X-Git-Url: https://git.openstreetmap.org./nominatim.git/blobdiff_plain/b07185c93450d097c6e3428f3c32475c69f21313..c32fa2172d7d688e3b816957cd1e92f6ca129244:/sql/functions.sql diff --git a/sql/functions.sql b/sql/functions.sql index 3029d015..b4657bf8 100644 --- a/sql/functions.sql +++ b/sql/functions.sql @@ -389,7 +389,7 @@ BEGIN result := '{}'::INTEGER[]; s := make_standard_name(src); - w := getorcreate_name_id(s); + w := getorcreate_name_id(s, src); IF NOT (ARRAY[w] <@ result) THEN result := result || w; @@ -407,6 +407,30 @@ BEGIN END LOOP; END IF; + words := regexp_split_to_array(src, E'[,;()]'); + IF array_upper(words, 1) != 1 THEN + FOR j IN 1..array_upper(words, 1) LOOP + s := make_standard_name(words[j]); + IF s != '' THEN + w := getorcreate_word_id(s); + IF NOT (ARRAY[w] <@ result) THEN + result := result || w; + END IF; + END IF; + END LOOP; + END IF; + + s := regexp_replace(src, '市$', ''); + IF s != src THEN + s := make_standard_name(s); + IF s != '' THEN + w := getorcreate_name_id(s, src); + IF NOT (ARRAY[w] <@ result) THEN + result := result || w; + END IF; + END IF; + END IF; + RETURN result; END; $$ @@ -742,7 +766,7 @@ $$ LANGUAGE plpgsql; -CREATE OR REPLACE FUNCTION create_interpolation(wayid INTEGER, interpolationtype TEXT) RETURNS INTEGER +CREATE OR REPLACE FUNCTION create_interpolation(wayid BIGINT, interpolationtype TEXT) RETURNS INTEGER AS $$ DECLARE @@ -793,7 +817,7 @@ BEGIN linestr := linestr||','||ST_X(nextnode.geometry)||' '||ST_Y(nextnode.geometry); endnumber := ('0'||substring(nextnode.housenumber,'[0-9]+'))::integer; - IF startnumber IS NOT NULL and startnumber > 0 AND endnumber IS NOT NULL and endnumber > 0 THEN + IF startnumber IS NOT NULL and startnumber > 0 AND endnumber IS NOT NULL and endnumber > 0 AND @(startnumber - endnumber) < 1000 THEN --RAISE WARNING 'interpolation end % % ',nextnode.place_id,endnumber; @@ -875,6 +899,7 @@ DECLARE country_code VARCHAR(2); default_language VARCHAR(10); diameter FLOAT; + classtable TEXT; BEGIN -- RAISE WARNING '%',NEW.osm_id; @@ -908,18 +933,23 @@ BEGIN NEW.place_id := nextval('seq_place'); NEW.indexed_status := 1; --STATUS_NEW - NEW.country_code := lower(get_country_code(NEW.geometry, NEW.country_code)); + IF NEW.rank_search >= 4 THEN + NEW.country_code := lower(get_country_code(NEW.geometry, NEW.country_code)); + ELSE + NEW.country_code := NULL; + END IF; + NEW.partition := get_partition(NEW.geometry, NEW.country_code); NEW.geometry_sector := geometry_sector(NEW.partition, NEW.geometry); -- copy 'name' to or from the default language (if there is a default language) - IF NEW.name is not null AND array_upper(%#NEW.name,1) > 1 THEN + IF NEW.name is not null AND array_upper(akeys(NEW.name),1) > 1 THEN default_language := get_country_language_code(NEW.country_code); IF default_language IS NOT NULL THEN IF NEW.name ? 'name' AND NOT NEW.name ? ('name:'||default_language) THEN NEW.name := NEW.name || (('name:'||default_language) => (NEW.name -> 'name')); ELSEIF NEW.name ? ('name:'||default_language) AND NOT NEW.name ? 'name' THEN - NEW.name := NEW.name || ('name' => (NEW.name -> 'name:'||default_language)); + NEW.name := NEW.name || ('name' => (NEW.name -> ('name:'||default_language))); END IF; END IF; END IF; @@ -939,7 +969,55 @@ BEGIN NEW.rank_address := NEW.rank_search; -- By doing in postgres we have the country available to us - currently only used for postcode - IF NEW.class = 'place' THEN + IF NEW.class in ('place','boundary') AND NEW.type in ('postcode','postal_code') THEN + + IF NEW.postcode IS NULL THEN + -- most likely just a part of a multipolygon postcode boundary, throw it away + RETURN NULL; + END IF; + + NEW.name := 'ref'=>NEW.postcode; + + IF NEW.country_code = 'gb' THEN + + IF NEW.postcode ~ '^([A-Z][A-Z]?[0-9][0-9A-Z]? [0-9][A-Z][A-Z])$' THEN + NEW.rank_search := 25; + NEW.rank_address := 5; + ELSEIF NEW.postcode ~ '^([A-Z][A-Z]?[0-9][0-9A-Z]? [0-9])$' THEN + NEW.rank_search := 23; + NEW.rank_address := 5; + ELSEIF NEW.postcode ~ '^([A-Z][A-Z]?[0-9][0-9A-Z])$' THEN + NEW.rank_search := 21; + NEW.rank_address := 5; + END IF; + + ELSEIF NEW.country_code = 'de' THEN + + IF NEW.postcode ~ '^([0-9]{5})$' THEN + NEW.rank_search := 21; + NEW.rank_address := 11; + END IF; + + ELSE + -- Guess at the postcode format and coverage (!) + IF upper(NEW.postcode) ~ '^[A-Z0-9]{1,5}$' THEN -- Probably too short to be very local + NEW.rank_search := 21; + NEW.rank_address := 11; + ELSE + -- Does it look splitable into and area and local code? + postcode := substring(upper(NEW.postcode) from '^([- :A-Z0-9]+)([- :][A-Z0-9]+)$'); + + IF postcode IS NOT NULL THEN + NEW.rank_search := 25; + NEW.rank_address := 11; + ELSEIF NEW.postcode ~ '^[- :A-Z0-9]{6,}$' THEN + NEW.rank_search := 21; + NEW.rank_address := 11; + END IF; + END IF; + END IF; + + ELSEIF NEW.class = 'place' THEN IF NEW.type in ('continent') THEN NEW.rank_search := 2; NEW.rank_address := NEW.rank_search; @@ -953,8 +1031,8 @@ BEGIN NEW.rank_search := 8; NEW.rank_address := NEW.rank_search; ELSEIF NEW.type in ('region') THEN - NEW.rank_search := 10; - NEW.rank_address := NEW.rank_search; + NEW.rank_search := 18; -- dropped from previous value of 10 + NEW.rank_address := 0; -- So badly miss-used that better to just drop it! ELSEIF NEW.type in ('county') THEN NEW.rank_search := 12; NEW.rank_address := NEW.rank_search; @@ -985,55 +1063,16 @@ BEGIN ELSEIF NEW.type in ('suburb','croft','subdivision') THEN NEW.rank_search := 20; NEW.rank_address := NEW.rank_search; - ELSEIF NEW.type in ('farm','locality','islet','isolated_dwelling') THEN + ELSEIF NEW.type in ('farm','locality','islet','isolated_dwelling','mountain_pass') THEN NEW.rank_search := 20; NEW.rank_address := 0; + -- Irish townlands, tagged as place=locality and locality=townland + IF (NEW.extratags -> 'locality') = 'townland' THEN + NEW.rank_address := 20; + END IF; ELSEIF NEW.type in ('hall_of_residence','neighbourhood','housing_estate','nature_reserve') THEN NEW.rank_search := 22; NEW.rank_address := 22; - ELSEIF NEW.type in ('postcode') THEN - - NEW.name := 'ref'=>NEW.postcode; - - IF NEW.country_code = 'gb' THEN - - IF NEW.postcode ~ '^([A-Z][A-Z]?[0-9][0-9A-Z]? [0-9][A-Z][A-Z])$' THEN - NEW.rank_search := 25; - NEW.rank_address := 5; - ELSEIF NEW.postcode ~ '^([A-Z][A-Z]?[0-9][0-9A-Z]? [0-9])$' THEN - NEW.rank_search := 23; - NEW.rank_address := 5; - ELSEIF NEW.postcode ~ '^([A-Z][A-Z]?[0-9][0-9A-Z])$' THEN - NEW.rank_search := 21; - NEW.rank_address := 5; - END IF; - - ELSEIF NEW.country_code = 'de' THEN - - IF NEW.postcode ~ '^([0-9]{5})$' THEN - NEW.rank_search := 21; - NEW.rank_address := 11; - END IF; - - ELSE - -- Guess at the postcode format and coverage (!) - IF upper(NEW.postcode) ~ '^[A-Z0-9]{1,5}$' THEN -- Probably too short to be very local - NEW.rank_search := 21; - NEW.rank_address := 11; - ELSE - -- Does it look splitable into and area and local code? - postcode := substring(upper(NEW.postcode) from '^([- :A-Z0-9]+)([- :][A-Z0-9]+)$'); - - IF postcode IS NOT NULL THEN - NEW.rank_search := 25; - NEW.rank_address := 11; - ELSEIF NEW.postcode ~ '^[- :A-Z0-9]{6,}$' THEN - NEW.rank_search := 21; - NEW.rank_address := 11; - END IF; - END IF; - END IF; - ELSEIF NEW.type in ('airport','street') THEN NEW.rank_search := 26; NEW.rank_address := NEW.rank_search; @@ -1114,11 +1153,12 @@ BEGIN IF st_area(NEW.geometry) < 1 THEN -- mark items within the geometry for re-indexing -- RAISE WARNING 'placex poly insert: % % % %',NEW.osm_type,NEW.osm_id,NEW.class,NEW.type; --- work around bug in postgis + + -- work around bug in postgis, this may have been fixed in 2.0.0 (see http://trac.osgeo.org/postgis/ticket/547) update placex set indexed_status = 2 where (ST_Contains(NEW.geometry, placex.geometry) OR ST_Intersects(NEW.geometry, placex.geometry)) - AND rank_search > NEW.rank_search and indexed_status = 0 and ST_geometrytype(placex.geometry) = 'ST_Point'; + AND rank_search > NEW.rank_search and indexed_status = 0 and ST_geometrytype(placex.geometry) = 'ST_Point' and (rank_search < 28 or name is not null); update placex set indexed_status = 2 where (ST_Contains(NEW.geometry, placex.geometry) OR ST_Intersects(NEW.geometry, placex.geometry)) - AND rank_search > NEW.rank_search and indexed_status = 0 and ST_geometrytype(placex.geometry) != 'ST_Point'; + AND rank_search > NEW.rank_search and indexed_status = 0 and ST_geometrytype(placex.geometry) != 'ST_Point' and (rank_search < 28 or name is not null); END IF; ELSE -- mark nearby items for re-indexing, where 'nearby' depends on the features rank_search and is a complete guess :( @@ -1143,11 +1183,22 @@ BEGIN END IF; IF diameter > 0 THEN -- RAISE WARNING 'placex point insert: % % % % %',NEW.osm_type,NEW.osm_id,NEW.class,NEW.type,diameter; - update placex set indexed_status = 2 where indexed_status = 0 and rank_search > NEW.rank_search and ST_DWithin(placex.geometry, NEW.geometry, diameter); + update placex set indexed_status = 2 where indexed_status = 0 and rank_search > NEW.rank_search and ST_DWithin(placex.geometry, NEW.geometry, diameter) and (rank_search < 28 or name is not null); END IF; END IF; + -- add to tables for special search + -- Note: won't work on initial import because the classtype tables + -- do not yet exist. It won't hurt either. + classtable := 'place_classtype_' || NEW.class || '_' || NEW.type; + SELECT count(*)>0 FROM pg_tables WHERE tablename = classtable INTO result; + IF result THEN + EXECUTE 'INSERT INTO ' || classtable::regclass || ' (place_id, centroid) VALUES ($1,$2)' + USING NEW.place_id, ST_Centroid(NEW.geometry); + END IF; + + -- IF NEW.rank_search < 26 THEN -- RAISE WARNING 'placex insert: % % % %',NEW.osm_type,NEW.osm_id,NEW.class,NEW.type; -- END IF; @@ -1176,6 +1227,9 @@ DECLARE location RECORD; way RECORD; relation RECORD; + relation_members TEXT[]; + relMember RECORD; + linkedplacex RECORD; search_diameter FLOAT; search_prevdiameter FLOAT; search_maxrank INTEGER; @@ -1191,9 +1245,12 @@ DECLARE tagpairid INTEGER; + default_language TEXT; name_vector INTEGER[]; nameaddress_vector INTEGER[]; + linked_node_id BIGINT; + result BOOLEAN; BEGIN @@ -1205,9 +1262,9 @@ BEGIN RETURN NEW; END IF; - -- defered delete + -- deferred delete IF OLD.indexed_status = 100 THEN - delete from placex where osm_type = OLD.osm_type and osm_id = OLD.osm_id and class = OLD.class and type = OLD.type; + delete from placex where place_id = OLD.place_id; RETURN NULL; END IF; @@ -1220,15 +1277,22 @@ BEGIN RETURN NEW; END IF; - result := deleteSearchName(NEW.partition, NEW.place_id); - DELETE FROM place_addressline WHERE place_id = NEW.place_id; - DELETE FROM place_boundingbox where place_id = NEW.place_id; - result := deleteRoad(NEW.partition, NEW.place_id); - result := deleteLocationArea(NEW.partition, NEW.place_id); + IF OLD.indexed_status > 1 THEN + result := deleteSearchName(NEW.partition, NEW.place_id); + DELETE FROM place_addressline WHERE place_id = NEW.place_id; + DELETE FROM place_boundingbox where place_id = NEW.place_id; + result := deleteRoad(NEW.partition, NEW.place_id); + result := deleteLocationArea(NEW.partition, NEW.place_id); + UPDATE placex set linked_place_id = null where linked_place_id = NEW.place_id; + END IF; -- reclaculate country and partition (should probably have a country_code and calculated_country_code as seperate fields) - SELECT country_code from place where osm_type = NEW.osm_type and osm_id = NEW.osm_id and class = NEW.class and type = NEW.type INTO NEW.country_code; - NEW.country_code := lower(get_country_code(NEW.geometry, NEW.country_code)); + IF NEW.rank_search >= 4 THEN + SELECT country_code from place where osm_type = NEW.osm_type and osm_id = NEW.osm_id and class = NEW.class and type = NEW.type INTO NEW.country_code; + NEW.country_code := lower(get_country_code(NEW.geometry, NEW.country_code)); + ELSE + NEW.country_code := NULL; + END IF; NEW.partition := get_partition(NEW.geometry, NEW.country_code); NEW.geometry_sector := geometry_sector(NEW.partition, NEW.geometry); @@ -1241,6 +1305,20 @@ BEGIN -- Speed up searches - just use the centroid of the feature -- cheaper but less acurate place_centroid := ST_Centroid(NEW.geometry); + NEW.centroid := null; + + -- Thought this wasn't needed but when we add new languages to the country_name table + -- we need to update the existing names + IF NEW.name is not null AND array_upper(akeys(NEW.name),1) > 1 THEN + default_language := get_country_language_code(NEW.country_code); + IF default_language IS NOT NULL THEN + IF NEW.name ? 'name' AND NOT NEW.name ? ('name:'||default_language) THEN + NEW.name := NEW.name || (('name:'||default_language) => (NEW.name -> 'name')); + ELSEIF NEW.name ? ('name:'||default_language) AND NOT NEW.name ? 'name' THEN + NEW.name := NEW.name || ('name' => (NEW.name -> ('name:'||default_language))); + END IF; + END IF; + END IF; -- Initialise the name vector using our name name_vector := make_keywords(NEW.name); @@ -1256,6 +1334,12 @@ BEGIN address_havelevel[i] := false; END LOOP; + NEW.importance := null; + select language||':'||title,importance from get_wikipedia_match(NEW.extratags) INTO NEW.wikipedia,NEW.importance; + IF NEW.importance IS NULL THEN + select language||':'||title,importance from wikipedia_article where osm_type = NEW.osm_type and osm_id = NEW.osm_id order by importance desc limit 1 INTO NEW.wikipedia,NEW.importance; + END IF; + --RAISE WARNING '% %', NEW.place_id, NEW.rank_search; -- For low level elements we inherit from our parent road @@ -1271,7 +1355,7 @@ BEGIN IF NEW.parent_place_id IS NULL AND NEW.osm_type = 'N' THEN -- Is this node part of a relation? - FOR relation IN select * from planet_osm_rels where parts @> ARRAY[NEW.osm_id::integer] and members @> ARRAY['n'||NEW.osm_id] + FOR relation IN select * from planet_osm_rels where parts @> ARRAY[NEW.osm_id] and members @> ARRAY['n'||NEW.osm_id] LOOP -- At the moment we only process one type of relation - associatedStreet IF relation.tags @> ARRAY['associatedStreet'] AND array_upper(relation.members, 1) IS NOT NULL THEN @@ -1287,7 +1371,7 @@ BEGIN --RAISE WARNING 'x1'; -- Is this node part of a way? - FOR way IN select id from planet_osm_ways where nodes @> ARRAY[NEW.osm_id::integer] LOOP + FOR way IN select id from planet_osm_ways where nodes @> ARRAY[NEW.osm_id] LOOP --RAISE WARNING '%', way; FOR location IN select * from placex where osm_type = 'W' and osm_id = way.id LOOP @@ -1300,7 +1384,7 @@ BEGIN -- Is the WAY part of a relation IF NEW.parent_place_id IS NULL THEN - FOR relation IN select * from planet_osm_rels where parts @> ARRAY[location.osm_id::integer] and members @> ARRAY['w'||location.osm_id] + FOR relation IN select * from planet_osm_rels where parts @> ARRAY[location.osm_id] and members @> ARRAY['w'||location.osm_id] LOOP -- At the moment we only process one type of relation - associatedStreet IF relation.tags @> ARRAY['associatedStreet'] AND array_upper(relation.members, 1) IS NOT NULL THEN @@ -1345,7 +1429,7 @@ BEGIN IF NEW.parent_place_id IS NULL AND NEW.osm_type = 'W' THEN -- Is this way part of a relation? - FOR relation IN select * from planet_osm_rels where parts @> ARRAY[NEW.osm_id::integer] and members @> ARRAY['w'||NEW.osm_id] + FOR relation IN select * from planet_osm_rels where parts @> ARRAY[NEW.osm_id] and members @> ARRAY['w'||NEW.osm_id] LOOP -- At the moment we only process one type of relation - associatedStreet IF relation.tags @> ARRAY['associatedStreet'] AND array_upper(relation.members, 1) IS NOT NULL THEN @@ -1423,6 +1507,114 @@ BEGIN -- RAISE WARNING ' INDEXING: %',NEW; + IF NEW.osm_type = 'R' AND NEW.rank_search < 26 THEN + + -- see if we have any special relation members + select members from planet_osm_rels where id = NEW.osm_id INTO relation_members; + + FOR relMember IN select get_osm_rel_members(relation_members,ARRAY['label']) as member LOOP + + select * from placex where osm_type = upper(substring(relMember.member,1,1)) + and osm_id = substring(relMember.member,2,10000)::integer order by rank_search desc limit 1 into linkedPlacex; + + -- If we don't already have one use this as the centre point of the geometry + IF NEW.centroid IS NULL THEN + NEW.centroid := coalesce(linkedPlacex.centroid,st_centroid(linkedPlacex.geometry)); + END IF; + + -- merge in the label name, re-init word vector + NEW.name := linkedPlacex.name || NEW.name; + name_vector := make_keywords(NEW.name); + + -- merge in extra tags + NEW.extratags := linkedPlacex.extratags || NEW.extratags; + + -- mark the linked place (excludes from search results) + UPDATE placex set linked_place_id = NEW.place_id where place_id = linkedPlacex.place_id; + + END LOOP; + + IF NEW.centroid IS NULL THEN + + FOR relMember IN select get_osm_rel_members(relation_members,ARRAY['admin_center','admin_centre']) as member LOOP + + select * from placex where osm_type = upper(substring(relMember.member,1,1)) + and osm_id = substring(relMember.member,2,10000)::integer order by rank_search desc limit 1 into linkedPlacex; + + -- For an admin centre we also want a name match - still not perfect, for example 'new york, new york' + -- But that can be fixed by explicitly setting the label in the data + IF make_standard_name(NEW.name->'name') = make_standard_name(linkedPlacex.name->'name') + AND NEW.rank_search = linkedPlacex.rank_search THEN + + -- If we don't already have one use this as the centre point of the geometry + IF NEW.centroid IS NULL THEN + NEW.centroid := coalesce(linkedPlacex.centroid,st_centroid(linkedPlacex.geometry)); + END IF; + + -- merge in the name, re-init word vector + NEW.name := linkedPlacex.name || NEW.name; + name_vector := make_keywords(NEW.name); + + -- merge in extra tags + NEW.extratags := linkedPlacex.extratags || NEW.extratags; + + -- mark the linked place (excludes from search results) + UPDATE placex set linked_place_id = NEW.place_id where place_id = linkedPlacex.place_id; + + -- keep a note of the node id in case we need it for wikipedia in a bit + linked_node_id := linkedPlacex.osm_id; + END IF; + + END LOOP; + + END IF; + + -- not found one yet? how about doing a name search + IF NEW.centroid IS NULL AND (NEW.name->'name') is not null and make_standard_name(NEW.name->'name') != '' THEN + + FOR linkedPlacex IN select placex.* from placex WHERE + make_standard_name(name->'name') = make_standard_name(NEW.name->'name') + AND placex.rank_search = NEW.rank_search + AND placex.place_id != NEW.place_id + AND placex.osm_type = 'N' AND placex.rank_search < 26 + AND st_contains(NEW.geometry, placex.geometry) + LOOP + + -- If we don't already have one use this as the centre point of the geometry + IF NEW.centroid IS NULL THEN + NEW.centroid := coalesce(linkedPlacex.centroid,st_centroid(linkedPlacex.geometry)); + END IF; + + -- merge in the name, re-init word vector + NEW.name := linkedPlacex.name || NEW.name; + name_vector := make_keywords(NEW.name); + + -- merge in extra tags + NEW.extratags := linkedPlacex.extratags || NEW.extratags; + + -- mark the linked place (excludes from search results) + UPDATE placex set linked_place_id = NEW.place_id where place_id = linkedPlacex.place_id; + + -- keep a note of the node id in case we need it for wikipedia in a bit + linked_node_id := linkedPlacex.osm_id; + END LOOP; + END IF; + + IF NEW.centroid IS NOT NULL THEN + place_centroid := NEW.centroid; + END IF; + + -- Did we gain a wikipedia tag in the process? then we need to recalculate our importance + IF NEW.importance is null THEN + select language||':'||title,importance from get_wikipedia_match(NEW.extratags) INTO NEW.wikipedia,NEW.importance; + END IF; + -- Still null? how about looking it up by the node id + IF NEW.importance IS NULL THEN + select language||':'||title,importance from wikipedia_article where osm_type = 'N' and osm_id = linked_node_id order by importance desc limit 1 INTO NEW.wikipedia,NEW.importance; + END IF; + + END IF; + NEW.parent_place_id = 0; parent_place_id_rank = 0; @@ -1547,6 +1739,11 @@ BEGIN -- INSERT INTO search_name values (NEW.place_id, NEW.rank_search, NEW.rank_search, 0, NEW.country_code, name_vector, nameaddress_vector, place_centroid); END IF; + -- If we've not managed to pick up a better one - default centroid + IF NEW.centroid IS NULL THEN + NEW.centroid := place_centroid; + END IF; + END IF; RETURN NEW; @@ -1558,8 +1755,12 @@ CREATE OR REPLACE FUNCTION placex_delete() RETURNS TRIGGER AS $$ DECLARE b BOOLEAN; + classtable TEXT; BEGIN + update placex set linked_place_id = null where linked_place_id = OLD.place_id; + update placex set indexed_status = 2 where linked_place_id = OLD.place_id and indexed_status = 0; + IF OLD.rank_address < 30 THEN -- mark everything linked to this place for re-indexing @@ -1584,6 +1785,13 @@ BEGIN DELETE FROM place_addressline where place_id = OLD.place_id; + -- remove from tables for special search + classtable := 'place_classtype_' || OLD.class || '_' || OLD.type; + SELECT count(*)>0 FROM pg_tables WHERE tablename = classtable INTO b; + IF b THEN + EXECUTE 'DELETE FROM ' || classtable::regclass || ' WHERE place_id = $1' USING OLD.place_id; + END IF; + RETURN OLD; END; @@ -1770,12 +1978,12 @@ BEGIN update placex set indexed_status = 2 where indexed_status = 0 and (ST_Contains(NEW.geometry, placex.geometry) OR ST_Intersects(NEW.geometry, placex.geometry)) AND NOT (ST_Contains(existinggeometry, placex.geometry) OR ST_Intersects(existinggeometry, placex.geometry)) - AND rank_search > existingplacex.rank_search; + AND rank_search > existingplacex.rank_search AND (rank_search < 28 or name is not null); update placex set indexed_status = 2 where indexed_status = 0 and (ST_Contains(existinggeometry, placex.geometry) OR ST_Intersects(existinggeometry, placex.geometry)) AND NOT (ST_Contains(NEW.geometry, placex.geometry) OR ST_Intersects(NEW.geometry, placex.geometry)) - AND rank_search > existingplacex.rank_search; + AND rank_search > existingplacex.rank_search AND (rank_search < 28 or name is not null); END IF; @@ -1795,7 +2003,8 @@ BEGIN IF st_area(NEW.geometry) < 0.5 THEN UPDATE placex set indexed_status = 2 from place_addressline where address_place_id = existingplacex.place_id - and placex.place_id = place_addressline.place_id and indexed_status = 0; + and placex.place_id = place_addressline.place_id and indexed_status = 0 + and (rank_search < 28 or name is not null); END IF; END IF; @@ -2371,13 +2580,6 @@ END; $$ LANGUAGE plpgsql; -CREATE AGGREGATE array_agg(INT[]) -( - sfunc = array_cat, - stype = INT[], - initcond = '{}' -); - CREATE OR REPLACE FUNCTION tigger_create_interpolation(linegeo GEOMETRY, in_startnumber INTEGER, in_endnumber INTEGER, interpolationtype TEXT, in_street TEXT, in_isin TEXT, in_postcode TEXT) RETURNS INTEGER @@ -2392,8 +2594,8 @@ DECLARE numberrange INTEGER; rangestartnumber INTEGER; place_centroid GEOMETRY; - partition INTEGER; - parent_place_id BIGINT; + out_partition INTEGER; + out_parent_place_id BIGINT; location RECORD; address_street_word_id INTEGER; @@ -2430,32 +2632,32 @@ BEGIN END IF; place_centroid := ST_Centroid(linegeo); - partition := get_partition(place_centroid, 'us'); - parent_place_id := null; + out_partition := get_partition(place_centroid, 'us'); + out_parent_place_id := null; address_street_word_id := get_name_id(make_standard_name(in_street)); IF address_street_word_id IS NOT NULL THEN - FOR location IN SELECT * from getNearestNamedRoadFeature(partition, place_centroid, address_street_word_id) LOOP - parent_place_id := location.place_id; + FOR location IN SELECT * from getNearestNamedRoadFeature(out_partition, place_centroid, address_street_word_id) LOOP + out_parent_place_id := location.place_id; END LOOP; END IF; - IF parent_place_id IS NULL THEN - FOR location IN SELECT place_id FROM getNearestParellelRoadFeature(partition, linegeo) LOOP - parent_place_id := location.place_id; + IF out_parent_place_id IS NULL THEN + FOR location IN SELECT place_id FROM getNearestParellelRoadFeature(out_partition, linegeo) LOOP + out_parent_place_id := location.place_id; END LOOP; END IF; - IF parent_place_id IS NULL THEN - FOR location IN SELECT place_id FROM getNearestRoadFeature(partition, place_centroid) LOOP - parent_place_id := location.place_id; + IF out_parent_place_id IS NULL THEN + FOR location IN SELECT place_id FROM getNearestRoadFeature(out_partition, place_centroid) LOOP + out_parent_place_id := location.place_id; END LOOP; END IF; newpoints := 0; FOR housenum IN startnumber..endnumber BY stepsize LOOP insert into location_property_tiger (place_id, partition, parent_place_id, housenumber, postcode, centroid) - values (nextval('seq_place'), partition, parent_place_id, housenum, in_postcode, + values (nextval('seq_place'), out_partition, out_parent_place_id, housenum, in_postcode, ST_Line_Interpolate_Point(linegeo, (housenum::float-rangestartnumber::float)/numberrange::float)); newpoints := newpoints + 1; END LOOP; @@ -2472,7 +2674,7 @@ DECLARE newpoints INTEGER; place_centroid GEOMETRY; - partition INTEGER; + out_partition INTEGER; out_parent_place_id BIGINT; location RECORD; address_street_word_id INTEGER; @@ -2481,18 +2683,18 @@ DECLARE BEGIN place_centroid := ST_Centroid(pointgeo); - partition := get_partition(place_centroid, in_countrycode); + out_partition := get_partition(place_centroid, in_countrycode); out_parent_place_id := null; address_street_word_id := get_name_id(make_standard_name(in_street)); IF address_street_word_id IS NOT NULL THEN - FOR location IN SELECT * from getNearestNamedRoadFeature(partition, place_centroid, address_street_word_id) LOOP + FOR location IN SELECT * from getNearestNamedRoadFeature(out_partition, place_centroid, address_street_word_id) LOOP out_parent_place_id := location.place_id; END LOOP; END IF; IF out_parent_place_id IS NULL THEN - FOR location IN SELECT place_id FROM getNearestRoadFeature(partition, place_centroid) LOOP + FOR location IN SELECT place_id FROM getNearestRoadFeature(out_partition, place_centroid) LOOP out_parent_place_id := location.place_id; END LOOP; END IF; @@ -2502,15 +2704,124 @@ BEGIN SELECT postcode from placex where place_id = out_parent_place_id INTO out_postcode; END IF; IF out_postcode IS NULL THEN - out_postcode := getNearestPostcode(partition, place_centroid); + out_postcode := getNearestPostcode(out_partition, place_centroid); END IF; newpoints := 0; insert into location_property_aux (place_id, partition, parent_place_id, housenumber, postcode, centroid) - values (nextval('seq_place'), partition, out_parent_place_id, in_housenumber, out_postcode, place_centroid); + values (nextval('seq_place'), out_partition, out_parent_place_id, in_housenumber, out_postcode, place_centroid); newpoints := newpoints + 1; RETURN newpoints; END; $$ LANGUAGE plpgsql; + +CREATE OR REPLACE FUNCTION get_osm_rel_members(members TEXT[], member TEXT) RETURNS TEXT[] + AS $$ +DECLARE + result TEXT[]; + i INTEGER; +BEGIN + + FOR i IN 1..ARRAY_UPPER(members,1) BY 2 LOOP + IF members[i+1] = member THEN + result := result || members[i]; + END IF; + END LOOP; + + return result; +END; +$$ +LANGUAGE plpgsql; + +CREATE OR REPLACE FUNCTION get_osm_rel_members(members TEXT[], memberLabels TEXT[]) RETURNS SETOF TEXT + AS $$ +DECLARE + i INTEGER; +BEGIN + + FOR i IN 1..ARRAY_UPPER(members,1) BY 2 LOOP + IF members[i+1] = ANY(memberLabels) THEN + RETURN NEXT members[i]; + END IF; + END LOOP; + + RETURN; +END; +$$ +LANGUAGE plpgsql; + +-- See: http://stackoverflow.com/questions/6410088/how-can-i-mimic-the-php-urldecode-function-in-postgresql +CREATE OR REPLACE FUNCTION decode_url_part(p varchar) RETURNS varchar + AS $$ +SELECT convert_from(CAST(E'\\x' || array_to_string(ARRAY( + SELECT CASE WHEN length(r.m[1]) = 1 THEN encode(convert_to(r.m[1], 'SQL_ASCII'), 'hex') ELSE substring(r.m[1] from 2 for 2) END + FROM regexp_matches($1, '%[0-9a-f][0-9a-f]|.', 'gi') AS r(m) +), '') AS bytea), 'UTF8'); +$$ +LANGUAGE SQL IMMUTABLE STRICT; + +CREATE OR REPLACE FUNCTION catch_decode_url_part(p varchar) RETURNS varchar + AS $$ +DECLARE +BEGIN + RETURN decode_url_part(p); +EXCEPTION + WHEN others THEN return null; +END; +$$ +LANGUAGE plpgsql IMMUTABLE; + +DROP TYPE wikipedia_article_match CASCADE; +create type wikipedia_article_match as ( + language TEXT, + title TEXT, + importance FLOAT +); + +CREATE OR REPLACE FUNCTION get_wikipedia_match(extratags HSTORE) RETURNS wikipedia_article_match + AS $$ +DECLARE + langs TEXT[]; + i INT; + wiki_article TEXT; + wiki_article_title TEXT; + wiki_article_language TEXT; + result wikipedia_article_match; +BEGIN + langs := ARRAY['','ar','bg','ca','cs','da','de','en','es','eo','eu','fa','fr','ko','hi','hr','id','it','he','lt','hu','ms','nl','ja','no','pl','pt','kk','ro','ru','sk','sl','sr','fi','sv','tr','uk','vi','vo','war','zh']; + i := 1; + WHILE langs[i] IS NOT NULL LOOP + wiki_article := extratags->(case when langs[i] = '' THEN 'wikipedia' ELSE 'wikipedia:'||langs[i] END); + IF wiki_article is not null THEN + wiki_article := replace(regexp_replace(wiki_article,E'(.*?)([a-z]+).wikipedia.org/wiki/',E'\\2:'),' ','_'); + wiki_article_title := split_part(wiki_article, ':', 2); + IF wiki_article_title IS NULL OR wiki_article_title = '' THEN + wiki_article_title := wiki_article; + wiki_article_language := CASE WHEN langs[i] = '' THEN 'en' ELSE langs[i] END; + ELSE + wiki_article_language := lower(split_part(wiki_article, ':', 1)); + END IF; + + select wikipedia_article.language,wikipedia_article.title,wikipedia_article.importance + from wikipedia_article + where language = wiki_article_language and + (title = wiki_article_title OR title = catch_decode_url_part(wiki_article_title) OR title = replace(catch_decode_url_part(wiki_article_title),E'\\','')) + UNION ALL + select wikipedia_article.language,wikipedia_article.title,wikipedia_article.importance + from wikipedia_redirect join wikipedia_article on (wikipedia_redirect.language = wikipedia_article.language and wikipedia_redirect.to_title = wikipedia_article.title) + where wikipedia_redirect.language = wiki_article_language and + (from_title = wiki_article_title OR from_title = catch_decode_url_part(wiki_article_title) OR from_title = replace(catch_decode_url_part(wiki_article_title),E'\\','')) + order by importance desc limit 1 INTO result; + + IF result.language is not null THEN + return result; + END IF; + END IF; + i := i + 1; + END LOOP; + RETURN NULL; +END; +$$ +LANGUAGE plpgsql;