENGTH (arabic_features); i++)
146 map->add_feature (arabic_features[i], 1, F_NONE);
147 map->add_gsub_pause (NULL);
149 /* "Standard typographic presentation" and "Positional feature application" */
150 for (unsigned int i = 0; i < ARRAY_LENGTH (other_features); i++)
151 map->add_feature (other_features[i], 1, F_GLOBAL | F_MANUAL_ZWJ);
154 struct use_shape_plan_t
160 arabic_shape_plan_t *arabic_plan;
164 has_arabic_joining (hb_script_t script)
166 /* List of scripts that have data in arabic-table. */
167 switch ((int) script)
169 /* Unicode-1.1 additions */
170 case HB_SCRIPT_ARABIC:
172 /* Unicode-3.0 additions */
173 case HB_SCRIPT_MONGOLIAN:
174 case HB_SCRIPT_SYRIAC:
176 /* Unicode-5.0 additions */
178 case HB_SCRIPT_PHAGS_PA:
180 /* Unicode-6.0 additions */
181 case HB_SCRIPT_MANDAIC:
183 /* Unicode-7.0 additions */
184 case HB_SCRIPT_MANICHAEAN:
185 case HB_SCRIPT_PSALTER_PAHLAVI:
187 /* Unicode-9.0 additions */
188 case HB_SCRIPT_ADLAM:
198 data_create_use (const hb_ot_shape_plan_t *plan)
200 use_shape_plan_t *use_plan = (use_shape_plan_t *) calloc (1, sizeof (use_shape_plan_t));
201 if (unlikely (!use_plan))
204 use_plan->rphf_mask = plan->map.get_1_mask (HB_TAG('r','p','h','f'));
206 if (has_arabic_joining (plan->props.script))
208 use_plan->arabic_plan = (arabic_shape_plan_t *) data_create_arabic (plan);
209 if (unlikely (!use_plan->arabic_plan))
220 data_destroy_use (void *data)
222 use_shape_plan_t *use_plan = (use_shape_plan_t *) data;
224 if (use_plan->arabic_plan)
225 data_destroy_arabic (use_plan->arabic_plan);
230 enum syllable_type_t {
232 virama_terminated_cluster,
234 number_joiner_terminated_cluster,
241 #include "hb-ot-shape-complex-use-machine.hh"
245 setup_masks_use (const hb_ot_shape_plan_t *plan,
247 hb_font_t *font HB_UNUSED)
249 const use_shape_plan_t *use_plan = (const use_shape_plan_t *) plan->data;
251 /* Do this before allocating use_category(). */
252 if (use_plan->arabic_plan)
254 setup_masks_arabic_plan (use_plan->arabic_plan, buffer, plan->props.script);
257 HB_BUFFER_ALLOCATE_VAR (buffer, use_category);
259 /* We cannot setup masks here. We save information about characters
260 * and setup masks later on in a pause-callback. */
262 unsigned int count = buffer->len;
263 hb_glyph_info_t *info = buffer->info;
264 for (unsigned int i = 0; i < count; i++)
265 info[i].use_category() = hb_use_get_categories (info[i].codepoint);
269 setup_rphf_mask (const hb_ot_shape_plan_t *plan,
272 const use_shape_plan_t *use_plan = (const use_shape_plan_t *) plan->data;
274 hb_mask_t mask = use_plan->rphf_mask;
277 hb_glyph_info_t *info = buffer->info;
279 foreach_syllable (buffer, start, end)
281 unsigned int limit = info[start].use_category() == USE_R ? 1 : MIN (3u, end - start);
282 for (unsigned int i = start; i < start + limit; i++)
283 info[i].mask |= mask;
288 setup_topographical_masks (const hb_ot_shape_plan_t *plan,
291 const use_shape_plan_t *use_plan = (const use_shape_plan_t *) plan->data;
292 if (use_plan->arabic_plan)
295 ASSERT_STATIC (INIT < 4 && ISOL < 4 && MEDI < 4 && FINA < 4);
296 hb_mask_t masks[4], all_masks = 0;
297 for (unsigned int i = 0; i < 4; i++)
299 masks[i] = plan->map.get_1_mask (arabic_features[i]);
300 if (masks[i] == plan->map.get_global_mask ())
302 all_masks |= masks[i];
306 hb_mask_t other_masks = ~all_masks;
308 unsigned int last_start = 0;
309 joining_form_t last_form = _NONE;
310 hb_glyph_info_t *info = buffer->info;
311 foreach_syllable (buffer, start, end)
313 syllable_type_t syllable_type = (syllable_type_t) (info[start].syllable() & 0x0F);
314 switch (syllable_type)
316 case independent_cluster:
319 /* These don't join. Nothing to do. */
323 case virama_terminated_cluster:
324 case standard_cluster:
325 case number_joiner_terminated_cluster:
326 case numeral_cluster:
329 bool join = last_form == FINA || last_form == ISOL;
333 /* Fixup previous syllable's form. */
334 last_form = last_form == FINA ? MEDI : INIT;
335 for (unsigned int i = last_start; i < start; i++)
336 info[i].mask = (info[i].mask & other_masks) | masks[last_form];
339 /* Form for this syllable. */
340 last_form = join ? FINA : ISOL;
341 for (unsigned int i = start; i < end; i++)
342 info[i].mask = (info[i].mask & other_masks) | masks[last_form];
352 setup_syllables (const hb_ot_shape_plan_t *plan,
353 hb_font_t *font HB_UNUSED,
356 find_syllables (buffer);
357 setup_rphf_mask (plan, buffer);
358 setup_topographical_masks (plan, buffer);
362 clear_substitution_flags (const hb_ot_shape_plan_t *plan,
363 hb_font_t *font HB_UNUSED,
366 hb_glyph_info_t *info = buffer->info;
367 unsigned int count = buffer->len;
368 for (unsigned int i = 0; i < count; i++)
369 _hb_glyph_info_clear_substituted (&info[i]);
373 record_rphf (const hb_ot_shape_plan_t *plan,
377 const use_shape_plan_t *use_plan = (const use_shape_plan_t *) plan->data;
379 hb_mask_t mask = use_plan->rphf_mask;
381 hb_glyph_info_t *info = buffer->info;
383 foreach_syllable (buffer, start, end)
385 /* Mark a substituted repha as USE_R. */
386 for (unsigned int i = start; i < end && (info[i].mask & mask); i++)
387 if (_hb_glyph_info_substituted (&info[i]))
389 info[i].use_category() = USE_R;
396 record_pref (const hb_ot_shape_plan_t *plan,
400 hb_glyph_info_t *info = buffer->info;
402 foreach_syllable (buffer, start, end)
404 /* Mark a substituted pref as VPre, as they behave the same way. */
405 for (unsigned int i = start; i < end; i++)
406 if (_hb_glyph_info_substituted (&info[i]))
408 info[i].use_category() = USE_VPre;
415 is_halant (const hb_glyph_info_t &info)
417 return info.use_category() == USE_H && !_hb_glyph_info_ligated (&info);
421 reorder_syllable (hb_buffer_t *buffer, unsigned int start, unsigned int end)
423 syllable_type_t syllable_type = (syllable_type_t) (buffer->info[start].syllable() & 0x0F);
424 /* Only a few syllable types need reordering. */
425 if (unlikely (!(FLAG_SAFE (syllable_type) &
426 (FLAG (virama_terminated_cluster) |
427 FLAG (standard_cluster) |
428 FLAG (broken_cluster) |
432 hb_glyph_info_t *info = buffer->info;
434 #define BASE_FLAGS (FLAG (USE_B) | FLAG (USE_GB))
436 /* Move things forward. */
437 if (info[start].use_category() == USE_R && end - start > 1)
439 /* Got a repha. Reorder it to after first base, before first halant. */
440 for (unsigned int i = start + 1; i < end; i++)
441 if ((FLAG_UNSAFE (info[i].use_category()) & (BASE_FLAGS)) || is_halant (info[i]))
443 /* If we hit a halant, move before it; otherwise it's a base: move to it's
444 * place, and shift things in between backward. */
446 if (is_halant (info[i]))
449 buffer->merge_clusters (start, i + 1);
450 hb_glyph_info_t t = info[start];
451 memmove (&info[start], &info[start + 1], (i - start) * sizeof (info[0]));
458 /* Move things back. */
459 unsigned int j = end;
460 for (unsigned int i = start; i < end; i++)
462 uint32_t flag = FLAG_UNSAFE (info[i].use_category());
463 if ((flag & (BASE_FLAGS)) || is_halant (info[i]))
465 /* If we hit a halant, move after it; otherwise it's a base: move to it's
466 * place, and shift things in between backward. */
467 if (is_halant (info[i]))
472 else if (((flag) & (FLAG (USE_VPre) | FLAG (USE_VMPre))) &&
473 /* Only move the first component of a MultipleSubst. */
474 0 == _hb_glyph_info_get_lig_comp (&info[i]) &&
477 buffer->merge_clusters (j, i + 1);
478 hb_glyph_info_t t = info[i];
479 memmove (&info[j + 1], &info[j], (i - j) * sizeof (info[0]));
486 insert_dotted_circles (const hb_ot_shape_plan_t *plan HB_UNUSED,
490 /* Note: This loop is extra overhead, but should not be measurable. */
491 bool has_broken_syllables = false;
492 unsigned int count = buffer->len;
493 hb_glyph_info_t *info = buffer->info;
494 for (unsigned int i = 0; i < count; i++)
495 if ((info[i].syllable() & 0x0F) == broken_cluster)
497 has_broken_syllables = true;
500 if (likely (!has_broken_syllables))
503 hb_glyph_info_t dottedcircle = {0};
504 if (!font->get_nominal_glyph (0x25CCu, &dottedcircle.codepoint))
506 dottedcircle.use_category() = hb_use_get_categories (0x25CC);
508 buffer->clear_output ();
511 unsigned int last_syllable = 0;
512 while (buffer->idx < buffer->len && !buffer->in_error)
514 unsigned int syllable = buffer->cur().syllable();
515 syllable_type_t syllable_type = (syllable_type_t) (syllable & 0x0F);
516 if (unlikely (last_syllable != syllable && syllable_type == broken_cluster))
518 last_syllable = syllable;
520 hb_glyph_info_t ginfo = dottedcircle;
521 ginfo.cluster = buffer->cur().cluster;
522 ginfo.mask = buffer->cur().mask;
523 ginfo.syllable() = buffer->cur().syllable();
524 /* TODO Set glyph_props? */
526 /* Insert dottedcircle after possible Repha. */
527 while (buffer->idx < buffer->len && !buffer->in_error &&
528 last_syllable == buffer->cur().syllable() &&
529 buffer->cur().use_category() == USE_R)
530 buffer->next_glyph ();
532 buffer->output_info (ginfo);
535 buffer->next_glyph ();
538 buffer->swap_buffers ();
542 reorder (const hb_ot_shape_plan_t *plan,
546 insert_dotted_circles (plan, font, buffer);
548 hb_glyph_info_t *info = buffer->info;
550 foreach_syllable (buffer, start, end)
551 reorder_syllable (buffer, start, end);
553 /* Zero syllables now... */
554 unsigned int count = buffer->len;
555 for (unsigned int i = 0; i < count; i++)
556 info[i].syllable() = 0;
558 HB_BUFFER_DEALLOCATE_VAR (buffer, use_category);
562 decompose_use (const hb_ot_shape_normalize_context_t *c,
570 * Special case where the Unicode decomp gives matras in the wrong order
571 * for cluster validation.
573 case 0x1112Eu : *a = 0x11127u; *b= 0x11131u; return true;
574 case 0x1112Fu : *a = 0x11127u; *b= 0x11132u; return true;
577 return (bool) c->unicode->decompose (ab, a, b);
581 compose_use (const hb_ot_shape_normalize_context_t *c,
586 /* Avoid recomposing split matras. */
587 if (HB_UNICODE_GENERAL_CATEGORY_IS_MARK (c->unicode->general_category (a)))
590 return (bool)c->unicode->compose (a, b, ab);
594 const hb_ot_complex_shaper_t _hb_ot_complex_shaper_use =
597 collect_features_use,
598 NULL, /* override_features */
601 NULL, /* preprocess_text */
602 NULL, /* postprocess_glyphs */
603 HB_OT_SHAPE_NORMALIZATION_MODE_COMPOSED_DIACRITICS_NO_SHORT_CIRCUIT,
607 NULL, /* disable_otl */
608 HB_OT_SHAPE_ZERO_WIDTH_MARKS_BY_GDEF_EARLY,
609 false, /* fallback_position */