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-digest-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_ot_apply_context_t;
98 hb_ot_layout_substitute_lookup (OT::hb_ot_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 fini 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 fini offsets. */
115 hb_ot_layout_position_finish_offsets (hb_font_t *font,
116 hb_buffer_t *buffer);
143 struct hb_ot_layout_lookup_accelerator_t
145 template <typename TLookup>
146 inline void init (const TLookup &lookup)
149 lookup.add_coverage (&digest);
152 inline void fini (void)
156 inline bool may_have (hb_codepoint_t g) const {
157 return digest.may_have (g);
161 hb_set_digest_t digest;
164 struct hb_ot_layout_t
166 hb_blob_t *gdef_blob;
167 hb_blob_t *gsub_blob;
168 hb_blob_t *gpos_blob;
170 const struct OT::GDEF *gdef;
171 const struct OT::GSUB *gsub;
172 const struct OT::GPOS *gpos;
174 /* TODO Move the following out of this struct. */
175 OT::hb_table_lazy_loader_t<struct OT::BASE> base;
176 OT::hb_table_lazy_loader_t<struct OT::MATH> math;
177 OT::hb_table_lazy_loader_t<struct OT::fvar> fvar;
178 OT::hb_table_lazy_loader_t<struct OT::avar> avar;
180 unsigned int gsub_lookup_count;
181 unsigned int gpos_lookup_count;
183 hb_ot_layout_lookup_accelerator_t *gsub_accels;
184 hb_ot_layout_lookup_accelerator_t *gpos_accels;
188 HB_INTERNAL hb_ot_layout_t *
189 _hb_ot_layout_create (hb_face_t *face);
192 _hb_ot_layout_destroy (hb_ot_layout_t *layout);
195 #define hb_ot_layout_from_face(face) ((hb_ot_layout_t *) face->shaper_data.ot)
199 * Buffer var routines.
202 /* buffer var allocations, used during the entire shaping process */
203 #define unicode_props() var2.u16[0]
205 /* buffer var allocations, used during the GSUB/GPOS processing */
206 #define glyph_props() var1.u16[0] /* GDEF glyph properties */
207 #define lig_props() var1.u8[2] /* GSUB/GPOS ligature tracking */
208 #define syllable() var1.u8[3] /* GSUB/GPOS shaping boundaries */
211 /* Loop over syllables. Based on foreach_cluster(). */
212 #define foreach_syllable(buffer, start, end) \
214 _count = buffer->len, \
215 start = 0, end = _count ? _next_syllable (buffer, 0) : 0; \
217 start = end, end = _next_syllable (buffer, start))
219 static inline unsigned int
220 _next_syllable (hb_buffer_t *buffer, unsigned int start)
222 hb_glyph_info_t *info = buffer->info;
223 unsigned int count = buffer->len;
225 unsigned int syllable = info[start].syllable();
226 while (++start < count && syllable == info[start].syllable())
236 * unicode_props() is a two-byte number. The low byte includes:
237 * - General_Category: 5 bits.
239 * * Is it Default_Ignorable(); we have a modified Default_Ignorable().
240 * * Whether it's one of the three Mongolian Free Variation Selectors,
241 * CGJ, or other characters that are hidden but should not be ignored
242 * like most other Default_Ignorable()s do during matching.
243 * * One free bit right now.
245 * The high-byte has different meanings, switched by the Gen-Cat:
246 * - For Mn,Mc,Me: the modified Combining_Class.
247 * - For Cf: whether it's ZWJ, ZWNJ, or something else.
248 * - For Ws: index of which space character this is, if space fallback
249 * is needed, ie. we don't set this by default, only if asked to.
252 enum hb_unicode_props_flags_t {
253 UPROPS_MASK_GEN_CAT = 0x001Fu,
254 UPROPS_MASK_IGNORABLE = 0x0020u,
255 UPROPS_MASK_HIDDEN = 0x0040u, /* MONGOLIAN FREE VARIATION SELECTOR 1..3,
256 * or TAG characters */
258 /* If GEN_CAT=FORMAT, top byte masks: */
259 UPROPS_MASK_Cf_ZWJ = 0x0100u,
260 UPROPS_MASK_Cf_ZWNJ = 0x0200u
262 HB_MARK_AS_FLAG_T (hb_unicode_props_flags_t);
265 _hb_glyph_info_set_unicode_props (hb_glyph_info_t *info, hb_buffer_t *buffer)
267 hb_unicode_funcs_t *unicode = buffer->unicode;
268 unsigned int u = info->codepoint;
269 unsigned int gen_cat = (unsigned int) unicode->general_category (u);
270 unsigned int props = gen_cat;
274 buffer->scratch_flags |= HB_BUFFER_SCRATCH_FLAG_HAS_NON_ASCII;
275 if (unlikely (unicode->is_default_ignorable (u)))
277 buffer->scratch_flags |= HB_BUFFER_SCRATCH_FLAG_HAS_DEFAULT_IGNORABLES;
278 props |= UPROPS_MASK_IGNORABLE;
279 if (u == 0x200Cu) props |= UPROPS_MASK_Cf_ZWNJ;
280 else if (u == 0x200Du) props |= UPROPS_MASK_Cf_ZWJ;
281 /* Mongolian Free Variation Selectors need to be remembered
282 * because although we need to hide them like default-ignorables,
283 * they need to non-ignorable during shaping. This is similar to
284 * what we do for joiners in Indic-like shapers, but since the
285 * FVSes are GC=Mn, we have use a separate bit to remember them.
287 * https://github.com/harfbuzz/harfbuzz/issues/234 */
288 else if (unlikely (hb_in_range (u, 0x180Bu, 0x180Du))) props |= UPROPS_MASK_HIDDEN;
289 /* TAG characters need similar treatment. Fixes:
290 * https://github.com/harfbuzz/harfbuzz/issues/463 */
291 else if (unlikely (hb_in_range (u, 0xE0020u, 0xE007Fu))) props |= UPROPS_MASK_HIDDEN;
292 /* COMBINING GRAPHEME JOINER should not be skipped; at least some times.
293 * https://github.com/harfbuzz/harfbuzz/issues/554 */
294 else if (unlikely (u == 0x034Fu))
296 buffer->scratch_flags |= HB_BUFFER_SCRATCH_FLAG_HAS_CGJ;
297 props |= UPROPS_MASK_HIDDEN;
300 else if (unlikely (HB_UNICODE_GENERAL_CATEGORY_IS_NON_ENCLOSING_MARK_OR_MODIFIER_SYMBOL (gen_cat)))
302 /* The above check is just an optimization to let in only things we need further
305 /* Only Mn and Mc can have non-zero ccc:
306 * https://unicode.org/policies/stability_policy.html#Property_Value
308 * Canonical_Combining_Class, General_Category
309 * All characters other than those with General_Category property values
310 * Spacing_Mark (Mc) and Nonspacing_Mark (Mn) have the Canonical_Combining_Class
315 * Also, all Mn's that are Default_Ignorable, have ccc=0, hence
318 props |= unicode->modified_combining_class (info->codepoint)<<8;
320 /* Recategorize emoji skin-tone modifiers as Unicode mark, so they
321 * behave correctly in non-native directionality. They originally
322 * are MODIFIER_SYMBOL. Fixes:
323 * https://github.com/harfbuzz/harfbuzz/issues/169
325 if (unlikely (hb_in_range (u, 0x1F3FBu, 0x1F3FFu)))
327 props = gen_cat = HB_UNICODE_GENERAL_CATEGORY_ENCLOSING_MARK;
332 info->unicode_props() = props;
336 _hb_glyph_info_set_general_category (hb_glyph_info_t *info,
337 hb_unicode_general_category_t gen_cat)
339 /* Clears top-byte. */
340 info->unicode_props() = (unsigned int) gen_cat | (info->unicode_props() & (0xFF & ~UPROPS_MASK_GEN_CAT));
343 static inline hb_unicode_general_category_t
344 _hb_glyph_info_get_general_category (const hb_glyph_info_t *info)
346 return (hb_unicode_general_category_t) (info->unicode_props() & UPROPS_MASK_GEN_CAT);
350 _hb_glyph_info_is_unicode_mark (const hb_glyph_info_t *info)
352 return HB_UNICODE_GENERAL_CATEGORY_IS_MARK (info->unicode_props() & UPROPS_MASK_GEN_CAT);
355 _hb_glyph_info_set_modified_combining_class (hb_glyph_info_t *info,
356 unsigned int modified_class)
358 if (unlikely (!_hb_glyph_info_is_unicode_mark (info)))
360 info->unicode_props() = (modified_class<<8) | (info->unicode_props() & 0xFF);
362 static inline unsigned int
363 _hb_glyph_info_get_modified_combining_class (const hb_glyph_info_t *info)
365 return _hb_glyph_info_is_unicode_mark (info) ? info->unicode_props()>>8 : 0;
369 /* Loop over grapheme. Based on foreach_cluster(). */
370 #define foreach_grapheme(buffer, start, end) \
372 _count = buffer->len, \
373 start = 0, end = _count ? _next_grapheme (buffer, 0) : 0; \
375 start = end, end = _next_grapheme (buffer, start))
377 static inline unsigned int
378 _next_grapheme (hb_buffer_t *buffer, unsigned int start)
380 hb_glyph_info_t *info = buffer->info;
381 unsigned int count = buffer->len;
383 while (++start < count && _hb_glyph_info_is_unicode_mark (&info[start]))
390 #define info_cc(info) (_hb_glyph_info_get_modified_combining_class (&(info)))
393 _hb_glyph_info_is_unicode_space (const hb_glyph_info_t *info)
395 return _hb_glyph_info_get_general_category (info) ==
396 HB_UNICODE_GENERAL_CATEGORY_SPACE_SEPARATOR;
399 _hb_glyph_info_set_unicode_space_fallback_type (hb_glyph_info_t *info, hb_unicode_funcs_t::space_t s)
401 if (unlikely (!_hb_glyph_info_is_unicode_space (info)))
403 info->unicode_props() = (((unsigned int) s)<<8) | (info->unicode_props() & 0xFF);
405 static inline hb_unicode_funcs_t::space_t
406 _hb_glyph_info_get_unicode_space_fallback_type (const hb_glyph_info_t *info)
408 return _hb_glyph_info_is_unicode_space (info) ?
409 (hb_unicode_funcs_t::space_t) (info->unicode_props()>>8) :
410 hb_unicode_funcs_t::NOT_SPACE;
413 static inline bool _hb_glyph_info_ligated (const hb_glyph_info_t *info);
415 static inline hb_bool_t
416 _hb_glyph_info_is_default_ignorable (const hb_glyph_info_t *info)
418 return (info->unicode_props() & UPROPS_MASK_IGNORABLE) &&
419 !_hb_glyph_info_ligated (info);
421 static inline hb_bool_t
422 _hb_glyph_info_is_default_ignorable_and_not_hidden (const hb_glyph_info_t *info)
424 return ((info->unicode_props() & (UPROPS_MASK_IGNORABLE|UPROPS_MASK_HIDDEN))
425 == UPROPS_MASK_IGNORABLE) &&
426 !_hb_glyph_info_ligated (info);
429 _hb_glyph_info_unhide (hb_glyph_info_t *info)
431 info->unicode_props() &= ~ UPROPS_MASK_HIDDEN;
435 _hb_glyph_info_is_unicode_format (const hb_glyph_info_t *info)
437 return _hb_glyph_info_get_general_category (info) ==
438 HB_UNICODE_GENERAL_CATEGORY_FORMAT;
440 static inline hb_bool_t
441 _hb_glyph_info_is_zwnj (const hb_glyph_info_t *info)
443 return _hb_glyph_info_is_unicode_format (info) && (info->unicode_props() & UPROPS_MASK_Cf_ZWNJ);
445 static inline hb_bool_t
446 _hb_glyph_info_is_zwj (const hb_glyph_info_t *info)
448 return _hb_glyph_info_is_unicode_format (info) && (info->unicode_props() & UPROPS_MASK_Cf_ZWJ);
450 static inline hb_bool_t
451 _hb_glyph_info_is_joiner (const hb_glyph_info_t *info)
453 return _hb_glyph_info_is_unicode_format (info) && (info->unicode_props() & (UPROPS_MASK_Cf_ZWNJ|UPROPS_MASK_Cf_ZWJ));
456 _hb_glyph_info_flip_joiners (hb_glyph_info_t *info)
458 if (!_hb_glyph_info_is_unicode_format (info))
460 info->unicode_props() ^= UPROPS_MASK_Cf_ZWNJ | UPROPS_MASK_Cf_ZWJ;
463 /* lig_props: aka lig_id / lig_comp
465 * When a ligature is formed:
467 * - The ligature glyph and any marks in between all the same newly allocated
469 * - The ligature glyph will get lig_num_comps set to the number of components
470 * - The marks get lig_comp > 0, reflecting which component of the ligature
471 * they were applied to.
472 * - This is used in GPOS to attach marks to the right component of a ligature
474 * - Note that when marks are ligated together, much of the above is skipped
475 * and the current lig_id reused.
477 * When a multiple-substitution is done:
479 * - All resulting glyphs will have lig_id = 0,
480 * - The resulting glyphs will have lig_comp = 0, 1, 2, ... respectively.
481 * - This is used in GPOS to attach marks to the first component of a
482 * multiple substitution in MarkBasePos.
484 * The numbers are also used in GPOS to do mark-to-mark positioning only
485 * to marks that belong to the same component of the same ligature.
489 _hb_glyph_info_clear_lig_props (hb_glyph_info_t *info)
491 info->lig_props() = 0;
494 #define IS_LIG_BASE 0x10
497 _hb_glyph_info_set_lig_props_for_ligature (hb_glyph_info_t *info,
499 unsigned int lig_num_comps)
501 info->lig_props() = (lig_id << 5) | IS_LIG_BASE | (lig_num_comps & 0x0F);
505 _hb_glyph_info_set_lig_props_for_mark (hb_glyph_info_t *info,
507 unsigned int lig_comp)
509 info->lig_props() = (lig_id << 5) | (lig_comp & 0x0F);
513 _hb_glyph_info_set_lig_props_for_component (hb_glyph_info_t *info, unsigned int comp)
515 _hb_glyph_info_set_lig_props_for_mark (info, 0, comp);
518 static inline unsigned int
519 _hb_glyph_info_get_lig_id (const hb_glyph_info_t *info)
521 return info->lig_props() >> 5;
525 _hb_glyph_info_ligated_internal (const hb_glyph_info_t *info)
527 return !!(info->lig_props() & IS_LIG_BASE);
530 static inline unsigned int
531 _hb_glyph_info_get_lig_comp (const hb_glyph_info_t *info)
533 if (_hb_glyph_info_ligated_internal (info))
536 return info->lig_props() & 0x0F;
539 static inline unsigned int
540 _hb_glyph_info_get_lig_num_comps (const hb_glyph_info_t *info)
542 if ((info->glyph_props() & HB_OT_LAYOUT_GLYPH_PROPS_LIGATURE) &&
543 _hb_glyph_info_ligated_internal (info))
544 return info->lig_props() & 0x0F;
549 static inline uint8_t
550 _hb_allocate_lig_id (hb_buffer_t *buffer) {
551 uint8_t lig_id = buffer->next_serial () & 0x07;
552 if (unlikely (!lig_id))
553 lig_id = _hb_allocate_lig_id (buffer); /* in case of overflow */
560 _hb_glyph_info_set_glyph_props (hb_glyph_info_t *info, unsigned int props)
562 info->glyph_props() = props;
565 static inline unsigned int
566 _hb_glyph_info_get_glyph_props (const hb_glyph_info_t *info)
568 return info->glyph_props();
572 _hb_glyph_info_is_base_glyph (const hb_glyph_info_t *info)
574 return !!(info->glyph_props() & HB_OT_LAYOUT_GLYPH_PROPS_BASE_GLYPH);
578 _hb_glyph_info_is_ligature (const hb_glyph_info_t *info)
580 return !!(info->glyph_props() & HB_OT_LAYOUT_GLYPH_PROPS_LIGATURE);
584 _hb_glyph_info_is_mark (const hb_glyph_info_t *info)
586 return !!(info->glyph_props() & HB_OT_LAYOUT_GLYPH_PROPS_MARK);
590 _hb_glyph_info_substituted (const hb_glyph_info_t *info)
592 return !!(info->glyph_props() & HB_OT_LAYOUT_GLYPH_PROPS_SUBSTITUTED);
596 _hb_glyph_info_ligated (const hb_glyph_info_t *info)
598 return !!(info->glyph_props() & HB_OT_LAYOUT_GLYPH_PROPS_LIGATED);
602 _hb_glyph_info_multiplied (const hb_glyph_info_t *info)
604 return !!(info->glyph_props() & HB_OT_LAYOUT_GLYPH_PROPS_MULTIPLIED);
608 _hb_glyph_info_ligated_and_didnt_multiply (const hb_glyph_info_t *info)
610 return _hb_glyph_info_ligated (info) && !_hb_glyph_info_multiplied (info);
614 _hb_glyph_info_clear_ligated_and_multiplied (hb_glyph_info_t *info)
616 info->glyph_props() &= ~(HB_OT_LAYOUT_GLYPH_PROPS_LIGATED |
617 HB_OT_LAYOUT_GLYPH_PROPS_MULTIPLIED);
621 _hb_glyph_info_clear_substituted (hb_glyph_info_t *info)
623 info->glyph_props() &= ~(HB_OT_LAYOUT_GLYPH_PROPS_SUBSTITUTED);
627 /* Allocation / deallocation. */
630 _hb_buffer_allocate_unicode_vars (hb_buffer_t *buffer)
632 HB_BUFFER_ALLOCATE_VAR (buffer, unicode_props);
636 _hb_buffer_deallocate_unicode_vars (hb_buffer_t *buffer)
638 HB_BUFFER_DEALLOCATE_VAR (buffer, unicode_props);
642 _hb_buffer_assert_unicode_vars (hb_buffer_t *buffer)
644 HB_BUFFER_ASSERT_VAR (buffer, unicode_props);
648 _hb_buffer_allocate_gsubgpos_vars (hb_buffer_t *buffer)
650 HB_BUFFER_ALLOCATE_VAR (buffer, glyph_props);
651 HB_BUFFER_ALLOCATE_VAR (buffer, lig_props);
652 HB_BUFFER_ALLOCATE_VAR (buffer, syllable);
656 _hb_buffer_deallocate_gsubgpos_vars (hb_buffer_t *buffer)
658 HB_BUFFER_DEALLOCATE_VAR (buffer, syllable);
659 HB_BUFFER_DEALLOCATE_VAR (buffer, lig_props);
660 HB_BUFFER_DEALLOCATE_VAR (buffer, glyph_props);
664 _hb_buffer_assert_gsubgpos_vars (hb_buffer_t *buffer)
666 HB_BUFFER_ASSERT_VAR (buffer, glyph_props);
667 HB_BUFFER_ASSERT_VAR (buffer, lig_props);
668 HB_BUFFER_ASSERT_VAR (buffer, syllable);
671 /* Make sure no one directly touches our props... */
672 #undef unicode_props0
673 #undef unicode_props1
677 #endif /* HB_OT_LAYOUT_PRIVATE_HH */