X-Git-Url: https://git.openstreetmap.org./nominatim.git/blobdiff_plain/49ac49228ee208f7b219ee8ea8e1b199d6486245..115c909fdf65966774940ef957cf0bd63ba844f2:/sql/functions.sql diff --git a/sql/functions.sql b/sql/functions.sql index 691ba85e..820bdbf8 100644 --- a/sql/functions.sql +++ b/sql/functions.sql @@ -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; +$$ +LANGUAGE plpgsql; + CREATE OR REPLACE FUNCTION getorcreate_country(lookup_word TEXT, lookup_country_code varchar(2)) RETURNS INTEGER AS $$ @@ -101,7 +121,7 @@ END; $$ LANGUAGE plpgsql; -CREATE OR REPLACE FUNCTION getorcreate_amenity(lookup_word TEXT, lookup_class text, lookup_type text) +CREATE OR REPLACE FUNCTION getorcreate_amenity(lookup_word TEXT, normalized_word TEXT, lookup_class text, lookup_type text) RETURNS INTEGER AS $$ DECLARE @@ -109,17 +129,17 @@ DECLARE return_word_id INTEGER; BEGIN lookup_token := ' '||trim(lookup_word); - SELECT min(word_id) FROM word WHERE word_token = lookup_token and class=lookup_class and type = lookup_type into return_word_id; + SELECT min(word_id) FROM word WHERE word_token = lookup_token and word=normalized_word and class=lookup_class and type = lookup_type 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, null, lookup_class, lookup_type, null, 0); + INSERT INTO word VALUES (return_word_id, lookup_token, normalized_word, lookup_class, lookup_type, null, 0); END IF; RETURN return_word_id; END; $$ LANGUAGE plpgsql; -CREATE OR REPLACE FUNCTION getorcreate_amenityoperator(lookup_word TEXT, lookup_class text, lookup_type text, op text) +CREATE OR REPLACE FUNCTION getorcreate_amenityoperator(lookup_word TEXT, normalized_word TEXT, lookup_class text, lookup_type text, op text) RETURNS INTEGER AS $$ DECLARE @@ -127,10 +147,10 @@ DECLARE return_word_id INTEGER; BEGIN lookup_token := ' '||trim(lookup_word); - SELECT min(word_id) FROM word WHERE word_token = lookup_token and class=lookup_class and type = lookup_type and operator = op into return_word_id; + SELECT min(word_id) FROM word WHERE word_token = lookup_token and word=normalized_word and class=lookup_class and type = lookup_type and operator = op 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, null, lookup_class, lookup_type, null, 0, op); + INSERT INTO word VALUES (return_word_id, lookup_token, normalized_word, lookup_class, lookup_type, null, 0, op); END IF; RETURN return_word_id; END; @@ -236,6 +256,99 @@ END; $$ LANGUAGE plpgsql IMMUTABLE; + +CREATE OR REPLACE FUNCTION get_postcode_rank(country_code VARCHAR(2), postcode TEXT, + OUT rank_search SMALLINT, OUT rank_address SMALLINT) +AS $$ +DECLARE + part TEXT; +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])$' THEN + rank_search := 25; + rank_address := 5; + ELSEIF postcode ~ '^([A-Z][A-Z]?[0-9][0-9A-Z]? [0-9])$' THEN + rank_search := 23; + rank_address := 5; + ELSEIF postcode ~ '^([A-Z][A-Z]?[0-9][0-9A-Z])$' THEN + rank_search := 21; + rank_address := 5; + END IF; + + ELSEIF country_code = 'sg' THEN + IF postcode ~ '^([0-9]{6})$' THEN + rank_search := 25; + rank_address := 11; + END IF; + + ELSEIF country_code = 'de' THEN + IF postcode ~ '^([0-9]{5})$' THEN + rank_search := 21; + rank_address := 11; + END IF; + + ELSE + -- Guess at the postcode format and coverage (!) + IF postcode ~ '^[A-Z0-9]{1,5}$' THEN -- Probably too short to be very local + rank_search := 21; + rank_address := 11; + ELSE + -- Does it look splitable into and area and local code? + part := substring(postcode from '^([- :A-Z0-9]+)([- :][A-Z0-9]+)$'); + + IF part IS NOT NULL THEN + rank_search := 25; + rank_address := 11; + ELSEIF postcode ~ '^[- :A-Z0-9]{6,}$' THEN + rank_search := 21; + rank_address := 11; + END IF; + END IF; + END IF; + +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; +$$ +LANGUAGE plpgsql; + + CREATE OR REPLACE FUNCTION create_country(src HSTORE, lookup_country_code varchar(2)) RETURNS VOID AS $$ DECLARE @@ -409,7 +522,7 @@ DECLARE BEGIN place_centre := ST_PointOnSurface(place); ---DEBUG: RAISE WARNING 'get_country_code, start: %', ST_AsText(place_centre); +-- RAISE WARNING 'get_country_code, start: %', ST_AsText(place_centre); -- Try for a OSM polygon FOR nearcountry IN select country_code from location_area_country where country_code is not null and not isguess and st_covers(geometry, place_centre) limit 1 @@ -417,7 +530,7 @@ BEGIN RETURN nearcountry.country_code; END LOOP; ---DEBUG: RAISE WARNING 'osm fallback: %', ST_AsText(place_centre); +-- RAISE WARNING 'osm fallback: %', ST_AsText(place_centre); -- Try for OSM fallback data -- The order is to deal with places like HongKong that are 'states' within another polygon @@ -426,7 +539,7 @@ BEGIN RETURN nearcountry.country_code; END LOOP; ---DEBUG: RAISE WARNING 'natural earth: %', ST_AsText(place_centre); +-- RAISE WARNING 'natural earth: %', ST_AsText(place_centre); -- Natural earth data FOR nearcountry IN select country_code from country_naturalearthdata where st_covers(geometry, place_centre) limit 1 @@ -434,7 +547,7 @@ BEGIN RETURN nearcountry.country_code; END LOOP; ---DEBUG: RAISE WARNING 'near osm fallback: %', ST_AsText(place_centre); +-- RAISE WARNING 'near osm fallback: %', ST_AsText(place_centre); -- FOR nearcountry IN select country_code from country_osm_grid where st_dwithin(geometry, place_centre, 0.5) order by st_distance(geometry, place_centre) asc, area asc limit 1 @@ -442,7 +555,7 @@ BEGIN RETURN nearcountry.country_code; END LOOP; ---DEBUG: RAISE WARNING 'near natural earth: %', ST_AsText(place_centre); +-- RAISE WARNING 'near natural earth: %', ST_AsText(place_centre); -- Natural earth data FOR nearcountry IN select country_code from country_naturalearthdata where st_dwithin(geometry, place_centre, 0.5) limit 1 @@ -515,36 +628,38 @@ 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 @@ -569,7 +684,7 @@ 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; @@ -722,54 +837,10 @@ BEGIN RETURN NULL; END IF; - NEW.postcode := NEW.address->'postcode'; - NEW.name := hstore('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 = 'sg' THEN + NEW.name := hstore('ref', NEW.address->'postcode'); - IF NEW.postcode ~ '^([0-9]{6})$' THEN - NEW.rank_search := 25; - NEW.rank_address := 11; - 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; + SELECT * FROM get_postcode_rank(NEW.country_code, NEW.address->'postcode') + INTO NEW.rank_search, NEW.rank_address; ELSEIF NEW.class = 'place' THEN IF NEW.type in ('continent') THEN @@ -844,6 +915,9 @@ 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; @@ -985,8 +1059,8 @@ DECLARE linegeo GEOMETRY; splitline GEOMETRY; sectiongeo GEOMETRY; + interpol_postcode TEXT; postcode TEXT; - seg_postcode TEXT; BEGIN -- deferred delete IF OLD.indexed_status = 100 THEN @@ -1005,9 +1079,11 @@ 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 @@ -1020,7 +1096,6 @@ BEGIN linegeo := NEW.linegeo; startnumber := NULL; - postcode := NEW.postcode; FOR nodeidpos in 1..array_upper(waynodes, 1) LOOP @@ -1053,15 +1128,24 @@ 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, @@ -1070,7 +1154,7 @@ 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; @@ -1094,7 +1178,42 @@ END; $$ LANGUAGE plpgsql; +-- Trigger for updates of location_postcode +-- +-- Computes the parent object the postcode most likely refers to. +-- This will be the place that determines the address displayed when +-- searching for this postcode. +CREATE OR REPLACE FUNCTION postcode_update() RETURNS +TRIGGER + AS $$ +DECLARE + partition SMALLINT; + location RECORD; +BEGIN + IF NEW.indexed_status != 0 OR OLD.indexed_status = 0 THEN + RETURN NEW; + END IF; + + NEW.indexed_date = now(); + + partition := get_partition(NEW.country_code); + + SELECT * FROM get_postcode_rank(NEW.country_code, NEW.postcode) + INTO NEW.rank_search, NEW.rank_address; + NEW.parent_place_id = 0; + FOR location IN + SELECT place_id + FROM getNearFeatures(partition, NEW.geometry, NEW.rank_search, '{}'::int[]) + WHERE NOT isguess ORDER BY rank_address DESC LIMIT 1 + LOOP + NEW.parent_place_id = location.place_id; + END LOOP; + + RETURN NEW; +END; +$$ +LANGUAGE plpgsql; CREATE OR REPLACE FUNCTION placex_update() RETURNS TRIGGER @@ -1140,12 +1259,14 @@ DECLARE nameaddress_vector INTEGER[]; linked_node_id BIGINT; + linked_importance FLOAT; + linked_wikipedia TEXT; result BOOLEAN; BEGIN -- deferred delete IF OLD.indexed_status = 100 THEN - --DEBUG: RAISE WARNING 'placex_update_delete % %',NEW.osm_type,NEW.osm_id; + --DEBUG: RAISE WARNING 'placex_update delete % %',NEW.osm_type,NEW.osm_id; delete from placex where place_id = OLD.place_id; RETURN NULL; END IF; @@ -1154,15 +1275,7 @@ BEGIN RETURN NEW; END IF; - --DEBUG: RAISE WARNING 'placex_update % %',NEW.osm_type,NEW.osm_id; - ---RAISE WARNING '%',NEW.place_id; ---RAISE WARNING '%', NEW; - - IF NEW.class = 'place' AND NEW.type = 'postcodearea' THEN - -- Silently do nothing - RETURN NEW; - END IF; + --DEBUG: RAISE WARNING 'placex_update % % (%)',NEW.osm_type,NEW.osm_id,NEW.place_id; NEW.indexed_date = now(); @@ -1175,9 +1288,11 @@ BEGIN -- update not necessary for osmline, cause linked_place_id does not exist IF NEW.linked_place_id is not null THEN + --DEBUG: RAISE WARNING 'place already linked to %', NEW.linked_place_id; RETURN NEW; END IF; + --DEBUG: RAISE WARNING 'Copy over address tags'; IF NEW.address is not NULL THEN IF NEW.address ? 'conscriptionnumber' THEN i := getorcreate_housenumber_id(make_standard_name(NEW.address->'conscriptionnumber')); @@ -1198,13 +1313,17 @@ BEGIN 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 IF NEW.rank_search = 4 AND NEW.address is not NULL AND NEW.address ? 'country' THEN @@ -1225,6 +1344,7 @@ BEGIN END IF; NEW.partition := get_partition(NEW.country_code); END IF; + --DEBUG: RAISE WARNING 'Country updated: "%"', NEW.country_code; -- waterway ways are linked when they are part of a relation and have the same class/type IF NEW.osm_type = 'R' and NEW.class = 'waterway' THEN @@ -1232,7 +1352,7 @@ BEGIN LOOP FOR i IN 1..array_upper(relation_members, 1) BY 2 LOOP IF relation_members[i+1] in ('', 'main_stream', 'side_stream') AND substring(relation_members[i],1,1) = 'w' THEN - --DEBUG: RAISE WARNING 'waterway parent %, child %/%', NEW.osm_id, i, relation.members[i]; + --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 @@ -1243,6 +1363,7 @@ BEGIN END IF; END LOOP; END LOOP; + --DEBUG: RAISE WARNING 'Waterway processed'; END IF; -- Adding ourselves to the list simplifies address calculations later @@ -1264,6 +1385,7 @@ BEGIN END IF; END IF; END IF; + --DEBUG: RAISE WARNING 'Local names updated'; -- Initialise the name vector using our name name_vector := make_keywords(NEW.name); @@ -1279,13 +1401,13 @@ BEGIN 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 'before low level% %', NEW.place_id, NEW.rank_search; +--DEBUG: RAISE WARNING 'Importance computed from wikipedia: %', NEW.importance; -- --------------------------------------------------------------------------- -- For low level elements we inherit from our parent road IF (NEW.rank_search > 27 OR (NEW.type = 'postcode' AND NEW.rank_search = 25)) THEN ---RAISE WARNING 'finding street for %', NEW; + --DEBUG: RAISE WARNING 'finding street for % %', NEW.osm_type, NEW.osm_id; -- We won't get a better centroid, besides these places are too small to care NEW.centroid := place_centroid; @@ -1305,6 +1427,7 @@ BEGIN NEW.housenumber := location.address->'housenumber'; addr_street := location.address->'street'; addr_place := location.address->'place'; + --DEBUG: RAISE WARNING 'Found surrounding building % %', location.osm_type, location.osm_id; END LOOP; END IF; @@ -1326,7 +1449,7 @@ BEGIN END LOOP; END IF; END LOOP; - + --DEBUG: RAISE WARNING 'Checked for street relation (%)', NEW.parent_place_id; -- Note that addr:street links can only be indexed once the street itself is indexed IF NEW.parent_place_id IS NULL AND addr_street IS NOT NULL THEN @@ -1337,6 +1460,7 @@ BEGIN END LOOP; END IF; END IF; + --DEBUG: RAISE WARNING 'Checked for addr:street (%)', NEW.parent_place_id; 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)); @@ -1346,6 +1470,7 @@ BEGIN END LOOP; 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 @@ -1357,6 +1482,7 @@ BEGIN NEW.parent_place_id := location.parent_place_id; END LOOP; 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 @@ -1364,12 +1490,14 @@ BEGIN 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) 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 --RAISE WARNING 'node in way that is a street %',location; NEW.parent_place_id := location.place_id; 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 @@ -1380,6 +1508,7 @@ BEGIN END LOOP; END IF; END IF; + --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')); @@ -1389,6 +1518,7 @@ BEGIN END LOOP; END IF; END IF; + --DEBUG: RAISE WARNING 'Checked for addr:place in way (%)', NEW.parent_place_id; -- Is the WAY part of a relation IF NEW.parent_place_id IS NULL THEN @@ -1406,21 +1536,20 @@ BEGIN END IF; END LOOP; END IF; + --DEBUG: RAISE WARNING 'Checked for street relation in way (%)', NEW.parent_place_id; END LOOP; END IF; ---RAISE WARNING 'x4 %',NEW.parent_place_id; -- 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; END IF; + --DEBUG: RAISE WARNING 'Checked for nearest way (%)', NEW.parent_place_id; ---return NEW; ---RAISE WARNING 'x6 %',NEW.parent_place_id; -- If we didn't find any road fallback to standard method IF NEW.parent_place_id IS NOT NULL THEN @@ -1428,30 +1557,23 @@ BEGIN -- Get the details of the parent road select * from search_name where 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 - 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; ---RAISE WARNING '%', NEW.name; -- If there is no name it isn't searchable, don't bother to create a search record IF NEW.name is NULL THEN + --DEBUG: RAISE WARNING 'Not a searchable place % %', NEW.osm_type, NEW.osm_id; return NEW; END IF; @@ -1464,33 +1586,35 @@ 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; result := insertSearchName(NEW.partition, NEW.place_id, NEW.country_code, name_vector, nameaddress_vector, NEW.rank_search, NEW.rank_address, NEW.importance, place_centroid, NEW.geometry); + --DEBUG: RAISE WARNING 'Place added to search table'; return NEW; END IF; END IF; --- RAISE WARNING ' INDEXING Started:'; --- RAISE WARNING ' INDEXING: %',NEW; - -- --------------------------------------------------------------------------- -- Full indexing + --DEBUG: RAISE WARNING 'Using full index mode for % %', NEW.osm_type, NEW.osm_id; 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; + --DEBUG: RAISE WARNING 'Got relation members'; --- RAISE WARNING 'get_osm_rel_members, label'; IF relation_members IS NOT NULL THEN FOR relMember IN select get_osm_rel_members(relation_members,ARRAY['label']) as member LOOP + --DEBUG: RAISE WARNING 'Found label member %', relMember.member; FOR linkedPlacex IN select * from placex where osm_type = upper(substring(relMember.member,1,1))::char(1) - and osm_id = substring(relMember.member,2,10000)::bigint order by rank_search desc limit 1 LOOP + and osm_id = substring(relMember.member,2,10000)::bigint + and class = 'place' order by rank_search desc limit 1 LOOP -- If we don't already have one use this as the centre point of the geometry IF NEW.centroid IS NULL THEN @@ -1511,6 +1635,8 @@ BEGIN -- keep a note of the node id in case we need it for wikipedia in a bit linked_node_id := linkedPlacex.osm_id; + select language||':'||title,importance from get_wikipedia_match(linkedPlacex.extratags, NEW.country_code) INTO linked_wikipedia,linked_importance; + --DEBUG: RAISE WARNING 'Linked label member'; END LOOP; END LOOP; @@ -1518,9 +1644,11 @@ BEGIN IF NEW.centroid IS NULL THEN FOR relMember IN select get_osm_rel_members(relation_members,ARRAY['admin_center','admin_centre']) as member LOOP + --DEBUG: RAISE WARNING 'Found admin_center member %', relMember.member; FOR linkedPlacex IN select * from placex where osm_type = upper(substring(relMember.member,1,1))::char(1) - and osm_id = substring(relMember.member,2,10000)::bigint order by rank_search desc limit 1 LOOP + and osm_id = substring(relMember.member,2,10000)::bigint + and class = 'place' order by rank_search desc limit 1 LOOP -- 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 @@ -1546,6 +1674,8 @@ BEGIN -- keep a note of the node id in case we need it for wikipedia in a bit linked_node_id := linkedPlacex.osm_id; + select language||':'||title,importance from get_wikipedia_match(linkedPlacex.extratags, NEW.country_code) INTO linked_wikipedia,linked_importance; + --DEBUG: RAISE WARNING 'Linked admin_center'; END IF; END LOOP; @@ -1563,6 +1693,7 @@ BEGIN -- 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 + --DEBUG: RAISE WARNING 'Looking for nodes with matching names'; FOR linkedPlacex IN select placex.* from placex WHERE make_standard_name(name->'name') = make_standard_name(NEW.name->'name') AND placex.rank_address = NEW.rank_address @@ -1570,7 +1701,7 @@ BEGIN AND placex.osm_type = 'N'::char(1) AND placex.rank_search < 26 AND st_covers(NEW.geometry, placex.geometry) LOOP - + --DEBUG: RAISE WARNING 'Found matching place node %', linkedPlacex.osm_id; -- 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)); @@ -1588,6 +1719,8 @@ BEGIN -- keep a note of the node id in case we need it for wikipedia in a bit linked_node_id := linkedPlacex.osm_id; + select language||':'||title,importance from get_wikipedia_match(linkedPlacex.extratags, NEW.country_code) INTO linked_wikipedia,linked_importance; + --DEBUG: RAISE WARNING 'Linked named place'; END LOOP; END IF; @@ -1606,14 +1739,18 @@ BEGIN END IF; END IF; END IF; + --DEBUG: RAISE WARNING 'Names updated from linked places'; 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, NEW.country_code) INTO NEW.wikipedia,NEW.importance; + -- Use the maximum importance if a one could be computed from the linked object. + IF linked_importance is not null AND + (NEW.importance is null or NEW.importance < linked_importance) THEN + NEW.importance = linked_importance; END IF; + -- Still null? how about looking it up by the node id IF NEW.importance IS NULL THEN + --DEBUG: RAISE WARNING 'Looking up importance by linked node id'; select language||':'||title,importance from wikipedia_article where osm_type = 'N'::char(1) and osm_id = linked_node_id order by importance desc limit 1 INTO NEW.wikipedia,NEW.importance; END IF; @@ -1622,6 +1759,7 @@ BEGIN -- 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 perform create_country(NEW.name, lower(NEW.country_code)); + --DEBUG: RAISE WARNING 'Country names updated'; END IF; NEW.parent_place_id = 0; @@ -1629,11 +1767,13 @@ BEGIN -- convert isin 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 + -- TODO further split terms with comma and semicolon 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]); @@ -1648,24 +1788,6 @@ BEGIN END LOOP; END IF; END IF; - 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; - - -- 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]); - END IF; - END LOOP; - END IF; - END IF; -- %NOTIGERDATA% IF 0 THEN -- for the USA we have an additional address table. Merge in zip codes from there too @@ -1680,6 +1802,7 @@ BEGIN nameaddress_vector := array_merge(nameaddress_vector, ARRAY[address_street_word_id]); END LOOP; END IF; + --DEBUG: RAISE WARNING 'Tiger postcodes collected'; -- %NOTIGERDATA% END IF; -- RAISE WARNING 'ISIN: %', isin_tokens; @@ -1690,7 +1813,7 @@ BEGIN location_parent := NULL; -- added ourself as address already address_havelevel[NEW.rank_address] := true; - -- RAISE WARNING ' getNearFeatures(%,''%'',%,''%'')',NEW.partition, place_centroid, search_maxrank, isin_tokens; + --DEBUG: RAISE WARNING ' getNearFeatures(%,''%'',%,''%'')',NEW.partition, place_centroid, search_maxrank, isin_tokens; FOR location IN SELECT * from getNearFeatures(NEW.partition, CASE WHEN NEW.rank_search >= 26 @@ -1699,9 +1822,6 @@ BEGIN ELSE place_centroid END, search_maxrank, isin_tokens) LOOP - ---RAISE WARNING ' AREA: %',location; - IF location.rank_address != location_rank_search THEN location_rank_search := location.rank_address; IF location.isguess THEN @@ -1739,6 +1859,11 @@ 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 @@ -1752,11 +1877,12 @@ BEGIN END IF; ---RAISE WARNING ' Terms: (%) %',location, nameaddress_vector; + --DEBUG: RAISE WARNING ' Terms: (%) %',location, nameaddress_vector; END IF; 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 @@ -1772,6 +1898,11 @@ BEGIN 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); + IF NEW.postcode is null AND location.postcode is not null + AND NOT address_havelevel[location.rank_address] THEN + NEW.postcode := location.postcode; + END IF; + address_havelevel[location.rank_address] := true; IF location.rank_address > parent_place_id_rank THEN @@ -1785,6 +1916,7 @@ BEGIN END LOOP; 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 @@ -1811,19 +1943,32 @@ BEGIN END LOOP; END IF; + --DEBUG: RAISE WARNING 'search terms for long ways added'; + + 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; + + IF NEW.postcode is null AND NEW.rank_search > 8 THEN + NEW.postcode := get_nearest_postcode(NEW.country_code, NEW.geometry); + END IF; -- 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; IF NEW.rank_search between 26 and 27 and NEW.class = 'highway' THEN result := insertLocationRoad(NEW.partition, NEW.place_id, NEW.country_code, NEW.geometry); + --DEBUG: RAISE WARNING 'insert into road location table (full)'; END IF; result := insertSearchName(NEW.partition, NEW.place_id, NEW.country_code, name_vector, nameaddress_vector, NEW.rank_search, NEW.rank_address, NEW.importance, place_centroid, NEW.geometry); + --DEBUG: RAISE WARNING 'added to serach name (full)'; END IF; @@ -1831,7 +1976,9 @@ BEGIN IF NEW.centroid IS NULL THEN NEW.centroid := place_centroid; END IF; - + + --DEBUG: RAISE WARNING 'place update % % finsihed.', NEW.osm_type, NEW.osm_id; + RETURN NEW; END; $$ @@ -2137,6 +2284,12 @@ BEGIN NEW.name := hstore('ref', NEW.address->'postcode'); END IF; + IF NEW.class in ('boundary') + AND ST_GeometryType(NEW.geometry) not in ('ST_Polygon','ST_MultiPolygon') THEN + DELETE FROM placex where place_id = existingplacex.place_id; + RETURN NULL; + END IF; + update placex set name = NEW.name, address = NEW.address, @@ -2154,6 +2307,20 @@ BEGIN update location_property_osmline p set indexed_status = 2 from planet_osm_ways w where p.linegeo && NEW.geometry and p.osm_id = w.id and NEW.osm_id = any(w.nodes); END IF; + -- linked places should get potential new naming and addresses + IF existingplacex.linked_place_id is not NULL THEN + update placex x set + name = p.name, + extratags = p.extratags, + indexed_status = 2 + from place p + where x.place_id = existingplacex.linked_place_id + and x.indexed_status = 0 + and x.osm_type = p.osm_type + and x.osm_id = p.osm_id + and x.class = p.class; + END IF; + END IF; -- Abort the add (we modified the existing place instead) @@ -2188,46 +2355,6 @@ END; $$ 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 $$ @@ -2311,11 +2438,19 @@ 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 @@ -2334,9 +2469,8 @@ 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 @@ -2346,13 +2480,9 @@ 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 @@ -2363,15 +2493,14 @@ 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 and linked_place_id is NULL as isaddress, @@ -2388,12 +2517,9 @@ 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 THEN - searchpostcode := location.postcode; - END IF; IF location.rank_address = 4 AND location.isaddress THEN hadcountry := true; END IF; @@ -2427,7 +2553,6 @@ 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; @@ -2573,9 +2698,7 @@ BEGIN IF out_postcode IS NULL THEN 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(out_partition, place_centroid); - END IF; + -- XXX look into postcode table newpoints := 0; insert into location_property_aux (place_id, partition, parent_place_id, housenumber, postcode, centroid) @@ -2833,7 +2956,7 @@ BEGIN IF ST_GeometryType(placegeom) in ('ST_Polygon','ST_MultiPolygon') THEN FOR geom IN select split_geometry(placegeom) FROM placex WHERE place_id = placeid LOOP update placex set indexed_status = 2 where (st_covers(geom, placex.geometry) OR ST_Intersects(geom, placex.geometry)) - AND rank_search > rank and indexed_status = 0 and ST_geometrytype(placex.geometry) = 'ST_Point' and (rank_search < 28 or name is not null or (rank >= 16 and address > 'place')); + AND rank_search > rank and indexed_status = 0 and ST_geometrytype(placex.geometry) = 'ST_Point' and (rank_search < 28 or name is not null or (rank >= 16 and address ? 'place')); update placex set indexed_status = 2 where (st_covers(geom, placex.geometry) OR ST_Intersects(geom, placex.geometry)) AND rank_search > rank and indexed_status = 0 and ST_geometrytype(placex.geometry) != 'ST_Point' and (rank_search < 28 or name is not null or (rank >= 16 and address ? 'place')); END LOOP;