switch word tokens to new word table layout

[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 92dd727283019ea3454b20ee7232f0234f583b0c..314bd27e43848eeb1633df377fe588659e8f0bfa 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) {
 
             // 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')
                         );
                         // 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(
                         // Unknown single word token with a number.
                         // Assume it is a house number.
                         $oValidTokens->addToken(


@@ -146,8 +146,11 @@ class Tokenizer
     private function addTokensFromDB(&$oValidTokens, $aTokens, $sNormQuery)
     {
         // Check which tokens we have, get the ID numbers
     private function addTokensFromDB(&$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->>'op' as operator,";
+        $sSQL .= "      info->>'class' as class, info->>'type' as type,";
+        $sSQL .= "      info->>'count' as count";

         $sSQL .= ' FROM word WHERE word_token in (';
         $sSQL .= join(',', $this->oDB->getDBQuotedList($aTokens)).')';
 
         $sSQL .= ' FROM word WHERE word_token in (';
         $sSQL .= join(',', $this->oDB->getDBQuotedList($aTokens)).')';
 


@@ -156,57 +159,67 @@ class Tokenizer
         $aDBWords = $this->oDB->getAll($sSQL, null, 'Could not get word tokens.');
 
         foreach ($aDBWords as $aWord) {
         $aDBWords = $this->oDB->getAll($sSQL, null, 'Could not get word tokens.');
 
         foreach ($aDBWords as $aWord) {

-            $oToken = null;

             $iId = (int) $aWord['word_id'];
 
             $iId = (int) $aWord['word_id'];
 

-            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']);
-                if ($aWord['word'] && strpos($sNormQuery, $sNormWord) === false) {
-                    continue;
-                }
-
-                if ($aWord['class'] == 'place' && $aWord['type'] == 'house') {
-                    $oToken = new Token\HouseNumber($iId, trim($aWord['word_token']));
-                } elseif ($aWord['class'] == 'place' && $aWord['type'] == 'postcode') {
-                    if ($aWord['word']
-                        && pg_escape_string($aWord['word']) == $aWord['word']
+            switch ($aWord['type']) {
+                'C':  // country name tokens
+                    if ($aWord['country'] === null
+                        || ($this->aCountryRestriction
+                            && !in_array($aWord['country'], $this->aCountryRestriction))

                     ) {
                     ) {

-                        $oToken = new Token\Postcode(
-                            $iId,
-                            $aWord['word'],
-                            $aWord['country_code']
-                        );
+                        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
+                    ) {
+                        continue;

                     }
                     }

-                } else {
-                    // near and in operator the same at the moment

                     $oToken = new Token\SpecialTerm(
                         $iId,
                         $aWord['class'],
                         $aWord['type'],
                     $oToken = new Token\SpecialTerm(
                         $iId,
                         $aWord['class'],
                         $aWord['type'],

-                        $aWord['operator'] ? Operator::NEAR : Operator::NONE
+                        $aWord['op'] ? Operator::NEAR : Operator::NONE

                     );
                     );

-                }
-            } elseif ($aWord['country_code']) {
-                // Filter country tokens that do not match restricted countries.
-                if (!$this->aCountryRestriction
-                    || in_array($aWord['country_code'], $this->aCountryRestriction)
-                ) {
-                    $oToken = new Token\Country($iId, $aWord['country_code']);
-                }
-            } else {
-                $oToken = new Token\Word(
-                    $iId,
-                    $aWord['word_token'][0] != ' ',
-                    (int) $aWord['count'],
-                    substr_count($aWord['word_token'], ' ')
-                );
+                    break;
+                'W': // full-word tokens
+                    $oToken = new Token\Word(
+                        $iId,
+                        (int) $aWord['count'],
+                        substr_count($aWord['word_token'], ' ')
+                    );
+                    break;
+                'w':  // partial word terms
+                    $oToken = new Token\Partial(
+                        $iId,
+                        $aWord['word_token'],
+                        (int) $aWord['count']
+                    );
+                    break;
+                default:
+                    continue;

             }
 
             }
 

-            if ($oToken) {
-                $oValidTokens->addToken($aWord['word_token'], $oToken);
-            }
+            $oValidTokens->addToken($aWord['word_token'], $oToken);

         }
     }
 
         }
     }
 


@@ -224,12 +237,10 @@ class Tokenizer
 
         for ($i = 0; $i < $iNumWords; $i++) {
             $sPhrase = $aWords[$i];
 
         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;
 
             for ($j = $i + 1; $j < $iNumWords; $j++) {
                 $sPhrase .= ' '.$aWords[$j];

-                $aTokens[' '.$sPhrase] = ' '.$sPhrase;

                 $aTokens[$sPhrase] = $sPhrase;
             }
         }
                 $aTokens[$sPhrase] = $sPhrase;
             }
         }