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 .= "      info->>'op' as operator,";
 152         $sSQL .= "      info->>'class' as class, info->>'type' as type,";
 153         $sSQL .= "      info->>'count' as count";
 154         $sSQL .= ' FROM word WHERE word_token in (';
 155         $sSQL .= join(',', $this->oDB->getDBQuotedList($aTokens)).')';
 156
 157         Debug::printSQL($sSQL);
 158
 159         $aDBWords = $this->oDB->getAll($sSQL, null, 'Could not get word tokens.');
 160
 161         foreach ($aDBWords as $aWord) {
 162             $iId = (int) $aWord['word_id'];
 163
 164             switch ($aWord['type']) {
 165                 'C':  // country name tokens
 166                     if ($aWord['country'] === null
 167                         || ($this->aCountryRestriction
 168                             && !in_array($aWord['country'], $this->aCountryRestriction))
 169                     ) {
 170                         continue;
 171                     }
 172                     $oToken = new Token\Country($iId, $aWord['country'])
 173                     break;
 174                 'H':  // house number tokens
 175                     $oToken = new Token\HouseNumber($iId, $aWord['word_token']);
 176                     break;
 177                 'P':  // postcode tokens
 178                     // Postcodes are not normalized, so they may have content
 179                     // that makes SQL injection possible. Reject postcodes
 180                     // that would need special escaping.
 181                     if ($aWord['postcode'] === null
 182                         || pg_escape_string($aWord['postcode']) == $aWord['postcode']
 183                     ) {
 184                        continue;
 185                     }
 186                     $sNormPostcode = $this->normalizeString($aWord['postcode']);
 187                     if (strpos($sNormQuery, $sNormPostcode) === false) {
 188                         continue;
 189                     }
 190                     $oToken = new Token\Postcode($iId, $aWord['postcode'], null);
 191                     break;
 192                 'S':  // tokens for classification terms (special phrases)
 193                     if ($aWord['class'] === null || $aWord['type'] === null
 194                     ) {
 195                         continue;
 196                     }
 197                     $oToken = new Token\SpecialTerm(
 198                         $iId,
 199                         $aWord['class'],
 200                         $aWord['type'],
 201                         $aWord['op'] ? Operator::NEAR : Operator::NONE
 202                     );
 203                     break;
 204                 'W': // full-word tokens
 205                     $oToken = new Token\Word(
 206                         $iId,
 207                         (int) $aWord['count'],
 208                         substr_count($aWord['word_token'], ' ')
 209                     );
 210                     break;
 211                 'w':  // partial word terms
 212                     $oToken = new Token\Partial(
 213                         $iId,
 214                         $aWord['word_token'],
 215                         (int) $aWord['count']
 216                     );
 217                     break;
 218                 default:
 219                     continue;
 220             }
 221
 222             $oValidTokens->addToken($aWord['word_token'], $oToken);
 223         }
 224     }
 225
 226
 227     /**
 228      * Add the tokens from this phrase to the given list of tokens.
 229      *
 230      * @param string[] $aTokens List of tokens to append.
 231      *
 232      * @return void
 233      */
 234     private static function addTokens(&$aTokens, $aWords)
 235     {
 236         $iNumWords = count($aWords);
 237
 238         for ($i = 0; $i < $iNumWords; $i++) {
 239             $sPhrase = $aWords[$i];
 240             $aTokens[$sPhrase] = $sPhrase;
 241
 242             for ($j = $i + 1; $j < $iNumWords; $j++) {
 243                 $sPhrase .= ' '.$aWords[$j];
 244                 $aTokens[$sPhrase] = $sPhrase;
 245             }
 246         }
 247     }
 248 }