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-private.hh" 30#include "hb-ot-shape-complex-arabic-private.hh" 31 32/* buffer var allocations */ 33#define use_category() complex_var_u8_0() 34 35 36/* 37 * Universal Shaping Engine. 38 * https://www.microsoft.com/typography/OpenTypeDev/USE/intro.htm 39 */ 40 41static const hb_tag_t 42basic_features[] = 43{ 44 /* 45 * Basic features. 46 * These features are applied all at once, before reordering. 47 */ 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'), 55}; 56static const hb_tag_t 57arabic_features[] = 58{ 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'), 68}; 69/* Same order as arabic_features. Don't need Syriac stuff.*/ 70enum joining_form_t { 71 ISOL, 72 INIT, 73 MEDI, 74 FINA, 75 _NONE 76}; 77static const hb_tag_t 78other_features[] = 79{ 80 /* 81 * Other features. 82 * These features are applied all at once, after reordering. 83 */ 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'), 93}; 94 95static void 96setup_syllables (const hb_ot_shape_plan_t *plan, 97 hb_font_t *font, 98 hb_buffer_t *buffer); 99static void 100clear_substitution_flags (const hb_ot_shape_plan_t *plan, 101 hb_font_t *font, 102 hb_buffer_t *buffer); 103static void 104record_rphf (const hb_ot_shape_plan_t *plan, 105 hb_font_t *font, 106 hb_buffer_t *buffer); 107static void 108record_pref (const hb_ot_shape_plan_t *plan, 109 hb_font_t *font, 110 hb_buffer_t *buffer); 111static void 112reorder (const hb_ot_shape_plan_t *plan, 113 hb_font_t *font, 114 hb_buffer_t *buffer); 115 116static void 117collect_features_use (hb_ot_shape_planner_t *plan) 118{ 119 hb_ot_map_builder_t *map = &plan->map; 120 121 /* Do this before any lookups have been applied. */ 122 map->add_gsub_pause (setup_syllables); 123 124 /* "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')); 129 130 /* "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); 137 138 /* "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); 141 142 map->add_gsub_pause (reorder); 143 144 /* "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); 148 149 /* "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); 152} 153 154struct use_shape_plan_t 155{ 156 ASSERT_POD (); 157 158 hb_mask_t rphf_mask; 159 160 arabic_shape_plan_t *arabic_plan; 161}; 162 163static bool 164has_arabic_joining (hb_script_t script) 165{ 166 /* List of scripts that have data in arabic-table. */ 167 switch ((int) script) 168 { 169 /* Unicode-1.1 additions */ 170 case HB_SCRIPT_ARABIC: 171 172 /* Unicode-3.0 additions */ 173 case HB_SCRIPT_MONGOLIAN: 174 case HB_SCRIPT_SYRIAC: 175 176 /* Unicode-5.0 additions */ 177 case HB_SCRIPT_NKO: 178 case HB_SCRIPT_PHAGS_PA: 179 180 /* Unicode-6.0 additions */ 181 case HB_SCRIPT_MANDAIC: 182 183 /* Unicode-7.0 additions */ 184 case HB_SCRIPT_MANICHAEAN: 185 case HB_SCRIPT_PSALTER_PAHLAVI: 186 187 return true; 188 189 default: 190 return false; 191 } 192} 193 194static void * 195data_create_use (const hb_ot_shape_plan_t *plan) 196{ 197 use_shape_plan_t *use_plan = (use_shape_plan_t *) calloc (1, sizeof (use_shape_plan_t)); 198 if (unlikely (!use_plan)) 199 return NULL; 200 201 use_plan->rphf_mask = plan->map.get_1_mask (HB_TAG('r','p','h','f')); 202 203 if (has_arabic_joining (plan->props.script)) 204 { 205 use_plan->arabic_plan = (arabic_shape_plan_t *) data_create_arabic (plan); 206 if (unlikely (!use_plan->arabic_plan)) 207 { 208 free (use_plan); 209 return NULL; 210 } 211 } 212 213 return use_plan; 214} 215 216static void 217data_destroy_use (void *data) 218{ 219 use_shape_plan_t *use_plan = (use_shape_plan_t *) data; 220 221 if (use_plan->arabic_plan) 222 data_destroy_arabic (use_plan->arabic_plan); 223 224 free (data); 225} 226 227enum syllable_type_t { 228 independent_cluster, 229 virama_terminated_cluster, 230 consonant_cluster, 231 vowel_cluster, 232 number_joiner_terminated_cluster, 233 numeral_cluster, 234 symbol_cluster, 235 broken_cluster, 236}; 237 238#include "hb-ot-shape-complex-use-machine.hh" 239 240 241static void 242setup_masks_use (const hb_ot_shape_plan_t *plan, 243 hb_buffer_t *buffer, 244 hb_font_t *font HB_UNUSED) 245{ 246 const use_shape_plan_t *use_plan = (const use_shape_plan_t *) plan->data; 247 248 /* Do this before allocating use_category(). */ 249 if (use_plan->arabic_plan) 250 { 251 setup_masks_arabic_plan (use_plan->arabic_plan, buffer, plan->props.script); 252 } 253 254 HB_BUFFER_ALLOCATE_VAR (buffer, use_category); 255 256 /* We cannot setup masks here. We save information about characters 257 * and setup masks later on in a pause-callback. */ 258 259 unsigned int count = buffer->len; 260 hb_glyph_info_t *info = buffer->info; 261 for (unsigned int i = 0; i < count; i++) 262 info[i].use_category() = hb_use_get_categories (info[i].codepoint); 263} 264 265static void 266setup_rphf_mask (const hb_ot_shape_plan_t *plan, 267 hb_buffer_t *buffer) 268{ 269 const use_shape_plan_t *use_plan = (const use_shape_plan_t *) plan->data; 270 271 hb_mask_t mask = use_plan->rphf_mask; 272 if (!mask) return; 273 274 hb_glyph_info_t *info = buffer->info; 275 276 foreach_syllable (buffer, start, end) 277 { 278 unsigned int limit = info[start].use_category() == USE_R ? 1 : MIN (3u, end - start); 279 for (unsigned int i = start; i < start + limit; i++) 280 info[i].mask |= mask; 281 } 282} 283 284static void 285setup_topographical_masks (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 if (use_plan->arabic_plan) 290 return; 291 292 ASSERT_STATIC (INIT < 4 && ISOL < 4 && MEDI < 4 && FINA < 4); 293 hb_mask_t masks[4], all_masks = 0; 294 for (unsigned int i = 0; i < 4; i++) 295 { 296 masks[i] = plan->map.get_1_mask (arabic_features[i]); 297 if (masks[i] == plan->map.get_global_mask ()) 298 masks[i] = 0; 299 all_masks |= masks[i]; 300 } 301 if (!all_masks) 302 return; 303 hb_mask_t other_masks = ~all_masks; 304 305 unsigned int last_start = 0; 306 joining_form_t last_form = _NONE; 307 hb_glyph_info_t *info = buffer->info; 308 foreach_syllable (buffer, start, end) 309 { 310 syllable_type_t syllable_type = (syllable_type_t) (info[start].syllable() & 0x0F); 311 switch (syllable_type) 312 { 313 case independent_cluster: 314 case symbol_cluster: 315 /* These don't join. Nothing to do. */ 316 last_form = _NONE; 317 break; 318 319 case virama_terminated_cluster: 320 case consonant_cluster: 321 case vowel_cluster: 322 case number_joiner_terminated_cluster: 323 case numeral_cluster: 324 case broken_cluster: 325 326 bool join = last_form == FINA || last_form == ISOL; 327 328 if (join) 329 { 330 /* Fixup previous syllable's form. */ 331 last_form = last_form == FINA ? MEDI : INIT; 332 for (unsigned int i = last_start; i < start; i++) 333 info[i].mask = (info[i].mask & other_masks) | masks[last_form]; 334 } 335 336 /* Form for this syllable. */ 337 last_form = join ? FINA : ISOL; 338 for (unsigned int i = start; i < end; i++) 339 info[i].mask = (info[i].mask & other_masks) | masks[last_form]; 340 341 break; 342 } 343 344 last_start = start; 345 } 346} 347 348static void 349setup_syllables (const hb_ot_shape_plan_t *plan, 350 hb_font_t *font HB_UNUSED, 351 hb_buffer_t *buffer) 352{ 353 find_syllables (buffer); 354 setup_rphf_mask (plan, buffer); 355 setup_topographical_masks (plan, buffer); 356} 357 358static void 359clear_substitution_flags (const hb_ot_shape_plan_t *plan, 360 hb_font_t *font HB_UNUSED, 361 hb_buffer_t *buffer) 362{ 363 hb_glyph_info_t *info = buffer->info; 364 unsigned int count = buffer->len; 365 for (unsigned int i = 0; i < count; i++) 366 _hb_glyph_info_clear_substituted (&info[i]); 367} 368 369static void 370record_rphf (const hb_ot_shape_plan_t *plan, 371 hb_font_t *font, 372 hb_buffer_t *buffer) 373{ 374 const use_shape_plan_t *use_plan = (const use_shape_plan_t *) plan->data; 375 376 hb_mask_t mask = use_plan->rphf_mask; 377 if (!mask) return; 378 hb_glyph_info_t *info = buffer->info; 379 380 foreach_syllable (buffer, start, end) 381 { 382 /* Mark a substituted repha as USE_R. */ 383 for (unsigned int i = start; i < end && (info[i].mask & mask); i++) 384 if (_hb_glyph_info_substituted (&info[i])) 385 { 386 info[i].use_category() = USE_R; 387 break; 388 } 389 } 390} 391 392static void 393record_pref (const hb_ot_shape_plan_t *plan, 394 hb_font_t *font, 395 hb_buffer_t *buffer) 396{ 397 hb_glyph_info_t *info = buffer->info; 398 399 foreach_syllable (buffer, start, end) 400 { 401 /* Mark a substituted pref as VPre, as they behave the same way. */ 402 for (unsigned int i = start; i < end; i++) 403 if (_hb_glyph_info_substituted (&info[i])) 404 { 405 info[i].use_category() = USE_VPre; 406 break; 407 } 408 } 409} 410 411static inline bool 412is_halant (const hb_glyph_info_t &info) 413{ 414 return info.use_category() == USE_H && !_hb_glyph_info_ligated (&info); 415} 416 417static void 418reorder_syllable (hb_buffer_t *buffer, unsigned int start, unsigned int end) 419{ 420 syllable_type_t syllable_type = (syllable_type_t) (buffer->info[start].syllable() & 0x0F); 421 /* Only a few syllable types need reordering. */ 422 if (unlikely (!(FLAG_SAFE (syllable_type) & 423 (FLAG (virama_terminated_cluster) | 424 FLAG (consonant_cluster) | 425 FLAG (vowel_cluster) | 426 FLAG (broken_cluster) | 427 0)))) 428 return; 429 430 hb_glyph_info_t *info = buffer->info; 431 432#define BASE_FLAGS (FLAG (USE_B) | FLAG (USE_GB) | FLAG (USE_IV)) 433 434 /* Move things forward. */ 435 if (info[start].use_category() == USE_R && end - start > 1) 436 { 437 /* Got a repha. Reorder it to after first base, before first halant. */ 438 for (unsigned int i = start + 1; i < end; i++) 439 if ((FLAG_UNSAFE (info[i].use_category()) & (BASE_FLAGS)) || is_halant (info[i])) 440 { 441 /* If we hit a halant, move before it; otherwise it's a base: move to it's 442 * place, and shift things in between backward. */ 443 444 if (is_halant (info[i])) 445 i--; 446 447 buffer->merge_clusters (start, i + 1); 448 hb_glyph_info_t t = info[start]; 449 memmove (&info[start], &info[start + 1], (i - start) * sizeof (info[0])); 450 info[i] = t; 451 452 break; 453 } 454 } 455 456 /* Move things back. */ 457 unsigned int j = end; 458 for (unsigned int i = start; i < end; i++) 459 { 460 uint32_t flag = FLAG_UNSAFE (info[i].use_category()); 461 if ((flag & (BASE_FLAGS)) || is_halant (info[i])) 462 { 463 /* If we hit a halant, move after it; otherwise it's a base: move to it's 464 * place, and shift things in between backward. */ 465 if (is_halant (info[i])) 466 j = i + 1; 467 else 468 j = i; 469 } 470 else if (((flag) & (FLAG (USE_VPre) | FLAG (USE_VMPre))) && 471 /* Only move the first component of a MultipleSubst. */ 472 0 == _hb_glyph_info_get_lig_comp (&info[i]) && 473 j < i) 474 { 475 buffer->merge_clusters (j, i + 1); 476 hb_glyph_info_t t = info[i]; 477 memmove (&info[j + 1], &info[j], (i - j) * sizeof (info[0])); 478 info[j] = t; 479 } 480 } 481} 482 483static inline void 484insert_dotted_circles (const hb_ot_shape_plan_t *plan HB_UNUSED, 485 hb_font_t *font, 486 hb_buffer_t *buffer) 487{ 488 /* Note: This loop is extra overhead, but should not be measurable. */ 489 bool has_broken_syllables = false; 490 unsigned int count = buffer->len; 491 hb_glyph_info_t *info = buffer->info; 492 for (unsigned int i = 0; i < count; i++) 493 if ((info[i].syllable() & 0x0F) == broken_cluster) 494 { 495 has_broken_syllables = true; 496 break; 497 } 498 if (likely (!has_broken_syllables)) 499 return; 500 501 hb_glyph_info_t dottedcircle = {0}; 502 if (!font->get_nominal_glyph (0x25CCu, &dottedcircle.codepoint)) 503 return; 504 dottedcircle.use_category() = hb_use_get_categories (0x25CC); 505 506 buffer->clear_output (); 507 508 buffer->idx = 0; 509 unsigned int last_syllable = 0; 510 while (buffer->idx < buffer->len && !buffer->in_error) 511 { 512 unsigned int syllable = buffer->cur().syllable(); 513 syllable_type_t syllable_type = (syllable_type_t) (syllable & 0x0F); 514 if (unlikely (last_syllable != syllable && syllable_type == broken_cluster)) 515 { 516 last_syllable = syllable; 517 518 hb_glyph_info_t ginfo = dottedcircle; 519 ginfo.cluster = buffer->cur().cluster; 520 ginfo.mask = buffer->cur().mask; 521 ginfo.syllable() = buffer->cur().syllable(); 522 /* TODO Set glyph_props? */ 523 524 /* Insert dottedcircle after possible Repha. */ 525 while (buffer->idx < buffer->len && !buffer->in_error && 526 last_syllable == buffer->cur().syllable() && 527 buffer->cur().use_category() == USE_R) 528 buffer->next_glyph (); 529 530 buffer->output_info (ginfo); 531 } 532 else 533 buffer->next_glyph (); 534 } 535 536 buffer->swap_buffers (); 537} 538 539static void 540reorder (const hb_ot_shape_plan_t *plan, 541 hb_font_t *font, 542 hb_buffer_t *buffer) 543{ 544 insert_dotted_circles (plan, font, buffer); 545 546 hb_glyph_info_t *info = buffer->info; 547 548 foreach_syllable (buffer, start, end) 549 reorder_syllable (buffer, start, end); 550 551 /* Zero syllables now... */ 552 unsigned int count = buffer->len; 553 for (unsigned int i = 0; i < count; i++) 554 info[i].syllable() = 0; 555 556 HB_BUFFER_DEALLOCATE_VAR (buffer, use_category); 557} 558 559static bool 560compose_use (const hb_ot_shape_normalize_context_t *c, 561 hb_codepoint_t a, 562 hb_codepoint_t b, 563 hb_codepoint_t *ab) 564{ 565 /* Avoid recomposing split matras. */ 566 if (HB_UNICODE_GENERAL_CATEGORY_IS_MARK (c->unicode->general_category (a))) 567 return false; 568 569 return (bool)c->unicode->compose (a, b, ab); 570} 571 572 573const hb_ot_complex_shaper_t _hb_ot_complex_shaper_use = 574{ 575 "use", 576 collect_features_use, 577 NULL, /* override_features */ 578 data_create_use, 579 data_destroy_use, 580 NULL, /* preprocess_text */ 581 NULL, /* postprocess_glyphs */ 582 HB_OT_SHAPE_NORMALIZATION_MODE_COMPOSED_DIACRITICS_NO_SHORT_CIRCUIT, 583 NULL, /* decompose */ 584 compose_use, 585 setup_masks_use, 586 HB_OT_SHAPE_ZERO_WIDTH_MARKS_BY_GDEF_EARLY, 587 false, /* fallback_position */ 588}; 589