]> git.openstreetmap.org Git - nominatim.git/blob - lib-php/tokenizer/legacy_icu_tokenizer.php
ca224a224861f5bd2e0f126cf10be5ea0edf5488
[nominatim.git] / lib-php / tokenizer / legacy_icu_tokenizer.php
1 <?php
2
3 namespace Nominatim;
4
5 require_once(CONST_LibDir.'/SimpleWordList.php');
6
7 class Tokenizer
8 {
9     private $oDB;
10
11     private $oNormalizer;
12     private $oTransliterator;
13
14     public function __construct(&$oDB)
15     {
16         $this->oDB =& $oDB;
17         $this->oNormalizer = \Transliterator::createFromRules(CONST_Term_Normalization_Rules);
18         $this->oTransliterator = \Transliterator::createFromRules(CONST_Transliteration);
19     }
20
21     public function checkStatus()
22     {
23         $sSQL = 'SELECT word_id FROM word WHERE word_id is not null limit 1';
24         $iWordID = $this->oDB->getOne($sSQL);
25         if ($iWordID === false) {
26             throw new \Exception('Query failed', 703);
27         }
28         if (!$iWordID) {
29             throw new \Exception('No value', 704);
30         }
31     }
32
33
34     public function normalizeString($sTerm)
35     {
36         if ($this->oNormalizer === null) {
37             return $sTerm;
38         }
39
40         return $this->oNormalizer->transliterate($sTerm);
41     }
42
43     private function makeStandardWord($sTerm)
44     {
45         return trim($this->oTransliterator->transliterate(' '.$sTerm.' '));
46     }
47
48
49     public function tokensForSpecialTerm($sTerm)
50     {
51         $aResults = array();
52
53         $sSQL = "SELECT word_id, info->>'class' as class, info->>'type' as type ";
54         $sSQL .= '   FROM word WHERE word_token = :term and type = \'S\'';
55
56         Debug::printVar('Term', $sTerm);
57         Debug::printSQL($sSQL);
58         $aSearchWords = $this->oDB->getAll($sSQL, array(':term' => $this->makeStandardWord($sTerm)));
59
60         Debug::printVar('Results', $aSearchWords);
61
62         foreach ($aSearchWords as $aSearchTerm) {
63             $aResults[] = new \Nominatim\Token\SpecialTerm(
64                 $aSearchTerm['word_id'],
65                 $aSearchTerm['class'],
66                 $aSearchTerm['type'],
67                 \Nominatim\Operator::TYPE
68             );
69         }
70
71         Debug::printVar('Special term tokens', $aResults);
72
73         return $aResults;
74     }
75
76
77     public function extractTokensFromPhrases(&$aPhrases)
78     {
79         $sNormQuery = '';
80         $aWordLists = array();
81         $aTokens = array();
82         foreach ($aPhrases as $iPhrase => $oPhrase) {
83             $sNormQuery .= ','.$this->normalizeString($oPhrase->getPhrase());
84             $sPhrase = $this->makeStandardWord($oPhrase->getPhrase());
85             Debug::printVar('Phrase', $sPhrase);
86
87             $oWordList = new SimpleWordList($sPhrase);
88             $aTokens = array_merge($aTokens, $oWordList->getTokens());
89             $aWordLists[] = $oWordList;
90         }
91
92         Debug::printVar('Tokens', $aTokens);
93         Debug::printVar('WordLists', $aWordLists);
94
95         $oValidTokens = $this->computeValidTokens($aTokens, $sNormQuery);
96
97         foreach ($aPhrases as $iPhrase => $oPhrase) {
98             $oPhrase->setWordSets($aWordLists[$iPhrase]->getWordSets($oValidTokens));
99         }
100
101         return $oValidTokens;
102     }
103
104
105     private function computeValidTokens($aTokens, $sNormQuery)
106     {
107         $oValidTokens = new TokenList();
108
109         if (!empty($aTokens)) {
110             $this->addTokensFromDB($oValidTokens, $aTokens, $sNormQuery);
111
112             // Try more interpretations for Tokens that could not be matched.
113             foreach ($aTokens as $sToken) {
114                 if ($sToken[0] != ' ' && !$oValidTokens->contains($sToken)) {
115                     if (preg_match('/^([0-9]{5}) [0-9]{4}$/', $sToken, $aData)) {
116                         // US ZIP+4 codes - merge in the 5-digit ZIP code
117                         $oValidTokens->addToken(
118                             $sToken,
119                             new Token\Postcode(null, $aData[1], 'us')
120                         );
121                     } elseif (preg_match('/^[0-9]+$/', $sToken)) {
122                         // Unknown single word token with a number.
123                         // Assume it is a house number.
124                         $oValidTokens->addToken(
125                             $sToken,
126                             new Token\HouseNumber(null, trim($sToken))
127                         );
128                     }
129                 }
130             }
131         }
132
133         return $oValidTokens;
134     }
135
136
137     private function addTokensFromDB(&$oValidTokens, $aTokens, $sNormQuery)
138     {
139         // Check which tokens we have, get the ID numbers
140         $sSQL = 'SELECT word_id, word_token, type, word,';
141         $sSQL .= "      info->>'op' as operator,";
142         $sSQL .= "      info->>'class' as class, info->>'type' as ctype,";
143         $sSQL .= "      info->>'count' as count";
144         $sSQL .= ' FROM word WHERE word_token in (';
145         $sSQL .= join(',', $this->oDB->getDBQuotedList($aTokens)).')';
146
147         Debug::printSQL($sSQL);
148
149         $aDBWords = $this->oDB->getAll($sSQL, null, 'Could not get word tokens.');
150
151         foreach ($aDBWords as $aWord) {
152             $iId = (int) $aWord['word_id'];
153             $sTok = $aWord['word_token'];
154
155             switch ($aWord['type']) {
156                 case 'C':  // country name tokens
157                     if ($aWord['word'] !== null) {
158                         $oValidTokens->addToken(
159                             $sTok,
160                             new Token\Country($iId, $aWord['word'])
161                         );
162                     }
163                     break;
164                 case 'H':  // house number tokens
165                     $oValidTokens->addToken($sTok, new Token\HouseNumber($iId, $aWord['word_token']));
166                     break;
167                 case 'P':  // postcode tokens
168                     // Postcodes are not normalized, so they may have content
169                     // that makes SQL injection possible. Reject postcodes
170                     // that would need special escaping.
171                     if ($aWord['word'] !== null
172                         && pg_escape_string($aWord['word']) == $aWord['word']
173                     ) {
174                         $sNormPostcode = $this->normalizeString($aWord['word']);
175                         if (strpos($sNormQuery, $sNormPostcode) !== false) {
176                             $oValidTokens->addToken(
177                                 $sTok,
178                                 new Token\Postcode($iId, $aWord['word'], null)
179                             );
180                         }
181                     }
182                     break;
183                 case 'S':  // tokens for classification terms (special phrases)
184                     if ($aWord['class'] !== null && $aWord['ctype'] !== null) {
185                         $oValidTokens->addToken($sTok, new Token\SpecialTerm(
186                             $iId,
187                             $aWord['class'],
188                             $aWord['ctype'],
189                             (isset($aWord['operator'])) ? Operator::NEAR : Operator::NONE
190                         ));
191                     }
192                     break;
193                 case 'W': // full-word tokens
194                     $oValidTokens->addToken($sTok, new Token\Word(
195                         $iId,
196                         (int) $aWord['count'],
197                         substr_count($aWord['word_token'], ' ')
198                     ));
199                     break;
200                 case 'w':  // partial word terms
201                     $oValidTokens->addToken($sTok, new Token\Partial(
202                         $iId,
203                         $aWord['word_token'],
204                         (int) $aWord['count']
205                     ));
206                     break;
207                 default:
208                     break;
209             }
210         }
211     }
212 }