icuSources/common/uprops.h

   1 // © 2016 and later: Unicode, Inc. and others.
   2 // License & terms of use: http://www.unicode.org/copyright.html
   3 /*
   4 *******************************************************************************
   5 *
   6 *   Copyright (C) 2002-2016, International Business Machines
   7 *   Corporation and others.  All Rights Reserved.
   8 *
   9 *******************************************************************************
  10 *   file name:  uprops.h
  11 *   encoding:   UTF-8
  12 *   tab size:   8 (not used)
  13 *   indentation:4
  14 *
  15 *   created on: 2002feb24
  16 *   created by: Markus W. Scherer
  17 *
  18 *   Constants for mostly non-core Unicode character properties
  19 *   stored in uprops.icu.
  20 */
  21
  22 #ifndef __UPROPS_H__
  23 #define __UPROPS_H__
  24
  25 #include "unicode/utypes.h"
  26 #include "unicode/uset.h"
  27 #include "uset_imp.h"
  28 #include "udataswp.h"
  29
  30 /* indexes[] entries */
  31 enum {
  32     UPROPS_PROPS32_INDEX,
  33     UPROPS_EXCEPTIONS_INDEX,
  34     UPROPS_EXCEPTIONS_TOP_INDEX,
  35
  36     UPROPS_ADDITIONAL_TRIE_INDEX,
  37     UPROPS_ADDITIONAL_VECTORS_INDEX,
  38     UPROPS_ADDITIONAL_VECTORS_COLUMNS_INDEX,
  39
  40     UPROPS_SCRIPT_EXTENSIONS_INDEX,
  41
  42     UPROPS_RESERVED_INDEX_7,
  43     UPROPS_RESERVED_INDEX_8,
  44
  45     /* size of the data file (number of 32-bit units after the header) */
  46     UPROPS_DATA_TOP_INDEX,
  47
  48     /* maximum values for code values in vector word 0 */
  49     UPROPS_MAX_VALUES_INDEX=10,
  50     /* maximum values for code values in vector word 2 */
  51     UPROPS_MAX_VALUES_2_INDEX,
  52
  53     UPROPS_INDEX_COUNT=16
  54 };
  55
  56 /* definitions for the main properties words */
  57 enum {
  58     /* general category shift==0                                0 (5 bits) */
  59     /* reserved                                                 5 (1 bit) */
  60     UPROPS_NUMERIC_TYPE_VALUE_SHIFT=6                       /*  6 (10 bits) */
  61 };
  62
  63 #define GET_CATEGORY(props) ((props)&0x1f)
  64 #define CAT_MASK(props) U_MASK(GET_CATEGORY(props))
  65
  66 #define GET_NUMERIC_TYPE_VALUE(props) ((props)>>UPROPS_NUMERIC_TYPE_VALUE_SHIFT)
  67
  68 /* constants for the storage form of numeric types and values */
  69 enum {
  70     /** No numeric value. */
  71     UPROPS_NTV_NONE=0,
  72     /** Decimal digits: nv=0..9 */
  73     UPROPS_NTV_DECIMAL_START=1,
  74     /** Other digits: nv=0..9 */
  75     UPROPS_NTV_DIGIT_START=11,
  76     /** Small integers: nv=0..154 */
  77     UPROPS_NTV_NUMERIC_START=21,
  78     /** Fractions: ((ntv>>4)-12) / ((ntv&0xf)+1) = -1..17 / 1..16 */
  79     UPROPS_NTV_FRACTION_START=0xb0,
  80     /**
  81      * Large integers:
  82      * ((ntv>>5)-14) * 10^((ntv&0x1f)+2) = (1..9)*(10^2..10^33)
  83      * (only one significant decimal digit)
  84      */
  85     UPROPS_NTV_LARGE_START=0x1e0,
  86     /**
  87      * Sexagesimal numbers:
  88      * ((ntv>>2)-0xbf) * 60^((ntv&3)+1) = (1..9)*(60^1..60^4)
  89      */
  90     UPROPS_NTV_BASE60_START=0x300,
  91     /**
  92      * Fraction-20 values:
  93      * frac20 = ntv-0x324 = 0..0x17 -> 1|3|5|7 / 20|40|80|160|320|640
  94      * numerator: num = 2*(frac20&3)+1
  95      * denominator: den = 20<<(frac20>>2)
  96      */
  97     UPROPS_NTV_FRACTION20_START=UPROPS_NTV_BASE60_START+36,  // 0x300+9*4=0x324
  98     /** No numeric value (yet). */
  99     UPROPS_NTV_RESERVED_START=UPROPS_NTV_FRACTION20_START+24,  // 0x324+6*4=0x34c
 100
 101     UPROPS_NTV_MAX_SMALL_INT=UPROPS_NTV_FRACTION_START-UPROPS_NTV_NUMERIC_START-1
 102 };
 103
 104 #define UPROPS_NTV_GET_TYPE(ntv) \
 105     ((ntv==UPROPS_NTV_NONE) ? U_NT_NONE : \
 106     (ntv<UPROPS_NTV_DIGIT_START) ?  U_NT_DECIMAL : \
 107     (ntv<UPROPS_NTV_NUMERIC_START) ? U_NT_DIGIT : \
 108     U_NT_NUMERIC)
 109
 110 /* number of properties vector words */
 111 #define UPROPS_VECTOR_WORDS     3
 112
 113 /*
 114  * Properties in vector word 0
 115  * Bits
 116  * 31..24   DerivedAge version major/minor one nibble each
 117  * 23..22   3..1: Bits 7..0 = Script_Extensions index
 118  *             3: Script value from Script_Extensions
 119  *             2: Script=Inherited
 120  *             1: Script=Common
 121  *             0: Script=bits 7..0
 122  * 21..20   reserved
 123  * 19..17   East Asian Width
 124  * 16.. 8   UBlockCode
 125  *  7.. 0   UScriptCode, or index to Script_Extensions
 126  */
 127
 128 /* derived age: one nibble each for major and minor version numbers */
 129 #define UPROPS_AGE_MASK         0xff000000
 130 #define UPROPS_AGE_SHIFT        24
 131
 132 /* Script_Extensions: mask includes Script */
 133 #define UPROPS_SCRIPT_X_MASK    0x00c000ff
 134 #define UPROPS_SCRIPT_X_SHIFT   22
 135
 136 #define UPROPS_EA_MASK          0x000e0000
 137 #define UPROPS_EA_SHIFT         17
 138
 139 #define UPROPS_BLOCK_MASK       0x0001ff00
 140 #define UPROPS_BLOCK_SHIFT      8
 141
 142 #define UPROPS_SCRIPT_MASK      0x000000ff
 143
 144 /* UPROPS_SCRIPT_X_WITH_COMMON must be the lowest value that involves Script_Extensions. */
 145 #define UPROPS_SCRIPT_X_WITH_COMMON     0x400000
 146 #define UPROPS_SCRIPT_X_WITH_INHERITED  0x800000
 147 #define UPROPS_SCRIPT_X_WITH_OTHER      0xc00000
 148
 149 /*
 150  * Properties in vector word 1
 151  * Each bit encodes one binary property.
 152  * The following constants represent the bit number, use 1<<UPROPS_XYZ.
 153  * UPROPS_BINARY_1_TOP<=32!
 154  *
 155  * Keep this list of property enums in sync with
 156  * propListNames[] in icu/source/tools/genprops/props2.c!
 157  *
 158  * ICU 2.6/uprops format version 3.2 stores full properties instead of "Other_".
 159  */
 160 enum {
 161     UPROPS_WHITE_SPACE,
 162     UPROPS_DASH,
 163     UPROPS_HYPHEN,
 164     UPROPS_QUOTATION_MARK,
 165     UPROPS_TERMINAL_PUNCTUATION,
 166     UPROPS_MATH,
 167     UPROPS_HEX_DIGIT,
 168     UPROPS_ASCII_HEX_DIGIT,
 169     UPROPS_ALPHABETIC,
 170     UPROPS_IDEOGRAPHIC,
 171     UPROPS_DIACRITIC,
 172     UPROPS_EXTENDER,
 173     UPROPS_NONCHARACTER_CODE_POINT,
 174     UPROPS_GRAPHEME_EXTEND,
 175     UPROPS_GRAPHEME_LINK,
 176     UPROPS_IDS_BINARY_OPERATOR,
 177     UPROPS_IDS_TRINARY_OPERATOR,
 178     UPROPS_RADICAL,
 179     UPROPS_UNIFIED_IDEOGRAPH,
 180     UPROPS_DEFAULT_IGNORABLE_CODE_POINT,
 181     UPROPS_DEPRECATED,
 182     UPROPS_LOGICAL_ORDER_EXCEPTION,
 183     UPROPS_XID_START,
 184     UPROPS_XID_CONTINUE,
 185     UPROPS_ID_START,                            /* ICU 2.6, uprops format version 3.2 */
 186     UPROPS_ID_CONTINUE,
 187     UPROPS_GRAPHEME_BASE,
 188     UPROPS_S_TERM,                              /* new in ICU 3.0 and Unicode 4.0.1 */
 189     UPROPS_VARIATION_SELECTOR,
 190     UPROPS_PATTERN_SYNTAX,                      /* new in ICU 3.4 and Unicode 4.1 */
 191     UPROPS_PATTERN_WHITE_SPACE,
 192     UPROPS_PREPENDED_CONCATENATION_MARK,        // new in ICU 60 and Unicode 10
 193     UPROPS_BINARY_1_TOP                         /* ==32 - full! */
 194 };
 195
 196 /*
 197  * Properties in vector word 2
 198  * Bits
 199  * 31..27   http://www.unicode.org/reports/tr51/#Emoji_Properties
 200  *     26   reserved
 201  * 25..20   Line Break
 202  * 19..15   Sentence Break
 203  * 14..10   Word Break
 204  *  9.. 5   Grapheme Cluster Break
 205  *  4.. 0   Decomposition Type
 206  */
 207 enum {
 208     UPROPS_2_EMOJI_COMPONENT=27,
 209     UPROPS_2_EMOJI,
 210     UPROPS_2_EMOJI_PRESENTATION,
 211     UPROPS_2_EMOJI_MODIFIER,
 212     UPROPS_2_EMOJI_MODIFIER_BASE
 213 };
 214
 215 #define UPROPS_LB_MASK          0x03f00000
 216 #define UPROPS_LB_SHIFT         20
 217
 218 #define UPROPS_SB_MASK          0x000f8000
 219 #define UPROPS_SB_SHIFT         15
 220
 221 #define UPROPS_WB_MASK          0x00007c00
 222 #define UPROPS_WB_SHIFT         10
 223
 224 #define UPROPS_GCB_MASK         0x000003e0
 225 #define UPROPS_GCB_SHIFT        5
 226
 227 #define UPROPS_DT_MASK          0x0000001f
 228
 229 /**
 230  * Gets the main properties value for a code point.
 231  * Implemented in uchar.c for uprops.cpp.
 232  */
 233 U_CFUNC uint32_t
 234 u_getMainProperties(UChar32 c);
 235
 236 /**
 237  * Get a properties vector word for a code point.
 238  * Implemented in uchar.c for uprops.cpp.
 239  * @return 0 if no data or illegal argument
 240  */
 241 U_CFUNC uint32_t
 242 u_getUnicodeProperties(UChar32 c, int32_t column);
 243
 244 /**
 245  * Get the the maximum values for some enum/int properties.
 246  * Use the same column numbers as for u_getUnicodeProperties().
 247  * The returned value will contain maximum values stored in the same bit fields
 248  * as where the enum values are stored in the u_getUnicodeProperties()
 249  * return values for the same columns.
 250  *
 251  * Valid columns are those for properties words that contain enumerated values.
 252  * (ICU 2.6: columns 0 and 2)
 253  * For other column numbers, this function will return 0.
 254  *
 255  * @internal
 256  */
 257 U_CFUNC int32_t
 258 uprv_getMaxValues(int32_t column);
 259
 260 /**
 261  * Checks if c is alphabetic, or a decimal digit; implements UCHAR_POSIX_ALNUM.
 262  * @internal
 263  */
 264 U_CFUNC UBool
 265 u_isalnumPOSIX(UChar32 c);
 266
 267 /**
 268  * Checks if c is in
 269  * [^\p{space}\p{gc=Control}\p{gc=Surrogate}\p{gc=Unassigned}]
 270  * with space=\p{Whitespace} and Control=Cc.
 271  * Implements UCHAR_POSIX_GRAPH.
 272  * @internal
 273  */
 274 U_CFUNC UBool
 275 u_isgraphPOSIX(UChar32 c);
 276
 277 /**
 278  * Checks if c is in \p{graph}\p{blank} - \p{cntrl}.
 279  * Implements UCHAR_POSIX_PRINT.
 280  * @internal
 281  */
 282 U_CFUNC UBool
 283 u_isprintPOSIX(UChar32 c);
 284
 285 /** Turn a bit index into a bit flag. @internal */
 286 #define FLAG(n) ((uint32_t)1<<(n))
 287
 288 /** Flags for general categories in the order of UCharCategory. @internal */
 289 #define _Cn     FLAG(U_GENERAL_OTHER_TYPES)
 290 #define _Lu     FLAG(U_UPPERCASE_LETTER)
 291 #define _Ll     FLAG(U_LOWERCASE_LETTER)
 292 #define _Lt     FLAG(U_TITLECASE_LETTER)
 293 #define _Lm     FLAG(U_MODIFIER_LETTER)
 294 /* #define _Lo     FLAG(U_OTHER_LETTER) -- conflicts with MS Visual Studio 9.0 xiosbase */
 295 #define _Mn     FLAG(U_NON_SPACING_MARK)
 296 #define _Me     FLAG(U_ENCLOSING_MARK)
 297 #define _Mc     FLAG(U_COMBINING_SPACING_MARK)
 298 #define _Nd     FLAG(U_DECIMAL_DIGIT_NUMBER)
 299 #define _Nl     FLAG(U_LETTER_NUMBER)
 300 #define _No     FLAG(U_OTHER_NUMBER)
 301 #define _Zs     FLAG(U_SPACE_SEPARATOR)
 302 #define _Zl     FLAG(U_LINE_SEPARATOR)
 303 #define _Zp     FLAG(U_PARAGRAPH_SEPARATOR)
 304 #define _Cc     FLAG(U_CONTROL_CHAR)
 305 #define _Cf     FLAG(U_FORMAT_CHAR)
 306 #define _Co     FLAG(U_PRIVATE_USE_CHAR)
 307 #define _Cs     FLAG(U_SURROGATE)
 308 #define _Pd     FLAG(U_DASH_PUNCTUATION)
 309 #define _Ps     FLAG(U_START_PUNCTUATION)
 310 /* #define _Pe     FLAG(U_END_PUNCTUATION) -- conflicts with MS Visual Studio 9.0 xlocnum */
 311 /* #define _Pc     FLAG(U_CONNECTOR_PUNCTUATION) -- conflicts with MS Visual Studio 9.0 streambuf */
 312 #define _Po     FLAG(U_OTHER_PUNCTUATION)
 313 #define _Sm     FLAG(U_MATH_SYMBOL)
 314 #define _Sc     FLAG(U_CURRENCY_SYMBOL)
 315 #define _Sk     FLAG(U_MODIFIER_SYMBOL)
 316 #define _So     FLAG(U_OTHER_SYMBOL)
 317 #define _Pi     FLAG(U_INITIAL_PUNCTUATION)
 318 /* #define _Pf     FLAG(U_FINAL_PUNCTUATION) -- conflicts with MS Visual Studio 9.0 streambuf */
 319
 320 /** Some code points. @internal */
 321 enum {
 322     TAB     =0x0009,
 323     LF      =0x000a,
 324     FF      =0x000c,
 325     CR      =0x000d,
 326     U_A     =0x0041,
 327     U_F     =0x0046,
 328     U_Z     =0x005a,
 329     U_a     =0x0061,
 330     U_f     =0x0066,
 331     U_z     =0x007a,
 332     DEL     =0x007f,
 333     NL      =0x0085,
 334     NBSP    =0x00a0,
 335     CGJ     =0x034f,
 336     FIGURESP=0x2007,
 337     HAIRSP  =0x200a,
 338     ZWNJ    =0x200c,
 339     ZWJ     =0x200d,
 340     RLM     =0x200f,
 341     NNBSP   =0x202f,
 342     WJ      =0x2060,
 343     INHSWAP =0x206a,
 344     NOMDIG  =0x206f,
 345     U_FW_A  =0xff21,
 346     U_FW_F  =0xff26,
 347     U_FW_Z  =0xff3a,
 348     U_FW_a  =0xff41,
 349     U_FW_f  =0xff46,
 350     U_FW_z  =0xff5a,
 351     ZWNBSP  =0xfeff
 352 };
 353
 354 /**
 355  * Get the maximum length of a (regular/1.0/extended) character name.
 356  * @return 0 if no character names available.
 357  */
 358 U_CAPI int32_t U_EXPORT2
 359 uprv_getMaxCharNameLength(void);
 360
 361 /**
 362  * Fills set with characters that are used in Unicode character names.
 363  * Includes all characters that are used in regular/Unicode 1.0/extended names.
 364  * Just empties the set if no character names are available.
 365  * @param sa USetAdder to receive characters.
 366  */
 367 U_CAPI void U_EXPORT2
 368 uprv_getCharNameCharacters(const USetAdder *sa);
 369
 370 /**
 371  * Constants for which data and implementation files provide which properties.
 372  * Used by UnicodeSet for service-specific property enumeration.
 373  * @internal
 374  */
 375 enum UPropertySource {
 376     /** No source, not a supported property. */
 377     UPROPS_SRC_NONE,
 378     /** From uchar.c/uprops.icu main trie */
 379     UPROPS_SRC_CHAR,
 380     /** From uchar.c/uprops.icu properties vectors trie */
 381     UPROPS_SRC_PROPSVEC,
 382     /** From unames.c/unames.icu */
 383     UPROPS_SRC_NAMES,
 384     /** From ucase.c/ucase.icu */
 385     UPROPS_SRC_CASE,
 386     /** From ubidi_props.c/ubidi.icu */
 387     UPROPS_SRC_BIDI,
 388     /** From uchar.c/uprops.icu main trie as well as properties vectors trie */
 389     UPROPS_SRC_CHAR_AND_PROPSVEC,
 390     /** From ucase.c/ucase.icu as well as unorm.cpp/unorm.icu */
 391     UPROPS_SRC_CASE_AND_NORM,
 392     /** From normalizer2impl.cpp/nfc.nrm */
 393     UPROPS_SRC_NFC,
 394     /** From normalizer2impl.cpp/nfkc.nrm */
 395     UPROPS_SRC_NFKC,
 396     /** From normalizer2impl.cpp/nfkc_cf.nrm */
 397     UPROPS_SRC_NFKC_CF,
 398     /** From normalizer2impl.cpp/nfc.nrm canonical iterator data */
 399     UPROPS_SRC_NFC_CANON_ITER,
 400     /** One more than the highest UPropertySource (UPROPS_SRC_) constant. */
 401     UPROPS_SRC_COUNT
 402 };
 403 typedef enum UPropertySource UPropertySource;
 404
 405 /**
 406  * @see UPropertySource
 407  * @internal
 408  */
 409 U_CFUNC UPropertySource U_EXPORT2
 410 uprops_getSource(UProperty which);
 411
 412 /**
 413  * Enumerate uprops.icu's main data trie and add the
 414  * start of each range of same properties to the set.
 415  * @internal
 416  */
 417 U_CFUNC void U_EXPORT2
 418 uchar_addPropertyStarts(const USetAdder *sa, UErrorCode *pErrorCode);
 419
 420 /**
 421  * Enumerate uprops.icu's properties vectors trie and add the
 422  * start of each range of same properties to the set.
 423  * @internal
 424  */
 425 U_CFUNC void U_EXPORT2
 426 upropsvec_addPropertyStarts(const USetAdder *sa, UErrorCode *pErrorCode);
 427
 428 /**
 429  * Return a set of characters for property enumeration.
 430  * For each two consecutive characters (start, limit) in the set,
 431  * all of the properties for start..limit-1 are all the same.
 432  *
 433  * @param sa USetAdder to receive result. Existing contents are lost.
 434  * @internal
 435  */
 436 /*U_CFUNC void U_EXPORT2
 437 uprv_getInclusions(const USetAdder *sa, UErrorCode *pErrorCode);
 438 */
 439
 440 /**
 441  * Swap the ICU Unicode character names file. See uchar.c.
 442  * @internal
 443  */
 444 U_CAPI int32_t U_EXPORT2
 445 uchar_swapNames(const UDataSwapper *ds,
 446                 const void *inData, int32_t length, void *outData,
 447                 UErrorCode *pErrorCode);
 448
 449 #ifdef __cplusplus
 450
 451 U_NAMESPACE_BEGIN
 452
 453 class UnicodeSet;
 454
 455 // implemented in uniset_props.cpp
 456 U_CFUNC UnicodeSet *
 457 uniset_getUnicode32Instance(UErrorCode &errorCode);
 458
 459 U_NAMESPACE_END
 460
 461 #endif
 462
 463 #endif