michael@0: /*
michael@0: *******************************************************************************
michael@0: *
michael@0: *   Copyright (C) 2002-2012, International Business Machines
michael@0: *   Corporation and others.  All Rights Reserved.
michael@0: *
michael@0: *******************************************************************************
michael@0: *   file name:  uprops.h
michael@0: *   encoding:   US-ASCII
michael@0: *   tab size:   8 (not used)
michael@0: *   indentation:4
michael@0: *
michael@0: *   created on: 2002feb24
michael@0: *   created by: Markus W. Scherer
michael@0: *
michael@0: *   Constants for mostly non-core Unicode character properties
michael@0: *   stored in uprops.icu.
michael@0: */
michael@0: 
michael@0: #ifndef __UPROPS_H__
michael@0: #define __UPROPS_H__
michael@0: 
michael@0: #include "unicode/utypes.h"
michael@0: #include "unicode/uset.h"
michael@0: #include "uset_imp.h"
michael@0: #include "udataswp.h"
michael@0: 
michael@0: /* indexes[] entries */
michael@0: enum {
michael@0:     UPROPS_PROPS32_INDEX,
michael@0:     UPROPS_EXCEPTIONS_INDEX,
michael@0:     UPROPS_EXCEPTIONS_TOP_INDEX,
michael@0: 
michael@0:     UPROPS_ADDITIONAL_TRIE_INDEX,
michael@0:     UPROPS_ADDITIONAL_VECTORS_INDEX,
michael@0:     UPROPS_ADDITIONAL_VECTORS_COLUMNS_INDEX,
michael@0: 
michael@0:     UPROPS_SCRIPT_EXTENSIONS_INDEX,
michael@0: 
michael@0:     UPROPS_RESERVED_INDEX_7,
michael@0:     UPROPS_RESERVED_INDEX_8,
michael@0: 
michael@0:     /* size of the data file (number of 32-bit units after the header) */
michael@0:     UPROPS_DATA_TOP_INDEX,
michael@0: 
michael@0:     /* maximum values for code values in vector word 0 */
michael@0:     UPROPS_MAX_VALUES_INDEX=10,
michael@0:     /* maximum values for code values in vector word 2 */
michael@0:     UPROPS_MAX_VALUES_2_INDEX,
michael@0: 
michael@0:     UPROPS_INDEX_COUNT=16
michael@0: };
michael@0: 
michael@0: /* definitions for the main properties words */
michael@0: enum {
michael@0:     /* general category shift==0                                0 (5 bits) */
michael@0:     /* reserved                                                 5 (1 bit) */
michael@0:     UPROPS_NUMERIC_TYPE_VALUE_SHIFT=6                       /*  6 (10 bits) */
michael@0: };
michael@0: 
michael@0: #define GET_CATEGORY(props) ((props)&0x1f)
michael@0: #define CAT_MASK(props) U_MASK(GET_CATEGORY(props))
michael@0: 
michael@0: #define GET_NUMERIC_TYPE_VALUE(props) ((props)>>UPROPS_NUMERIC_TYPE_VALUE_SHIFT)
michael@0: 
michael@0: /* constants for the storage form of numeric types and values */
michael@0: enum {
michael@0:     /** No numeric value. */
michael@0:     UPROPS_NTV_NONE=0,
michael@0:     /** Decimal digits: nv=0..9 */
michael@0:     UPROPS_NTV_DECIMAL_START=1,
michael@0:     /** Other digits: nv=0..9 */
michael@0:     UPROPS_NTV_DIGIT_START=11,
michael@0:     /** Small integers: nv=0..154 */
michael@0:     UPROPS_NTV_NUMERIC_START=21,
michael@0:     /** Fractions: ((ntv>>4)-12) / ((ntv&0xf)+1) = -1..17 / 1..16 */
michael@0:     UPROPS_NTV_FRACTION_START=0xb0,
michael@0:     /**
michael@0:      * Large integers:
michael@0:      * ((ntv>>5)-14) * 10^((ntv&0x1f)+2) = (1..9)*(10^2..10^33)
michael@0:      * (only one significant decimal digit)
michael@0:      */
michael@0:     UPROPS_NTV_LARGE_START=0x1e0,
michael@0:     /**
michael@0:      * Sexagesimal numbers:
michael@0:      * ((ntv>>2)-0xbf) * 60^((ntv&3)+1) = (1..9)*(60^1..60^4)
michael@0:      */
michael@0:     UPROPS_NTV_BASE60_START=0x300,
michael@0:     /** No numeric value (yet). */
michael@0:     UPROPS_NTV_RESERVED_START=UPROPS_NTV_BASE60_START+36,  /* 0x300+9*4=0x324 */
michael@0: 
michael@0:     UPROPS_NTV_MAX_SMALL_INT=UPROPS_NTV_FRACTION_START-UPROPS_NTV_NUMERIC_START-1
michael@0: };
michael@0: 
michael@0: #define UPROPS_NTV_GET_TYPE(ntv) \
michael@0:     ((ntv==UPROPS_NTV_NONE) ? U_NT_NONE : \
michael@0:     (ntv<UPROPS_NTV_DIGIT_START) ?  U_NT_DECIMAL : \
michael@0:     (ntv<UPROPS_NTV_NUMERIC_START) ? U_NT_DIGIT : \
michael@0:     U_NT_NUMERIC)
michael@0: 
michael@0: /* number of properties vector words */
michael@0: #define UPROPS_VECTOR_WORDS     3
michael@0: 
michael@0: /*
michael@0:  * Properties in vector word 0
michael@0:  * Bits
michael@0:  * 31..24   DerivedAge version major/minor one nibble each
michael@0:  * 23..22   3..1: Bits 7..0 = Script_Extensions index
michael@0:  *             3: Script value from Script_Extensions
michael@0:  *             2: Script=Inherited
michael@0:  *             1: Script=Common
michael@0:  *             0: Script=bits 7..0
michael@0:  * 21..20   reserved
michael@0:  * 19..17   East Asian Width
michael@0:  * 16.. 8   UBlockCode
michael@0:  *  7.. 0   UScriptCode, or index to Script_Extensions
michael@0:  */
michael@0: 
michael@0: /* derived age: one nibble each for major and minor version numbers */
michael@0: #define UPROPS_AGE_MASK         0xff000000
michael@0: #define UPROPS_AGE_SHIFT        24
michael@0: 
michael@0: /* Script_Extensions: mask includes Script */
michael@0: #define UPROPS_SCRIPT_X_MASK    0x00c000ff
michael@0: #define UPROPS_SCRIPT_X_SHIFT   22
michael@0: 
michael@0: #define UPROPS_EA_MASK          0x000e0000
michael@0: #define UPROPS_EA_SHIFT         17
michael@0: 
michael@0: #define UPROPS_BLOCK_MASK       0x0001ff00
michael@0: #define UPROPS_BLOCK_SHIFT      8
michael@0: 
michael@0: #define UPROPS_SCRIPT_MASK      0x000000ff
michael@0: 
michael@0: /* UPROPS_SCRIPT_X_WITH_COMMON must be the lowest value that involves Script_Extensions. */
michael@0: #define UPROPS_SCRIPT_X_WITH_COMMON     0x400000
michael@0: #define UPROPS_SCRIPT_X_WITH_INHERITED  0x800000
michael@0: #define UPROPS_SCRIPT_X_WITH_OTHER      0xc00000
michael@0: 
michael@0: /*
michael@0:  * Properties in vector word 1
michael@0:  * Each bit encodes one binary property.
michael@0:  * The following constants represent the bit number, use 1<<UPROPS_XYZ.
michael@0:  * UPROPS_BINARY_1_TOP<=32!
michael@0:  *
michael@0:  * Keep this list of property enums in sync with
michael@0:  * propListNames[] in icu/source/tools/genprops/props2.c!
michael@0:  *
michael@0:  * ICU 2.6/uprops format version 3.2 stores full properties instead of "Other_".
michael@0:  */
michael@0: enum {
michael@0:     UPROPS_WHITE_SPACE,
michael@0:     UPROPS_DASH,
michael@0:     UPROPS_HYPHEN,
michael@0:     UPROPS_QUOTATION_MARK,
michael@0:     UPROPS_TERMINAL_PUNCTUATION,
michael@0:     UPROPS_MATH,
michael@0:     UPROPS_HEX_DIGIT,
michael@0:     UPROPS_ASCII_HEX_DIGIT,
michael@0:     UPROPS_ALPHABETIC,
michael@0:     UPROPS_IDEOGRAPHIC,
michael@0:     UPROPS_DIACRITIC,
michael@0:     UPROPS_EXTENDER,
michael@0:     UPROPS_NONCHARACTER_CODE_POINT,
michael@0:     UPROPS_GRAPHEME_EXTEND,
michael@0:     UPROPS_GRAPHEME_LINK,
michael@0:     UPROPS_IDS_BINARY_OPERATOR,
michael@0:     UPROPS_IDS_TRINARY_OPERATOR,
michael@0:     UPROPS_RADICAL,
michael@0:     UPROPS_UNIFIED_IDEOGRAPH,
michael@0:     UPROPS_DEFAULT_IGNORABLE_CODE_POINT,
michael@0:     UPROPS_DEPRECATED,
michael@0:     UPROPS_LOGICAL_ORDER_EXCEPTION,
michael@0:     UPROPS_XID_START,
michael@0:     UPROPS_XID_CONTINUE,
michael@0:     UPROPS_ID_START,                            /* ICU 2.6, uprops format version 3.2 */
michael@0:     UPROPS_ID_CONTINUE,
michael@0:     UPROPS_GRAPHEME_BASE,
michael@0:     UPROPS_S_TERM,                              /* new in ICU 3.0 and Unicode 4.0.1 */
michael@0:     UPROPS_VARIATION_SELECTOR,
michael@0:     UPROPS_PATTERN_SYNTAX,                      /* new in ICU 3.4 and Unicode 4.1 */
michael@0:     UPROPS_PATTERN_WHITE_SPACE,
michael@0:     UPROPS_RESERVED,                            /* reserved & unused */
michael@0:     UPROPS_BINARY_1_TOP                         /* ==32 - full! */
michael@0: };
michael@0: 
michael@0: /*
michael@0:  * Properties in vector word 2
michael@0:  * Bits
michael@0:  * 31..26   reserved
michael@0:  * 25..20   Line Break
michael@0:  * 19..15   Sentence Break
michael@0:  * 14..10   Word Break
michael@0:  *  9.. 5   Grapheme Cluster Break
michael@0:  *  4.. 0   Decomposition Type
michael@0:  */
michael@0: #define UPROPS_LB_MASK          0x03f00000
michael@0: #define UPROPS_LB_SHIFT         20
michael@0: 
michael@0: #define UPROPS_SB_MASK          0x000f8000
michael@0: #define UPROPS_SB_SHIFT         15
michael@0: 
michael@0: #define UPROPS_WB_MASK          0x00007c00
michael@0: #define UPROPS_WB_SHIFT         10
michael@0: 
michael@0: #define UPROPS_GCB_MASK         0x000003e0
michael@0: #define UPROPS_GCB_SHIFT        5
michael@0: 
michael@0: #define UPROPS_DT_MASK          0x0000001f
michael@0: 
michael@0: /**
michael@0:  * Gets the main properties value for a code point.
michael@0:  * Implemented in uchar.c for uprops.cpp.
michael@0:  */
michael@0: U_CFUNC uint32_t
michael@0: u_getMainProperties(UChar32 c);
michael@0: 
michael@0: /**
michael@0:  * Get a properties vector word for a code point.
michael@0:  * Implemented in uchar.c for uprops.cpp.
michael@0:  * @return 0 if no data or illegal argument
michael@0:  */
michael@0: U_CFUNC uint32_t
michael@0: u_getUnicodeProperties(UChar32 c, int32_t column);
michael@0: 
michael@0: /**
michael@0:  * Get the the maximum values for some enum/int properties.
michael@0:  * Use the same column numbers as for u_getUnicodeProperties().
michael@0:  * The returned value will contain maximum values stored in the same bit fields
michael@0:  * as where the enum values are stored in the u_getUnicodeProperties()
michael@0:  * return values for the same columns.
michael@0:  *
michael@0:  * Valid columns are those for properties words that contain enumerated values.
michael@0:  * (ICU 2.6: columns 0 and 2)
michael@0:  * For other column numbers, this function will return 0.
michael@0:  *
michael@0:  * @internal
michael@0:  */
michael@0: U_CFUNC int32_t
michael@0: uprv_getMaxValues(int32_t column);
michael@0: 
michael@0: /**
michael@0:  * Checks if c is alphabetic, or a decimal digit; implements UCHAR_POSIX_ALNUM.
michael@0:  * @internal
michael@0:  */
michael@0: U_CFUNC UBool
michael@0: u_isalnumPOSIX(UChar32 c);
michael@0: 
michael@0: /**
michael@0:  * Checks if c is in
michael@0:  * [^\p{space}\p{gc=Control}\p{gc=Surrogate}\p{gc=Unassigned}]
michael@0:  * with space=\p{Whitespace} and Control=Cc.
michael@0:  * Implements UCHAR_POSIX_GRAPH.
michael@0:  * @internal
michael@0:  */
michael@0: U_CFUNC UBool
michael@0: u_isgraphPOSIX(UChar32 c);
michael@0: 
michael@0: /**
michael@0:  * Checks if c is in \p{graph}\p{blank} - \p{cntrl}.
michael@0:  * Implements UCHAR_POSIX_PRINT.
michael@0:  * @internal
michael@0:  */
michael@0: U_CFUNC UBool
michael@0: u_isprintPOSIX(UChar32 c);
michael@0: 
michael@0: /** Turn a bit index into a bit flag. @internal */
michael@0: #define FLAG(n) ((uint32_t)1<<(n))
michael@0: 
michael@0: /** Flags for general categories in the order of UCharCategory. @internal */
michael@0: #define _Cn     FLAG(U_GENERAL_OTHER_TYPES)
michael@0: #define _Lu     FLAG(U_UPPERCASE_LETTER)
michael@0: #define _Ll     FLAG(U_LOWERCASE_LETTER)
michael@0: #define _Lt     FLAG(U_TITLECASE_LETTER)
michael@0: #define _Lm     FLAG(U_MODIFIER_LETTER)
michael@0: /* #define _Lo     FLAG(U_OTHER_LETTER) -- conflicts with MS Visual Studio 9.0 xiosbase */
michael@0: #define _Mn     FLAG(U_NON_SPACING_MARK)
michael@0: #define _Me     FLAG(U_ENCLOSING_MARK)
michael@0: #define _Mc     FLAG(U_COMBINING_SPACING_MARK)
michael@0: #define _Nd     FLAG(U_DECIMAL_DIGIT_NUMBER)
michael@0: #define _Nl     FLAG(U_LETTER_NUMBER)
michael@0: #define _No     FLAG(U_OTHER_NUMBER)
michael@0: #define _Zs     FLAG(U_SPACE_SEPARATOR)
michael@0: #define _Zl     FLAG(U_LINE_SEPARATOR)
michael@0: #define _Zp     FLAG(U_PARAGRAPH_SEPARATOR)
michael@0: #define _Cc     FLAG(U_CONTROL_CHAR)
michael@0: #define _Cf     FLAG(U_FORMAT_CHAR)
michael@0: #define _Co     FLAG(U_PRIVATE_USE_CHAR)
michael@0: #define _Cs     FLAG(U_SURROGATE)
michael@0: #define _Pd     FLAG(U_DASH_PUNCTUATION)
michael@0: #define _Ps     FLAG(U_START_PUNCTUATION)
michael@0: /* #define _Pe     FLAG(U_END_PUNCTUATION) -- conflicts with MS Visual Studio 9.0 xlocnum */
michael@0: /* #define _Pc     FLAG(U_CONNECTOR_PUNCTUATION) -- conflicts with MS Visual Studio 9.0 streambuf */
michael@0: #define _Po     FLAG(U_OTHER_PUNCTUATION)
michael@0: #define _Sm     FLAG(U_MATH_SYMBOL)
michael@0: #define _Sc     FLAG(U_CURRENCY_SYMBOL)
michael@0: #define _Sk     FLAG(U_MODIFIER_SYMBOL)
michael@0: #define _So     FLAG(U_OTHER_SYMBOL)
michael@0: #define _Pi     FLAG(U_INITIAL_PUNCTUATION)
michael@0: /* #define _Pf     FLAG(U_FINAL_PUNCTUATION) -- conflicts with MS Visual Studio 9.0 streambuf */
michael@0: 
michael@0: /** Some code points. @internal */
michael@0: enum {
michael@0:     TAB     =0x0009,
michael@0:     LF      =0x000a,
michael@0:     FF      =0x000c,
michael@0:     CR      =0x000d,
michael@0:     U_A     =0x0041,
michael@0:     U_F     =0x0046,
michael@0:     U_Z     =0x005a,
michael@0:     U_a     =0x0061,
michael@0:     U_f     =0x0066,
michael@0:     U_z     =0x007a,
michael@0:     DEL     =0x007f,
michael@0:     NL      =0x0085,
michael@0:     NBSP    =0x00a0,
michael@0:     CGJ     =0x034f,
michael@0:     FIGURESP=0x2007,
michael@0:     HAIRSP  =0x200a,
michael@0:     ZWNJ    =0x200c,
michael@0:     ZWJ     =0x200d,
michael@0:     RLM     =0x200f,
michael@0:     NNBSP   =0x202f,
michael@0:     WJ      =0x2060,
michael@0:     INHSWAP =0x206a,
michael@0:     NOMDIG  =0x206f,
michael@0:     U_FW_A  =0xff21,
michael@0:     U_FW_F  =0xff26,
michael@0:     U_FW_Z  =0xff3a,
michael@0:     U_FW_a  =0xff41,
michael@0:     U_FW_f  =0xff46,
michael@0:     U_FW_z  =0xff5a,
michael@0:     ZWNBSP  =0xfeff
michael@0: };
michael@0: 
michael@0: /**
michael@0:  * Get the maximum length of a (regular/1.0/extended) character name.
michael@0:  * @return 0 if no character names available.
michael@0:  */
michael@0: U_CAPI int32_t U_EXPORT2
michael@0: uprv_getMaxCharNameLength(void);
michael@0: 
michael@0: /**
michael@0:  * Fills set with characters that are used in Unicode character names.
michael@0:  * Includes all characters that are used in regular/Unicode 1.0/extended names.
michael@0:  * Just empties the set if no character names are available.
michael@0:  * @param sa USetAdder to receive characters.
michael@0:  */
michael@0: U_CAPI void U_EXPORT2
michael@0: uprv_getCharNameCharacters(const USetAdder *sa);
michael@0: 
michael@0: /**
michael@0:  * Constants for which data and implementation files provide which properties.
michael@0:  * Used by UnicodeSet for service-specific property enumeration.
michael@0:  * @internal
michael@0:  */
michael@0: enum UPropertySource {
michael@0:     /** No source, not a supported property. */
michael@0:     UPROPS_SRC_NONE,
michael@0:     /** From uchar.c/uprops.icu main trie */
michael@0:     UPROPS_SRC_CHAR,
michael@0:     /** From uchar.c/uprops.icu properties vectors trie */
michael@0:     UPROPS_SRC_PROPSVEC,
michael@0:     /** From unames.c/unames.icu */
michael@0:     UPROPS_SRC_NAMES,
michael@0:     /** From ucase.c/ucase.icu */
michael@0:     UPROPS_SRC_CASE,
michael@0:     /** From ubidi_props.c/ubidi.icu */
michael@0:     UPROPS_SRC_BIDI,
michael@0:     /** From uchar.c/uprops.icu main trie as well as properties vectors trie */
michael@0:     UPROPS_SRC_CHAR_AND_PROPSVEC,
michael@0:     /** From ucase.c/ucase.icu as well as unorm.cpp/unorm.icu */
michael@0:     UPROPS_SRC_CASE_AND_NORM,
michael@0:     /** From normalizer2impl.cpp/nfc.nrm */
michael@0:     UPROPS_SRC_NFC,
michael@0:     /** From normalizer2impl.cpp/nfkc.nrm */
michael@0:     UPROPS_SRC_NFKC,
michael@0:     /** From normalizer2impl.cpp/nfkc_cf.nrm */
michael@0:     UPROPS_SRC_NFKC_CF,
michael@0:     /** From normalizer2impl.cpp/nfc.nrm canonical iterator data */
michael@0:     UPROPS_SRC_NFC_CANON_ITER,
michael@0:     /** One more than the highest UPropertySource (UPROPS_SRC_) constant. */
michael@0:     UPROPS_SRC_COUNT
michael@0: };
michael@0: typedef enum UPropertySource UPropertySource;
michael@0: 
michael@0: /**
michael@0:  * @see UPropertySource
michael@0:  * @internal
michael@0:  */
michael@0: U_CFUNC UPropertySource U_EXPORT2
michael@0: uprops_getSource(UProperty which);
michael@0: 
michael@0: /**
michael@0:  * Enumerate uprops.icu's main data trie and add the
michael@0:  * start of each range of same properties to the set.
michael@0:  * @internal
michael@0:  */
michael@0: U_CFUNC void U_EXPORT2
michael@0: uchar_addPropertyStarts(const USetAdder *sa, UErrorCode *pErrorCode);
michael@0: 
michael@0: /**
michael@0:  * Enumerate uprops.icu's properties vectors trie and add the
michael@0:  * start of each range of same properties to the set.
michael@0:  * @internal
michael@0:  */
michael@0: U_CFUNC void U_EXPORT2
michael@0: upropsvec_addPropertyStarts(const USetAdder *sa, UErrorCode *pErrorCode);
michael@0: 
michael@0: /**
michael@0:  * Return a set of characters for property enumeration.
michael@0:  * For each two consecutive characters (start, limit) in the set,
michael@0:  * all of the properties for start..limit-1 are all the same.
michael@0:  *
michael@0:  * @param sa USetAdder to receive result. Existing contents are lost.
michael@0:  * @internal
michael@0:  */
michael@0: /*U_CFUNC void U_EXPORT2
michael@0: uprv_getInclusions(const USetAdder *sa, UErrorCode *pErrorCode);
michael@0: */
michael@0: 
michael@0: /**
michael@0:  * Swap the ICU Unicode character names file. See uchar.c.
michael@0:  * @internal
michael@0:  */
michael@0: U_CAPI int32_t U_EXPORT2
michael@0: uchar_swapNames(const UDataSwapper *ds,
michael@0:                 const void *inData, int32_t length, void *outData,
michael@0:                 UErrorCode *pErrorCode);
michael@0: 
michael@0: #ifdef __cplusplus
michael@0: 
michael@0: U_NAMESPACE_BEGIN
michael@0: 
michael@0: class UnicodeSet;
michael@0: 
michael@0: // implemented in uniset_props.cpp
michael@0: U_CFUNC UnicodeSet *
michael@0: uniset_getUnicode32Instance(UErrorCode &errorCode);
michael@0: 
michael@0: U_NAMESPACE_END
michael@0: 
michael@0: #endif
michael@0: 
michael@0: #endif