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: *   file name:  bocsu.c
michael@0: *   encoding:   US-ASCII
michael@0: *   tab size:   8 (not used)
michael@0: *   indentation:4
michael@0: *
michael@0: *   Author: Markus W. Scherer
michael@0: *
michael@0: *   Modification history:
michael@0: *   05/18/2001  weiv    Made into separate module
michael@0: */
michael@0: 
michael@0: #ifndef BOCSU_H
michael@0: #define BOCSU_H
michael@0: 
michael@0: #include "unicode/utypes.h"
michael@0: 
michael@0: #if !UCONFIG_NO_COLLATION
michael@0: 
michael@0: U_NAMESPACE_BEGIN
michael@0: 
michael@0: class ByteSink;
michael@0: 
michael@0: U_NAMESPACE_END
michael@0: 
michael@0: /*
michael@0:  * "BOCSU"
michael@0:  * Binary Ordered Compression Scheme for Unicode
michael@0:  *
michael@0:  * Specific application:
michael@0:  *
michael@0:  * Encode a Unicode string for the identical level of a sort key.
michael@0:  * Restrictions:
michael@0:  * - byte stream (unsigned 8-bit bytes)
michael@0:  * - lexical order of the identical-level run must be
michael@0:  *   the same as code point order for the string
michael@0:  * - avoid byte values 0, 1, 2
michael@0:  *
michael@0:  * Method: Slope Detection
michael@0:  * Remember the previous code point (initial 0).
michael@0:  * For each cp in the string, encode the difference to the previous one.
michael@0:  *
michael@0:  * With a compact encoding of differences, this yields good results for
michael@0:  * small scripts and UTF-like results otherwise.
michael@0:  *
michael@0:  * Encoding of differences:
michael@0:  * - Similar to a UTF, encoding the length of the byte sequence in the lead bytes.
michael@0:  * - Does not need to be friendly for decoding or random access
michael@0:  *   (trail byte values may overlap with lead/single byte values).
michael@0:  * - The signedness must be encoded as the most significant part.
michael@0:  *
michael@0:  * We encode differences with few bytes if their absolute values are small.
michael@0:  * For correct ordering, we must treat the entire value range -10ffff..+10ffff
michael@0:  * in ascending order, which forbids encoding the sign and the absolute value separately.
michael@0:  * Instead, we split the lead byte range in the middle and encode non-negative values
michael@0:  * going up and negative values going down.
michael@0:  *
michael@0:  * For very small absolute values, the difference is added to a middle byte value
michael@0:  * for single-byte encoded differences.
michael@0:  * For somewhat larger absolute values, the difference is divided by the number
michael@0:  * of byte values available, the modulo is used for one trail byte, and the remainder
michael@0:  * is added to a lead byte avoiding the single-byte range.
michael@0:  * For large absolute values, the difference is similarly encoded in three bytes.
michael@0:  *
michael@0:  * This encoding does not use byte values 0, 1, 2, but uses all other byte values
michael@0:  * for lead/single bytes so that the middle range of single bytes is as large
michael@0:  * as possible.
michael@0:  * Note that the lead byte ranges overlap some, but that the sequences as a whole
michael@0:  * are well ordered. I.e., even if the lead byte is the same for sequences of different
michael@0:  * lengths, the trail bytes establish correct order.
michael@0:  * It would be possible to encode slightly larger ranges for each length (>1) by
michael@0:  * subtracting the lower bound of the range. However, that would also slow down the
michael@0:  * calculation.
michael@0:  *
michael@0:  * For the actual string encoding, an optimization moves the previous code point value
michael@0:  * to the middle of its Unicode script block to minimize the differences in
michael@0:  * same-script text runs.
michael@0:  */
michael@0: 
michael@0: /* Do not use byte values 0, 1, 2 because they are separators in sort keys. */
michael@0: #define SLOPE_MIN           3
michael@0: #define SLOPE_MAX           0xff
michael@0: #define SLOPE_MIDDLE        0x81
michael@0: 
michael@0: #define SLOPE_TAIL_COUNT    (SLOPE_MAX-SLOPE_MIN+1)
michael@0: 
michael@0: #define SLOPE_MAX_BYTES     4
michael@0: 
michael@0: /*
michael@0:  * Number of lead bytes:
michael@0:  * 1        middle byte for 0
michael@0:  * 2*80=160 single bytes for !=0
michael@0:  * 2*42=84  for double-byte values
michael@0:  * 2*3=6    for 3-byte values
michael@0:  * 2*1=2    for 4-byte values
michael@0:  *
michael@0:  * The sum must be <=SLOPE_TAIL_COUNT.
michael@0:  *
michael@0:  * Why these numbers?
michael@0:  * - There should be >=128 single-byte values to cover 128-blocks
michael@0:  *   with small scripts.
michael@0:  * - There should be >=20902 single/double-byte values to cover Unihan.
michael@0:  * - It helps CJK Extension B some if there are 3-byte values that cover
michael@0:  *   the distance between them and Unihan.
michael@0:  *   This also helps to jump among distant places in the BMP.
michael@0:  * - Four-byte values are necessary to cover the rest of Unicode.
michael@0:  *
michael@0:  * Symmetrical lead byte counts are for convenience.
michael@0:  * With an equal distribution of even and odd differences there is also
michael@0:  * no advantage to asymmetrical lead byte counts.
michael@0:  */
michael@0: #define SLOPE_SINGLE        80
michael@0: #define SLOPE_LEAD_2        42
michael@0: #define SLOPE_LEAD_3        3
michael@0: #define SLOPE_LEAD_4        1
michael@0: 
michael@0: /* The difference value range for single-byters. */
michael@0: #define SLOPE_REACH_POS_1   SLOPE_SINGLE
michael@0: #define SLOPE_REACH_NEG_1   (-SLOPE_SINGLE)
michael@0: 
michael@0: /* The difference value range for double-byters. */
michael@0: #define SLOPE_REACH_POS_2   (SLOPE_LEAD_2*SLOPE_TAIL_COUNT+(SLOPE_LEAD_2-1))
michael@0: #define SLOPE_REACH_NEG_2   (-SLOPE_REACH_POS_2-1)
michael@0: 
michael@0: /* The difference value range for 3-byters. */
michael@0: #define SLOPE_REACH_POS_3   (SLOPE_LEAD_3*SLOPE_TAIL_COUNT*SLOPE_TAIL_COUNT+(SLOPE_LEAD_3-1)*SLOPE_TAIL_COUNT+(SLOPE_TAIL_COUNT-1))
michael@0: #define SLOPE_REACH_NEG_3   (-SLOPE_REACH_POS_3-1)
michael@0: 
michael@0: /* The lead byte start values. */
michael@0: #define SLOPE_START_POS_2   (SLOPE_MIDDLE+SLOPE_SINGLE+1)
michael@0: #define SLOPE_START_POS_3   (SLOPE_START_POS_2+SLOPE_LEAD_2)
michael@0: 
michael@0: #define SLOPE_START_NEG_2   (SLOPE_MIDDLE+SLOPE_REACH_NEG_1)
michael@0: #define SLOPE_START_NEG_3   (SLOPE_START_NEG_2-SLOPE_LEAD_2)
michael@0: 
michael@0: /*
michael@0:  * Integer division and modulo with negative numerators
michael@0:  * yields negative modulo results and quotients that are one more than
michael@0:  * what we need here.
michael@0:  */
michael@0: #define NEGDIVMOD(n, d, m) { \
michael@0:     (m)=(n)%(d); \
michael@0:     (n)/=(d); \
michael@0:     if((m)<0) { \
michael@0:         --(n); \
michael@0:         (m)+=(d); \
michael@0:     } \
michael@0: }
michael@0: 
michael@0: U_CFUNC void
michael@0: u_writeIdenticalLevelRun(const UChar *s, int32_t length, icu::ByteSink &sink);
michael@0: 
michael@0: U_CFUNC int32_t
michael@0: u_writeIdenticalLevelRunTwoChars(UChar32 first, UChar32 second, uint8_t *p);
michael@0: 
michael@0: U_CFUNC uint8_t *
michael@0: u_writeDiff(int32_t diff, uint8_t *p);
michael@0: 
michael@0: #endif /* #if !UCONFIG_NO_COLLATION */
michael@0: 
michael@0: #endif