switch special phrases to new word table format

[nominatim.git] / lib-php / tokenizer / legacy_icu_tokenizer.php

diff --git a/lib-php/tokenizer/legacy_icu_tokenizer.php b/lib-php/tokenizer/legacy_icu_tokenizer.php
index 8cff6f322410366d2e0ca2ceaf143d2b2035ce64..70358976f15a0e7f6a67d485d8fe5eb6977bb1bb 100644 (file)
--- a/lib-php/tokenizer/legacy_icu_tokenizer.php
+++ b/lib-php/tokenizer/legacy_icu_tokenizer.php
@@ -120,14 +120,14 @@ class Tokenizer
 
             // Try more interpretations for Tokens that could not be matched.
             foreach ($aTokens as $sToken) {
-                if ($sToken[0] == ' ' && !$oValidTokens->contains($sToken)) {
-                    if (preg_match('/^ ([0-9]{5}) [0-9]{4}$/', $sToken, $aData)) {
+                if ($sToken[0] != ' ' && !$oValidTokens->contains($sToken)) {
+                    if (preg_match('/^([0-9]{5}) [0-9]{4}$/', $sToken, $aData)) {
                         // US ZIP+4 codes - merge in the 5-digit ZIP code
                         $oValidTokens->addToken(
                             $sToken,
                             new Token\Postcode(null, $aData[1], 'us')
                         );
-                    } elseif (preg_match('/^ [0-9]+$/', $sToken)) {
+                    } elseif (preg_match('/^[0-9]+$/', $sToken)) {
                         // Unknown single word token with a number.
                         // Assume it is a house number.
                         $oValidTokens->addToken(
@@ -146,8 +146,10 @@ class Tokenizer
     private function addTokensFromDB(&$oValidTokens, $aTokens, $sNormQuery)
     {
         // Check which tokens we have, get the ID numbers
-        $sSQL = 'SELECT word_id, word_token, word, class, type, country_code,';
-        $sSQL .= ' operator, coalesce(search_name_count, 0) as count';
+        $sSQL = 'SELECT word_id, word_token, type';
+        $sSQL .= "      info->>'cc' as country, info->>'postcode' as postcode,";
+        $sSQL .= "      info->>'word' as word, info->>'op' as operator,";
+        $sSQL .= "      info->>'class' as class, info->>'type' as type";
         $sSQL .= ' FROM word WHERE word_token in (';
         $sSQL .= join(',', $this->oDB->getDBQuotedList($aTokens)).')';
 
@@ -156,10 +158,54 @@ class Tokenizer
         $aDBWords = $this->oDB->getAll($sSQL, null, 'Could not get word tokens.');
 
         foreach ($aDBWords as $aWord) {
-            $oToken = null;
             $iId = (int) $aWord['word_id'];
 
-            if ($aWord['class']) {
+            switch ($aWord['type']) {
+                'C':  // country name tokens
+                    if ($aWord['country'] === null
+                        || ($this->aCountryRestriction
+                            && !in_array($aWord['country'], $this->aCountryRestriction))
+                    ) {
+                        continue;
+                    }
+                    $oToken = new Token\Country($iId, $aWord['country'])
+                    break;
+                'H':  // house number tokens
+                    $oToken = new Token\HouseNumber($iId, $aWord['word_token']);
+                    break;
+                'P':  // postcode tokens
+                    // Postcodes are not normalized, so they may have content
+                    // that makes SQL injection possible. Reject postcodes
+                    // that would need special escaping.
+                    if ($aWord['postcode'] === null
+                        || pg_escape_string($aWord['postcode']) == $aWord['postcode']
+                    ) {
+                       continue;
+                    }
+                    $sNormPostcode = $this->normalizeString($aWord['postcode']);
+                    if (strpos($sNormQuery, $sNormPostcode) === false) {
+                        continue;
+                    }
+                    $oToken = new Token\Postcode($iId, $aWord['postcode'], null);
+                    break;
+                'S':  // tokens for classification terms (special phrases)
+                    if ($aWord['class'] === null || $aWord['type'] === null
+                        || $aWord['word'] === null
+                        || strpos($sNormQuery, $aWord['word']) === false
+                    ) {
+                        continue;
+                    }
+                    $oToken = new Token\SpecialTerm(
+                        $iId,
+                        $aWord['class'],
+                        $aWord['type'],
+                        $aWord['op'] ? Operator::NEAR : Operator::NONE
+                    );
+                    break;
+                default:
+                    continue;
+            }
+/*          if ($aWord['class']) {
                 // Special terms need to appear in their normalized form.
                 // (postcodes are not normalized in the word table)
                 $sNormWord = $this->normalizeString($aWord['word']);
@@ -198,25 +244,18 @@ class Tokenizer
             } elseif ($aWord['word_token'][0] == ' ') {
                  $oToken = new Token\Word(
                      $iId,
-                     $aWord['word_token'][0] != ' ',
                      (int) $aWord['count'],
                      substr_count($aWord['word_token'], ' ')
                  );
             } else {
                 $oToken = new Token\Partial(
                     $iId,
+                    $aWord['word_token'],
                     (int) $aWord['count']
                 );
-            }
+            }*/
 
-            if ($oToken) {
-                // remove any leading spaces
-                if ($aWord['word_token'][0] == ' ') {
-                    $oValidTokens->addToken(substr($aWord['word_token'], 1), $oToken);
-                } else {
-                    $oValidTokens->addToken($aWord['word_token'], $oToken);
-                }
-            }
+            $oValidTokens->addToken($aWord['word_token'], $oToken);
         }
     }
 
@@ -234,12 +273,10 @@ class Tokenizer
 
         for ($i = 0; $i < $iNumWords; $i++) {
             $sPhrase = $aWords[$i];
-            $aTokens[' '.$sPhrase] = ' '.$sPhrase;
             $aTokens[$sPhrase] = $sPhrase;
 
             for ($j = $i + 1; $j < $iNumWords; $j++) {
                 $sPhrase .= ' '.$aWords[$j];
-                $aTokens[' '.$sPhrase] = ' '.$sPhrase;
                 $aTokens[$sPhrase] = $sPhrase;
             }
         }