|
1 /* |
|
2 ******************************************************************************* |
|
3 * Copyright (C) 2013, International Business Machines |
|
4 * Corporation and others. All Rights Reserved. |
|
5 ******************************************************************************* |
|
6 * dictionarydata.h |
|
7 * |
|
8 * created on: 2012may31 |
|
9 * created by: Markus W. Scherer & Maxime Serrano |
|
10 */ |
|
11 |
|
12 #ifndef __DICTIONARYDATA_H__ |
|
13 #define __DICTIONARYDATA_H__ |
|
14 |
|
15 #include "unicode/utypes.h" |
|
16 |
|
17 #if !UCONFIG_NO_BREAK_ITERATION |
|
18 |
|
19 #include "unicode/utext.h" |
|
20 #include "unicode/udata.h" |
|
21 #include "udataswp.h" |
|
22 #include "unicode/uobject.h" |
|
23 #include "unicode/ustringtrie.h" |
|
24 |
|
25 U_NAMESPACE_BEGIN |
|
26 |
|
27 class UCharsTrie; |
|
28 class BytesTrie; |
|
29 |
|
30 class U_COMMON_API DictionaryData : public UMemory { |
|
31 public: |
|
32 static const int32_t TRIE_TYPE_BYTES; // = 0; |
|
33 static const int32_t TRIE_TYPE_UCHARS; // = 1; |
|
34 static const int32_t TRIE_TYPE_MASK; // = 7; |
|
35 static const int32_t TRIE_HAS_VALUES; // = 8; |
|
36 |
|
37 static const int32_t TRANSFORM_NONE; // = 0; |
|
38 static const int32_t TRANSFORM_TYPE_OFFSET; // = 0x1000000; |
|
39 static const int32_t TRANSFORM_TYPE_MASK; // = 0x7f000000; |
|
40 static const int32_t TRANSFORM_OFFSET_MASK; // = 0x1fffff; |
|
41 |
|
42 enum { |
|
43 // Byte offsets from the start of the data, after the generic header. |
|
44 IX_STRING_TRIE_OFFSET, |
|
45 IX_RESERVED1_OFFSET, |
|
46 IX_RESERVED2_OFFSET, |
|
47 IX_TOTAL_SIZE, |
|
48 |
|
49 // Trie type: TRIE_HAS_VALUES | TRIE_TYPE_BYTES etc. |
|
50 IX_TRIE_TYPE, |
|
51 // Transform specification: TRANSFORM_TYPE_OFFSET | 0xe00 etc. |
|
52 IX_TRANSFORM, |
|
53 |
|
54 IX_RESERVED6, |
|
55 IX_RESERVED7, |
|
56 IX_COUNT |
|
57 }; |
|
58 }; |
|
59 |
|
60 /** |
|
61 * Wrapper class around generic dictionaries, implementing matches(). |
|
62 * getType() should return a TRIE_TYPE_??? constant from DictionaryData. |
|
63 * |
|
64 * All implementations of this interface must be thread-safe if they are to be used inside of the |
|
65 * dictionary-based break iteration code. |
|
66 */ |
|
67 class U_COMMON_API DictionaryMatcher : public UMemory { |
|
68 public: |
|
69 virtual ~DictionaryMatcher(); |
|
70 // this should emulate CompactTrieDictionary::matches() |
|
71 virtual int32_t matches(UText *text, int32_t maxLength, int32_t *lengths, int32_t &count, |
|
72 int32_t limit, int32_t *values = NULL) const = 0; |
|
73 /** @return DictionaryData::TRIE_TYPE_XYZ */ |
|
74 virtual int32_t getType() const = 0; |
|
75 }; |
|
76 |
|
77 // Implementation of the DictionaryMatcher interface for a UCharsTrie dictionary |
|
78 class U_COMMON_API UCharsDictionaryMatcher : public DictionaryMatcher { |
|
79 public: |
|
80 // constructs a new UCharsDictionaryMatcher. |
|
81 // The UDataMemory * will be closed on this object's destruction. |
|
82 UCharsDictionaryMatcher(const UChar *c, UDataMemory *f) : characters(c), file(f) { } |
|
83 virtual ~UCharsDictionaryMatcher(); |
|
84 virtual int32_t matches(UText *text, int32_t maxLength, int32_t *lengths, int32_t &count, |
|
85 int32_t limit, int32_t *values = NULL) const; |
|
86 virtual int32_t getType() const; |
|
87 private: |
|
88 const UChar *characters; |
|
89 UDataMemory *file; |
|
90 }; |
|
91 |
|
92 // Implementation of the DictionaryMatcher interface for a BytesTrie dictionary |
|
93 class U_COMMON_API BytesDictionaryMatcher : public DictionaryMatcher { |
|
94 public: |
|
95 // constructs a new BytesTrieDictionaryMatcher |
|
96 // the transform constant should be the constant read from the file, not a masked version! |
|
97 // the UDataMemory * fed in here will be closed on this object's destruction |
|
98 BytesDictionaryMatcher(const char *c, int32_t t, UDataMemory *f) |
|
99 : characters(c), transformConstant(t), file(f) { } |
|
100 virtual ~BytesDictionaryMatcher(); |
|
101 virtual int32_t matches(UText *text, int32_t maxLength, int32_t *lengths, int32_t &count, |
|
102 int32_t limit, int32_t *values = NULL) const; |
|
103 virtual int32_t getType() const; |
|
104 private: |
|
105 UChar32 transform(UChar32 c) const; |
|
106 |
|
107 const char *characters; |
|
108 int32_t transformConstant; |
|
109 UDataMemory *file; |
|
110 }; |
|
111 |
|
112 U_NAMESPACE_END |
|
113 |
|
114 U_CAPI int32_t U_EXPORT2 |
|
115 udict_swap(const UDataSwapper *ds, const void *inData, int32_t length, void *outData, UErrorCode *pErrorCode); |
|
116 |
|
117 /** |
|
118 * Format of dictionary .dict data files. |
|
119 * Format version 1.0. |
|
120 * |
|
121 * A dictionary .dict data file contains a byte-serialized BytesTrie or |
|
122 * a UChars-serialized UCharsTrie. |
|
123 * Such files are used in dictionary-based break iteration (DBBI). |
|
124 * |
|
125 * For a BytesTrie, a transformation type is specified for |
|
126 * transforming Unicode strings into byte sequences. |
|
127 * |
|
128 * A .dict file begins with a standard ICU data file header |
|
129 * (DataHeader, see ucmndata.h and unicode/udata.h). |
|
130 * The UDataInfo.dataVersion field is currently unused (set to 0.0.0.0). |
|
131 * |
|
132 * After the header, the file contains the following parts. |
|
133 * Constants are defined in the DictionaryData class. |
|
134 * |
|
135 * For the data structure of BytesTrie & UCharsTrie see |
|
136 * http://site.icu-project.org/design/struct/tries |
|
137 * and the bytestrie.h and ucharstrie.h header files. |
|
138 * |
|
139 * int32_t indexes[indexesLength]; -- indexesLength=indexes[IX_STRING_TRIE_OFFSET]/4; |
|
140 * |
|
141 * The first four indexes are byte offsets in ascending order. |
|
142 * Each byte offset marks the start of the next part in the data file, |
|
143 * and the end of the previous one. |
|
144 * When two consecutive byte offsets are the same, then the corresponding part is empty. |
|
145 * Byte offsets are offsets from after the header, |
|
146 * that is, from the beginning of the indexes[]. |
|
147 * Each part starts at an offset with proper alignment for its data. |
|
148 * If necessary, the previous part may include padding bytes to achieve this alignment. |
|
149 * |
|
150 * trieType=indexes[IX_TRIE_TYPE] defines the trie type. |
|
151 * transform=indexes[IX_TRANSFORM] defines the Unicode-to-bytes transformation. |
|
152 * If the transformation type is TRANSFORM_TYPE_OFFSET, |
|
153 * then the lower 21 bits contain the offset code point. |
|
154 * Each code point c is mapped to byte b = (c - offset). |
|
155 * Code points outside the range offset..(offset+0xff) cannot be mapped |
|
156 * and do not occur in the dictionary. |
|
157 * |
|
158 * stringTrie; -- a serialized BytesTrie or UCharsTrie |
|
159 * |
|
160 * The dictionary maps strings to specific values (TRIE_HAS_VALUES bit set in trieType), |
|
161 * or it maps all strings to 0 (TRIE_HAS_VALUES bit not set). |
|
162 */ |
|
163 |
|
164 #endif /* !UCONFIG_NO_BREAK_ITERATION */ |
|
165 #endif /* __DICTIONARYDATA_H__ */ |