icuSources/common/uprops.h

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