1 /* 2 * 3 * (C) Copyright IBM Corp. 1998-2007 - All Rights Reserved 4 * 5 * This file is a modification of the ICU file IndicReordering.cpp 6 * by Jens Herden and Javier Sola for Khmer language 7 * 8 */ 9 10 #include "LETypes.h" 11 #include "OpenTypeTables.h" 12 #include "KhmerReordering.h" 13 #include "LEGlyphStorage.h" 14 15 16 U_NAMESPACE_BEGIN 17 18 // Characters that get refered to by name... 19 enum 20 { 21 C_SIGN_ZWNJ = 0x200C, 22 C_SIGN_ZWJ = 0x200D, 23 C_DOTTED_CIRCLE = 0x25CC, 24 C_RO = 0x179A, 25 C_VOWEL_AA = 0x17B6, 26 C_SIGN_NIKAHIT = 0x17C6, 27 C_VOWEL_E = 0x17C1, 28 C_COENG = 0x17D2 29 }; 30 31 32 enum 33 { 34 // simple classes, they are used in the statetable (in this file) to control the length of a syllable 35 // they are also used to know where a character should be placed (location in reference to the base character) 36 // and also to know if a character, when independtly displayed, should be displayed with a dotted-circle to 37 // indicate error in syllable construction 38 _xx = KhmerClassTable::CC_RESERVED, 39 _sa = KhmerClassTable::CC_SIGN_ABOVE | KhmerClassTable::CF_DOTTED_CIRCLE | KhmerClassTable::CF_POS_ABOVE, 40 _sp = KhmerClassTable::CC_SIGN_AFTER | KhmerClassTable::CF_DOTTED_CIRCLE| KhmerClassTable::CF_POS_AFTER, 41 _c1 = KhmerClassTable::CC_CONSONANT | KhmerClassTable::CF_CONSONANT, 42 _c2 = KhmerClassTable::CC_CONSONANT2 | KhmerClassTable::CF_CONSONANT, 43 _c3 = KhmerClassTable::CC_CONSONANT3 | KhmerClassTable::CF_CONSONANT, 44 _rb = KhmerClassTable::CC_ROBAT | KhmerClassTable::CF_POS_ABOVE | KhmerClassTable::CF_DOTTED_CIRCLE, 45 _cs = KhmerClassTable::CC_CONSONANT_SHIFTER | KhmerClassTable::CF_DOTTED_CIRCLE | KhmerClassTable::CF_SHIFTER, 46 _dl = KhmerClassTable::CC_DEPENDENT_VOWEL | KhmerClassTable::CF_POS_BEFORE | KhmerClassTable::CF_DOTTED_CIRCLE, 47 _db = KhmerClassTable::CC_DEPENDENT_VOWEL | KhmerClassTable::CF_POS_BELOW | KhmerClassTable::CF_DOTTED_CIRCLE, 48 _da = KhmerClassTable::CC_DEPENDENT_VOWEL | KhmerClassTable::CF_POS_ABOVE | KhmerClassTable::CF_DOTTED_CIRCLE | KhmerClassTable::CF_ABOVE_VOWEL, 49 _dr = KhmerClassTable::CC_DEPENDENT_VOWEL | KhmerClassTable::CF_POS_AFTER | KhmerClassTable::CF_DOTTED_CIRCLE, 50 _co = KhmerClassTable::CC_COENG | KhmerClassTable::CF_COENG | KhmerClassTable::CF_DOTTED_CIRCLE, 51 52 // split vowel 53 _va = _da | KhmerClassTable::CF_SPLIT_VOWEL, 54 _vr = _dr | KhmerClassTable::CF_SPLIT_VOWEL 55 }; 56 57 58 // Character class tables 59 // _xx character does not combine into syllable, such as numbers, puntuation marks, non-Khmer signs... 60 // _sa Sign placed above the base 61 // _sp Sign placed after the base 62 // _c1 Consonant of type 1 or independent vowel (independent vowels behave as type 1 consonants) 63 // _c2 Consonant of type 2 (only RO) 64 // _c3 Consonant of type 3 65 // _rb Khmer sign robat u17CC. combining mark for subscript consonants 66 // _cd Consonant-shifter 67 // _dl Dependent vowel placed before the base (left of the base) 68 // _db Dependent vowel placed below the base 69 // _da Dependent vowel placed above the base 70 // _dr Dependent vowel placed behind the base (right of the base) 71 // _co Khmer combining mark COENG u17D2, combines with the consonant or independent vowel following 72 // it to create a subscript consonant or independent vowel 73 // _va Khmer split vowel in wich the first part is before the base and the second one above the base 74 // _vr Khmer split vowel in wich the first part is before the base and the second one behind (right of) the base 75 76 static const KhmerClassTable::CharClass khmerCharClasses[] = 77 { 78 _c1, _c1, _c1, _c3, _c1, _c1, _c1, _c1, _c3, _c1, _c1, _c1, _c1, _c3, _c1, _c1, // 1780 - 178F 79 _c1, _c1, _c1, _c1, _c3, _c1, _c1, _c1, _c1, _c3, _c2, _c1, _c1, _c1, _c3, _c3, // 1790 - 179F 80 _c1, _c3, _c1, _c1, _c1, _c1, _c1, _c1, _c1, _c1, _c1, _c1, _c1, _c1, _c1, _c1, // 17A0 - 17AF 81 _c1, _c1, _c1, _c1, _dr, _dr, _dr, _da, _da, _da, _da, _db, _db, _db, _va, _vr, // 17B0 - 17BF 82 _vr, _dl, _dl, _dl, _vr, _vr, _sa, _sp, _sp, _cs, _cs, _sa, _rb, _sa, _sa, _sa, // 17C0 - 17CF 83 _sa, _sa, _co, _sa, _xx, _xx, _xx, _xx, _xx, _xx, _xx, _xx, _xx, _sa, _xx, _xx, // 17D0 - 17DF 84 }; 85 86 87 // 88 // Khmer Class Tables 89 // 90 91 // 92 // The range of characters defined in the above table is defined here. FOr Khmer 1780 to 17DF 93 // Even if the Khmer range is bigger, all other characters are not combinable, and therefore treated 94 // as _xx 95 static const KhmerClassTable khmerClassTable = {0x1780, 0x17df, khmerCharClasses}; 96 97 98 // Below we define how a character in the input string is either in the khmerCharClasses table 99 // (in which case we get its type back), a ZWJ or ZWNJ (two characters that may appear 100 // within the syllable, but are not in the table) we also get their type back, or an unknown object 101 // in which case we get _xx (CC_RESERVED) back 102 KhmerClassTable::CharClass KhmerClassTable::getCharClass(LEUnicode ch) const 103 { 104 105 if (ch == C_SIGN_ZWJ) { 106 return CC_ZERO_WIDTH_J_MARK; 107 } 108 109 if (ch == C_SIGN_ZWNJ) { 110 return CC_ZERO_WIDTH_NJ_MARK; 111 } 112 113 if (ch < firstChar || ch > lastChar) { 114 return CC_RESERVED; 115 } 116 117 return classTable[ch - firstChar]; 118 } 119 120 const KhmerClassTable *KhmerClassTable::getKhmerClassTable() 121 { 122 return &khmerClassTable; 123 } 124 125 126 127 class KhmerReorderingOutput : public UMemory { 128 private: 129 le_int32 fSyllableCount; 130 le_int32 fOutIndex; 131 LEUnicode *fOutChars; 132 133 LEGlyphStorage &fGlyphStorage; 134 135 136 public: 137 KhmerReorderingOutput(LEUnicode *outChars, LEGlyphStorage &glyphStorage) 138 : fSyllableCount(0), fOutIndex(0), fOutChars(outChars), fGlyphStorage(glyphStorage) 139 { 140 // nothing else to do... 141 } 142 143 ~KhmerReorderingOutput() 144 { 145 // nothing to do here... 146 } 147 148 void reset() 149 { 150 fSyllableCount += 1; 151 } 152 153 void writeChar(LEUnicode ch, le_uint32 charIndex, FeatureMask charFeatures) 154 { 155 LEErrorCode success = LE_NO_ERROR; 156 157 fOutChars[fOutIndex] = ch; 158 159 fGlyphStorage.setCharIndex(fOutIndex, charIndex, success); 160 fGlyphStorage.setAuxData(fOutIndex, charFeatures | (fSyllableCount & LE_GLYPH_GROUP_MASK), success); 161 162 fOutIndex += 1; 163 } 164 165 le_int32 getOutputIndex() 166 { 167 return fOutIndex; 168 } 169 }; 170 171 172 #define blwfFeatureTag LE_BLWF_FEATURE_TAG 173 #define pstfFeatureTag LE_PSTF_FEATURE_TAG 174 #define presFeatureTag LE_PRES_FEATURE_TAG 175 #define blwsFeatureTag LE_BLWS_FEATURE_TAG 176 #define abvsFeatureTag LE_ABVS_FEATURE_TAG 177 #define pstsFeatureTag LE_PSTS_FEATURE_TAG 178 179 #define blwmFeatureTag LE_BLWM_FEATURE_TAG 180 #define abvmFeatureTag LE_ABVM_FEATURE_TAG 181 #define distFeatureTag LE_DIST_FEATURE_TAG 182 183 #define prefFeatureTag LE_PREF_FEATURE_TAG 184 #define abvfFeatureTag LE_ABVF_FEATURE_TAG 185 #define cligFeatureTag LE_CLIG_FEATURE_TAG 186 #define mkmkFeatureTag LE_MKMK_FEATURE_TAG 187 188 #define prefFeatureMask 0x80000000UL 189 #define blwfFeatureMask 0x40000000UL 190 #define abvfFeatureMask 0x20000000UL 191 #define pstfFeatureMask 0x10000000UL 192 #define presFeatureMask 0x08000000UL 193 #define blwsFeatureMask 0x04000000UL 194 #define abvsFeatureMask 0x02000000UL 195 #define pstsFeatureMask 0x01000000UL 196 #define cligFeatureMask 0x00800000UL 197 #define distFeatureMask 0x00400000UL 198 #define blwmFeatureMask 0x00200000UL 199 #define abvmFeatureMask 0x00100000UL 200 #define mkmkFeatureMask 0x00080000UL 201 202 #define tagPref (prefFeatureMask | presFeatureMask | cligFeatureMask | distFeatureMask) 203 #define tagAbvf (abvfFeatureMask | abvsFeatureMask | cligFeatureMask | distFeatureMask | abvmFeatureMask | mkmkFeatureMask) 204 #define tagPstf (blwfFeatureMask | blwsFeatureMask | prefFeatureMask | presFeatureMask | pstfFeatureMask | pstsFeatureMask | cligFeatureMask | distFeatureMask | blwmFeatureMask) 205 #define tagBlwf (blwfFeatureMask | blwsFeatureMask | cligFeatureMask | distFeatureMask | blwmFeatureMask | mkmkFeatureMask) 206 #define tagDefault (prefFeatureMask | blwfFeatureMask | presFeatureMask | blwsFeatureMask | cligFeatureMask | distFeatureMask | abvmFeatureMask | blwmFeatureMask | mkmkFeatureMask) 207 208 209 210 // These are in the order in which the features need to be applied 211 // for correct processing 212 static const FeatureMap featureMap[] = 213 { 214 // Shaping features 215 {prefFeatureTag, prefFeatureMask}, 216 {blwfFeatureTag, blwfFeatureMask}, 217 {abvfFeatureTag, abvfFeatureMask}, 218 {pstfFeatureTag, pstfFeatureMask}, 219 {presFeatureTag, presFeatureMask}, 220 {blwsFeatureTag, blwsFeatureMask}, 221 {abvsFeatureTag, abvsFeatureMask}, 222 {pstsFeatureTag, pstsFeatureMask}, 223 {cligFeatureTag, cligFeatureMask}, 224 225 // Positioning features 226 {distFeatureTag, distFeatureMask}, 227 {blwmFeatureTag, blwmFeatureMask}, 228 {abvmFeatureTag, abvmFeatureMask}, 229 {mkmkFeatureTag, mkmkFeatureMask}, 230 }; 231 232 static const le_int32 featureMapCount = LE_ARRAY_SIZE(featureMap); 233 234 // The stateTable is used to calculate the end (the length) of a well 235 // formed Khmer Syllable. 236 // 237 // Each horizontal line is ordered exactly the same way as the values in KhmerClassTable 238 // CharClassValues in KhmerReordering.h This coincidence of values allows the 239 // follow up of the table. 240 // 241 // Each line corresponds to a state, which does not necessarily need to be a type 242 // of component... for example, state 2 is a base, with is always a first character 243 // in the syllable, but the state could be produced a consonant of any type when 244 // it is the first character that is analysed (in ground state). 245 // 246 // Differentiating 3 types of consonants is necessary in order to 247 // forbid the use of certain combinations, such as having a second 248 // coeng after a coeng RO, 249 // The inexistent possibility of having a type 3 after another type 3 is permitted, 250 // eliminating it would very much complicate the table, and it does not create typing 251 // problems, as the case above. 252 // 253 // The table is quite complex, in order to limit the number of coeng consonants 254 // to 2 (by means of the table). 255 // 256 // There a peculiarity, as far as Unicode is concerned: 257 // - The consonant-shifter is considered in two possible different 258 // locations, the one considered in Unicode 3.0 and the one considered in 259 // Unicode 4.0. (there is a backwards compatibility problem in this standard). 260 261 262 // xx independent character, such as a number, punctuation sign or non-khmer char 263 // 264 // c1 Khmer consonant of type 1 or an independent vowel 265 // that is, a letter in which the subscript for is only under the 266 // base, not taking any space to the right or to the left 267 // 268 // c2 Khmer consonant of type 2, the coeng form takes space under 269 // and to the left of the base (only RO is of this type) 270 // 271 // c3 Khmer consonant of type 3. Its subscript form takes space under 272 // and to the right of the base. 273 // 274 // cs Khmer consonant shifter 275 // 276 // rb Khmer robat 277 // 278 // co coeng character (u17D2) 279 // 280 // dv dependent vowel (including split vowels, they are treated in the same way). 281 // even if dv is not defined above, the component that is really tested for is 282 // KhmerClassTable::CC_DEPENDENT_VOWEL, which is common to all dependent vowels 283 // 284 // zwj Zero Width joiner 285 // 286 // zwnj Zero width non joiner 287 // 288 // sa above sign 289 // 290 // sp post sign 291 // 292 // there are lines with equal content but for an easier understanding 293 // (and maybe change in the future) we did not join them 294 // 295 static const le_int8 khmerStateTable[][KhmerClassTable::CC_COUNT] = 296 { 297 298 // xx c1 c2 c3 zwnj cs rb co dv sa sp zwj 299 { 1, 2, 2, 2, 1, 1, 1, 6, 1, 1, 1, 2}, // 0 - ground state 300 {-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1}, // 1 - exit state (or sign to the right of the syllable) 301 {-1, -1, -1, -1, 3, 4, 5, 6, 16, 17, 1, -1}, // 2 - Base consonant 302 {-1, -1, -1, -1, -1, 4, -1, -1, 16, -1, -1, -1}, // 3 - First ZWNJ before a register shifter 303 // It can only be followed by a shifter or a vowel 304 {-1, -1, -1, -1, 15, -1, -1, 6, 16, 17, 1, 14}, // 4 - First register shifter 305 {-1, -1, -1, -1, -1, -1, -1, -1, 20, -1, 1, -1}, // 5 - Robat 306 {-1, 7, 8, 9, -1, -1, -1, -1, -1, -1, -1, -1}, // 6 - First Coeng 307 {-1, -1, -1, -1, 12, 13, -1, 10, 16, 17, 1, 14}, // 7 - First consonant of type 1 after coeng 308 {-1, -1, -1, -1, 12, 13, -1, -1, 16, 17, 1, 14}, // 8 - First consonant of type 2 after coeng 309 {-1, -1, -1, -1, 12, 13, -1, 10, 16, 17, 1, 14}, // 9 - First consonant or type 3 after ceong 310 {-1, 11, 11, 11, -1, -1, -1, -1, -1, -1, -1, -1}, // 10 - Second Coeng (no register shifter before) 311 {-1, -1, -1, -1, 15, -1, -1, -1, 16, 17, 1, 14}, // 11 - Second coeng consonant (or ind. vowel) no register shifter before 312 {-1, -1, -1, -1, -1, 13, -1, -1, 16, -1, -1, -1}, // 12 - Second ZWNJ before a register shifter 313 {-1, -1, -1, -1, 15, -1, -1, -1, 16, 17, 1, 14}, // 13 - Second register shifter 314 {-1, -1, -1, -1, -1, -1, -1, -1, 16, -1, -1, -1}, // 14 - ZWJ before vowel 315 {-1, -1, -1, -1, -1, -1, -1, -1, 16, -1, -1, -1}, // 15 - ZWNJ before vowel 316 {-1, -1, -1, -1, -1, -1, -1, -1, -1, 17, 1, 18}, // 16 - dependent vowel 317 {-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 1, 18}, // 17 - sign above 318 {-1, -1, -1, -1, -1, -1, -1, 19, -1, -1, -1, -1}, // 18 - ZWJ after vowel 319 {-1, 1, -1, 1, -1, -1, -1, -1, -1, -1, -1, -1}, // 19 - Third coeng 320 {-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 1, -1}, // 20 - dependent vowel after a Robat 321 322 }; 323 324 325 const FeatureMap *KhmerReordering::getFeatureMap(le_int32 &count) 326 { 327 count = featureMapCount; 328 329 return featureMap; 330 } 331 332 333 // Given an input string of characters and a location in which to start looking 334 // calculate, using the state table, which one is the last character of the syllable 335 // that starts in the starting position. 336 le_int32 KhmerReordering::findSyllable(const KhmerClassTable *classTable, const LEUnicode *chars, le_int32 prev, le_int32 charCount) 337 { 338 le_int32 cursor = prev; 339 le_int8 state = 0; 340 341 while (cursor < charCount) { 342 KhmerClassTable::CharClass charClass = (classTable->getCharClass(chars[cursor]) & KhmerClassTable::CF_CLASS_MASK); 343 344 state = khmerStateTable[state][charClass]; 345 346 if (state < 0) { 347 break; 348 } 349 350 cursor += 1; 351 } 352 353 return cursor; 354 } 355 356 357 // This is the real reordering function as applied to the Khmer language 358 359 le_int32 KhmerReordering::reorder(const LEUnicode *chars, le_int32 charCount, le_int32 /*scriptCode*/, 360 LEUnicode *outChars, LEGlyphStorage &glyphStorage) 361 { 362 const KhmerClassTable *classTable = KhmerClassTable::getKhmerClassTable(); 363 364 KhmerReorderingOutput output(outChars, glyphStorage); 365 KhmerClassTable::CharClass charClass; 366 le_int32 i, prev = 0, coengRo; 367 368 369 // This loop only exits when we reach the end of a run, which may contain 370 // several syllables. 371 while (prev < charCount) { 372 le_int32 syllable = findSyllable(classTable, chars, prev, charCount); 373 374 output.reset(); 375 376 // write a pre vowel or the pre part of a split vowel first 377 // and look out for coeng + ro. RO is the only vowel of type 2, and 378 // therefore the only one that requires saving space before the base. 379 coengRo = -1; // There is no Coeng Ro, if found this value will change 380 for (i = prev; i < syllable; i += 1) { 381 charClass = classTable->getCharClass(chars[i]); 382 383 // if a split vowel, write the pre part. In Khmer the pre part 384 // is the same for all split vowels, same glyph as pre vowel C_VOWEL_E 385 if (charClass & KhmerClassTable::CF_SPLIT_VOWEL) { 386 output.writeChar(C_VOWEL_E, i, tagPref); 387 break; // there can be only one vowel 388 } 389 390 // if a vowel with pos before write it out 391 if (charClass & KhmerClassTable::CF_POS_BEFORE) { 392 output.writeChar(chars[i], i, tagPref); 393 break; // there can be only one vowel 394 } 395 396 // look for coeng + ro and remember position 397 // works because coeng + ro is always in front of a vowel (if there is a vowel) 398 // and because CC_CONSONANT2 is enough to identify it, as it is the only consonant 399 // with this flag 400 if ( (charClass & KhmerClassTable::CF_COENG) && (i + 1 < syllable) && 401 ( (classTable->getCharClass(chars[i + 1]) & KhmerClassTable::CF_CLASS_MASK) == KhmerClassTable::CC_CONSONANT2) ) 402 { 403 coengRo = i; 404 } 405 } 406 407 // write coeng + ro if found 408 if (coengRo > -1) { 409 output.writeChar(C_COENG, coengRo, tagPref); 410 output.writeChar(C_RO, coengRo + 1, tagPref); 411 } 412 413 // shall we add a dotted circle? 414 // If in the position in which the base should be (first char in the string) there is 415 // a character that has the Dotted circle flag (a character that cannot be a base) 416 // then write a dotted circle 417 if (classTable->getCharClass(chars[prev]) & KhmerClassTable::CF_DOTTED_CIRCLE) { 418 output.writeChar(C_DOTTED_CIRCLE, prev, tagDefault); 419 } 420 421 // copy what is left to the output, skipping before vowels and coeng Ro if they are present 422 for (i = prev; i < syllable; i += 1) { 423 charClass = classTable->getCharClass(chars[i]); 424 425 // skip a before vowel, it was already processed 426 if (charClass & KhmerClassTable::CF_POS_BEFORE) { 427 continue; 428 } 429 430 // skip coeng + ro, it was already processed 431 if (i == coengRo) { 432 i += 1; 433 continue; 434 } 435 436 switch (charClass & KhmerClassTable::CF_POS_MASK) { 437 case KhmerClassTable::CF_POS_ABOVE : 438 output.writeChar(chars[i], i, tagAbvf); 439 break; 440 441 case KhmerClassTable::CF_POS_AFTER : 442 output.writeChar(chars[i], i, tagPstf); 443 break; 444 445 case KhmerClassTable::CF_POS_BELOW : 446 output.writeChar(chars[i], i, tagBlwf); 447 break; 448 449 default: 450 // assign the correct flags to a coeng consonant 451 // Consonants of type 3 are taged as Post forms and those type 1 as below forms 452 if ( (charClass & KhmerClassTable::CF_COENG) && i + 1 < syllable ) { 453 if ( (classTable->getCharClass(chars[i + 1]) & KhmerClassTable::CF_CLASS_MASK) 454 == KhmerClassTable::CC_CONSONANT3) { 455 output.writeChar(chars[i], i, tagPstf); 456 i += 1; 457 output.writeChar(chars[i], i, tagPstf); 458 } 459 else { 460 output.writeChar(chars[i], i, tagBlwf); 461 i += 1; 462 output.writeChar(chars[i], i, tagBlwf); 463 } 464 break; 465 } 466 // if a shifter is followed by an above vowel change the shifter to below form, 467 // an above vowel can have two possible positions i + 1 or i + 3 468 // (position i+1 corresponds to unicode 3, position i+3 to Unicode 4) 469 // and there is an extra rule for C_VOWEL_AA + C_SIGN_NIKAHIT also for two 470 // different positions, right after the shifter or after a vowel (Unicode 4) 471 if ( (charClass & KhmerClassTable::CF_SHIFTER) && (i + 1 < syllable) ) { 472 if ((classTable->getCharClass(chars[i + 1]) & KhmerClassTable::CF_ABOVE_VOWEL) 473 || (i + 2 < syllable 474 && ( (classTable->getCharClass(chars[i + 1]) & KhmerClassTable::CF_CLASS_MASK) == C_VOWEL_AA) 475 && ( (classTable->getCharClass(chars[i + 2]) & KhmerClassTable::CF_CLASS_MASK) == C_SIGN_NIKAHIT)) 476 || (i + 3 < syllable && (classTable->getCharClass(chars[i + 3]) & KhmerClassTable::CF_ABOVE_VOWEL)) 477 || (i + 4 < syllable 478 && ( (classTable->getCharClass(chars[i + 3]) & KhmerClassTable::CF_CLASS_MASK) == C_VOWEL_AA) 479 && ( (classTable->getCharClass(chars[i + 4]) & KhmerClassTable::CF_CLASS_MASK) == C_SIGN_NIKAHIT) ) ) 480 { 481 output.writeChar(chars[i], i, tagBlwf); 482 break; 483 } 484 485 } 486 // default - any other characters 487 output.writeChar(chars[i], i, tagDefault); 488 break; 489 } // switch 490 } // for 491 492 prev = syllable; // move the pointer to the start of next syllable 493 } 494 495 return output.getOutputIndex(); 496 } 497 498 499 U_NAMESPACE_END 500
