2 * Copyright © 2010,2012 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"
31 /* buffer var allocations */
32 #define arabic_shaping_action() complex_var_u8_0() /* arabic shaping action */
36 * Bits used in the joining tables
42 JOINING_TYPE_C = JOINING_TYPE_D,
43 JOINING_GROUP_ALAPH = 3,
44 JOINING_GROUP_DALATH_RISH = 4,
45 NUM_STATE_MACHINE_COLS = 5,
47 /* We deliberately don't have a JOINING_TYPE_L since that's unused in Unicode. */
50 JOINING_TYPE_X = 7 /* means: use general-category to choose between U or T. */
57 #include "hb-ot-shape-complex-arabic-table.hh"
59 static unsigned int get_joining_type (hb_codepoint_t u, hb_unicode_general_category_t gen_cat)
61 if (likely (hb_in_range<hb_codepoint_t> (u, JOINING_TABLE_FIRST, JOINING_TABLE_LAST))) {
62 unsigned int j_type = joining_table[u - JOINING_TABLE_FIRST];
63 if (likely (j_type != JOINING_TYPE_X))
67 /* Mongolian joining data is not in ArabicJoining.txt yet. */
68 if (unlikely (hb_in_range<hb_codepoint_t> (u, 0x1800, 0x18AF)))
70 if (unlikely (hb_in_range<hb_codepoint_t> (u, 0x1880, 0x1886)))
71 return JOINING_TYPE_U;
73 /* All letters, SIBE SYLLABLE BOUNDARY MARKER, and NIRUGU are D */
74 if ((FLAG(gen_cat) & (FLAG (HB_UNICODE_GENERAL_CATEGORY_OTHER_LETTER) |
75 FLAG (HB_UNICODE_GENERAL_CATEGORY_MODIFIER_LETTER)))
76 || u == 0x1807 || u == 0x180A)
77 return JOINING_TYPE_D;
80 /* 'Phags-pa joining data is not in ArabicJoining.txt yet. */
81 if (unlikely (hb_in_range<hb_codepoint_t> (u, 0xA840, 0xA872)))
83 if (unlikely (u == 0xA872))
84 /* XXX Looks like this should be TYPE_L, but we don't support that yet! */
85 return JOINING_TYPE_R;
87 return JOINING_TYPE_D;
90 if (unlikely (hb_in_range<hb_codepoint_t> (u, 0x200C, 0x200D)))
92 return u == 0x200C ? JOINING_TYPE_U : JOINING_TYPE_C;
95 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))) ?
96 JOINING_TYPE_T : JOINING_TYPE_U;
99 static const hb_tag_t arabic_features[] =
101 HB_TAG('i','n','i','t'),
102 HB_TAG('m','e','d','i'),
103 HB_TAG('f','i','n','a'),
104 HB_TAG('i','s','o','l'),
106 HB_TAG('m','e','d','2'),
107 HB_TAG('f','i','n','2'),
108 HB_TAG('f','i','n','3'),
113 /* Same order as the feature array */
127 ARABIC_NUM_FEATURES = NONE
130 static const struct arabic_state_table_entry {
134 } arabic_state_table[][NUM_STATE_MACHINE_COLS] =
136 /* jt_U, jt_R, jt_D, jg_ALAPH, jg_DALATH_RISH */
138 /* State 0: prev was U, not willing to join. */
139 { {NONE,NONE,0}, {NONE,ISOL,1}, {NONE,ISOL,2}, {NONE,ISOL,1}, {NONE,ISOL,6}, },
141 /* State 1: prev was R or ISOL/ALAPH, not willing to join. */
142 { {NONE,NONE,0}, {NONE,ISOL,1}, {NONE,ISOL,2}, {NONE,FIN2,5}, {NONE,ISOL,6}, },
144 /* State 2: prev was D/ISOL, willing to join. */
145 { {NONE,NONE,0}, {INIT,FINA,1}, {INIT,FINA,3}, {INIT,FINA,4}, {INIT,FINA,6}, },
147 /* State 3: prev was D/FINA, willing to join. */
148 { {NONE,NONE,0}, {MEDI,FINA,1}, {MEDI,FINA,3}, {MEDI,FINA,4}, {MEDI,FINA,6}, },
150 /* State 4: prev was FINA ALAPH, not willing to join. */
151 { {NONE,NONE,0}, {MED2,ISOL,1}, {MED2,ISOL,2}, {MED2,FIN2,5}, {MED2,ISOL,6}, },
153 /* State 5: prev was FIN2/FIN3 ALAPH, not willing to join. */
154 { {NONE,NONE,0}, {ISOL,ISOL,1}, {ISOL,ISOL,2}, {ISOL,FIN2,5}, {ISOL,ISOL,6}, },
156 /* State 6: prev was DALATH/RISH, not willing to join. */
157 { {NONE,NONE,0}, {NONE,ISOL,1}, {NONE,ISOL,2}, {NONE,FIN3,5}, {NONE,ISOL,6}, }
162 arabic_fallback_shape (const hb_ot_shape_plan_t *plan,
164 hb_buffer_t *buffer);
167 collect_features_arabic (hb_ot_shape_planner_t *plan)
169 hb_ot_map_builder_t *map = &plan->map;
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);
186 for (unsigned int i = 0; i < ARABIC_NUM_FEATURES; i++)
187 map->add_bool_feature (arabic_features[i], false, i < 4); /* The first four features have fallback. */
189 map->add_gsub_pause (NULL);
191 map->add_bool_feature (HB_TAG('r','l','i','g'), true, true);
192 map->add_gsub_pause (arabic_fallback_shape);
194 map->add_bool_feature (HB_TAG('c','a','l','t'));
195 map->add_gsub_pause (NULL);
197 map->add_bool_feature (HB_TAG('c','s','w','h'));
198 map->add_bool_feature (HB_TAG('d','l','i','g'));
199 map->add_bool_feature (HB_TAG('m','s','e','t'));
202 #include "hb-ot-shape-complex-arabic-fallback.hh"
204 struct arabic_shape_plan_t
208 /* The "+ 1" in the next array is to accommodate for the "NONE" command,
209 * which is not an OpenType feature, but this simplifies the code by not
210 * having to do a "if (... < NONE) ..." and just rely on the fact that
211 * mask_array[NONE] == 0. */
212 hb_mask_t mask_array[ARABIC_NUM_FEATURES + 1];
215 arabic_fallback_plan_t *fallback_plan;
219 data_create_arabic (const hb_ot_shape_plan_t *plan)
221 arabic_shape_plan_t *arabic_plan = (arabic_shape_plan_t *) calloc (1, sizeof (arabic_shape_plan_t));
222 if (unlikely (!arabic_plan))
225 arabic_plan->do_fallback = plan->props.script == HB_SCRIPT_ARABIC;
226 for (unsigned int i = 0; i < ARABIC_NUM_FEATURES; i++) {
227 arabic_plan->mask_array[i] = plan->map.get_1_mask (arabic_features[i]);
229 arabic_plan->do_fallback = arabic_plan->do_fallback && plan->map.needs_fallback (arabic_features[i]);
236 data_destroy_arabic (void *data)
238 arabic_shape_plan_t *arabic_plan = (arabic_shape_plan_t *) data;
240 arabic_fallback_plan_destroy (arabic_plan->fallback_plan);
246 arabic_joining (hb_buffer_t *buffer)
248 unsigned int count = buffer->len;
249 unsigned int prev = (unsigned int) -1, state = 0;
251 HB_BUFFER_ALLOCATE_VAR (buffer, arabic_shaping_action);
253 /* Check pre-context */
254 if (!(buffer->flags & HB_BUFFER_FLAG_BOT))
255 for (unsigned int i = 0; i < buffer->context_len[0]; i++)
257 unsigned int this_type = get_joining_type (buffer->context[0][i], buffer->unicode->general_category (buffer->context[0][i]));
259 if (unlikely (this_type == JOINING_TYPE_T))
262 const arabic_state_table_entry *entry = &arabic_state_table[state][this_type];
263 state = entry->next_state;
267 for (unsigned int i = 0; i < count; i++)
269 unsigned int this_type = get_joining_type (buffer->info[i].codepoint, _hb_glyph_info_get_general_category (&buffer->info[i]));
271 if (unlikely (this_type == JOINING_TYPE_T)) {
272 buffer->info[i].arabic_shaping_action() = NONE;
276 const arabic_state_table_entry *entry = &arabic_state_table[state][this_type];
278 if (entry->prev_action != NONE && prev != (unsigned int) -1)
279 buffer->info[prev].arabic_shaping_action() = entry->prev_action;
281 buffer->info[i].arabic_shaping_action() = entry->curr_action;
284 state = entry->next_state;
287 if (!(buffer->flags & HB_BUFFER_FLAG_EOT))
288 for (unsigned int i = 0; i < buffer->context_len[1]; i++)
290 unsigned int this_type = get_joining_type (buffer->context[1][i], buffer->unicode->general_category (buffer->context[1][i]));
292 if (unlikely (this_type == JOINING_TYPE_T))
295 const arabic_state_table_entry *entry = &arabic_state_table[state][this_type];
296 if (entry->prev_action != NONE && prev != (unsigned int) -1)
297 buffer->info[prev].arabic_shaping_action() = entry->prev_action;
302 HB_BUFFER_DEALLOCATE_VAR (buffer, arabic_shaping_action);
306 setup_masks_arabic (const hb_ot_shape_plan_t *plan,
308 hb_font_t *font HB_UNUSED)
310 const arabic_shape_plan_t *arabic_plan = (const arabic_shape_plan_t *) plan->data;
312 arabic_joining (buffer);
313 unsigned int count = buffer->len;
314 for (unsigned int i = 0; i < count; i++)
315 buffer->info[i].mask |= arabic_plan->mask_array[buffer->info[i].arabic_shaping_action()];
320 arabic_fallback_shape (const hb_ot_shape_plan_t *plan,
324 const arabic_shape_plan_t *arabic_plan = (const arabic_shape_plan_t *) plan->data;
326 if (!arabic_plan->do_fallback)
330 arabic_fallback_plan_t *fallback_plan = (arabic_fallback_plan_t *) hb_atomic_ptr_get (&arabic_plan->fallback_plan);
331 if (unlikely (!fallback_plan))
333 /* This sucks. We need a font to build the fallback plan... */
334 fallback_plan = arabic_fallback_plan_create (plan, font);
335 if (unlikely (!hb_atomic_ptr_cmpexch (&(const_cast<arabic_shape_plan_t *> (arabic_plan))->fallback_plan, NULL, fallback_plan))) {
336 arabic_fallback_plan_destroy (fallback_plan);
341 arabic_fallback_plan_shape (fallback_plan, font, buffer);
345 const hb_ot_complex_shaper_t _hb_ot_complex_shaper_arabic =
348 collect_features_arabic,
349 NULL, /* override_features */
352 NULL, /* preprocess_text_arabic */
353 NULL, /* normalization_preference */
354 NULL, /* decompose */
357 true, /* zero_width_attached_marks */
358 true, /* fallback_position */