X-Git-Url: https://git.openstreetmap.org./nominatim.git/blobdiff_plain/30511fd3ab9d731b0afec202386fc2425a0b1b9f..9ef2370a2a6e6383b2e166896bfa379e5faff485:/lib/Geocode.php diff --git a/lib/Geocode.php b/lib/Geocode.php index a6baa96b..9115be0b 100644 --- a/lib/Geocode.php +++ b/lib/Geocode.php @@ -3,6 +3,7 @@ namespace Nominatim; 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'); @@ -37,9 +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; @@ -184,26 +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); @@ -218,29 +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) @@ -319,7 +276,8 @@ class Geocode $aRoute = $oParams->getStringList('route'); $fRouteWidth = $oParams->getFloat('routewidth'); if ($aRoute && $fRouteWidth) { - $this->setRoute($aRoute, $fRouteWidth); + $this->aRoutePoints = $aRoute; + $this->aRouteWidth = $fRouteWidth; } } } @@ -407,7 +365,7 @@ class Geocode return false; } - public function getDetails($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(); @@ -419,16 +377,8 @@ class Geocode // 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 "; $sSQL .= " osm_type,"; @@ -448,17 +398,21 @@ class Geocode if ($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) "; @@ -507,17 +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) "; @@ -555,17 +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 ("; $sSQL .= " SELECT place_id, "; // interpolate the Tiger housenumbers here @@ -605,18 +567,22 @@ class Geocode if ($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 ("; $sSQL .= " SELECT "; @@ -664,17 +630,21 @@ class Geocode if ($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) "; @@ -699,7 +669,7 @@ class Geocode return $aSearchResults; } - public function getGroupedSearches($aSearches, $aPhraseTypes, $aPhrases, $aValidTokens, $aWordFrequencyScores, $bStructuredPhrases, $sNormQuery) + public function getGroupedSearches($aSearches, $aPhrases, $aValidTokens, $bIsStructured, $sNormQuery) { /* Calculate all searches using aValidTokens i.e. @@ -714,15 +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; @@ -777,9 +743,8 @@ class Geocode foreach ($aValidTokens[$sToken] as $aSearchTerm) { $aNewSearches = $oCurrentSearch->extendWithPartialTerm( $aSearchTerm, - $bStructuredPhrases, + $bIsStructured, $iPhrase, - $aWordFrequencyScores, isset($aValidTokens[' '.$sToken]) ? $aValidTokens[' '.$sToken] : array() ); @@ -890,14 +855,27 @@ class Geocode $oCtx = new SearchContext(); + if ($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) { + $oCtx->setCountryList($this->aCountryCodes); + } + $sNormQuery = $this->normTerm($this->sQuery); $sLanguagePrefArraySQL = getArraySQL( array_map("getDBQuoted", $this->aLangPrefOrder) ); - $sCountryCodesSQL = false; - if ($this->aCountryCodes) { - $sCountryCodesSQL = join(',', array_map('addQuotes', $this->aCountryCodes)); - } $sQuery = $this->sQuery; if (!preg_match('//u', $sQuery)) { @@ -911,22 +889,6 @@ 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? $sQuery = $oCtx->setNearPointFromQuery($sQuery); @@ -989,10 +951,10 @@ 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; } @@ -1001,25 +963,19 @@ 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'; @@ -1033,7 +989,6 @@ 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 @@ -1048,7 +1003,6 @@ class Geocode } 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); @@ -1080,23 +1034,18 @@ class Geocode // 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, $sNormQuery); if ($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, $sNormQuery); foreach ($aGroupedSearches as $aSearches) { foreach ($aSearches as $aSearch) { @@ -1138,124 +1087,38 @@ class Geocode if (CONST_Debug) _debugDumpGroupedSearches($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 "