1// © 2016 and later: Unicode, Inc. and others.
2// License & terms of use: http://www.unicode.org/copyright.html
3/*
4*******************************************************************************
5* Copyright (C) 2010-2016, International Business Machines
6* Corporation and others. All Rights Reserved.
7*******************************************************************************
8* file name: bytestriebuilder.h
9* encoding: UTF-8
10* tab size: 8 (not used)
11* indentation:4
12*
13* created on: 2010sep25
14* created by: Markus W. Scherer
15*/
16
17/**
18 * \file
19 * \brief C++ API: Builder for icu::BytesTrie
20 */
21
22#ifndef __BYTESTRIEBUILDER_H__
23#define __BYTESTRIEBUILDER_H__
24
25#include "unicode/utypes.h"
26
27#if U_SHOW_CPLUSPLUS_API
28
29#include "unicode/bytestrie.h"
30#include "unicode/stringpiece.h"
31#include "unicode/stringtriebuilder.h"
32
33class BytesTrieTest;
34
35U_NAMESPACE_BEGIN
36
37class BytesTrieElement;
38class CharString;
39/**
40 * Builder class for BytesTrie.
41 *
42 * This class is not intended for public subclassing.
43 * @stable ICU 4.8
44 */
45class U_COMMON_API BytesTrieBuilder : public StringTrieBuilder {
46public:
47 /**
48 * Constructs an empty builder.
49 * @param errorCode Standard ICU error code.
50 * @stable ICU 4.8
51 */
52 BytesTrieBuilder(UErrorCode &errorCode);
53
54 /**
55 * Destructor.
56 * @stable ICU 4.8
57 */
58 virtual ~BytesTrieBuilder();
59
60 /**
61 * Adds a (byte sequence, value) pair.
62 * The byte sequence must be unique.
63 * The bytes will be copied; the builder does not keep
64 * a reference to the input StringPiece or its data().
65 * @param s The input byte sequence.
66 * @param value The value associated with this byte sequence.
67 * @param errorCode Standard ICU error code. Its input value must
68 * pass the U_SUCCESS() test, or else the function returns
69 * immediately. Check for U_FAILURE() on output or use with
70 * function chaining. (See User Guide for details.)
71 * @return *this
72 * @stable ICU 4.8
73 */
74 BytesTrieBuilder &add(StringPiece s, int32_t value, UErrorCode &errorCode);
75
76 /**
77 * Builds a BytesTrie for the add()ed data.
78 * Once built, no further data can be add()ed until clear() is called.
79 *
80 * A BytesTrie cannot be empty. At least one (byte sequence, value) pair
81 * must have been add()ed.
82 *
83 * This method passes ownership of the builder's internal result array to the new trie object.
84 * Another call to any build() variant will re-serialize the trie.
85 * After clear() has been called, a new array will be used as well.
86 * @param buildOption Build option, see UStringTrieBuildOption.
87 * @param errorCode Standard ICU error code. Its input value must
88 * pass the U_SUCCESS() test, or else the function returns
89 * immediately. Check for U_FAILURE() on output or use with
90 * function chaining. (See User Guide for details.)
91 * @return A new BytesTrie for the add()ed data.
92 * @stable ICU 4.8
93 */
94 BytesTrie *build(UStringTrieBuildOption buildOption, UErrorCode &errorCode);
95
96 /**
97 * Builds a BytesTrie for the add()ed data and byte-serializes it.
98 * Once built, no further data can be add()ed until clear() is called.
99 *
100 * A BytesTrie cannot be empty. At least one (byte sequence, value) pair
101 * must have been add()ed.
102 *
103 * Multiple calls to buildStringPiece() return StringPieces referring to the
104 * builder's same byte array, without rebuilding.
105 * If buildStringPiece() is called after build(), the trie will be
106 * re-serialized into a new array (because build() passes on ownership).
107 * If build() is called after buildStringPiece(), the trie object returned
108 * by build() will become the owner of the underlying string for the
109 * previously returned StringPiece.
110 * After clear() has been called, a new array will be used as well.
111 * @param buildOption Build option, see UStringTrieBuildOption.
112 * @param errorCode Standard ICU error code. Its input value must
113 * pass the U_SUCCESS() test, or else the function returns
114 * immediately. Check for U_FAILURE() on output or use with
115 * function chaining. (See User Guide for details.)
116 * @return A StringPiece which refers to the byte-serialized BytesTrie for the add()ed data.
117 * @stable ICU 4.8
118 */
119 StringPiece buildStringPiece(UStringTrieBuildOption buildOption, UErrorCode &errorCode);
120
121 /**
122 * Removes all (byte sequence, value) pairs.
123 * New data can then be add()ed and a new trie can be built.
124 * @return *this
125 * @stable ICU 4.8
126 */
127 BytesTrieBuilder &clear();
128
129private:
130 friend class ::BytesTrieTest;
131
132 BytesTrieBuilder(const BytesTrieBuilder &other) = delete; // no copy constructor
133 BytesTrieBuilder &operator=(const BytesTrieBuilder &other) = delete; // no assignment operator
134
135 void buildBytes(UStringTrieBuildOption buildOption, UErrorCode &errorCode);
136
137 virtual int32_t getElementStringLength(int32_t i) const override;
138 virtual char16_t getElementUnit(int32_t i, int32_t byteIndex) const override;
139 virtual int32_t getElementValue(int32_t i) const override;
140
141 virtual int32_t getLimitOfLinearMatch(int32_t first, int32_t last, int32_t byteIndex) const override;
142
143 virtual int32_t countElementUnits(int32_t start, int32_t limit, int32_t byteIndex) const override;
144 virtual int32_t skipElementsBySomeUnits(int32_t i, int32_t byteIndex, int32_t count) const override;
145 virtual int32_t indexOfElementWithNextUnit(int32_t i, int32_t byteIndex, char16_t byte) const override;
146
147 virtual UBool matchNodesCanHaveValues() const override { return false; }
148
149 virtual int32_t getMaxBranchLinearSubNodeLength() const override { return BytesTrie::kMaxBranchLinearSubNodeLength; }
150 virtual int32_t getMinLinearMatch() const override { return BytesTrie::kMinLinearMatch; }
151 virtual int32_t getMaxLinearMatchLength() const override { return BytesTrie::kMaxLinearMatchLength; }
152
153 /**
154 * @internal (private)
155 */
156 class BTLinearMatchNode : public LinearMatchNode {
157 public:
158 BTLinearMatchNode(const char *units, int32_t len, Node *nextNode);
159 virtual bool operator==(const Node &other) const override;
160 virtual void write(StringTrieBuilder &builder) override;
161 private:
162 const char *s;
163 };
164
165 virtual Node *createLinearMatchNode(int32_t i, int32_t byteIndex, int32_t length,
166 Node *nextNode) const override;
167
168 UBool ensureCapacity(int32_t length);
169 virtual int32_t write(int32_t byte) override;
170 int32_t write(const char *b, int32_t length);
171 virtual int32_t writeElementUnits(int32_t i, int32_t byteIndex, int32_t length) override;
172 virtual int32_t writeValueAndFinal(int32_t i, UBool isFinal) override;
173 virtual int32_t writeValueAndType(UBool hasValue, int32_t value, int32_t node) override;
174 virtual int32_t writeDeltaTo(int32_t jumpTarget) override;
175 static int32_t internalEncodeDelta(int32_t i, char intBytes[]);
176
177 CharString *strings; // Pointer not object so we need not #include internal charstr.h.
178 BytesTrieElement *elements;
179 int32_t elementsCapacity;
180 int32_t elementsLength;
181
182 // Byte serialization of the trie.
183 // Grows from the back: bytesLength measures from the end of the buffer!
184 char *bytes;
185 int32_t bytesCapacity;
186 int32_t bytesLength;
187};
188
189U_NAMESPACE_END
190
191#endif /* U_SHOW_CPLUSPLUS_API */
192
193#endif // __BYTESTRIEBUILDER_H__
194