lib-php/tokenizer/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
  44     public function mostFrequentWords($iNum)
  45     {
  46         $sSQL = "SELECT word FROM word WHERE type = 'W'";
  47         $sSQL .= "ORDER BY info->'count' DESC LIMIT ".$iNum;
  48         return $this->oDB->getCol($sSQL);
  49     }
  50
  51
  52     private function makeStandardWord($sTerm)
  53     {
  54         return trim($this->oTransliterator->transliterate(' '.$sTerm.' '));
  55     }
  56
  57
  58     public function tokensForSpecialTerm($sTerm)
  59     {
  60         $aResults = array();
  61
  62         $sSQL = "SELECT word_id, info->>'class' as class, info->>'type' as type ";
  63         $sSQL .= '   FROM word WHERE word_token = :term and type = \'S\'';
  64
  65         Debug::printVar('Term', $sTerm);
  66         Debug::printSQL($sSQL);
  67         $aSearchWords = $this->oDB->getAll($sSQL, array(':term' => $this->makeStandardWord($sTerm)));
  68
  69         Debug::printVar('Results', $aSearchWords);
  70
  71         foreach ($aSearchWords as $aSearchTerm) {
  72             $aResults[] = new \Nominatim\Token\SpecialTerm(
  73                 $aSearchTerm['word_id'],
  74                 $aSearchTerm['class'],
  75                 $aSearchTerm['type'],
  76                 \Nominatim\Operator::TYPE
  77             );
  78         }
  79
  80         Debug::printVar('Special term tokens', $aResults);
  81
  82         return $aResults;
  83     }
  84
  85
  86     public function extractTokensFromPhrases(&$aPhrases)
  87     {
  88         $sNormQuery = '';
  89         $aWordLists = array();
  90         $aTokens = array();
  91         foreach ($aPhrases as $iPhrase => $oPhrase) {
  92             $sNormQuery .= ','.$this->normalizeString($oPhrase->getPhrase());
  93             $sPhrase = $this->makeStandardWord($oPhrase->getPhrase());
  94             Debug::printVar('Phrase', $sPhrase);
  95
  96             $oWordList = new SimpleWordList($sPhrase);
  97             $aTokens = array_merge($aTokens, $oWordList->getTokens());
  98             $aWordLists[] = $oWordList;
  99         }
 100
 101         Debug::printVar('Tokens', $aTokens);
 102         Debug::printVar('WordLists', $aWordLists);
 103
 104         $oValidTokens = $this->computeValidTokens($aTokens, $sNormQuery);
 105
 106         foreach ($aPhrases as $iPhrase => $oPhrase) {
 107             $oPhrase->setWordSets($aWordLists[$iPhrase]->getWordSets($oValidTokens));
 108         }
 109
 110         return $oValidTokens;
 111     }
 112
 113
 114     private function computeValidTokens($aTokens, $sNormQuery)
 115     {
 116         $oValidTokens = new TokenList();
 117
 118         if (!empty($aTokens)) {
 119             $this->addTokensFromDB($oValidTokens, $aTokens, $sNormQuery);
 120
 121             // Try more interpretations for Tokens that could not be matched.
 122             foreach ($aTokens as $sToken) {
 123                 if ($sToken[0] != ' ' && !$oValidTokens->contains($sToken)) {
 124                     if (preg_match('/^([0-9]{5}) [0-9]{4}$/', $sToken, $aData)) {
 125                         // US ZIP+4 codes - merge in the 5-digit ZIP code
 126                         $oValidTokens->addToken(
 127                             $sToken,
 128                             new Token\Postcode(null, $aData[1], 'us')
 129                         );
 130                     } elseif (preg_match('/^[0-9]+$/', $sToken)) {
 131                         // Unknown single word token with a number.
 132                         // Assume it is a house number.
 133                         $oValidTokens->addToken(
 134                             $sToken,
 135                             new Token\HouseNumber(null, trim($sToken))
 136                         );
 137                     }
 138                 }
 139             }
 140         }
 141
 142         return $oValidTokens;
 143     }
 144
 145
 146     private function addTokensFromDB(&$oValidTokens, $aTokens, $sNormQuery)
 147     {
 148         // Check which tokens we have, get the ID numbers
 149         $sSQL = 'SELECT word_id, word_token, type, word,';
 150         $sSQL .= "      info->>'op' as operator,";
 151         $sSQL .= "      info->>'class' as class, info->>'type' as ctype,";
 152         $sSQL .= "      info->>'count' as count";
 153         $sSQL .= ' FROM word WHERE word_token in (';
 154         $sSQL .= join(',', $this->oDB->getDBQuotedList($aTokens)).')';
 155
 156         Debug::printSQL($sSQL);
 157
 158         $aDBWords = $this->oDB->getAll($sSQL, null, 'Could not get word tokens.');
 159
 160         foreach ($aDBWords as $aWord) {
 161             $iId = (int) $aWord['word_id'];
 162             $sTok = $aWord['word_token'];
 163
 164             switch ($aWord['type']) {
 165                 case 'C':  // country name tokens
 166                     if ($aWord['word'] !== null) {
 167                         $oValidTokens->addToken(
 168                             $sTok,
 169                             new Token\Country($iId, $aWord['word'])
 170                         );
 171                     }
 172                     break;
 173                 case 'H':  // house number tokens
 174                     $oValidTokens->addToken($sTok, new Token\HouseNumber($iId, $aWord['word_token']));
 175                     break;
 176                 case 'P':  // postcode tokens
 177                     // Postcodes are not normalized, so they may have content
 178                     // that makes SQL injection possible. Reject postcodes
 179                     // that would need special escaping.
 180                     if ($aWord['word'] !== null
 181                         && pg_escape_string($aWord['word']) == $aWord['word']
 182                     ) {
 183                         $sNormPostcode = $this->normalizeString($aWord['word']);
 184                         if (strpos($sNormQuery, $sNormPostcode) !== false) {
 185                             $oValidTokens->addToken(
 186                                 $sTok,
 187                                 new Token\Postcode($iId, $aWord['word'], null)
 188                             );
 189                         }
 190                     }
 191                     break;
 192                 case 'S':  // tokens for classification terms (special phrases)
 193                     if ($aWord['class'] !== null && $aWord['ctype'] !== null) {
 194                         $oValidTokens->addToken($sTok, new Token\SpecialTerm(
 195                             $iId,
 196                             $aWord['class'],
 197                             $aWord['ctype'],
 198                             (isset($aWord['operator'])) ? Operator::NEAR : Operator::NONE
 199                         ));
 200                     }
 201                     break;
 202                 case 'W': // full-word tokens
 203                     $oValidTokens->addToken($sTok, new Token\Word(
 204                         $iId,
 205                         (int) $aWord['count'],
 206                         substr_count($aWord['word_token'], ' ')
 207                     ));
 208                     break;
 209                 case 'w':  // partial word terms
 210                     $oValidTokens->addToken($sTok, new Token\Partial(
 211                         $iId,
 212                         $aWord['word_token'],
 213                         (int) $aWord['count']
 214                     ));
 215                     break;
 216                 default:
 217                     break;
 218             }
 219         }
 220     }
 221 }