1 /* 2 * Copyright 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 #ifndef HB_OT_SHAPE_COMPLEX_INDIC_HH 28 #define HB_OT_SHAPE_COMPLEX_INDIC_HH 29 30 #include "hb.hh" 31 32 #include "hb-ot-shape-complex.hh" 33 34 35 /* buffer var allocations */ 36 #define indic_category() complex_var_u8_0() /* indic_category_t */ 37 #define indic_position() complex_var_u8_1() /* indic_position_t */ 38 39 40 #define INDIC_TABLE_ELEMENT_TYPE uint16_t 41 42 /* Cateories used in the OpenType spec: 43 * https://docs.microsoft.com/en-us/typography/script-development/devanagari 44 */ 45 /* Note: This enum is duplicated in the -machine.rl source file. 46 * Not sure how to avoid duplication. */ 47 enum indic_category_t { 48 OT_X = 0, 49 OT_C = 1, 50 OT_V = 2, 51 OT_N = 3, 52 OT_H = 4, 53 OT_ZWNJ = 5, 54 OT_ZWJ = 6, 55 OT_M = 7, 56 OT_SM = 8, 57 /* OT_VD = 9, UNUSED; we use OT_A instead. */ 58 OT_A = 10, 59 OT_PLACEHOLDER = 11, 60 OT_DOTTEDCIRCLE = 12, 61 OT_RS = 13, /* Register Shifter, used in Khmer OT spec. */ 62 OT_Coeng = 14, /* Khmer-style Virama. */ 63 OT_Repha = 15, /* Atomically-encoded logical or visual repha. */ 64 OT_Ra = 16, 65 OT_CM = 17, /* Consonant-Medial; Unused by Indic shaper. */ 66 OT_Symbol = 18, /* Avagraha, etc that take marks (SM,A,VD). */ 67 OT_CS = 19 68 }; 69 70 /* Note: 71 * 72 * We treat Vowels and placeholders as if they were consonants. This is safe because Vowels 73 * cannot happen in a consonant syllable. The plus side however is, we can call the 74 * consonant syllable logic from the vowel syllable function and get it all right! */ 75 #define CONSONANT_FLAGS (FLAG (OT_C) | FLAG (OT_CS) | FLAG (OT_Ra) | FLAG (OT_V) | FLAG (OT_PLACEHOLDER) | FLAG (OT_DOTTEDCIRCLE)) 76 #define JOINER_FLAGS (FLAG (OT_ZWJ) | FLAG (OT_ZWNJ)) 77 78 79 /* Visual positions in a syllable from left to right. */ 80 enum indic_position_t { 81 POS_START = 0, 82 83 POS_RA_TO_BECOME_REPH = 1, 84 POS_PRE_M = 2, 85 POS_PRE_C = 3, 86 87 POS_BASE_C = 4, 88 POS_AFTER_MAIN = 5, 89 90 POS_ABOVE_C = 6, 91 92 POS_BEFORE_SUB = 7, 93 POS_BELOW_C = 8, 94 POS_AFTER_SUB = 9, 95 96 POS_BEFORE_POST = 10, 97 POS_POST_C = 11, 98 POS_AFTER_POST = 12, 99 100 POS_FINAL_C = 13, 101 POS_SMVD = 14, 102 103 POS_END = 15 104 }; 105 106 /* Categories used in IndicSyllabicCategory.txt from UCD. */ 107 enum indic_syllabic_category_t { 108 INDIC_SYLLABIC_CATEGORY_OTHER = OT_X, 109 110 INDIC_SYLLABIC_CATEGORY_AVAGRAHA = OT_Symbol, 111 INDIC_SYLLABIC_CATEGORY_BINDU = OT_SM, 112 INDIC_SYLLABIC_CATEGORY_BRAHMI_JOINING_NUMBER = OT_PLACEHOLDER, /* Don't care. */ 113 INDIC_SYLLABIC_CATEGORY_CANTILLATION_MARK = OT_A, 114 INDIC_SYLLABIC_CATEGORY_CONSONANT = OT_C, 115 INDIC_SYLLABIC_CATEGORY_CONSONANT_DEAD = OT_C, 116 INDIC_SYLLABIC_CATEGORY_CONSONANT_FINAL = OT_CM, 117 INDIC_SYLLABIC_CATEGORY_CONSONANT_HEAD_LETTER = OT_C, 118 INDIC_SYLLABIC_CATEGORY_CONSONANT_KILLER = OT_M, /* U+17CD only. */ 119 INDIC_SYLLABIC_CATEGORY_CONSONANT_MEDIAL = OT_CM, 120 INDIC_SYLLABIC_CATEGORY_CONSONANT_PLACEHOLDER = OT_PLACEHOLDER, 121 INDIC_SYLLABIC_CATEGORY_CONSONANT_PRECEDING_REPHA = OT_Repha, 122 INDIC_SYLLABIC_CATEGORY_CONSONANT_PREFIXED = OT_X, /* Don't care. */ 123 INDIC_SYLLABIC_CATEGORY_CONSONANT_SUBJOINED = OT_CM, 124 INDIC_SYLLABIC_CATEGORY_CONSONANT_SUCCEEDING_REPHA = OT_CM, 125 INDIC_SYLLABIC_CATEGORY_CONSONANT_WITH_STACKER = OT_CS, 126 INDIC_SYLLABIC_CATEGORY_GEMINATION_MARK = OT_SM, /* https://github.com/harfbuzz/harfbuzz/issues/552 */ 127 INDIC_SYLLABIC_CATEGORY_INVISIBLE_STACKER = OT_Coeng, 128 INDIC_SYLLABIC_CATEGORY_JOINER = OT_ZWJ, 129 INDIC_SYLLABIC_CATEGORY_MODIFYING_LETTER = OT_X, 130 INDIC_SYLLABIC_CATEGORY_NON_JOINER = OT_ZWNJ, 131 INDIC_SYLLABIC_CATEGORY_NUKTA = OT_N, 132 INDIC_SYLLABIC_CATEGORY_NUMBER = OT_PLACEHOLDER, 133 INDIC_SYLLABIC_CATEGORY_NUMBER_JOINER = OT_PLACEHOLDER, /* Don't care. */ 134 INDIC_SYLLABIC_CATEGORY_PURE_KILLER = OT_M, /* Is like a vowel matra. */ 135 INDIC_SYLLABIC_CATEGORY_REGISTER_SHIFTER = OT_RS, 136 INDIC_SYLLABIC_CATEGORY_SYLLABLE_MODIFIER = OT_SM, 137 INDIC_SYLLABIC_CATEGORY_TONE_LETTER = OT_X, 138 INDIC_SYLLABIC_CATEGORY_TONE_MARK = OT_N, 139 INDIC_SYLLABIC_CATEGORY_VIRAMA = OT_H, 140 INDIC_SYLLABIC_CATEGORY_VISARGA = OT_SM, 141 INDIC_SYLLABIC_CATEGORY_VOWEL = OT_V, 142 INDIC_SYLLABIC_CATEGORY_VOWEL_DEPENDENT = OT_M, 143 INDIC_SYLLABIC_CATEGORY_VOWEL_INDEPENDENT = OT_V 144 }; 145 146 /* Categories used in IndicSMatraCategory.txt from UCD */ 147 enum indic_matra_category_t { 148 INDIC_MATRA_CATEGORY_NOT_APPLICABLE = POS_END, 149 150 INDIC_MATRA_CATEGORY_LEFT = POS_PRE_C, 151 INDIC_MATRA_CATEGORY_TOP = POS_ABOVE_C, 152 INDIC_MATRA_CATEGORY_BOTTOM = POS_BELOW_C, 153 INDIC_MATRA_CATEGORY_RIGHT = POS_POST_C, 154 155 /* These should resolve to the position of the last part of the split sequence. */ 156 INDIC_MATRA_CATEGORY_BOTTOM_AND_RIGHT = INDIC_MATRA_CATEGORY_RIGHT, 157 INDIC_MATRA_CATEGORY_LEFT_AND_RIGHT = INDIC_MATRA_CATEGORY_RIGHT, 158 INDIC_MATRA_CATEGORY_TOP_AND_BOTTOM = INDIC_MATRA_CATEGORY_BOTTOM, 159 INDIC_MATRA_CATEGORY_TOP_AND_BOTTOM_AND_RIGHT = INDIC_MATRA_CATEGORY_RIGHT, 160 INDIC_MATRA_CATEGORY_TOP_AND_LEFT = INDIC_MATRA_CATEGORY_TOP, 161 INDIC_MATRA_CATEGORY_TOP_AND_LEFT_AND_RIGHT = INDIC_MATRA_CATEGORY_RIGHT, 162 INDIC_MATRA_CATEGORY_TOP_AND_RIGHT = INDIC_MATRA_CATEGORY_RIGHT, 163 164 INDIC_MATRA_CATEGORY_OVERSTRUCK = POS_AFTER_MAIN, 165 INDIC_MATRA_CATEGORY_VISUAL_ORDER_LEFT = POS_PRE_M 166 }; 167 168 #define INDIC_COMBINE_CATEGORIES(S,M) \ 169 ( \ 170 ASSERT_STATIC_EXPR_ZERO (S < 255 && M < 255) + \ 171 ( S | \ 172 ( \ 173 ( \ 174 S == INDIC_SYLLABIC_CATEGORY_CONSONANT_MEDIAL || \ 175 S == INDIC_SYLLABIC_CATEGORY_GEMINATION_MARK || \ 176 S == INDIC_SYLLABIC_CATEGORY_REGISTER_SHIFTER || \ 177 S == INDIC_SYLLABIC_CATEGORY_CONSONANT_SUCCEEDING_REPHA || \ 178 S == INDIC_SYLLABIC_CATEGORY_VIRAMA || \ 179 S == INDIC_SYLLABIC_CATEGORY_VOWEL_DEPENDENT || \ 180 false \ 181 ? M : INDIC_MATRA_CATEGORY_NOT_APPLICABLE \ 182 ) << 8 \ 183 ) \ 184 ) \ 185 ) 186 187 HB_INTERNAL INDIC_TABLE_ELEMENT_TYPE 188 hb_indic_get_categories (hb_codepoint_t u); 189 190 191 static inline bool 192 is_one_of (const hb_glyph_info_t &info, unsigned int flags) 193 { 194 /* If it ligated, all bets are off. */ 195 if (_hb_glyph_info_ligated (&info)) return false; 196 return !!(FLAG_UNSAFE (info.indic_category()) & flags); 197 } 198 199 static inline bool 200 is_joiner (const hb_glyph_info_t &info) 201 { 202 return is_one_of (info, JOINER_FLAGS); 203 } 204 205 static inline bool 206 is_consonant (const hb_glyph_info_t &info) 207 { 208 return is_one_of (info, CONSONANT_FLAGS); 209 } 210 211 static inline bool 212 is_halant (const hb_glyph_info_t &info) 213 { 214 return is_one_of (info, FLAG (OT_H)); 215 } 216 217 #define IN_HALF_BLOCK(u, Base) (((u) & ~0x7Fu) == (Base)) 218 219 #define IS_DEVA(u) (IN_HALF_BLOCK (u, 0x0900u)) 220 #define IS_BENG(u) (IN_HALF_BLOCK (u, 0x0980u)) 221 #define IS_GURU(u) (IN_HALF_BLOCK (u, 0x0A00u)) 222 #define IS_GUJR(u) (IN_HALF_BLOCK (u, 0x0A80u)) 223 #define IS_ORYA(u) (IN_HALF_BLOCK (u, 0x0B00u)) 224 #define IS_TAML(u) (IN_HALF_BLOCK (u, 0x0B80u)) 225 #define IS_TELU(u) (IN_HALF_BLOCK (u, 0x0C00u)) 226 #define IS_KNDA(u) (IN_HALF_BLOCK (u, 0x0C80u)) 227 #define IS_MLYM(u) (IN_HALF_BLOCK (u, 0x0D00u)) 228 #define IS_SINH(u) (IN_HALF_BLOCK (u, 0x0D80u)) 229 230 231 #define MATRA_POS_LEFT(u) POS_PRE_M 232 #define MATRA_POS_RIGHT(u) ( \ 233 IS_DEVA(u) ? POS_AFTER_SUB : \ 234 IS_BENG(u) ? POS_AFTER_POST : \ 235 IS_GURU(u) ? POS_AFTER_POST : \ 236 IS_GUJR(u) ? POS_AFTER_POST : \ 237 IS_ORYA(u) ? POS_AFTER_POST : \ 238 IS_TAML(u) ? POS_AFTER_POST : \ 239 IS_TELU(u) ? (u <= 0x0C42u ? POS_BEFORE_SUB : POS_AFTER_SUB) : \ 240 IS_KNDA(u) ? (u < 0x0CC3u || u > 0xCD6u ? POS_BEFORE_SUB : POS_AFTER_SUB) : \ 241 IS_MLYM(u) ? POS_AFTER_POST : \ 242 IS_SINH(u) ? POS_AFTER_SUB : \ 243 /*default*/ POS_AFTER_SUB \ 244 ) 245 #define MATRA_POS_TOP(u) ( /* BENG and MLYM don't have top matras. */ \ 246 IS_DEVA(u) ? POS_AFTER_SUB : \ 247 IS_GURU(u) ? POS_AFTER_POST : /* Deviate from spec */ \ 248 IS_GUJR(u) ? POS_AFTER_SUB : \ 249 IS_ORYA(u) ? POS_AFTER_MAIN : \ 250 IS_TAML(u) ? POS_AFTER_SUB : \ 251 IS_TELU(u) ? POS_BEFORE_SUB : \ 252 IS_KNDA(u) ? POS_BEFORE_SUB : \ 253 IS_SINH(u) ? POS_AFTER_SUB : \ 254 /*default*/ POS_AFTER_SUB \ 255 ) 256 #define MATRA_POS_BOTTOM(u) ( \ 257 IS_DEVA(u) ? POS_AFTER_SUB : \ 258 IS_BENG(u) ? POS_AFTER_SUB : \ 259 IS_GURU(u) ? POS_AFTER_POST : \ 260 IS_GUJR(u) ? POS_AFTER_POST : \ 261 IS_ORYA(u) ? POS_AFTER_SUB : \ 262 IS_TAML(u) ? POS_AFTER_POST : \ 263 IS_TELU(u) ? POS_BEFORE_SUB : \ 264 IS_KNDA(u) ? POS_BEFORE_SUB : \ 265 IS_MLYM(u) ? POS_AFTER_POST : \ 266 IS_SINH(u) ? POS_AFTER_SUB : \ 267 /*default*/ POS_AFTER_SUB \ 268 ) 269 270 static inline indic_position_t 271 matra_position_indic (hb_codepoint_t u, indic_position_t side) 272 { 273 switch ((int) side) 274 { 275 case POS_PRE_C: return MATRA_POS_LEFT (u); 276 case POS_POST_C: return MATRA_POS_RIGHT (u); 277 case POS_ABOVE_C: return MATRA_POS_TOP (u); 278 case POS_BELOW_C: return MATRA_POS_BOTTOM (u); 279 }; 280 return side; 281 } 282 283 /* XXX 284 * This is a hack for now. We should move this data into the main Indic table. 285 * Or completely remove it and just check in the tables. 286 */ 287 static const hb_codepoint_t ra_chars[] = { 288 0x0930u, /* Devanagari */ 289 0x09B0u, /* Bengali */ 290 0x09F0u, /* Bengali */ 291 0x0A30u, /* Gurmukhi */ /* No Reph */ 292 0x0AB0u, /* Gujarati */ 293 0x0B30u, /* Oriya */ 294 0x0BB0u, /* Tamil */ /* No Reph */ 295 0x0C30u, /* Telugu */ /* Reph formed only with ZWJ */ 296 0x0CB0u, /* Kannada */ 297 0x0D30u, /* Malayalam */ /* No Reph, Logical Repha */ 298 299 0x0DBBu, /* Sinhala */ /* Reph formed only with ZWJ */ 300 301 0x179Au, /* Khmer */ 302 }; 303 304 static inline bool 305 is_ra (hb_codepoint_t u) 306 { 307 for (unsigned int i = 0; i < ARRAY_LENGTH (ra_chars); i++) 308 if (u == ra_chars[i]) 309 return true; 310 return false; 311 } 312 313 static inline void 314 set_indic_properties (hb_glyph_info_t &info) 315 { 316 hb_codepoint_t u = info.codepoint; 317 unsigned int type = hb_indic_get_categories (u); 318 indic_category_t cat = (indic_category_t) (type & 0x7Fu); 319 indic_position_t pos = (indic_position_t) (type >> 8); 320 321 322 /* 323 * Re-assign category 324 */ 325 326 /* The following act more like the Bindus. */ 327 if (unlikely (hb_in_range<hb_codepoint_t> (u, 0x0953u, 0x0954u))) 328 cat = OT_SM; 329 /* The following act like consonants. */ 330 else if (unlikely (hb_in_ranges<hb_codepoint_t> (u, 0x0A72u, 0x0A73u, 331 0x1CF5u, 0x1CF6u))) 332 cat = OT_C; 333 /* TODO: The following should only be allowed after a Visarga. 334 * For now, just treat them like regular tone marks. */ 335 else if (unlikely (hb_in_range<hb_codepoint_t> (u, 0x1CE2u, 0x1CE8u))) 336 cat = OT_A; 337 /* TODO: The following should only be allowed after some of 338 * the nasalization marks, maybe only for U+1CE9..U+1CF1. 339 * For now, just treat them like tone marks. */ 340 else if (unlikely (u == 0x1CEDu)) 341 cat = OT_A; 342 /* The following take marks in standalone clusters, similar to Avagraha. */ 343 else if (unlikely (hb_in_ranges<hb_codepoint_t> (u, 0xA8F2u, 0xA8F7u, 344 0x1CE9u, 0x1CECu, 345 0x1CEEu, 0x1CF1u))) 346 { 347 cat = OT_Symbol; 348 static_assert (((int) INDIC_SYLLABIC_CATEGORY_AVAGRAHA == OT_Symbol), ""); 349 } 350 else if (unlikely (u == 0x0A51u)) 351 { 352 /* https://github.com/harfbuzz/harfbuzz/issues/524 */ 353 cat = OT_M; 354 pos = POS_BELOW_C; 355 } 356 357 /* According to ScriptExtensions.txt, these Grantha marks may also be used in Tamil, 358 * so the Indic shaper needs to know their categories. */ 359 else if (unlikely (u == 0x11301u || u == 0x11303u)) cat = OT_SM; 360 else if (unlikely (u == 0x1133cu)) cat = OT_N; 361 362 else if (unlikely (u == 0x0AFBu)) cat = OT_N; /* https://github.com/harfbuzz/harfbuzz/issues/552 */ 363 364 else if (unlikely (u == 0x0980u)) cat = OT_PLACEHOLDER; /* https://github.com/harfbuzz/harfbuzz/issues/538 */ 365 else if (unlikely (u == 0x0C80u)) cat = OT_PLACEHOLDER; /* https://github.com/harfbuzz/harfbuzz/pull/623 */ 366 else if (unlikely (hb_in_range<hb_codepoint_t> (u, 0x2010u, 0x2011u))) 367 cat = OT_PLACEHOLDER; 368 else if (unlikely (u == 0x25CCu)) cat = OT_DOTTEDCIRCLE; 369 370 371 /* 372 * Re-assign position. 373 */ 374 375 if ((FLAG_UNSAFE (cat) & CONSONANT_FLAGS)) 376 { 377 pos = POS_BASE_C; 378 if (is_ra (u)) 379 cat = OT_Ra; 380 } 381 else if (cat == OT_M) 382 { 383 pos = matra_position_indic (u, pos); 384 } 385 else if ((FLAG_UNSAFE (cat) & (FLAG (OT_SM) /* | FLAG (OT_VD) */ | FLAG (OT_A) | FLAG (OT_Symbol)))) 386 { 387 pos = POS_SMVD; 388 } 389 390 if (unlikely (u == 0x0B01u)) pos = POS_BEFORE_SUB; /* Oriya Bindu is BeforeSub in the spec. */ 391 392 393 394 info.indic_category() = cat; 395 info.indic_position() = pos; 396 } 397 398 399 #endif /* HB_OT_SHAPE_COMPLEX_INDIC_HH */ 400
