2 * Copyright © 2010 Google, Inc.
4 * This is part of HarfBuzz, a text shaping library.
6 * Permission is hereby granted, without written agreement and without
7 * license or royalty fees, to use, copy, modify, and distribute this
8 * software and its documentation for any purpose, provided that the
9 * above copyright notice and the following two paragraphs appear in
10 * all copies of this software.
12 * IN NO EVENT SHALL THE COPYRIGHT HOLDER BE LIABLE TO ANY PARTY FOR
13 * DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES
14 * ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS DOCUMENTATION, EVEN
15 * IF THE COPYRIGHT HOLDER HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH
18 * THE COPYRIGHT HOLDER SPECIFICALLY DISCLAIMS ANY WARRANTIES, INCLUDING,
19 * BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
20 * FITNESS FOR A PARTICULAR PURPOSE. THE SOFTWARE PROVIDED HEREUNDER IS
21 * ON AN "AS IS" BASIS, AND THE COPYRIGHT HOLDER HAS NO OBLIGATION TO
22 * PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
24 * Google Author(s): Behdad Esfahbod
27 #include "hb-ot-shape-complex-private.hh"
28 #include "hb-ot-shape-private.hh"
32 /* buffer var allocations */
33 #define arabic_shaping_action() complex_var_temporary_u8() /* arabic shaping action */
37 * Bits used in the joining tables
43 JOINING_TYPE_C = JOINING_TYPE_D,
44 JOINING_GROUP_ALAPH = 3,
45 JOINING_GROUP_DALATH_RISH = 4,
46 NUM_STATE_MACHINE_COLS = 5,
48 /* We deliberately don't have a JOINING_TYPE_L since that's unused in Unicode. */
51 JOINING_TYPE_X = 7 /* means: use general-category to choose between U or T. */
58 #include "hb-ot-shape-complex-arabic-table.hh"
60 static unsigned int get_joining_type (hb_codepoint_t u, hb_unicode_general_category_t gen_cat)
62 if (likely (hb_in_range<hb_codepoint_t> (u, JOINING_TABLE_FIRST, JOINING_TABLE_LAST))) {
63 unsigned int j_type = joining_table[u - JOINING_TABLE_FIRST];
64 if (likely (j_type != JOINING_TYPE_X))
68 /* Mongolian joining data is not in ArabicJoining.txt yet */
69 if (unlikely (hb_in_range<hb_codepoint_t> (u, 0x1800, 0x18AF)))
71 /* All letters, SIBE SYLLABLE BOUNDARY MARKER, and NIRUGU are D */
72 if (gen_cat == HB_UNICODE_GENERAL_CATEGORY_OTHER_LETTER || u == 0x1807 || u == 0x180A)
73 return JOINING_TYPE_D;
76 if (unlikely (hb_in_range<hb_codepoint_t> (u, 0x200C, 0x200D))) {
77 return u == 0x200C ? JOINING_TYPE_U : JOINING_TYPE_C;
80 return (FLAG(gen_cat) & (FLAG(HB_UNICODE_GENERAL_CATEGORY_NON_SPACING_MARK) | FLAG(HB_UNICODE_GENERAL_CATEGORY_ENCLOSING_MARK) | FLAG(HB_UNICODE_GENERAL_CATEGORY_FORMAT))) ?
81 JOINING_TYPE_T : JOINING_TYPE_U;
84 static hb_codepoint_t get_arabic_shape (hb_codepoint_t u, unsigned int shape)
86 if (likely (hb_in_range<hb_codepoint_t> (u, SHAPING_TABLE_FIRST, SHAPING_TABLE_LAST)) && shape < 4)
87 return shaping_table[u - SHAPING_TABLE_FIRST][shape];
91 static uint16_t get_ligature (hb_codepoint_t first, hb_codepoint_t second)
93 if (unlikely (!second)) return 0;
94 for (unsigned i = 0; i < ARRAY_LENGTH (ligature_table); i++)
95 if (ligature_table[i].first == first)
96 for (unsigned j = 0; j < ARRAY_LENGTH (ligature_table[i].ligatures); j++)
97 if (ligature_table[i].ligatures[j].second == second)
98 return ligature_table[i].ligatures[j].ligature;
102 static const hb_tag_t arabic_syriac_features[] =
104 HB_TAG('i','n','i','t'),
105 HB_TAG('m','e','d','i'),
106 HB_TAG('f','i','n','a'),
107 HB_TAG('i','s','o','l'),
109 HB_TAG('m','e','d','2'),
110 HB_TAG('f','i','n','2'),
111 HB_TAG('f','i','n','3'),
116 /* Same order as the feature array */
130 COMMON_NUM_FEATURES = 4,
131 SYRIAC_NUM_FEATURES = 7,
132 TOTAL_NUM_FEATURES = NONE
135 static const struct arabic_state_table_entry {
139 } arabic_state_table[][NUM_STATE_MACHINE_COLS] =
141 /* jt_U, jt_R, jt_D, jg_ALAPH, jg_DALATH_RISH */
143 /* State 0: prev was U, not willing to join. */
144 { {NONE,NONE,0}, {NONE,ISOL,1}, {NONE,ISOL,2}, {NONE,ISOL,1}, {NONE,ISOL,6}, },
146 /* State 1: prev was R or ISOL/ALAPH, not willing to join. */
147 { {NONE,NONE,0}, {NONE,ISOL,1}, {NONE,ISOL,2}, {NONE,FIN2,5}, {NONE,ISOL,6}, },
149 /* State 2: prev was D/ISOL, willing to join. */
150 { {NONE,NONE,0}, {INIT,FINA,1}, {INIT,FINA,3}, {INIT,FINA,4}, {INIT,FINA,6}, },
152 /* State 3: prev was D/FINA, willing to join. */
153 { {NONE,NONE,0}, {MEDI,FINA,1}, {MEDI,FINA,3}, {MEDI,FINA,4}, {MEDI,FINA,6}, },
155 /* State 4: prev was FINA ALAPH, not willing to join. */
156 { {NONE,NONE,0}, {MED2,ISOL,1}, {MED2,ISOL,2}, {MED2,FIN2,5}, {MED2,ISOL,6}, },
158 /* State 5: prev was FIN2/FIN3 ALAPH, not willing to join. */
159 { {NONE,NONE,0}, {ISOL,ISOL,1}, {ISOL,ISOL,2}, {ISOL,FIN2,5}, {ISOL,ISOL,6}, },
161 /* State 6: prev was DALATH/RISH, not willing to join. */
162 { {NONE,NONE,0}, {NONE,ISOL,1}, {NONE,ISOL,2}, {NONE,FIN3,5}, {NONE,ISOL,6}, }
168 _hb_ot_shape_complex_collect_features_arabic (hb_ot_map_builder_t *map,
169 const hb_segment_properties_t *props)
171 /* For Language forms (in ArabicOT speak), we do the iso/fina/medi/init together,
172 * then rlig and calt each in their own stage. This makes IranNastaliq's ALLAH
173 * ligature work correctly. It's unfortunate though...
175 * This also makes Arial Bold in Windows7 work. See:
176 * https://bugzilla.mozilla.org/show_bug.cgi?id=644184
178 * TODO: Add test cases for these two.
181 map->add_bool_feature (HB_TAG('c','c','m','p'));
182 map->add_bool_feature (HB_TAG('l','o','c','l'));
184 map->add_gsub_pause (NULL, NULL);
186 unsigned int num_features = props->script == HB_SCRIPT_SYRIAC ? SYRIAC_NUM_FEATURES : COMMON_NUM_FEATURES;
187 for (unsigned int i = 0; i < num_features; i++)
188 map->add_bool_feature (arabic_syriac_features[i], false);
190 map->add_gsub_pause (NULL, NULL);
192 map->add_bool_feature (HB_TAG('r','l','i','g'));
193 map->add_gsub_pause (NULL, NULL);
195 map->add_bool_feature (HB_TAG('c','a','l','t'));
196 map->add_gsub_pause (NULL, NULL);
198 /* ArabicOT spec enables 'cswh' for Arabic where as for basic shaper it's disabled by default. */
199 map->add_bool_feature (HB_TAG('c','s','w','h'));
202 hb_ot_shape_normalization_mode_t
203 _hb_ot_shape_complex_normalization_preference_arabic (void)
205 return HB_OT_SHAPE_NORMALIZATION_MODE_COMPOSED_DIACRITICS;
210 arabic_fallback_shape (hb_font_t *font, hb_buffer_t *buffer)
212 unsigned int count = buffer->len;
213 hb_codepoint_t glyph;
215 /* Shape to presentation forms */
216 for (unsigned int i = 0; i < count; i++) {
217 hb_codepoint_t u = buffer->info[i].codepoint;
218 hb_codepoint_t shaped = get_arabic_shape (u, buffer->info[i].arabic_shaping_action());
219 if (shaped != u && hb_font_get_glyph (font, shaped, 0, &glyph))
220 buffer->info[i].codepoint = shaped;
223 /* Mandatory ligatures */
224 buffer->clear_output ();
225 for (buffer->idx = 0; buffer->idx + 1 < count;) {
226 hb_codepoint_t ligature = get_ligature (buffer->cur().codepoint,
227 buffer->cur(+1).codepoint);
228 if (likely (!ligature) || !(hb_font_get_glyph (font, ligature, 0, &glyph))) {
229 buffer->next_glyph ();
233 buffer->replace_glyphs (2, 1, &ligature);
235 /* Technically speaking we can skip marks and stuff, like the GSUB path does.
236 * But who cares, we're in fallback! */
238 for (; buffer->idx < count;)
239 buffer->next_glyph ();
240 buffer->swap_buffers ();
244 _hb_ot_shape_complex_setup_masks_arabic (hb_ot_map_t *map,
248 unsigned int count = buffer->len;
249 unsigned int prev = 0, state = 0;
251 HB_BUFFER_ALLOCATE_VAR (buffer, arabic_shaping_action);
253 for (unsigned int i = 0; i < count; i++)
255 unsigned int this_type = get_joining_type (buffer->info[i].codepoint, _hb_glyph_info_get_general_category (&buffer->info[i]));
257 if (unlikely (this_type == JOINING_TYPE_T)) {
258 buffer->info[i].arabic_shaping_action() = NONE;
262 const arabic_state_table_entry *entry = &arabic_state_table[state][this_type];
264 if (entry->prev_action != NONE)
265 buffer->info[prev].arabic_shaping_action() = entry->prev_action;
267 buffer->info[i].arabic_shaping_action() = entry->curr_action;
270 state = entry->next_state;
273 hb_mask_t mask_array[TOTAL_NUM_FEATURES + 1] = {0};
274 hb_mask_t total_masks = 0;
275 unsigned int num_masks = buffer->props.script == HB_SCRIPT_SYRIAC ? SYRIAC_NUM_FEATURES : COMMON_NUM_FEATURES;
276 for (unsigned int i = 0; i < num_masks; i++) {
277 mask_array[i] = map->get_1_mask (arabic_syriac_features[i]);
278 total_masks |= mask_array[i];
282 /* Has OpenType tables */
283 for (unsigned int i = 0; i < count; i++)
284 buffer->info[i].mask |= mask_array[buffer->info[i].arabic_shaping_action()];
285 } else if (buffer->props.script == HB_SCRIPT_ARABIC) {
286 /* Fallback Arabic shaping to Presentation Forms */
288 * - This path fires if user force-set init/medi/fina/isol off,
289 * - If font does not declare script 'arab', well, what to do?
290 * Most probably it's safe to assume that init/medi/fina/isol
291 * still mean Arabic shaping, although they do not have to.
293 arabic_fallback_shape (font, buffer);
296 HB_BUFFER_DEALLOCATE_VAR (buffer, arabic_shaping_action);