lib-php/tokenizer/legacy_icu_tokenizer.php

   1 <?php
   2
   3 namespace Nominatim;
   4
   5 class Tokenizer
   6 {
   7     private $oDB;
   8
   9     private $oNormalizer;
  10     private $oTransliterator;
  11     private $aCountryRestriction;
  12
  13     public function __construct(&$oDB)
  14     {
  15         $this->oDB =& $oDB;
  16         $this->oNormalizer = \Transliterator::createFromRules(CONST_Term_Normalization_Rules);
  17         $this->oTransliterator = \Transliterator::createFromRules(CONST_Transliteration);
  18     }
  19
  20     public function checkStatus()
  21     {
  22         $sSQL = 'SELECT word_id FROM word WHERE word_id is not null limit 1';
  23         $iWordID = $this->oDB->getOne($sSQL);
  24         if ($iWordID === false) {
  25             throw new \Exception('Query failed', 703);
  26         }
  27         if (!$iWordID) {
  28             throw new \Exception('No value', 704);
  29         }
  30     }
  31
  32
  33     public function setCountryRestriction($aCountries)
  34     {
  35         $this->aCountryRestriction = $aCountries;
  36     }
  37
  38
  39     public function normalizeString($sTerm)
  40     {
  41         if ($this->oNormalizer === null) {
  42             return $sTerm;
  43         }
  44
  45         return $this->oNormalizer->transliterate($sTerm);
  46     }
  47
  48     private function makeStandardWord($sTerm)
  49     {
  50         return trim($this->oTransliterator->transliterate(' '.$sTerm.' '));
  51     }
  52
  53
  54     public function tokensForSpecialTerm($sTerm)
  55     {
  56         $aResults = array();
  57
  58         $sSQL = "SELECT word_id, info->>'class' as class, info->>'type' as type ";
  59         $sSQL .= '   FROM word WHERE word_token = :term and type = \'S\'';
  60
  61         Debug::printVar('Term', $sTerm);
  62         Debug::printSQL($sSQL);
  63         $aSearchWords = $this->oDB->getAll($sSQL, array(':term' => $this->makeStandardWord($sTerm)));
  64
  65         Debug::printVar('Results', $aSearchWords);
  66
  67         foreach ($aSearchWords as $aSearchTerm) {
  68             $aResults[] = new \Nominatim\Token\SpecialTerm(
  69                 $aSearchTerm['word_id'],
  70                 $aSearchTerm['class'],
  71                 $aSearchTerm['type'],
  72                 \Nominatim\Operator::TYPE
  73             );
  74         }
  75
  76         Debug::printVar('Special term tokens', $aResults);
  77
  78         return $aResults;
  79     }
  80
  81
  82     public function extractTokensFromPhrases(&$aPhrases)
  83     {
  84         $sNormQuery = '';
  85         $aWordLists = array();
  86         $aTokens = array();
  87         foreach ($aPhrases as $iPhrase => $oPhrase) {
  88             $sNormQuery .= ','.$this->normalizeString($oPhrase->getPhrase());
  89             $sPhrase = $this->makeStandardWord($oPhrase->getPhrase());
  90             Debug::printVar('Phrase', $sPhrase);
  91             if (strlen($sPhrase) > 0) {
  92                 $aWords = explode(' ', $sPhrase);
  93                 Tokenizer::addTokens($aTokens, $aWords);
  94                 $aWordLists[] = $aWords;
  95             } else {
  96                 $aWordLists[] = array();
  97             }
  98         }
  99
 100         Debug::printVar('Tokens', $aTokens);
 101         Debug::printVar('WordLists', $aWordLists);
 102
 103         $oValidTokens = $this->computeValidTokens($aTokens, $sNormQuery);
 104
 105         foreach ($aPhrases as $iPhrase => $oPhrase) {
 106             $oPhrase->computeWordSets($aWordLists[$iPhrase], $oValidTokens);
 107         }
 108
 109         return $oValidTokens;
 110     }
 111
 112
 113     private function computeValidTokens($aTokens, $sNormQuery)
 114     {
 115         $oValidTokens = new TokenList();
 116
 117         if (!empty($aTokens)) {
 118             $this->addTokensFromDB($oValidTokens, $aTokens, $sNormQuery);
 119
 120             // Try more interpretations for Tokens that could not be matched.
 121             foreach ($aTokens as $sToken) {
 122                 if ($sToken[0] != ' ' && !$oValidTokens->contains($sToken)) {
 123                     if (preg_match('/^([0-9]{5}) [0-9]{4}$/', $sToken, $aData)) {
 124                         // US ZIP+4 codes - merge in the 5-digit ZIP code
 125                         $oValidTokens->addToken(
 126                             $sToken,
 127                             new Token\Postcode(null, $aData[1], 'us')
 128                         );
 129                     } elseif (preg_match('/^[0-9]+$/', $sToken)) {
 130                         // Unknown single word token with a number.
 131                         // Assume it is a house number.
 132                         $oValidTokens->addToken(
 133                             $sToken,
 134                             new Token\HouseNumber(null, trim($sToken))
 135                         );
 136                     }
 137                 }
 138             }
 139         }
 140
 141         return $oValidTokens;
 142     }
 143
 144
 145     private function addTokensFromDB(&$oValidTokens, $aTokens, $sNormQuery)
 146     {
 147         // Check which tokens we have, get the ID numbers
 148         $sSQL = 'SELECT word_id, word_token, type, word,';
 149         $sSQL .= "      info->>'op' as operator,";
 150         $sSQL .= "      info->>'class' as class, info->>'type' as ctype,";
 151         $sSQL .= "      info->>'count' as count";
 152         $sSQL .= ' FROM word WHERE word_token in (';
 153         $sSQL .= join(',', $this->oDB->getDBQuotedList($aTokens)).')';
 154
 155         Debug::printSQL($sSQL);
 156
 157         $aDBWords = $this->oDB->getAll($sSQL, null, 'Could not get word tokens.');
 158
 159         foreach ($aDBWords as $aWord) {
 160             $iId = (int) $aWord['word_id'];
 161             $sTok = $aWord['word_token'];
 162
 163             switch ($aWord['type']) {
 164                 case 'C':  // country name tokens
 165                     if ($aWord['word'] !== null
 166                         && (!$this->aCountryRestriction
 167                             || in_array($aWord['word'], $this->aCountryRestriction))
 168                     ) {
 169                         $oValidTokens->addToken(
 170                             $sTok,
 171                             new Token\Country($iId, $aWord['word'])
 172                         );
 173                     }
 174                     break;
 175                 case 'H':  // house number tokens
 176                     $oValidTokens->addToken($sTok, new Token\HouseNumber($iId, $aWord['word_token']));
 177                     break;
 178                 case 'P':  // postcode tokens
 179                     // Postcodes are not normalized, so they may have content
 180                     // that makes SQL injection possible. Reject postcodes
 181                     // that would need special escaping.
 182                     if ($aWord['word'] !== null
 183                         && pg_escape_string($aWord['word']) == $aWord['word']
 184                     ) {
 185                         $sNormPostcode = $this->normalizeString($aWord['word']);
 186                         if (strpos($sNormQuery, $sNormPostcode) !== false) {
 187                             $oValidTokens->addToken(
 188                                 $sTok,
 189                                 new Token\Postcode($iId, $aWord['word'], null)
 190                             );
 191                         }
 192                     }
 193                     break;
 194                 case 'S':  // tokens for classification terms (special phrases)
 195                     if ($aWord['class'] !== null && $aWord['ctype'] !== null) {
 196                         $oValidTokens->addToken($sTok, new Token\SpecialTerm(
 197                             $iId,
 198                             $aWord['class'],
 199                             $aWord['ctype'],
 200                             (isset($aWord['operator'])) ? Operator::NEAR : Operator::NONE
 201                         ));
 202                     }
 203                     break;
 204                 case 'W': // full-word tokens
 205                     $oValidTokens->addToken($sTok, new Token\Word(
 206                         $iId,
 207                         (int) $aWord['count'],
 208                         substr_count($aWord['word_token'], ' ')
 209                     ));
 210                     break;
 211                 case 'w':  // partial word terms
 212                     $oValidTokens->addToken($sTok, new Token\Partial(
 213                         $iId,
 214                         $aWord['word_token'],
 215                         (int) $aWord['count']
 216                     ));
 217                     break;
 218                 default:
 219                     break;
 220             }
 221         }
 222     }
 223
 224
 225     /**
 226      * Add the tokens from this phrase to the given list of tokens.
 227      *
 228      * @param string[] $aTokens List of tokens to append.
 229      *
 230      * @return void
 231      */
 232     private static function addTokens(&$aTokens, $aWords)
 233     {
 234         $iNumWords = count($aWords);
 235
 236         for ($i = 0; $i < $iNumWords; $i++) {
 237             $sPhrase = $aWords[$i];
 238             $aTokens[$sPhrase] = $sPhrase;
 239
 240             for ($j = $i + 1; $j < $iNumWords; $j++) {
 241                 $sPhrase .= ' '.$aWords[$j];
 242                 $aTokens[$sPhrase] = $sPhrase;
 243             }
 244         }
 245     }
 246 }