sl@0: /*
sl@0: *******************************************************************************
sl@0: *
sl@0: *   Copyright (C) 2002-2005, International Business Machines
sl@0: *   Corporation and others.  All Rights Reserved.
sl@0: *
sl@0: *******************************************************************************
sl@0: *   file name:  uprops.h
sl@0: *   encoding:   US-ASCII
sl@0: *   tab size:   8 (not used)
sl@0: *   indentation:4
sl@0: *
sl@0: *   created on: 2002feb24
sl@0: *   created by: Markus W. Scherer
sl@0: *
sl@0: *   Constants for mostly non-core Unicode character properties
sl@0: *   stored in uprops.icu.
sl@0: */
sl@0: 
sl@0: #ifndef __UPROPS_H__
sl@0: #define __UPROPS_H__
sl@0: 
sl@0: #include "unicode/utypes.h"
sl@0: #include "unicode/uset.h"
sl@0: #include "uset_imp.h"
sl@0: #include "udataswp.h"
sl@0: 
sl@0: /* indexes[] entries */
sl@0: enum {
sl@0:     UPROPS_PROPS32_INDEX,
sl@0:     UPROPS_EXCEPTIONS_INDEX,
sl@0:     UPROPS_EXCEPTIONS_TOP_INDEX,
sl@0: 
sl@0:     UPROPS_ADDITIONAL_TRIE_INDEX,
sl@0:     UPROPS_ADDITIONAL_VECTORS_INDEX,
sl@0:     UPROPS_ADDITIONAL_VECTORS_COLUMNS_INDEX,
sl@0: 
sl@0:     UPROPS_RESERVED_INDEX, /* 6 */
sl@0: 
sl@0:     /* maximum values for code values in vector word 0 */
sl@0:     UPROPS_MAX_VALUES_INDEX=10,
sl@0:     /* maximum values for code values in vector word 2 */
sl@0:     UPROPS_MAX_VALUES_2_INDEX,
sl@0: 
sl@0:     UPROPS_INDEX_COUNT=16
sl@0: };
sl@0: 
sl@0: /* definitions for the main properties words */
sl@0: enum {
sl@0:     /* general category shift==0                                0 (5 bits) */
sl@0:     UPROPS_NUMERIC_TYPE_SHIFT=5,                            /*  5 (3 bits) */
sl@0:     UPROPS_NUMERIC_VALUE_SHIFT=8                            /*  8 (8 bits) */
sl@0: };
sl@0: 
sl@0: #define GET_CATEGORY(props) ((props)&0x1f)
sl@0: #define CAT_MASK(props) U_MASK(GET_CATEGORY(props))
sl@0: 
sl@0: #define GET_NUMERIC_TYPE(props) (((props)>>UPROPS_NUMERIC_TYPE_SHIFT)&7)
sl@0: #define GET_NUMERIC_VALUE(props) (((props)>>UPROPS_NUMERIC_VALUE_SHIFT)&0xff)
sl@0: 
sl@0: /* internal numeric pseudo-types for special encodings of numeric values */
sl@0: enum {
sl@0:     UPROPS_NT_FRACTION=4, /* ==U_NT_COUNT, must not change unless binary format version changes */
sl@0:     UPROPS_NT_LARGE,
sl@0:     UPROPS_NT_COUNT
sl@0: };
sl@0: 
sl@0: /* encoding of fractional and large numbers */
sl@0: enum {
sl@0:     UPROPS_MAX_SMALL_NUMBER=0xff,
sl@0: 
sl@0:     UPROPS_FRACTION_NUM_SHIFT=3,        /* numerator: bits 7..3 */
sl@0:     UPROPS_FRACTION_DEN_MASK=7,         /* denominator: bits 2..0 */
sl@0: 
sl@0:     UPROPS_FRACTION_MAX_NUM=31,
sl@0:     UPROPS_FRACTION_DEN_OFFSET=2,       /* denominator values are 2..9 */
sl@0: 
sl@0:     UPROPS_FRACTION_MIN_DEN=UPROPS_FRACTION_DEN_OFFSET,
sl@0:     UPROPS_FRACTION_MAX_DEN=UPROPS_FRACTION_MIN_DEN+UPROPS_FRACTION_DEN_MASK,
sl@0: 
sl@0:     UPROPS_LARGE_MANT_SHIFT=4,          /* mantissa: bits 7..4 */
sl@0:     UPROPS_LARGE_EXP_MASK=0xf,          /* exponent: bits 3..0 */
sl@0:     UPROPS_LARGE_EXP_OFFSET=2,          /* regular exponents 2..17 */
sl@0:     UPROPS_LARGE_EXP_OFFSET_EXTRA=18,   /* extra large exponents 18..33 */
sl@0: 
sl@0:     UPROPS_LARGE_MIN_EXP=UPROPS_LARGE_EXP_OFFSET,
sl@0:     UPROPS_LARGE_MAX_EXP=UPROPS_LARGE_MIN_EXP+UPROPS_LARGE_EXP_MASK,
sl@0:     UPROPS_LARGE_MAX_EXP_EXTRA=UPROPS_LARGE_EXP_OFFSET_EXTRA+UPROPS_LARGE_EXP_MASK
sl@0: };
sl@0: 
sl@0: /* number of properties vector words */
sl@0: #define UPROPS_VECTOR_WORDS     3
sl@0: 
sl@0: /*
sl@0:  * Properties in vector word 0
sl@0:  * Bits
sl@0:  * 31..24   DerivedAge version major/minor one nibble each
sl@0:  * 23..18   Line Break
sl@0:  * 17..15   East Asian Width
sl@0:  * 14.. 7   UBlockCode
sl@0:  *  6.. 0   UScriptCode
sl@0:  */
sl@0: 
sl@0: /* derived age: one nibble each for major and minor version numbers */
sl@0: #define UPROPS_AGE_MASK         0xff000000
sl@0: #define UPROPS_AGE_SHIFT        24
sl@0: 
sl@0: #define UPROPS_LB_MASK          0x00FC0000
sl@0: #define UPROPS_LB_SHIFT         18
sl@0: 
sl@0: #define UPROPS_EA_MASK          0x00038000
sl@0: #define UPROPS_EA_SHIFT         15
sl@0: 
sl@0: #define UPROPS_BLOCK_MASK       0x00007f80
sl@0: #define UPROPS_BLOCK_SHIFT      7
sl@0: 
sl@0: #define UPROPS_SCRIPT_MASK      0x0000007f
sl@0: 
sl@0: /*
sl@0:  * Properties in vector word 1
sl@0:  * Each bit encodes one binary property.
sl@0:  * The following constants represent the bit number, use 1<<UPROPS_XYZ.
sl@0:  * UPROPS_BINARY_1_TOP<=32!
sl@0:  *
sl@0:  * Keep this list of property enums in sync with
sl@0:  * propListNames[] in icu/source/tools/genprops/props2.c!
sl@0:  *
sl@0:  * ICU 2.6/uprops format version 3.2 stores full properties instead of "Other_".
sl@0:  */
sl@0: enum {
sl@0:     UPROPS_WHITE_SPACE,
sl@0:         UPROPS_WAS_BIDI_CONTROL,                /* reserved, was used in format version 3 */
sl@0:         UPROPS_WAS_JOIN_CONTROL,
sl@0:     UPROPS_DASH,
sl@0:     UPROPS_HYPHEN,
sl@0:     UPROPS_QUOTATION_MARK,
sl@0:     UPROPS_TERMINAL_PUNCTUATION,
sl@0:     UPROPS_MATH,
sl@0:     UPROPS_HEX_DIGIT,
sl@0:     UPROPS_ASCII_HEX_DIGIT,
sl@0:     UPROPS_ALPHABETIC,
sl@0:     UPROPS_IDEOGRAPHIC,
sl@0:     UPROPS_DIACRITIC,
sl@0:     UPROPS_EXTENDER,
sl@0:         UPROPS_WAS_LOWERCASE,                   /* reserved, was used in format version 3 */
sl@0:         UPROPS_WAS_UPPERCASE,
sl@0:     UPROPS_NONCHARACTER_CODE_POINT,
sl@0:     UPROPS_GRAPHEME_EXTEND,
sl@0:     UPROPS_GRAPHEME_LINK,
sl@0:     UPROPS_IDS_BINARY_OPERATOR,
sl@0:     UPROPS_IDS_TRINARY_OPERATOR,
sl@0:     UPROPS_RADICAL,
sl@0:     UPROPS_UNIFIED_IDEOGRAPH,
sl@0:     UPROPS_DEFAULT_IGNORABLE_CODE_POINT,
sl@0:     UPROPS_DEPRECATED,
sl@0:         UPROPS_WAS_SOFT_DOTTED,                 /* reserved, was used in format version 3 */
sl@0:     UPROPS_LOGICAL_ORDER_EXCEPTION,
sl@0:     UPROPS_XID_START,
sl@0:     UPROPS_XID_CONTINUE,
sl@0:     UPROPS_ID_START,                            /* ICU 2.6, uprops format version 3.2 */
sl@0:     UPROPS_ID_CONTINUE,
sl@0:     UPROPS_GRAPHEME_BASE,
sl@0:     UPROPS_BINARY_1_TOP                         /* ==32 - full! */
sl@0: };
sl@0: 
sl@0: /*
sl@0:  * Properties in vector word 2
sl@0:  * Bits
sl@0:  * 31..24   More binary properties
sl@0:  * 23..19   reserved
sl@0:  * 18..14   Sentence Break
sl@0:  * 13..10   Word Break
sl@0:  *  9.. 5   Grapheme Cluster Break
sl@0:  *  4.. 0   Decomposition Type
sl@0:  */
sl@0: #define UPROPS_SB_MASK          0x0007c000
sl@0: #define UPROPS_SB_SHIFT         14
sl@0: 
sl@0: #define UPROPS_WB_MASK          0x00003c00
sl@0: #define UPROPS_WB_SHIFT         10
sl@0: 
sl@0: #define UPROPS_GCB_MASK         0x000003e0
sl@0: #define UPROPS_GCB_SHIFT        5
sl@0: 
sl@0: #define UPROPS_DT_MASK          0x0000001f
sl@0: 
sl@0: enum {
sl@0:     UPROPS_V2_S_TERM=24,                        /* new in ICU 3.0 and Unicode 4.0.1 */
sl@0:     UPROPS_V2_VARIATION_SELECTOR,
sl@0:     UPROPS_V2_PATTERN_SYNTAX,                   /* new in ICU 3.4 and Unicode 4.1 */
sl@0:     UPROPS_V2_PATTERN_WHITE_SPACE,
sl@0:     UPROPS_V2_TOP                               /* must be <=32 */
sl@0: };
sl@0: 
sl@0: /**
sl@0:  * Get a properties vector word for a code point.
sl@0:  * Implemented in uchar.c for uprops.c.
sl@0:  * column==-1 gets the 32-bit main properties word instead.
sl@0:  * @return 0 if no data or illegal argument
sl@0:  */
sl@0: U_CFUNC uint32_t
sl@0: u_getUnicodeProperties(UChar32 c, int32_t column);
sl@0: 
sl@0: /**
sl@0:  * Get the the maximum values for some enum/int properties.
sl@0:  * Use the same column numbers as for u_getUnicodeProperties().
sl@0:  * The returned value will contain maximum values stored in the same bit fields
sl@0:  * as where the enum values are stored in the u_getUnicodeProperties()
sl@0:  * return values for the same columns.
sl@0:  *
sl@0:  * Valid columns are those for properties words that contain enumerated values.
sl@0:  * (ICU 2.6: columns 0 and 2)
sl@0:  * For other column numbers, this function will return 0.
sl@0:  *
sl@0:  * @internal
sl@0:  */
sl@0: U_CFUNC int32_t
sl@0: uprv_getMaxValues(int32_t column);
sl@0: 
sl@0: /**
sl@0:  * Get the Hangul Syllable Type for c.
sl@0:  * @internal
sl@0:  */
sl@0: U_CFUNC UHangulSyllableType
sl@0: uchar_getHST(UChar32 c);
sl@0: 
sl@0: /**
sl@0:  * Checks if c is alphabetic, or a decimal digit; implements UCHAR_POSIX_ALNUM.
sl@0:  * @internal
sl@0:  */
sl@0: U_CFUNC UBool
sl@0: u_isalnumPOSIX(UChar32 c);
sl@0: 
sl@0: /**
sl@0:  * Checks if c is in
sl@0:  * [^\p{space}\p{gc=Control}\p{gc=Surrogate}\p{gc=Unassigned}]
sl@0:  * with space=\p{Whitespace} and Control=Cc.
sl@0:  * Implements UCHAR_POSIX_GRAPH.
sl@0:  * @internal
sl@0:  */
sl@0: U_CFUNC UBool
sl@0: u_isgraphPOSIX(UChar32 c);
sl@0: 
sl@0: /**
sl@0:  * Checks if c is in \p{graph}\p{blank} - \p{cntrl}.
sl@0:  * Implements UCHAR_POSIX_PRINT.
sl@0:  * @internal
sl@0:  */
sl@0: U_CFUNC UBool
sl@0: u_isprintPOSIX(UChar32 c);
sl@0: 
sl@0: /** Turn a bit index into a bit flag. @internal */
sl@0: #define FLAG(n) ((uint32_t)1<<(n))
sl@0: 
sl@0: /** Flags for general categories in the order of UCharCategory. @internal */
sl@0: #define _Cn     FLAG(U_GENERAL_OTHER_TYPES)
sl@0: #define _Lu     FLAG(U_UPPERCASE_LETTER)
sl@0: #define _Ll     FLAG(U_LOWERCASE_LETTER)
sl@0: #define _Lt     FLAG(U_TITLECASE_LETTER)
sl@0: #define _Lm     FLAG(U_MODIFIER_LETTER)
sl@0: #define _Lo     FLAG(U_OTHER_LETTER)
sl@0: #define _Mn     FLAG(U_NON_SPACING_MARK)
sl@0: #define _Me     FLAG(U_ENCLOSING_MARK)
sl@0: #define _Mc     FLAG(U_COMBINING_SPACING_MARK)
sl@0: #define _Nd     FLAG(U_DECIMAL_DIGIT_NUMBER)
sl@0: #define _Nl     FLAG(U_LETTER_NUMBER)
sl@0: #define _No     FLAG(U_OTHER_NUMBER)
sl@0: #define _Zs     FLAG(U_SPACE_SEPARATOR)
sl@0: #define _Zl     FLAG(U_LINE_SEPARATOR)
sl@0: #define _Zp     FLAG(U_PARAGRAPH_SEPARATOR)
sl@0: #define _Cc     FLAG(U_CONTROL_CHAR)
sl@0: #define _Cf     FLAG(U_FORMAT_CHAR)
sl@0: #define _Co     FLAG(U_PRIVATE_USE_CHAR)
sl@0: #define _Cs     FLAG(U_SURROGATE)
sl@0: #define _Pd     FLAG(U_DASH_PUNCTUATION)
sl@0: #define _Ps     FLAG(U_START_PUNCTUATION)
sl@0: #define _Pe     FLAG(U_END_PUNCTUATION)
sl@0: #define _Pc     FLAG(U_CONNECTOR_PUNCTUATION)
sl@0: #define _Po     FLAG(U_OTHER_PUNCTUATION)
sl@0: #define _Sm     FLAG(U_MATH_SYMBOL)
sl@0: #define _Sc     FLAG(U_CURRENCY_SYMBOL)
sl@0: #define _Sk     FLAG(U_MODIFIER_SYMBOL)
sl@0: #define _So     FLAG(U_OTHER_SYMBOL)
sl@0: #define _Pi     FLAG(U_INITIAL_PUNCTUATION)
sl@0: #define _Pf     FLAG(U_FINAL_PUNCTUATION)
sl@0: 
sl@0: /** Some code points. @internal */
sl@0: enum {
sl@0:     TAB     =0x0009,
sl@0:     LF      =0x000a,
sl@0:     FF      =0x000c,
sl@0:     CR      =0x000d,
sl@0:     U_A     =0x0041,
sl@0:     U_F     =0x0046,
sl@0:     U_Z     =0x005a,
sl@0:     U_a     =0x0061,
sl@0:     U_f     =0x0066,
sl@0:     U_z     =0x007a,
sl@0:     DEL     =0x007f,
sl@0:     NL      =0x0085,
sl@0:     NBSP    =0x00a0,
sl@0:     CGJ     =0x034f,
sl@0:     FIGURESP=0x2007,
sl@0:     HAIRSP  =0x200a,
sl@0:     ZWNJ    =0x200c,
sl@0:     ZWJ     =0x200d,
sl@0:     RLM     =0x200f,
sl@0:     NNBSP   =0x202f,
sl@0:     WJ      =0x2060,
sl@0:     INHSWAP =0x206a,
sl@0:     NOMDIG  =0x206f,
sl@0:     U_FW_A  =0xff21,
sl@0:     U_FW_F  =0xff26,
sl@0:     U_FW_Z  =0xff3a,
sl@0:     U_FW_a  =0xff41,
sl@0:     U_FW_f  =0xff46,
sl@0:     U_FW_z  =0xff5a,
sl@0:     ZWNBSP  =0xfeff
sl@0: };
sl@0: 
sl@0: /**
sl@0:  * Get the maximum length of a (regular/1.0/extended) character name.
sl@0:  * @return 0 if no character names available.
sl@0:  */
sl@0: U_CAPI int32_t U_EXPORT2
sl@0: uprv_getMaxCharNameLength(void);
sl@0: 
sl@0: #if 0
sl@0: /* 
sl@0: Currently not used but left for future use. Probably by UnicodeSet. 
sl@0: urename.h and unames.c changed accordingly. 
sl@0: */
sl@0: /**
sl@0:  * Get the maximum length of an ISO comment.
sl@0:  * @return 0 if no ISO comments available.
sl@0:  */
sl@0: U_CAPI int32_t U_EXPORT2
sl@0: uprv_getMaxISOCommentLength();
sl@0: #endif
sl@0: 
sl@0: /**
sl@0:  * Fills set with characters that are used in Unicode character names.
sl@0:  * Includes all characters that are used in regular/Unicode 1.0/extended names.
sl@0:  * Just empties the set if no character names are available.
sl@0:  * @param sa USetAdder to receive characters.
sl@0:  */
sl@0: U_CAPI void U_EXPORT2
sl@0: uprv_getCharNameCharacters(const USetAdder *sa);
sl@0: 
sl@0: #if 0
sl@0: /* 
sl@0: Currently not used but left for future use. Probably by UnicodeSet. 
sl@0: urename.h and unames.c changed accordingly. 
sl@0: */
sl@0: /**
sl@0:  * Fills set with characters that are used in Unicode character names.
sl@0:  * Just empties the set if no ISO comments are available.
sl@0:  * @param sa USetAdder to receive characters.
sl@0:  */
sl@0: U_CAPI void U_EXPORT2
sl@0: uprv_getISOCommentCharacters(const USetAdder *sa);
sl@0: */
sl@0: #endif
sl@0: 
sl@0: /**
sl@0:  * Constants for which data and implementation files provide which properties.
sl@0:  * Used by UnicodeSet for service-specific property enumeration.
sl@0:  * @internal
sl@0:  */
sl@0: enum UPropertySource {
sl@0:     /** No source, not a supported property. */
sl@0:     UPROPS_SRC_NONE,
sl@0:     /** From uchar.c/uprops.icu main trie */
sl@0:     UPROPS_SRC_CHAR,
sl@0:     /** From uchar.c/uprops.icu properties vectors trie */
sl@0:     UPROPS_SRC_PROPSVEC,
sl@0:     /** Hangul_Syllable_Type, from uchar.c/uprops.icu */
sl@0:     UPROPS_SRC_HST,
sl@0:     /** From unames.c/unames.icu */
sl@0:     UPROPS_SRC_NAMES,
sl@0:     /** From unorm.cpp/unorm.icu */
sl@0:     UPROPS_SRC_NORM,
sl@0:     /** From ucase.c/ucase.icu */
sl@0:     UPROPS_SRC_CASE,
sl@0:     /** From ubidi_props.c/ubidi.icu */
sl@0:     UPROPS_SRC_BIDI,
sl@0:     /** From uchar.c/uprops.icu main trie as well as properties vectors trie */
sl@0:     UPROPS_SRC_CHAR_AND_PROPSVEC,
sl@0:     /** One more than the highest UPropertySource (UPROPS_SRC_) constant. */
sl@0:     UPROPS_SRC_COUNT
sl@0: };
sl@0: typedef enum UPropertySource UPropertySource;
sl@0: 
sl@0: /**
sl@0:  * @see UPropertySource
sl@0:  * @internal
sl@0:  */
sl@0: U_CAPI UPropertySource U_EXPORT2
sl@0: uprops_getSource(UProperty which);
sl@0: 
sl@0: /**
sl@0:  * Enumerate uprops.icu's main data trie and add the
sl@0:  * start of each range of same properties to the set.
sl@0:  * @internal
sl@0:  */
sl@0: U_CAPI void U_EXPORT2
sl@0: uchar_addPropertyStarts(const USetAdder *sa, UErrorCode *pErrorCode);
sl@0: 
sl@0: /**
sl@0:  * Enumerate uprops.icu's properties vectors trie and add the
sl@0:  * start of each range of same properties to the set.
sl@0:  * @internal
sl@0:  */
sl@0: U_CAPI void U_EXPORT2
sl@0: upropsvec_addPropertyStarts(const USetAdder *sa, UErrorCode *pErrorCode);
sl@0: 
sl@0: /**
sl@0:  * Same as uchar_addPropertyStarts() but only for Hangul_Syllable_Type.
sl@0:  * @internal
sl@0:  */
sl@0: U_CAPI void U_EXPORT2
sl@0: uhst_addPropertyStarts(const USetAdder *sa, UErrorCode *pErrorCode);
sl@0: 
sl@0: /**
sl@0:  * Return a set of characters for property enumeration.
sl@0:  * For each two consecutive characters (start, limit) in the set,
sl@0:  * all of the properties for start..limit-1 are all the same.
sl@0:  *
sl@0:  * @param sa USetAdder to receive result. Existing contents are lost.
sl@0:  * @internal
sl@0:  */
sl@0: U_CAPI void U_EXPORT2
sl@0: uprv_getInclusions(const USetAdder *sa, UErrorCode *pErrorCode);
sl@0: 
sl@0: /**
sl@0:  * Swap the ICU Unicode properties file. See uchar.c.
sl@0:  * @internal
sl@0:  */
sl@0: U_CAPI int32_t U_EXPORT2
sl@0: uprops_swap(const UDataSwapper *ds,
sl@0:             const void *inData, int32_t length, void *outData,
sl@0:             UErrorCode *pErrorCode);
sl@0: 
sl@0: /**
sl@0:  * Swap the ICU Unicode character names file. See uchar.c.
sl@0:  * @internal
sl@0:  */
sl@0: U_CAPI int32_t U_EXPORT2
sl@0: uchar_swapNames(const UDataSwapper *ds,
sl@0:                 const void *inData, int32_t length, void *outData,
sl@0:                 UErrorCode *pErrorCode);
sl@0: 
sl@0: #endif