| 1 | /* |
| 2 | * Copyright © 2011,2012 Google, Inc. |
| 3 | * |
| 4 | * This is part of HarfBuzz, a text shaping library. |
| 5 | * |
| 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. |
| 11 | * |
| 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 |
| 16 | * DAMAGE. |
| 17 | * |
| 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. |
| 23 | * |
| 24 | * Google Author(s): Behdad Esfahbod |
| 25 | */ |
| 26 | |
| 27 | #include "hb.hh" |
| 28 | |
| 29 | #ifndef HB_NO_OT_SHAPE |
| 30 | |
| 31 | #include "hb-ot-shaper-khmer-machine.hh" |
| 32 | #include "hb-ot-shaper-indic.hh" |
| 33 | #include "hb-ot-layout.hh" |
| 34 | |
| 35 | |
| 36 | /* |
| 37 | * Khmer shaper. |
| 38 | */ |
| 39 | |
| 40 | |
| 41 | static const hb_ot_map_feature_t |
| 42 | khmer_features[] = |
| 43 | { |
| 44 | /* |
| 45 | * Basic features. |
| 46 | * These features are applied all at once, before reordering, constrained |
| 47 | * to the syllable. |
| 48 | */ |
| 49 | {HB_TAG('p','r','e','f'), F_MANUAL_JOINERS | F_PER_SYLLABLE}, |
| 50 | {HB_TAG('b','l','w','f'), F_MANUAL_JOINERS | F_PER_SYLLABLE}, |
| 51 | {HB_TAG('a','b','v','f'), F_MANUAL_JOINERS | F_PER_SYLLABLE}, |
| 52 | {HB_TAG('p','s','t','f'), F_MANUAL_JOINERS | F_PER_SYLLABLE}, |
| 53 | {HB_TAG('c','f','a','r'), F_MANUAL_JOINERS | F_PER_SYLLABLE}, |
| 54 | /* |
| 55 | * Other features. |
| 56 | * These features are applied all at once after clearing syllables. |
| 57 | */ |
| 58 | {HB_TAG('p','r','e','s'), F_GLOBAL_MANUAL_JOINERS}, |
| 59 | {HB_TAG('a','b','v','s'), F_GLOBAL_MANUAL_JOINERS}, |
| 60 | {HB_TAG('b','l','w','s'), F_GLOBAL_MANUAL_JOINERS}, |
| 61 | {HB_TAG('p','s','t','s'), F_GLOBAL_MANUAL_JOINERS}, |
| 62 | }; |
| 63 | |
| 64 | /* |
| 65 | * Must be in the same order as the khmer_features array. |
| 66 | */ |
| 67 | enum { |
| 68 | KHMER_PREF, |
| 69 | KHMER_BLWF, |
| 70 | KHMER_ABVF, |
| 71 | KHMER_PSTF, |
| 72 | KHMER_CFAR, |
| 73 | |
| 74 | _KHMER_PRES, |
| 75 | _KHMER_ABVS, |
| 76 | _KHMER_BLWS, |
| 77 | _KHMER_PSTS, |
| 78 | |
| 79 | KHMER_NUM_FEATURES, |
| 80 | KHMER_BASIC_FEATURES = _KHMER_PRES, /* Don't forget to update this! */ |
| 81 | }; |
| 82 | |
| 83 | static inline void |
| 84 | set_khmer_properties (hb_glyph_info_t &info) |
| 85 | { |
| 86 | hb_codepoint_t u = info.codepoint; |
| 87 | unsigned int type = hb_indic_get_categories (u); |
| 88 | |
| 89 | info.khmer_category() = (khmer_category_t) (type & 0xFFu); |
| 90 | } |
| 91 | |
| 92 | static bool |
| 93 | setup_syllables_khmer (const hb_ot_shape_plan_t *plan, |
| 94 | hb_font_t *font, |
| 95 | hb_buffer_t *buffer); |
| 96 | static bool |
| 97 | reorder_khmer (const hb_ot_shape_plan_t *plan, |
| 98 | hb_font_t *font, |
| 99 | hb_buffer_t *buffer); |
| 100 | |
| 101 | static void |
| 102 | collect_features_khmer (hb_ot_shape_planner_t *plan) |
| 103 | { |
| 104 | hb_ot_map_builder_t *map = &plan->map; |
| 105 | |
| 106 | /* Do this before any lookups have been applied. */ |
| 107 | map->add_gsub_pause (setup_syllables_khmer); |
| 108 | map->add_gsub_pause (reorder_khmer); |
| 109 | |
| 110 | /* Testing suggests that Uniscribe does NOT pause between basic |
| 111 | * features. Test with KhmerUI.ttf and the following three |
| 112 | * sequences: |
| 113 | * |
| 114 | * U+1789,U+17BC |
| 115 | * U+1789,U+17D2,U+1789 |
| 116 | * U+1789,U+17D2,U+1789,U+17BC |
| 117 | * |
| 118 | * https://github.com/harfbuzz/harfbuzz/issues/974 |
| 119 | */ |
| 120 | map->enable_feature (HB_TAG('l','o','c','l'), F_PER_SYLLABLE); |
| 121 | map->enable_feature (HB_TAG('c','c','m','p'), F_PER_SYLLABLE); |
| 122 | |
| 123 | unsigned int i = 0; |
| 124 | for (; i < KHMER_BASIC_FEATURES; i++) |
| 125 | map->add_feature (khmer_features[i]); |
| 126 | |
| 127 | /* https://github.com/harfbuzz/harfbuzz/issues/3531 */ |
| 128 | map->add_gsub_pause (hb_syllabic_clear_var); // Don't need syllables anymore, use stop to free buffer var |
| 129 | |
| 130 | for (; i < KHMER_NUM_FEATURES; i++) |
| 131 | map->add_feature (khmer_features[i]); |
| 132 | } |
| 133 | |
| 134 | static void |
| 135 | override_features_khmer (hb_ot_shape_planner_t *plan) |
| 136 | { |
| 137 | hb_ot_map_builder_t *map = &plan->map; |
| 138 | |
| 139 | /* Khmer spec has 'clig' as part of required shaping features: |
| 140 | * "Apply feature 'clig' to form ligatures that are desired for |
| 141 | * typographical correctness.", hence in overrides... */ |
| 142 | map->enable_feature (HB_TAG('c','l','i','g')); |
| 143 | |
| 144 | /* Uniscribe does not apply 'kern' in Khmer. */ |
| 145 | if (hb_options ().uniscribe_bug_compatible) |
| 146 | { |
| 147 | map->disable_feature (HB_TAG('k','e','r','n')); |
| 148 | } |
| 149 | |
| 150 | map->disable_feature (HB_TAG('l','i','g','a')); |
| 151 | } |
| 152 | |
| 153 | |
| 154 | struct khmer_shape_plan_t |
| 155 | { |
| 156 | hb_mask_t mask_array[KHMER_NUM_FEATURES]; |
| 157 | }; |
| 158 | |
| 159 | static void * |
| 160 | data_create_khmer (const hb_ot_shape_plan_t *plan) |
| 161 | { |
| 162 | khmer_shape_plan_t *khmer_plan = (khmer_shape_plan_t *) hb_calloc (1, sizeof (khmer_shape_plan_t)); |
| 163 | if (unlikely (!khmer_plan)) |
| 164 | return nullptr; |
| 165 | |
| 166 | for (unsigned int i = 0; i < ARRAY_LENGTH (khmer_plan->mask_array); i++) |
| 167 | khmer_plan->mask_array[i] = (khmer_features[i].flags & F_GLOBAL) ? |
| 168 | 0 : plan->map.get_1_mask (khmer_features[i].tag); |
| 169 | |
| 170 | return khmer_plan; |
| 171 | } |
| 172 | |
| 173 | static void |
| 174 | data_destroy_khmer (void *data) |
| 175 | { |
| 176 | hb_free (data); |
| 177 | } |
| 178 | |
| 179 | static void |
| 180 | setup_masks_khmer (const hb_ot_shape_plan_t *plan HB_UNUSED, |
| 181 | hb_buffer_t *buffer, |
| 182 | hb_font_t *font HB_UNUSED) |
| 183 | { |
| 184 | HB_BUFFER_ALLOCATE_VAR (buffer, khmer_category); |
| 185 | |
| 186 | /* We cannot setup masks here. We save information about characters |
| 187 | * and setup masks later on in a pause-callback. */ |
| 188 | |
| 189 | unsigned int count = buffer->len; |
| 190 | hb_glyph_info_t *info = buffer->info; |
| 191 | for (unsigned int i = 0; i < count; i++) |
| 192 | set_khmer_properties (info[i]); |
| 193 | } |
| 194 | |
| 195 | static bool |
| 196 | setup_syllables_khmer (const hb_ot_shape_plan_t *plan HB_UNUSED, |
| 197 | hb_font_t *font HB_UNUSED, |
| 198 | hb_buffer_t *buffer) |
| 199 | { |
| 200 | HB_BUFFER_ALLOCATE_VAR (buffer, syllable); |
| 201 | find_syllables_khmer (buffer); |
| 202 | foreach_syllable (buffer, start, end) |
| 203 | buffer->unsafe_to_break (start, end); |
| 204 | return false; |
| 205 | } |
| 206 | |
| 207 | |
| 208 | /* Rules from: |
| 209 | * https://docs.microsoft.com/en-us/typography/script-development/devanagari */ |
| 210 | |
| 211 | static void |
| 212 | reorder_consonant_syllable (const hb_ot_shape_plan_t *plan, |
| 213 | hb_face_t *face HB_UNUSED, |
| 214 | hb_buffer_t *buffer, |
| 215 | unsigned int start, unsigned int end) |
| 216 | { |
| 217 | const khmer_shape_plan_t *khmer_plan = (const khmer_shape_plan_t *) plan->data; |
| 218 | hb_glyph_info_t *info = buffer->info; |
| 219 | |
| 220 | /* Setup masks. */ |
| 221 | { |
| 222 | /* Post-base */ |
| 223 | hb_mask_t mask = khmer_plan->mask_array[KHMER_BLWF] | |
| 224 | khmer_plan->mask_array[KHMER_ABVF] | |
| 225 | khmer_plan->mask_array[KHMER_PSTF]; |
| 226 | for (unsigned int i = start + 1; i < end; i++) |
| 227 | info[i].mask |= mask; |
| 228 | } |
| 229 | |
| 230 | unsigned int num_coengs = 0; |
| 231 | for (unsigned int i = start + 1; i < end; i++) |
| 232 | { |
| 233 | /* """ |
| 234 | * When a COENG + (Cons | IndV) combination are found (and subscript count |
| 235 | * is less than two) the character combination is handled according to the |
| 236 | * subscript type of the character following the COENG. |
| 237 | * |
| 238 | * ... |
| 239 | * |
| 240 | * Subscript Type 2 - The COENG + RO characters are reordered to immediately |
| 241 | * before the base glyph. Then the COENG + RO characters are assigned to have |
| 242 | * the 'pref' OpenType feature applied to them. |
| 243 | * """ |
| 244 | */ |
| 245 | if (info[i].khmer_category() == K_Cat(H) && num_coengs <= 2 && i + 1 < end) |
| 246 | { |
| 247 | num_coengs++; |
| 248 | |
| 249 | if (info[i + 1].khmer_category() == K_Cat(Ra)) |
| 250 | { |
| 251 | for (unsigned int j = 0; j < 2; j++) |
| 252 | info[i + j].mask |= khmer_plan->mask_array[KHMER_PREF]; |
| 253 | |
| 254 | /* Move the Coeng,Ro sequence to the start. */ |
| 255 | buffer->merge_clusters (start, i + 2); |
| 256 | hb_glyph_info_t t0 = info[i]; |
| 257 | hb_glyph_info_t t1 = info[i + 1]; |
| 258 | memmove (&info[start + 2], &info[start], (i - start) * sizeof (info[0])); |
| 259 | info[start] = t0; |
| 260 | info[start + 1] = t1; |
| 261 | |
| 262 | /* Mark the subsequent stuff with 'cfar'. Used in Khmer. |
| 263 | * Read the feature spec. |
| 264 | * This allows distinguishing the following cases with MS Khmer fonts: |
| 265 | * U+1784,U+17D2,U+179A,U+17D2,U+1782 |
| 266 | * U+1784,U+17D2,U+1782,U+17D2,U+179A |
| 267 | */ |
| 268 | if (khmer_plan->mask_array[KHMER_CFAR]) |
| 269 | for (unsigned int j = i + 2; j < end; j++) |
| 270 | info[j].mask |= khmer_plan->mask_array[KHMER_CFAR]; |
| 271 | |
| 272 | num_coengs = 2; /* Done. */ |
| 273 | } |
| 274 | } |
| 275 | |
| 276 | /* Reorder left matra piece. */ |
| 277 | else if (info[i].khmer_category() == K_Cat(VPre)) |
| 278 | { |
| 279 | /* Move to the start. */ |
| 280 | buffer->merge_clusters (start, i + 1); |
| 281 | hb_glyph_info_t t = info[i]; |
| 282 | memmove (&info[start + 1], &info[start], (i - start) * sizeof (info[0])); |
| 283 | info[start] = t; |
| 284 | } |
| 285 | } |
| 286 | } |
| 287 | |
| 288 | static void |
| 289 | reorder_syllable_khmer (const hb_ot_shape_plan_t *plan, |
| 290 | hb_face_t *face, |
| 291 | hb_buffer_t *buffer, |
| 292 | unsigned int start, unsigned int end) |
| 293 | { |
| 294 | khmer_syllable_type_t syllable_type = (khmer_syllable_type_t) (buffer->info[start].syllable() & 0x0F); |
| 295 | switch (syllable_type) |
| 296 | { |
| 297 | case khmer_broken_cluster: /* We already inserted dotted-circles, so just call the consonant_syllable. */ |
| 298 | case khmer_consonant_syllable: |
| 299 | reorder_consonant_syllable (plan, face, buffer, start, end); |
| 300 | break; |
| 301 | |
| 302 | case khmer_non_khmer_cluster: |
| 303 | break; |
| 304 | } |
| 305 | } |
| 306 | |
| 307 | static bool |
| 308 | reorder_khmer (const hb_ot_shape_plan_t *plan, |
| 309 | hb_font_t *font, |
| 310 | hb_buffer_t *buffer) |
| 311 | { |
| 312 | bool ret = false; |
| 313 | if (buffer->message (font, "start reordering khmer" )) |
| 314 | { |
| 315 | if (hb_syllabic_insert_dotted_circles (font, buffer, |
| 316 | khmer_broken_cluster, |
| 317 | K_Cat(DOTTEDCIRCLE), |
| 318 | (unsigned) -1)) |
| 319 | ret = true; |
| 320 | |
| 321 | foreach_syllable (buffer, start, end) |
| 322 | reorder_syllable_khmer (plan, font->face, buffer, start, end); |
| 323 | (void) buffer->message (font, "end reordering khmer" ); |
| 324 | } |
| 325 | HB_BUFFER_DEALLOCATE_VAR (buffer, khmer_category); |
| 326 | |
| 327 | return ret; |
| 328 | } |
| 329 | |
| 330 | |
| 331 | static bool |
| 332 | decompose_khmer (const hb_ot_shape_normalize_context_t *c, |
| 333 | hb_codepoint_t ab, |
| 334 | hb_codepoint_t *a, |
| 335 | hb_codepoint_t *b) |
| 336 | { |
| 337 | switch (ab) |
| 338 | { |
| 339 | /* |
| 340 | * Decompose split matras that don't have Unicode decompositions. |
| 341 | */ |
| 342 | |
| 343 | /* Khmer */ |
| 344 | case 0x17BEu : *a = 0x17C1u; *b= 0x17BEu; return true; |
| 345 | case 0x17BFu : *a = 0x17C1u; *b= 0x17BFu; return true; |
| 346 | case 0x17C0u : *a = 0x17C1u; *b= 0x17C0u; return true; |
| 347 | case 0x17C4u : *a = 0x17C1u; *b= 0x17C4u; return true; |
| 348 | case 0x17C5u : *a = 0x17C1u; *b= 0x17C5u; return true; |
| 349 | } |
| 350 | |
| 351 | return (bool) c->unicode->decompose (ab, a, b); |
| 352 | } |
| 353 | |
| 354 | static bool |
| 355 | compose_khmer (const hb_ot_shape_normalize_context_t *c, |
| 356 | hb_codepoint_t a, |
| 357 | hb_codepoint_t b, |
| 358 | hb_codepoint_t *ab) |
| 359 | { |
| 360 | /* Avoid recomposing split matras. */ |
| 361 | if (HB_UNICODE_GENERAL_CATEGORY_IS_MARK (c->unicode->general_category (a))) |
| 362 | return false; |
| 363 | |
| 364 | return (bool) c->unicode->compose (a, b, ab); |
| 365 | } |
| 366 | |
| 367 | |
| 368 | const hb_ot_shaper_t _hb_ot_shaper_khmer = |
| 369 | { |
| 370 | collect_features_khmer, |
| 371 | override_features_khmer, |
| 372 | data_create_khmer, |
| 373 | data_destroy_khmer, |
| 374 | nullptr, /* preprocess_text */ |
| 375 | nullptr, /* postprocess_glyphs */ |
| 376 | decompose_khmer, |
| 377 | compose_khmer, |
| 378 | setup_masks_khmer, |
| 379 | nullptr, /* reorder_marks */ |
| 380 | HB_TAG_NONE, /* gpos_tag */ |
| 381 | HB_OT_SHAPE_NORMALIZATION_MODE_COMPOSED_DIACRITICS_NO_SHORT_CIRCUIT, |
| 382 | HB_OT_SHAPE_ZERO_WIDTH_MARKS_NONE, |
| 383 | false, /* fallback_position */ |
| 384 | }; |
| 385 | |
| 386 | |
| 387 | #endif |
| 388 | |