Victor Chang | 7322950 | 2020-09-17 13:39:19 +0100 | [diff] [blame] | 1 | // © 2016 and later: Unicode, Inc. and others. |
| 2 | // License & terms of use: http://www.unicode.org/copyright.html |
| 3 | /* |
| 4 | ******************************************************************************* |
| 5 | * Copyright (C) 2014, International Business Machines |
| 6 | * Corporation and others. All Rights Reserved. |
| 7 | ******************************************************************************* |
| 8 | * dictionarydata.h |
| 9 | * |
| 10 | * created on: 2012may31 |
| 11 | * created by: Markus W. Scherer & Maxime Serrano |
| 12 | */ |
| 13 | |
| 14 | #ifndef __DICTIONARYDATA_H__ |
| 15 | #define __DICTIONARYDATA_H__ |
| 16 | |
| 17 | #include "unicode/utypes.h" |
| 18 | |
| 19 | #if !UCONFIG_NO_BREAK_ITERATION |
| 20 | |
| 21 | #include "unicode/utext.h" |
| 22 | #include "unicode/udata.h" |
| 23 | #include "udataswp.h" |
| 24 | #include "unicode/uobject.h" |
| 25 | #include "unicode/ustringtrie.h" |
| 26 | |
| 27 | U_NAMESPACE_BEGIN |
| 28 | |
| 29 | class UCharsTrie; |
| 30 | class BytesTrie; |
| 31 | |
| 32 | class U_COMMON_API DictionaryData : public UMemory { |
| 33 | public: |
| 34 | static const int32_t TRIE_TYPE_BYTES; // = 0; |
| 35 | static const int32_t TRIE_TYPE_UCHARS; // = 1; |
| 36 | static const int32_t TRIE_TYPE_MASK; // = 7; |
| 37 | static const int32_t TRIE_HAS_VALUES; // = 8; |
| 38 | |
| 39 | static const int32_t TRANSFORM_NONE; // = 0; |
| 40 | static const int32_t TRANSFORM_TYPE_OFFSET; // = 0x1000000; |
| 41 | static const int32_t TRANSFORM_TYPE_MASK; // = 0x7f000000; |
| 42 | static const int32_t TRANSFORM_OFFSET_MASK; // = 0x1fffff; |
| 43 | |
| 44 | enum { |
| 45 | // Byte offsets from the start of the data, after the generic header. |
| 46 | IX_STRING_TRIE_OFFSET, |
| 47 | IX_RESERVED1_OFFSET, |
| 48 | IX_RESERVED2_OFFSET, |
| 49 | IX_TOTAL_SIZE, |
| 50 | |
| 51 | // Trie type: TRIE_HAS_VALUES | TRIE_TYPE_BYTES etc. |
| 52 | IX_TRIE_TYPE, |
| 53 | // Transform specification: TRANSFORM_TYPE_OFFSET | 0xe00 etc. |
| 54 | IX_TRANSFORM, |
| 55 | |
| 56 | IX_RESERVED6, |
| 57 | IX_RESERVED7, |
| 58 | IX_COUNT |
| 59 | }; |
| 60 | }; |
| 61 | |
| 62 | /** |
| 63 | * Wrapper class around generic dictionaries, implementing matches(). |
| 64 | * getType() should return a TRIE_TYPE_??? constant from DictionaryData. |
| 65 | * |
| 66 | * All implementations of this interface must be thread-safe if they are to be used inside of the |
| 67 | * dictionary-based break iteration code. |
| 68 | */ |
| 69 | class U_COMMON_API DictionaryMatcher : public UMemory { |
| 70 | public: |
| 71 | DictionaryMatcher() {} |
| 72 | virtual ~DictionaryMatcher(); |
| 73 | // this should emulate CompactTrieDictionary::matches() |
| 74 | /* @param text The text in which to look for matching words. Matching begins |
| 75 | * at the current position of the UText. |
| 76 | * @param maxLength The max length of match to consider. Units are the native indexing |
| 77 | * units of the UText. |
| 78 | * @param limit Capacity of output arrays, which is also the maximum number of |
| 79 | * matching words to be found. |
| 80 | * @param lengths output array, filled with the lengths of the matches, in order, |
| 81 | * from shortest to longest. Lengths are in native indexing units |
| 82 | * of the UText. May be NULL. |
| 83 | * @param cpLengths output array, filled with the lengths of the matches, in order, |
| 84 | * from shortest to longest. Lengths are the number of Unicode code points. |
| 85 | * May be NULL. |
| 86 | * @param values Output array, filled with the values associated with the words found. |
| 87 | * May be NULL. |
| 88 | * @param prefix Output parameter, the code point length of the prefix match, even if that |
| 89 | * prefix didn't lead to a complete word. Will always be >= the cpLength |
| 90 | * of the longest complete word matched. May be NULL. |
| 91 | * @return Number of matching words found. |
| 92 | */ |
| 93 | virtual int32_t matches(UText *text, int32_t maxLength, int32_t limit, |
| 94 | int32_t *lengths, int32_t *cpLengths, int32_t *values, |
| 95 | int32_t *prefix) const = 0; |
| 96 | |
| 97 | /** @return DictionaryData::TRIE_TYPE_XYZ */ |
| 98 | virtual int32_t getType() const = 0; |
| 99 | }; |
| 100 | |
| 101 | // Implementation of the DictionaryMatcher interface for a UCharsTrie dictionary |
| 102 | class U_COMMON_API UCharsDictionaryMatcher : public DictionaryMatcher { |
| 103 | public: |
| 104 | // constructs a new UCharsDictionaryMatcher. |
| 105 | // The UDataMemory * will be closed on this object's destruction. |
| 106 | UCharsDictionaryMatcher(const UChar *c, UDataMemory *f) : characters(c), file(f) { } |
| 107 | virtual ~UCharsDictionaryMatcher(); |
| 108 | virtual int32_t matches(UText *text, int32_t maxLength, int32_t limit, |
| 109 | int32_t *lengths, int32_t *cpLengths, int32_t *values, |
| 110 | int32_t *prefix) const; |
| 111 | virtual int32_t getType() const; |
| 112 | private: |
| 113 | const UChar *characters; |
| 114 | UDataMemory *file; |
| 115 | }; |
| 116 | |
| 117 | // Implementation of the DictionaryMatcher interface for a BytesTrie dictionary |
| 118 | class U_COMMON_API BytesDictionaryMatcher : public DictionaryMatcher { |
| 119 | public: |
| 120 | // constructs a new BytesTrieDictionaryMatcher |
| 121 | // the transform constant should be the constant read from the file, not a masked version! |
| 122 | // the UDataMemory * fed in here will be closed on this object's destruction |
| 123 | BytesDictionaryMatcher(const char *c, int32_t t, UDataMemory *f) |
| 124 | : characters(c), transformConstant(t), file(f) { } |
| 125 | virtual ~BytesDictionaryMatcher(); |
| 126 | virtual int32_t matches(UText *text, int32_t maxLength, int32_t limit, |
| 127 | int32_t *lengths, int32_t *cpLengths, int32_t *values, |
| 128 | int32_t *prefix) const; |
| 129 | virtual int32_t getType() const; |
| 130 | private: |
| 131 | UChar32 transform(UChar32 c) const; |
| 132 | |
| 133 | const char *characters; |
| 134 | int32_t transformConstant; |
| 135 | UDataMemory *file; |
| 136 | }; |
| 137 | |
| 138 | U_NAMESPACE_END |
| 139 | |
| 140 | U_CAPI int32_t U_EXPORT2 |
| 141 | udict_swap(const UDataSwapper *ds, const void *inData, int32_t length, void *outData, UErrorCode *pErrorCode); |
| 142 | |
| 143 | /** |
| 144 | * Format of dictionary .dict data files. |
| 145 | * Format version 1.0. |
| 146 | * |
| 147 | * A dictionary .dict data file contains a byte-serialized BytesTrie or |
| 148 | * a UChars-serialized UCharsTrie. |
| 149 | * Such files are used in dictionary-based break iteration (DBBI). |
| 150 | * |
| 151 | * For a BytesTrie, a transformation type is specified for |
| 152 | * transforming Unicode strings into byte sequences. |
| 153 | * |
| 154 | * A .dict file begins with a standard ICU data file header |
| 155 | * (DataHeader, see ucmndata.h and unicode/udata.h). |
| 156 | * The UDataInfo.dataVersion field is currently unused (set to 0.0.0.0). |
| 157 | * |
| 158 | * After the header, the file contains the following parts. |
| 159 | * Constants are defined in the DictionaryData class. |
| 160 | * |
| 161 | * For the data structure of BytesTrie & UCharsTrie see |
| 162 | * http://site.icu-project.org/design/struct/tries |
| 163 | * and the bytestrie.h and ucharstrie.h header files. |
| 164 | * |
| 165 | * int32_t indexes[indexesLength]; -- indexesLength=indexes[IX_STRING_TRIE_OFFSET]/4; |
| 166 | * |
| 167 | * The first four indexes are byte offsets in ascending order. |
| 168 | * Each byte offset marks the start of the next part in the data file, |
| 169 | * and the end of the previous one. |
| 170 | * When two consecutive byte offsets are the same, then the corresponding part is empty. |
| 171 | * Byte offsets are offsets from after the header, |
| 172 | * that is, from the beginning of the indexes[]. |
| 173 | * Each part starts at an offset with proper alignment for its data. |
| 174 | * If necessary, the previous part may include padding bytes to achieve this alignment. |
| 175 | * |
| 176 | * trieType=indexes[IX_TRIE_TYPE] defines the trie type. |
| 177 | * transform=indexes[IX_TRANSFORM] defines the Unicode-to-bytes transformation. |
| 178 | * If the transformation type is TRANSFORM_TYPE_OFFSET, |
| 179 | * then the lower 21 bits contain the offset code point. |
| 180 | * Each code point c is mapped to byte b = (c - offset). |
| 181 | * Code points outside the range offset..(offset+0xff) cannot be mapped |
| 182 | * and do not occur in the dictionary. |
| 183 | * |
| 184 | * stringTrie; -- a serialized BytesTrie or UCharsTrie |
| 185 | * |
| 186 | * The dictionary maps strings to specific values (TRIE_HAS_VALUES bit set in trieType), |
| 187 | * or it maps all strings to 0 (TRIE_HAS_VALUES bit not set). |
| 188 | */ |
| 189 | |
| 190 | #endif /* !UCONFIG_NO_BREAK_ITERATION */ |
| 191 | #endif /* __DICTIONARYDATA_H__ */ |