author | hgs |
Fri, 15 Oct 2010 12:09:28 +0530 | |
changeset 24 | 65456528cac2 |
parent 0 | 671dee74050a |
permissions | -rw-r--r-- |
0
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
1 |
/*------------------------------------------------------------------------------ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
2 |
* Copyright (C) 2003-2006 Ben van Klinken and the CLucene Team |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
3 |
* |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
4 |
* Distributable under the terms of either the Apache License (Version 2.0) or |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
5 |
* the GNU Lesser General Public License, as specified in the COPYING file. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
6 |
------------------------------------------------------------------------------*/ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
7 |
#ifndef _lucene_analysis_Analyzers_ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
8 |
#define _lucene_analysis_Analyzers_ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
9 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
10 |
#if defined(_LUCENE_PRAGMA_ONCE) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
11 |
# pragma once |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
12 |
#endif |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
13 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
14 |
#include "CLucene/util/Reader.h" |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
15 |
#include "CLucene/analysis/AnalysisHeader.h" |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
16 |
#include "clucene/util/misc.h" |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
17 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
18 |
CL_NS_DEF(analysis) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
19 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
20 |
/** An abstract base class for simple, character-oriented tokenizers.*/ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
21 |
class CharTokenizer:public Tokenizer { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
22 |
private: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
23 |
int32_t offset, bufferIndex, dataLen; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
24 |
TCHAR buffer[LUCENE_MAX_WORD_LEN+1]; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
25 |
const TCHAR* ioBuffer; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
26 |
protected: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
27 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
28 |
/** Returns true iff a character should be included in a token. This |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
29 |
* tokenizer generates as tokens adjacent sequences of characters which |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
30 |
* satisfy this predicate. Characters for which this is false are used to |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
31 |
* define token boundaries and are not included in tokens. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
32 |
virtual bool isTokenChar(const TCHAR c) const = 0; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
33 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
34 |
/** Called on each token character to normalize it before it is added to the |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
35 |
* token. The default implementation does nothing. Subclasses may use this |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
36 |
* to, e.g., lowercase tokens. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
37 |
virtual TCHAR normalize(const TCHAR c) const; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
38 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
39 |
public: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
40 |
CharTokenizer(CL_NS(util)::Reader* in); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
41 |
virtual ~CharTokenizer(){ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
42 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
43 |
bool next(Token* token); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
44 |
}; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
45 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
46 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
47 |
/** A LetterTokenizer is a tokenizer that divides text at non-letters. That's |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
48 |
to say, it defines tokens as maximal strings of adjacent letters, as defined |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
49 |
by java.lang.Character.isLetter() predicate. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
50 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
51 |
Note: this does a decent job for most European languages, but does a terrible |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
52 |
job for some Asian languages, where words are not separated by spaces. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
53 |
class LetterTokenizer:public CharTokenizer { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
54 |
public: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
55 |
// Construct a new LetterTokenizer. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
56 |
LetterTokenizer(CL_NS(util)::Reader* in): |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
57 |
CharTokenizer(in) {} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
58 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
59 |
~LetterTokenizer(){} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
60 |
protected: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
61 |
/** Collects only characters which satisfy _istalpha.*/ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
62 |
bool isTokenChar(const TCHAR c) const; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
63 |
}; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
64 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
65 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
66 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
67 |
/** |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
68 |
* LowerCaseTokenizer performs the function of LetterTokenizer |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
69 |
* and LowerCaseFilter together. It divides text at non-letters and converts |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
70 |
* them to lower case. While it is functionally equivalent to the combination |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
71 |
* of LetterTokenizer and LowerCaseFilter, there is a performance advantage |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
72 |
* to doing the two tasks at once, hence this (redundant) implementation. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
73 |
* <P> |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
74 |
* Note: this does a decent job for most European languages, but does a terrible |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
75 |
* job for some Asian languages, where words are not separated by spaces. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
76 |
*/ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
77 |
class LowerCaseTokenizer:public LetterTokenizer { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
78 |
public: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
79 |
/** Construct a new LowerCaseTokenizer. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
80 |
LowerCaseTokenizer(CL_NS(util)::Reader* in): |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
81 |
LetterTokenizer(in) {} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
82 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
83 |
~LowerCaseTokenizer(){} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
84 |
protected: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
85 |
/** Collects only characters which satisfy _totlower. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
86 |
TCHAR normalize(const TCHAR chr) const; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
87 |
}; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
88 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
89 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
90 |
/** A WhitespaceTokenizer is a tokenizer that divides text at whitespace. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
91 |
* Adjacent sequences of non-Whitespace characters form tokens. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
92 |
class WhitespaceTokenizer: public CharTokenizer { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
93 |
public: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
94 |
/** Construct a new WhitespaceTokenizer. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
95 |
WhitespaceTokenizer(CL_NS(util)::Reader* in):CharTokenizer(in) {} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
96 |
~WhitespaceTokenizer(){} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
97 |
protected: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
98 |
/** Collects only characters which do not satisfy _istspace. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
99 |
*/ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
100 |
bool isTokenChar(const TCHAR c) const; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
101 |
}; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
102 |
|
24 | 103 |
/** A PhoneNumberTokenizer is a tokenizer that strips a number to its subset. |
104 |
* ex: A number 567 is tokenized as 567, 67 and 7. This is introduced to make |
|
105 |
* number/word searchable from middle*/ |
|
106 |
class PhoneNumberTokenizer: public Tokenizer { |
|
107 |
private: |
|
108 |
const wchar_t *termText; |
|
109 |
int32_t tokenLen; |
|
110 |
int32_t termLen; |
|
111 |
public: |
|
112 |
/** Construct a new PhoneNumberTokenizer. */ |
|
113 |
PhoneNumberTokenizer(CL_NS(util)::Reader* input); |
|
114 |
~PhoneNumberTokenizer(){} |
|
115 |
bool next(Token* token); |
|
116 |
}; |
|
0
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
117 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
118 |
/** An Analyzer that uses WhitespaceTokenizer. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
119 |
class WhitespaceAnalyzer: public Analyzer { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
120 |
public: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
121 |
TokenStream* tokenStream(const TCHAR* fieldName, CL_NS(util)::Reader* reader); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
122 |
~WhitespaceAnalyzer(){} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
123 |
}; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
124 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
125 |
/** An Analyzer that filters LetterTokenizer with LowerCaseFilter. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
126 |
class SimpleAnalyzer: public Analyzer { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
127 |
public: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
128 |
TokenStream* tokenStream(const TCHAR* fieldName, CL_NS(util)::Reader* reader); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
129 |
~SimpleAnalyzer(){} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
130 |
}; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
131 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
132 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
133 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
134 |
/** |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
135 |
* Normalizes token text to lower case. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
136 |
*/ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
137 |
class LowerCaseFilter: public TokenFilter { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
138 |
public: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
139 |
LowerCaseFilter(TokenStream* in, bool deleteTokenStream):TokenFilter(in,deleteTokenStream) {} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
140 |
~LowerCaseFilter(){} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
141 |
bool next(Token* token); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
142 |
}; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
143 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
144 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
145 |
/** |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
146 |
* Removes stop words from a token stream. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
147 |
*/ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
148 |
class StopFilter: public TokenFilter { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
149 |
private: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
150 |
//bvk: i found this to work faster with a non-hash table. the number of items |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
151 |
//in the stop table is not like to make it worth having hashing. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
152 |
CL_NS(util)::CLSetList<const TCHAR*>* table; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
153 |
bool ownTable; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
154 |
public: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
155 |
// Constructs a filter which removes words from the input |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
156 |
// TokenStream that are named in the array of words. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
157 |
StopFilter(TokenStream* in, bool deleteTokenStream, const TCHAR** stopWords); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
158 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
159 |
~StopFilter(); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
160 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
161 |
/** Constructs a filter which removes words from the input |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
162 |
* TokenStream that are named in the CLSetList. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
163 |
*/ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
164 |
StopFilter(TokenStream* in, bool deleteTokenStream, CL_NS(util)::CLSetList<const TCHAR*>* stopTable): |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
165 |
TokenFilter(in, deleteTokenStream), |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
166 |
table(stopTable), |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
167 |
ownTable(false) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
168 |
{} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
169 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
170 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
171 |
/** |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
172 |
* Builds a Hashtable from an array of stop words, appropriate for passing |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
173 |
* into the StopFilter constructor. This permits this table construction to |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
174 |
* be cached once when an Analyzer is constructed. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
175 |
* Note: the stopWords list must be a static list because the strings are not copied |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
176 |
*/ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
177 |
static void fillStopTable(CL_NS(util)::CLSetList<const TCHAR*>* stopTable, |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
178 |
const TCHAR** stopWords); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
179 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
180 |
/** |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
181 |
* Returns the next input Token whose termText() is not a stop word. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
182 |
*/ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
183 |
bool next(Token* token); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
184 |
}; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
185 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
186 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
187 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
188 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
189 |
/** Filters LetterTokenizer with LowerCaseFilter and StopFilter. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
190 |
class StopAnalyzer: public Analyzer { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
191 |
CL_NS(util)::CLSetList<const TCHAR*> stopTable; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
192 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
193 |
public: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
194 |
/** Builds an analyzer which removes words in ENGLISH_STOP_WORDS. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
195 |
StopAnalyzer(); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
196 |
~StopAnalyzer(); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
197 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
198 |
/** Builds an analyzer which removes words in the provided array. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
199 |
StopAnalyzer( const TCHAR** stopWords ); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
200 |
/** Filters LowerCaseTokenizer with StopFilter. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
201 |
TokenStream* tokenStream(const TCHAR* fieldName, CL_NS(util)::Reader* reader); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
202 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
203 |
/** An array containing some common English words that are not usually useful |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
204 |
for searching. */ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
205 |
static const TCHAR* ENGLISH_STOP_WORDS[]; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
206 |
}; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
207 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
208 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
209 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
210 |
/** |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
211 |
* This analyzer is used to facilitate scenarios where different |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
212 |
* fields require different analysis techniques. Use {@link #addAnalyzer} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
213 |
* to add a non-default analyzer on a field name basis. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
214 |
* |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
215 |
* <p>Example usage: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
216 |
* |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
217 |
* <pre> |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
218 |
* PerFieldAnalyzerWrapper aWrapper = |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
219 |
* new PerFieldAnalyzerWrapper(new StandardAnalyzer()); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
220 |
* aWrapper.addAnalyzer("firstname", new KeywordAnalyzer()); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
221 |
* aWrapper.addAnalyzer("lastname", new KeywordAnalyzer()); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
222 |
* </pre> |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
223 |
* |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
224 |
* <p>In this example, StandardAnalyzer will be used for all fields except "firstname" |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
225 |
* and "lastname", for which KeywordAnalyzer will be used. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
226 |
* |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
227 |
* <p>A PerFieldAnalyzerWrapper can be used like any other analyzer, for both indexing |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
228 |
* and query parsing. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
229 |
*/ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
230 |
class PerFieldAnalyzerWrapper : public Analyzer { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
231 |
private: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
232 |
Analyzer* defaultAnalyzer; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
233 |
CL_NS(util)::LHashMap<const TCHAR*, Analyzer*, CL_NS(util)::Compare::TChar, |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
234 |
CL_NS(util)::Equals::TChar, CL_NS(util)::Deletor::tcArray,CL_NS(util)::Deletor::Void<Analyzer> > analyzerMap; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
235 |
public: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
236 |
/** |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
237 |
* Constructs with default analyzer. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
238 |
* |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
239 |
* @param defaultAnalyzer Any fields not specifically |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
240 |
* defined to use a different analyzer will use the one provided here. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
241 |
*/ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
242 |
PerFieldAnalyzerWrapper(Analyzer* defaultAnalyzer); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
243 |
~PerFieldAnalyzerWrapper(); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
244 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
245 |
/** |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
246 |
* Defines an analyzer to use for the specified field. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
247 |
* |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
248 |
* @param fieldName field name requiring a non-default analyzer |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
249 |
* @param analyzer non-default analyzer to use for field |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
250 |
*/ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
251 |
void addAnalyzer(const TCHAR* fieldName, Analyzer* analyzer); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
252 |
TokenStream* tokenStream(const TCHAR* fieldName, CL_NS(util)::Reader* reader); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
253 |
}; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
254 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
255 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
256 |
/** |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
257 |
* A filter that replaces accented characters in the ISO Latin 1 character set |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
258 |
* (ISO-8859-1) by their unaccented equivalent. The case will not be altered. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
259 |
* <p> |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
260 |
* For instance, 'à' will be replaced by 'a'. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
261 |
* <p> |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
262 |
*/ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
263 |
class ISOLatin1AccentFilter: public TokenFilter { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
264 |
public: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
265 |
ISOLatin1AccentFilter(TokenStream* input, bool deleteTs): |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
266 |
TokenFilter(input,deleteTs) |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
267 |
{ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
268 |
} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
269 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
270 |
/** |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
271 |
* To replace accented characters in a |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
272 |
* String by unaccented equivalents. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
273 |
*/ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
274 |
bool next(Token* token); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
275 |
}; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
276 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
277 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
278 |
/** |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
279 |
* Emits the entire input as a single token. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
280 |
*/ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
281 |
class KeywordTokenizer: public Tokenizer { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
282 |
private: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
283 |
LUCENE_STATIC_CONSTANT(int, DEFAULT_BUFFER_SIZE = 256); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
284 |
bool done; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
285 |
int bufferSize; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
286 |
public: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
287 |
KeywordTokenizer(CL_NS(util)::Reader* input, int bufferSize=-1); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
288 |
virtual ~KeywordTokenizer(); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
289 |
bool next(Token* token); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
290 |
}; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
291 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
292 |
/** |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
293 |
* "Tokenizes" the entire stream as a single token. This is useful |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
294 |
* for data like zip codes, ids, and some product names. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
295 |
*/ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
296 |
class KeywordAnalyzer: public Analyzer { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
297 |
public: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
298 |
TokenStream* tokenStream(const TCHAR* fieldName, CL_NS(util)::Reader* reader); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
299 |
virtual ~KeywordAnalyzer(){} |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
300 |
}; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
301 |
|
24 | 302 |
|
303 |
/** An Analyzer that uses PhoneNumberTokenizer. */ |
|
304 |
class PhoneNumberAnalyzer: public Analyzer { |
|
305 |
public: |
|
306 |
TokenStream* tokenStream(const TCHAR* fieldName, CL_NS(util)::Reader* reader); |
|
307 |
virtual ~PhoneNumberAnalyzer(){} |
|
308 |
}; |
|
0
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
309 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
310 |
/** |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
311 |
* Removes words that are too long and too short from the stream. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
312 |
* |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
313 |
*/ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
314 |
class LengthFilter: public TokenFilter { |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
315 |
private: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
316 |
int _min; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
317 |
int _max; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
318 |
public: |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
319 |
/** |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
320 |
* Build a filter that removes words that are too long or too |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
321 |
* short from the text. |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
322 |
*/ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
323 |
LengthFilter(TokenStream* in, int _min, int _max); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
324 |
LengthFilter(TokenStream* in, bool deleteTs, int _min, int _max); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
325 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
326 |
/** |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
327 |
* Returns the next input Token whose termText() is the right len |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
328 |
*/ |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
329 |
bool next(Token* token); |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
330 |
}; |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
331 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
332 |
|
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
333 |
CL_NS_END |
671dee74050a
Revision: 201011
Dremov Kirill (Nokia-D-MSW/Tampere) <kirill.dremov@nokia.com>
parents:
diff
changeset
|
334 |
#endif |