1/*
2*******************************************************************************
3*   Copyright (C) 2010-2011, International Business Machines
4*   Corporation and others.  All Rights Reserved.
5*******************************************************************************
6*   file name:  bytestriebuilder.h
7*   encoding:   US-ASCII
8*   tab size:   8 (not used)
9*   indentation:4
10*
11*   created on: 2010sep25
12*   created by: Markus W. Scherer
13*/
14
15#ifndef __BYTESTRIEBUILDER_H__
16#define __BYTESTRIEBUILDER_H__
17
18#include "unicode/utypes.h"
19#include "unicode/bytestrie.h"
20#include "unicode/stringpiece.h"
21#include "unicode/stringtriebuilder.h"
22
23U_NAMESPACE_BEGIN
24
25class BytesTrieElement;
26class CharString;
27
28/**
29 * Builder class for BytesTrie.
30 *
31 * This class is not intended for public subclassing.
32 * @draft ICU 4.8
33 */
34class U_COMMON_API BytesTrieBuilder : public StringTrieBuilder {
35public:
36    /**
37     * Constructs an empty builder.
38     * @param errorCode Standard ICU error code.
39     * @draft ICU 4.8
40     */
41    BytesTrieBuilder(UErrorCode &errorCode);
42
43    /**
44     * Destructor.
45     * @draft ICU 4.8
46     */
47    virtual ~BytesTrieBuilder();
48
49    /**
50     * Adds a (byte sequence, value) pair.
51     * The byte sequence must be unique.
52     * The bytes will be copied; the builder does not keep
53     * a reference to the input StringPiece or its data().
54     * @param s The input byte sequence.
55     * @param value The value associated with this byte sequence.
56     * @param errorCode Standard ICU error code. Its input value must
57     *                  pass the U_SUCCESS() test, or else the function returns
58     *                  immediately. Check for U_FAILURE() on output or use with
59     *                  function chaining. (See User Guide for details.)
60     * @return *this
61     * @draft ICU 4.8
62     */
63    BytesTrieBuilder &add(const StringPiece &s, int32_t value, UErrorCode &errorCode);
64
65    /**
66     * Builds a BytesTrie for the add()ed data.
67     * Once built, no further data can be add()ed until clear() is called.
68     *
69     * This method passes ownership of the builder's internal result array to the new trie object.
70     * Another call to any build() variant will re-serialize the trie.
71     * After clear() has been called, a new array will be used as well.
72     * @param buildOption Build option, see UStringTrieBuildOption.
73     * @param errorCode Standard ICU error code. Its input value must
74     *                  pass the U_SUCCESS() test, or else the function returns
75     *                  immediately. Check for U_FAILURE() on output or use with
76     *                  function chaining. (See User Guide for details.)
77     * @return A new BytesTrie for the add()ed data.
78     * @draft ICU 4.8
79     */
80    BytesTrie *build(UStringTrieBuildOption buildOption, UErrorCode &errorCode);
81
82    /**
83     * Builds a BytesTrie for the add()ed data and byte-serializes it.
84     * Once built, no further data can be add()ed until clear() is called.
85     *
86     * Multiple calls to buildStringPiece() return StringPieces referring to the
87     * builder's same byte array, without rebuilding.
88     * If buildStringPiece() is called after build(), the trie will be
89     * re-serialized into a new array.
90     * If build() is called after buildStringPiece(), the trie object will become
91     * the owner of the previously returned array.
92     * After clear() has been called, a new array will be used as well.
93     * @param buildOption Build option, see UStringTrieBuildOption.
94     * @param errorCode Standard ICU error code. Its input value must
95     *                  pass the U_SUCCESS() test, or else the function returns
96     *                  immediately. Check for U_FAILURE() on output or use with
97     *                  function chaining. (See User Guide for details.)
98     * @return A StringPiece which refers to the byte-serialized BytesTrie for the add()ed data.
99     * @draft ICU 4.8
100     */
101    StringPiece buildStringPiece(UStringTrieBuildOption buildOption, UErrorCode &errorCode);
102
103    /**
104     * Removes all (byte sequence, value) pairs.
105     * New data can then be add()ed and a new trie can be built.
106     * @return *this
107     * @draft ICU 4.8
108     */
109    BytesTrieBuilder &clear();
110
111private:
112    BytesTrieBuilder(const BytesTrieBuilder &other);  // no copy constructor
113    BytesTrieBuilder &operator=(const BytesTrieBuilder &other);  // no assignment operator
114
115    void buildBytes(UStringTrieBuildOption buildOption, UErrorCode &errorCode);
116
117    virtual int32_t getElementStringLength(int32_t i) const;
118    virtual UChar getElementUnit(int32_t i, int32_t byteIndex) const;
119    virtual int32_t getElementValue(int32_t i) const;
120
121    virtual int32_t getLimitOfLinearMatch(int32_t first, int32_t last, int32_t byteIndex) const;
122
123    virtual int32_t countElementUnits(int32_t start, int32_t limit, int32_t byteIndex) const;
124    virtual int32_t skipElementsBySomeUnits(int32_t i, int32_t byteIndex, int32_t count) const;
125    virtual int32_t indexOfElementWithNextUnit(int32_t i, int32_t byteIndex, UChar byte) const;
126
127    virtual UBool matchNodesCanHaveValues() const { return FALSE; }
128
129    virtual int32_t getMaxBranchLinearSubNodeLength() const { return BytesTrie::kMaxBranchLinearSubNodeLength; }
130    virtual int32_t getMinLinearMatch() const { return BytesTrie::kMinLinearMatch; }
131    virtual int32_t getMaxLinearMatchLength() const { return BytesTrie::kMaxLinearMatchLength; }
132
133    /**
134     * @internal
135     */
136    class BTLinearMatchNode : public LinearMatchNode {
137    public:
138        BTLinearMatchNode(const char *units, int32_t len, Node *nextNode);
139        virtual UBool operator==(const Node &other) const;
140        virtual void write(StringTrieBuilder &builder);
141    private:
142        const char *s;
143    };
144
145    virtual Node *createLinearMatchNode(int32_t i, int32_t byteIndex, int32_t length,
146                                        Node *nextNode) const;
147
148    UBool ensureCapacity(int32_t length);
149    virtual int32_t write(int32_t byte);
150    int32_t write(const char *b, int32_t length);
151    virtual int32_t writeElementUnits(int32_t i, int32_t byteIndex, int32_t length);
152    virtual int32_t writeValueAndFinal(int32_t i, UBool isFinal);
153    virtual int32_t writeValueAndType(UBool hasValue, int32_t value, int32_t node);
154    virtual int32_t writeDeltaTo(int32_t jumpTarget);
155
156    CharString *strings;  // Pointer not object so we need not #include internal charstr.h.
157    BytesTrieElement *elements;
158    int32_t elementsCapacity;
159    int32_t elementsLength;
160
161    // Byte serialization of the trie.
162    // Grows from the back: bytesLength measures from the end of the buffer!
163    char *bytes;
164    int32_t bytesCapacity;
165    int32_t bytesLength;
166};
167
168U_NAMESPACE_END
169
170#endif  // __BYTESTRIEBUILDER_H__
171