154dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius/*
254dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius*******************************************************************************
359d709d503bab6e2b61931737e662dd293b40578ccornelius* Copyright (C) 2013, International Business Machines
454dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius* Corporation and others.  All Rights Reserved.
554dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius*******************************************************************************
654dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius* dictionarydata.h
754dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius*
854dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius* created on: 2012may31
954dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius* created by: Markus W. Scherer & Maxime Serrano
1054dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius*/
1154dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius
1254dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius#ifndef __DICTIONARYDATA_H__
1354dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius#define __DICTIONARYDATA_H__
1454dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius
1554dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius#include "unicode/utypes.h"
1654dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius
1754dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius#if !UCONFIG_NO_BREAK_ITERATION
1854dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius
1954dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius#include "unicode/utext.h"
2054dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius#include "unicode/udata.h"
2154dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius#include "udataswp.h"
2254dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius#include "unicode/uobject.h"
2354dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius#include "unicode/ustringtrie.h"
2454dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius
2554dcd9b6a06071f647dac967e9e267abb9410720Craig CorneliusU_NAMESPACE_BEGIN
2654dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius
2754dcd9b6a06071f647dac967e9e267abb9410720Craig Corneliusclass UCharsTrie;
2854dcd9b6a06071f647dac967e9e267abb9410720Craig Corneliusclass BytesTrie;
2954dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius
3054dcd9b6a06071f647dac967e9e267abb9410720Craig Corneliusclass U_COMMON_API DictionaryData : public UMemory {
3154dcd9b6a06071f647dac967e9e267abb9410720Craig Corneliuspublic:
3259d709d503bab6e2b61931737e662dd293b40578ccornelius    static const int32_t TRIE_TYPE_BYTES; // = 0;
3359d709d503bab6e2b61931737e662dd293b40578ccornelius    static const int32_t TRIE_TYPE_UCHARS; // = 1;
3459d709d503bab6e2b61931737e662dd293b40578ccornelius    static const int32_t TRIE_TYPE_MASK; // = 7;
3559d709d503bab6e2b61931737e662dd293b40578ccornelius    static const int32_t TRIE_HAS_VALUES; // = 8;
3659d709d503bab6e2b61931737e662dd293b40578ccornelius
3759d709d503bab6e2b61931737e662dd293b40578ccornelius    static const int32_t TRANSFORM_NONE; // = 0;
3859d709d503bab6e2b61931737e662dd293b40578ccornelius    static const int32_t TRANSFORM_TYPE_OFFSET; // = 0x1000000;
3959d709d503bab6e2b61931737e662dd293b40578ccornelius    static const int32_t TRANSFORM_TYPE_MASK; // = 0x7f000000;
4059d709d503bab6e2b61931737e662dd293b40578ccornelius    static const int32_t TRANSFORM_OFFSET_MASK; // = 0x1fffff;
4154dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius
4254dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    enum {
4354dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius        // Byte offsets from the start of the data, after the generic header.
4454dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius        IX_STRING_TRIE_OFFSET,
4554dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius        IX_RESERVED1_OFFSET,
4654dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius        IX_RESERVED2_OFFSET,
4754dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius        IX_TOTAL_SIZE,
4854dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius
4954dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius        // Trie type: TRIE_HAS_VALUES | TRIE_TYPE_BYTES etc.
5054dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius        IX_TRIE_TYPE,
5154dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius        // Transform specification: TRANSFORM_TYPE_OFFSET | 0xe00 etc.
5254dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius        IX_TRANSFORM,
5354dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius
5454dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius        IX_RESERVED6,
5554dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius        IX_RESERVED7,
5654dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius        IX_COUNT
5754dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    };
5854dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius};
5954dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius
6054dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius/**
6154dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius * Wrapper class around generic dictionaries, implementing matches().
6254dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius * getType() should return a TRIE_TYPE_??? constant from DictionaryData.
6354dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *
6454dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius * All implementations of this interface must be thread-safe if they are to be used inside of the
6554dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius * dictionary-based break iteration code.
6654dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius */
6754dcd9b6a06071f647dac967e9e267abb9410720Craig Corneliusclass U_COMMON_API DictionaryMatcher : public UMemory {
6854dcd9b6a06071f647dac967e9e267abb9410720Craig Corneliuspublic:
6954dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    virtual ~DictionaryMatcher();
7054dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    // this should emulate CompactTrieDictionary::matches()
7154dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    virtual int32_t matches(UText *text, int32_t maxLength, int32_t *lengths, int32_t &count,
7254dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius                            int32_t limit, int32_t *values = NULL) const = 0;
7354dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    /** @return DictionaryData::TRIE_TYPE_XYZ */
7454dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    virtual int32_t getType() const = 0;
7554dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius};
7654dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius
7754dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius// Implementation of the DictionaryMatcher interface for a UCharsTrie dictionary
7854dcd9b6a06071f647dac967e9e267abb9410720Craig Corneliusclass U_COMMON_API UCharsDictionaryMatcher : public DictionaryMatcher {
7954dcd9b6a06071f647dac967e9e267abb9410720Craig Corneliuspublic:
8054dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    // constructs a new UCharsDictionaryMatcher.
8154dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    // The UDataMemory * will be closed on this object's destruction.
8254dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    UCharsDictionaryMatcher(const UChar *c, UDataMemory *f) : characters(c), file(f) { }
8354dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    virtual ~UCharsDictionaryMatcher();
8454dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    virtual int32_t matches(UText *text, int32_t maxLength, int32_t *lengths, int32_t &count,
8554dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius                            int32_t limit, int32_t *values = NULL) const;
8654dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    virtual int32_t getType() const;
8754dcd9b6a06071f647dac967e9e267abb9410720Craig Corneliusprivate:
8854dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    const UChar *characters;
8954dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    UDataMemory *file;
9054dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius};
9154dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius
9254dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius// Implementation of the DictionaryMatcher interface for a BytesTrie dictionary
9354dcd9b6a06071f647dac967e9e267abb9410720Craig Corneliusclass U_COMMON_API BytesDictionaryMatcher : public DictionaryMatcher {
9454dcd9b6a06071f647dac967e9e267abb9410720Craig Corneliuspublic:
9554dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    // constructs a new BytesTrieDictionaryMatcher
9654dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    // the transform constant should be the constant read from the file, not a masked version!
9754dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    // the UDataMemory * fed in here will be closed on this object's destruction
9854dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    BytesDictionaryMatcher(const char *c, int32_t t, UDataMemory *f)
9954dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius            : characters(c), transformConstant(t), file(f) { }
10054dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    virtual ~BytesDictionaryMatcher();
10154dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    virtual int32_t matches(UText *text, int32_t maxLength, int32_t *lengths, int32_t &count,
10254dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius                            int32_t limit, int32_t *values = NULL) const;
10354dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    virtual int32_t getType() const;
10454dcd9b6a06071f647dac967e9e267abb9410720Craig Corneliusprivate:
10554dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    UChar32 transform(UChar32 c) const;
10654dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius
10754dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    const char *characters;
10854dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    int32_t transformConstant;
10954dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius    UDataMemory *file;
11054dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius};
11154dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius
11254dcd9b6a06071f647dac967e9e267abb9410720Craig CorneliusU_NAMESPACE_END
11354dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius
11454dcd9b6a06071f647dac967e9e267abb9410720Craig CorneliusU_CAPI int32_t U_EXPORT2
11554dcd9b6a06071f647dac967e9e267abb9410720Craig Corneliusudict_swap(const UDataSwapper *ds, const void *inData, int32_t length, void *outData, UErrorCode *pErrorCode);
11654dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius
11754dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius/**
11854dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius * Format of dictionary .dict data files.
11954dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius * Format version 1.0.
12054dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *
12154dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius * A dictionary .dict data file contains a byte-serialized BytesTrie or
12254dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius * a UChars-serialized UCharsTrie.
12354dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius * Such files are used in dictionary-based break iteration (DBBI).
12454dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *
12554dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius * For a BytesTrie, a transformation type is specified for
12654dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius * transforming Unicode strings into byte sequences.
12754dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *
12854dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius * A .dict file begins with a standard ICU data file header
12954dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius * (DataHeader, see ucmndata.h and unicode/udata.h).
13054dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius * The UDataInfo.dataVersion field is currently unused (set to 0.0.0.0).
13154dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *
13254dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius * After the header, the file contains the following parts.
13354dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius * Constants are defined in the DictionaryData class.
13454dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *
13554dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius * For the data structure of BytesTrie & UCharsTrie see
13654dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius * http://site.icu-project.org/design/struct/tries
13754dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius * and the bytestrie.h and ucharstrie.h header files.
13854dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *
13954dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius * int32_t indexes[indexesLength]; -- indexesLength=indexes[IX_STRING_TRIE_OFFSET]/4;
14054dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *
14154dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *      The first four indexes are byte offsets in ascending order.
14254dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *      Each byte offset marks the start of the next part in the data file,
14354dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *      and the end of the previous one.
14454dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *      When two consecutive byte offsets are the same, then the corresponding part is empty.
14554dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *      Byte offsets are offsets from after the header,
14654dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *      that is, from the beginning of the indexes[].
14754dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *      Each part starts at an offset with proper alignment for its data.
14854dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *      If necessary, the previous part may include padding bytes to achieve this alignment.
14954dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *
15054dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *      trieType=indexes[IX_TRIE_TYPE] defines the trie type.
15154dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *      transform=indexes[IX_TRANSFORM] defines the Unicode-to-bytes transformation.
15254dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *          If the transformation type is TRANSFORM_TYPE_OFFSET,
15354dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *          then the lower 21 bits contain the offset code point.
15454dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *          Each code point c is mapped to byte b = (c - offset).
15554dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *          Code points outside the range offset..(offset+0xff) cannot be mapped
15654dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *          and do not occur in the dictionary.
15754dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *
15854dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius * stringTrie; -- a serialized BytesTrie or UCharsTrie
15954dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *
16054dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *      The dictionary maps strings to specific values (TRIE_HAS_VALUES bit set in trieType),
16154dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius *      or it maps all strings to 0 (TRIE_HAS_VALUES bit not set).
16254dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius */
16354dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius
16454dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius#endif  /* !UCONFIG_NO_BREAK_ITERATION */
16554dcd9b6a06071f647dac967e9e267abb9410720Craig Cornelius#endif  /* __DICTIONARYDATA_H__ */
166