adapt BDD tests for legacy tokenizer + Python frontend

[nominatim.git] / lib-php / Geocode.php

diff --git a/lib-php/Geocode.php b/lib-php/Geocode.php
index eda6df5492c0e9c160e5c4ae7507b2fa1c96cd9b..0881d20a216f5adeb77f22cf75dbc15cab0b4cc0 100644 (file)
--- a/lib-php/Geocode.php
+++ b/lib-php/Geocode.php
@@ -1,4 +1,12 @@
 <?php
 <?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;
 
 
 namespace Nominatim;
 


@@ -7,6 +15,7 @@ require_once(CONST_LibDir.'/Phrase.php');
 require_once(CONST_LibDir.'/ReverseGeocode.php');
 require_once(CONST_LibDir.'/SearchDescription.php');
 require_once(CONST_LibDir.'/SearchContext.php');
 require_once(CONST_LibDir.'/ReverseGeocode.php');
 require_once(CONST_LibDir.'/SearchDescription.php');
 require_once(CONST_LibDir.'/SearchContext.php');

+require_once(CONST_LibDir.'/SearchPosition.php');

 require_once(CONST_LibDir.'/TokenList.php');
 require_once(CONST_TokenizerDir.'/tokenizer.php');
 
 require_once(CONST_LibDir.'/TokenList.php');
 require_once(CONST_TokenizerDir.'/tokenizer.php');
 


@@ -94,7 +103,7 @@ class Geocode
         }
 
         $this->iFinalLimit = $iLimit;
         }
 
         $this->iFinalLimit = $iLimit;

-        $this->iLimit = $iLimit + min($iLimit, 10);
+        $this->iLimit = $iLimit + max($iLimit, 10);

     }
 
     public function setFeatureType($sFeatureType)
     }
 
     public function setFeatureType($sFeatureType)


@@ -181,7 +190,7 @@ class Geocode
 
         $this->bFallback = $oParams->getBool('fallback', $this->bFallback);
 
 
         $this->bFallback = $oParams->getBool('fallback', $this->bFallback);
 

-        // List of excluded Place IDs - used for more acurate pageing
+        // List of excluded Place IDs - used for more accurate pageing

         $sExcluded = $oParams->getStringList('exclude_place_ids');
         if ($sExcluded) {
             foreach ($sExcluded as $iExcludedPlaceID) {
         $sExcluded = $oParams->getStringList('exclude_place_ids');
         if ($sExcluded) {
             foreach ($sExcluded as $iExcludedPlaceID) {


@@ -345,7 +354,11 @@ class Geocode
          */
         foreach ($aPhrases as $iPhrase => $oPhrase) {
             $aNewPhraseSearches = array();
          */
         foreach ($aPhrases as $iPhrase => $oPhrase) {
             $aNewPhraseSearches = array();

-            $sPhraseType = $oPhrase->getPhraseType();
+            $oPosition = new SearchPosition(
+                $oPhrase->getPhraseType(),
+                $iPhrase,
+                count($aPhrases)
+            );

 
             foreach ($oPhrase->getWordSets() as $aWordset) {
                 $aWordsetSearches = $aSearches;
 
             foreach ($oPhrase->getWordSets() as $aWordset) {
                 $aWordsetSearches = $aSearches;


@@ -353,22 +366,20 @@ class Geocode
                 // Add all words from this wordset
                 foreach ($aWordset as $iToken => $sToken) {
                     $aNewWordsetSearches = array();
                 // Add all words from this wordset
                 foreach ($aWordset as $iToken => $sToken) {
                     $aNewWordsetSearches = array();

+                    $oPosition->setTokenPosition($iToken, count($aWordset));

 
                     foreach ($aWordsetSearches as $oCurrentSearch) {
                         foreach ($oValidTokens->get($sToken) as $oSearchTerm) {
 
                     foreach ($aWordsetSearches as $oCurrentSearch) {
                         foreach ($oValidTokens->get($sToken) as $oSearchTerm) {

-                            $aNewSearches = $oCurrentSearch->extendWithSearchTerm(
-                                $sToken,
-                                $oSearchTerm,
-                                $sPhraseType,
-                                $iToken == 0 && $iPhrase == 0,
-                                $iToken + 1 == count($aWordset)
-                                  && $iPhrase + 1 == count($aPhrases),
-                                $iPhrase
-                            );
-
-                            foreach ($aNewSearches as $oSearch) {
-                                if ($oSearch->getRank() < $this->iMaxRank) {
-                                    $aNewWordsetSearches[] = $oSearch;
+                            if ($oSearchTerm->isExtendable($oCurrentSearch, $oPosition)) {
+                                $aNewSearches = $oSearchTerm->extendSearch(
+                                    $oCurrentSearch,
+                                    $oPosition
+                                );
+
+                                foreach ($aNewSearches as $oSearch) {
+                                    if ($oSearch->getRank() < $this->iMaxRank) {
+                                        $aNewWordsetSearches[] = $oSearch;
+                                    }

                                 }
                             }
                         }
                                 }
                             }
                         }


@@ -495,7 +506,6 @@ class Geocode
         if ($this->aCountryCodes) {
             $oCtx->setCountryList($this->aCountryCodes);
         }
         if ($this->aCountryCodes) {
             $oCtx->setCountryList($this->aCountryCodes);
         }

-        $this->oTokenizer->setCountryRestriction($this->aCountryCodes);

 
         Debug::newSection('Query Preprocessing');
 
 
         Debug::newSection('Query Preprocessing');
 


@@ -504,13 +514,6 @@ class Geocode
             userError('Query string is not UTF-8 encoded.');
         }
 
             userError('Query string is not UTF-8 encoded.');
         }
 

-        // Conflicts between US state abreviations and various words for 'the' in different languages
-        if (isset($this->aLangPrefOrder['name:en'])) {
-            $sQuery = preg_replace('/(^|,)\s*il\s*(,|$)/i', '\1illinois\2', $sQuery);
-            $sQuery = preg_replace('/(^|,)\s*al\s*(,|$)/i', '\1alabama\2', $sQuery);
-            $sQuery = preg_replace('/(^|,)\s*la\s*(,|$)/i', '\1louisiana\2', $sQuery);
-        }
-

         // Do we have anything that looks like a lat/lon pair?
         $sQuery = $oCtx->setNearPointFromQuery($sQuery);
 
         // Do we have anything that looks like a lat/lon pair?
         $sQuery = $oCtx->setNearPointFromQuery($sQuery);
 


@@ -552,15 +555,15 @@ class Geocode
 
                 if (!empty($aTokens)) {
                     $aNewSearches = array();
 
                 if (!empty($aTokens)) {
                     $aNewSearches = array();

+                    $oPosition = new SearchPosition('', 0, 1);
+                    $oPosition->setTokenPosition(0, 1);
+

                     foreach ($aSearches as $oSearch) {
                         foreach ($aTokens as $oToken) {
                     foreach ($aSearches as $oSearch) {
                         foreach ($aTokens as $oToken) {

-                            $oNewSearch = clone $oSearch;
-                            $oNewSearch->setPoiSearch(
-                                $oToken->iOperator,
-                                $oToken->sClass,
-                                $oToken->sType
+                            $aNewSearches = array_merge(
+                                $aNewSearches,
+                                $oToken->extendSearch($oSearch, $oPosition)

                             );
                             );

-                            $aNewSearches[] = $oNewSearch;

                         }
                     }
                     $aSearches = $aNewSearches;
                         }
                     }
                     $aSearches = $aNewSearches;


@@ -614,16 +617,15 @@ class Geocode
                     }
                     $aReverseGroupedSearches = $this->getGroupedSearches($aSearches, $aPhrases, $oValidTokens);
 
                     }
                     $aReverseGroupedSearches = $this->getGroupedSearches($aSearches, $aPhrases, $oValidTokens);
 

-                    foreach ($aGroupedSearches as $aSearches) {
+                    foreach ($aReverseGroupedSearches as $aSearches) {

                         foreach ($aSearches as $aSearch) {
                         foreach ($aSearches as $aSearch) {

-                            if (!isset($aReverseGroupedSearches[$aSearch->getRank()])) {
-                                $aReverseGroupedSearches[$aSearch->getRank()] = array();
+                            if (!isset($aGroupedSearches[$aSearch->getRank()])) {
+                                $aGroupedSearches[$aSearch->getRank()] = array();

                             }
                             }

-                            $aReverseGroupedSearches[$aSearch->getRank()][] = $aSearch;
+                            $aGroupedSearches[$aSearch->getRank()][] = $aSearch;

                         }
                     }
 
                         }
                     }
 

-                    $aGroupedSearches = $aReverseGroupedSearches;

                     ksort($aGroupedSearches);
                 }
             } else {
                     ksort($aGroupedSearches);
                 }
             } else {


@@ -841,7 +843,9 @@ class Geocode
                 $aResult['importance'] = 0.001;
                 $aResult['foundorder'] = $aResult['addressimportance'];
             } else {
                 $aResult['importance'] = 0.001;
                 $aResult['foundorder'] = $aResult['addressimportance'];
             } else {

-                $aResult['importance'] = max(0.001, $aResult['importance']);
+                if ($aResult['importance'] == 0) {
+                    $aResult['importance'] = 0.0001;
+                }

                 $aResult['importance'] *= $this->viewboxImportanceFactor(
                     $aResult['lon'],
                     $aResult['lat']
                 $aResult['importance'] *= $this->viewboxImportanceFactor(
                     $aResult['lon'],
                     $aResult['lat']


@@ -870,7 +874,7 @@ class Geocode
                 $iCountWords = 0;
                 $sAddress = $aResult['langaddress'];
                 foreach ($aRecheckWords as $i => $sWord) {
                 $iCountWords = 0;
                 $sAddress = $aResult['langaddress'];
                 foreach ($aRecheckWords as $i => $sWord) {

-                    if (stripos($sAddress, $sWord)!==false) {
+                    if (grapheme_stripos($sAddress, $sWord)!==false) {

                         $iCountWords++;
                         if (preg_match('/(^|,)\s*'.preg_quote($sWord, '/').'\s*(,|$)/', $sAddress)) {
                             $iCountWords += 0.1;
                         $iCountWords++;
                         if (preg_match('/(^|,)\s*'.preg_quote($sWord, '/').'\s*(,|$)/', $sAddress)) {
                             $iCountWords += 0.1;