Merge remote-tracking branch 'upstream/master'

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

diff --git a/lib-php/tokenizer/icu_tokenizer.php b/lib-php/tokenizer/icu_tokenizer.php
index cebdac476829f240cf1183ca2371f91fbc604ba0..ccce99ca1330d7a42a6976d7fb7c9eaf3d8a84d7 100644 (file)
--- a/lib-php/tokenizer/icu_tokenizer.php
+++ b/lib-php/tokenizer/icu_tokenizer.php
@@ -1,4 +1,12 @@
 <?php
+/**
+ * SPDX-License-Identifier: GPL-2.0-only
+ *
+ * This file is part of Nominatim. (https://nominatim.org)
+ *
+ * Copyright (C) 2022 by the Nominatim developer community.
+ * For a full list of authors see the git log.
+ */
 
 namespace Nominatim;
 
@@ -89,26 +97,19 @@ class Tokenizer
         $aWordLists = array();
         $aTokens = array();
         foreach ($aPhrases as $iPhrase => $oPhrase) {
-            $sNormPhrase = $this->normalizeString($oPhrase->getPhrase());
-            Debug::printVar('Phrase', $sNormPhrase);
-
-            $oWordList = new SimpleWordList($sNormPhrase);
-
-            foreach ($oWordList->getTokens() as $sToken) {
-                $sTransToken = $this->makeStandardWord($sToken);
-                if (!isset($aTokens[$sTransToken])) {
-                    $aTokens[$sTransToken] = array();
-                }
-                $aTokens[$sTransToken][$sToken] = $sToken;
-            }
+            $sNormQuery .= ','.$this->normalizeString($oPhrase->getPhrase());
+            $sPhrase = $this->makeStandardWord($oPhrase->getPhrase());
+            Debug::printVar('Phrase', $sPhrase);
 
+            $oWordList = new SimpleWordList($sPhrase);
+            $aTokens = array_merge($aTokens, $oWordList->getTokens());
             $aWordLists[] = $oWordList;
         }
 
         Debug::printVar('Tokens', $aTokens);
         Debug::printVar('WordLists', $aWordLists);
 
-        $oValidTokens = $this->computeValidTokens($aTokens);
+        $oValidTokens = $this->computeValidTokens($aTokens, $sNormQuery);
 
         foreach ($aPhrases as $iPhrase => $oPhrase) {
             $oPhrase->setWordSets($aWordLists[$iPhrase]->getWordSets($oValidTokens));
@@ -118,16 +119,16 @@ class Tokenizer
     }
 
 
-    private function computeValidTokens($aTokens)
+    private function computeValidTokens($aTokens, $sNormQuery)
     {
         $oValidTokens = new TokenList();
 
         if (!empty($aTokens)) {
-            $this->addTokensFromDB($oValidTokens, $aTokens);
+            $this->addTokensFromDB($oValidTokens, $aTokens, $sNormQuery);
 
             // Try more interpretations for Tokens that could not be matched.
-            foreach ($aTokens as $sToken => $aNormalized) {
-                if (!$oValidTokens->contains($sToken)) {
+            foreach ($aTokens as $sToken) {
+                if ($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(
@@ -150,15 +151,16 @@ class Tokenizer
     }
 
 
-    private function addTokensFromDB(&$oValidTokens, $aTokens)
+    private function addTokensFromDB(&$oValidTokens, $aTokens, $sNormQuery)
     {
         // Check which tokens we have, get the ID numbers
         $sSQL = 'SELECT word_id, word_token, type, word,';
         $sSQL .= "      info->>'op' as operator,";
         $sSQL .= "      info->>'class' as class, info->>'type' as ctype,";
-        $sSQL .= "      info->>'count' as count";
+        $sSQL .= "      info->>'count' as count,";
+        $sSQL .= "      info->>'lookup' as lookup";
         $sSQL .= ' FROM word WHERE word_token in (';
-        $sSQL .= join(',', $this->oDB->getDBQuotedList(array_keys($aTokens))).')';
+        $sSQL .= join(',', $this->oDB->getDBQuotedList($aTokens)).')';
 
         Debug::printSQL($sSQL);
 
@@ -167,23 +169,19 @@ class Tokenizer
         foreach ($aDBWords as $aWord) {
             $iId = (int) $aWord['word_id'];
             $sTok = $aWord['word_token'];
-            $aNorms = $aTokens[$sTok];
 
             switch ($aWord['type']) {
                 case 'C':  // country name tokens
                     if ($aWord['word'] !== null) {
-                        foreach ($aNorms as $sNorm) {
-                            $oValidTokens->addToken(
-                                $sNorm,
-                                new Token\Country($iId, $aWord['word'])
-                            );
-                        }
+                        $oValidTokens->addToken(
+                            $sTok,
+                            new Token\Country($iId, $aWord['word'])
+                        );
                     }
                     break;
                 case 'H':  // house number tokens
-                    foreach ($aNorms as $sNorm) {
-                        $oValidTokens->addToken($sNorm, new Token\HouseNumber($iId, $aWord['word_token']));
-                    }
+                    $sLookup = $aWord['lookup'] ?? $aWord['word_token'];
+                    $oValidTokens->addToken($sTok, new Token\HouseNumber($iId, $sLookup));
                     break;
                 case 'P':  // postcode tokens
                     // Postcodes are not normalized, so they may have content
@@ -193,48 +191,37 @@ class Tokenizer
                         && pg_escape_string($aWord['word']) == $aWord['word']
                     ) {
                         $sNormPostcode = $this->normalizeString($aWord['word']);
-                        foreach ($aNorms as $sNorm) {
-                            if ($sNormPostcode == $sNorm) {
-                                $oValidTokens->addToken(
-                                    $sNorm,
-                                    new Token\Postcode($iId, $aWord['word'], null)
-                                );
-                            }
+                        if (strpos($sNormQuery, $sNormPostcode) !== false) {
+                            $oValidTokens->addToken(
+                                $sTok,
+                                new Token\Postcode($iId, $aWord['word'], null)
+                            );
                         }
                     }
                     break;
                 case 'S':  // tokens for classification terms (special phrases)
                     if ($aWord['class'] !== null && $aWord['ctype'] !== null) {
-                        foreach ($aNorms as $sNorm) {
-                            if ($aWord['word'] == $sNorm) {
-                                $oValidTokens->addToken($sTok, new Token\SpecialTerm(
-                                    $iId,
-                                    $aWord['class'],
-                                    $aWord['ctype'],
-                                    (isset($aWord['operator'])) ? Operator::NEAR : Operator::NONE
-                                ));
-                            }
-                        }
-                    }
-                    break;
-                case 'W': // full-word tokens
-                    foreach ($aNorms as $sNorm) {
-                        $oValidTokens->addToken($sNorm, new Token\Word(
+                        $oValidTokens->addToken($sTok, new Token\SpecialTerm(
                             $iId,
-                            (int) $aWord['count'],
-                            substr_count($aWord['word_token'], ' ') + 1,
-                            levenshtein($aWord['word'], $sNorm) + 1
+                            $aWord['class'],
+                            $aWord['ctype'],
+                            (isset($aWord['operator'])) ? Operator::NEAR : Operator::NONE
                         ));
                     }
                     break;
+                case 'W': // full-word tokens
+                    $oValidTokens->addToken($sTok, new Token\Word(
+                        $iId,
+                        (int) $aWord['count'],
+                        substr_count($aWord['word_token'], ' ')
+                    ));
+                    break;
                 case 'w':  // partial word terms
-                    foreach ($aNorms as $sNorm) {
-                        $oValidTokens->addToken($sNorm, new Token\Partial(
-                            $iId,
-                            $aWord['word_token'],
-                            (int) $aWord['count']
-                        ));
-                    }
+                    $oValidTokens->addToken($sTok, new Token\Partial(
+                        $iId,
+                        $aWord['word_token'],
+                        (int) $aWord['count']
+                    ));
                     break;
                 default:
                     break;