collation.h (19435B)
1 // © 2016 and later: Unicode, Inc. and others. 2 // License & terms of use: http://www.unicode.org/copyright.html 3 /* 4 ******************************************************************************* 5 * Copyright (C) 2010-2015, International Business Machines 6 * Corporation and others. All Rights Reserved. 7 ******************************************************************************* 8 * collation.h 9 * 10 * created on: 2010oct27 11 * created by: Markus W. Scherer 12 */ 13 14 #ifndef __COLLATION_H__ 15 #define __COLLATION_H__ 16 17 #include "unicode/utypes.h" 18 19 #if !UCONFIG_NO_COLLATION 20 21 U_NAMESPACE_BEGIN 22 23 /** 24 * Collation v2 basic definitions and static helper functions. 25 * 26 * Data structures except for expansion tables store 32-bit CEs which are 27 * either specials (see tags below) or are compact forms of 64-bit CEs. 28 */ 29 class U_I18N_API Collation { 30 public: 31 // Special sort key bytes for all levels. 32 static const uint8_t TERMINATOR_BYTE = 0; 33 static const uint8_t LEVEL_SEPARATOR_BYTE = 1; 34 35 /** The secondary/tertiary lower limit for tailoring before any root elements. */ 36 static const uint32_t BEFORE_WEIGHT16 = 0x0100; 37 38 /** 39 * Merge-sort-key separator. 40 * Same as the unique primary and identical-level weights of U+FFFE. 41 * Must not be used as primary compression low terminator. 42 * Otherwise usable. 43 */ 44 static const uint8_t MERGE_SEPARATOR_BYTE = 2; 45 static const uint32_t MERGE_SEPARATOR_PRIMARY = 0x02000000; // U+FFFE 46 static const uint32_t MERGE_SEPARATOR_CE32 = 0x02000505; // U+FFFE 47 48 /** 49 * Primary compression low terminator, must be greater than MERGE_SEPARATOR_BYTE. 50 * Reserved value in primary second byte if the lead byte is compressible. 51 * Otherwise usable in all CE weight bytes. 52 */ 53 static const uint8_t PRIMARY_COMPRESSION_LOW_BYTE = 3; 54 /** 55 * Primary compression high terminator. 56 * Reserved value in primary second byte if the lead byte is compressible. 57 * Otherwise usable in all CE weight bytes. 58 */ 59 static const uint8_t PRIMARY_COMPRESSION_HIGH_BYTE = 0xff; 60 61 /** Default secondary/tertiary weight lead byte. */ 62 static const uint8_t COMMON_BYTE = 5; 63 static const uint32_t COMMON_WEIGHT16 = 0x0500; 64 /** Middle 16 bits of a CE with a common secondary weight. */ 65 static const uint32_t COMMON_SECONDARY_CE = 0x05000000; 66 /** Lower 16 bits of a CE with a common tertiary weight. */ 67 static const uint32_t COMMON_TERTIARY_CE = 0x0500; 68 /** Lower 32 bits of a CE with common secondary and tertiary weights. */ 69 static const uint32_t COMMON_SEC_AND_TER_CE = 0x05000500; 70 71 static const uint32_t SECONDARY_MASK = 0xffff0000; 72 static const uint32_t CASE_MASK = 0xc000; 73 static const uint32_t SECONDARY_AND_CASE_MASK = SECONDARY_MASK | CASE_MASK; 74 /** Only the 2*6 bits for the pure tertiary weight. */ 75 static const uint32_t ONLY_TERTIARY_MASK = 0x3f3f; 76 /** Only the secondary & tertiary bits; no case, no quaternary. */ 77 static const uint32_t ONLY_SEC_TER_MASK = SECONDARY_MASK | ONLY_TERTIARY_MASK; 78 /** Case bits and tertiary bits. */ 79 static const uint32_t CASE_AND_TERTIARY_MASK = CASE_MASK | ONLY_TERTIARY_MASK; 80 static const uint32_t QUATERNARY_MASK = 0xc0; 81 /** Case bits and quaternary bits. */ 82 static const uint32_t CASE_AND_QUATERNARY_MASK = CASE_MASK | QUATERNARY_MASK; 83 84 static const uint8_t UNASSIGNED_IMPLICIT_BYTE = 0xfe; // compressible 85 /** 86 * First unassigned: AlphabeticIndex overflow boundary. 87 * We want a 3-byte primary so that it fits into the root elements table. 88 * 89 * This 3-byte primary will not collide with 90 * any unassigned-implicit 4-byte primaries because 91 * the first few hundred Unicode code points all have real mappings. 92 */ 93 static const uint32_t FIRST_UNASSIGNED_PRIMARY = 0xfe040200; 94 95 static const uint8_t TRAIL_WEIGHT_BYTE = 0xff; // not compressible 96 static const uint32_t FIRST_TRAILING_PRIMARY = 0xff020200; // [first trailing] 97 static const uint32_t MAX_PRIMARY = 0xffff0000; // U+FFFF 98 static const uint32_t MAX_REGULAR_CE32 = 0xffff0505; // U+FFFF 99 100 // CE32 value for U+FFFD as well as illegal UTF-8 byte sequences (which behave like U+FFFD). 101 // We use the third-highest primary weight for U+FFFD (as in UCA 6.3+). 102 static const uint32_t FFFD_PRIMARY = MAX_PRIMARY - 0x20000; 103 static const uint32_t FFFD_CE32 = MAX_REGULAR_CE32 - 0x20000; 104 105 /** 106 * A CE32 is special if its low byte is this or greater. 107 * Impossible case bits 11 mark special CE32s. 108 * This value itself is used to indicate a fallback to the base collator. 109 */ 110 static const uint8_t SPECIAL_CE32_LOW_BYTE = 0xc0; 111 static const uint32_t FALLBACK_CE32 = SPECIAL_CE32_LOW_BYTE; 112 /** 113 * Low byte of a long-primary special CE32. 114 */ 115 static const uint8_t LONG_PRIMARY_CE32_LOW_BYTE = 0xc1; // SPECIAL_CE32_LOW_BYTE | LONG_PRIMARY_TAG 116 117 static const uint32_t UNASSIGNED_CE32 = 0xffffffff; // Compute an unassigned-implicit CE. 118 119 static const uint32_t NO_CE32 = 1; 120 121 /** No CE: End of input. Only used in runtime code, not stored in data. */ 122 static const uint32_t NO_CE_PRIMARY = 1; // not a left-adjusted weight 123 static const uint32_t NO_CE_WEIGHT16 = 0x0100; // weight of LEVEL_SEPARATOR_BYTE 124 static const int64_t NO_CE = INT64_C(0x101000100); // NO_CE_PRIMARY, NO_CE_WEIGHT16, NO_CE_WEIGHT16 125 126 /** Sort key levels. */ 127 enum Level { 128 /** Unspecified level. */ 129 NO_LEVEL, 130 PRIMARY_LEVEL, 131 SECONDARY_LEVEL, 132 CASE_LEVEL, 133 TERTIARY_LEVEL, 134 QUATERNARY_LEVEL, 135 IDENTICAL_LEVEL, 136 /** Beyond sort key bytes. */ 137 ZERO_LEVEL 138 }; 139 140 /** 141 * Sort key level flags: xx_FLAG = 1 << xx_LEVEL. 142 * In Java, use enum Level with flag() getters, or use EnumSet rather than hand-made bit sets. 143 */ 144 static const uint32_t NO_LEVEL_FLAG = 1; 145 static const uint32_t PRIMARY_LEVEL_FLAG = 2; 146 static const uint32_t SECONDARY_LEVEL_FLAG = 4; 147 static const uint32_t CASE_LEVEL_FLAG = 8; 148 static const uint32_t TERTIARY_LEVEL_FLAG = 0x10; 149 static const uint32_t QUATERNARY_LEVEL_FLAG = 0x20; 150 static const uint32_t IDENTICAL_LEVEL_FLAG = 0x40; 151 static const uint32_t ZERO_LEVEL_FLAG = 0x80; 152 153 /** 154 * Special-CE32 tags, from bits 3..0 of a special 32-bit CE. 155 * Bits 31..8 are available for tag-specific data. 156 * Bits 5..4: Reserved. May be used in the future to indicate lccc!=0 and tccc!=0. 157 */ 158 enum { 159 /** 160 * Fall back to the base collator. 161 * This is the tag value in SPECIAL_CE32_LOW_BYTE and FALLBACK_CE32. 162 * Bits 31..8: Unused, 0. 163 */ 164 FALLBACK_TAG = 0, 165 /** 166 * Long-primary CE with COMMON_SEC_AND_TER_CE. 167 * Bits 31..8: Three-byte primary. 168 */ 169 LONG_PRIMARY_TAG = 1, 170 /** 171 * Long-secondary CE with zero primary. 172 * Bits 31..16: Secondary weight. 173 * Bits 15.. 8: Tertiary weight. 174 */ 175 LONG_SECONDARY_TAG = 2, 176 /** 177 * Unused. 178 * May be used in the future for single-byte secondary CEs (SHORT_SECONDARY_TAG), 179 * storing the secondary in bits 31..24, the ccc in bits 23..16, 180 * and the tertiary in bits 15..8. 181 */ 182 RESERVED_TAG_3 = 3, 183 /** 184 * Latin mini expansions of two simple CEs [pp, 05, tt] [00, ss, 05]. 185 * Bits 31..24: Single-byte primary weight pp of the first CE. 186 * Bits 23..16: Tertiary weight tt of the first CE. 187 * Bits 15.. 8: Secondary weight ss of the second CE. 188 */ 189 LATIN_EXPANSION_TAG = 4, 190 /** 191 * Points to one or more simple/long-primary/long-secondary 32-bit CE32s. 192 * Bits 31..13: Index into uint32_t table. 193 * Bits 12.. 8: Length=1..31. 194 */ 195 EXPANSION32_TAG = 5, 196 /** 197 * Points to one or more 64-bit CEs. 198 * Bits 31..13: Index into CE table. 199 * Bits 12.. 8: Length=1..31. 200 */ 201 EXPANSION_TAG = 6, 202 /** 203 * Builder data, used only in the CollationDataBuilder, not in runtime data. 204 * 205 * If bit 8 is 0: Builder context, points to a list of context-sensitive mappings. 206 * Bits 31..13: Index to the builder's list of ConditionalCE32 for this character. 207 * Bits 12.. 9: Unused, 0. 208 * 209 * If bit 8 is 1 (IS_BUILDER_JAMO_CE32): Builder-only jamoCE32 value. 210 * The builder fetches the Jamo CE32 from the trie. 211 * Bits 31..13: Jamo code point. 212 * Bits 12.. 9: Unused, 0. 213 */ 214 BUILDER_DATA_TAG = 7, 215 /** 216 * Points to prefix trie. 217 * Bits 31..13: Index into prefix/contraction data. 218 * Bits 12.. 8: Unused, 0. 219 */ 220 PREFIX_TAG = 8, 221 /** 222 * Points to contraction data. 223 * Bits 31..13: Index into prefix/contraction data. 224 * Bit 12: Unused, 0. 225 * Bit 11: CONTRACT_HAS_STARTER flag. (Used by ICU4X only.) 226 * Bit 10: CONTRACT_TRAILING_CCC flag. 227 * Bit 9: CONTRACT_NEXT_CCC flag. 228 * Bit 8: CONTRACT_SINGLE_CP_NO_MATCH flag. 229 */ 230 CONTRACTION_TAG = 9, 231 /** 232 * Decimal digit. 233 * Bits 31..13: Index into uint32_t table for non-numeric-collation CE32. 234 * Bit 12: Unused, 0. 235 * Bits 11.. 8: Digit value 0..9. 236 */ 237 DIGIT_TAG = 10, 238 /** 239 * Tag for U+0000, for moving the NUL-termination handling 240 * from the regular fastpath into specials-handling code. 241 * Bits 31..8: Unused, 0. 242 */ 243 U0000_TAG = 11, 244 /** 245 * Tag for a Hangul syllable. 246 * Bits 31..9: Unused, 0. 247 * Bit 8: HANGUL_NO_SPECIAL_JAMO flag. 248 */ 249 HANGUL_TAG = 12, 250 /** 251 * Tag for a lead surrogate code unit. 252 * Optional optimization for UTF-16 string processing. 253 * Bits 31..10: Unused, 0. 254 * 9.. 8: =0: All associated supplementary code points are unassigned-implicit. 255 * =1: All associated supplementary code points fall back to the base data. 256 * else: (Normally 2) Look up the data for the supplementary code point. 257 */ 258 LEAD_SURROGATE_TAG = 13, 259 /** 260 * Tag for CEs with primary weights in code point order. 261 * Bits 31..13: Index into CE table, for one data "CE". 262 * Bits 12.. 8: Unused, 0. 263 * 264 * This data "CE" has the following bit fields: 265 * Bits 63..32: Three-byte primary pppppp00. 266 * 31.. 8: Start/base code point of the in-order range. 267 * 7: Flag isCompressible primary. 268 * 6.. 0: Per-code point primary-weight increment. 269 */ 270 OFFSET_TAG = 14, 271 /** 272 * Implicit CE tag. Compute an unassigned-implicit CE. 273 * All bits are set (UNASSIGNED_CE32=0xffffffff). 274 */ 275 IMPLICIT_TAG = 15 276 }; 277 278 static UBool isAssignedCE32(uint32_t ce32) { 279 return ce32 != FALLBACK_CE32 && ce32 != UNASSIGNED_CE32; 280 } 281 282 /** 283 * We limit the number of CEs in an expansion 284 * so that we can use a small number of length bits in the data structure, 285 * and so that an implementation can copy CEs at runtime without growing a destination buffer. 286 */ 287 static const int32_t MAX_EXPANSION_LENGTH = 31; 288 static const int32_t MAX_INDEX = 0x7ffff; 289 290 /** 291 * Set if there is no match for the single (no-suffix) character itself. 292 * This is only possible if there is a prefix. 293 * In this case, discontiguous contraction matching cannot add combining marks 294 * starting from an empty suffix. 295 * The default CE32 is used anyway if there is no suffix match. 296 */ 297 static const uint32_t CONTRACT_SINGLE_CP_NO_MATCH = 0x100; 298 /** Set if the first character of every contraction suffix has lccc!=0. */ 299 static const uint32_t CONTRACT_NEXT_CCC = 0x200; 300 /** Set if any contraction suffix ends with lccc!=0. */ 301 static const uint32_t CONTRACT_TRAILING_CCC = 0x400; 302 /** Set if any contraction suffix contains a starter. (Used by ICU4X only.) */ 303 static const uint32_t CONTRACT_HAS_STARTER = 0x800; 304 305 /** For HANGUL_TAG: None of its Jamo CE32s isSpecialCE32(). */ 306 static const uint32_t HANGUL_NO_SPECIAL_JAMO = 0x100; 307 308 static const uint32_t LEAD_ALL_UNASSIGNED = 0; 309 static const uint32_t LEAD_ALL_FALLBACK = 0x100; 310 static const uint32_t LEAD_MIXED = 0x200; 311 static const uint32_t LEAD_TYPE_MASK = 0x300; 312 313 static uint32_t makeLongPrimaryCE32(uint32_t p) { return p | LONG_PRIMARY_CE32_LOW_BYTE; } 314 315 /** Turns the long-primary CE32 into a primary weight pppppp00. */ 316 static inline uint32_t primaryFromLongPrimaryCE32(uint32_t ce32) { 317 return ce32 & 0xffffff00; 318 } 319 static inline int64_t ceFromLongPrimaryCE32(uint32_t ce32) { 320 return (static_cast<int64_t>(ce32 & 0xffffff00) << 32) | COMMON_SEC_AND_TER_CE; 321 } 322 323 static uint32_t makeLongSecondaryCE32(uint32_t lower32) { 324 return lower32 | SPECIAL_CE32_LOW_BYTE | LONG_SECONDARY_TAG; 325 } 326 static inline int64_t ceFromLongSecondaryCE32(uint32_t ce32) { 327 return ce32 & 0xffffff00; 328 } 329 330 /** Makes a special CE32 with tag, index and length. */ 331 static uint32_t makeCE32FromTagIndexAndLength(int32_t tag, int32_t index, int32_t length) { 332 return (index << 13) | (length << 8) | SPECIAL_CE32_LOW_BYTE | tag; 333 } 334 /** Makes a special CE32 with only tag and index. */ 335 static uint32_t makeCE32FromTagAndIndex(int32_t tag, int32_t index) { 336 return (index << 13) | SPECIAL_CE32_LOW_BYTE | tag; 337 } 338 339 static inline UBool isSpecialCE32(uint32_t ce32) { 340 return (ce32 & 0xff) >= SPECIAL_CE32_LOW_BYTE; 341 } 342 343 static inline int32_t tagFromCE32(uint32_t ce32) { 344 return static_cast<int32_t>(ce32 & 0xf); 345 } 346 347 static inline UBool hasCE32Tag(uint32_t ce32, int32_t tag) { 348 return isSpecialCE32(ce32) && tagFromCE32(ce32) == tag; 349 } 350 351 static inline UBool isLongPrimaryCE32(uint32_t ce32) { 352 return hasCE32Tag(ce32, LONG_PRIMARY_TAG); 353 } 354 355 static UBool isSimpleOrLongCE32(uint32_t ce32) { 356 return !isSpecialCE32(ce32) || 357 tagFromCE32(ce32) == LONG_PRIMARY_TAG || 358 tagFromCE32(ce32) == LONG_SECONDARY_TAG; 359 } 360 361 /** 362 * @return true if the ce32 yields one or more CEs without further data lookups 363 */ 364 static UBool isSelfContainedCE32(uint32_t ce32) { 365 return !isSpecialCE32(ce32) || 366 tagFromCE32(ce32) == LONG_PRIMARY_TAG || 367 tagFromCE32(ce32) == LONG_SECONDARY_TAG || 368 tagFromCE32(ce32) == LATIN_EXPANSION_TAG; 369 } 370 371 static inline UBool isPrefixCE32(uint32_t ce32) { 372 return hasCE32Tag(ce32, PREFIX_TAG); 373 } 374 375 static inline UBool isContractionCE32(uint32_t ce32) { 376 return hasCE32Tag(ce32, CONTRACTION_TAG); 377 } 378 379 static inline UBool ce32HasContext(uint32_t ce32) { 380 return isSpecialCE32(ce32) && 381 (tagFromCE32(ce32) == PREFIX_TAG || 382 tagFromCE32(ce32) == CONTRACTION_TAG); 383 } 384 385 /** 386 * Get the first of the two Latin-expansion CEs encoded in ce32. 387 * @see LATIN_EXPANSION_TAG 388 */ 389 static inline int64_t latinCE0FromCE32(uint32_t ce32) { 390 return (static_cast<int64_t>(ce32 & 0xff000000) << 32) | COMMON_SECONDARY_CE | ((ce32 & 0xff0000) >> 8); 391 } 392 393 /** 394 * Get the second of the two Latin-expansion CEs encoded in ce32. 395 * @see LATIN_EXPANSION_TAG 396 */ 397 static inline int64_t latinCE1FromCE32(uint32_t ce32) { 398 return ((ce32 & 0xff00) << 16) | COMMON_TERTIARY_CE; 399 } 400 401 /** 402 * Returns the data index from a special CE32. 403 */ 404 static inline int32_t indexFromCE32(uint32_t ce32) { 405 return static_cast<int32_t>(ce32 >> 13); 406 } 407 408 /** 409 * Returns the data length from a ce32. 410 */ 411 static inline int32_t lengthFromCE32(uint32_t ce32) { 412 return (ce32 >> 8) & 31; 413 } 414 415 /** 416 * Returns the digit value from a DIGIT_TAG ce32. 417 */ 418 static inline char digitFromCE32(uint32_t ce32) { 419 return static_cast<char>((ce32 >> 8) & 0xf); 420 } 421 422 /** Returns a 64-bit CE from a simple CE32 (not special). */ 423 static inline int64_t ceFromSimpleCE32(uint32_t ce32) { 424 // normal form ppppsstt -> pppp0000ss00tt00 425 // assert (ce32 & 0xff) < SPECIAL_CE32_LOW_BYTE 426 return (static_cast<int64_t>(ce32 & 0xffff0000) << 32) | ((ce32 & 0xff00) << 16) | ((ce32 & 0xff) << 8); 427 } 428 429 /** Returns a 64-bit CE from a simple/long-primary/long-secondary CE32. */ 430 static inline int64_t ceFromCE32(uint32_t ce32) { 431 uint32_t tertiary = ce32 & 0xff; 432 if(tertiary < SPECIAL_CE32_LOW_BYTE) { 433 // normal form ppppsstt -> pppp0000ss00tt00 434 return (static_cast<int64_t>(ce32 & 0xffff0000) << 32) | ((ce32 & 0xff00) << 16) | (tertiary << 8); 435 } else { 436 ce32 -= tertiary; 437 if((tertiary & 0xf) == LONG_PRIMARY_TAG) { 438 // long-primary form ppppppC1 -> pppppp00050000500 439 return (static_cast<int64_t>(ce32) << 32) | COMMON_SEC_AND_TER_CE; 440 } else { 441 // long-secondary form ssssttC2 -> 00000000sssstt00 442 // assert (tertiary & 0xf) == LONG_SECONDARY_TAG 443 return ce32; 444 } 445 } 446 } 447 448 /** Creates a CE from a primary weight. */ 449 static inline int64_t makeCE(uint32_t p) { 450 return (static_cast<int64_t>(p) << 32) | COMMON_SEC_AND_TER_CE; 451 } 452 /** 453 * Creates a CE from a primary weight, 454 * 16-bit secondary/tertiary weights, and a 2-bit quaternary. 455 */ 456 static inline int64_t makeCE(uint32_t p, uint32_t s, uint32_t t, uint32_t q) { 457 return (static_cast<int64_t>(p) << 32) | (s << 16) | t | (q << 6); 458 } 459 460 /** 461 * Increments a 2-byte primary by a code point offset. 462 */ 463 static uint32_t incTwoBytePrimaryByOffset(uint32_t basePrimary, UBool isCompressible, 464 int32_t offset); 465 466 /** 467 * Increments a 3-byte primary by a code point offset. 468 */ 469 static uint32_t incThreeBytePrimaryByOffset(uint32_t basePrimary, UBool isCompressible, 470 int32_t offset); 471 472 /** 473 * Decrements a 2-byte primary by one range step (1..0x7f). 474 */ 475 static uint32_t decTwoBytePrimaryByOneStep(uint32_t basePrimary, UBool isCompressible, int32_t step); 476 477 /** 478 * Decrements a 3-byte primary by one range step (1..0x7f). 479 */ 480 static uint32_t decThreeBytePrimaryByOneStep(uint32_t basePrimary, UBool isCompressible, int32_t step); 481 482 /** 483 * Computes a 3-byte primary for c's OFFSET_TAG data "CE". 484 */ 485 static uint32_t getThreeBytePrimaryForOffsetData(UChar32 c, int64_t dataCE); 486 487 /** 488 * Returns the unassigned-character implicit primary weight for any valid code point c. 489 */ 490 static uint32_t unassignedPrimaryFromCodePoint(UChar32 c); 491 492 static inline int64_t unassignedCEFromCodePoint(UChar32 c) { 493 return makeCE(unassignedPrimaryFromCodePoint(c)); 494 } 495 496 private: 497 Collation() = delete; // No instantiation. 498 }; 499 500 U_NAMESPACE_END 501 502 #endif // !UCONFIG_NO_COLLATION 503 #endif // __COLLATION_H__