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_token IN (' a')";
  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, class, type FROM word ';
  59         $sSQL .= '   WHERE word_token = \' \' || :term';
  60         $sSQL .= '   AND class is not null AND class not in (\'place\')';
  61
  62         Debug::printVar('Term', $sTerm);
  63         Debug::printSQL($sSQL);
  64         $aSearchWords = $this->oDB->getAll($sSQL, array(':term' => $this->makeStandardWord($sTerm)));
  65
  66         Debug::printVar('Results', $aSearchWords);
  67
  68         foreach ($aSearchWords as $aSearchTerm) {
  69             $aResults[] = new \Nominatim\Token\SpecialTerm(
  70                 $aSearchTerm['word_id'],
  71                 $aSearchTerm['class'],
  72                 $aSearchTerm['type'],
  73                 \Nominatim\Operator::TYPE
  74             );
  75         }
  76
  77         Debug::printVar('Special term tokens', $aResults);
  78
  79         return $aResults;
  80     }
  81
  82
  83     public function extractTokensFromPhrases(&$aPhrases)
  84     {
  85         $sNormQuery = '';
  86         $aWordLists = array();
  87         $aTokens = array();
  88         foreach ($aPhrases as $iPhrase => $oPhrase) {
  89             $sNormQuery .= ','.$this->normalizeString($oPhrase->getPhrase());
  90             $sPhrase = $this->makeStandardWord($oPhrase->getPhrase());
  91             Debug::printVar('Phrase', $sPhrase);
  92             if (strlen($sPhrase) > 0) {
  93                 $aWords = explode(' ', $sPhrase);
  94                 Tokenizer::addTokens($aTokens, $aWords);
  95                 $aWordLists[] = $aWords;
  96             } else {
  97                 $aWordLists[] = array();
  98             }
  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             $oPhrase->computeWordSets($aWordLists[$iPhrase], $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';
 150         $sSQL .= "      info->>'cc' as country, info->>'postcode' as postcode";
 151         $sSQL .= ' FROM word WHERE word_token in (';
 152         $sSQL .= join(',', $this->oDB->getDBQuotedList($aTokens)).')';
 153
 154         Debug::printSQL($sSQL);
 155
 156         $aDBWords = $this->oDB->getAll($sSQL, null, 'Could not get word tokens.');
 157
 158         foreach ($aDBWords as $aWord) {
 159             $iId = (int) $aWord['word_id'];
 160
 161             switch ($aWord['type']) {
 162                 'C':  // country name tokens
 163                     if ($aWord['country'] === null
 164                         || ($this->aCountryRestriction
 165                             && !in_array($aWord['country'], $this->aCountryRestriction))
 166                     ) {
 167                         continue;
 168                     }
 169                     $oToken = new Token\Country($iId, $aWord['country'])
 170                     break;
 171                 'H':  // house number tokens
 172                     $oToken = new Token\HouseNumber($iId, $aWord['word_token']);
 173                     break;
 174                 'P':  // postcode tokens
 175                     // Postcodes are not normalized, so they may have content
 176                     // that makes SQL injection possible. Reject postcodes
 177                     // that would need special escaping.
 178                     if ($aWord['postcode'] === null
 179                         || pg_escape_string($aWord['postcode']) == $aWord['postcode']
 180                     ) {
 181                        continue;
 182                     }
 183                     $oToken = new Token\Postcode($iId, $aWord['postcode'], null);
 184                 default:
 185                     continue;
 186             }
 187 /*          if ($aWord['class']) {
 188                 // Special terms need to appear in their normalized form.
 189                 // (postcodes are not normalized in the word table)
 190                 $sNormWord = $this->normalizeString($aWord['word']);
 191                 if ($aWord['word'] && strpos($sNormQuery, $sNormWord) === false) {
 192                     continue;
 193                 }
 194
 195                 if ($aWord['class'] == 'place' && $aWord['type'] == 'house') {
 196                     $oToken = new Token\HouseNumber($iId, trim($aWord['word_token']));
 197                 } elseif ($aWord['class'] == 'place' && $aWord['type'] == 'postcode') {
 198                     if ($aWord['word']
 199                         && pg_escape_string($aWord['word']) == $aWord['word']
 200                     ) {
 201                         $oToken = new Token\Postcode(
 202                             $iId,
 203                             $aWord['word'],
 204                             $aWord['country_code']
 205                         );
 206                     }
 207                 } else {
 208                     // near and in operator the same at the moment
 209                     $oToken = new Token\SpecialTerm(
 210                         $iId,
 211                         $aWord['class'],
 212                         $aWord['type'],
 213                         $aWord['operator'] ? Operator::NEAR : Operator::NONE
 214                     );
 215                 }
 216             } elseif ($aWord['country_code']) {
 217                 // Filter country tokens that do not match restricted countries.
 218                 if (!$this->aCountryRestriction
 219                     || in_array($aWord['country_code'], $this->aCountryRestriction)
 220                 ) {
 221                     $oToken = new Token\Country($iId, $aWord['country_code']);
 222                 }
 223             } elseif ($aWord['word_token'][0] == ' ') {
 224                  $oToken = new Token\Word(
 225                      $iId,
 226                      (int) $aWord['count'],
 227                      substr_count($aWord['word_token'], ' ')
 228                  );
 229             } else {
 230                 $oToken = new Token\Partial(
 231                     $iId,
 232                     $aWord['word_token'],
 233                     (int) $aWord['count']
 234                 );
 235             }*/
 236
 237             $oValidTokens->addToken($aWord['word_token'], $oToken);
 238         }
 239     }
 240
 241
 242     /**
 243      * Add the tokens from this phrase to the given list of tokens.
 244      *
 245      * @param string[] $aTokens List of tokens to append.
 246      *
 247      * @return void
 248      */
 249     private static function addTokens(&$aTokens, $aWords)
 250     {
 251         $iNumWords = count($aWords);
 252
 253         for ($i = 0; $i < $iNumWords; $i++) {
 254             $sPhrase = $aWords[$i];
 255             $aTokens[$sPhrase] = $sPhrase;
 256
 257             for ($j = $i + 1; $j < $iNumWords; $j++) {
 258                 $sPhrase .= ' '.$aWords[$j];
 259                 $aTokens[$sPhrase] = $sPhrase;
 260             }
 261         }
 262     }
 263 }