1/* 2 * Copyright �� 2007,2008,2009 Red Hat, Inc. 3 * Copyright �� 2012,2013 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 * Red Hat Author(s): Behdad Esfahbod 26 * Google Author(s): Behdad Esfahbod 27 */ 28 29#ifndef HB_OT_LAYOUT_PRIVATE_HH 30#define HB_OT_LAYOUT_PRIVATE_HH 31 32#include "hb-private.hh" 33 34#include "hb-font-private.hh" 35#include "hb-buffer-private.hh" 36#include "hb-set-private.hh" 37 38 39/* Private API corresponding to hb-ot-layout.h: */ 40 41HB_INTERNAL hb_bool_t 42hb_ot_layout_table_find_feature (hb_face_t *face, 43 hb_tag_t table_tag, 44 hb_tag_t feature_tag, 45 unsigned int *feature_index); 46 47 48/* 49 * GDEF 50 */ 51 52enum hb_ot_layout_glyph_props_flags_t 53{ 54 /* The following three match LookupFlags::Ignore* numbers. */ 55 HB_OT_LAYOUT_GLYPH_PROPS_BASE_GLYPH = 0x02u, 56 HB_OT_LAYOUT_GLYPH_PROPS_LIGATURE = 0x04u, 57 HB_OT_LAYOUT_GLYPH_PROPS_MARK = 0x08u, 58 59 /* The following are used internally; not derived from GDEF. */ 60 HB_OT_LAYOUT_GLYPH_PROPS_SUBSTITUTED = 0x10u, 61 HB_OT_LAYOUT_GLYPH_PROPS_LIGATED = 0x20u, 62 HB_OT_LAYOUT_GLYPH_PROPS_MULTIPLIED = 0x40u, 63 64 HB_OT_LAYOUT_GLYPH_PROPS_PRESERVE = HB_OT_LAYOUT_GLYPH_PROPS_SUBSTITUTED | 65 HB_OT_LAYOUT_GLYPH_PROPS_LIGATED | 66 HB_OT_LAYOUT_GLYPH_PROPS_MULTIPLIED 67}; 68HB_MARK_AS_FLAG_T (hb_ot_layout_glyph_props_flags_t); 69 70 71/* 72 * GSUB/GPOS 73 */ 74 75HB_INTERNAL hb_bool_t 76hb_ot_layout_lookup_would_substitute_fast (hb_face_t *face, 77 unsigned int lookup_index, 78 const hb_codepoint_t *glyphs, 79 unsigned int glyphs_length, 80 hb_bool_t zero_context); 81 82 83/* Should be called before all the substitute_lookup's are done. */ 84HB_INTERNAL void 85hb_ot_layout_substitute_start (hb_font_t *font, 86 hb_buffer_t *buffer); 87 88 89struct hb_ot_layout_lookup_accelerator_t; 90 91namespace OT { 92 struct hb_apply_context_t; 93 struct SubstLookup; 94} 95 96HB_INTERNAL void 97hb_ot_layout_substitute_lookup (OT::hb_apply_context_t *c, 98 const OT::SubstLookup &lookup, 99 const hb_ot_layout_lookup_accelerator_t &accel); 100 101 102/* Should be called before all the position_lookup's are done. */ 103HB_INTERNAL void 104hb_ot_layout_position_start (hb_font_t *font, 105 hb_buffer_t *buffer); 106 107/* Should be called after all the position_lookup's are done, to finish advances. */ 108HB_INTERNAL void 109hb_ot_layout_position_finish_advances (hb_font_t *font, 110 hb_buffer_t *buffer); 111 112/* Should be called after hb_ot_layout_position_finish_advances, to finish offsets. */ 113HB_INTERNAL void 114hb_ot_layout_position_finish_offsets (hb_font_t *font, 115 hb_buffer_t *buffer); 116 117 118 119/* 120 * hb_ot_layout_t 121 */ 122 123namespace OT { 124 struct GDEF; 125 struct GSUB; 126 struct GPOS; 127 struct MATH; 128} 129 130struct hb_ot_layout_lookup_accelerator_t 131{ 132 template <typename TLookup> 133 inline void init (const TLookup &lookup) 134 { 135 digest.init (); 136 lookup.add_coverage (&digest); 137 } 138 139 inline void fini (void) 140 { 141 } 142 143 inline bool may_have (hb_codepoint_t g) const { 144 return digest.may_have (g); 145 } 146 147 private: 148 hb_set_digest_t digest; 149}; 150 151struct hb_ot_layout_t 152{ 153 hb_blob_t *gdef_blob; 154 hb_blob_t *gsub_blob; 155 hb_blob_t *gpos_blob; 156 hb_blob_t *math_blob; 157 158 const struct OT::GDEF *gdef; 159 const struct OT::GSUB *gsub; 160 const struct OT::GPOS *gpos; 161 const struct OT::MATH *math; 162 163 unsigned int gsub_lookup_count; 164 unsigned int gpos_lookup_count; 165 166 hb_ot_layout_lookup_accelerator_t *gsub_accels; 167 hb_ot_layout_lookup_accelerator_t *gpos_accels; 168}; 169 170 171HB_INTERNAL hb_ot_layout_t * 172_hb_ot_layout_create (hb_face_t *face); 173 174HB_INTERNAL void 175_hb_ot_layout_destroy (hb_ot_layout_t *layout); 176 177 178#define hb_ot_layout_from_face(face) ((hb_ot_layout_t *) face->shaper_data.ot) 179 180 181/* 182 * Buffer var routines. 183 */ 184 185/* buffer var allocations, used during the entire shaping process */ 186#define unicode_props() var2.u16[0] 187 188/* buffer var allocations, used during the GSUB/GPOS processing */ 189#define glyph_props() var1.u16[0] /* GDEF glyph properties */ 190#define lig_props() var1.u8[2] /* GSUB/GPOS ligature tracking */ 191#define syllable() var1.u8[3] /* GSUB/GPOS shaping boundaries */ 192 193 194/* loop over syllables */ 195 196#define foreach_syllable(buffer, start, end) \ 197 for (unsigned int \ 198 _count = buffer->len, \ 199 start = 0, end = _count ? _next_syllable (buffer, 0) : 0; \ 200 start < _count; \ 201 start = end, end = _next_syllable (buffer, start)) 202 203static inline unsigned int 204_next_syllable (hb_buffer_t *buffer, unsigned int start) 205{ 206 hb_glyph_info_t *info = buffer->info; 207 unsigned int count = buffer->len; 208 209 unsigned int syllable = info[start].syllable(); 210 while (++start < count && syllable == info[start].syllable()) 211 ; 212 213 return start; 214} 215 216 217/* unicode_props */ 218 219/* Design: 220 * unicode_props() is a two-byte number. The low byte includes: 221 * - General_Category: 5 bits. 222 * - A bit each for: 223 * * Is it Default_Ignorable(); we have a modified Default_Ignorable(). 224 * * Whether it's one of the three Mongolian Free Variation Selectors. 225 * * One free bit right now. 226 * 227 * The high-byte has different meanings, switched by the Gen-Cat: 228 * - For Mn,Mc,Me: the modified Combining_Class. 229 * - For Cf: whether it's ZWJ, ZWNJ, or something else. 230 * - For Ws: index of which space character this is, if space fallback 231 * is needed, ie. we don't set this by default, only if asked to. 232 */ 233 234enum hb_unicode_props_flags_t { 235 UPROPS_MASK_GEN_CAT = 0x001Fu, 236 UPROPS_MASK_IGNORABLE = 0x0020u, 237 UPROPS_MASK_FVS = 0x0040u, /* MONGOLIAN FREE VARIATION SELECTOR 1..3 */ 238 239 /* If GEN_CAT=FORMAT, top byte masks: */ 240 UPROPS_MASK_Cf_ZWJ = 0x0100u, 241 UPROPS_MASK_Cf_ZWNJ = 0x0200u 242}; 243HB_MARK_AS_FLAG_T (hb_unicode_props_flags_t); 244 245static inline void 246_hb_glyph_info_set_unicode_props (hb_glyph_info_t *info, hb_buffer_t *buffer) 247{ 248 hb_unicode_funcs_t *unicode = buffer->unicode; 249 unsigned int u = info->codepoint; 250 unsigned int gen_cat = (unsigned int) unicode->general_category (u); 251 unsigned int props = gen_cat; 252 253 if (u >= 0x80) 254 { 255 buffer->scratch_flags |= HB_BUFFER_SCRATCH_FLAG_HAS_NON_ASCII; 256 if (unlikely (unicode->is_default_ignorable (u))) 257 { 258 buffer->scratch_flags |= HB_BUFFER_SCRATCH_FLAG_HAS_DEFAULT_IGNORABLES; 259 props |= UPROPS_MASK_IGNORABLE; 260 if (u == 0x200Cu) props |= UPROPS_MASK_Cf_ZWNJ; 261 if (u == 0x200Du) props |= UPROPS_MASK_Cf_ZWJ; 262 /* Mongolian Free Variation Selectors need to be remembered 263 * because although we need to hide them like default-ignorables, 264 * they need to non-ignorable during shaping. This is similar to 265 * what we do for joiners in Indic-like shapers, but since the 266 * FVSes are GC=Mn, we have use a separate bit to remember them. 267 * Fixes: 268 * https://github.com/behdad/harfbuzz/issues/234 269 */ 270 if (unlikely (hb_in_range (u, 0x180Bu, 0x180Du))) props |= UPROPS_MASK_FVS; 271 } 272 else if (unlikely (HB_UNICODE_GENERAL_CATEGORY_IS_NON_ENCLOSING_MARK_OR_MODIFIER_SYMBOL (gen_cat))) 273 { 274 /* The above check is just an optimization to let in only things we need further 275 * processing on. */ 276 277 /* Only Mn and Mc can have non-zero ccc: 278 * http://www.unicode.org/policies/stability_policy.html#Property_Value 279 * """ 280 * Canonical_Combining_Class, General_Category 281 * All characters other than those with General_Category property values 282 * Spacing_Mark (Mc) and Nonspacing_Mark (Mn) have the Canonical_Combining_Class 283 * property value 0. 284 * 1.1.5+ 285 * """ 286 * 287 * Also, all Mn's that are Default_Ignorable, have ccc=0, hence 288 * the "else if". 289 */ 290 props |= unicode->modified_combining_class (info->codepoint)<<8; 291 292 /* Recategorize emoji skin-tone modifiers as Unicode mark, so they 293 * behave correctly in non-native directionality. They originally 294 * are MODIFIER_SYMBOL. Fixes: 295 * https://github.com/behdad/harfbuzz/issues/169 296 */ 297 if (unlikely (hb_in_range (u, 0x1F3FBu, 0x1F3FFu))) 298 { 299 props = gen_cat = HB_UNICODE_GENERAL_CATEGORY_ENCLOSING_MARK; 300 } 301 } 302 } 303 304 info->unicode_props() = props; 305} 306 307static inline void 308_hb_glyph_info_set_general_category (hb_glyph_info_t *info, 309 hb_unicode_general_category_t gen_cat) 310{ 311 /* Clears top-byte. */ 312 info->unicode_props() = (unsigned int) gen_cat | (info->unicode_props() & (0xFF & ~UPROPS_MASK_GEN_CAT)); 313} 314 315static inline hb_unicode_general_category_t 316_hb_glyph_info_get_general_category (const hb_glyph_info_t *info) 317{ 318 return (hb_unicode_general_category_t) (info->unicode_props() & UPROPS_MASK_GEN_CAT); 319} 320 321static inline bool 322_hb_glyph_info_is_unicode_mark (const hb_glyph_info_t *info) 323{ 324 return HB_UNICODE_GENERAL_CATEGORY_IS_MARK (info->unicode_props() & UPROPS_MASK_GEN_CAT); 325} 326static inline void 327_hb_glyph_info_set_modified_combining_class (hb_glyph_info_t *info, 328 unsigned int modified_class) 329{ 330 if (unlikely (!_hb_glyph_info_is_unicode_mark (info))) 331 return; 332 info->unicode_props() = (modified_class<<8) | (info->unicode_props() & 0xFF); 333} 334static inline unsigned int 335_hb_glyph_info_get_modified_combining_class (const hb_glyph_info_t *info) 336{ 337 return _hb_glyph_info_is_unicode_mark (info) ? info->unicode_props()>>8 : 0; 338} 339 340static inline bool 341_hb_glyph_info_is_unicode_space (const hb_glyph_info_t *info) 342{ 343 return _hb_glyph_info_get_general_category (info) == 344 HB_UNICODE_GENERAL_CATEGORY_SPACE_SEPARATOR; 345} 346static inline void 347_hb_glyph_info_set_unicode_space_fallback_type (hb_glyph_info_t *info, hb_unicode_funcs_t::space_t s) 348{ 349 if (unlikely (!_hb_glyph_info_is_unicode_space (info))) 350 return; 351 info->unicode_props() = (((unsigned int) s)<<8) | (info->unicode_props() & 0xFF); 352} 353static inline hb_unicode_funcs_t::space_t 354_hb_glyph_info_get_unicode_space_fallback_type (const hb_glyph_info_t *info) 355{ 356 return _hb_glyph_info_is_unicode_space (info) ? 357 (hb_unicode_funcs_t::space_t) (info->unicode_props()>>8) : 358 hb_unicode_funcs_t::NOT_SPACE; 359} 360 361static inline bool _hb_glyph_info_ligated (const hb_glyph_info_t *info); 362 363static inline hb_bool_t 364_hb_glyph_info_is_default_ignorable (const hb_glyph_info_t *info) 365{ 366 return (info->unicode_props() & UPROPS_MASK_IGNORABLE) && 367 !_hb_glyph_info_ligated (info); 368} 369static inline hb_bool_t 370_hb_glyph_info_is_default_ignorable_and_not_fvs (const hb_glyph_info_t *info) 371{ 372 return ((info->unicode_props() & (UPROPS_MASK_IGNORABLE|UPROPS_MASK_FVS)) 373 == UPROPS_MASK_IGNORABLE) && 374 !_hb_glyph_info_ligated (info); 375} 376 377static inline bool 378_hb_glyph_info_is_unicode_format (const hb_glyph_info_t *info) 379{ 380 return _hb_glyph_info_get_general_category (info) == 381 HB_UNICODE_GENERAL_CATEGORY_FORMAT; 382} 383static inline hb_bool_t 384_hb_glyph_info_is_zwnj (const hb_glyph_info_t *info) 385{ 386 return _hb_glyph_info_is_unicode_format (info) && (info->unicode_props() & UPROPS_MASK_Cf_ZWNJ); 387} 388static inline hb_bool_t 389_hb_glyph_info_is_zwj (const hb_glyph_info_t *info) 390{ 391 return _hb_glyph_info_is_unicode_format (info) && (info->unicode_props() & UPROPS_MASK_Cf_ZWJ); 392} 393static inline hb_bool_t 394_hb_glyph_info_is_joiner (const hb_glyph_info_t *info) 395{ 396 return _hb_glyph_info_is_unicode_format (info) && (info->unicode_props() & (UPROPS_MASK_Cf_ZWNJ|UPROPS_MASK_Cf_ZWJ)); 397} 398static inline void 399_hb_glyph_info_flip_joiners (hb_glyph_info_t *info) 400{ 401 if (!_hb_glyph_info_is_unicode_format (info)) 402 return; 403 info->unicode_props() ^= UPROPS_MASK_Cf_ZWNJ | UPROPS_MASK_Cf_ZWJ; 404} 405 406/* lig_props: aka lig_id / lig_comp 407 * 408 * When a ligature is formed: 409 * 410 * - The ligature glyph and any marks in between all the same newly allocated 411 * lig_id, 412 * - The ligature glyph will get lig_num_comps set to the number of components 413 * - The marks get lig_comp > 0, reflecting which component of the ligature 414 * they were applied to. 415 * - This is used in GPOS to attach marks to the right component of a ligature 416 * in MarkLigPos, 417 * - Note that when marks are ligated together, much of the above is skipped 418 * and the current lig_id reused. 419 * 420 * When a multiple-substitution is done: 421 * 422 * - All resulting glyphs will have lig_id = 0, 423 * - The resulting glyphs will have lig_comp = 0, 1, 2, ... respectively. 424 * - This is used in GPOS to attach marks to the first component of a 425 * multiple substitution in MarkBasePos. 426 * 427 * The numbers are also used in GPOS to do mark-to-mark positioning only 428 * to marks that belong to the same component of the same ligature. 429 */ 430 431static inline void 432_hb_glyph_info_clear_lig_props (hb_glyph_info_t *info) 433{ 434 info->lig_props() = 0; 435} 436 437#define IS_LIG_BASE 0x10 438 439static inline void 440_hb_glyph_info_set_lig_props_for_ligature (hb_glyph_info_t *info, 441 unsigned int lig_id, 442 unsigned int lig_num_comps) 443{ 444 info->lig_props() = (lig_id << 5) | IS_LIG_BASE | (lig_num_comps & 0x0F); 445} 446 447static inline void 448_hb_glyph_info_set_lig_props_for_mark (hb_glyph_info_t *info, 449 unsigned int lig_id, 450 unsigned int lig_comp) 451{ 452 info->lig_props() = (lig_id << 5) | (lig_comp & 0x0F); 453} 454 455static inline void 456_hb_glyph_info_set_lig_props_for_component (hb_glyph_info_t *info, unsigned int comp) 457{ 458 _hb_glyph_info_set_lig_props_for_mark (info, 0, comp); 459} 460 461static inline unsigned int 462_hb_glyph_info_get_lig_id (const hb_glyph_info_t *info) 463{ 464 return info->lig_props() >> 5; 465} 466 467static inline bool 468_hb_glyph_info_ligated_internal (const hb_glyph_info_t *info) 469{ 470 return !!(info->lig_props() & IS_LIG_BASE); 471} 472 473static inline unsigned int 474_hb_glyph_info_get_lig_comp (const hb_glyph_info_t *info) 475{ 476 if (_hb_glyph_info_ligated_internal (info)) 477 return 0; 478 else 479 return info->lig_props() & 0x0F; 480} 481 482static inline unsigned int 483_hb_glyph_info_get_lig_num_comps (const hb_glyph_info_t *info) 484{ 485 if ((info->glyph_props() & HB_OT_LAYOUT_GLYPH_PROPS_LIGATURE) && 486 _hb_glyph_info_ligated_internal (info)) 487 return info->lig_props() & 0x0F; 488 else 489 return 1; 490} 491 492static inline uint8_t 493_hb_allocate_lig_id (hb_buffer_t *buffer) { 494 uint8_t lig_id = buffer->next_serial () & 0x07; 495 if (unlikely (!lig_id)) 496 lig_id = _hb_allocate_lig_id (buffer); /* in case of overflow */ 497 return lig_id; 498} 499 500/* glyph_props: */ 501 502static inline void 503_hb_glyph_info_set_glyph_props (hb_glyph_info_t *info, unsigned int props) 504{ 505 info->glyph_props() = props; 506} 507 508static inline unsigned int 509_hb_glyph_info_get_glyph_props (const hb_glyph_info_t *info) 510{ 511 return info->glyph_props(); 512} 513 514static inline bool 515_hb_glyph_info_is_base_glyph (const hb_glyph_info_t *info) 516{ 517 return !!(info->glyph_props() & HB_OT_LAYOUT_GLYPH_PROPS_BASE_GLYPH); 518} 519 520static inline bool 521_hb_glyph_info_is_ligature (const hb_glyph_info_t *info) 522{ 523 return !!(info->glyph_props() & HB_OT_LAYOUT_GLYPH_PROPS_LIGATURE); 524} 525 526static inline bool 527_hb_glyph_info_is_mark (const hb_glyph_info_t *info) 528{ 529 return !!(info->glyph_props() & HB_OT_LAYOUT_GLYPH_PROPS_MARK); 530} 531 532static inline bool 533_hb_glyph_info_substituted (const hb_glyph_info_t *info) 534{ 535 return !!(info->glyph_props() & HB_OT_LAYOUT_GLYPH_PROPS_SUBSTITUTED); 536} 537 538static inline bool 539_hb_glyph_info_ligated (const hb_glyph_info_t *info) 540{ 541 return !!(info->glyph_props() & HB_OT_LAYOUT_GLYPH_PROPS_LIGATED); 542} 543 544static inline bool 545_hb_glyph_info_multiplied (const hb_glyph_info_t *info) 546{ 547 return !!(info->glyph_props() & HB_OT_LAYOUT_GLYPH_PROPS_MULTIPLIED); 548} 549 550static inline bool 551_hb_glyph_info_ligated_and_didnt_multiply (const hb_glyph_info_t *info) 552{ 553 return _hb_glyph_info_ligated (info) && !_hb_glyph_info_multiplied (info); 554} 555 556static inline void 557_hb_glyph_info_clear_ligated_and_multiplied (hb_glyph_info_t *info) 558{ 559 info->glyph_props() &= ~(HB_OT_LAYOUT_GLYPH_PROPS_LIGATED | 560 HB_OT_LAYOUT_GLYPH_PROPS_MULTIPLIED); 561} 562 563static inline void 564_hb_glyph_info_clear_substituted (hb_glyph_info_t *info) 565{ 566 info->glyph_props() &= ~(HB_OT_LAYOUT_GLYPH_PROPS_SUBSTITUTED); 567} 568 569 570/* Allocation / deallocation. */ 571 572static inline void 573_hb_buffer_allocate_unicode_vars (hb_buffer_t *buffer) 574{ 575 HB_BUFFER_ALLOCATE_VAR (buffer, unicode_props); 576} 577 578static inline void 579_hb_buffer_deallocate_unicode_vars (hb_buffer_t *buffer) 580{ 581 HB_BUFFER_DEALLOCATE_VAR (buffer, unicode_props); 582} 583 584static inline void 585_hb_buffer_assert_unicode_vars (hb_buffer_t *buffer) 586{ 587 HB_BUFFER_ASSERT_VAR (buffer, unicode_props); 588} 589 590static inline void 591_hb_buffer_allocate_gsubgpos_vars (hb_buffer_t *buffer) 592{ 593 HB_BUFFER_ALLOCATE_VAR (buffer, glyph_props); 594 HB_BUFFER_ALLOCATE_VAR (buffer, lig_props); 595 HB_BUFFER_ALLOCATE_VAR (buffer, syllable); 596} 597 598static inline void 599_hb_buffer_deallocate_gsubgpos_vars (hb_buffer_t *buffer) 600{ 601 HB_BUFFER_DEALLOCATE_VAR (buffer, syllable); 602 HB_BUFFER_DEALLOCATE_VAR (buffer, lig_props); 603 HB_BUFFER_DEALLOCATE_VAR (buffer, glyph_props); 604} 605 606static inline void 607_hb_buffer_assert_gsubgpos_vars (hb_buffer_t *buffer) 608{ 609 HB_BUFFER_ASSERT_VAR (buffer, glyph_props); 610 HB_BUFFER_ASSERT_VAR (buffer, lig_props); 611 HB_BUFFER_ASSERT_VAR (buffer, syllable); 612} 613 614/* Make sure no one directly touches our props... */ 615#undef unicode_props0 616#undef unicode_props1 617#undef lig_props 618#undef glyph_props 619 620 621#endif /* HB_OT_LAYOUT_PRIVATE_HH */ 622