X-Git-Url: https://git.openstreetmap.org./nominatim.git/blobdiff_plain/eb0b3bfa4cd74ee9e425c54868a30d8859e694cd..9905bbf733542af5f0ac1e3d827fbb353b446beb:/utils/importWikipedia.php diff --git a/utils/importWikipedia.php b/utils/importWikipedia.php old mode 100755 new mode 100644 index a4b8cb77..2e256e35 --- a/utils/importWikipedia.php +++ b/utils/importWikipedia.php @@ -1,13 +1,11 @@ -#!/usr/bin/php -Cq connect(); if ($aCMDResult['drop-tables']) { @@ -90,12 +89,14 @@ EOD; $oDB->query($sSQL); } + function degreesAndMinutesToDecimal($iDegrees, $iMinutes = 0, $fSeconds = 0, $sNSEW = 'N') { $sNSEW = strtoupper($sNSEW); return ($sNSEW == 'S' || $sNSEW == 'W'?-1:1) * ((float)$iDegrees + (float)$iMinutes/60 + (float)$fSeconds/3600); } + function _parseWikipediaContent($sPageText) { $sPageText = str_replace("\n", ' ', $sPageText); @@ -113,83 +114,83 @@ function _parseWikipediaContent($sPageText) $aState = array('body'); foreach ($aPageText as $i => $sPart) { switch ($sPart) { - case '{{': - array_unshift($aTemplateStack, array('', array())); - array_unshift($aState, 'template'); - break; - case '}}': - if ($aState[0] == 'template' || $aState[0] == 'templateparam') { - $aTemplate = array_shift($aTemplateStack); - array_shift($aState); - - $aTemplates[] = $aTemplate; - } - break; - case '[[': - $sLinkPage = ''; - $sLinkSyn = ''; - array_unshift($aState, 'link'); - break; - case ']]': - if ($aState[0] == 'link' || $aState[0] == 'linksynonim') { - if (!$sLinkSyn) $sLinkSyn = $sLinkPage; - if (substr($sLinkPage, 0, 6) == 'Image:') $sLinkSyn = substr($sLinkPage, 6); - - $aLinks[] = array($sLinkPage, $sLinkSyn); + case '{{': + array_unshift($aTemplateStack, array('', array())); + array_unshift($aState, 'template'); + break; + case '}}': + if ($aState[0] == 'template' || $aState[0] == 'templateparam') { + $aTemplate = array_shift($aTemplateStack); + array_shift($aState); - array_shift($aState); - switch ($aState[0]) { - case 'template': - $aTemplateStack[0][0] .= trim($sPart); - break; - case 'templateparam': - $aTemplateStack[0][1][0] .= $sLinkSyn; - break; - case 'link': - $sLinkPage .= trim($sPart); - break; - case 'linksynonim': - $sLinkSyn .= $sPart; - break; - case 'body': - $sPageBody .= $sLinkSyn; - break; - default: - var_dump($aState, $sPageName, $aTemplateStack, $sPart, $aPageText); - fail('unknown state'); + $aTemplates[] = $aTemplate; } - } - break; - case '|': - if ($aState[0] == 'template' || $aState[0] == 'templateparam') { - // Create a new template paramater - $aState[0] = 'templateparam'; - array_unshift($aTemplateStack[0][1], ''); - } - if ($aState[0] == 'link') $aState[0] = 'linksynonim'; - break; - default: - switch ($aState[0]) { - case 'template': - $aTemplateStack[0][0] .= trim($sPart); - break; - case 'templateparam': - $aTemplateStack[0][1][0] .= $sPart; break; - case 'link': - $sLinkPage .= trim($sPart); + case '[[': + $sLinkPage = ''; + $sLinkSyn = ''; + array_unshift($aState, 'link'); break; - case 'linksynonim': - $sLinkSyn .= $sPart; + case ']]': + if ($aState[0] == 'link' || $aState[0] == 'linksynonim') { + if (!$sLinkSyn) $sLinkSyn = $sLinkPage; + if (substr($sLinkPage, 0, 6) == 'Image:') $sLinkSyn = substr($sLinkPage, 6); + + $aLinks[] = array($sLinkPage, $sLinkSyn); + + array_shift($aState); + switch ($aState[0]) { + case 'template': + $aTemplateStack[0][0] .= trim($sPart); + break; + case 'templateparam': + $aTemplateStack[0][1][0] .= $sLinkSyn; + break; + case 'link': + $sLinkPage .= trim($sPart); + break; + case 'linksynonim': + $sLinkSyn .= $sPart; + break; + case 'body': + $sPageBody .= $sLinkSyn; + break; + default: + var_dump($aState, $sPageName, $aTemplateStack, $sPart, $aPageText); + fail('unknown state'); + } + } break; - case 'body': - $sPageBody .= $sPart; + case '|': + if ($aState[0] == 'template' || $aState[0] == 'templateparam') { + // Create a new template paramater + $aState[0] = 'templateparam'; + array_unshift($aTemplateStack[0][1], ''); + } + if ($aState[0] == 'link') $aState[0] = 'linksynonim'; break; default: - var_dump($aState, $aPageText); - fail('unknown state'); - } - break; + switch ($aState[0]) { + case 'template': + $aTemplateStack[0][0] .= trim($sPart); + break; + case 'templateparam': + $aTemplateStack[0][1][0] .= $sPart; + break; + case 'link': + $sLinkPage .= trim($sPart); + break; + case 'linksynonim': + $sLinkSyn .= $sPart; + break; + case 'body': + $sPageBody .= $sPart; + break; + default: + var_dump($aState, $aPageText); + fail('unknown state'); + } + break; } } return $aTemplates; @@ -201,7 +202,7 @@ function _templatesToProperties($aTemplates) foreach ($aTemplates as $iTemplate => $aTemplate) { $aParams = array(); foreach (array_reverse($aTemplate[1]) as $iParam => $sParam) { - if (($iPos = strpos($sParam, '=')) === FALSE) { + if (($iPos = strpos($sParam, '=')) === false) { $aParams[] = trim($sParam); } else { $aParams[trim(substr($sParam, 0, $iPos))] = trim(substr($sParam, $iPos+1)); @@ -224,7 +225,7 @@ function _templatesToProperties($aTemplates) if (!isset($aPageProperties['sWebsite']) && isset($aParams['website']) && $aParams['website']) { if (preg_match('#^\\[?([^ \\]]+)[^\\]]*\\]?$#', $aParams['website'], $aMatch)) { $aPageProperties['sWebsite'] = $aMatch[1]; - if (strpos($aPageProperties['sWebsite'], ':/'.'/') === FALSE) { + if (strpos($aPageProperties['sWebsite'], ':/'.'/') === false) { $aPageProperties['sWebsite'] = 'http:/'.'/'.$aPageProperties['sWebsite']; } } @@ -239,7 +240,7 @@ function _templatesToProperties($aTemplates) } // Assume the first template with lots of params is the type (fallback for infobox) - if (!isset($aPageProperties['sPossibleInfoboxType']) && sizeof($aParams) > 10) { + if (!isset($aPageProperties['sPossibleInfoboxType']) && count($aParams) > 10) { $aPageProperties['sPossibleInfoboxType'] = trim($aTemplate[0]); // $aPageProperties['aInfoboxParams'] = $aParams; } @@ -304,9 +305,17 @@ function _templatesToProperties($aTemplates) } if (isset($aCMDResult['parse-wikipedia'])) { - $oDB =& getDB(); - $aArticleNames = $oDB->getCol('select page_title from content where page_namespace = 0 and page_id %10 = '.$aCMDResult['parse-wikipedia'].' and (page_content ilike \'%{{Coord%\' or (page_content ilike \'%lat%\' and page_content ilike \'%lon%\'))'); -// $aArticleNames = $oDB->getCol($sSQL = 'select page_title from content where page_namespace = 0 and (page_content ilike \'%{{Coord%\' or (page_content ilike \'%lat%\' and page_content ilike \'%lon%\')) and page_title in (\'Virginia\')'); + $oDB = new Nominatim\DB(); + $oDB->connect(); + + $sSQL = 'select page_title from content where page_namespace = 0 and page_id %10 = '; + $sSQL .= $aCMDResult['parse-wikipedia']; + $sSQL .= ' and (page_content ilike \'%{{Coord%\' or (page_content ilike \'%lat%\' and page_content ilike \'%lon%\'))'; + $aArticleNames = $oDB->getCol($sSQL); + /* $aArticleNames = $oDB->getCol($sSQL = 'select page_title from content where page_namespace = 0 + and (page_content ilike \'%{{Coord%\' or (page_content ilike \'%lat%\' + and page_content ilike \'%lon%\')) and page_title in (\'Virginia\')'); + */ foreach ($aArticleNames as $sArticleName) { $sPageText = $oDB->getOne('select page_content from content where page_namespace = 0 and page_title = \''.pg_escape_string($sArticleName).'\''); $aP = _templatesToProperties(_parseWikipediaContent($sPageText)); @@ -342,23 +351,27 @@ if (isset($aCMDResult['parse-wikipedia'])) { } } + function nominatimXMLStart($hParser, $sName, $aAttr) { - global $aNominatRecords; - switch ($sName) { + global $aNominatRecords; + switch ($sName) { case 'PLACE': - $aNominatRecords[] = $aAttr; - break; - } + $aNominatRecords[] = $aAttr; + break; + } } + function nominatimXMLEnd($hParser, $sName) { } if (isset($aCMDResult['link'])) { - $oDB =& getDB(); + $oDB = new Nominatim\DB(); + $oDB->connect(); + $aWikiArticles = $oDB->getAll("select * from wikipedia_article where language = 'en' and lat is not null and osm_type is null and totalcount < 31 order by importance desc limit 200000"); // If you point this script at production OSM you will be blocked @@ -369,106 +382,108 @@ if (isset($aCMDResult['link'])) { $sURL = $sNominatimBaseURL.'?format=xml&accept-language=en'; - echo "\n-- ".$aRecord['name'].", ".$aRecord['infobox_type']."\n"; + echo "\n-- ".$aRecord['name'].', '.$aRecord['infobox_type']."\n"; $fMaxDist = 0.0000001; $bUnknown = false; switch (strtolower($aRecord['infobox_type'])) { - case 'former country': - continue 2; - case 'sea': - $fMaxDist = 60; // effectively turn it off - $sURL .= "&viewbox=".($aRecord['lon']-$fMaxDist).",".($aRecord['lat']+$fMaxDist).",".($aRecord['lon']+$fMaxDist).",".($aRecord['lat']-$fMaxDist); - break; - case 'country': - case 'island': - case 'islands': - case 'continent': - $fMaxDist = 60; // effectively turn it off - $sURL .= "&featuretype=country"; - $sURL .= "&viewbox=".($aRecord['lon']-$fMaxDist).",".($aRecord['lat']+$fMaxDist).",".($aRecord['lon']+$fMaxDist).",".($aRecord['lat']-$fMaxDist); - break; - case 'prefecture japan': - $aRecord['name'] = trim(str_replace(' Prefecture', ' ', $aRecord['name'])); - case 'state': - case '#us state': - case 'county': - case 'u.s. state': - case 'u.s. state symbols': - case 'german state': - case 'province or territory of canada'; - case 'indian jurisdiction'; - case 'province'; - case 'french region': - case 'region of italy': - case 'kommune': - case '#australia state or territory': - case 'russian federal subject': - $fMaxDist = 4; - $sURL .= "&featuretype=state"; - $sURL .= "&viewbox=".($aRecord['lon']-$fMaxDist).",".($aRecord['lat']+$fMaxDist).",".($aRecord['lon']+$fMaxDist).",".($aRecord['lat']-$fMaxDist); - break; - case 'protected area': - $fMaxDist = 1; - $sURL .= "&nearlat=".$aRecord['lat']; - $sURL .= "&nearlon=".$aRecord['lon']; - $sURL .= "&viewbox=".($aRecord['lon']-$fMaxDist).",".($aRecord['lat']+$fMaxDist).",".($aRecord['lon']+$fMaxDist).",".($aRecord['lat']-$fMaxDist); - break; - case 'settlement': - $bUnknown = true; - case 'french commune': - case 'italian comune': - case 'uk place': - case 'italian comune': - case 'australian place': - case 'german place': - case '#geobox': - case 'u.s. county': - case 'municipality': - case 'city japan': - case 'russian inhabited locality': - case 'finnish municipality/land area': - case 'england county': - case 'israel municipality': - case 'russian city': - case 'city': - $fMaxDist = 0.2; - $sURL .= "&featuretype=settlement"; - $sURL .= "&viewbox=".($aRecord['lon']-0.5).",".($aRecord['lat']+0.5).",".($aRecord['lon']+0.5).",".($aRecord['lat']-0.5); - break; - case 'mountain': - case 'mountain pass': - case 'river': - case 'lake': - case 'airport': - $fMaxDist = 0.2; - $sURL .= "&viewbox=".($aRecord['lon']-0.5).",".($aRecord['lat']+0.5).",".($aRecord['lon']+0.5).",".($aRecord['lat']-0.5); - - case 'ship begin': - $fMaxDist = 0.1; - $aTypes = array('wreck'); - $sURL .= "&viewbox=".($aRecord['lon']-0.01).",".($aRecord['lat']+0.01).",".($aRecord['lon']+0.01).",".($aRecord['lat']-0.01); - $sURL .= "&nearlat=".$aRecord['lat']; - $sURL .= "&nearlon=".$aRecord['lon']; - break; - case 'road': - case 'university': - case 'company': - case 'department': - $fMaxDist = 0.005; - $sURL .= "&viewbox=".($aRecord['lon']-0.01).",".($aRecord['lat']+0.01).",".($aRecord['lon']+0.01).",".($aRecord['lat']-0.01); - $sURL .= "&bounded=1"; - $sURL .= "&nearlat=".$aRecord['lat']; - $sURL .= "&nearlon=".$aRecord['lon']; - break; - default: - $bUnknown = true; - $fMaxDist = 0.005; - $sURL .= "&viewbox=".($aRecord['lon']-0.01).",".($aRecord['lat']+0.01).",".($aRecord['lon']+0.01).",".($aRecord['lat']-0.01); -// $sURL .= "&bounded=1"; - $sURL .= "&nearlat=".$aRecord['lat']; - $sURL .= "&nearlon=".$aRecord['lon']; - echo "-- Unknown: ".$aRecord['infobox_type']."\n"; - break; + case 'former country': + continue 2; + case 'sea': + $fMaxDist = 60; // effectively turn it off + $sURL .= '&viewbox='.($aRecord['lon']-$fMaxDist).','.($aRecord['lat']+$fMaxDist).','.($aRecord['lon']+$fMaxDist).','.($aRecord['lat']-$fMaxDist); + break; + case 'country': + case 'island': + case 'islands': + case 'continent': + $fMaxDist = 60; // effectively turn it off + $sURL .= '&featuretype=country'; + $sURL .= '&viewbox='.($aRecord['lon']-$fMaxDist).','.($aRecord['lat']+$fMaxDist).','.($aRecord['lon']+$fMaxDist).','.($aRecord['lat']-$fMaxDist); + break; + case 'prefecture japan': + $aRecord['name'] = trim(str_replace(' Prefecture', ' ', $aRecord['name'])); + // intentionally no break + case 'state': + case '#us state': + case 'county': + case 'u.s. state': + case 'u.s. state symbols': + case 'german state': + case 'province or territory of canada': + case 'indian jurisdiction': + case 'province': + case 'french region': + case 'region of italy': + case 'kommune': + case '#australia state or territory': + case 'russian federal subject': + $fMaxDist = 4; + $sURL .= '&featuretype=state'; + $sURL .= '&viewbox='.($aRecord['lon']-$fMaxDist).','.($aRecord['lat']+$fMaxDist).','.($aRecord['lon']+$fMaxDist).','.($aRecord['lat']-$fMaxDist); + break; + case 'protected area': + $fMaxDist = 1; + $sURL .= '&nearlat='.$aRecord['lat']; + $sURL .= '&nearlon='.$aRecord['lon']; + $sURL .= '&viewbox='.($aRecord['lon']-$fMaxDist).','.($aRecord['lat']+$fMaxDist).','.($aRecord['lon']+$fMaxDist).','.($aRecord['lat']-$fMaxDist); + break; + case 'settlement': + $bUnknown = true; + // intentionally no break + case 'french commune': + case 'italian comune': + case 'uk place': + case 'italian comune': + case 'australian place': + case 'german place': + case '#geobox': + case 'u.s. county': + case 'municipality': + case 'city japan': + case 'russian inhabited locality': + case 'finnish municipality/land area': + case 'england county': + case 'israel municipality': + case 'russian city': + case 'city': + $fMaxDist = 0.2; + $sURL .= '&featuretype=settlement'; + $sURL .= '&viewbox='.($aRecord['lon']-0.5).','.($aRecord['lat']+0.5).','.($aRecord['lon']+0.5).','.($aRecord['lat']-0.5); + break; + case 'mountain': + case 'mountain pass': + case 'river': + case 'lake': + case 'airport': + $fMaxDist = 0.2; + $sURL .= '&viewbox='.($aRecord['lon']-0.5).','.($aRecord['lat']+0.5).','.($aRecord['lon']+0.5).','.($aRecord['lat']-0.5); + break; + case 'ship begin': + $fMaxDist = 0.1; + $aTypes = array('wreck'); + $sURL .= '&viewbox='.($aRecord['lon']-0.01).','.($aRecord['lat']+0.01).','.($aRecord['lon']+0.01).','.($aRecord['lat']-0.01); + $sURL .= '&nearlat='.$aRecord['lat']; + $sURL .= '&nearlon='.$aRecord['lon']; + break; + case 'road': + case 'university': + case 'company': + case 'department': + $fMaxDist = 0.005; + $sURL .= '&viewbox='.($aRecord['lon']-0.01).','.($aRecord['lat']+0.01).','.($aRecord['lon']+0.01).','.($aRecord['lat']-0.01); + $sURL .= '&bounded=1'; + $sURL .= '&nearlat='.$aRecord['lat']; + $sURL .= '&nearlon='.$aRecord['lon']; + break; + default: + $bUnknown = true; + $fMaxDist = 0.005; + $sURL .= '&viewbox='.($aRecord['lon']-0.01).','.($aRecord['lat']+0.01).','.($aRecord['lon']+0.01).','.($aRecord['lat']-0.01); + // $sURL .= "&bounded=1"; + $sURL .= '&nearlat='.$aRecord['lat']; + $sURL .= '&nearlon='.$aRecord['lon']; + echo '-- Unknown: '.$aRecord['infobox_type']."\n"; + break; } $sNameURL = $sURL.'&q='.urlencode($aRecord['name']); @@ -483,7 +498,7 @@ if (isset($aCMDResult['link'])) { if (!isset($aNominatRecords[0])) { $aNameParts = preg_split('#[(,]#', $aRecord['name']); - if (sizeof($aNameParts) > 1) { + if (count($aNameParts) > 1) { $sNameURL = $sURL.'&q='.urlencode(trim($aNameParts[0])); var_Dump($sNameURL); $sXML = file_get_contents($sNameURL); @@ -492,12 +507,12 @@ if (isset($aCMDResult['link'])) { $hXMLParser = xml_parser_create(); xml_set_element_handler($hXMLParser, 'nominatimXMLStart', 'nominatimXMLEnd'); xml_parse($hXMLParser, $sXML, true); - xml_parser_free($hXMLParser);# + xml_parser_free($hXMLParser); } } // assume first is best/right - for ($i = 0; $i < sizeof($aNominatRecords); $i++) { + for ($i = 0; $i < count($aNominatRecords); $i++) { $fDiff = ($aRecord['lat']-$aNominatRecords[$i]['LAT']) * ($aRecord['lat']-$aNominatRecords[$i]['LAT']); $fDiff += ($aRecord['lon']-$aNominatRecords[$i]['LON']) * ($aRecord['lon']-$aNominatRecords[$i]['LON']); $fDiff = sqrt($fDiff); @@ -514,17 +529,26 @@ if (isset($aCMDResult['link'])) { elseif ($iRank <= 26) $fMaxDist = 0.001; else $fMaxDist = 0.001; } - echo "-- FOUND \"".substr($aNominatRecords[$i]['DISPLAY_NAME'], 0, 50)."\", ".$aNominatRecords[$i]['CLASS'].", ".$aNominatRecords[$i]['TYPE'].", ".$aNominatRecords[$i]['PLACE_RANK'].", ".$aNominatRecords[$i]['OSM_TYPE']." (dist:$fDiff, max:$fMaxDist)\n"; + echo '-- FOUND "'.substr($aNominatRecords[$i]['DISPLAY_NAME'], 0, 50); + echo '", '.$aNominatRecords[$i]['CLASS'].', '.$aNominatRecords[$i]['TYPE']; + echo ', '.$aNominatRecords[$i]['PLACE_RANK'].', '.$aNominatRecords[$i]['OSM_TYPE']; + echo " (dist:$fDiff, max:$fMaxDist)\n"; if ($fDiff > $fMaxDist) { echo "-- Diff too big $fDiff (max: $fMaxDist)".$aRecord['lat'].','.$aNominatRecords[$i]['LAT'].' & '.$aRecord['lon'].','.$aNominatRecords[$i]['LON']." \n"; } else { - $sSQL = "update wikipedia_article set osm_type="; + $sSQL = 'update wikipedia_article set osm_type='; switch ($aNominatRecords[$i]['OSM_TYPE']) { - case 'relation': $sSQL .= "'R'"; break; - case 'way': $sSQL .= "'W'"; break; - case 'node': $sSQL .= "'N'"; break; + case 'relation': + $sSQL .= "'R'"; + break; + case 'way': + $sSQL .= "'W'"; + break; + case 'node': + $sSQL .= "'N'"; + break; } - $sSQL .= ", osm_id=".$aNominatRecords[$i]['OSM_ID']." where language = '".pg_escape_string($aRecord['language'])."' and title = '".pg_escape_string($aRecord['title'])."'"; + $sSQL .= ', osm_id='.$aNominatRecords[$i]['OSM_ID']." where language = '".pg_escape_string($aRecord['language'])."' and title = '".pg_escape_string($aRecord['title'])."'"; $oDB->query($sSQL); break; }