X-Git-Url: https://git.openstreetmap.org./nominatim.git/blobdiff_plain/96b6a1a41892224b79fb99917981843aef6a4465..00265af528652d6c1bb32cf0694d71e8c5603f39:/lib/Geocode.php diff --git a/lib/Geocode.php b/lib/Geocode.php index f65a485a..f7f97593 100644 --- a/lib/Geocode.php +++ b/lib/Geocode.php @@ -2,9 +2,11 @@ namespace Nominatim; -require_once(CONST_BasePath.'/lib/NearPoint.php'); require_once(CONST_BasePath.'/lib/PlaceLookup.php'); +require_once(CONST_BasePath.'/lib/Phrase.php'); require_once(CONST_BasePath.'/lib/ReverseGeocode.php'); +require_once(CONST_BasePath.'/lib/SearchDescription.php'); +require_once(CONST_BasePath.'/lib/SearchContext.php'); class Geocode { @@ -36,9 +38,8 @@ class Geocode protected $bBoundedSearch = false; protected $aViewBox = false; - protected $sViewboxCentreSQL = false; - protected $sViewboxSmallSQL = false; - protected $sViewboxLargeSQL = false; + protected $aRoutePoints = false; + protected $aRouteWidth = false; protected $iMaxRank = 20; protected $iMinAddressRank = 0; @@ -183,26 +184,6 @@ class Geocode $this->iMaxAddressRank = $iMax; } - public function setRoute($aRoutePoints, $fRouteWidth) - { - $this->aViewBox = false; - - $this->sViewboxCentreSQL = "ST_SetSRID('LINESTRING("; - $sSep = ''; - foreach ($aRoutePoints as $aPoint) { - $fPoint = (float)$aPoint; - $this->sViewboxCentreSQL .= $sSep.$fPoint; - $sSep = ($sSep == ' ') ? ',' : ' '; - } - $this->sViewboxCentreSQL .= ")'::geometry,4326)"; - - $this->sViewboxSmallSQL = 'ST_BUFFER('.$this->sViewboxCentreSQL; - $this->sViewboxSmallSQL .= ','.($fRouteWidth/69).')'; - - $this->sViewboxLargeSQL = 'ST_BUFFER('.$this->sViewboxCentreSQL; - $this->sViewboxLargeSQL .= ','.($fRouteWidth/30).')'; - } - public function setViewbox($aViewbox) { $this->aViewBox = array_map('floatval', $aViewbox); @@ -217,29 +198,6 @@ class Geocode ) { userError("Bad parameter 'viewbox'. Not a box."); } - - $fHeight = $this->aViewBox[0] - $this->aViewBox[2]; - $fWidth = $this->aViewBox[1] - $this->aViewBox[3]; - $aBigViewBox[0] = $this->aViewBox[0] + $fHeight; - $aBigViewBox[2] = $this->aViewBox[2] - $fHeight; - $aBigViewBox[1] = $this->aViewBox[1] + $fWidth; - $aBigViewBox[3] = $this->aViewBox[3] - $fWidth; - - $this->sViewboxCentreSQL = false; - $this->sViewboxSmallSQL = sprintf( - 'ST_SetSRID(ST_MakeBox2D(ST_Point(%F,%F),ST_Point(%F,%F)),4326)', - $this->aViewBox[0], - $this->aViewBox[1], - $this->aViewBox[2], - $this->aViewBox[3] - ); - $this->sViewboxLargeSQL = sprintf( - 'ST_SetSRID(ST_MakeBox2D(ST_Point(%F,%F),ST_Point(%F,%F)),4326)', - $aBigViewBox[0], - $aBigViewBox[1], - $aBigViewBox[2], - $aBigViewBox[3] - ); } public function setQuery($sQueryString) @@ -318,7 +276,8 @@ class Geocode $aRoute = $oParams->getStringList('route'); $fRouteWidth = $oParams->getFloat('routewidth'); if ($aRoute && $fRouteWidth) { - $this->setRoute($aRoute, $fRouteWidth); + $this->aRoutePoints = $aRoute; + $this->aRouteWidth = $fRouteWidth; } } } @@ -367,7 +326,7 @@ class Geocode $this->aAddressRankList = array(); $this->aStructuredQuery = array(); - $this->sAllowedTypesSQLList = False; + $this->sAllowedTypesSQLList = false; $this->loadStructuredAddressElement($sAmenity, 'amenity', 26, 30, false); $this->loadStructuredAddressElement($sStreet, 'street', 26, 30, false); @@ -406,26 +365,20 @@ class Geocode return false; } - public function getDetails($aPlaceIDs) + public function getDetails($aPlaceIDs, $oCtx) { //$aPlaceIDs is an array with key: placeID and value: tiger-housenumber, if found, else -1 if (sizeof($aPlaceIDs) == 0) return array(); - $sLanguagePrefArraySQL = "ARRAY[".join(',', array_map("getDBQuoted", $this->aLangPrefOrder))."]"; + $sLanguagePrefArraySQL = getArraySQL( + array_map("getDBQuoted", $this->aLangPrefOrder) + ); // Get the details for display (is this a redundant extra step?) $sPlaceIDs = join(',', array_keys($aPlaceIDs)); - $sImportanceSQL = ''; - $sImportanceSQLGeom = ''; - if ($this->sViewboxSmallSQL) { - $sImportanceSQL .= " CASE WHEN ST_Contains($this->sViewboxSmallSQL, ST_Collect(centroid)) THEN 1 ELSE 0.75 END * "; - $sImportanceSQLGeom .= " CASE WHEN ST_Contains($this->sViewboxSmallSQL, geometry) THEN 1 ELSE 0.75 END * "; - } - if ($this->sViewboxLargeSQL) { - $sImportanceSQL .= " CASE WHEN ST_Contains($this->sViewboxLargeSQL, ST_Collect(centroid)) THEN 1 ELSE 0.75 END * "; - $sImportanceSQLGeom .= " CASE WHEN ST_Contains($this->sViewboxLargeSQL, geometry) THEN 1 ELSE 0.75 END * "; - } + $sImportanceSQL = $oCtx->viewboxImportanceSQL('ST_Collect(centroid)'); + $sImportanceSQLGeom = $oCtx->viewboxImportanceSQL('geometry'); $sSQL = "SELECT "; $sSQL .= " osm_type,"; @@ -445,17 +398,21 @@ class Geocode if ($this->bIncludeNameDetails) $sSQL .= "hstore_to_json(name)::text AS names,"; $sSQL .= " avg(ST_X(centroid)) AS lon, "; $sSQL .= " avg(ST_Y(centroid)) AS lat, "; - $sSQL .= " ".$sImportanceSQL."COALESCE(importance,0.75-(rank_search::float/40)) AS importance, "; - $sSQL .= " ( "; - $sSQL .= " SELECT max(p.importance*(p.rank_address+2))"; - $sSQL .= " FROM "; - $sSQL .= " place_addressline s, "; - $sSQL .= " placex p"; - $sSQL .= " WHERE s.place_id = min(CASE WHEN placex.rank_search < 28 THEN placex.place_id ELSE placex.parent_place_id END)"; - $sSQL .= " AND p.place_id = s.address_place_id "; - $sSQL .= " AND s.isaddress "; - $sSQL .= " AND p.importance is not null "; - $sSQL .= " ) AS addressimportance, "; + $sSQL .= " COALESCE(importance,0.75-(rank_search::float/40)) $sImportanceSQL AS importance, "; + if ($oCtx->hasNearPoint()) { + $sSQL .= $oCtx->distanceSQL('ST_Collect(centroid)')." AS addressimportance,"; + } else { + $sSQL .= " ( "; + $sSQL .= " SELECT max(p.importance*(p.rank_address+2))"; + $sSQL .= " FROM "; + $sSQL .= " place_addressline s, "; + $sSQL .= " placex p"; + $sSQL .= " WHERE s.place_id = min(CASE WHEN placex.rank_search < 28 THEN placex.place_id ELSE placex.parent_place_id END)"; + $sSQL .= " AND p.place_id = s.address_place_id "; + $sSQL .= " AND s.isaddress "; + $sSQL .= " AND p.importance is not null "; + $sSQL .= " ) AS addressimportance, "; + } $sSQL .= " (extratags->'place') AS extra_place "; $sSQL .= " FROM placex"; $sSQL .= " WHERE place_id in ($sPlaceIDs) "; @@ -504,17 +461,21 @@ class Geocode if ($this->bIncludeExtraTags) $sSQL .= "null AS extra,"; if ($this->bIncludeNameDetails) $sSQL .= "null AS names,"; $sSQL .= " ST_x(st_centroid(geometry)) AS lon, ST_y(st_centroid(geometry)) AS lat,"; - $sSQL .= $sImportanceSQLGeom."(0.75-(rank_search::float/40)) AS importance, "; - $sSQL .= " ("; - $sSQL .= " SELECT max(p.importance*(p.rank_address+2))"; - $sSQL .= " FROM "; - $sSQL .= " place_addressline s, "; - $sSQL .= " placex p"; - $sSQL .= " WHERE s.place_id = lp.parent_place_id"; - $sSQL .= " AND p.place_id = s.address_place_id "; - $sSQL .= " AND s.isaddress"; - $sSQL .= " AND p.importance is not null"; - $sSQL .= " ) AS addressimportance, "; + $sSQL .= " (0.75-(rank_search::float/40)) $sImportanceSQLGeom AS importance, "; + if ($oCtx->hasNearPoint()) { + $sSQL .= $oCtx->distanceSQL('geometry')." AS addressimportance,"; + } else { + $sSQL .= " ("; + $sSQL .= " SELECT max(p.importance*(p.rank_address+2))"; + $sSQL .= " FROM "; + $sSQL .= " place_addressline s, "; + $sSQL .= " placex p"; + $sSQL .= " WHERE s.place_id = lp.parent_place_id"; + $sSQL .= " AND p.place_id = s.address_place_id "; + $sSQL .= " AND s.isaddress"; + $sSQL .= " AND p.importance is not null"; + $sSQL .= " ) AS addressimportance, "; + } $sSQL .= " null AS extra_place "; $sSQL .= "FROM location_postcode lp"; $sSQL .= " WHERE place_id in ($sPlaceIDs) "; @@ -552,17 +513,21 @@ class Geocode if ($this->bIncludeNameDetails) $sSQL .= "null AS names,"; $sSQL .= " avg(st_x(centroid)) AS lon, "; $sSQL .= " avg(st_y(centroid)) AS lat,"; - $sSQL .= " ".$sImportanceSQL."-1.15 AS importance, "; - $sSQL .= " ("; - $sSQL .= " SELECT max(p.importance*(p.rank_address+2))"; - $sSQL .= " FROM "; - $sSQL .= " place_addressline s, "; - $sSQL .= " placex p"; - $sSQL .= " WHERE s.place_id = min(blub.parent_place_id)"; - $sSQL .= " AND p.place_id = s.address_place_id "; - $sSQL .= " AND s.isaddress"; - $sSQL .= " AND p.importance is not null"; - $sSQL .= " ) AS addressimportance, "; + $sSQL .= " -1.15".$sImportanceSQL." AS importance, "; + if ($oCtx->hasNearPoint()) { + $sSQL .= $oCtx->distanceSQL('ST_Collect(centroid)')." AS addressimportance,"; + } else { + $sSQL .= " ("; + $sSQL .= " SELECT max(p.importance*(p.rank_address+2))"; + $sSQL .= " FROM "; + $sSQL .= " place_addressline s, "; + $sSQL .= " placex p"; + $sSQL .= " WHERE s.place_id = min(blub.parent_place_id)"; + $sSQL .= " AND p.place_id = s.address_place_id "; + $sSQL .= " AND s.isaddress"; + $sSQL .= " AND p.importance is not null"; + $sSQL .= " ) AS addressimportance, "; + } $sSQL .= " null AS extra_place "; $sSQL .= " FROM ("; $sSQL .= " SELECT place_id, "; // interpolate the Tiger housenumbers here @@ -602,18 +567,22 @@ class Geocode if ($this->bIncludeNameDetails) $sSQL .= "null AS names, "; $sSQL .= " AVG(st_x(centroid)) AS lon, "; $sSQL .= " AVG(st_y(centroid)) AS lat, "; - $sSQL .= " ".$sImportanceSQL."-0.1 AS importance, "; // slightly smaller than the importance for normal houses with rank 30, which is 0 - $sSQL .= " ("; - $sSQL .= " SELECT "; - $sSQL .= " MAX(p.importance*(p.rank_address+2)) "; - $sSQL .= " FROM"; - $sSQL .= " place_addressline s, "; - $sSQL .= " placex p"; - $sSQL .= " WHERE s.place_id = min(blub.parent_place_id) "; - $sSQL .= " AND p.place_id = s.address_place_id "; - $sSQL .= " AND s.isaddress "; - $sSQL .= " AND p.importance is not null"; - $sSQL .= " ) AS addressimportance,"; + $sSQL .= " -0.1".$sImportanceSQL." AS importance, "; // slightly smaller than the importance for normal houses with rank 30, which is 0 + if ($oCtx->hasNearPoint()) { + $sSQL .= $oCtx->distanceSQL('ST_Collect(centroid)')." AS addressimportance,"; + } else { + $sSQL .= " ("; + $sSQL .= " SELECT "; + $sSQL .= " MAX(p.importance*(p.rank_address+2)) "; + $sSQL .= " FROM"; + $sSQL .= " place_addressline s, "; + $sSQL .= " placex p"; + $sSQL .= " WHERE s.place_id = min(blub.parent_place_id) "; + $sSQL .= " AND p.place_id = s.address_place_id "; + $sSQL .= " AND s.isaddress "; + $sSQL .= " AND p.importance is not null"; + $sSQL .= " ) AS addressimportance,"; + } $sSQL .= " null AS extra_place "; $sSQL .= " FROM ("; $sSQL .= " SELECT "; @@ -661,17 +630,21 @@ class Geocode if ($this->bIncludeNameDetails) $sSQL .= "null AS names, "; $sSQL .= " avg(ST_X(centroid)) AS lon, "; $sSQL .= " avg(ST_Y(centroid)) AS lat, "; - $sSQL .= " ".$sImportanceSQL."-1.10 AS importance, "; - $sSQL .= " ( "; - $sSQL .= " SELECT max(p.importance*(p.rank_address+2))"; - $sSQL .= " FROM "; - $sSQL .= " place_addressline s, "; - $sSQL .= " placex p"; - $sSQL .= " WHERE s.place_id = min(location_property_aux.parent_place_id)"; - $sSQL .= " AND p.place_id = s.address_place_id "; - $sSQL .= " AND s.isaddress"; - $sSQL .= " AND p.importance is not null"; - $sSQL .= " ) AS addressimportance, "; + $sSQL .= " -1.10".$sImportanceSQL." AS importance, "; + if ($oCtx->hasNearPoint()) { + $sSQL .= $oCtx->distanceSQL('ST_Collect(centroid)')." AS addressimportance,"; + } else { + $sSQL .= " ( "; + $sSQL .= " SELECT max(p.importance*(p.rank_address+2))"; + $sSQL .= " FROM "; + $sSQL .= " place_addressline s, "; + $sSQL .= " placex p"; + $sSQL .= " WHERE s.place_id = min(location_property_aux.parent_place_id)"; + $sSQL .= " AND p.place_id = s.address_place_id "; + $sSQL .= " AND s.isaddress"; + $sSQL .= " AND p.importance is not null"; + $sSQL .= " ) AS addressimportance, "; + } $sSQL .= " null AS extra_place "; $sSQL .= " FROM location_property_aux "; $sSQL .= " WHERE place_id in ($sPlaceIDs) "; @@ -696,7 +669,7 @@ class Geocode return $aSearchResults; } - public function getGroupedSearches($aSearches, $aPhraseTypes, $aPhrases, $aValidTokens, $aWordFrequencyScores, $bStructuredPhrases, $sNormQuery) + public function getGroupedSearches($aSearches, $aPhrases, $aValidTokens, $bIsStructured) { /* Calculate all searches using aValidTokens i.e. @@ -711,12 +684,11 @@ class Geocode */ $iGlobalRank = 0; - foreach ($aPhrases as $iPhrase => $aPhrase) { + foreach ($aPhrases as $iPhrase => $oPhrase) { $aNewPhraseSearches = array(); - if ($bStructuredPhrases) $sPhraseType = $aPhraseTypes[$iPhrase]; - else $sPhraseType = ''; + $sPhraseType = $bIsStructured ? $oPhrase->getPhraseType() : ''; - foreach ($aPhrase['wordsets'] as $iWordSet => $aWordset) { + foreach ($oPhrase->getWordSets() as $iWordSet => $aWordset) { // Too many permutations - too expensive if ($iWordSet > 120) break; @@ -727,169 +699,62 @@ class Geocode //echo "
$sToken"; $aNewWordsetSearches = array(); - foreach ($aWordsetSearches as $aCurrentSearch) { + foreach ($aWordsetSearches as $oCurrentSearch) { //echo ""; - //var_dump($aCurrentSearch); + //var_dump($oCurrentSearch); //echo ""; // If the token is valid if (isset($aValidTokens[' '.$sToken])) { - // TODO variable should go into aCurrentSearch - $bHavePostcode = false; foreach ($aValidTokens[' '.$sToken] as $aSearchTerm) { - $aSearch = $aCurrentSearch; - $aSearch['iSearchRank']++; - if (($sPhraseType == '' || $sPhraseType == 'country') && !empty($aSearchTerm['country_code']) && $aSearchTerm['country_code'] != '0') { - if ($aSearch['sCountryCode'] === false) { - $aSearch['sCountryCode'] = strtolower($aSearchTerm['country_code']); - // Country is almost always at the end of the string - increase score for finding it anywhere else (optimisation) - if (($iToken+1 != sizeof($aWordset) || $iPhrase+1 != sizeof($aPhrases))) { - $aSearch['iSearchRank'] += 5; - } - if ($aSearch['iSearchRank'] < $this->iMaxRank) $aNewWordsetSearches[] = $aSearch; - // If it is at the beginning, we can be almost sure that this is the wrong order - // Increase score for all searches. - if ($iToken == 0 && $iPhrase == 0) { - $iGlobalRank++; - } + $aNewSearches = $oCurrentSearch->extendWithFullTerm( + $aSearchTerm, + isset($aValidTokens[$sToken]) + && strpos($sToken, ' ') === false, + $sPhraseType, + $iToken == 0 && $iPhrase == 0, + $iPhrase == 0, + $iToken + 1 == sizeof($aWordset) + && $iPhrase + 1 == sizeof($aPhrases), + $iGlobalRank + ); + + foreach ($aNewSearches as $oSearch) { + if ($oSearch->getRank() < $this->iMaxRank) { + $aNewWordsetSearches[] = $oSearch; } - } elseif (($sPhraseType == '' || $sPhraseType == 'postalcode') && $aSearchTerm['class'] == 'place' && $aSearchTerm['type'] == 'postcode') { - // We need to try the case where the postal code is the primary element (i.e. no way to tell if it is (postalcode, city) OR (city, postalcode) so try both - if ($aSearch['sPostcode'] === '' && - isset($aSearchTerm['word']) && $aSearchTerm['word'] && strpos($sNormQuery, $this->normTerm($aSearchTerm['word'])) !== false) { - // If we have structured search or this is the first term, - // make the postcode the primary search element. - if (!$bHavePostcode && $aSearch['sOperator'] === '' && ($sPhraseType == 'postalcode' || ($iToken == 0 && $iPhrase == 0))) { - $aNewSearch = $aSearch; - $aNewSearch['sOperator'] = 'postcode'; - $aNewSearch['aAddress'] = array_merge($aNewSearch['aAddress'], $aNewSearch['aName']); - $aNewSearch['aName'] = array($aSearchTerm['word_id'] => $aSearchTerm['word']); - if ($aSearch['iSearchRank'] < $this->iMaxRank) $aNewWordsetSearches[] = $aNewSearch; - $bHavePostcode = true; - } - - // If we have a structured search or this is not the first term, - // add the postcode as an addendum. - if ($aSearch['sOperator'] !== 'postcode' && ($sPhraseType == 'postalcode' || sizeof($aSearch['aName']))) { - $aSearch['sPostcode'] = $aSearchTerm['word']; - if ($aSearch['iSearchRank'] < $this->iMaxRank) $aNewWordsetSearches[] = $aSearch; - } - } - } elseif (($sPhraseType == '' || $sPhraseType == 'street') && $aSearchTerm['class'] == 'place' && $aSearchTerm['type'] == 'house') { - if ($aSearch['sHouseNumber'] === '' && $aSearch['sOperator'] !== 'postcode') { - $aSearch['sHouseNumber'] = $sToken; - // sanity check: if the housenumber is not mainly made - // up of numbers, add a penalty - if (preg_match_all("/[^0-9]/", $sToken, $aMatches) > 2) $aSearch['iSearchRank']++; - // also must not appear in the middle of the address - if ($aSearch['aAddress'] || $aSearch['aAddressNonSearch']) $aSearch['iSearchRank'] += 1; - if ($aSearch['iSearchRank'] < $this->iMaxRank) $aNewWordsetSearches[] = $aSearch; - /* - // Fall back to not searching for this item (better than nothing) - $aSearch = $aCurrentSearch; - $aSearch['iSearchRank'] += 1; - if ($aSearch['iSearchRank'] < $this->iMaxRank) $aNewWordsetSearches[] = $aSearch; - */ - } - } elseif ($sPhraseType == '' && $aSearchTerm['class'] !== '' && $aSearchTerm['class'] !== null) { - // require a normalized exact match of the term - // if we have the normalizer version of the query - // available - if ($aSearch['sOperator'] === '' - && ($sNormQuery === null || !($aSearchTerm['word'] && strpos($sNormQuery, $aSearchTerm['word']) === false))) { - $aSearch['sClass'] = $aSearchTerm['class']; - $aSearch['sType'] = $aSearchTerm['type']; - if ($aSearchTerm['operator'] == '') { - $aSearch['sOperator'] = sizeof($aSearch['aName']) ? 'name' : 'near'; - $aSearch['iSearchRank'] += 2; - } else { - $aSearch['sOperator'] = 'near'; // near = in for the moment - } - - if ($aSearch['iSearchRank'] < $this->iMaxRank) $aNewWordsetSearches[] = $aSearch; - } - } elseif (isset($aSearchTerm['word_id']) && $aSearchTerm['word_id']) { - if (sizeof($aSearch['aName'])) { - if ((!$bStructuredPhrases || $iPhrase > 0) && $sPhraseType != 'country' && (!isset($aValidTokens[$sToken]) || strpos($sToken, ' ') !== false)) { - $aSearch['aAddress'][$aSearchTerm['word_id']] = $aSearchTerm['word_id']; - } else { - $aCurrentSearch['aFullNameAddress'][$aSearchTerm['word_id']] = $aSearchTerm['word_id']; - $aSearch['iSearchRank'] += 1000; // skip; - } - } else { - $aSearch['aName'][$aSearchTerm['word_id']] = $aSearchTerm['word_id']; - //$aSearch['iNamePhrase'] = $iPhrase; - } - if ($aSearch['iSearchRank'] < $this->iMaxRank) $aNewWordsetSearches[] = $aSearch; } } } // Look for partial matches. // Note that there is no point in adding country terms here - // because country are omitted in the address. + // because country is omitted in the address. if (isset($aValidTokens[$sToken]) && $sPhraseType != 'country') { // Allow searching for a word - but at extra cost foreach ($aValidTokens[$sToken] as $aSearchTerm) { - if (isset($aSearchTerm['word_id']) && $aSearchTerm['word_id']) { - if ((!$bStructuredPhrases || $iPhrase > 0) && sizeof($aCurrentSearch['aName']) && strpos($sToken, ' ') === false) { - $aSearch = $aCurrentSearch; - $aSearch['iSearchRank'] += 1; - if ($aWordFrequencyScores[$aSearchTerm['word_id']] < CONST_Max_Word_Frequency) { - $aSearch['aAddress'][$aSearchTerm['word_id']] = $aSearchTerm['word_id']; - if ($aSearch['iSearchRank'] < $this->iMaxRank) $aNewWordsetSearches[] = $aSearch; - } elseif (isset($aValidTokens[' '.$sToken])) { // revert to the token version? - $aSearch['aAddressNonSearch'][$aSearchTerm['word_id']] = $aSearchTerm['word_id']; - $aSearch['iSearchRank'] += 1; - if ($aSearch['iSearchRank'] < $this->iMaxRank) $aNewWordsetSearches[] = $aSearch; - foreach ($aValidTokens[' '.$sToken] as $aSearchTermToken) { - if (empty($aSearchTermToken['country_code']) - && empty($aSearchTermToken['lat']) - && empty($aSearchTermToken['class']) - ) { - $aSearch = $aCurrentSearch; - $aSearch['iSearchRank'] += 1; - $aSearch['aAddress'][$aSearchTermToken['word_id']] = $aSearchTermToken['word_id']; - if ($aSearch['iSearchRank'] < $this->iMaxRank) $aNewWordsetSearches[] = $aSearch; - } - } - } else { - $aSearch['aAddressNonSearch'][$aSearchTerm['word_id']] = $aSearchTerm['word_id']; - if (preg_match('#^[0-9]+$#', $sToken)) $aSearch['iSearchRank'] += 2; - if ($aSearch['iSearchRank'] < $this->iMaxRank) $aNewWordsetSearches[] = $aSearch; - } - } - - if ((!$aCurrentSearch['sPostcode'] && !$aCurrentSearch['aAddress'] && !$aCurrentSearch['aAddressNonSearch']) - && (!sizeof($aCurrentSearch['aName']) || $aCurrentSearch['iNamePhrase'] == $iPhrase)) { - $aSearch = $aCurrentSearch; - $aSearch['iSearchRank'] += 1; - if (!sizeof($aCurrentSearch['aName'])) $aSearch['iSearchRank'] += 1; - if (preg_match('#^[0-9]+$#', $sToken)) $aSearch['iSearchRank'] += 2; - if ($aWordFrequencyScores[$aSearchTerm['word_id']] < CONST_Max_Word_Frequency) { - $aSearch['aName'][$aSearchTerm['word_id']] = $aSearchTerm['word_id']; - } else { - $aSearch['aNameNonSearch'][$aSearchTerm['word_id']] = $aSearchTerm['word_id']; - } - $aSearch['iNamePhrase'] = $iPhrase; - if ($aSearch['iSearchRank'] < $this->iMaxRank) $aNewWordsetSearches[] = $aSearch; + $aNewSearches = $oCurrentSearch->extendWithPartialTerm( + $aSearchTerm, + $bIsStructured, + $iPhrase, + isset($aValidTokens[' '.$sToken]) ? $aValidTokens[' '.$sToken] : array() + ); + + foreach ($aNewSearches as $oSearch) { + if ($oSearch->getRank() < $this->iMaxRank) { + $aNewWordsetSearches[] = $oSearch; } } } - } else { - // Allow skipping a word - but at EXTREAM cost - //$aSearch = $aCurrentSearch; - //$aSearch['iSearchRank']+=100; - //$aNewWordsetSearches[] = $aSearch; } } // Sort and cut - usort($aNewWordsetSearches, 'bySearchRank'); + usort($aNewWordsetSearches, array('Nominatim\SearchDescription', 'bySearchRank')); $aWordsetSearches = array_slice($aNewWordsetSearches, 0, 50); } //var_Dump('
',sizeof($aWordsetSearches)); exit; $aNewPhraseSearches = array_merge($aNewPhraseSearches, $aNewWordsetSearches); - usort($aNewPhraseSearches, 'bySearchRank'); + usort($aNewPhraseSearches, array('Nominatim\SearchDescription', 'bySearchRank')); $aSearchHash = array(); foreach ($aNewPhraseSearches as $iSearch => $aSearch) { @@ -904,9 +769,12 @@ class Geocode // Re-group the searches by their score, junk anything over 20 as just not worth trying $aGroupedSearches = array(); foreach ($aNewPhraseSearches as $aSearch) { - if ($aSearch['iSearchRank'] < $this->iMaxRank) { - if (!isset($aGroupedSearches[$aSearch['iSearchRank']])) $aGroupedSearches[$aSearch['iSearchRank']] = array(); - $aGroupedSearches[$aSearch['iSearchRank']][] = $aSearch; + $iRank = $aSearch->getRank(); + if ($iRank < $this->iMaxRank) { + if (!isset($aGroupedSearches[$iRank])) { + $aGroupedSearches[$iRank] = array(); + } + $aGroupedSearches[$iRank][] = $aSearch; } } ksort($aGroupedSearches); @@ -924,19 +792,16 @@ class Geocode // Revisit searches, drop bad searches and give penalty to unlikely combinations. $aGroupedSearches = array(); - foreach ($aSearches as $aSearch) { - if (!$aSearch['aName']) { - if ($aSearch['sHouseNumber']) { - continue; - } - } - if ($this->aCountryCodes && $aSearch['sCountryCode'] - && !in_array($aSearch['sCountryCode'], $this->aCountryCodes)) { + foreach ($aSearches as $oSearch) { + if (!$oSearch->isValidSearch()) { continue; } - $aSearch['iSearchRank'] += $iGlobalRank; - $aGroupedSearches[$aSearch['iSearchRank']][] = $aSearch; + $iRank = $oSearch->addToRank($iGlobalRank); + if (!isset($aGroupedSearches[$iRank])) { + $aGroupedSearches[$iRank] = array(); + } + $aGroupedSearches[$iRank][] = $oSearch; } ksort($aGroupedSearches); @@ -979,13 +844,30 @@ class Geocode { if (!$this->sQuery && !$this->aStructuredQuery) return array(); - $sNormQuery = $this->normTerm($this->sQuery); - $sLanguagePrefArraySQL = "ARRAY[".join(',', array_map("getDBQuoted", $this->aLangPrefOrder))."]"; - $sCountryCodesSQL = false; + $oCtx = new SearchContext(); + + if ($this->aRoutePoints) { + $oCtx->setViewboxFromRoute( + $this->oDB, + $this->aRoutePoints, + $this->aRouteWidth, + $this->bBoundedSearch + ); + } elseif ($this->aViewBox) { + $oCtx->setViewboxFromBox($this->aViewBox, $this->bBoundedSearch); + } + if ($this->aExcludePlaceIDs) { + $oCtx->setExcludeList($this->aExcludePlaceIDs); + } if ($this->aCountryCodes) { - $sCountryCodesSQL = join(',', array_map('addQuotes', $this->aCountryCodes)); + $oCtx->setCountryList($this->aCountryCodes); } + $sNormQuery = $this->normTerm($this->sQuery); + $sLanguagePrefArraySQL = getArraySQL( + array_map("getDBQuoted", $this->aLangPrefOrder) + ); + $sQuery = $this->sQuery; if (!preg_match('//u', $sQuery)) { userError("Query string is not UTF-8 encoded."); @@ -998,37 +880,13 @@ class Geocode $sQuery = preg_replace('/(^|,)\s*la\s*(,|$)/', '\1louisiana\2', $sQuery); } - $bBoundingBoxSearch = $this->bBoundedSearch && $this->sViewboxSmallSQL; - if ($this->sViewboxCentreSQL) { - // For complex viewboxes (routes) precompute the bounding geometry - $sGeom = chksql( - $this->oDB->getOne("select ".$this->sViewboxSmallSQL), - "Could not get small viewbox" - ); - $this->sViewboxSmallSQL = "'".$sGeom."'::geometry"; - - $sGeom = chksql( - $this->oDB->getOne("select ".$this->sViewboxLargeSQL), - "Could not get large viewbox" - ); - $this->sViewboxLargeSQL = "'".$sGeom."'::geometry"; - } - // Do we have anything that looks like a lat/lon pair? - $oNearPoint = false; - if ($aLooksLike = NearPoint::extractFromQuery($sQuery)) { - $oNearPoint = $aLooksLike['pt']; - $sQuery = $aLooksLike['query']; - } + $sQuery = $oCtx->setNearPointFromQuery($sQuery); $aSearchResults = array(); if ($sQuery || $this->aStructuredQuery) { // Start with a single blank search - $aSearches = array(new SearchDescription()); - - if ($oNearPoint) { - $aSearches[0]->setNear($oNearPoint); - } + $aSearches = array(new SearchDescription($oCtx)); if ($sQuery) { $sQuery = $aSearches[0]->extractKeyValuePairs($sQuery); @@ -1073,7 +931,7 @@ class Geocode $oNewSearch->setPoiSearch( Operator::TYPE, $aSearchTerm['class'], - $aSearchTerm['type'], + $aSearchTerm['type'] ); $aNewSearches[] = $oNewSearch; } @@ -1084,10 +942,10 @@ class Geocode // Split query into phrases // Commas are used to reduce the search space by indicating where phrases split if ($this->aStructuredQuery) { - $aPhrases = $this->aStructuredQuery; + $aInPhrases = $this->aStructuredQuery; $bStructuredPhrases = true; } else { - $aPhrases = explode(',', $sQuery); + $aInPhrases = explode(',', $sQuery); $bStructuredPhrases = false; } @@ -1096,25 +954,19 @@ class Geocode // Get all 'sets' of words // Generate a complete list of all $aTokens = array(); - foreach ($aPhrases as $iPhrase => $sPhrase) { - $aPhrase = chksql( - $this->oDB->getRow("SELECT make_standard_name('".pg_escape_string($sPhrase)."') as string"), + $aPhrases = array(); + foreach ($aInPhrases as $iPhrase => $sPhrase) { + $sPhrase = chksql( + $this->oDB->getOne('SELECT make_standard_name('.getDBQuoted($sPhrase).')'), "Cannot normalize query string (is it a UTF-8 string?)" ); - if (trim($aPhrase['string'])) { - $aPhrases[$iPhrase] = $aPhrase; - $aPhrases[$iPhrase]['words'] = explode(' ', $aPhrases[$iPhrase]['string']); - $aPhrases[$iPhrase]['wordsets'] = getWordSets($aPhrases[$iPhrase]['words'], 0); - $aTokens = array_merge($aTokens, getTokensFromSets($aPhrases[$iPhrase]['wordsets'])); - } else { - unset($aPhrases[$iPhrase]); + if (trim($sPhrase)) { + $oPhrase = new Phrase($sPhrase, is_string($iPhrase) ? $iPhrase : ''); + $oPhrase->addTokens($aTokens); + $aPhrases[] = $oPhrase; } } - // Reindex phrases - we make assumptions later on that they are numerically keyed in order - $aPhraseTypes = array_keys($aPhrases); - $aPhrases = array_values($aPhrases); - if (sizeof($aTokens)) { // Check which tokens we have, get the ID numbers $sSQL = 'SELECT word_id, word_token, word, class, type, country_code, operator, search_name_count'; @@ -1128,22 +980,29 @@ class Geocode $this->oDB->getAll($sSQL), "Could not get word tokens." ); - $aPossibleMainWordIDs = array(); $aWordFrequencyScores = array(); foreach ($aDatabaseWords as $aToken) { - // Very special case - require 2 letter country param to match the country code found - if ($bStructuredPhrases && $aToken['country_code'] && !empty($this->aStructuredQuery['country']) - && strlen($this->aStructuredQuery['country']) == 2 && strtolower($this->aStructuredQuery['country']) != $aToken['country_code'] + // Filter country tokens that do not match restricted countries. + if ($this->aCountryCodes + && $aToken['country_code'] + && !in_array($aToken['country_code'], $this->aCountryCodes) ) { continue; } + // Special terms need to appear in their normalized form. + if ($aToken['word'] && $aToken['class']) { + $sNormWord = $this->normTerm($aToken['word']); + if (strpos($sNormQuery, $sNormWord) === false) { + continue; + } + } + if (isset($aValidTokens[$aToken['word_token']])) { $aValidTokens[$aToken['word_token']][] = $aToken; } else { $aValidTokens[$aToken['word_token']] = array($aToken); } - if (!$aToken['class'] && !$aToken['country_code']) $aPossibleMainWordIDs[$aToken['word_id']] = 1; $aWordFrequencyScores[$aToken['word_id']] = $aToken['search_name_count'] + 1; } if (CONST_Debug) var_Dump($aPhrases, $aValidTokens); @@ -1167,31 +1026,26 @@ class Geocode foreach ($aTokens as $sToken) { // Unknown single word token with a number - assume it is a house number - if (!isset($aValidTokens[' '.$sToken]) && strpos($sToken, ' ') === false && preg_match('/[0-9]/', $sToken)) { - $aValidTokens[' '.$sToken] = array(array('class' => 'place', 'type' => 'house')); + if (!isset($aValidTokens[' '.$sToken]) && strpos($sToken, ' ') === false && preg_match('/^[0-9]+$/', $sToken)) { + $aValidTokens[' '.$sToken] = array(array('class' => 'place', 'type' => 'house', 'word_token' => ' '.$sToken)); } } // Any words that have failed completely? // TODO: suggestions - // Start the search process - // array with: placeid => -1 | tiger-housenumber - $aResultPlaceIDs = array(); - - $aGroupedSearches = $this->getGroupedSearches($aSearches, $aPhraseTypes, $aPhrases, $aValidTokens, $aWordFrequencyScores, $bStructuredPhrases, $sNormQuery); + $aGroupedSearches = $this->getGroupedSearches($aSearches, $aPhrases, $aValidTokens, $bStructuredPhrases); if ($this->bReverseInPlan) { // Reverse phrase array and also reverse the order of the wordsets in // the first and final phrase. Don't bother about phrases in the middle // because order in the address doesn't matter. $aPhrases = array_reverse($aPhrases); - $aPhrases[0]['wordsets'] = getInverseWordSets($aPhrases[0]['words'], 0); + $aPhrases[0]->invertWordSets(); if (sizeof($aPhrases) > 1) { - $aFinalPhrase = end($aPhrases); - $aPhrases[sizeof($aPhrases)-1]['wordsets'] = getInverseWordSets($aFinalPhrase['words'], 0); + $aPhrases[sizeof($aPhrases)-1]->invertWordSets(); } - $aReverseGroupedSearches = $this->getGroupedSearches($aSearches, null, $aPhrases, $aValidTokens, $aWordFrequencyScores, false, $sNormQuery); + $aReverseGroupedSearches = $this->getGroupedSearches($aSearches, $aPhrases, $aValidTokens, false); foreach ($aGroupedSearches as $aSearches) { foreach ($aSearches as $aSearch) { @@ -1233,124 +1087,38 @@ class Geocode if (CONST_Debug) _debugDumpGroupedSearches($aGroupedSearches, $aValidTokens); + // Start the search process + // array with: placeid => -1 | tiger-housenumber + $aResultPlaceIDs = array(); $iGroupLoop = 0; $iQueryLoop = 0; foreach ($aGroupedSearches as $iGroupedRank => $aSearches) { $iGroupLoop++; foreach ($aSearches as $oSearch) { $iQueryLoop++; - $searchedHousenumber = -1; - - if (CONST_Debug) echo "
Search Loop, group $iGroupLoop, loop $iQueryLoop"; - if (CONST_Debug) _debugDumpGroupedSearches(array($iGroupedRank => array($oSearch)), $aValidTokens); - - $aPlaceIDs = array(); - if ($oSearch->isCountrySearch()) { - // Just looking for a country - look it up - if (4 >= $this->iMinAddressRank && 4 <= $this->iMaxAddressRank) { - $aPlaceIDs = $oSearch->queryCountry( - $this->oDB, - $bBoundingBoxSearch ? $this->sViewboxSmallSQL : '' - ); - } - } elseif (!$oSearch->isNamedSearch()) { - // looking for a POI in a geographic area - if (!$bBoundingBoxSearch && !$oSearch->isNearSearch()) { - continue; - } - - $aPlaceIDs = $oSearch->queryNearbyPoi( - $this->oDB, - $sCountryCodesSQL, - $bBoundingBoxSearch ? $this->sViewboxSmallSQL : '', - $sViewboxCentreSQL, - $this->aExcludePlaceIDs ? join(',', $this->aExcludePlaceIDs) : '', - $this->iLimit - ); - } elseif ($oSearch->isOperator(Operator::POSTCODE)) { - $aPlaceIDs = $oSearch->queryPostcode( - $oDB, - $sCountryCodesSQL, - $this->iLimit - ); - } else { - // Ordinary search: - // First search for places according to name and address. - $aNamedPlaceIDs = $oSearch->queryNamedPlace( - $this->oDB, - $aWordFrequencyScores, - $sCountryCodesSQL, - $this->iMinAddressRank, - $this->iMaxAddressRank, - $this->aExcludePlaceIDs ? join(',', $this->aExcludePlaceIDs) : '', - $bBoundingBoxSearch ? $this->sViewboxSmallSQL : '', - $bBoundingBoxSearch ? $this->sViewboxLargeSQL : '', - $this->iLimit - ); - - if (sizeof($aNamedPlaceIDs)) { - foreach ($aNamedPlaceIDs as $aRow) { - $aPlaceIDs[] = $aRow['place_id']; - $this->exactMatchCache[$aRow['place_id']] = $aRow['exactmatch']; - } - } - - //now search for housenumber, if housenumber provided - if ($oSearch->hasHouseNumber() && sizeof($aPlaceIDs)) { - $aResult = $oSearch->queryHouseNumber( - $this->oDB, - $aPlaceIDs, - $this->aExcludePlaceIDs ? join(',', $this->aExcludePlaceIDs) : '' - $this->iLimit - ); - - if (sizeof($aResult)) { - $searchedHousenumber = $aResult['iHouseNumber']; - $aPlaceIDs = $aResults['aPlaceIDs']; - } elseif (!$oSearch->looksLikeFullAddress()) { - $aPlaceIDs = array(); - } - } - - // finally get POIs if requested - if ($oSearch->isPoiSearch() && sizeof($aPlaceIDs)) { - $aPlaceIDs = $oSearch->queryPoiByOperator( - $this->oDB, - $aPlaceIDs, - $this->aExcludePlaceIDs ? join(',', $this->aExcludePlaceIDs) : '' - $this->iLimit - ); - } - } if (CONST_Debug) { - echo "
Place IDs: "; - var_Dump($aPlaceIDs); + echo "
Search Loop, group $iGroupLoop, loop $iQueryLoop"; + _debugDumpGroupedSearches(array($iGroupedRank => array($oSearch)), $aValidTokens); } - if (sizeof($aPlaceIDs) && $oSearch->getPostcode()) { - $sSQL = 'SELECT place_id FROM placex'; - $sSQL .= ' WHERE place_id in ('.join(',', $aPlaceIDs).')'; - $sSQL .= " AND postcode = '".$oSearch->getPostcode()."'"; - if (CONST_Debug) var_dump($sSQL); - $aFilteredPlaceIDs = chksql($this->oDB->getCol($sSQL)); - if ($aFilteredPlaceIDs) { - $aPlaceIDs = $aFilteredPlaceIDs; - if (CONST_Debug) { - echo "
Place IDs after postcode filtering: "; - var_Dump($aPlaceIDs); - } - } - } + $aRes = $oSearch->query( + $this->oDB, + $aWordFrequencyScores, + $this->exactMatchCache, + $this->iMinAddressRank, + $this->iMaxAddressRank, + $this->iLimit + ); - foreach ($aPlaceIDs as $iPlaceID) { + foreach ($aRes['IDs'] as $iPlaceID) { // array for placeID => -1 | Tiger housenumber - $aResultPlaceIDs[$iPlaceID] = $searchedHousenumber; + $aResultPlaceIDs[$iPlaceID] = $aRes['houseNumber']; } if ($iQueryLoop > 20) break; } - if (isset($aResultPlaceIDs) && sizeof($aResultPlaceIDs) && ($this->iMinAddressRank != 0 || $this->iMaxAddressRank != 30)) { + if (sizeof($aResultPlaceIDs) && ($this->iMinAddressRank != 0 || $this->iMaxAddressRank != 30)) { // Need to verify passes rank limits before dropping out of the loop (yuk!) // reduces the number of place ids, like a filter // rank_address is 30 for interpolated housenumbers @@ -1393,31 +1161,26 @@ class Geocode $aResultPlaceIDs = $tempIDs; } - //exit; - if (isset($aResultPlaceIDs) && sizeof($aResultPlaceIDs)) break; + if (sizeof($aResultPlaceIDs)) break; if ($iGroupLoop > 4) break; if ($iQueryLoop > 30) break; } // Did we find anything? - if (isset($aResultPlaceIDs) && sizeof($aResultPlaceIDs)) { - $aSearchResults = $this->getDetails($aResultPlaceIDs); + if (sizeof($aResultPlaceIDs)) { + $aSearchResults = $this->getDetails($aResultPlaceIDs, $oCtx); } } else { // Just interpret as a reverse geocode $oReverse = new ReverseGeocode($this->oDB); $oReverse->setZoom(18); - $aLookup = $oReverse->lookup( - $oNearPoint->lat(), - $oNearPoint->lon(), - false - ); + $aLookup = $oReverse->lookupPoint($oCtx->sqlNear, false); if (CONST_Debug) var_dump("Reverse search", $aLookup); if ($aLookup['place_id']) { - $aSearchResults = $this->getDetails(array($aLookup['place_id'] => -1)); + $aSearchResults = $this->getDetails(array($aLookup['place_id'] => -1), $oCtx); $aResultPlaceIDs[$aLookup['place_id']] = -1; } else { $aSearchResults = array(); @@ -1510,36 +1273,43 @@ class Geocode } } - // Adjust importance for the number of exact string matches in the result - $aResult['importance'] = max(0.001, $aResult['importance']); - $iCountWords = 0; - $sAddress = $aResult['langaddress']; - foreach ($aRecheckWords as $i => $sWord) { - if (stripos($sAddress, $sWord)!==false) { - $iCountWords++; - if (preg_match("/(^|,)\s*".preg_quote($sWord, '/')."\s*(,|$)/", $sAddress)) $iCountWords += 0.1; + $aResult['name'] = $aResult['langaddress']; + + if ($oCtx->hasNearPoint()) + { + $aResult['importance'] = 0.001; + $aResult['foundorder'] = $aResult['addressimportance']; + } else { + // Adjust importance for the number of exact string matches in the result + $aResult['importance'] = max(0.001, $aResult['importance']); + $iCountWords = 0; + $sAddress = $aResult['langaddress']; + foreach ($aRecheckWords as $i => $sWord) { + if (stripos($sAddress, $sWord)!==false) { + $iCountWords++; + if (preg_match("/(^|,)\s*".preg_quote($sWord, '/')."\s*(,|$)/", $sAddress)) $iCountWords += 0.1; + } } - } - $aResult['importance'] = $aResult['importance'] + ($iCountWords*0.1); // 0.1 is a completely arbitrary number but something in the range 0.1 to 0.5 would seem right + $aResult['importance'] = $aResult['importance'] + ($iCountWords*0.1); // 0.1 is a completely arbitrary number but something in the range 0.1 to 0.5 would seem right - $aResult['name'] = $aResult['langaddress']; - // secondary ordering (for results with same importance (the smaller the better): - // - approximate importance of address parts - $aResult['foundorder'] = -$aResult['addressimportance']/10; - // - number of exact matches from the query - if (isset($this->exactMatchCache[$aResult['place_id']])) { - $aResult['foundorder'] -= $this->exactMatchCache[$aResult['place_id']]; - } elseif (isset($this->exactMatchCache[$aResult['parent_place_id']])) { - $aResult['foundorder'] -= $this->exactMatchCache[$aResult['parent_place_id']]; - } - // - importance of the class/type - if (isset($aClassType[$aResult['class'].':'.$aResult['type']]['importance']) - && $aClassType[$aResult['class'].':'.$aResult['type']]['importance'] - ) { - $aResult['foundorder'] += 0.0001 * $aClassType[$aResult['class'].':'.$aResult['type']]['importance']; - } else { - $aResult['foundorder'] += 0.01; + // secondary ordering (for results with same importance (the smaller the better): + // - approximate importance of address parts + $aResult['foundorder'] = -$aResult['addressimportance']/10; + // - number of exact matches from the query + if (isset($this->exactMatchCache[$aResult['place_id']])) { + $aResult['foundorder'] -= $this->exactMatchCache[$aResult['place_id']]; + } elseif (isset($this->exactMatchCache[$aResult['parent_place_id']])) { + $aResult['foundorder'] -= $this->exactMatchCache[$aResult['parent_place_id']]; + } + // - importance of the class/type + if (isset($aClassType[$aResult['class'].':'.$aResult['type']]['importance']) + && $aClassType[$aResult['class'].':'.$aResult['type']]['importance'] + ) { + $aResult['foundorder'] += 0.0001 * $aClassType[$aResult['class'].':'.$aResult['type']]['importance']; + } else { + $aResult['foundorder'] += 0.01; + } } if (CONST_Debug) var_dump($aResult); $aSearchResults[$iResNum] = $aResult;