X-Git-Url: https://git.openstreetmap.org./nominatim.git/blobdiff_plain/71ddeb40a66e0fef307350cdcd7cb823c90eb132..09935d41484d5e9ec450bb16170785849029f3da:/sql/functions.sql diff --git a/sql/functions.sql b/sql/functions.sql index fd71c6fd..33a5e6b9 100644 --- a/sql/functions.sql +++ b/sql/functions.sql @@ -35,7 +35,7 @@ CREATE OR REPLACE FUNCTION make_standard_name(name TEXT) RETURNS TEXT DECLARE o TEXT; BEGIN - o := gettokenstring(transliteration(name)); + o := public.gettokenstring(public.transliteration(name)); RETURN trim(substr(o,1,length(o))); END; $$ @@ -83,6 +83,26 @@ END; $$ LANGUAGE plpgsql; +CREATE OR REPLACE FUNCTION getorcreate_postcode_id(postcode TEXT) + RETURNS INTEGER + AS $$ +DECLARE + lookup_token TEXT; + lookup_word TEXT; + return_word_id INTEGER; +BEGIN + lookup_word := upper(trim(postcode)); + lookup_token := ' ' || make_standard_name(lookup_word); + SELECT min(word_id) FROM word WHERE word_token = lookup_token and class='place' and type='postcode' into return_word_id; + IF return_word_id IS NULL THEN + return_word_id := nextval('seq_word'); + INSERT INTO word VALUES (return_word_id, lookup_token, lookup_word, 'place', 'postcode', null, 0); + END IF; + RETURN return_word_id; +END; +$$ +LANGUAGE plpgsql; + CREATE OR REPLACE FUNCTION getorcreate_country(lookup_word TEXT, lookup_country_code varchar(2)) RETURNS INTEGER AS $$ @@ -236,6 +256,28 @@ END; $$ LANGUAGE plpgsql IMMUTABLE; +CREATE OR REPLACE FUNCTION reverse_place_diameter(rank_search SMALLINT) + RETURNS FLOAT + AS $$ +BEGIN + IF rank_search <= 4 THEN + RETURN 5.0; + ELSIF rank_search <= 8 THEN + RETURN 1.8; + ELSIF rank_search <= 12 THEN + RETURN 0.6; + ELSIF rank_search <= 17 THEN + RETURN 0.16; + ELSIF rank_search <= 18 THEN + RETURN 0.08; + ELSIF rank_search <= 19 THEN + RETURN 0.04; + END IF; + + RETURN 0.02; +END; +$$ +LANGUAGE plpgsql IMMUTABLE; CREATE OR REPLACE FUNCTION get_postcode_rank(country_code VARCHAR(2), postcode TEXT, OUT rank_search SMALLINT, OUT rank_address SMALLINT) @@ -245,6 +287,7 @@ DECLARE BEGIN rank_search := 30; rank_address := 30; + postcode := upper(postcode); IF country_code = 'gb' THEN IF postcode ~ '^([A-Z][A-Z]?[0-9][0-9A-Z]? [0-9][A-Z][A-Z])$' THEN @@ -293,6 +336,39 @@ END; $$ LANGUAGE plpgsql IMMUTABLE; +-- Find the nearest artificial postcode for the given geometry. +-- TODO For areas there should not be more than two inside the geometry. +CREATE OR REPLACE FUNCTION get_nearest_postcode(country VARCHAR(2), geom GEOMETRY) RETURNS TEXT + AS $$ +DECLARE + outcode TEXT; + cnt INTEGER; +BEGIN + -- If the geometry is an area then only one postcode must be within + -- that area, otherwise consider the area as not having a postcode. + IF ST_GeometryType(geom) in ('ST_Polygon','ST_MultiPolygon') THEN + SELECT min(postcode), count(*) FROM + (SELECT postcode FROM location_postcode + WHERE ST_Contains(geom, location_postcode.geometry) LIMIT 2) sub + INTO outcode, cnt; + + IF cnt = 1 THEN + RETURN outcode; + ELSE + RETURN null; + END IF; + END IF; + + SELECT postcode FROM location_postcode + WHERE ST_DWithin(geom, location_postcode.geometry, 0.05) + AND location_postcode.country_code = country + ORDER BY ST_Distance(geom, location_postcode.geometry) LIMIT 1 + INTO outcode; + + RETURN outcode; +END; +$$ +LANGUAGE plpgsql; CREATE OR REPLACE FUNCTION create_country(src HSTORE, lookup_country_code varchar(2)) RETURNS VOID @@ -574,36 +650,38 @@ CREATE OR REPLACE FUNCTION add_location( keywords INTEGER[], rank_search INTEGER, rank_address INTEGER, + in_postcode TEXT, geometry GEOMETRY ) RETURNS BOOLEAN AS $$ DECLARE locationid INTEGER; - isarea BOOLEAN; centroid GEOMETRY; diameter FLOAT; x BOOLEAN; splitGeom RECORD; secgeo GEOMETRY; + postcode TEXT; BEGIN IF rank_search > 25 THEN RAISE EXCEPTION 'Adding location with rank > 25 (% rank %)', place_id, rank_search; END IF; --- RAISE WARNING 'Adding location with rank > 25 (% rank %)', place_id, rank_search; - x := deleteLocationArea(partition, place_id, rank_search); - isarea := false; - IF (ST_GeometryType(geometry) in ('ST_Polygon','ST_MultiPolygon') AND ST_IsValid(geometry)) THEN + -- add postcode only if it contains a single entry, i.e. ignore postcode lists + postcode := NULL; + IF in_postcode is not null AND in_postcode not similar to '%(,|;)%' THEN + postcode := upper(trim (in_postcode)); + END IF; - isArea := true; + IF ST_GeometryType(geometry) in ('ST_Polygon','ST_MultiPolygon') THEN centroid := ST_Centroid(geometry); FOR secgeo IN select split_geometry(geometry) AS geom LOOP - x := insertLocationAreaLarge(partition, place_id, country_code, keywords, rank_search, rank_address, false, centroid, secgeo); + x := insertLocationAreaLarge(partition, place_id, country_code, keywords, rank_search, rank_address, false, postcode, centroid, secgeo); END LOOP; ELSE @@ -628,7 +706,7 @@ BEGIN -- RAISE WARNING 'adding % diameter %', place_id, diameter; secgeo := ST_Buffer(geometry, diameter); - x := insertLocationAreaLarge(partition, place_id, country_code, keywords, rank_search, rank_address, true, ST_Centroid(geometry), secgeo); + x := insertLocationAreaLarge(partition, place_id, country_code, keywords, rank_search, rank_address, true, postcode, ST_Centroid(geometry), secgeo); END IF; @@ -781,12 +859,15 @@ BEGIN RETURN NULL; END IF; - NEW.postcode := NEW.address->'postcode'; - NEW.name := hstore('ref', NEW.postcode); + NEW.name := hstore('ref', NEW.address->'postcode'); - SELECT * FROM get_postcode_rank(NEW.country_code, NEW.postcode) + SELECT * FROM get_postcode_rank(NEW.country_code, NEW.address->'postcode') INTO NEW.rank_search, NEW.rank_address; + IF NOT ST_GeometryType(NEW.geometry) IN ('ST_Polygon','ST_MultiPolygon') THEN + NEW.rank_address := 0; + END IF; + ELSEIF NEW.class = 'place' THEN IF NEW.type in ('continent') THEN NEW.rank_search := 2; @@ -860,6 +941,9 @@ BEGIN ELSE NEW.rank_address := 0; END IF; + ELSEIF NEW.class = 'leisure' and NEW.type in ('park') THEN + NEW.rank_search := 24; + NEW.rank_address := 0; ELSEIF NEW.class = 'natural' and NEW.type in ('peak','volcano','mountain_range') THEN NEW.rank_search := 18; NEW.rank_address := 0; @@ -1001,8 +1085,8 @@ DECLARE linegeo GEOMETRY; splitline GEOMETRY; sectiongeo GEOMETRY; + interpol_postcode TEXT; postcode TEXT; - seg_postcode TEXT; BEGIN -- deferred delete IF OLD.indexed_status = 100 THEN @@ -1021,9 +1105,11 @@ BEGIN NEW.address->'place', NEW.partition, place_centroid, NEW.linegeo); - - IF NEW.address is not NULL and NEW.address ? 'postcode' THEN - NEW.postcode = NEW.address->'postcode'; + IF NEW.address is not NULL AND NEW.address ? 'postcode' AND NEW.address->'postcode' not similar to '%(,|;)%' THEN + interpol_postcode := NEW.address->'postcode'; + housenum := getorcreate_postcode_id(NEW.address->'postcode'); + ELSE + interpol_postcode := NULL; END IF; -- if the line was newly inserted, split the line as necessary @@ -1036,7 +1122,6 @@ BEGIN linegeo := NEW.linegeo; startnumber := NULL; - postcode := NEW.postcode; FOR nodeidpos in 1..array_upper(waynodes, 1) LOOP @@ -1069,15 +1154,24 @@ BEGIN sectiongeo := ST_Reverse(sectiongeo); END IF; - seg_postcode := coalesce(postcode, - prevnode.address->'postcode', - nextnode.address->'postcode'); + -- determine postcode + postcode := coalesce(interpol_postcode, + prevnode.address->'postcode', + nextnode.address->'postcode', + postcode); + + IF postcode is NULL THEN + SELECT placex.postcode FROM placex WHERE place_id = NEW.parent_place_id INTO postcode; + END IF; + IF postcode is NULL THEN + postcode := get_nearest_postcode(NEW.country_code, nextnode.geometry); + END IF; IF NEW.startnumber IS NULL THEN NEW.startnumber := startnumber; NEW.endnumber := endnumber; NEW.linegeo := sectiongeo; - NEW.postcode := seg_postcode; + NEW.postcode := upper(trim(postcode)); ELSE insert into location_property_osmline (linegeo, partition, osm_id, parent_place_id, @@ -1086,7 +1180,7 @@ BEGIN geometry_sector, indexed_status) values (sectiongeo, NEW.partition, NEW.osm_id, NEW.parent_place_id, startnumber, endnumber, NEW.interpolationtype, - NEW.address, seg_postcode, + NEW.address, postcode, NEW.country_code, NEW.geometry_sector, 0); END IF; END IF; @@ -1166,6 +1260,7 @@ DECLARE relation_members TEXT[]; relMember RECORD; linkedplacex RECORD; + addr_item RECORD; search_diameter FLOAT; search_prevdiameter FLOAT; search_maxrank INTEGER; @@ -1209,11 +1304,6 @@ BEGIN --DEBUG: RAISE WARNING 'placex_update % % (%)',NEW.osm_type,NEW.osm_id,NEW.place_id; - IF NEW.class = 'place' AND NEW.type = 'postcodearea' THEN - -- Silently do nothing - RETURN NEW; - END IF; - NEW.indexed_date = now(); result := deleteSearchName(NEW.partition, NEW.place_id); @@ -1247,16 +1337,19 @@ BEGIN i := getorcreate_housenumber_id(make_standard_name(NEW.housenumber)); END IF; - addr_street = NEW.address->'street'; - addr_place = NEW.address->'place'; + addr_street := NEW.address->'street'; + addr_place := NEW.address->'place'; - NEW.postcode = NEW.address->'postcode'; + IF NEW.address ? 'postcode' and NEW.address->'postcode' not similar to '%(,|;)%' THEN + i := getorcreate_postcode_id(NEW.address->'postcode'); + END IF; END IF; -- Speed up searches - just use the centroid of the feature -- cheaper but less acurate place_centroid := ST_PointOnSurface(NEW.geometry); NEW.centroid := null; + NEW.postcode := null; --DEBUG: RAISE WARNING 'Computing preliminary centroid at %',ST_AsText(place_centroid); -- recalculate country and partition @@ -1289,7 +1382,7 @@ BEGIN --DEBUG: RAISE WARNING 'waterway parent %, child %/%', NEW.osm_id, i, relation_members[i]; FOR linked_node_id IN SELECT place_id FROM placex WHERE osm_type = 'W' and osm_id = substring(relation_members[i],2,200)::bigint - and class = NEW.class and type = NEW.type + and class = NEW.class and type in ('river', 'stream', 'canal', 'drain', 'ditch') and ( relation_members[i+1] != 'side_stream' or NEW.name->'name' = name->'name') LOOP UPDATE placex SET linked_place_id = NEW.place_id WHERE place_id = linked_node_id; @@ -1352,7 +1445,7 @@ BEGIN -- see if we can get it from a surrounding building IF NEW.osm_type = 'N' AND addr_street IS NULL AND addr_place IS NULL AND NEW.housenumber IS NULL THEN - FOR location IN select * from placex where ST_Covers(geometry, place_centroid) + FOR location IN select address from placex where ST_Covers(geometry, place_centroid) and address is not null and (address ? 'housenumber' or address ? 'street' or address ? 'place') and rank_search > 28 AND ST_GeometryType(geometry) in ('ST_Polygon','ST_MultiPolygon') @@ -1389,9 +1482,7 @@ BEGIN IF NEW.parent_place_id IS NULL AND addr_street IS NOT NULL THEN address_street_word_ids := get_name_ids(make_standard_name(addr_street)); IF address_street_word_ids IS NOT NULL THEN - FOR location IN SELECT * from getNearestNamedRoadFeature(NEW.partition, place_centroid, address_street_word_ids) LOOP - NEW.parent_place_id := location.place_id; - END LOOP; + SELECT place_id from getNearestNamedRoadFeature(NEW.partition, place_centroid, address_street_word_ids) INTO NEW.parent_place_id; END IF; END IF; --DEBUG: RAISE WARNING 'Checked for addr:street (%)', NEW.parent_place_id; @@ -1399,88 +1490,80 @@ BEGIN IF NEW.parent_place_id IS NULL AND addr_place IS NOT NULL THEN address_street_word_ids := get_name_ids(make_standard_name(addr_place)); IF address_street_word_ids IS NOT NULL THEN - FOR location IN SELECT * from getNearestNamedPlaceFeature(NEW.partition, place_centroid, address_street_word_ids) LOOP - NEW.parent_place_id := location.place_id; - END LOOP; + SELECT place_id from getNearestNamedPlaceFeature(NEW.partition, place_centroid, address_street_word_ids) INTO NEW.parent_place_id; END IF; END IF; --DEBUG: RAISE WARNING 'Checked for addr:place (%)', NEW.parent_place_id; -- Is this node part of an interpolation? IF NEW.parent_place_id IS NULL AND NEW.osm_type = 'N' THEN - FOR location IN - SELECT q.parent_place_id FROM location_property_osmline q, planet_osm_ways x - WHERE q.linegeo && NEW.geometry and x.id = q.osm_id and NEW.osm_id = any(x.nodes) - LIMIT 1 - LOOP - NEW.parent_place_id := location.parent_place_id; - END LOOP; + SELECT q.parent_place_id FROM location_property_osmline q, planet_osm_ways x + WHERE q.linegeo && NEW.geometry and x.id = q.osm_id and NEW.osm_id = any(x.nodes) + LIMIT 1 INTO NEW.parent_place_id; END IF; --DEBUG: RAISE WARNING 'Checked for interpolation (%)', NEW.parent_place_id; -- Is this node part of a way? IF NEW.parent_place_id IS NULL AND NEW.osm_type = 'N' THEN - FOR location IN select p.place_id, p.osm_id, p.parent_place_id, p.rank_search, p.address from placex p, planet_osm_ways w - where p.osm_type = 'W' and p.rank_search >= 26 and p.geometry && NEW.geometry and w.id = p.osm_id and NEW.osm_id = any(w.nodes) + FOR location IN + SELECT p.place_id, p.osm_id, p.rank_search, p.address from placex p, planet_osm_ways w + WHERE p.osm_type = 'W' and p.rank_search >= 26 and p.geometry && NEW.geometry and w.id = p.osm_id and NEW.osm_id = any(w.nodes) LOOP --DEBUG: RAISE WARNING 'Node is part of way % ', location.osm_id; -- Way IS a road then we are on it - that must be our road - IF location.rank_search < 28 AND NEW.parent_place_id IS NULL THEN + IF location.rank_search < 28 THEN --RAISE WARNING 'node in way that is a street %',location; NEW.parent_place_id := location.place_id; + EXIT; END IF; --DEBUG: RAISE WARNING 'Checked if way is street (%)', NEW.parent_place_id; -- If the way mentions a street or place address, try that for parenting. - IF NEW.parent_place_id IS NULL AND location.address ? 'street' THEN - address_street_word_ids := get_name_ids(make_standard_name(location.address->'street')); - IF address_street_word_ids IS NOT NULL THEN - FOR linkedplacex IN SELECT place_id from getNearestNamedRoadFeature(NEW.partition, place_centroid, address_street_word_ids) LOOP - NEW.parent_place_id := linkedplacex.place_id; - END LOOP; + IF location.address is not null THEN + IF location.address ? 'street' THEN + address_street_word_ids := get_name_ids(make_standard_name(location.address->'street')); + IF address_street_word_ids IS NOT NULL THEN + SELECT place_id from getNearestNamedRoadFeature(NEW.partition, place_centroid, address_street_word_ids) INTO NEW.parent_place_id; + EXIT WHEN NEW.parent_place_id is not NULL; + END IF; END IF; - END IF; - --DEBUG: RAISE WARNING 'Checked for addr:street in way (%)', NEW.parent_place_id; + --DEBUG: RAISE WARNING 'Checked for addr:street in way (%)', NEW.parent_place_id; - IF NEW.parent_place_id IS NULL AND location.address ? 'place' THEN - address_street_word_ids := get_name_ids(make_standard_name(location.address->'place')); - IF address_street_word_ids IS NOT NULL THEN - FOR linkedplacex IN SELECT place_id from getNearestNamedPlaceFeature(NEW.partition, place_centroid, address_street_word_ids) LOOP - NEW.parent_place_id := linkedplacex.place_id; - END LOOP; + IF location.address ? 'place' THEN + address_street_word_ids := get_name_ids(make_standard_name(location.address->'place')); + IF address_street_word_ids IS NOT NULL THEN + SELECT place_id from getNearestNamedPlaceFeature(NEW.partition, place_centroid, address_street_word_ids) INTO NEW.parent_place_id; + EXIT WHEN NEW.parent_place_id is not NULL; + END IF; END IF; - END IF; --DEBUG: RAISE WARNING 'Checked for addr:place in way (%)', NEW.parent_place_id; + END IF; -- 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] 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 - FOR i IN 1..array_upper(relation.members, 1) BY 2 LOOP - IF NEW.parent_place_id IS NULL AND relation.members[i+1] = 'street' THEN - --RAISE WARNING 'node in way that is in a relation %',relation; - SELECT place_id from placex where osm_type='W' and osm_id = substring(relation.members[i],2,200)::bigint - and rank_search = 26 and name is not null INTO NEW.parent_place_id; - END IF; - END LOOP; + 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 + FOR i IN 1..array_upper(relation.members, 1) BY 2 LOOP + IF NEW.parent_place_id IS NULL AND relation.members[i+1] = 'street' THEN +--RAISE WARNING 'node in way that is in a relation %',relation; + SELECT place_id from placex where osm_type='W' and osm_id = substring(relation.members[i],2,200)::bigint + and rank_search = 26 and name is not null INTO NEW.parent_place_id; END IF; END LOOP; - END IF; + END IF; + END LOOP; + EXIT WHEN NEW.parent_place_id is not null; --DEBUG: RAISE WARNING 'Checked for street relation in way (%)', NEW.parent_place_id; END LOOP; - END IF; -- Still nothing, just use the nearest road IF NEW.parent_place_id IS NULL THEN - FOR location IN SELECT place_id FROM getNearestRoadFeature(NEW.partition, place_centroid) LOOP - NEW.parent_place_id := location.place_id; - END LOOP; + SELECT place_id FROM getNearestRoadFeature(NEW.partition, place_centroid) INTO NEW.parent_place_id; END IF; --DEBUG: RAISE WARNING 'Checked for nearest way (%)', NEW.parent_place_id; @@ -1489,29 +1572,21 @@ BEGIN IF NEW.parent_place_id IS NOT NULL THEN -- Get the details of the parent road - select * from search_name where place_id = NEW.parent_place_id INTO location; + select s.country_code, s.name_vector, s.nameaddress_vector from search_name s + where s.place_id = NEW.parent_place_id INTO location; NEW.country_code := location.country_code; --DEBUG: RAISE WARNING 'Got parent details from search name'; - -- Merge the postcode into the parent's address if necessary - IF NEW.postcode IS NOT NULL THEN - --DEBUG: RAISE WARNING 'Merging postcode into parent'; - isin_tokens := '{}'::int[]; - address_street_word_id := getorcreate_word_id(make_standard_name(NEW.postcode)); - IF address_street_word_id is not null - and not ARRAY[address_street_word_id] <@ location.nameaddress_vector THEN - isin_tokens := isin_tokens || address_street_word_id; - END IF; - address_street_word_id := getorcreate_name_id(make_standard_name(NEW.postcode)); - IF address_street_word_id is not null - and not ARRAY[address_street_word_id] <@ location.nameaddress_vector THEN - isin_tokens := isin_tokens || address_street_word_id; - END IF; - IF isin_tokens != '{}'::int[] THEN - UPDATE search_name - SET nameaddress_vector = search_name.nameaddress_vector || isin_tokens - WHERE place_id = NEW.parent_place_id; - END IF; + -- determine postcode + IF NEW.rank_search > 4 THEN + IF NEW.address is not null AND NEW.address ? 'postcode' THEN + NEW.postcode = upper(trim(NEW.address->'postcode')); + ELSE + SELECT postcode FROM placex WHERE place_id = NEW.parent_place_id INTO NEW.postcode; + END IF; + IF NEW.postcode is null THEN + NEW.postcode := get_nearest_postcode(NEW.country_code, place_centroid); + END IF; END IF; -- If there is no name it isn't searchable, don't bother to create a search record @@ -1529,7 +1604,7 @@ BEGIN -- Just be happy with inheriting from parent road only IF NEW.rank_search <= 25 and NEW.rank_address > 0 THEN - result := add_location(NEW.place_id, NEW.country_code, NEW.partition, name_vector, NEW.rank_search, NEW.rank_address, NEW.geometry); + result := add_location(NEW.place_id, NEW.country_code, NEW.partition, name_vector, NEW.rank_search, NEW.rank_address, upper(trim(NEW.address->'postcode')), NEW.geometry); --DEBUG: RAISE WARNING 'Place added to location table'; END IF; @@ -1700,7 +1775,7 @@ BEGIN END IF; -- make sure all names are in the word table - IF NEW.admin_level = 2 AND NEW.class = 'boundary' AND NEW.type = 'administrative' AND NEW.country_code IS NOT NULL THEN + IF NEW.admin_level = 2 AND NEW.class = 'boundary' AND NEW.type = 'administrative' AND NEW.country_code IS NOT NULL AND NEW.osm_type = 'R' THEN perform create_country(NEW.name, lower(NEW.country_code)); --DEBUG: RAISE WARNING 'Country names updated'; END IF; @@ -1709,63 +1784,43 @@ BEGIN parent_place_id_rank = 0; - -- convert isin to array of tokenids + -- convert address store to array of tokenids --DEBUG: RAISE WARNING 'Starting address search'; isin_tokens := '{}'::int[]; IF NEW.address IS NOT NULL THEN - isin := avals(NEW.address); - IF array_upper(isin, 1) IS NOT NULL THEN - FOR i IN 1..array_upper(isin, 1) LOOP - address_street_word_id := get_name_id(make_standard_name(isin[i])); + FOR addr_item IN SELECT * FROM each(NEW.address) + LOOP + IF addr_item.key IN ('city', 'tiger:county', 'state', 'suburb', 'province', 'district', 'region', 'county', 'municipality', 'hamlet', 'village', 'subdistrict', 'town', 'neighbourhood', 'quarter', 'parish') THEN + address_street_word_id := get_name_id(make_standard_name(addr_item.value)); IF address_street_word_id IS NOT NULL AND NOT(ARRAY[address_street_word_id] <@ isin_tokens) THEN - nameaddress_vector := array_merge(nameaddress_vector, ARRAY[address_street_word_id]); isin_tokens := isin_tokens || address_street_word_id; END IF; - - -- merge word into address vector - address_street_word_id := get_word_id(make_standard_name(isin[i])); + address_street_word_id := get_word_id(make_standard_name(addr_item.value)); IF address_street_word_id IS NOT NULL THEN nameaddress_vector := array_merge(nameaddress_vector, ARRAY[address_street_word_id]); END IF; - END LOOP; - END IF; - END IF; - --DEBUG: RAISE WARNING '"address:* tokens collected'; - IF NEW.postcode IS NOT NULL THEN - isin := regexp_split_to_array(NEW.postcode, E'[;,]'); - IF array_upper(isin, 1) IS NOT NULL THEN - FOR i IN 1..array_upper(isin, 1) LOOP - address_street_word_id := get_name_id(make_standard_name(isin[i])); - IF address_street_word_id IS NOT NULL AND NOT(ARRAY[address_street_word_id] <@ isin_tokens) THEN - nameaddress_vector := array_merge(nameaddress_vector, ARRAY[address_street_word_id]); - isin_tokens := isin_tokens || address_street_word_id; - END IF; + END IF; + IF addr_item.key = 'is_in' THEN + -- is_in items need splitting + isin := regexp_split_to_array(addr_item.value, E'[;,]'); + IF array_upper(isin, 1) IS NOT NULL THEN + FOR i IN 1..array_upper(isin, 1) LOOP + address_street_word_id := get_name_id(make_standard_name(isin[i])); + IF address_street_word_id IS NOT NULL AND NOT(ARRAY[address_street_word_id] <@ isin_tokens) THEN + isin_tokens := isin_tokens || address_street_word_id; + END IF; - -- merge into address vector - address_street_word_id := get_word_id(make_standard_name(isin[i])); - IF address_street_word_id IS NOT NULL THEN - nameaddress_vector := array_merge(nameaddress_vector, ARRAY[address_street_word_id]); + -- merge word into address vector + address_street_word_id := get_word_id(make_standard_name(isin[i])); + IF address_street_word_id IS NOT NULL THEN + nameaddress_vector := array_merge(nameaddress_vector, ARRAY[address_street_word_id]); + END IF; + END LOOP; END IF; - END LOOP; - END IF; - END IF; - --DEBUG: RAISE WARNING 'postcode tokens collected'; - - -- %NOTIGERDATA% IF 0 THEN - -- for the USA we have an additional address table. Merge in zip codes from there too - IF NEW.rank_search = 26 AND NEW.country_code = 'us' THEN - FOR location IN SELECT distinct postcode from location_property_tiger where parent_place_id = NEW.place_id LOOP - address_street_word_id := get_name_id(make_standard_name(location.postcode)); - nameaddress_vector := array_merge(nameaddress_vector, ARRAY[address_street_word_id]); - isin_tokens := isin_tokens || address_street_word_id; - - -- also merge in the single word version - address_street_word_id := get_word_id(make_standard_name(location.postcode)); - nameaddress_vector := array_merge(nameaddress_vector, ARRAY[address_street_word_id]); + END IF; END LOOP; END IF; - --DEBUG: RAISE WARNING 'Tiger postcodes collected'; - -- %NOTIGERDATA% END IF; + nameaddress_vector := array_merge(nameaddress_vector, isin_tokens); -- RAISE WARNING 'ISIN: %', isin_tokens; @@ -1821,6 +1876,11 @@ BEGIN VALUES (NEW.place_id, location.place_id, true, location_isaddress, location.distance, location.rank_address); IF location_isaddress THEN + -- add postcode if we have one + -- (If multiple postcodes are available, we end up with the highest ranking one.) + IF location.postcode is not null THEN + NEW.postcode = location.postcode; + END IF; address_havelevel[location.rank_address] := true; IF NOT location.isguess THEN @@ -1841,67 +1901,20 @@ BEGIN END LOOP; --DEBUG: RAISE WARNING 'address computed'; - -- try using the isin value to find parent places - IF array_upper(isin_tokens, 1) IS NOT NULL THEN - FOR i IN 1..array_upper(isin_tokens, 1) LOOP ---RAISE WARNING ' getNearestNamedFeature: % % % %',NEW.partition, place_centroid, search_maxrank, isin_tokens[i]; - IF NOT ARRAY[isin_tokens[i]] <@ nameaddress_vector THEN - - FOR location IN SELECT * from getNearestNamedFeature(NEW.partition, place_centroid, search_maxrank, isin_tokens[i]) LOOP - ---RAISE WARNING ' ISIN: %',location; - - IF location.rank_search > 4 THEN - nameaddress_vector := array_merge(nameaddress_vector, location.keywords::integer[]); - INSERT INTO place_addressline (place_id, address_place_id, fromarea, isaddress, distance, cached_rank_address) - VALUES (NEW.place_id, location.place_id, false, NOT address_havelevel[location.rank_address], location.distance, location.rank_address); - address_havelevel[location.rank_address] := true; - - IF location.rank_address > parent_place_id_rank THEN - NEW.parent_place_id = location.place_id; - parent_place_id_rank = location.rank_address; - END IF; - END IF; - END LOOP; - - END IF; - - END LOOP; + IF NEW.address is not null AND NEW.address ? 'postcode' + AND NEW.address->'postcode' not similar to '%(,|;)%' THEN + NEW.postcode := upper(trim(NEW.address->'postcode')); END IF; - --DEBUG: RAISE WARNING 'isin tokens processed'; - - -- for long ways we should add search terms for the entire length - IF st_length(NEW.geometry) > 0.05 THEN - - location_rank_search := 0; - location_distance := 0; - - FOR location IN SELECT * from getNearFeatures(NEW.partition, NEW.geometry, search_maxrank, isin_tokens) LOOP - - IF location.rank_address != location_rank_search THEN - location_rank_search := location.rank_address; - location_distance := location.distance * 1.5; - END IF; - - IF location.rank_search > 4 AND location.distance < location_distance THEN - - -- Add it to the list of search terms - nameaddress_vector := array_merge(nameaddress_vector, location.keywords::integer[]); - INSERT INTO place_addressline (place_id, address_place_id, fromarea, isaddress, distance, cached_rank_address) - VALUES (NEW.place_id, location.place_id, true, false, location.distance, location.rank_address); - - END IF; - - END LOOP; + IF NEW.postcode is null AND NEW.rank_search > 8 THEN + NEW.postcode := get_nearest_postcode(NEW.country_code, NEW.geometry); END IF; - --DEBUG: RAISE WARNING 'search terms for long ways added'; -- if we have a name add this to the name search table IF NEW.name IS NOT NULL THEN IF NEW.rank_search <= 25 and NEW.rank_address > 0 THEN - result := add_location(NEW.place_id, NEW.country_code, NEW.partition, name_vector, NEW.rank_search, NEW.rank_address, NEW.geometry); + result := add_location(NEW.place_id, NEW.country_code, NEW.partition, name_vector, NEW.rank_search, NEW.rank_address, upper(trim(NEW.address->'postcode')), NEW.geometry); --DEBUG: RAISE WARNING 'added to location (full)'; END IF; @@ -2143,7 +2156,9 @@ BEGIN -- To paraphrase, if there isn't an existing item, OR if the admin level has changed IF existingplacex.osm_type IS NULL OR - (coalesce(existingplacex.admin_level, 15) != coalesce(NEW.admin_level, 15) AND existingplacex.class = 'boundary' AND existingplacex.type = 'administrative') + (existingplacex.class = 'boundary' AND + ((coalesce(existingplacex.admin_level, 15) != coalesce(NEW.admin_level, 15) AND existingplacex.type = 'administrative') OR + (existingplacex.type != NEW.type))) THEN IF existingplacex.osm_type IS NOT NULL THEN @@ -2298,46 +2313,6 @@ END; $$ LANGUAGE plpgsql IMMUTABLE; - -CREATE OR REPLACE FUNCTION get_address_postcode(for_place_id BIGINT) RETURNS TEXT - AS $$ -DECLARE - result TEXT[]; - search TEXT[]; - for_postcode TEXT; - found INTEGER; - location RECORD; -BEGIN - - found := 1000; - search := ARRAY['ref']; - result := '{}'; - - select postcode from placex where place_id = for_place_id limit 1 into for_postcode; - - FOR location IN - select rank_address,name,distance,length(name::text) as namelength - from place_addressline join placex on (address_place_id = placex.place_id) - where place_addressline.place_id = for_place_id and rank_address in (5,11) - order by rank_address desc,rank_search desc,fromarea desc,distance asc,namelength desc - LOOP - IF array_upper(search, 1) IS NOT NULL AND array_upper(location.name, 1) IS NOT NULL THEN - FOR j IN 1..array_upper(search, 1) LOOP - FOR k IN 1..array_upper(location.name, 1) LOOP - IF (found > location.rank_address AND location.name[k].key = search[j] AND location.name[k].value != '') AND NOT result @> ARRAY[trim(location.name[k].value)] AND (for_postcode IS NULL OR location.name[k].value ilike for_postcode||'%') THEN - result[(100 - location.rank_address)] := trim(location.name[k].value); - found := location.rank_address; - END IF; - END LOOP; - END LOOP; - END IF; - END LOOP; - - RETURN array_to_string(result,', '); -END; -$$ -LANGUAGE plpgsql; - --housenumber only needed for tiger data CREATE OR REPLACE FUNCTION get_address_by_language(for_place_id BIGINT, housenumber INTEGER, languagepref TEXT[]) RETURNS TEXT AS $$ @@ -2421,11 +2396,19 @@ BEGIN -- %NOAUXDATA% IF 0 THEN IF for_place_id IS NULL THEN select parent_place_id,'us', housenumber, 30, postcode, null, 'place', 'house' from location_property_aux - WHERE place_id = in_place_id + WHERE place_id = in_place_id INTO for_place_id,searchcountrycode, searchhousenumber, searchrankaddress, searchpostcode, searchhousename, searchclass, searchtype; END IF; -- %NOAUXDATA% END IF; + -- postcode table + IF for_place_id IS NULL THEN + select parent_place_id, country_code, rank_address, postcode, 'place', 'postcode' + FROM location_postcode + WHERE place_id = in_place_id + INTO for_place_id, searchcountrycode, searchrankaddress, searchpostcode, searchclass, searchtype; + END IF; + IF for_place_id IS NULL THEN select parent_place_id, country_code, housenumber, rank_search, postcode, name, class, type from placex WHERE place_id = in_place_id and rank_search > 27 @@ -2444,9 +2427,8 @@ BEGIN found := 1000; hadcountry := false; FOR location IN - select placex.place_id, osm_type, osm_id, - CASE WHEN class = 'place' and type = 'postcode' THEN hstore('name', postcode) ELSE name END as name, - class, type, admin_level, true as fromarea, true as isaddress, + select placex.place_id, osm_type, osm_id, name, + class, type, admin_level, true as isaddress, CASE WHEN rank_address = 0 THEN 100 WHEN rank_address = 11 THEN 5 ELSE rank_address END as rank_address, 0 as distance, country_code, postcode from placex @@ -2456,13 +2438,9 @@ BEGIN IF searchcountrycode IS NULL AND location.country_code IS NOT NULL THEN searchcountrycode := location.country_code; END IF; - IF searchpostcode IS NOT NULL and location.type = 'postcode' THEN + IF location.type in ('postcode', 'postal_code') THEN location.isaddress := FALSE; - END IF; - IF searchpostcode IS NULL and location.postcode IS NOT NULL THEN - searchpostcode := location.postcode; - END IF; - IF location.rank_address = 4 AND location.isaddress THEN + ELSEIF location.rank_address = 4 THEN hadcountry := true; END IF; IF location.rank_address < 4 AND NOT hadcountry THEN @@ -2473,24 +2451,23 @@ BEGIN END IF; END IF; countrylocation := ROW(location.place_id, location.osm_type, location.osm_id, location.name, location.class, - location.type, location.admin_level, location.fromarea, location.isaddress, location.rank_address, + location.type, location.admin_level, true, location.isaddress, location.rank_address, location.distance)::addressline; RETURN NEXT countrylocation; found := location.rank_address; END LOOP; FOR location IN - select placex.place_id, osm_type, osm_id, - CASE WHEN class = 'place' and type = 'postcode' THEN hstore('name', postcode) ELSE name END as name, + select placex.place_id, osm_type, osm_id, name, CASE WHEN extratags ? 'place' THEN 'place' ELSE class END as class, CASE WHEN extratags ? 'place' THEN extratags->'place' ELSE type END as type, - admin_level, fromarea, isaddress, + admin_level, fromarea, isaddress and linked_place_id is NULL as isaddress, CASE WHEN address_place_id = for_place_id AND rank_address = 0 THEN 100 WHEN rank_address = 11 THEN 5 ELSE rank_address END as rank_address, distance,country_code,postcode from place_addressline join placex on (address_place_id = placex.place_id) where place_addressline.place_id = for_place_id and (cached_rank_address > 0 AND cached_rank_address < searchrankaddress) - and address_place_id != for_place_id + and address_place_id != for_place_id and linked_place_id is null and (placex.country_code IS NULL OR searchcountrycode IS NULL OR placex.country_code = searchcountrycode) order by rank_address desc,isaddress desc,fromarea desc,distance asc,rank_search desc LOOP @@ -2498,12 +2475,9 @@ BEGIN IF searchcountrycode IS NULL AND location.country_code IS NOT NULL THEN searchcountrycode := location.country_code; END IF; - IF searchpostcode IS NOT NULL and location.type = 'postcode' THEN + IF location.type in ('postcode', 'postal_code') THEN location.isaddress := FALSE; END IF; - IF searchpostcode IS NULL and location.isaddress and location.type != 'postcode' and location.postcode IS NOT NULL and location.postcode not similar to '%(,|;)%' THEN - searchpostcode := location.postcode; - END IF; IF location.rank_address = 4 AND location.isaddress THEN hadcountry := true; END IF; @@ -2537,7 +2511,6 @@ BEGIN IF searchhousename IS NOT NULL THEN location := ROW(in_place_id, null, null, searchhousename, searchclass, searchtype, null, true, true, 29, 0)::addressline; --- location := ROW(in_place_id, null, null, searchhousename, 'place', 'house_name', null, true, true, 29, 0)::addressline; RETURN NEXT location; END IF;