From f055442716ec7543ed156d4789955b19c11a5255 Mon Sep 17 00:00:00 2001 From: Behdad Esfahbod Date: Thu, 19 Jul 2012 16:20:21 -0400 Subject: [PATCH] [Indic] Lookup consonant position in the font Fixes most failures of Oriya, and improves others a bit. --- src/hb-ot-map-private.hh | 13 ++- src/hb-ot-shape-complex-indic-private.hh | 105 ----------------------- src/hb-ot-shape-complex-indic.cc | 63 ++++++++++---- 3 files changed, 54 insertions(+), 127 deletions(-) diff --git a/src/hb-ot-map-private.hh b/src/hb-ot-map-private.hh index 3811206d..d098c750 100644 --- a/src/hb-ot-map-private.hh +++ b/src/hb-ot-map-private.hh @@ -50,17 +50,22 @@ struct hb_ot_map_t inline hb_mask_t get_global_mask (void) const { return global_mask; } - inline hb_mask_t get_mask (hb_tag_t tag, unsigned int *shift = NULL) const { - const feature_map_t *map = features.bsearch (&tag); + inline hb_mask_t get_mask (hb_tag_t feature_tag, unsigned int *shift = NULL) const { + const feature_map_t *map = features.bsearch (&feature_tag); if (shift) *shift = map ? map->shift : 0; return map ? map->mask : 0; } - inline hb_mask_t get_1_mask (hb_tag_t tag) const { - const feature_map_t *map = features.bsearch (&tag); + inline hb_mask_t get_1_mask (hb_tag_t feature_tag) const { + const feature_map_t *map = features.bsearch (&feature_tag); return map ? map->_1_mask : 0; } + inline hb_mask_t get_feature_index (unsigned int table_index, hb_tag_t feature_tag) const { + const feature_map_t *map = features.bsearch (&feature_tag); + return map ? map->index[table_index] : HB_OT_LAYOUT_NO_FEATURE_INDEX; + } + inline hb_tag_t get_chosen_script (unsigned int table_index) const { return chosen_script[table_index]; } diff --git a/src/hb-ot-shape-complex-indic-private.hh b/src/hb-ot-shape-complex-indic-private.hh index e64213b8..bbf54840 100644 --- a/src/hb-ot-shape-complex-indic-private.hh +++ b/src/hb-ot-shape-complex-indic-private.hh @@ -164,111 +164,6 @@ enum indic_matra_category_t { #include "hb-ot-shape-complex-indic-table.hh" -/* XXX - * This is a hack for now. We should: - * 1. Move this data into the main Indic table, - * and/or - * 2. Probe font lookups to determine consonant positions. - */ -static const struct consonant_position_t { - hb_codepoint_t u; - indic_position_t position; -} consonant_positions[] = { - {0x0930, POS_BELOW_C}, - {0x09AC, POS_BELOW_C}, - {0x09AF, POS_POST_C}, - {0x09B0, POS_BELOW_C}, - {0x09F0, POS_BELOW_C}, - {0x0A2F, POS_POST_C}, - {0x0A30, POS_BELOW_C}, - {0x0A35, POS_BELOW_C}, - {0x0A39, POS_BELOW_C}, - {0x0AB0, POS_BELOW_C}, - {0x0B24, POS_BELOW_C}, - {0x0B28, POS_BELOW_C}, - {0x0B2C, POS_BELOW_C}, - {0x0B2D, POS_BELOW_C}, - {0x0B2E, POS_BELOW_C}, - {0x0B2F, POS_POST_C}, - {0x0B30, POS_BELOW_C}, - {0x0B32, POS_BELOW_C}, - {0x0B33, POS_BELOW_C}, - {0x0B5F, POS_POST_C}, - {0x0B71, POS_BELOW_C}, - {0x0C15, POS_BELOW_C}, - {0x0C16, POS_BELOW_C}, - {0x0C17, POS_BELOW_C}, - {0x0C18, POS_BELOW_C}, - {0x0C19, POS_BELOW_C}, - {0x0C1A, POS_BELOW_C}, - {0x0C1B, POS_BELOW_C}, - {0x0C1C, POS_BELOW_C}, - {0x0C1D, POS_BELOW_C}, - {0x0C1E, POS_BELOW_C}, - {0x0C1F, POS_BELOW_C}, - {0x0C20, POS_BELOW_C}, - {0x0C21, POS_BELOW_C}, - {0x0C22, POS_BELOW_C}, - {0x0C23, POS_BELOW_C}, - {0x0C24, POS_BELOW_C}, - {0x0C25, POS_BELOW_C}, - {0x0C26, POS_BELOW_C}, - {0x0C27, POS_BELOW_C}, - {0x0C28, POS_BELOW_C}, - {0x0C2A, POS_BELOW_C}, - {0x0C2B, POS_BELOW_C}, - {0x0C2C, POS_BELOW_C}, - {0x0C2D, POS_BELOW_C}, - {0x0C2E, POS_BELOW_C}, - {0x0C2F, POS_BELOW_C}, - {0x0C30, POS_BELOW_C}, - {0x0C32, POS_BELOW_C}, - {0x0C33, POS_BELOW_C}, - {0x0C35, POS_BELOW_C}, - {0x0C36, POS_BELOW_C}, - {0x0C37, POS_BELOW_C}, - {0x0C38, POS_BELOW_C}, - {0x0C39, POS_BELOW_C}, - {0x0C95, POS_BELOW_C}, - {0x0C96, POS_BELOW_C}, - {0x0C97, POS_BELOW_C}, - {0x0C98, POS_BELOW_C}, - {0x0C99, POS_BELOW_C}, - {0x0C9A, POS_BELOW_C}, - {0x0C9B, POS_BELOW_C}, - {0x0C9C, POS_BELOW_C}, - {0x0C9D, POS_BELOW_C}, - {0x0C9E, POS_BELOW_C}, - {0x0C9F, POS_BELOW_C}, - {0x0CA0, POS_BELOW_C}, - {0x0CA1, POS_BELOW_C}, - {0x0CA2, POS_BELOW_C}, - {0x0CA3, POS_BELOW_C}, - {0x0CA4, POS_BELOW_C}, - {0x0CA5, POS_BELOW_C}, - {0x0CA6, POS_BELOW_C}, - {0x0CA7, POS_BELOW_C}, - {0x0CA8, POS_BELOW_C}, - {0x0CAA, POS_BELOW_C}, - {0x0CAB, POS_BELOW_C}, - {0x0CAC, POS_BELOW_C}, - {0x0CAD, POS_BELOW_C}, - {0x0CAE, POS_BELOW_C}, - {0x0CAF, POS_BELOW_C}, - {0x0CB0, POS_BELOW_C}, - {0x0CB2, POS_BELOW_C}, - {0x0CB3, POS_BELOW_C}, - {0x0CB5, POS_BELOW_C}, - {0x0CB6, POS_BELOW_C}, - {0x0CB7, POS_BELOW_C}, - {0x0CB8, POS_BELOW_C}, - {0x0CB9, POS_BELOW_C}, - {0x0CDE, POS_BELOW_C}, - {0x0D2F, POS_POST_C}, - {0x0D30, POS_POST_C}, - {0x0D32, POS_BELOW_C}, - {0x0D35, POS_POST_C}, -}; /* XXX * This is a hack for now. We should move this data into the main Indic table. diff --git a/src/hb-ot-shape-complex-indic.cc b/src/hb-ot-shape-complex-indic.cc index 4254d734..ea40100b 100644 --- a/src/hb-ot-shape-complex-indic.cc +++ b/src/hb-ot-shape-complex-indic.cc @@ -87,21 +87,48 @@ compare_codepoint (const void *pa, const void *pb) return a < b ? -1 : a == b ? 0 : +1; } -static indic_position_t -consonant_position (hb_codepoint_t u) +static bool +would_substitute (hb_codepoint_t *glyphs, unsigned int glyphs_count, + hb_tag_t feature_tag, hb_ot_map_t *map, hb_face_t *face) { - consonant_position_t *record; - - /* Khmer does not have pre-base half forms. */ - if (0x1780 <= u && u <= 0x17FF) - return POS_BELOW_C; - - record = (consonant_position_t *) bsearch (&u, consonant_positions, - ARRAY_LENGTH (consonant_positions), - sizeof (consonant_positions[0]), - compare_codepoint); + unsigned int lookup_indices[32]; + unsigned int offset, len; + + offset = 0; + do { + len = ARRAY_LENGTH (lookup_indices); + hb_ot_layout_feature_get_lookup_indexes (face, HB_OT_TAG_GSUB, + map->get_feature_index (0/*GSUB*/, feature_tag), + offset, + &len, + lookup_indices); + + for (unsigned int i = 0; i < len; i++) + if (hb_ot_layout_would_substitute_lookup (face, glyphs, glyphs_count, lookup_indices[i])) + return true; + + offset += len; + } while (len == ARRAY_LENGTH (lookup_indices)); + + return false; +} - return record ? record->position : POS_BASE_C; +static indic_position_t +consonant_position (hb_codepoint_t u, hb_ot_map_t *map, hb_font_t *font) +{ + hb_codepoint_t virama = (u & ~0x007F) | 0x004D; + if ((u & ~0x007F) == 0x0D80) virama = 0x0DCA; /* Sinahla */ + if ((u & ~0x007F) == 0x1780) virama = 0x17D2; /* Khmer */ + hb_codepoint_t glyphs[2]; + + hb_font_get_glyph (font, virama, 0, &glyphs[0]); + hb_font_get_glyph (font, u, 0, &glyphs[1]); + + hb_face_t *face = hb_font_get_face (font); + if (would_substitute (glyphs, ARRAY_LENGTH (glyphs), HB_TAG('p','r','e','f'), map, face)) return POS_BELOW_C; + if (would_substitute (glyphs, ARRAY_LENGTH (glyphs), HB_TAG('b','l','w','f'), map, face)) return POS_BELOW_C; + if (would_substitute (glyphs, ARRAY_LENGTH (glyphs), HB_TAG('p','s','t','f'), map, face)) return POS_POST_C; + return POS_BASE_C; } #define MATRA_POS_LEFT(u) POS_PRE_M @@ -193,7 +220,7 @@ is_halant_or_coeng (const hb_glyph_info_t &info) } static inline void -set_indic_properties (hb_glyph_info_t &info) +set_indic_properties (hb_glyph_info_t &info, hb_ot_map_t *map, hb_font_t *font) { hb_codepoint_t u = info.codepoint; unsigned int type = get_indic_categories (u); @@ -247,7 +274,7 @@ set_indic_properties (hb_glyph_info_t &info) if ((FLAG (cat) & CONSONANT_FLAGS)) { - pos = consonant_position (u); + pos = consonant_position (u, map, font); if (is_ra (u)) cat = OT_Ra; } @@ -380,9 +407,9 @@ _hb_ot_shape_complex_normalization_preference_indic (void) void -_hb_ot_shape_complex_setup_masks_indic (hb_ot_map_t *map HB_UNUSED, +_hb_ot_shape_complex_setup_masks_indic (hb_ot_map_t *map, hb_buffer_t *buffer, - hb_font_t *font HB_UNUSED) + hb_font_t *font) { HB_BUFFER_ALLOCATE_VAR (buffer, indic_category); HB_BUFFER_ALLOCATE_VAR (buffer, indic_position); @@ -392,7 +419,7 @@ _hb_ot_shape_complex_setup_masks_indic (hb_ot_map_t *map HB_UNUSED, unsigned int count = buffer->len; for (unsigned int i = 0; i < count; i++) - set_indic_properties (buffer->info[i]); + set_indic_properties (buffer->info[i], map, font); } static int -- GitLab