1 /*
2 *******************************************************************************
3 * Copyright (C) 2010-2014, International Business Machines
4 * Corporation and others. All Rights Reserved.
5 *******************************************************************************
6 * collation.cpp
7 *
8 * created on: 2010oct27
9 * created by: Markus W. Scherer
10 */
11
12 #include "unicode/utypes.h"
13
14 #if !UCONFIG_NO_COLLATION
15
16 #include "collation.h"
17 #include "uassert.h"
18
19 U_NAMESPACE_BEGIN
20
21 // Some compilers don't care if constants are defined in the .cpp file.
22 // MS Visual C++ does not like it, but gcc requires it. clang does not care.
23 #ifndef _MSC_VER
24 const uint32_t Collation::ONLY_TERTIARY_MASK;
25 const uint32_t Collation::CASE_AND_TERTIARY_MASK;
26 #endif
27
28 uint32_t
incTwoBytePrimaryByOffset(uint32_t basePrimary,UBool isCompressible,int32_t offset)29 Collation::incTwoBytePrimaryByOffset(uint32_t basePrimary, UBool isCompressible, int32_t offset) {
30 // Extract the second byte, minus the minimum byte value,
31 // plus the offset, modulo the number of usable byte values, plus the minimum.
32 // Reserve the PRIMARY_COMPRESSION_LOW_BYTE and high byte if necessary.
33 uint32_t primary;
34 if(isCompressible) {
35 offset += ((int32_t)(basePrimary >> 16) & 0xff) - 4;
36 primary = (uint32_t)((offset % 251) + 4) << 16;
37 offset /= 251;
38 } else {
39 offset += ((int32_t)(basePrimary >> 16) & 0xff) - 2;
40 primary = (uint32_t)((offset % 254) + 2) << 16;
41 offset /= 254;
42 }
43 // First byte, assume no further overflow.
44 return primary | ((basePrimary & 0xff000000) + (uint32_t)(offset << 24));
45 }
46
47 uint32_t
incThreeBytePrimaryByOffset(uint32_t basePrimary,UBool isCompressible,int32_t offset)48 Collation::incThreeBytePrimaryByOffset(uint32_t basePrimary, UBool isCompressible, int32_t offset) {
49 // Extract the third byte, minus the minimum byte value,
50 // plus the offset, modulo the number of usable byte values, plus the minimum.
51 offset += ((int32_t)(basePrimary >> 8) & 0xff) - 2;
52 uint32_t primary = (uint32_t)((offset % 254) + 2) << 8;
53 offset /= 254;
54 // Same with the second byte,
55 // but reserve the PRIMARY_COMPRESSION_LOW_BYTE and high byte if necessary.
56 if(isCompressible) {
57 offset += ((int32_t)(basePrimary >> 16) & 0xff) - 4;
58 primary |= (uint32_t)((offset % 251) + 4) << 16;
59 offset /= 251;
60 } else {
61 offset += ((int32_t)(basePrimary >> 16) & 0xff) - 2;
62 primary |= (uint32_t)((offset % 254) + 2) << 16;
63 offset /= 254;
64 }
65 // First byte, assume no further overflow.
66 return primary | ((basePrimary & 0xff000000) + (uint32_t)(offset << 24));
67 }
68
69 uint32_t
decTwoBytePrimaryByOneStep(uint32_t basePrimary,UBool isCompressible,int32_t step)70 Collation::decTwoBytePrimaryByOneStep(uint32_t basePrimary, UBool isCompressible, int32_t step) {
71 // Extract the second byte, minus the minimum byte value,
72 // minus the step, modulo the number of usable byte values, plus the minimum.
73 // Reserve the PRIMARY_COMPRESSION_LOW_BYTE and high byte if necessary.
74 // Assume no further underflow for the first byte.
75 U_ASSERT(0 < step && step <= 0x7f);
76 int32_t byte2 = ((int32_t)(basePrimary >> 16) & 0xff) - step;
77 if(isCompressible) {
78 if(byte2 < 4) {
79 byte2 += 251;
80 basePrimary -= 0x1000000;
81 }
82 } else {
83 if(byte2 < 2) {
84 byte2 += 254;
85 basePrimary -= 0x1000000;
86 }
87 }
88 return (basePrimary & 0xff000000) | ((uint32_t)byte2 << 16);
89 }
90
91 uint32_t
decThreeBytePrimaryByOneStep(uint32_t basePrimary,UBool isCompressible,int32_t step)92 Collation::decThreeBytePrimaryByOneStep(uint32_t basePrimary, UBool isCompressible, int32_t step) {
93 // Extract the third byte, minus the minimum byte value,
94 // minus the step, modulo the number of usable byte values, plus the minimum.
95 U_ASSERT(0 < step && step <= 0x7f);
96 int32_t byte3 = ((int32_t)(basePrimary >> 8) & 0xff) - step;
97 if(byte3 >= 2) {
98 return (basePrimary & 0xffff0000) | ((uint32_t)byte3 << 8);
99 }
100 byte3 += 254;
101 // Same with the second byte,
102 // but reserve the PRIMARY_COMPRESSION_LOW_BYTE and high byte if necessary.
103 int32_t byte2 = ((int32_t)(basePrimary >> 16) & 0xff) - 1;
104 if(isCompressible) {
105 if(byte2 < 4) {
106 byte2 = 0xfe;
107 basePrimary -= 0x1000000;
108 }
109 } else {
110 if(byte2 < 2) {
111 byte2 = 0xff;
112 basePrimary -= 0x1000000;
113 }
114 }
115 // First byte, assume no further underflow.
116 return (basePrimary & 0xff000000) | ((uint32_t)byte2 << 16) | ((uint32_t)byte3 << 8);
117 }
118
119 uint32_t
getThreeBytePrimaryForOffsetData(UChar32 c,int64_t dataCE)120 Collation::getThreeBytePrimaryForOffsetData(UChar32 c, int64_t dataCE) {
121 uint32_t p = (uint32_t)(dataCE >> 32); // three-byte primary pppppp00
122 int32_t lower32 = (int32_t)dataCE; // base code point b & step s: bbbbbbss (bit 7: isCompressible)
123 int32_t offset = (c - (lower32 >> 8)) * (lower32 & 0x7f); // delta * increment
124 UBool isCompressible = (lower32 & 0x80) != 0;
125 return Collation::incThreeBytePrimaryByOffset(p, isCompressible, offset);
126 }
127
128 uint32_t
unassignedPrimaryFromCodePoint(UChar32 c)129 Collation::unassignedPrimaryFromCodePoint(UChar32 c) {
130 // Create a gap before U+0000. Use c=-1 for [first unassigned].
131 ++c;
132 // Fourth byte: 18 values, every 14th byte value (gap of 13).
133 uint32_t primary = 2 + (c % 18) * 14;
134 c /= 18;
135 // Third byte: 254 values.
136 primary |= (2 + (c % 254)) << 8;
137 c /= 254;
138 // Second byte: 251 values 04..FE excluding the primary compression bytes.
139 primary |= (4 + (c % 251)) << 16;
140 // One lead byte covers all code points (c < 0x1182B4 = 1*251*254*18).
141 return primary | (UNASSIGNED_IMPLICIT_BYTE << 24);
142 }
143
144 U_NAMESPACE_END
145
146 #endif // !UCONFIG_NO_COLLATION
147