2 * Copyright © 2007,2008,2009 Red Hat, Inc.
3 * Copyright © 2012,2013 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 * Red Hat Author(s): Behdad Esfahbod
26 * Google Author(s): Behdad Esfahbod
29 #ifndef HB_OT_LAYOUT_PRIVATE_HH
30 #define HB_OT_LAYOUT_PRIVATE_HH
32 #include "hb-private.hh"
34 #include "hb-font-private.hh"
35 #include "hb-buffer-private.hh"
36 #include "hb-set-private.hh"
37 #include "hb-open-type-private.hh"
40 /* Private API corresponding to hb-ot-layout.h: */
43 hb_ot_layout_table_find_feature (hb_face_t *face,
46 unsigned int *feature_index);
53 enum hb_ot_layout_glyph_props_flags_t
55 /* The following three match LookupFlags::Ignore* numbers. */
56 HB_OT_LAYOUT_GLYPH_PROPS_BASE_GLYPH = 0x02u,
57 HB_OT_LAYOUT_GLYPH_PROPS_LIGATURE = 0x04u,
58 HB_OT_LAYOUT_GLYPH_PROPS_MARK = 0x08u,
60 /* The following are used internally; not derived from GDEF. */
61 HB_OT_LAYOUT_GLYPH_PROPS_SUBSTITUTED = 0x10u,
62 HB_OT_LAYOUT_GLYPH_PROPS_LIGATED = 0x20u,
63 HB_OT_LAYOUT_GLYPH_PROPS_MULTIPLIED = 0x40u,
65 HB_OT_LAYOUT_GLYPH_PROPS_PRESERVE = HB_OT_LAYOUT_GLYPH_PROPS_SUBSTITUTED |
66 HB_OT_LAYOUT_GLYPH_PROPS_LIGATED |
67 HB_OT_LAYOUT_GLYPH_PROPS_MULTIPLIED
69 HB_MARK_AS_FLAG_T (hb_ot_layout_glyph_props_flags_t);
77 hb_ot_layout_lookup_would_substitute_fast (hb_face_t *face,
78 unsigned int lookup_index,
79 const hb_codepoint_t *glyphs,
80 unsigned int glyphs_length,
81 hb_bool_t zero_context);
84 /* Should be called before all the substitute_lookup's are done. */
86 hb_ot_layout_substitute_start (hb_font_t *font,
90 struct hb_ot_layout_lookup_accelerator_t;
93 struct hb_apply_context_t;
98 hb_ot_layout_substitute_lookup (OT::hb_apply_context_t *c,
99 const OT::SubstLookup &lookup,
100 const hb_ot_layout_lookup_accelerator_t &accel);
103 /* Should be called before all the position_lookup's are done. */
105 hb_ot_layout_position_start (hb_font_t *font,
106 hb_buffer_t *buffer);
108 /* Should be called after all the position_lookup's are done, to finish advances. */
110 hb_ot_layout_position_finish_advances (hb_font_t *font,
111 hb_buffer_t *buffer);
113 /* Should be called after hb_ot_layout_position_finish_advances, to finish offsets. */
115 hb_ot_layout_position_finish_offsets (hb_font_t *font,
116 hb_buffer_t *buffer);
133 struct hb_ot_layout_lookup_accelerator_t
135 template <typename TLookup>
136 inline void init (const TLookup &lookup)
139 lookup.add_coverage (&digest);
142 inline void fini (void)
146 inline bool may_have (hb_codepoint_t g) const {
147 return digest.may_have (g);
151 hb_set_digest_t digest;
154 struct hb_ot_layout_t
156 hb_blob_t *gdef_blob;
157 hb_blob_t *gsub_blob;
158 hb_blob_t *gpos_blob;
160 const struct OT::GDEF *gdef;
161 const struct OT::GSUB *gsub;
162 const struct OT::GPOS *gpos;
164 /* TODO Move the following out of this struct. */
165 OT::hb_lazy_table_loader_t<struct OT::MATH> math;
166 OT::hb_lazy_table_loader_t<struct OT::fvar> fvar;
167 OT::hb_lazy_table_loader_t<struct OT::avar> avar;
169 unsigned int gsub_lookup_count;
170 unsigned int gpos_lookup_count;
172 hb_ot_layout_lookup_accelerator_t *gsub_accels;
173 hb_ot_layout_lookup_accelerator_t *gpos_accels;
177 HB_INTERNAL hb_ot_layout_t *
178 _hb_ot_layout_create (hb_face_t *face);
181 _hb_ot_layout_destroy (hb_ot_layout_t *layout);
184 #define hb_ot_layout_from_face(face) ((hb_ot_layout_t *) face->shaper_data.ot)
188 * Buffer var routines.
191 /* buffer var allocations, used during the entire shaping process */
192 #define unicode_props() var2.u16[0]
194 /* buffer var allocations, used during the GSUB/GPOS processing */
195 #define glyph_props() var1.u16[0] /* GDEF glyph properties */
196 #define lig_props() var1.u8[2] /* GSUB/GPOS ligature tracking */
197 #define syllable() var1.u8[3] /* GSUB/GPOS shaping boundaries */
200 /* loop over syllables */
202 #define foreach_syllable(buffer, start, end) \
204 _count = buffer->len, \
205 start = 0, end = _count ? _next_syllable (buffer, 0) : 0; \
207 start = end, end = _next_syllable (buffer, start))
209 static inline unsigned int
210 _next_syllable (hb_buffer_t *buffer, unsigned int start)
212 hb_glyph_info_t *info = buffer->info;
213 unsigned int count = buffer->len;
215 unsigned int syllable = info[start].syllable();
216 while (++start < count && syllable == info[start].syllable())
226 * unicode_props() is a two-byte number. The low byte includes:
227 * - General_Category: 5 bits.
229 * * Is it Default_Ignorable(); we have a modified Default_Ignorable().
230 * * Whether it's one of the three Mongolian Free Variation Selectors.
231 * * One free bit right now.
233 * The high-byte has different meanings, switched by the Gen-Cat:
234 * - For Mn,Mc,Me: the modified Combining_Class.
235 * - For Cf: whether it's ZWJ, ZWNJ, or something else.
236 * - For Ws: index of which space character this is, if space fallback
237 * is needed, ie. we don't set this by default, only if asked to.
240 enum hb_unicode_props_flags_t {
241 UPROPS_MASK_GEN_CAT = 0x001Fu,
242 UPROPS_MASK_IGNORABLE = 0x0020u,
243 UPROPS_MASK_FVS = 0x0040u, /* MONGOLIAN FREE VARIATION SELECTOR 1..3 */
245 /* If GEN_CAT=FORMAT, top byte masks: */
246 UPROPS_MASK_Cf_ZWJ = 0x0100u,
247 UPROPS_MASK_Cf_ZWNJ = 0x0200u
249 HB_MARK_AS_FLAG_T (hb_unicode_props_flags_t);
252 _hb_glyph_info_set_unicode_props (hb_glyph_info_t *info, hb_buffer_t *buffer)
254 hb_unicode_funcs_t *unicode = buffer->unicode;
255 unsigned int u = info->codepoint;
256 unsigned int gen_cat = (unsigned int) unicode->general_category (u);
257 unsigned int props = gen_cat;
261 buffer->scratch_flags |= HB_BUFFER_SCRATCH_FLAG_HAS_NON_ASCII;
262 if (unlikely (unicode->is_default_ignorable (u)))
264 buffer->scratch_flags |= HB_BUFFER_SCRATCH_FLAG_HAS_DEFAULT_IGNORABLES;
265 props |= UPROPS_MASK_IGNORABLE;
266 if (u == 0x200Cu) props |= UPROPS_MASK_Cf_ZWNJ;
267 if (u == 0x200Du) props |= UPROPS_MASK_Cf_ZWJ;
268 /* Mongolian Free Variation Selectors need to be remembered
269 * because although we need to hide them like default-ignorables,
270 * they need to non-ignorable during shaping. This is similar to
271 * what we do for joiners in Indic-like shapers, but since the
272 * FVSes are GC=Mn, we have use a separate bit to remember them.
274 * https://github.com/behdad/harfbuzz/issues/234
276 if (unlikely (hb_in_range (u, 0x180Bu, 0x180Du))) props |= UPROPS_MASK_FVS;
278 else if (unlikely (HB_UNICODE_GENERAL_CATEGORY_IS_NON_ENCLOSING_MARK_OR_MODIFIER_SYMBOL (gen_cat)))
280 /* The above check is just an optimization to let in only things we need further
283 /* Only Mn and Mc can have non-zero ccc:
284 * http://www.unicode.org/policies/stability_policy.html#Property_Value
286 * Canonical_Combining_Class, General_Category
287 * All characters other than those with General_Category property values
288 * Spacing_Mark (Mc) and Nonspacing_Mark (Mn) have the Canonical_Combining_Class
293 * Also, all Mn's that are Default_Ignorable, have ccc=0, hence
296 props |= unicode->modified_combining_class (info->codepoint)<<8;
298 /* Recategorize emoji skin-tone modifiers as Unicode mark, so they
299 * behave correctly in non-native directionality. They originally
300 * are MODIFIER_SYMBOL. Fixes:
301 * https://github.com/behdad/harfbuzz/issues/169
303 if (unlikely (hb_in_range (u, 0x1F3FBu, 0x1F3FFu)))
305 props = gen_cat = HB_UNICODE_GENERAL_CATEGORY_ENCLOSING_MARK;
310 info->unicode_props() = props;
314 _hb_glyph_info_set_general_category (hb_glyph_info_t *info,
315 hb_unicode_general_category_t gen_cat)
317 /* Clears top-byte. */
318 info->unicode_props() = (unsigned int) gen_cat | (info->unicode_props() & (0xFF & ~UPROPS_MASK_GEN_CAT));
321 static inline hb_unicode_general_category_t
322 _hb_glyph_info_get_general_category (const hb_glyph_info_t *info)
324 return (hb_unicode_general_category_t) (info->unicode_props() & UPROPS_MASK_GEN_CAT);
328 _hb_glyph_info_is_unicode_mark (const hb_glyph_info_t *info)
330 return HB_UNICODE_GENERAL_CATEGORY_IS_MARK (info->unicode_props() & UPROPS_MASK_GEN_CAT);
333 _hb_glyph_info_set_modified_combining_class (hb_glyph_info_t *info,
334 unsigned int modified_class)
336 if (unlikely (!_hb_glyph_info_is_unicode_mark (info)))
338 info->unicode_props() = (modified_class<<8) | (info->unicode_props() & 0xFF);
340 static inline unsigned int
341 _hb_glyph_info_get_modified_combining_class (const hb_glyph_info_t *info)
343 return _hb_glyph_info_is_unicode_mark (info) ? info->unicode_props()>>8 : 0;
347 _hb_glyph_info_is_unicode_space (const hb_glyph_info_t *info)
349 return _hb_glyph_info_get_general_category (info) ==
350 HB_UNICODE_GENERAL_CATEGORY_SPACE_SEPARATOR;
353 _hb_glyph_info_set_unicode_space_fallback_type (hb_glyph_info_t *info, hb_unicode_funcs_t::space_t s)
355 if (unlikely (!_hb_glyph_info_is_unicode_space (info)))
357 info->unicode_props() = (((unsigned int) s)<<8) | (info->unicode_props() & 0xFF);
359 static inline hb_unicode_funcs_t::space_t
360 _hb_glyph_info_get_unicode_space_fallback_type (const hb_glyph_info_t *info)
362 return _hb_glyph_info_is_unicode_space (info) ?
363 (hb_unicode_funcs_t::space_t) (info->unicode_props()>>8) :
364 hb_unicode_funcs_t::NOT_SPACE;
367 static inline bool _hb_glyph_info_ligated (const hb_glyph_info_t *info);
369 static inline hb_bool_t
370 _hb_glyph_info_is_default_ignorable (const hb_glyph_info_t *info)
372 return (info->unicode_props() & UPROPS_MASK_IGNORABLE) &&
373 !_hb_glyph_info_ligated (info);
375 static inline hb_bool_t
376 _hb_glyph_info_is_default_ignorable_and_not_fvs (const hb_glyph_info_t *info)
378 return ((info->unicode_props() & (UPROPS_MASK_IGNORABLE|UPROPS_MASK_FVS))
379 == UPROPS_MASK_IGNORABLE) &&
380 !_hb_glyph_info_ligated (info);
384 _hb_glyph_info_is_unicode_format (const hb_glyph_info_t *info)
386 return _hb_glyph_info_get_general_category (info) ==
387 HB_UNICODE_GENERAL_CATEGORY_FORMAT;
389 static inline hb_bool_t
390 _hb_glyph_info_is_zwnj (const hb_glyph_info_t *info)
392 return _hb_glyph_info_is_unicode_format (info) && (info->unicode_props() & UPROPS_MASK_Cf_ZWNJ);
394 static inline hb_bool_t
395 _hb_glyph_info_is_zwj (const hb_glyph_info_t *info)
397 return _hb_glyph_info_is_unicode_format (info) && (info->unicode_props() & UPROPS_MASK_Cf_ZWJ);
399 static inline hb_bool_t
400 _hb_glyph_info_is_joiner (const hb_glyph_info_t *info)
402 return _hb_glyph_info_is_unicode_format (info) && (info->unicode_props() & (UPROPS_MASK_Cf_ZWNJ|UPROPS_MASK_Cf_ZWJ));
405 _hb_glyph_info_flip_joiners (hb_glyph_info_t *info)
407 if (!_hb_glyph_info_is_unicode_format (info))
409 info->unicode_props() ^= UPROPS_MASK_Cf_ZWNJ | UPROPS_MASK_Cf_ZWJ;
412 /* lig_props: aka lig_id / lig_comp
414 * When a ligature is formed:
416 * - The ligature glyph and any marks in between all the same newly allocated
418 * - The ligature glyph will get lig_num_comps set to the number of components
419 * - The marks get lig_comp > 0, reflecting which component of the ligature
420 * they were applied to.
421 * - This is used in GPOS to attach marks to the right component of a ligature
423 * - Note that when marks are ligated together, much of the above is skipped
424 * and the current lig_id reused.
426 * When a multiple-substitution is done:
428 * - All resulting glyphs will have lig_id = 0,
429 * - The resulting glyphs will have lig_comp = 0, 1, 2, ... respectively.
430 * - This is used in GPOS to attach marks to the first component of a
431 * multiple substitution in MarkBasePos.
433 * The numbers are also used in GPOS to do mark-to-mark positioning only
434 * to marks that belong to the same component of the same ligature.
438 _hb_glyph_info_clear_lig_props (hb_glyph_info_t *info)
440 info->lig_props() = 0;
443 #define IS_LIG_BASE 0x10
446 _hb_glyph_info_set_lig_props_for_ligature (hb_glyph_info_t *info,
448 unsigned int lig_num_comps)
450 info->lig_props() = (lig_id << 5) | IS_LIG_BASE | (lig_num_comps & 0x0F);
454 _hb_glyph_info_set_lig_props_for_mark (hb_glyph_info_t *info,
456 unsigned int lig_comp)
458 info->lig_props() = (lig_id << 5) | (lig_comp & 0x0F);
462 _hb_glyph_info_set_lig_props_for_component (hb_glyph_info_t *info, unsigned int comp)
464 _hb_glyph_info_set_lig_props_for_mark (info, 0, comp);
467 static inline unsigned int
468 _hb_glyph_info_get_lig_id (const hb_glyph_info_t *info)
470 return info->lig_props() >> 5;
474 _hb_glyph_info_ligated_internal (const hb_glyph_info_t *info)
476 return !!(info->lig_props() & IS_LIG_BASE);
479 static inline unsigned int
480 _hb_glyph_info_get_lig_comp (const hb_glyph_info_t *info)
482 if (_hb_glyph_info_ligated_internal (info))
485 return info->lig_props() & 0x0F;
488 static inline unsigned int
489 _hb_glyph_info_get_lig_num_comps (const hb_glyph_info_t *info)
491 if ((info->glyph_props() & HB_OT_LAYOUT_GLYPH_PROPS_LIGATURE) &&
492 _hb_glyph_info_ligated_internal (info))
493 return info->lig_props() & 0x0F;
498 static inline uint8_t
499 _hb_allocate_lig_id (hb_buffer_t *buffer) {
500 uint8_t lig_id = buffer->next_serial () & 0x07;
501 if (unlikely (!lig_id))
502 lig_id = _hb_allocate_lig_id (buffer); /* in case of overflow */
509 _hb_glyph_info_set_glyph_props (hb_glyph_info_t *info, unsigned int props)
511 info->glyph_props() = props;
514 static inline unsigned int
515 _hb_glyph_info_get_glyph_props (const hb_glyph_info_t *info)
517 return info->glyph_props();
521 _hb_glyph_info_is_base_glyph (const hb_glyph_info_t *info)
523 return !!(info->glyph_props() & HB_OT_LAYOUT_GLYPH_PROPS_BASE_GLYPH);
527 _hb_glyph_info_is_ligature (const hb_glyph_info_t *info)
529 return !!(info->glyph_props() & HB_OT_LAYOUT_GLYPH_PROPS_LIGATURE);
533 _hb_glyph_info_is_mark (const hb_glyph_info_t *info)
535 return !!(info->glyph_props() & HB_OT_LAYOUT_GLYPH_PROPS_MARK);
539 _hb_glyph_info_substituted (const hb_glyph_info_t *info)
541 return !!(info->glyph_props() & HB_OT_LAYOUT_GLYPH_PROPS_SUBSTITUTED);
545 _hb_glyph_info_ligated (const hb_glyph_info_t *info)
547 return !!(info->glyph_props() & HB_OT_LAYOUT_GLYPH_PROPS_LIGATED);
551 _hb_glyph_info_multiplied (const hb_glyph_info_t *info)
553 return !!(info->glyph_props() & HB_OT_LAYOUT_GLYPH_PROPS_MULTIPLIED);
557 _hb_glyph_info_ligated_and_didnt_multiply (const hb_glyph_info_t *info)
559 return _hb_glyph_info_ligated (info) && !_hb_glyph_info_multiplied (info);
563 _hb_glyph_info_clear_ligated_and_multiplied (hb_glyph_info_t *info)
565 info->glyph_props() &= ~(HB_OT_LAYOUT_GLYPH_PROPS_LIGATED |
566 HB_OT_LAYOUT_GLYPH_PROPS_MULTIPLIED);
570 _hb_glyph_info_clear_substituted (hb_glyph_info_t *info)
572 info->glyph_props() &= ~(HB_OT_LAYOUT_GLYPH_PROPS_SUBSTITUTED);
576 /* Allocation / deallocation. */
579 _hb_buffer_allocate_unicode_vars (hb_buffer_t *buffer)
581 HB_BUFFER_ALLOCATE_VAR (buffer, unicode_props);
585 _hb_buffer_deallocate_unicode_vars (hb_buffer_t *buffer)
587 HB_BUFFER_DEALLOCATE_VAR (buffer, unicode_props);
591 _hb_buffer_assert_unicode_vars (hb_buffer_t *buffer)
593 HB_BUFFER_ASSERT_VAR (buffer, unicode_props);
597 _hb_buffer_allocate_gsubgpos_vars (hb_buffer_t *buffer)
599 HB_BUFFER_ALLOCATE_VAR (buffer, glyph_props);
600 HB_BUFFER_ALLOCATE_VAR (buffer, lig_props);
601 HB_BUFFER_ALLOCATE_VAR (buffer, syllable);
605 _hb_buffer_deallocate_gsubgpos_vars (hb_buffer_t *buffer)
607 HB_BUFFER_DEALLOCATE_VAR (buffer, syllable);
608 HB_BUFFER_DEALLOCATE_VAR (buffer, lig_props);
609 HB_BUFFER_DEALLOCATE_VAR (buffer, glyph_props);
613 _hb_buffer_assert_gsubgpos_vars (hb_buffer_t *buffer)
615 HB_BUFFER_ASSERT_VAR (buffer, glyph_props);
616 HB_BUFFER_ASSERT_VAR (buffer, lig_props);
617 HB_BUFFER_ASSERT_VAR (buffer, syllable);
620 /* Make sure no one directly touches our props... */
621 #undef unicode_props0
622 #undef unicode_props1
626 #endif /* HB_OT_LAYOUT_PRIVATE_HH */