michael@0: /*
michael@0: ******************************************************************************
michael@0: *
michael@0: *   Copyright (C) 2001-2011, International Business Machines
michael@0: *   Corporation and others.  All Rights Reserved.
michael@0: *
michael@0: ******************************************************************************
michael@0: *   file name:  utrie.h
michael@0: *   encoding:   US-ASCII
michael@0: *   tab size:   8 (not used)
michael@0: *   indentation:4
michael@0: *
michael@0: *   created on: 2001nov08
michael@0: *   created by: Markus W. Scherer
michael@0: */
michael@0: 
michael@0: #ifndef __UTRIE_H__
michael@0: #define __UTRIE_H__
michael@0: 
michael@0: #include "unicode/utypes.h"
michael@0: #include "unicode/utf16.h"
michael@0: #include "udataswp.h"
michael@0: 
michael@0: U_CDECL_BEGIN
michael@0: 
michael@0: /**
michael@0:  * \file
michael@0:  *
michael@0:  * This is a common implementation of a "folded" trie.
michael@0:  * It is a kind of compressed, serializable table of 16- or 32-bit values associated with
michael@0:  * Unicode code points (0..0x10ffff).
michael@0:  *
michael@0:  * This implementation is optimized for getting values while walking forward
michael@0:  * through a UTF-16 string.
michael@0:  * Therefore, the simplest and fastest access macros are the
michael@0:  * _FROM_LEAD() and _FROM_OFFSET_TRAIL() macros.
michael@0:  *
michael@0:  * The _FROM_BMP() macros are a little more complicated; they get values
michael@0:  * even for lead surrogate code _points_, while the _FROM_LEAD() macros
michael@0:  * get special "folded" values for lead surrogate code _units_ if
michael@0:  * there is relevant data associated with them.
michael@0:  * From such a folded value, an offset needs to be extracted to supply
michael@0:  * to the _FROM_OFFSET_TRAIL() macros.
michael@0:  *
michael@0:  * Most of the more complex (and more convenient) functions/macros call a callback function
michael@0:  * to get that offset from the folded value for a lead surrogate unit.
michael@0:  */
michael@0: 
michael@0: /**
michael@0:  * Trie constants, defining shift widths, index array lengths, etc.
michael@0:  */
michael@0: enum {
michael@0:     /** Shift size for shifting right the input index. 1..9 */
michael@0:     UTRIE_SHIFT=5,
michael@0: 
michael@0:     /** Number of data values in a stage 2 (data array) block. 2, 4, 8, .., 0x200 */
michael@0:     UTRIE_DATA_BLOCK_LENGTH=1<<UTRIE_SHIFT,
michael@0: 
michael@0:     /** Mask for getting the lower bits from the input index. */
michael@0:     UTRIE_MASK=UTRIE_DATA_BLOCK_LENGTH-1,
michael@0: 
michael@0:     /**
michael@0:      * Lead surrogate code points' index displacement in the index array.
michael@0:      * 0x10000-0xd800=0x2800
michael@0:      */
michael@0:     UTRIE_LEAD_INDEX_DISP=0x2800>>UTRIE_SHIFT,
michael@0: 
michael@0:     /**
michael@0:      * Shift size for shifting left the index array values.
michael@0:      * Increases possible data size with 16-bit index values at the cost
michael@0:      * of compactability.
michael@0:      * This requires blocks of stage 2 data to be aligned by UTRIE_DATA_GRANULARITY.
michael@0:      * 0..UTRIE_SHIFT
michael@0:      */
michael@0:     UTRIE_INDEX_SHIFT=2,
michael@0: 
michael@0:     /** The alignment size of a stage 2 data block. Also the granularity for compaction. */
michael@0:     UTRIE_DATA_GRANULARITY=1<<UTRIE_INDEX_SHIFT,
michael@0: 
michael@0:     /** Number of bits of a trail surrogate that are used in index table lookups. */
michael@0:     UTRIE_SURROGATE_BLOCK_BITS=10-UTRIE_SHIFT,
michael@0: 
michael@0:     /**
michael@0:      * Number of index (stage 1) entries per lead surrogate.
michael@0:      * Same as number of index entries for 1024 trail surrogates,
michael@0:      * ==0x400>>UTRIE_SHIFT
michael@0:      */
michael@0:     UTRIE_SURROGATE_BLOCK_COUNT=(1<<UTRIE_SURROGATE_BLOCK_BITS),
michael@0: 
michael@0:     /** Length of the BMP portion of the index (stage 1) array. */
michael@0:     UTRIE_BMP_INDEX_LENGTH=0x10000>>UTRIE_SHIFT
michael@0: };
michael@0: 
michael@0: /**
michael@0:  * Length of the index (stage 1) array before folding.
michael@0:  * Maximum number of Unicode code points (0x110000) shifted right by UTRIE_SHIFT.
michael@0:  */
michael@0: #define UTRIE_MAX_INDEX_LENGTH (0x110000>>UTRIE_SHIFT)
michael@0: 
michael@0: /**
michael@0:  * Maximum length of the runtime data (stage 2) array.
michael@0:  * Limited by 16-bit index values that are left-shifted by UTRIE_INDEX_SHIFT.
michael@0:  */
michael@0: #define UTRIE_MAX_DATA_LENGTH (0x10000<<UTRIE_INDEX_SHIFT)
michael@0: 
michael@0: /**
michael@0:  * Maximum length of the build-time data (stage 2) array.
michael@0:  * The maximum length is 0x110000+UTRIE_DATA_BLOCK_LENGTH+0x400.
michael@0:  * (Number of Unicode code points + one all-initial-value block +
michael@0:  *  possible duplicate entries for 1024 lead surrogates.)
michael@0:  */
michael@0: #define UTRIE_MAX_BUILD_TIME_DATA_LENGTH (0x110000+UTRIE_DATA_BLOCK_LENGTH+0x400)
michael@0: 
michael@0: /**
michael@0:  * Number of bytes for a dummy trie.
michael@0:  * A dummy trie is an empty runtime trie, used when a real data trie cannot
michael@0:  * be loaded.
michael@0:  * The number of bytes works for Latin-1-linear tries with 32-bit data
michael@0:  * (worst case).
michael@0:  *
michael@0:  * Calculation:
michael@0:  *   BMP index + 1 index block for lead surrogate code points +
michael@0:  *   Latin-1-linear array + 1 data block for lead surrogate code points
michael@0:  *
michael@0:  * Latin-1: if(UTRIE_SHIFT<=8) { 256 } else { included in first data block }
michael@0:  *
michael@0:  * @see utrie_unserializeDummy
michael@0:  */
michael@0: #define UTRIE_DUMMY_SIZE ((UTRIE_BMP_INDEX_LENGTH+UTRIE_SURROGATE_BLOCK_COUNT)*2+(UTRIE_SHIFT<=8?256:UTRIE_DATA_BLOCK_LENGTH)*4+UTRIE_DATA_BLOCK_LENGTH*4)
michael@0: 
michael@0: /**
michael@0:  * Runtime UTrie callback function.
michael@0:  * Extract from a lead surrogate's data the
michael@0:  * index array offset of the indexes for that lead surrogate.
michael@0:  *
michael@0:  * @param data data value for a surrogate from the trie, including the folding offset
michael@0:  * @return offset>=UTRIE_BMP_INDEX_LENGTH, or 0 if there is no data for the lead surrogate
michael@0:  */
michael@0: typedef int32_t U_CALLCONV
michael@0: UTrieGetFoldingOffset(uint32_t data);
michael@0: 
michael@0: /**
michael@0:  * Run-time Trie structure.
michael@0:  *
michael@0:  * Either the data table is 16 bits wide and accessed via the index
michael@0:  * pointer, with each index item increased by indexLength;
michael@0:  * in this case, data32==NULL.
michael@0:  *
michael@0:  * Or the data table is 32 bits wide and accessed via the data32 pointer.
michael@0:  */
michael@0: struct UTrie {
michael@0:     const uint16_t *index;
michael@0:     const uint32_t *data32; /* NULL if 16b data is used via index */
michael@0: 
michael@0:     /**
michael@0:      * This function is not used in _FROM_LEAD, _FROM_BMP, and _FROM_OFFSET_TRAIL macros.
michael@0:      * If convenience macros like _GET16 or _NEXT32 are used, this function must be set.
michael@0:      *
michael@0:      * utrie_unserialize() sets a default function which simply returns
michael@0:      * the lead surrogate's value itself - which is the inverse of the default
michael@0:      * folding function used by utrie_serialize().
michael@0:      *
michael@0:      * @see UTrieGetFoldingOffset
michael@0:      */
michael@0:     UTrieGetFoldingOffset *getFoldingOffset;
michael@0: 
michael@0:     int32_t indexLength, dataLength;
michael@0:     uint32_t initialValue;
michael@0:     UBool isLatin1Linear;
michael@0: };
michael@0: 
michael@0: #ifndef __UTRIE2_H__
michael@0: typedef struct UTrie UTrie;
michael@0: #endif
michael@0: 
michael@0: /** Internal trie getter from an offset (0 if c16 is a BMP/lead units) and a 16-bit unit */
michael@0: #define _UTRIE_GET_RAW(trie, data, offset, c16) \
michael@0:     (trie)->data[ \
michael@0:         ((int32_t)((trie)->index[(offset)+((c16)>>UTRIE_SHIFT)])<<UTRIE_INDEX_SHIFT)+ \
michael@0:         ((c16)&UTRIE_MASK) \
michael@0:     ]
michael@0: 
michael@0: /** Internal trie getter from a pair of surrogates */
michael@0: #define _UTRIE_GET_FROM_PAIR(trie, data, c, c2, result, resultType) { \
michael@0:     int32_t __offset; \
michael@0: \
michael@0:     /* get data for lead surrogate */ \
michael@0:     (result)=_UTRIE_GET_RAW((trie), data, 0, (c)); \
michael@0:     __offset=(trie)->getFoldingOffset(result); \
michael@0: \
michael@0:     /* get the real data from the folded lead/trail units */ \
michael@0:     if(__offset>0) { \
michael@0:         (result)=_UTRIE_GET_RAW((trie), data, __offset, (c2)&0x3ff); \
michael@0:     } else { \
michael@0:         (result)=(resultType)((trie)->initialValue); \
michael@0:     } \
michael@0: }
michael@0: 
michael@0: /** Internal trie getter from a BMP code point, treating a lead surrogate as a normal code point */
michael@0: #define _UTRIE_GET_FROM_BMP(trie, data, c16) \
michael@0:     _UTRIE_GET_RAW(trie, data, 0xd800<=(c16) && (c16)<=0xdbff ? UTRIE_LEAD_INDEX_DISP : 0, c16);
michael@0: 
michael@0: /**
michael@0:  * Internal trie getter from a code point.
michael@0:  * Could be faster(?) but longer with
michael@0:  *   if((c32)<=0xd7ff) { (result)=_UTRIE_GET_RAW(trie, data, 0, c32); }
michael@0:  */
michael@0: #define _UTRIE_GET(trie, data, c32, result, resultType) \
michael@0:     if((uint32_t)(c32)<=0xffff) { \
michael@0:         /* BMP code points */ \
michael@0:         (result)=_UTRIE_GET_FROM_BMP(trie, data, c32); \
michael@0:     } else if((uint32_t)(c32)<=0x10ffff) { \
michael@0:         /* supplementary code point */ \
michael@0:         UChar __lead16=U16_LEAD(c32); \
michael@0:         _UTRIE_GET_FROM_PAIR(trie, data, __lead16, c32, result, resultType); \
michael@0:     } else { \
michael@0:         /* out of range */ \
michael@0:         (result)=(resultType)((trie)->initialValue); \
michael@0:     }
michael@0: 
michael@0: /** Internal next-post-increment: get the next code point (c, c2) and its data */
michael@0: #define _UTRIE_NEXT(trie, data, src, limit, c, c2, result, resultType) { \
michael@0:     (c)=*(src)++; \
michael@0:     if(!U16_IS_LEAD(c)) { \
michael@0:         (c2)=0; \
michael@0:         (result)=_UTRIE_GET_RAW((trie), data, 0, (c)); \
michael@0:     } else if((src)!=(limit) && U16_IS_TRAIL((c2)=*(src))) { \
michael@0:         ++(src); \
michael@0:         _UTRIE_GET_FROM_PAIR((trie), data, (c), (c2), (result), resultType); \
michael@0:     } else { \
michael@0:         /* unpaired lead surrogate code point */ \
michael@0:         (c2)=0; \
michael@0:         (result)=_UTRIE_GET_RAW((trie), data, UTRIE_LEAD_INDEX_DISP, (c)); \
michael@0:     } \
michael@0: }
michael@0: 
michael@0: /** Internal previous: get the previous code point (c, c2) and its data */
michael@0: #define _UTRIE_PREVIOUS(trie, data, start, src, c, c2, result, resultType) { \
michael@0:     (c)=*--(src); \
michael@0:     if(!U16_IS_SURROGATE(c)) { \
michael@0:         (c2)=0; \
michael@0:         (result)=_UTRIE_GET_RAW((trie), data, 0, (c)); \
michael@0:     } else if(!U16_IS_SURROGATE_LEAD(c)) { \
michael@0:         /* trail surrogate */ \
michael@0:         if((start)!=(src) && U16_IS_LEAD((c2)=*((src)-1))) { \
michael@0:             --(src); \
michael@0:             (result)=(c); (c)=(c2); (c2)=(UChar)(result); /* swap c, c2 */ \
michael@0:             _UTRIE_GET_FROM_PAIR((trie), data, (c), (c2), (result), resultType); \
michael@0:         } else { \
michael@0:             /* unpaired trail surrogate code point */ \
michael@0:             (c2)=0; \
michael@0:             (result)=_UTRIE_GET_RAW((trie), data, 0, (c)); \
michael@0:         } \
michael@0:     } else { \
michael@0:         /* unpaired lead surrogate code point */ \
michael@0:         (c2)=0; \
michael@0:         (result)=_UTRIE_GET_RAW((trie), data, UTRIE_LEAD_INDEX_DISP, (c)); \
michael@0:     } \
michael@0: }
michael@0: 
michael@0: /* Public UTrie API ---------------------------------------------------------*/
michael@0: 
michael@0: /**
michael@0:  * Get a pointer to the contiguous part of the data array
michael@0:  * for the Latin-1 range (U+0000..U+00ff).
michael@0:  * Must be used only if the Latin-1 range is in fact linear
michael@0:  * (trie->isLatin1Linear).
michael@0:  *
michael@0:  * @param trie (const UTrie *, in) a pointer to the runtime trie structure
michael@0:  * @return (const uint16_t *) pointer to values for Latin-1 code points
michael@0:  */
michael@0: #define UTRIE_GET16_LATIN1(trie) ((trie)->index+(trie)->indexLength+UTRIE_DATA_BLOCK_LENGTH)
michael@0: 
michael@0: /**
michael@0:  * Get a pointer to the contiguous part of the data array
michael@0:  * for the Latin-1 range (U+0000..U+00ff).
michael@0:  * Must be used only if the Latin-1 range is in fact linear
michael@0:  * (trie->isLatin1Linear).
michael@0:  *
michael@0:  * @param trie (const UTrie *, in) a pointer to the runtime trie structure
michael@0:  * @return (const uint32_t *) pointer to values for Latin-1 code points
michael@0:  */
michael@0: #define UTRIE_GET32_LATIN1(trie) ((trie)->data32+UTRIE_DATA_BLOCK_LENGTH)
michael@0: 
michael@0: /**
michael@0:  * Get a 16-bit trie value from a BMP code point (UChar, <=U+ffff).
michael@0:  * c16 may be a lead surrogate, which may have a value including a folding offset.
michael@0:  *
michael@0:  * @param trie (const UTrie *, in) a pointer to the runtime trie structure
michael@0:  * @param c16 (UChar, in) the input BMP code point
michael@0:  * @return (uint16_t) trie lookup result
michael@0:  */
michael@0: #define UTRIE_GET16_FROM_LEAD(trie, c16) _UTRIE_GET_RAW(trie, index, 0, c16)
michael@0: 
michael@0: /**
michael@0:  * Get a 32-bit trie value from a BMP code point (UChar, <=U+ffff).
michael@0:  * c16 may be a lead surrogate, which may have a value including a folding offset.
michael@0:  *
michael@0:  * @param trie (const UTrie *, in) a pointer to the runtime trie structure
michael@0:  * @param c16 (UChar, in) the input BMP code point
michael@0:  * @return (uint32_t) trie lookup result
michael@0:  */
michael@0: #define UTRIE_GET32_FROM_LEAD(trie, c16) _UTRIE_GET_RAW(trie, data32, 0, c16)
michael@0: 
michael@0: /**
michael@0:  * Get a 16-bit trie value from a BMP code point (UChar, <=U+ffff).
michael@0:  * Even lead surrogate code points are treated as normal code points,
michael@0:  * with unfolded values that may differ from _FROM_LEAD() macro results for them.
michael@0:  *
michael@0:  * @param trie (const UTrie *, in) a pointer to the runtime trie structure
michael@0:  * @param c16 (UChar, in) the input BMP code point
michael@0:  * @return (uint16_t) trie lookup result
michael@0:  */
michael@0: #define UTRIE_GET16_FROM_BMP(trie, c16) _UTRIE_GET_FROM_BMP(trie, index, c16)
michael@0: 
michael@0: /**
michael@0:  * Get a 32-bit trie value from a BMP code point (UChar, <=U+ffff).
michael@0:  * Even lead surrogate code points are treated as normal code points,
michael@0:  * with unfolded values that may differ from _FROM_LEAD() macro results for them.
michael@0:  *
michael@0:  * @param trie (const UTrie *, in) a pointer to the runtime trie structure
michael@0:  * @param c16 (UChar, in) the input BMP code point
michael@0:  * @return (uint32_t) trie lookup result
michael@0:  */
michael@0: #define UTRIE_GET32_FROM_BMP(trie, c16) _UTRIE_GET_FROM_BMP(trie, data32, c16)
michael@0: 
michael@0: /**
michael@0:  * Get a 16-bit trie value from a code point.
michael@0:  * Even lead surrogate code points are treated as normal code points,
michael@0:  * with unfolded values that may differ from _FROM_LEAD() macro results for them.
michael@0:  *
michael@0:  * @param trie (const UTrie *, in) a pointer to the runtime trie structure
michael@0:  * @param c32 (UChar32, in) the input code point
michael@0:  * @param result (uint16_t, out) uint16_t variable for the trie lookup result
michael@0:  */
michael@0: #define UTRIE_GET16(trie, c32, result) _UTRIE_GET(trie, index, c32, result, uint16_t)
michael@0: 
michael@0: /**
michael@0:  * Get a 32-bit trie value from a code point.
michael@0:  * Even lead surrogate code points are treated as normal code points,
michael@0:  * with unfolded values that may differ from _FROM_LEAD() macro results for them.
michael@0:  *
michael@0:  * @param trie (const UTrie *, in) a pointer to the runtime trie structure
michael@0:  * @param c32 (UChar32, in) the input code point
michael@0:  * @param result (uint32_t, out) uint32_t variable for the trie lookup result
michael@0:  */
michael@0: #define UTRIE_GET32(trie, c32, result) _UTRIE_GET(trie, data32, c32, result, uint32_t)
michael@0: 
michael@0: /**
michael@0:  * Get the next code point (c, c2), post-increment src,
michael@0:  * and get a 16-bit value from the trie.
michael@0:  *
michael@0:  * @param trie (const UTrie *, in) a pointer to the runtime trie structure
michael@0:  * @param src (const UChar *, in/out) the source text pointer
michael@0:  * @param limit (const UChar *, in) the limit pointer for the text, or NULL
michael@0:  * @param c (UChar, out) variable for the BMP or lead code unit
michael@0:  * @param c2 (UChar, out) variable for 0 or the trail code unit
michael@0:  * @param result (uint16_t, out) uint16_t variable for the trie lookup result
michael@0:  */
michael@0: #define UTRIE_NEXT16(trie, src, limit, c, c2, result) _UTRIE_NEXT(trie, index, src, limit, c, c2, result, uint16_t)
michael@0: 
michael@0: /**
michael@0:  * Get the next code point (c, c2), post-increment src,
michael@0:  * and get a 32-bit value from the trie.
michael@0:  *
michael@0:  * @param trie (const UTrie *, in) a pointer to the runtime trie structure
michael@0:  * @param src (const UChar *, in/out) the source text pointer
michael@0:  * @param limit (const UChar *, in) the limit pointer for the text, or NULL
michael@0:  * @param c (UChar, out) variable for the BMP or lead code unit
michael@0:  * @param c2 (UChar, out) variable for 0 or the trail code unit
michael@0:  * @param result (uint32_t, out) uint32_t variable for the trie lookup result
michael@0:  */
michael@0: #define UTRIE_NEXT32(trie, src, limit, c, c2, result) _UTRIE_NEXT(trie, data32, src, limit, c, c2, result, uint32_t)
michael@0: 
michael@0: /**
michael@0:  * Get the previous code point (c, c2), pre-decrement src,
michael@0:  * and get a 16-bit value from the trie.
michael@0:  *
michael@0:  * @param trie (const UTrie *, in) a pointer to the runtime trie structure
michael@0:  * @param start (const UChar *, in) the start pointer for the text, or NULL
michael@0:  * @param src (const UChar *, in/out) the source text pointer
michael@0:  * @param c (UChar, out) variable for the BMP or lead code unit
michael@0:  * @param c2 (UChar, out) variable for 0 or the trail code unit
michael@0:  * @param result (uint16_t, out) uint16_t variable for the trie lookup result
michael@0:  */
michael@0: #define UTRIE_PREVIOUS16(trie, start, src, c, c2, result) _UTRIE_PREVIOUS(trie, index, start, src, c, c2, result, uint16_t)
michael@0: 
michael@0: /**
michael@0:  * Get the previous code point (c, c2), pre-decrement src,
michael@0:  * and get a 32-bit value from the trie.
michael@0:  *
michael@0:  * @param trie (const UTrie *, in) a pointer to the runtime trie structure
michael@0:  * @param start (const UChar *, in) the start pointer for the text, or NULL
michael@0:  * @param src (const UChar *, in/out) the source text pointer
michael@0:  * @param c (UChar, out) variable for the BMP or lead code unit
michael@0:  * @param c2 (UChar, out) variable for 0 or the trail code unit
michael@0:  * @param result (uint32_t, out) uint32_t variable for the trie lookup result
michael@0:  */
michael@0: #define UTRIE_PREVIOUS32(trie, start, src, c, c2, result) _UTRIE_PREVIOUS(trie, data32, start, src, c, c2, result, uint32_t)
michael@0: 
michael@0: /**
michael@0:  * Get a 16-bit trie value from a pair of surrogates.
michael@0:  *
michael@0:  * @param trie (const UTrie *, in) a pointer to the runtime trie structure
michael@0:  * @param c (UChar, in) a lead surrogate
michael@0:  * @param c2 (UChar, in) a trail surrogate
michael@0:  * @param result (uint16_t, out) uint16_t variable for the trie lookup result
michael@0:  */
michael@0: #define UTRIE_GET16_FROM_PAIR(trie, c, c2, result) _UTRIE_GET_FROM_PAIR(trie, index, c, c2, result, uint16_t)
michael@0: 
michael@0: /**
michael@0:  * Get a 32-bit trie value from a pair of surrogates.
michael@0:  *
michael@0:  * @param trie (const UTrie *, in) a pointer to the runtime trie structure
michael@0:  * @param c (UChar, in) a lead surrogate
michael@0:  * @param c2 (UChar, in) a trail surrogate
michael@0:  * @param result (uint32_t, out) uint32_t variable for the trie lookup result
michael@0:  */
michael@0: #define UTRIE_GET32_FROM_PAIR(trie, c, c2, result) _UTRIE_GET_FROM_PAIR(trie, data32, c, c2, result, uint32_t)
michael@0: 
michael@0: /**
michael@0:  * Get a 16-bit trie value from a folding offset (from the value of a lead surrogate)
michael@0:  * and a trail surrogate.
michael@0:  *
michael@0:  * @param trie (const UTrie *, in) a pointer to the runtime trie structure
michael@0:  * @param offset (int32_t, in) the folding offset from the value of a lead surrogate
michael@0:  * @param c2 (UChar, in) a trail surrogate (only the 10 low bits are significant)
michael@0:  * @return (uint16_t) trie lookup result
michael@0:  */
michael@0: #define UTRIE_GET16_FROM_OFFSET_TRAIL(trie, offset, c2) _UTRIE_GET_RAW(trie, index, offset, (c2)&0x3ff)
michael@0: 
michael@0: /**
michael@0:  * Get a 32-bit trie value from a folding offset (from the value of a lead surrogate)
michael@0:  * and a trail surrogate.
michael@0:  *
michael@0:  * @param trie (const UTrie *, in) a pointer to the runtime trie structure
michael@0:  * @param offset (int32_t, in) the folding offset from the value of a lead surrogate
michael@0:  * @param c2 (UChar, in) a trail surrogate (only the 10 low bits are significant)
michael@0:  * @return (uint32_t) trie lookup result
michael@0:  */
michael@0: #define UTRIE_GET32_FROM_OFFSET_TRAIL(trie, offset, c2) _UTRIE_GET_RAW(trie, data32, offset, (c2)&0x3ff)
michael@0: 
michael@0: /* enumeration callback types */
michael@0: 
michael@0: /**
michael@0:  * Callback from utrie_enum(), extracts a uint32_t value from a
michael@0:  * trie value. This value will be passed on to the UTrieEnumRange function.
michael@0:  *
michael@0:  * @param context an opaque pointer, as passed into utrie_enum()
michael@0:  * @param value a value from the trie
michael@0:  * @return the value that is to be passed on to the UTrieEnumRange function
michael@0:  */
michael@0: typedef uint32_t U_CALLCONV
michael@0: UTrieEnumValue(const void *context, uint32_t value);
michael@0: 
michael@0: /**
michael@0:  * Callback from utrie_enum(), is called for each contiguous range
michael@0:  * of code points with the same value as retrieved from the trie and
michael@0:  * transformed by the UTrieEnumValue function.
michael@0:  *
michael@0:  * The callback function can stop the enumeration by returning FALSE.
michael@0:  *
michael@0:  * @param context an opaque pointer, as passed into utrie_enum()
michael@0:  * @param start the first code point in a contiguous range with value
michael@0:  * @param limit one past the last code point in a contiguous range with value
michael@0:  * @param value the value that is set for all code points in [start..limit[
michael@0:  * @return FALSE to stop the enumeration
michael@0:  */
michael@0: typedef UBool U_CALLCONV
michael@0: UTrieEnumRange(const void *context, UChar32 start, UChar32 limit, uint32_t value);
michael@0: 
michael@0: /**
michael@0:  * Enumerate efficiently all values in a trie.
michael@0:  * For each entry in the trie, the value to be delivered is passed through
michael@0:  * the UTrieEnumValue function.
michael@0:  * The value is unchanged if that function pointer is NULL.
michael@0:  *
michael@0:  * For each contiguous range of code points with a given value,
michael@0:  * the UTrieEnumRange function is called.
michael@0:  *
michael@0:  * @param trie a pointer to the runtime trie structure
michael@0:  * @param enumValue a pointer to a function that may transform the trie entry value,
michael@0:  *                  or NULL if the values from the trie are to be used directly
michael@0:  * @param enumRange a pointer to a function that is called for each contiguous range
michael@0:  *                  of code points with the same value
michael@0:  * @param context an opaque pointer that is passed on to the callback functions
michael@0:  */
michael@0: U_CAPI void U_EXPORT2
michael@0: utrie_enum(const UTrie *trie,
michael@0:            UTrieEnumValue *enumValue, UTrieEnumRange *enumRange, const void *context);
michael@0: 
michael@0: /**
michael@0:  * Unserialize a trie from 32-bit-aligned memory.
michael@0:  * Inverse of utrie_serialize().
michael@0:  * Fills the UTrie runtime trie structure with the settings for the trie data.
michael@0:  *
michael@0:  * @param trie a pointer to the runtime trie structure
michael@0:  * @param data a pointer to 32-bit-aligned memory containing trie data
michael@0:  * @param length the number of bytes available at data
michael@0:  * @param pErrorCode an in/out ICU UErrorCode
michael@0:  * @return the number of bytes at data taken up by the trie data
michael@0:  */
michael@0: U_CAPI int32_t U_EXPORT2
michael@0: utrie_unserialize(UTrie *trie, const void *data, int32_t length, UErrorCode *pErrorCode);
michael@0: 
michael@0: /**
michael@0:  * "Unserialize" a dummy trie.
michael@0:  * A dummy trie is an empty runtime trie, used when a real data trie cannot
michael@0:  * be loaded.
michael@0:  *
michael@0:  * The input memory is filled so that the trie always returns the initialValue,
michael@0:  * or the leadUnitValue for lead surrogate code points.
michael@0:  * The Latin-1 part is always set up to be linear.
michael@0:  *
michael@0:  * @param trie a pointer to the runtime trie structure
michael@0:  * @param data a pointer to 32-bit-aligned memory to be filled with the dummy trie data
michael@0:  * @param length the number of bytes available at data (recommended to use UTRIE_DUMMY_SIZE)
michael@0:  * @param initialValue the initial value that is set for all code points
michael@0:  * @param leadUnitValue the value for lead surrogate code _units_ that do not
michael@0:  *                      have associated supplementary data
michael@0:  * @param pErrorCode an in/out ICU UErrorCode
michael@0:  *
michael@0:  * @see UTRIE_DUMMY_SIZE
michael@0:  * @see utrie_open
michael@0:  */
michael@0: U_CAPI int32_t U_EXPORT2
michael@0: utrie_unserializeDummy(UTrie *trie,
michael@0:                        void *data, int32_t length,
michael@0:                        uint32_t initialValue, uint32_t leadUnitValue,
michael@0:                        UBool make16BitTrie,
michael@0:                        UErrorCode *pErrorCode);
michael@0: 
michael@0: /**
michael@0:  * Default implementation for UTrie.getFoldingOffset, set automatically by
michael@0:  * utrie_unserialize().
michael@0:  * Simply returns the lead surrogate's value itself - which is the inverse
michael@0:  * of the default folding function used by utrie_serialize().
michael@0:  * Exported for static const UTrie structures.
michael@0:  *
michael@0:  * @see UTrieGetFoldingOffset
michael@0:  */
michael@0: U_CAPI int32_t U_EXPORT2
michael@0: utrie_defaultGetFoldingOffset(uint32_t data);
michael@0: 
michael@0: /* Building a trie ----------------------------------------------------------*/
michael@0: 
michael@0: /**
michael@0:  * Build-time trie structure.
michael@0:  * Opaque definition, here only to make fillIn parameters possible
michael@0:  * for utrie_open() and utrie_clone().
michael@0:  */
michael@0: struct UNewTrie {
michael@0:     /**
michael@0:      * Index values at build-time are 32 bits wide for easier processing.
michael@0:      * Bit 31 is set if the data block is used by multiple index values (from utrie_setRange()).
michael@0:      */
michael@0:     int32_t index[UTRIE_MAX_INDEX_LENGTH];
michael@0:     uint32_t *data;
michael@0: 
michael@0:     uint32_t leadUnitValue;
michael@0:     int32_t indexLength, dataCapacity, dataLength;
michael@0:     UBool isAllocated, isDataAllocated;
michael@0:     UBool isLatin1Linear, isCompacted;
michael@0: 
michael@0:     /**
michael@0:      * Map of adjusted indexes, used in utrie_compact().
michael@0:      * Maps from original indexes to new ones.
michael@0:      */
michael@0:     int32_t map[UTRIE_MAX_BUILD_TIME_DATA_LENGTH>>UTRIE_SHIFT];
michael@0: };
michael@0: 
michael@0: typedef struct UNewTrie UNewTrie;
michael@0: 
michael@0: /**
michael@0:  * Build-time trie callback function, used with utrie_serialize().
michael@0:  * This function calculates a lead surrogate's value including a folding offset
michael@0:  * from the 1024 supplementary code points [start..start+1024[ .
michael@0:  * It is U+10000 <= start <= U+10fc00 and (start&0x3ff)==0.
michael@0:  *
michael@0:  * The folding offset is provided by the caller.
michael@0:  * It is offset=UTRIE_BMP_INDEX_LENGTH+n*UTRIE_SURROGATE_BLOCK_COUNT with n=0..1023.
michael@0:  * Instead of the offset itself, n can be stored in 10 bits -
michael@0:  * or fewer if it can be assumed that few lead surrogates have associated data.
michael@0:  *
michael@0:  * The returned value must be
michael@0:  * - not zero if and only if there is relevant data
michael@0:  *   for the corresponding 1024 supplementary code points
michael@0:  * - such that UTrie.getFoldingOffset(UNewTrieGetFoldedValue(..., offset))==offset
michael@0:  *
michael@0:  * @return a folded value, or 0 if there is no relevant data for the lead surrogate.
michael@0:  */
michael@0: typedef uint32_t U_CALLCONV
michael@0: UNewTrieGetFoldedValue(UNewTrie *trie, UChar32 start, int32_t offset);
michael@0: 
michael@0: /**
michael@0:  * Open a build-time trie structure.
michael@0:  * The size of the build-time data array is specified to avoid allocating a large
michael@0:  * array in all cases. The array itself can also be passed in.
michael@0:  *
michael@0:  * Although the trie is never fully expanded to a linear array, especially when
michael@0:  * utrie_setRange32() is used, the data array could be large during build time.
michael@0:  * The maximum length is
michael@0:  * UTRIE_MAX_BUILD_TIME_DATA_LENGTH=0x110000+UTRIE_DATA_BLOCK_LENGTH+0x400.
michael@0:  * (Number of Unicode code points + one all-initial-value block +
michael@0:  *  possible duplicate entries for 1024 lead surrogates.)
michael@0:  * (UTRIE_DATA_BLOCK_LENGTH<=0x200 in all cases.)
michael@0:  *
michael@0:  * @param fillIn a pointer to a UNewTrie structure to be initialized (will not be released), or
michael@0:  *               NULL if one is to be allocated
michael@0:  * @param aliasData a pointer to a data array to be used (will not be released), or
michael@0:  *                  NULL if one is to be allocated
michael@0:  * @param maxDataLength the capacity of aliasData (if not NULL) or
michael@0:  *                      the length of the data array to be allocated
michael@0:  * @param initialValue the initial value that is set for all code points
michael@0:  * @param leadUnitValue the value for lead surrogate code _units_ that do not
michael@0:  *                      have associated supplementary data
michael@0:  * @param latin1Linear a flag indicating whether the Latin-1 range is to be allocated and
michael@0:  *                     kept in a linear, contiguous part of the data array
michael@0:  * @return a pointer to the initialized fillIn or the allocated and initialized new UNewTrie
michael@0:  */
michael@0: U_CAPI UNewTrie * U_EXPORT2
michael@0: utrie_open(UNewTrie *fillIn,
michael@0:            uint32_t *aliasData, int32_t maxDataLength,
michael@0:            uint32_t initialValue, uint32_t leadUnitValue,
michael@0:            UBool latin1Linear);
michael@0: 
michael@0: /**
michael@0:  * Clone a build-time trie structure with all entries.
michael@0:  *
michael@0:  * @param fillIn like in utrie_open()
michael@0:  * @param other the build-time trie structure to clone
michael@0:  * @param aliasData like in utrie_open(),
michael@0:  *                  used if aliasDataLength>=(capacity of other's data array)
michael@0:  * @param aliasDataLength the length of aliasData
michael@0:  * @return a pointer to the initialized fillIn or the allocated and initialized new UNewTrie
michael@0:  */
michael@0: U_CAPI UNewTrie * U_EXPORT2
michael@0: utrie_clone(UNewTrie *fillIn, const UNewTrie *other, uint32_t *aliasData, int32_t aliasDataLength);
michael@0: 
michael@0: /**
michael@0:  * Close a build-time trie structure, and release memory
michael@0:  * that was allocated by utrie_open() or utrie_clone().
michael@0:  *
michael@0:  * @param trie the build-time trie
michael@0:  */
michael@0: U_CAPI void U_EXPORT2
michael@0: utrie_close(UNewTrie *trie);
michael@0: 
michael@0: /**
michael@0:  * Get the data array of a build-time trie.
michael@0:  * The data may be modified, but entries that are equal before
michael@0:  * must still be equal after modification.
michael@0:  *
michael@0:  * @param trie the build-time trie
michael@0:  * @param pLength (out) a pointer to a variable that receives the number
michael@0:  *                of entries in the data array
michael@0:  * @return the data array
michael@0:  */
michael@0: U_CAPI uint32_t * U_EXPORT2
michael@0: utrie_getData(UNewTrie *trie, int32_t *pLength);
michael@0: 
michael@0: /**
michael@0:  * Set a value for a code point.
michael@0:  *
michael@0:  * @param trie the build-time trie
michael@0:  * @param c the code point
michael@0:  * @param value the value
michael@0:  * @return FALSE if a failure occurred (illegal argument or data array overrun)
michael@0:  */
michael@0: U_CAPI UBool U_EXPORT2
michael@0: utrie_set32(UNewTrie *trie, UChar32 c, uint32_t value);
michael@0: 
michael@0: /**
michael@0:  * Get a value from a code point as stored in the build-time trie.
michael@0:  *
michael@0:  * @param trie the build-time trie
michael@0:  * @param c the code point
michael@0:  * @param pInBlockZero if not NULL, then *pInBlockZero is set to TRUE
michael@0:  *                     iff the value is retrieved from block 0;
michael@0:  *                     block 0 is the all-initial-value initial block
michael@0:  * @return the value
michael@0:  */
michael@0: U_CAPI uint32_t U_EXPORT2
michael@0: utrie_get32(UNewTrie *trie, UChar32 c, UBool *pInBlockZero);
michael@0: 
michael@0: /**
michael@0:  * Set a value in a range of code points [start..limit[.
michael@0:  * All code points c with start<=c<limit will get the value if
michael@0:  * overwrite is TRUE or if the old value is 0.
michael@0:  *
michael@0:  * @param trie the build-time trie
michael@0:  * @param start the first code point to get the value
michael@0:  * @param limit one past the last code point to get the value
michael@0:  * @param value the value
michael@0:  * @param overwrite flag for whether old non-initial values are to be overwritten
michael@0:  * @return FALSE if a failure occurred (illegal argument or data array overrun)
michael@0:  */
michael@0: U_CAPI UBool U_EXPORT2
michael@0: utrie_setRange32(UNewTrie *trie, UChar32 start, UChar32 limit, uint32_t value, UBool overwrite);
michael@0: 
michael@0: /**
michael@0:  * Compact the build-time trie after all values are set, and then
michael@0:  * serialize it into 32-bit aligned memory.
michael@0:  *
michael@0:  * After this, the trie can only be serizalized again and/or closed;
michael@0:  * no further values can be added.
michael@0:  *
michael@0:  * @see utrie_unserialize()
michael@0:  *
michael@0:  * @param trie the build-time trie
michael@0:  * @param data a pointer to 32-bit-aligned memory for the trie data
michael@0:  * @param capacity the number of bytes available at data
michael@0:  * @param getFoldedValue a callback function that calculates the value for
michael@0:  *                       a lead surrogate from all of its supplementary code points
michael@0:  *                       and the folding offset;
michael@0:  *                       if NULL, then a default function is used which returns just
michael@0:  *                       the input offset when there are any non-initial-value entries
michael@0:  * @param reduceTo16Bits flag for whether the values are to be reduced to a
michael@0:  *                       width of 16 bits for serialization and runtime
michael@0:  * @param pErrorCode a UErrorCode argument; among other possible error codes:
michael@0:  * - U_BUFFER_OVERFLOW_ERROR if the data storage block is too small for serialization
michael@0:  * - U_MEMORY_ALLOCATION_ERROR if the trie data array is too small
michael@0:  * - U_INDEX_OUTOFBOUNDS_ERROR if the index or data arrays are too long after compaction for serialization
michael@0:  *
michael@0:  * @return the number of bytes written for the trie
michael@0:  */
michael@0: U_CAPI int32_t U_EXPORT2
michael@0: utrie_serialize(UNewTrie *trie, void *data, int32_t capacity,
michael@0:                 UNewTrieGetFoldedValue *getFoldedValue,
michael@0:                 UBool reduceTo16Bits,
michael@0:                 UErrorCode *pErrorCode);
michael@0: 
michael@0: /**
michael@0:  * Swap a serialized UTrie.
michael@0:  * @internal
michael@0:  */
michael@0: U_CAPI int32_t U_EXPORT2
michael@0: utrie_swap(const UDataSwapper *ds,
michael@0:            const void *inData, int32_t length, void *outData,
michael@0:            UErrorCode *pErrorCode);
michael@0: 
michael@0: /* serialization ------------------------------------------------------------ */
michael@0: 
michael@0: /**
michael@0:  * Trie data structure in serialized form:
michael@0:  *
michael@0:  * UTrieHeader header;
michael@0:  * uint16_t index[header.indexLength];
michael@0:  * uint16_t data[header.dataLength];
michael@0:  * @internal
michael@0:  */
michael@0: typedef struct UTrieHeader {
michael@0:     /** "Trie" in big-endian US-ASCII (0x54726965) */
michael@0:     uint32_t signature;
michael@0: 
michael@0:     /**
michael@0:      * options bit field:
michael@0:      *     9    1=Latin-1 data is stored linearly at data+UTRIE_DATA_BLOCK_LENGTH
michael@0:      *     8    0=16-bit data, 1=32-bit data
michael@0:      *  7..4    UTRIE_INDEX_SHIFT   // 0..UTRIE_SHIFT
michael@0:      *  3..0    UTRIE_SHIFT         // 1..9
michael@0:      */
michael@0:     uint32_t options;
michael@0: 
michael@0:     /** indexLength is a multiple of UTRIE_SURROGATE_BLOCK_COUNT */
michael@0:     int32_t indexLength;
michael@0: 
michael@0:     /** dataLength>=UTRIE_DATA_BLOCK_LENGTH */
michael@0:     int32_t dataLength;
michael@0: } UTrieHeader;
michael@0: 
michael@0: /**
michael@0:  * Constants for use with UTrieHeader.options.
michael@0:  * @internal
michael@0:  */
michael@0: enum {
michael@0:     /** Mask to get the UTRIE_SHIFT value from options. */
michael@0:     UTRIE_OPTIONS_SHIFT_MASK=0xf,
michael@0: 
michael@0:     /** Shift options right this much to get the UTRIE_INDEX_SHIFT value. */
michael@0:     UTRIE_OPTIONS_INDEX_SHIFT=4,
michael@0: 
michael@0:     /** If set, then the data (stage 2) array is 32 bits wide. */
michael@0:     UTRIE_OPTIONS_DATA_IS_32_BIT=0x100,
michael@0: 
michael@0:     /**
michael@0:      * If set, then Latin-1 data (for U+0000..U+00ff) is stored in the data (stage 2) array
michael@0:      * as a simple, linear array at data+UTRIE_DATA_BLOCK_LENGTH.
michael@0:      */
michael@0:     UTRIE_OPTIONS_LATIN1_IS_LINEAR=0x200
michael@0: };
michael@0: 
michael@0: U_CDECL_END
michael@0: 
michael@0: #endif