1 | // © 2016 and later: Unicode, Inc. and others. |
2 | // License & terms of use: http://www.unicode.org/copyright.html |
3 | /* |
4 | ******************************************************************************* |
5 | * Copyright (C) 2010-2015, International Business Machines |
6 | * Corporation and others. All Rights Reserved. |
7 | ******************************************************************************* |
8 | * collationdata.h |
9 | * |
10 | * created on: 2010oct27 |
11 | * created by: Markus W. Scherer |
12 | */ |
13 | |
14 | #ifndef __COLLATIONDATA_H__ |
15 | #define __COLLATIONDATA_H__ |
16 | |
17 | #include "unicode/utypes.h" |
18 | |
19 | #if !UCONFIG_NO_COLLATION |
20 | |
21 | #include "unicode/ucol.h" |
22 | #include "unicode/uniset.h" |
23 | #include "collation.h" |
24 | #include "normalizer2impl.h" |
25 | #include "utrie2.h" |
26 | |
27 | struct UDataMemory; |
28 | |
29 | U_NAMESPACE_BEGIN |
30 | |
31 | class UVector32; |
32 | |
33 | /** |
34 | * Collation data container. |
35 | * Immutable data created by a CollationDataBuilder, or loaded from a file, |
36 | * or deserialized from API-provided binary data. |
37 | * |
38 | * Includes data for the collation base (root/default), aliased if this is not the base. |
39 | */ |
40 | struct U_I18N_API CollationData : public UMemory { |
41 | // Note: The ucadata.icu loader could discover the reserved ranges by setting an array |
42 | // parallel with the ranges, and resetting ranges that are indexed. |
43 | // The reordering builder code could clone the resulting template array. |
44 | enum { |
45 | REORDER_RESERVED_BEFORE_LATIN = UCOL_REORDER_CODE_FIRST + 14, |
46 | REORDER_RESERVED_AFTER_LATIN |
47 | }; |
48 | |
49 | enum { |
50 | MAX_NUM_SPECIAL_REORDER_CODES = 8, |
51 | /** C++ only, data reader check scriptStartsLength. */ |
52 | MAX_NUM_SCRIPT_RANGES = 256 |
53 | }; |
54 | |
55 | CollationData(const Normalizer2Impl &nfc) |
56 | : trie(NULL), |
57 | ce32s(NULL), ces(NULL), contexts(NULL), base(NULL), |
58 | jamoCE32s(NULL), |
59 | nfcImpl(nfc), |
60 | numericPrimary(0x12000000), |
61 | ce32sLength(0), cesLength(0), contextsLength(0), |
62 | compressibleBytes(NULL), |
63 | unsafeBackwardSet(NULL), |
64 | fastLatinTable(NULL), fastLatinTableLength(0), |
65 | numScripts(0), scriptsIndex(NULL), scriptStarts(NULL), scriptStartsLength(0), |
66 | rootElements(NULL), rootElementsLength(0) {} |
67 | |
68 | uint32_t getCE32(UChar32 c) const { |
69 | return UTRIE2_GET32(trie, c); |
70 | } |
71 | |
72 | uint32_t getCE32FromSupplementary(UChar32 c) const { |
73 | return UTRIE2_GET32_FROM_SUPP(trie, c); |
74 | } |
75 | |
76 | UBool isDigit(UChar32 c) const { |
77 | return c < 0x660 ? c <= 0x39 && 0x30 <= c : |
78 | Collation::hasCE32Tag(getCE32(c), Collation::DIGIT_TAG); |
79 | } |
80 | |
81 | UBool isUnsafeBackward(UChar32 c, UBool numeric) const { |
82 | return unsafeBackwardSet->contains(c) || (numeric && isDigit(c)); |
83 | } |
84 | |
85 | UBool isCompressibleLeadByte(uint32_t b) const { |
86 | return compressibleBytes[b]; |
87 | } |
88 | |
89 | inline UBool isCompressiblePrimary(uint32_t p) const { |
90 | return isCompressibleLeadByte(p >> 24); |
91 | } |
92 | |
93 | /** |
94 | * Returns the CE32 from two contexts words. |
95 | * Access to the defaultCE32 for contraction and prefix matching. |
96 | */ |
97 | static uint32_t readCE32(const UChar *p) { |
98 | return ((uint32_t)p[0] << 16) | p[1]; |
99 | } |
100 | |
101 | /** |
102 | * Returns the CE32 for an indirect special CE32 (e.g., with DIGIT_TAG). |
103 | * Requires that ce32 is special. |
104 | */ |
105 | uint32_t getIndirectCE32(uint32_t ce32) const; |
106 | /** |
107 | * Returns the CE32 for an indirect special CE32 (e.g., with DIGIT_TAG), |
108 | * if ce32 is special. |
109 | */ |
110 | uint32_t getFinalCE32(uint32_t ce32) const; |
111 | |
112 | /** |
113 | * Computes a CE from c's ce32 which has the OFFSET_TAG. |
114 | */ |
115 | int64_t getCEFromOffsetCE32(UChar32 c, uint32_t ce32) const { |
116 | int64_t dataCE = ces[Collation::indexFromCE32(ce32)]; |
117 | return Collation::makeCE(Collation::getThreeBytePrimaryForOffsetData(c, dataCE)); |
118 | } |
119 | |
120 | /** |
121 | * Returns the single CE that c maps to. |
122 | * Sets U_UNSUPPORTED_ERROR if c does not map to a single CE. |
123 | */ |
124 | int64_t getSingleCE(UChar32 c, UErrorCode &errorCode) const; |
125 | |
126 | /** |
127 | * Returns the FCD16 value for code point c. c must be >= 0. |
128 | */ |
129 | uint16_t getFCD16(UChar32 c) const { |
130 | return nfcImpl.getFCD16(c); |
131 | } |
132 | |
133 | /** |
134 | * Returns the first primary for the script's reordering group. |
135 | * @return the primary with only the first primary lead byte of the group |
136 | * (not necessarily an actual root collator primary weight), |
137 | * or 0 if the script is unknown |
138 | */ |
139 | uint32_t getFirstPrimaryForGroup(int32_t script) const; |
140 | |
141 | /** |
142 | * Returns the last primary for the script's reordering group. |
143 | * @return the last primary of the group |
144 | * (not an actual root collator primary weight), |
145 | * or 0 if the script is unknown |
146 | */ |
147 | uint32_t getLastPrimaryForGroup(int32_t script) const; |
148 | |
149 | /** |
150 | * Finds the reordering group which contains the primary weight. |
151 | * @return the first script of the group, or -1 if the weight is beyond the last group |
152 | */ |
153 | int32_t getGroupForPrimary(uint32_t p) const; |
154 | |
155 | int32_t getEquivalentScripts(int32_t script, |
156 | int32_t dest[], int32_t capacity, UErrorCode &errorCode) const; |
157 | |
158 | /** |
159 | * Writes the permutation of primary-weight ranges |
160 | * for the given reordering of scripts and groups. |
161 | * The caller checks for illegal arguments and |
162 | * takes care of [DEFAULT] and memory allocation. |
163 | * |
164 | * Each list element will be a (limit, offset) pair as described |
165 | * for the CollationSettings::reorderRanges. |
166 | * The list will be empty if no ranges are reordered. |
167 | */ |
168 | void makeReorderRanges(const int32_t *reorder, int32_t length, |
169 | UVector32 &ranges, UErrorCode &errorCode) const; |
170 | |
171 | /** @see jamoCE32s */ |
172 | static const int32_t JAMO_CE32S_LENGTH = 19 + 21 + 27; |
173 | |
174 | /** Main lookup trie. */ |
175 | const UTrie2 *trie; |
176 | /** |
177 | * Array of CE32 values. |
178 | * At index 0 there must be CE32(U+0000) |
179 | * to support U+0000's special-tag for NUL-termination handling. |
180 | */ |
181 | const uint32_t *ce32s; |
182 | /** Array of CE values for expansions and OFFSET_TAG. */ |
183 | const int64_t *ces; |
184 | /** Array of prefix and contraction-suffix matching data. */ |
185 | const UChar *contexts; |
186 | /** Base collation data, or NULL if this data itself is a base. */ |
187 | const CollationData *base; |
188 | /** |
189 | * Simple array of JAMO_CE32S_LENGTH=19+21+27 CE32s, one per canonical Jamo L/V/T. |
190 | * They are normally simple CE32s, rarely expansions. |
191 | * For fast handling of HANGUL_TAG. |
192 | */ |
193 | const uint32_t *jamoCE32s; |
194 | const Normalizer2Impl &nfcImpl; |
195 | /** The single-byte primary weight (xx000000) for numeric collation. */ |
196 | uint32_t numericPrimary; |
197 | |
198 | int32_t ce32sLength; |
199 | int32_t cesLength; |
200 | int32_t contextsLength; |
201 | |
202 | /** 256 flags for which primary-weight lead bytes are compressible. */ |
203 | const UBool *compressibleBytes; |
204 | /** |
205 | * Set of code points that are unsafe for starting string comparison after an identical prefix, |
206 | * or in backwards CE iteration. |
207 | */ |
208 | const UnicodeSet *unsafeBackwardSet; |
209 | |
210 | /** |
211 | * Fast Latin table for common-Latin-text string comparisons. |
212 | * Data structure see class CollationFastLatin. |
213 | */ |
214 | const uint16_t *fastLatinTable; |
215 | int32_t fastLatinTableLength; |
216 | |
217 | /** |
218 | * Data for scripts and reordering groups. |
219 | * Uses include building a reordering permutation table and |
220 | * providing script boundaries to AlphabeticIndex. |
221 | */ |
222 | int32_t numScripts; |
223 | /** |
224 | * The length of scriptsIndex is numScripts+16. |
225 | * It maps from a UScriptCode or a special reorder code to an entry in scriptStarts. |
226 | * 16 special reorder codes (not all used) are mapped starting at numScripts. |
227 | * Up to MAX_NUM_SPECIAL_REORDER_CODES are codes for special groups like space/punct/digit. |
228 | * There are special codes at the end for reorder-reserved primary ranges. |
229 | * |
230 | * Multiple scripts may share a range and index, for example Hira & Kana. |
231 | */ |
232 | const uint16_t *scriptsIndex; |
233 | /** |
234 | * Start primary weight (top 16 bits only) for a group/script/reserved range |
235 | * indexed by scriptsIndex. |
236 | * The first range (separators & terminators) and the last range (trailing weights) |
237 | * are not reorderable, and no scriptsIndex entry points to them. |
238 | */ |
239 | const uint16_t *scriptStarts; |
240 | int32_t scriptStartsLength; |
241 | |
242 | /** |
243 | * Collation elements in the root collator. |
244 | * Used by the CollationRootElements class. The data structure is described there. |
245 | * NULL in a tailoring. |
246 | */ |
247 | const uint32_t *rootElements; |
248 | int32_t rootElementsLength; |
249 | |
250 | private: |
251 | int32_t getScriptIndex(int32_t script) const; |
252 | void makeReorderRanges(const int32_t *reorder, int32_t length, |
253 | UBool latinMustMove, |
254 | UVector32 &ranges, UErrorCode &errorCode) const; |
255 | int32_t addLowScriptRange(uint8_t table[], int32_t index, int32_t lowStart) const; |
256 | int32_t addHighScriptRange(uint8_t table[], int32_t index, int32_t highLimit) const; |
257 | }; |
258 | |
259 | U_NAMESPACE_END |
260 | |
261 | #endif // !UCONFIG_NO_COLLATION |
262 | #endif // __COLLATIONDATA_H__ |
263 | |