|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace Onoi\Tesa; |
|
4
|
|
|
|
|
5
|
|
|
use Onoi\Tesa\StopwordAnalyzer\StopwordAnalyzer; |
|
6
|
|
|
use Onoi\Tesa\StopwordAnalyzer\NullStopwordAnalyzer; |
|
7
|
|
|
use Onoi\Tesa\StopwordAnalyzer\CdbStopwordAnalyzer; |
|
8
|
|
|
use Onoi\Tesa\StopwordAnalyzer\ArrayStopwordAnalyzer; |
|
9
|
|
|
use Onoi\Tesa\Synonymizer\Synonymizer; |
|
10
|
|
|
use Onoi\Tesa\Synonymizer\NullSynonymizer; |
|
11
|
|
|
use Onoi\Tesa\LanguageDetector\NullLanguageDetector; |
|
12
|
|
|
use Onoi\Tesa\LanguageDetector\TextCatLanguageDetector; |
|
13
|
|
|
use Onoi\Tesa\Tokenizer\CJKSimpleCharacterRegExTokenizer; |
|
14
|
|
|
use Onoi\Tesa\Tokenizer\Tokenizer; |
|
15
|
|
|
use Onoi\Tesa\Tokenizer\GenericRegExTokenizer; |
|
16
|
|
|
use Onoi\Tesa\Tokenizer\JaCompoundGroupTokenizer; |
|
17
|
|
|
use Onoi\Tesa\Tokenizer\IcuWordBoundaryTokenizer; |
|
18
|
|
|
use Onoi\Tesa\Tokenizer\NGramTokenizer; |
|
19
|
|
|
use Onoi\Tesa\Tokenizer\JaTinySegmenterTokenizer; |
|
20
|
|
|
use Onoi\Tesa\Tokenizer\PunctuationRegExTokenizer; |
|
21
|
|
|
|
|
22
|
|
|
/** |
|
23
|
|
|
* @license GNU GPL v2+ |
|
24
|
|
|
* @since 0.1 |
|
25
|
|
|
* |
|
26
|
|
|
* @author mwjames |
|
27
|
|
|
*/ |
|
28
|
|
|
class SanitizerFactory { |
|
29
|
|
|
|
|
30
|
|
|
/** |
|
31
|
|
|
* @since 0.1 |
|
32
|
|
|
* |
|
33
|
|
|
* @return Sanitizer |
|
34
|
|
|
*/ |
|
35
|
|
|
public function newSanitizer( $text = '' ) { |
|
36
|
|
|
return new Sanitizer( $text ); |
|
37
|
|
|
} |
|
38
|
|
|
|
|
39
|
|
|
/* StopwordAnalyzer */ |
|
40
|
|
|
|
|
41
|
|
|
/** |
|
42
|
|
|
* @since 0.1 |
|
43
|
|
|
* |
|
44
|
|
|
* @param string|null $languageCode |
|
45
|
|
|
* |
|
46
|
|
|
* @return StopwordAnalyzer |
|
47
|
|
|
*/ |
|
48
|
|
|
public function newStopwordAnalyzerByLanguage( $languageCode = null ) { |
|
49
|
|
|
|
|
50
|
|
|
if ( $languageCode === null ) { |
|
51
|
|
|
return $this->newNullStopwordAnalyzer(); |
|
52
|
|
|
} |
|
53
|
|
|
|
|
54
|
|
|
$cdbStopwordAnalyzer = $this->newCdbStopwordAnalyzer( |
|
55
|
|
|
$languageCode |
|
56
|
|
|
); |
|
57
|
|
|
|
|
58
|
|
|
return $cdbStopwordAnalyzer->isAvailable() ? $cdbStopwordAnalyzer : $this->newNullStopwordAnalyzer();; |
|
59
|
|
|
} |
|
60
|
|
|
|
|
61
|
|
|
/** |
|
62
|
|
|
* @since 0.1 |
|
63
|
|
|
* |
|
64
|
|
|
* @return StopwordAnalyzer |
|
65
|
|
|
*/ |
|
66
|
|
|
public function newCdbStopwordAnalyzer( $languageCode = null ) { |
|
67
|
|
|
return new CdbStopwordAnalyzer( CdbStopwordAnalyzer::getTargetByLanguage( $languageCode ) ); |
|
68
|
|
|
} |
|
69
|
|
|
|
|
70
|
|
|
/** |
|
71
|
|
|
* @since 0.1 |
|
72
|
|
|
* |
|
73
|
|
|
* @param array $stopwords; |
|
|
|
|
|
|
74
|
|
|
* |
|
75
|
|
|
* @return StopwordAnalyzer |
|
76
|
|
|
*/ |
|
77
|
|
|
public function newArrayStopwordAnalyzer( array $stopwords = array() ) { |
|
78
|
|
|
return new ArrayStopwordAnalyzer( $stopwords ); |
|
79
|
|
|
} |
|
80
|
|
|
|
|
81
|
|
|
/** |
|
82
|
|
|
* @since 0.1 |
|
83
|
|
|
* |
|
84
|
|
|
* @return StopwordAnalyzer |
|
85
|
|
|
*/ |
|
86
|
|
|
public function newNullStopwordAnalyzer() { |
|
87
|
|
|
return new NullStopwordAnalyzer(); |
|
88
|
|
|
} |
|
89
|
|
|
|
|
90
|
|
|
/** |
|
91
|
|
|
* @since 0.1 |
|
92
|
|
|
* |
|
93
|
|
|
* @param string|null $languageCode |
|
94
|
|
|
* |
|
95
|
|
|
* @return Synonymizer |
|
96
|
|
|
*/ |
|
97
|
|
|
public function newSynonymizerByLanguage( $languageCode = null ) { |
|
98
|
|
|
|
|
99
|
|
|
if ( $languageCode === null ) { |
|
100
|
|
|
return $this->newNullSynonymizer(); |
|
101
|
|
|
} |
|
102
|
|
|
|
|
103
|
|
|
return $this->newNullSynonymizer();; |
|
104
|
|
|
} |
|
105
|
|
|
|
|
106
|
|
|
/* Synonymizer */ |
|
107
|
|
|
|
|
108
|
|
|
/** |
|
109
|
|
|
* @since 0.1 |
|
110
|
|
|
* |
|
111
|
|
|
* @return Synonymizer |
|
112
|
|
|
*/ |
|
113
|
|
|
public function newNullSynonymizer() { |
|
114
|
|
|
return new NullSynonymizer(); |
|
115
|
|
|
} |
|
116
|
|
|
|
|
117
|
|
|
/* LanguageDetector */ |
|
118
|
|
|
|
|
119
|
|
|
/** |
|
120
|
|
|
* @since 0.1 |
|
121
|
|
|
* |
|
122
|
|
|
* @return NullLanguageDetector |
|
123
|
|
|
*/ |
|
124
|
|
|
public function newNullLanguageDetector() { |
|
125
|
|
|
return new NullLanguageDetector(); |
|
126
|
|
|
} |
|
127
|
|
|
|
|
128
|
|
|
/** |
|
129
|
|
|
* @since 0.1 |
|
130
|
|
|
* |
|
131
|
|
|
* @return TextCatLanguageDetector |
|
132
|
|
|
*/ |
|
133
|
|
|
public function newTextCatLanguageDetector() { |
|
134
|
|
|
return new TextCatLanguageDetector(); |
|
135
|
|
|
} |
|
136
|
|
|
|
|
137
|
|
|
/* Tokenizer */ |
|
138
|
|
|
|
|
139
|
|
|
/** |
|
140
|
|
|
* @since 0.1 |
|
141
|
|
|
* |
|
142
|
|
|
* @param string $text |
|
143
|
|
|
* @param string|null $languageCode |
|
144
|
|
|
* |
|
145
|
|
|
* @return Tokenizer |
|
146
|
|
|
*/ |
|
147
|
|
|
public function newPreferredTokenizerByLanguage( $text, $languageCode = null ) { |
|
148
|
|
|
|
|
149
|
|
|
$tokenizer = $this->newIcuWordBoundaryTokenizer(); |
|
150
|
|
|
|
|
151
|
|
|
if ( !$tokenizer->isAvailable() && CharacterExaminer::contains( CharacterExaminer::CJK_UNIFIED, $text ) ) { |
|
152
|
|
|
return $this->newCJKMatchableTokenizer( $text ); |
|
153
|
|
|
} elseif( !$tokenizer->isAvailable() ) { |
|
154
|
|
|
return $this->newGenericRegExTokenizer( $tokenizer ); |
|
155
|
|
|
} |
|
156
|
|
|
|
|
157
|
|
|
$tokenizer->setLocale( $languageCode ); |
|
158
|
|
|
|
|
159
|
|
|
$tokenizer->setWordTokenizerAttribute( |
|
|
|
|
|
|
160
|
|
|
!CharacterExaminer::contains( CharacterExaminer::CJK_UNIFIED, $text ) |
|
161
|
|
|
); |
|
162
|
|
|
|
|
163
|
|
|
return $this->newGenericRegExTokenizer( $tokenizer ); |
|
164
|
|
|
} |
|
165
|
|
|
|
|
166
|
|
|
/** |
|
167
|
|
|
* @since 0.1 |
|
168
|
|
|
* |
|
169
|
|
|
* @param string $text |
|
170
|
|
|
* |
|
171
|
|
|
* @return Tokenizer |
|
172
|
|
|
*/ |
|
173
|
|
|
public function newCJKMatchableTokenizer( $text ) { |
|
174
|
|
|
|
|
175
|
|
|
$tokenizer = null; |
|
176
|
|
|
|
|
177
|
|
|
if ( CharacterExaminer::contains( CharacterExaminer::HIRAGANA_KATAKANA, $text ) ) { |
|
178
|
|
|
$tokenizer = $this->newJaTinySegmenterTokenizer(); |
|
179
|
|
|
} else { |
|
180
|
|
|
$tokenizer = $this->newNGramTokenizer( $tokenizer ); |
|
181
|
|
|
} |
|
182
|
|
|
|
|
183
|
|
|
$tokenizer = $this->newCJKSimpleCharacterRegExTokenizer( $tokenizer ); |
|
184
|
|
|
|
|
185
|
|
|
return $this->newGenericRegExTokenizer( $tokenizer ); |
|
186
|
|
|
} |
|
187
|
|
|
|
|
188
|
|
|
/** |
|
189
|
|
|
* @since 0.1 |
|
190
|
|
|
* |
|
191
|
|
|
* @param Tokenizer|null $tokenizer |
|
192
|
|
|
* |
|
193
|
|
|
* @return Tokenizer |
|
194
|
|
|
*/ |
|
195
|
|
|
public function newIcuWordBoundaryTokenizer( Tokenizer $tokenizer = null ) { |
|
196
|
|
|
return new IcuWordBoundaryTokenizer( $tokenizer ); |
|
197
|
|
|
} |
|
198
|
|
|
|
|
199
|
|
|
/** |
|
200
|
|
|
* @since 0.1 |
|
201
|
|
|
* |
|
202
|
|
|
* @param Tokenizer|null $tokenizer |
|
203
|
|
|
* |
|
204
|
|
|
* @return Tokenizer |
|
205
|
|
|
*/ |
|
206
|
|
|
public function newGenericRegExTokenizer( Tokenizer $tokenizer = null ) { |
|
207
|
|
|
return new GenericRegExTokenizer( $tokenizer ); |
|
208
|
|
|
} |
|
209
|
|
|
|
|
210
|
|
|
/** |
|
211
|
|
|
* @since 0.1 |
|
212
|
|
|
* |
|
213
|
|
|
* @param Tokenizer|null $tokenizer |
|
214
|
|
|
* |
|
215
|
|
|
* @return Tokenizer |
|
216
|
|
|
*/ |
|
217
|
|
|
public function newPunctuationRegExTokenizer( Tokenizer $tokenizer = null ) { |
|
218
|
|
|
return new PunctuationRegExTokenizer( $tokenizer ); |
|
219
|
|
|
} |
|
220
|
|
|
|
|
221
|
|
|
/** |
|
222
|
|
|
* @since 0.1 |
|
223
|
|
|
* |
|
224
|
|
|
* @return Tokenizer |
|
225
|
|
|
*/ |
|
226
|
|
|
public function newJaCompoundGroupTokenizer( Tokenizer $tokinizer = null ) { |
|
227
|
|
|
return new JaCompoundGroupTokenizer( $tokinizer ); |
|
228
|
|
|
} |
|
229
|
|
|
|
|
230
|
|
|
/** |
|
231
|
|
|
* @since 0.1 |
|
232
|
|
|
* |
|
233
|
|
|
* @return Tokenizer |
|
234
|
|
|
*/ |
|
235
|
|
|
public function newJaTinySegmenterTokenizer( Tokenizer $tokinizer = null ) { |
|
236
|
|
|
return new JaTinySegmenterTokenizer( $tokinizer ); |
|
237
|
|
|
} |
|
238
|
|
|
|
|
239
|
|
|
/** |
|
240
|
|
|
* @since 0.1 |
|
241
|
|
|
* |
|
242
|
|
|
* @return Tokenizer |
|
243
|
|
|
*/ |
|
244
|
|
|
public function newCJKSimpleCharacterRegExTokenizer( Tokenizer $tokinizer = null ) { |
|
245
|
|
|
return new CJKSimpleCharacterRegExTokenizer( $tokinizer ); |
|
246
|
|
|
} |
|
247
|
|
|
|
|
248
|
|
|
/** |
|
249
|
|
|
* @since 0.1 |
|
250
|
|
|
* |
|
251
|
|
|
* @return Tokenizer |
|
252
|
|
|
*/ |
|
253
|
|
|
public function newNGramTokenizer( Tokenizer $tokinizer = null, $ngram = 2 ) { |
|
254
|
|
|
return new NGramTokenizer( $tokinizer, $ngram ); |
|
255
|
|
|
} |
|
256
|
|
|
|
|
257
|
|
|
} |
|
258
|
|
|
|
This check looks for PHPDoc comments describing methods or function parameters that do not exist on the corresponding method or function. It has, however, found a similar but not annotated parameter which might be a good fit.
Consider the following example. The parameter
$irelandis not defined by the methodfinale(...).The most likely cause is that the parameter was changed, but the annotation was not.