1 /*
2 *******************************************************************************
3 *   Copyright (C) 2010-2014, 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 /**
16  * \file
17  * \brief C++ API: Builder for icu::BytesTrie
18  */
19 
20 #ifndef __BYTESTRIEBUILDER_H__
21 #define __BYTESTRIEBUILDER_H__
22 
23 #include "unicode/utypes.h"
24 #include "unicode/bytestrie.h"
25 #include "unicode/stringpiece.h"
26 #include "unicode/stringtriebuilder.h"
27 
28 U_NAMESPACE_BEGIN
29 
30 class BytesTrieElement;
31 class CharString;
32 
33 /**
34  * Builder class for BytesTrie.
35  *
36  * This class is not intended for public subclassing.
37  * @stable ICU 4.8
38  */
39 class U_COMMON_API BytesTrieBuilder : public StringTrieBuilder {
40 public:
41     /**
42      * Constructs an empty builder.
43      * @param errorCode Standard ICU error code.
44      * @stable ICU 4.8
45      */
46     BytesTrieBuilder(UErrorCode &errorCode);
47 
48     /**
49      * Destructor.
50      * @stable ICU 4.8
51      */
52     virtual ~BytesTrieBuilder();
53 
54     /**
55      * Adds a (byte sequence, value) pair.
56      * The byte sequence must be unique.
57      * The bytes will be copied; the builder does not keep
58      * a reference to the input StringPiece or its data().
59      * @param s The input byte sequence.
60      * @param value The value associated with this byte sequence.
61      * @param errorCode Standard ICU error code. Its input value must
62      *                  pass the U_SUCCESS() test, or else the function returns
63      *                  immediately. Check for U_FAILURE() on output or use with
64      *                  function chaining. (See User Guide for details.)
65      * @return *this
66      * @stable ICU 4.8
67      */
68     BytesTrieBuilder &add(const StringPiece &s, int32_t value, UErrorCode &errorCode);
69 
70     /**
71      * Builds a BytesTrie for the add()ed data.
72      * Once built, no further data can be add()ed until clear() is called.
73      *
74      * A BytesTrie cannot be empty. At least one (byte sequence, value) pair
75      * must have been add()ed.
76      *
77      * This method passes ownership of the builder's internal result array to the new trie object.
78      * Another call to any build() variant will re-serialize the trie.
79      * After clear() has been called, a new array will be used as well.
80      * @param buildOption Build option, see UStringTrieBuildOption.
81      * @param errorCode Standard ICU error code. Its input value must
82      *                  pass the U_SUCCESS() test, or else the function returns
83      *                  immediately. Check for U_FAILURE() on output or use with
84      *                  function chaining. (See User Guide for details.)
85      * @return A new BytesTrie for the add()ed data.
86      * @stable ICU 4.8
87      */
88     BytesTrie *build(UStringTrieBuildOption buildOption, UErrorCode &errorCode);
89 
90     /**
91      * Builds a BytesTrie for the add()ed data and byte-serializes it.
92      * Once built, no further data can be add()ed until clear() is called.
93      *
94      * A BytesTrie cannot be empty. At least one (byte sequence, value) pair
95      * must have been add()ed.
96      *
97      * Multiple calls to buildStringPiece() return StringPieces referring to the
98      * builder's same byte array, without rebuilding.
99      * If buildStringPiece() is called after build(), the trie will be
100      * re-serialized into a new array.
101      * If build() is called after buildStringPiece(), the trie object will become
102      * the owner of the previously returned array.
103      * After clear() has been called, a new array will be used as well.
104      * @param buildOption Build option, see UStringTrieBuildOption.
105      * @param errorCode Standard ICU error code. Its input value must
106      *                  pass the U_SUCCESS() test, or else the function returns
107      *                  immediately. Check for U_FAILURE() on output or use with
108      *                  function chaining. (See User Guide for details.)
109      * @return A StringPiece which refers to the byte-serialized BytesTrie for the add()ed data.
110      * @stable ICU 4.8
111      */
112     StringPiece buildStringPiece(UStringTrieBuildOption buildOption, UErrorCode &errorCode);
113 
114     /**
115      * Removes all (byte sequence, value) pairs.
116      * New data can then be add()ed and a new trie can be built.
117      * @return *this
118      * @stable ICU 4.8
119      */
120     BytesTrieBuilder &clear();
121 
122 private:
123     BytesTrieBuilder(const BytesTrieBuilder &other);  // no copy constructor
124     BytesTrieBuilder &operator=(const BytesTrieBuilder &other);  // no assignment operator
125 
126     void buildBytes(UStringTrieBuildOption buildOption, UErrorCode &errorCode);
127 
128     virtual int32_t getElementStringLength(int32_t i) const;
129     virtual UChar getElementUnit(int32_t i, int32_t byteIndex) const;
130     virtual int32_t getElementValue(int32_t i) const;
131 
132     virtual int32_t getLimitOfLinearMatch(int32_t first, int32_t last, int32_t byteIndex) const;
133 
134     virtual int32_t countElementUnits(int32_t start, int32_t limit, int32_t byteIndex) const;
135     virtual int32_t skipElementsBySomeUnits(int32_t i, int32_t byteIndex, int32_t count) const;
136     virtual int32_t indexOfElementWithNextUnit(int32_t i, int32_t byteIndex, UChar byte) const;
137 
matchNodesCanHaveValues()138     virtual UBool matchNodesCanHaveValues() const { return FALSE; }
139 
getMaxBranchLinearSubNodeLength()140     virtual int32_t getMaxBranchLinearSubNodeLength() const { return BytesTrie::kMaxBranchLinearSubNodeLength; }
getMinLinearMatch()141     virtual int32_t getMinLinearMatch() const { return BytesTrie::kMinLinearMatch; }
getMaxLinearMatchLength()142     virtual int32_t getMaxLinearMatchLength() const { return BytesTrie::kMaxLinearMatchLength; }
143 
144     /**
145      * @internal
146      */
147     class BTLinearMatchNode : public LinearMatchNode {
148     public:
149         BTLinearMatchNode(const char *units, int32_t len, Node *nextNode);
150         virtual UBool operator==(const Node &other) const;
151         virtual void write(StringTrieBuilder &builder);
152     private:
153         const char *s;
154     };
155 
156     virtual Node *createLinearMatchNode(int32_t i, int32_t byteIndex, int32_t length,
157                                         Node *nextNode) const;
158 
159     UBool ensureCapacity(int32_t length);
160     virtual int32_t write(int32_t byte);
161     int32_t write(const char *b, int32_t length);
162     virtual int32_t writeElementUnits(int32_t i, int32_t byteIndex, int32_t length);
163     virtual int32_t writeValueAndFinal(int32_t i, UBool isFinal);
164     virtual int32_t writeValueAndType(UBool hasValue, int32_t value, int32_t node);
165     virtual int32_t writeDeltaTo(int32_t jumpTarget);
166 
167     CharString *strings;  // Pointer not object so we need not #include internal charstr.h.
168     BytesTrieElement *elements;
169     int32_t elementsCapacity;
170     int32_t elementsLength;
171 
172     // Byte serialization of the trie.
173     // Grows from the back: bytesLength measures from the end of the buffer!
174     char *bytes;
175     int32_t bytesCapacity;
176     int32_t bytesLength;
177 };
178 
179 U_NAMESPACE_END
180 
181 #endif  // __BYTESTRIEBUILDER_H__
182