michael@0: /*
michael@0: *******************************************************************************
michael@0: *
michael@0: *   Copyright (C) 2002-2010, International Business Machines
michael@0: *   Corporation and others.  All Rights Reserved.
michael@0: *
michael@0: *******************************************************************************
michael@0: *   file name:  propsvec.h
michael@0: *   encoding:   US-ASCII
michael@0: *   tab size:   8 (not used)
michael@0: *   indentation:4
michael@0: *
michael@0: *   created on: 2002feb22
michael@0: *   created by: Markus W. Scherer
michael@0: *
michael@0: *   Store bits (Unicode character properties) in bit set vectors.
michael@0: */
michael@0: 
michael@0: #ifndef __UPROPSVEC_H__
michael@0: #define __UPROPSVEC_H__
michael@0: 
michael@0: #include "unicode/utypes.h"
michael@0: #include "utrie.h"
michael@0: #include "utrie2.h"
michael@0: 
michael@0: U_CDECL_BEGIN
michael@0: 
michael@0: /**
michael@0:  * Unicode Properties Vectors associated with code point ranges.
michael@0:  *
michael@0:  * Rows of uint32_t integers in a contiguous array store
michael@0:  * the range limits and the properties vectors.
michael@0:  *
michael@0:  * Logically, each row has a certain number of uint32_t values,
michael@0:  * which is set via the upvec_open() "columns" parameter.
michael@0:  *
michael@0:  * Internally, two additional columns are stored.
michael@0:  * In each internal row,
michael@0:  * row[0] contains the start code point and
michael@0:  * row[1] contains the limit code point,
michael@0:  * which is the start of the next range.
michael@0:  *
michael@0:  * Initially, there is only one "normal" row for
michael@0:  * range [0..0x110000[ with values 0.
michael@0:  * There are additional rows for special purposes, see UPVEC_FIRST_SPECIAL_CP.
michael@0:  *
michael@0:  * It would be possible to store only one range boundary per row,
michael@0:  * but self-contained rows allow to later sort them by contents.
michael@0:  */
michael@0: struct UPropsVectors;
michael@0: typedef struct UPropsVectors UPropsVectors;
michael@0: 
michael@0: /*
michael@0:  * Special pseudo code points for storing the initialValue and the errorValue,
michael@0:  * which are used to initialize a UTrie2 or similar.
michael@0:  */
michael@0: #define UPVEC_FIRST_SPECIAL_CP 0x110000
michael@0: #define UPVEC_INITIAL_VALUE_CP 0x110000
michael@0: #define UPVEC_ERROR_VALUE_CP 0x110001
michael@0: #define UPVEC_MAX_CP 0x110001
michael@0: 
michael@0: /*
michael@0:  * Special pseudo code point used in upvec_compact() signalling the end of
michael@0:  * delivering special values and the beginning of delivering real ones.
michael@0:  * Stable value, unlike UPVEC_MAX_CP which might grow over time.
michael@0:  */
michael@0: #define UPVEC_START_REAL_VALUES_CP 0x200000
michael@0: 
michael@0: /*
michael@0:  * Open a UPropsVectors object.
michael@0:  * @param columns Number of value integers (uint32_t) per row.
michael@0:  */
michael@0: U_CAPI UPropsVectors * U_EXPORT2
michael@0: upvec_open(int32_t columns, UErrorCode *pErrorCode);
michael@0: 
michael@0: U_CAPI void U_EXPORT2
michael@0: upvec_close(UPropsVectors *pv);
michael@0: 
michael@0: /*
michael@0:  * In rows for code points [start..end], select the column,
michael@0:  * reset the mask bits and set the value bits (ANDed with the mask).
michael@0:  *
michael@0:  * Will set U_NO_WRITE_PERMISSION if called after upvec_compact().
michael@0:  */
michael@0: U_CAPI void U_EXPORT2
michael@0: upvec_setValue(UPropsVectors *pv,
michael@0:                UChar32 start, UChar32 end,
michael@0:                int32_t column,
michael@0:                uint32_t value, uint32_t mask,
michael@0:                UErrorCode *pErrorCode);
michael@0: 
michael@0: /*
michael@0:  * Logically const but must not be used on the same pv concurrently!
michael@0:  * Always returns 0 if called after upvec_compact().
michael@0:  */
michael@0: U_CAPI uint32_t U_EXPORT2
michael@0: upvec_getValue(const UPropsVectors *pv, UChar32 c, int32_t column);
michael@0: 
michael@0: /*
michael@0:  * pRangeStart and pRangeEnd can be NULL.
michael@0:  * @return NULL if rowIndex out of range and for illegal arguments,
michael@0:  *         or if called after upvec_compact()
michael@0:  */
michael@0: U_CAPI uint32_t * U_EXPORT2
michael@0: upvec_getRow(const UPropsVectors *pv, int32_t rowIndex,
michael@0:              UChar32 *pRangeStart, UChar32 *pRangeEnd);
michael@0: 
michael@0: /*
michael@0:  * Compact the vectors:
michael@0:  * - modify the memory
michael@0:  * - keep only unique vectors
michael@0:  * - store them contiguously from the beginning of the memory
michael@0:  * - for each (non-unique) row, call the handler function
michael@0:  *
michael@0:  * The handler's rowIndex is the index of the row in the compacted
michael@0:  * memory block.
michael@0:  * (Therefore, it starts at 0 increases in increments of the columns value.)
michael@0:  *
michael@0:  * In a first phase, only special values are delivered (each exactly once),
michael@0:  * with start==end both equalling a special pseudo code point.
michael@0:  * Then the handler is called once more with start==end==UPVEC_START_REAL_VALUES_CP
michael@0:  * where rowIndex is the length of the compacted array,
michael@0:  * and the row is arbitrary (but not NULL).
michael@0:  * Then, in the second phase, the handler is called for each row of real values.
michael@0:  */
michael@0: typedef void U_CALLCONV
michael@0: UPVecCompactHandler(void *context,
michael@0:                     UChar32 start, UChar32 end,
michael@0:                     int32_t rowIndex, uint32_t *row, int32_t columns,
michael@0:                     UErrorCode *pErrorCode);
michael@0: 
michael@0: U_CAPI void U_EXPORT2
michael@0: upvec_compact(UPropsVectors *pv, UPVecCompactHandler *handler, void *context, UErrorCode *pErrorCode);
michael@0: 
michael@0: /*
michael@0:  * Get the vectors array after calling upvec_compact().
michael@0:  * The caller must not modify nor release the returned array.
michael@0:  * Returns NULL if called before upvec_compact().
michael@0:  */
michael@0: U_CAPI const uint32_t * U_EXPORT2
michael@0: upvec_getArray(const UPropsVectors *pv, int32_t *pRows, int32_t *pColumns);
michael@0: 
michael@0: /*
michael@0:  * Get a clone of the vectors array after calling upvec_compact().
michael@0:  * The caller owns the returned array and must uprv_free() it.
michael@0:  * Returns NULL if called before upvec_compact().
michael@0:  */
michael@0: U_CAPI uint32_t * U_EXPORT2
michael@0: upvec_cloneArray(const UPropsVectors *pv,
michael@0:                  int32_t *pRows, int32_t *pColumns, UErrorCode *pErrorCode);
michael@0: 
michael@0: /*
michael@0:  * Call upvec_compact(), create a 16-bit UTrie2 with indexes into the compacted
michael@0:  * vectors array, and freeze the trie.
michael@0:  */
michael@0: U_CAPI UTrie2 * U_EXPORT2
michael@0: upvec_compactToUTrie2WithRowIndexes(UPropsVectors *pv, UErrorCode *pErrorCode);
michael@0: 
michael@0: struct UPVecToUTrie2Context {
michael@0:     UTrie2 *trie;
michael@0:     int32_t initialValue;
michael@0:     int32_t errorValue;
michael@0:     int32_t maxValue;
michael@0: };
michael@0: typedef struct UPVecToUTrie2Context UPVecToUTrie2Context;
michael@0: 
michael@0: /* context=UPVecToUTrie2Context, creates the trie and stores the rowIndex values */
michael@0: U_CAPI void U_CALLCONV
michael@0: upvec_compactToUTrie2Handler(void *context,
michael@0:                              UChar32 start, UChar32 end,
michael@0:                              int32_t rowIndex, uint32_t *row, int32_t columns,
michael@0:                              UErrorCode *pErrorCode);
michael@0: 
michael@0: U_CDECL_END
michael@0: 
michael@0: #endif