1 /* 2 ******************************************************************************* 3 * Copyright (C) 2001-2011, International Business Machines 4 * Corporation and others. All Rights Reserved. 5 ******************************************************************************* 6 * file name: bocsu.c 7 * encoding: US-ASCII 8 * tab size: 8 (not used) 9 * indentation:4 10 * 11 * Author: Markus W. Scherer 12 * 13 * Modification history: 14 * 05/18/2001 weiv Made into separate module 15 */ 16 17 #ifndef BOCSU_H 18 #define BOCSU_H 19 20 #include "unicode/utypes.h" 21 22 #if !UCONFIG_NO_COLLATION 23 24 U_NAMESPACE_BEGIN 25 26 class ByteSink; 27 28 U_NAMESPACE_END 29 30 /* 31 * "BOCSU" 32 * Binary Ordered Compression Scheme for Unicode 33 * 34 * Specific application: 35 * 36 * Encode a Unicode string for the identical level of a sort key. 37 * Restrictions: 38 * - byte stream (unsigned 8-bit bytes) 39 * - lexical order of the identical-level run must be 40 * the same as code point order for the string 41 * - avoid byte values 0, 1, 2 42 * 43 * Method: Slope Detection 44 * Remember the previous code point (initial 0). 45 * For each cp in the string, encode the difference to the previous one. 46 * 47 * With a compact encoding of differences, this yields good results for 48 * small scripts and UTF-like results otherwise. 49 * 50 * Encoding of differences: 51 * - Similar to a UTF, encoding the length of the byte sequence in the lead bytes. 52 * - Does not need to be friendly for decoding or random access 53 * (trail byte values may overlap with lead/single byte values). 54 * - The signedness must be encoded as the most significant part. 55 * 56 * We encode differences with few bytes if their absolute values are small. 57 * For correct ordering, we must treat the entire value range -10ffff..+10ffff 58 * in ascending order, which forbids encoding the sign and the absolute value separately. 59 * Instead, we split the lead byte range in the middle and encode non-negative values 60 * going up and negative values going down. 61 * 62 * For very small absolute values, the difference is added to a middle byte value 63 * for single-byte encoded differences. 64 * For somewhat larger absolute values, the difference is divided by the number 65 * of byte values available, the modulo is used for one trail byte, and the remainder 66 * is added to a lead byte avoiding the single-byte range. 67 * For large absolute values, the difference is similarly encoded in three bytes. 68 * 69 * This encoding does not use byte values 0, 1, 2, but uses all other byte values 70 * for lead/single bytes so that the middle range of single bytes is as large 71 * as possible. 72 * Note that the lead byte ranges overlap some, but that the sequences as a whole 73 * are well ordered. I.e., even if the lead byte is the same for sequences of different 74 * lengths, the trail bytes establish correct order. 75 * It would be possible to encode slightly larger ranges for each length (>1) by 76 * subtracting the lower bound of the range. However, that would also slow down the 77 * calculation. 78 * 79 * For the actual string encoding, an optimization moves the previous code point value 80 * to the middle of its Unicode script block to minimize the differences in 81 * same-script text runs. 82 */ 83 84 /* Do not use byte values 0, 1, 2 because they are separators in sort keys. */ 85 #define SLOPE_MIN 3 86 #define SLOPE_MAX 0xff 87 #define SLOPE_MIDDLE 0x81 88 89 #define SLOPE_TAIL_COUNT (SLOPE_MAX-SLOPE_MIN+1) 90 91 #define SLOPE_MAX_BYTES 4 92 93 /* 94 * Number of lead bytes: 95 * 1 middle byte for 0 96 * 2*80=160 single bytes for !=0 97 * 2*42=84 for double-byte values 98 * 2*3=6 for 3-byte values 99 * 2*1=2 for 4-byte values 100 * 101 * The sum must be <=SLOPE_TAIL_COUNT. 102 * 103 * Why these numbers? 104 * - There should be >=128 single-byte values to cover 128-blocks 105 * with small scripts. 106 * - There should be >=20902 single/double-byte values to cover Unihan. 107 * - It helps CJK Extension B some if there are 3-byte values that cover 108 * the distance between them and Unihan. 109 * This also helps to jump among distant places in the BMP. 110 * - Four-byte values are necessary to cover the rest of Unicode. 111 * 112 * Symmetrical lead byte counts are for convenience. 113 * With an equal distribution of even and odd differences there is also 114 * no advantage to asymmetrical lead byte counts. 115 */ 116 #define SLOPE_SINGLE 80 117 #define SLOPE_LEAD_2 42 118 #define SLOPE_LEAD_3 3 119 #define SLOPE_LEAD_4 1 120 121 /* The difference value range for single-byters. */ 122 #define SLOPE_REACH_POS_1 SLOPE_SINGLE 123 #define SLOPE_REACH_NEG_1 (-SLOPE_SINGLE) 124 125 /* The difference value range for double-byters. */ 126 #define SLOPE_REACH_POS_2 (SLOPE_LEAD_2*SLOPE_TAIL_COUNT+(SLOPE_LEAD_2-1)) 127 #define SLOPE_REACH_NEG_2 (-SLOPE_REACH_POS_2-1) 128 129 /* The difference value range for 3-byters. */ 130 #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)) 131 #define SLOPE_REACH_NEG_3 (-SLOPE_REACH_POS_3-1) 132 133 /* The lead byte start values. */ 134 #define SLOPE_START_POS_2 (SLOPE_MIDDLE+SLOPE_SINGLE+1) 135 #define SLOPE_START_POS_3 (SLOPE_START_POS_2+SLOPE_LEAD_2) 136 137 #define SLOPE_START_NEG_2 (SLOPE_MIDDLE+SLOPE_REACH_NEG_1) 138 #define SLOPE_START_NEG_3 (SLOPE_START_NEG_2-SLOPE_LEAD_2) 139 140 /* 141 * Integer division and modulo with negative numerators 142 * yields negative modulo results and quotients that are one more than 143 * what we need here. 144 */ 145 #define NEGDIVMOD(n, d, m) { \ 146 (m)=(n)%(d); \ 147 (n)/=(d); \ 148 if((m)<0) { \ 149 --(n); \ 150 (m)+=(d); \ 151 } \ 152 } 153 154 U_CFUNC void 155 u_writeIdenticalLevelRun(const UChar *s, int32_t length, U_NAMESPACE_QUALIFIER ByteSink &sink); 156 157 U_CFUNC int32_t 158 u_writeIdenticalLevelRunTwoChars(UChar32 first, UChar32 second, uint8_t *p); 159 160 U_CFUNC uint8_t * 161 u_writeDiff(int32_t diff, uint8_t *p); 162 163 #endif /* #if !UCONFIG_NO_COLLATION */ 164 165 #endif 166