Merge pull request #2396 from lonvia/partial-word-token

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

diff --git a/lib-php/tokenizer/legacy_tokenizer.php b/lib-php/tokenizer/legacy_tokenizer.php
index ec2d7e68cbeb5ed4baa011db99ec13295f9cc1aa..064b41667a9322bb6cb164dd6f7bb041490d1257 100644 (file)
--- a/lib-php/tokenizer/legacy_tokenizer.php
+++ b/lib-php/tokenizer/legacy_tokenizer.php
@@ -137,14 +137,14 @@ 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(
@@ -218,9 +218,12 @@ class Tokenizer
                     (int) $aWord['count'],
                     substr_count($aWord['word_token'], ' ')
                 );
-            } else {
+            // For backward compatibility: ignore all partial tokens with more
+            // than one word.
+            } elseif (strpos($aWord['word_token'], ' ') === false) {
                 $oToken = new Token\Partial(
                     $iId,
+                    $aWord['word_token'],
                     (int) $aWord['count']
                 );
             }