X-Git-Url: https://git.openstreetmap.org./nominatim.git/blobdiff_plain/b99a043c78f756f58b8e106ba60e999ea9538bb3..d6f4448ed33298e563c520fc654fe5d445ff7f57:/sql/functions.sql diff --git a/sql/functions.sql b/sql/functions.sql index 9a613152..d86f4a9b 100644 --- a/sql/functions.sql +++ b/sql/functions.sql @@ -5,7 +5,6 @@ -- score integer --); - CREATE OR REPLACE FUNCTION getclasstypekey(c text, t text) RETURNS TEXT AS $$ DECLARE @@ -52,35 +51,14 @@ DECLARE NEWgeometry geometry; BEGIN -- RAISE WARNING '%',place; - NEWgeometry := place; - IF ST_IsEmpty(NEWgeometry) OR NOT ST_IsValid(NEWgeometry) OR ST_X(ST_Centroid(NEWgeometry))::text in ('NaN','Infinity','-Infinity') OR ST_Y(ST_Centroid(NEWgeometry))::text in ('NaN','Infinity','-Infinity') THEN - NEWgeometry := ST_buffer(NEWgeometry,0); - IF ST_IsEmpty(NEWgeometry) OR NOT ST_IsValid(NEWgeometry) OR ST_X(ST_Centroid(NEWgeometry))::text in ('NaN','Infinity','-Infinity') OR ST_Y(ST_Centroid(NEWgeometry))::text in ('NaN','Infinity','-Infinity') THEN - RETURN 0; - END IF; - END IF; - RETURN (partition*1000000) + (500-ST_X(ST_Centroid(NEWgeometry))::integer)*1000 + (500-ST_Y(ST_Centroid(NEWgeometry))::integer); -END; -$$ -LANGUAGE plpgsql IMMUTABLE; - -CREATE OR REPLACE FUNCTION debug_geometry_sector(osmid integer, place geometry) RETURNS INTEGER - AS $$ -DECLARE - NEWgeometry geometry; -BEGIN --- RAISE WARNING '%',osmid; - IF osmid = 61315 THEN - return null; - END IF; - NEWgeometry := place; - IF ST_IsEmpty(NEWgeometry) OR NOT ST_IsValid(NEWgeometry) OR ST_X(ST_Centroid(NEWgeometry))::text in ('NaN','Infinity','-Infinity') OR ST_Y(ST_Centroid(NEWgeometry))::text in ('NaN','Infinity','-Infinity') THEN - NEWgeometry := ST_buffer(NEWgeometry,0); - IF ST_IsEmpty(NEWgeometry) OR NOT ST_IsValid(NEWgeometry) OR ST_X(ST_Centroid(NEWgeometry))::text in ('NaN','Infinity','-Infinity') OR ST_Y(ST_Centroid(NEWgeometry))::text in ('NaN','Infinity','-Infinity') THEN - RETURN NULL; - END IF; - END IF; - RETURN (500-ST_X(ST_Centroid(NEWgeometry))::integer)*1000 + (500-ST_Y(ST_Centroid(NEWgeometry))::integer); + NEWgeometry := ST_PointOnSurface(place); +-- IF ST_IsEmpty(NEWgeometry) OR NOT ST_IsValid(NEWgeometry) OR ST_X(ST_Centroid(NEWgeometry))::text in ('NaN','Infinity','-Infinity') OR ST_Y(ST_Centroid(NEWgeometry))::text in ('NaN','Infinity','-Infinity') THEN +-- NEWgeometry := ST_buffer(NEWgeometry,0); +-- IF ST_IsEmpty(NEWgeometry) OR NOT ST_IsValid(NEWgeometry) OR ST_X(ST_Centroid(NEWgeometry))::text in ('NaN','Infinity','-Infinity') OR ST_Y(ST_Centroid(NEWgeometry))::text in ('NaN','Infinity','-Infinity') THEN +-- RETURN 0; +-- END IF; +-- END IF; + RETURN (partition*1000000) + (500-ST_X(NEWgeometry)::integer)*1000 + (500-ST_Y(NEWgeometry)::integer); END; $$ LANGUAGE plpgsql IMMUTABLE; @@ -104,18 +82,24 @@ END; $$ LANGUAGE 'plpgsql' IMMUTABLE; +-- returns NULL if the word is too common CREATE OR REPLACE FUNCTION getorcreate_word_id(lookup_word TEXT) RETURNS INTEGER AS $$ DECLARE lookup_token TEXT; return_word_id INTEGER; + count INTEGER; BEGIN lookup_token := trim(lookup_word); - SELECT min(word_id) FROM word WHERE word_token = lookup_token and class is null and type is null into return_word_id; + SELECT min(word_id), max(search_name_count) FROM word WHERE word_token = lookup_token and class is null and type is null into return_word_id, count; IF return_word_id IS NULL THEN return_word_id := nextval('seq_word'); - INSERT INTO word VALUES (return_word_id, lookup_token, regexp_replace(lookup_token,E'([^0-9])\\1+',E'\\1','g'), null, null, null, null, 0, null); + INSERT INTO word VALUES (return_word_id, lookup_token, null, null, null, null, 0); + ELSE + IF count > get_maxwordfreq() THEN + return_word_id := NULL; + END IF; END IF; RETURN return_word_id; END; @@ -133,7 +117,7 @@ BEGIN SELECT min(word_id) FROM word WHERE word_token = lookup_token and class='place' and type='house' 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, null, 'place', 'house', null, 0, null); + INSERT INTO word VALUES (return_word_id, lookup_token, null, 'place', 'house', null, 0); END IF; RETURN return_word_id; END; @@ -151,7 +135,7 @@ BEGIN SELECT min(word_id) FROM word WHERE word_token = lookup_token and country_code=lookup_country_code 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, null, null, null, lookup_country_code, 0, null); + INSERT INTO word VALUES (return_word_id, lookup_token, null, null, null, lookup_country_code, 0); END IF; RETURN return_word_id; END; @@ -169,7 +153,7 @@ BEGIN SELECT min(word_id) FROM word WHERE word_token = lookup_token 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, null, lookup_class, lookup_type, null, 0, null); + INSERT INTO word VALUES (return_word_id, lookup_token, null, lookup_class, lookup_type, null, 0); END IF; RETURN return_word_id; END; @@ -187,7 +171,7 @@ BEGIN SELECT min(word_id) FROM word WHERE word_token = lookup_token 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, null, null, null, null, 0, null); + INSERT INTO word VALUES (return_word_id, lookup_token, null, null, null, null, 0); END IF; RETURN return_word_id; END; @@ -219,7 +203,7 @@ BEGIN 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; IF return_word_id IS NULL THEN return_word_id := nextval('seq_word'); - INSERT INTO word VALUES (return_word_id, lookup_token, null, null, lookup_class, lookup_type, null, 0, op, null); + INSERT INTO word VALUES (return_word_id, lookup_token, null, lookup_class, lookup_type, null, 0, op); END IF; RETURN return_word_id; END; @@ -238,7 +222,7 @@ BEGIN SELECT min(word_id) FROM word WHERE word_token = lookup_token and class is null and type is null 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, regexp_replace(lookup_token,E'([^0-9])\\1+',E'\\1','g'), src_word, null, null, null, 0, null); + INSERT INTO word VALUES (return_word_id, lookup_token, src_word, null, null, null, 0); -- nospace_lookup_token := replace(replace(lookup_token, '-',''), ' ',''); -- IF ' '||nospace_lookup_token != lookup_token THEN -- INSERT INTO word VALUES (return_word_id, '-'||nospace_lookup_token, null, src_word, null, null, null, 0, null); @@ -287,6 +271,20 @@ END; $$ LANGUAGE plpgsql IMMUTABLE; +CREATE OR REPLACE FUNCTION get_name_ids(lookup_word TEXT) + RETURNS INTEGER[] + AS $$ +DECLARE + lookup_token TEXT; + return_word_ids INTEGER[]; +BEGIN + lookup_token := ' '||trim(lookup_word); + SELECT array_agg(word_id) FROM word WHERE word_token = lookup_token and class is null and type is null into return_word_ids; + RETURN return_word_ids; +END; +$$ +LANGUAGE plpgsql IMMUTABLE; + CREATE OR REPLACE FUNCTION array_merge(a INTEGER[], b INTEGER[]) RETURNS INTEGER[] AS $$ @@ -311,6 +309,34 @@ END; $$ LANGUAGE plpgsql IMMUTABLE; +CREATE OR REPLACE FUNCTION create_country(src HSTORE, lookup_country_code varchar(2)) RETURNS VOID + AS $$ +DECLARE + s TEXT; + w INTEGER; + words TEXT[]; + item RECORD; + j INTEGER; +BEGIN + FOR item IN SELECT (each(src)).* LOOP + + s := make_standard_name(item.value); + w := getorcreate_country(s, lookup_country_code); + + words := regexp_split_to_array(item.value, 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_country(s, lookup_country_code); + END IF; + END LOOP; + END IF; + END LOOP; +END; +$$ +LANGUAGE plpgsql; + CREATE OR REPLACE FUNCTION make_keywords(src HSTORE) RETURNS INTEGER[] AS $$ DECLARE @@ -333,12 +359,18 @@ BEGIN result := result || w; END IF; + w := getorcreate_word_id(s); + + IF w IS NOT NULL AND NOT (ARRAY[w] <@ result) THEN + result := result || w; + END IF; + words := string_to_array(s, ' '); IF array_upper(words, 1) IS NOT NULL THEN FOR j IN 1..array_upper(words, 1) LOOP IF (words[j] != '') THEN w = getorcreate_word_id(words[j]); - IF NOT (ARRAY[w] <@ result) THEN + IF w IS NOT NULL AND NOT (ARRAY[w] <@ result) THEN result := result || w; END IF; END IF; @@ -351,7 +383,7 @@ BEGIN s := make_standard_name(words[j]); IF s != '' THEN w := getorcreate_word_id(s); - IF NOT (ARRAY[w] <@ result) THEN + IF w IS NOT NULL AND NOT (ARRAY[w] <@ result) THEN result := result || w; END IF; END IF; @@ -389,24 +421,54 @@ 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; END IF; + w := getorcreate_word_id(s); + + IF w IS NOT NULL AND NOT (ARRAY[w] <@ result) THEN + result := result || w; + END IF; + words := string_to_array(s, ' '); IF array_upper(words, 1) IS NOT NULL THEN FOR j IN 1..array_upper(words, 1) LOOP IF (words[j] != '') THEN w = getorcreate_word_id(words[j]); - IF NOT (ARRAY[w] <@ result) THEN + IF w IS NOT NULL AND NOT (ARRAY[w] <@ result) THEN + result := result || w; + END IF; + END IF; + 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 w IS NOT NULL AND 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; $$ @@ -443,92 +505,100 @@ DECLARE place_centre GEOMETRY; nearcountry RECORD; BEGIN - place_centre := ST_Centroid(place); + place_centre := ST_PointOnSurface(place); ---RAISE WARNING 'start: %', ST_AsText(place_centre); +--DEBUG: RAISE WARNING 'get_country_code, start: %', ST_AsText(place_centre); - -- Try for a OSM polygon first - FOR nearcountry IN select country_code from location_area_country where country_code is not null and not isguess and st_contains(geometry, place_centre) limit 1 + -- 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 LOOP RETURN nearcountry.country_code; END LOOP; ---RAISE WARNING 'osm fallback: %', ST_AsText(place_centre); +--DEBUG: RAISE WARNING 'osm fallback: %', ST_AsText(place_centre); -- Try for OSM fallback data - FOR nearcountry IN select country_code from country_osm_grid where st_contains(geometry, place_centre) limit 1 + -- The order is to deal with places like HongKong that are 'states' within another polygon + FOR nearcountry IN select country_code from country_osm_grid where st_covers(geometry, place_centre) order by area asc limit 1 LOOP RETURN nearcountry.country_code; END LOOP; ---RAISE WARNING 'natural earth: %', ST_AsText(place_centre); +--DEBUG: RAISE WARNING 'natural earth: %', ST_AsText(place_centre); - -- Natural earth data (first fallback) - FOR nearcountry IN select country_code from country_naturalearthdata where st_contains(geometry, place_centre) limit 1 + -- Natural earth data + FOR nearcountry IN select country_code from country_naturalearthdata where st_covers(geometry, place_centre) limit 1 LOOP RETURN nearcountry.country_code; END LOOP; - -- Natural earth data (first fallback) - FOR nearcountry IN select country_code from country_naturalearthdata where st_distance(geometry, place_centre) < 0.5 limit 1 +--DEBUG: 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 LOOP RETURN nearcountry.country_code; END LOOP; ---RAISE WARNING 'in country: %', ST_AsText(place_centre); +--DEBUG: RAISE WARNING 'near natural earth: %', ST_AsText(place_centre); - -- WorldBoundaries data (second fallback - think there might be something broken in this data) - FOR nearcountry IN select country_code from country where st_contains(geometry, place_centre) limit 1 + -- Natural earth data + FOR nearcountry IN select country_code from country_naturalearthdata where st_dwithin(geometry, place_centre, 0.5) limit 1 LOOP RETURN nearcountry.country_code; END LOOP; + -- WorldBoundaries data (second fallback - think there might be something broken in this data) +-- FOR nearcountry IN select country_code from country where st_covers(geometry, place_centre) limit 1 +-- LOOP +-- RETURN nearcountry.country_code; +-- END LOOP; + --RAISE WARNING 'near country: %', ST_AsText(place_centre); -- Still not in a country - try nearest within ~12 miles of a country - FOR nearcountry IN select country_code from country where st_distance(geometry, place_centre) < 0.5 - order by st_distance(geometry, place) limit 1 - LOOP - RETURN nearcountry.country_code; - END LOOP; +-- FOR nearcountry IN select country_code from country where st_distance(geometry, place_centre) < 0.5 +-- order by st_distance(geometry, place) limit 1 +-- LOOP +-- RETURN nearcountry.country_code; +-- END LOOP; RETURN NULL; END; $$ LANGUAGE plpgsql IMMUTABLE; -CREATE OR REPLACE FUNCTION get_country_code(place geometry, in_country_code VARCHAR(2)) RETURNS TEXT +CREATE OR REPLACE FUNCTION get_country_language_code(search_country_code VARCHAR(2)) RETURNS TEXT AS $$ DECLARE nearcountry RECORD; BEGIN - FOR nearcountry IN select country_code from country_name where country_code = lower(in_country_code) + FOR nearcountry IN select distinct country_default_language_code from country_name where country_code = search_country_code limit 1 LOOP - RETURN nearcountry.country_code; + RETURN lower(nearcountry.country_default_language_code); END LOOP; - RETURN get_country_code(place); + RETURN NULL; END; $$ LANGUAGE plpgsql IMMUTABLE; -CREATE OR REPLACE FUNCTION get_country_language_code(search_country_code VARCHAR(2)) RETURNS TEXT +CREATE OR REPLACE FUNCTION get_country_language_codes(search_country_code VARCHAR(2)) RETURNS TEXT[] AS $$ DECLARE nearcountry RECORD; BEGIN - FOR nearcountry IN select distinct country_default_language_code from country_name where country_code = search_country_code limit 1 + FOR nearcountry IN select country_default_language_codes from country_name where country_code = search_country_code limit 1 LOOP - RETURN lower(nearcountry.country_default_language_code); + RETURN lower(nearcountry.country_default_language_codes); END LOOP; RETURN NULL; END; $$ LANGUAGE plpgsql IMMUTABLE; -CREATE OR REPLACE FUNCTION get_partition(place geometry, in_country_code VARCHAR(10)) RETURNS INTEGER +CREATE OR REPLACE FUNCTION get_partition(in_country_code VARCHAR(10)) RETURNS INTEGER AS $$ DECLARE - place_centre GEOMETRY; nearcountry RECORD; BEGIN FOR nearcountry IN select partition from country_name where country_code = in_country_code @@ -565,17 +635,11 @@ CREATE OR REPLACE FUNCTION add_location( DECLARE locationid INTEGER; isarea BOOLEAN; - xmin INTEGER; - ymin INTEGER; - xmax INTEGER; - ymax INTEGER; - lon INTEGER; - lat INTEGER; centroid GEOMETRY; - secgeo GEOMETRY; - secbox GEOMETRY; diameter FLOAT; x BOOLEAN; + splitGeom RECORD; + secgeo GEOMETRY; BEGIN IF rank_search > 25 THEN @@ -584,7 +648,7 @@ BEGIN -- RAISE WARNING 'Adding location with rank > 25 (% rank %)', place_id, rank_search; - x := deleteLocationArea(partition, place_id); + x := deleteLocationArea(partition, place_id, rank_search); isarea := false; IF (ST_GeometryType(geometry) in ('ST_Polygon','ST_MultiPolygon') AND ST_IsValid(geometry)) THEN @@ -592,27 +656,9 @@ BEGIN isArea := true; centroid := ST_Centroid(geometry); - xmin := floor(st_xmin(geometry)); - xmax := ceil(st_xmax(geometry)); - ymin := floor(st_ymin(geometry)); - ymax := ceil(st_ymax(geometry)); - - IF xmin = xmax OR ymin = ymax OR (xmax-xmin < 2 AND ymax-ymin < 2) THEN - x := insertLocationAreaLarge(partition, place_id, country_code, keywords, rank_search, rank_address, false, centroid, geometry); - ELSE --- RAISE WARNING 'Spliting geometry: % to %, % to %', xmin, xmax, ymin, ymax; - FOR lon IN xmin..(xmax-1) LOOP - FOR lat IN ymin..(ymax-1) LOOP - secbox := ST_SetSRID(ST_MakeBox2D(ST_Point(lon,lat),ST_Point(lon+1,lat+1)),4326); - IF st_intersects(geometry, secbox) THEN - secgeo := st_intersection(geometry, secbox); - IF NOT ST_IsEmpty(secgeo) AND ST_GeometryType(secgeo) in ('ST_Polygon','ST_MultiPolygon') THEN - x := insertLocationAreaLarge(partition, place_id, country_code, keywords, rank_search, rank_address, false, centroid, secgeo); - END IF; - END IF; - END LOOP; - END LOOP; - END IF; + 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); + END LOOP; ELSEIF rank_search < 26 THEN @@ -669,7 +715,7 @@ CREATE OR REPLACE FUNCTION update_location( DECLARE b BOOLEAN; BEGIN - b := deleteLocationArea(partition, place_id); + b := deleteLocationArea(partition, place_id, rank_search); -- result := add_location(NEW.place_id, NEW.country_code, NEW.partition, name_vector, NEW.rank_search, NEW.rank_address, NEW.geometry); RETURN add_location(place_id, place_country_code, name, rank_search, rank_address, geometry); END; @@ -742,13 +788,13 @@ $$ 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 newpoints INTEGER; - waynodes integer[]; - nodeid INTEGER; + waynodes BIGINT[]; + nodeid BIGINT; prevnode RECORD; nextnode RECORD; startnumber INTEGER; @@ -776,11 +822,22 @@ BEGIN FOR nodeidpos in 1..array_upper(waynodes, 1) LOOP - select min(place_id) from placex where osm_type = 'N' and osm_id = waynodes[nodeidpos]::INTEGER and type = 'house' INTO search_place_id; + -- If there is a place of a type other than place/house, use that because + -- it is guaranteed to be the original node. For place/house types use the + -- one with the smallest id because the original node was created first. + -- Ignore all nodes marked for deletion. (Might happen when the type changes.) + select place_id from placex where osm_type = 'N' and osm_id = waynodes[nodeidpos]::BIGINT and indexed_status < 100 order by (type = 'house'),place_id limit 1 INTO search_place_id; IF search_place_id IS NULL THEN - -- null record of right type - select * from placex where osm_type = 'N' and osm_id = waynodes[nodeidpos]::INTEGER and type = 'house' limit 1 INTO nextnode; + -- if no such node exists, create a record of the right type + select * from placex where osm_type = 'N' and osm_id = waynodes[nodeidpos]::BIGINT and type = 'house' limit 1 INTO nextnode; select ST_SetSRID(ST_Point(lon::float/10000000,lat::float/10000000),4326) from planet_osm_nodes where id = waynodes[nodeidpos] INTO nextnode.geometry; + IF nextnode.geometry IS NULL THEN + -- we don't have any information about this point, most likely + -- because an excerpt was updated and the node never imported + -- because the interpolation is outside the region of the excerpt. + -- Give up. + RETURN newpoints; + END IF; ELSE select * from placex where place_id = search_place_id INTO nextnode; END IF; @@ -834,10 +891,10 @@ BEGIN FOR housenum IN startnumber..endnumber BY stepsize LOOP -- this should really copy postcodes but it puts a huge burdon on the system for no big benefit -- ideally postcodes should move up to the way - insert into placex (osm_type, osm_id, class, type, admin_level, housenumber, street, isin, postcode, + insert into placex (osm_type, osm_id, class, type, admin_level, housenumber, street, addr_place, isin, postcode, country_code, parent_place_id, rank_address, rank_search, indexed_status, geometry) - values ('N',prevnode.osm_id, prevnode.class, prevnode.type, prevnode.admin_level, housenum, prevnode.street, prevnode.isin, coalesce(prevnode.postcode, defpostalcode), - prevnode.country_code, prevnode.parent_place_id, prevnode.rank_address, prevnode.rank_search, 1, ST_Line_Interpolate_Point(linegeo, (housenum::float-orginalstartnumber::float)/originalnumberrange::float)); + values ('N',prevnode.osm_id, 'place', 'house', prevnode.admin_level, housenum, prevnode.street, prevnode.addr_place, prevnode.isin, coalesce(prevnode.postcode, defpostalcode), + prevnode.country_code, prevnode.parent_place_id, prevnode.rank_address, prevnode.rank_search, 1, ST_LineInterpolatePoint(linegeo, (housenum::float-orginalstartnumber::float)/originalnumberrange::float)); newpoints := newpoints + 1; --RAISE WARNING 'interpolation number % % ',prevnode.place_id,housenum; END LOOP; @@ -877,13 +934,14 @@ DECLARE diameter FLOAT; classtable TEXT; BEGIN --- RAISE WARNING '%',NEW.osm_id; + --DEBUG: RAISE WARNING '% %',NEW.osm_type,NEW.osm_id; - -- just block these - IF NEW.class = 'highway' and NEW.type in ('turning_circle','traffic_signals','mini_roundabout','noexit','crossing') THEN --- RAISE WARNING 'bad highway %',NEW.osm_id; - RETURN null; + -- remove operator tag for most places, messes too much with search_name indexes + IF NEW.class not in ('amenity', 'shop') THEN + NEW.name := delete(NEW.name, 'operator'); END IF; + + -- just block these IF NEW.class in ('landuse','natural') and NEW.name is null THEN -- RAISE WARNING 'empty landuse %',NEW.osm_id; RETURN null; @@ -891,7 +949,7 @@ BEGIN IF ST_IsEmpty(NEW.geometry) OR NOT ST_IsValid(NEW.geometry) OR ST_X(ST_Centroid(NEW.geometry))::text in ('NaN','Infinity','-Infinity') OR ST_Y(ST_Centroid(NEW.geometry))::text in ('NaN','Infinity','-Infinity') THEN -- block all invalid geometary - just not worth the risk. seg faults are causing serious problems. --- RAISE WARNING 'invalid geometry %',NEW.osm_id; + RAISE WARNING 'invalid geometry %',NEW.osm_id; RETURN NULL; -- Dead code @@ -901,26 +959,29 @@ BEGIN END IF; NEW.geometry := ST_buffer(NEW.geometry,0); IF ST_IsEmpty(NEW.geometry) OR NOT ST_IsValid(NEW.geometry) OR ST_X(ST_Centroid(NEW.geometry))::text in ('NaN','Infinity','-Infinity') OR ST_Y(ST_Centroid(NEW.geometry))::text in ('NaN','Infinity','-Infinity') THEN --- RAISE WARNING 'Invalid geometary, rejecting: % %', NEW.osm_type, NEW.osm_id; + RAISE WARNING 'Invalid geometary, rejecting: % %', NEW.osm_type, NEW.osm_id; RETURN NULL; END IF; END IF; + --DEBUG: RAISE WARNING '% % % %',NEW.osm_type,NEW.osm_id,NEW.class,NEW.type; + NEW.place_id := nextval('seq_place'); NEW.indexed_status := 1; --STATUS_NEW - NEW.country_code := lower(get_country_code(NEW.geometry, NEW.country_code)); - NEW.partition := get_partition(NEW.geometry, NEW.country_code); + NEW.calculated_country_code := lower(get_country_code(NEW.geometry)); + + NEW.partition := get_partition(NEW.calculated_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 - default_language := get_country_language_code(NEW.country_code); + IF NEW.name is not null AND array_upper(akeys(NEW.name),1) > 1 THEN + default_language := get_country_language_code(NEW.calculated_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')); + NEW.name := NEW.name || hstore(('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 || hstore('name', (NEW.name -> ('name:'||default_language))); END IF; END IF; END IF; @@ -942,9 +1003,14 @@ BEGIN -- By doing in postgres we have the country available to us - currently only used for postcode IF NEW.class in ('place','boundary') AND NEW.type in ('postcode','postal_code') THEN - NEW.name := 'ref'=>NEW.postcode; + 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 := hstore('ref', NEW.postcode); - IF NEW.country_code = 'gb' THEN + IF NEW.calculated_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; @@ -957,7 +1023,7 @@ BEGIN NEW.rank_address := 5; END IF; - ELSEIF NEW.country_code = 'de' THEN + ELSEIF NEW.calculated_country_code = 'de' THEN IF NEW.postcode ~ '^([0-9]{5})$' THEN NEW.rank_search := 21; @@ -987,9 +1053,11 @@ BEGIN IF NEW.type in ('continent') THEN NEW.rank_search := 2; NEW.rank_address := NEW.rank_search; + NEW.calculated_country_code := NULL; ELSEIF NEW.type in ('sea') THEN NEW.rank_search := 2; NEW.rank_address := 0; + NEW.calculated_country_code := NULL; ELSEIF NEW.type in ('country') THEN NEW.rank_search := 4; NEW.rank_address := NEW.rank_search; @@ -997,8 +1065,8 @@ 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; @@ -1014,30 +1082,19 @@ BEGIN ELSEIF NEW.type in ('village','hamlet','municipality','district','unincorporated_area','borough') THEN NEW.rank_search := 19; NEW.rank_address := 16; - ELSEIF NEW.type in ('airport') AND ST_GeometryType(NEW.geometry) in ('ST_Polygon','ST_MultiPolygon') THEN - NEW.rank_search := 18; - NEW.rank_address := 17; - ELSEIF NEW.type in ('moor') AND ST_GeometryType(NEW.geometry) in ('ST_Polygon','ST_MultiPolygon') THEN - NEW.rank_search := 17; - NEW.rank_address := 18; - ELSEIF NEW.type in ('moor') THEN - NEW.rank_search := 17; - NEW.rank_address := 0; - ELSEIF NEW.type in ('national_park') THEN - NEW.rank_search := 18; - NEW.rank_address := 18; - ELSEIF NEW.type in ('suburb','croft','subdivision') THEN + ELSEIF NEW.type in ('suburb','croft','subdivision','isolated_dwelling') 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','mountain_pass') THEN NEW.rank_search := 20; NEW.rank_address := 0; - ELSEIF NEW.type in ('hall_of_residence','neighbourhood','housing_estate','nature_reserve') THEN + -- 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 ('neighbourhood') THEN NEW.rank_search := 22; NEW.rank_address := 22; - ELSEIF NEW.type in ('airport','street') THEN - NEW.rank_search := 26; - NEW.rank_address := NEW.rank_search; ELSEIF NEW.type in ('house','building') THEN NEW.rank_search := 30; NEW.rank_address := NEW.rank_search; @@ -1055,10 +1112,22 @@ BEGIN return NULL; END IF; NEW.rank_search := NEW.admin_level * 2; - NEW.rank_address := NEW.rank_search; + IF NEW.type = 'administrative' THEN + NEW.rank_address := NEW.rank_search; + ELSE + NEW.rank_address := 0; + END IF; ELSEIF NEW.class = 'landuse' AND ST_GeometryType(NEW.geometry) in ('ST_Polygon','ST_MultiPolygon') THEN NEW.rank_search := 22; NEW.rank_address := NEW.rank_search; + ELSEIF NEW.class = 'natural' and NEW.type in ('peak','volcano','mountain_range') THEN + NEW.rank_search := 18; + NEW.rank_address := 0; + ELSEIF NEW.class = 'natural' and NEW.type = 'sea' THEN + NEW.rank_search := 4; + NEW.rank_address := NEW.rank_search; + ELSEIF NEW.class = 'natural' and NEW.type in ('coastline') THEN + RETURN NULL; -- any feature more than 5 square miles is probably worth indexing ELSEIF ST_GeometryType(NEW.geometry) in ('ST_Polygon','ST_MultiPolygon') AND ST_Area(NEW.geometry) > 0.1 THEN NEW.rank_search := 22; @@ -1072,21 +1141,21 @@ BEGIN ELSEIF NEW.class = 'waterway' AND NEW.name is NULL THEN RETURN NULL; ELSEIF NEW.class = 'waterway' THEN - NEW.rank_address := 17; + IF NEW.osm_type = 'R' THEN + NEW.rank_search := 16; + ELSE + NEW.rank_search := 17; + END IF; + NEW.rank_address := 0; ELSEIF NEW.class = 'highway' AND NEW.osm_type != 'N' AND NEW.type in ('service','cycleway','path','footway','steps','bridleway','motorway_link','primary_link','trunk_link','secondary_link','tertiary_link') THEN NEW.rank_search := 27; NEW.rank_address := NEW.rank_search; ELSEIF NEW.class = 'highway' AND NEW.osm_type != 'N' THEN NEW.rank_search := 26; NEW.rank_address := NEW.rank_search; - ELSEIF NEW.class = 'natural' and NEW.type = 'sea' THEN - NEW.rank_search := 4; - NEW.rank_address := NEW.rank_search; - ELSEIF NEW.class = 'natural' and NEW.type in ('coastline') THEN - RETURN NULL; - ELSEIF NEW.class = 'natural' and NEW.type in ('peak','volcano') THEN - NEW.rank_search := 18; - NEW.rank_address := 0; + ELSEIF NEW.class = 'mountain_pass' THEN + NEW.rank_search := 20; + NEW.rank_address := 0; END IF; END IF; @@ -1100,7 +1169,12 @@ BEGIN END IF; IF (NEW.extratags -> 'capital') = 'yes' THEN - NEW.rank_search := NEW.rank_search -1; + NEW.rank_search := NEW.rank_search - 1; + END IF; + + -- a country code make no sense below rank 4 (country) + IF NEW.rank_search < 4 THEN + NEW.calculated_country_code := NULL; END IF; -- Block import below rank 22 @@ -1108,53 +1182,65 @@ BEGIN -- RETURN NULL; -- END IF; - RETURN NEW; -- The following is not needed until doing diff updates, and slows the main index process down + --DEBUG: RAISE WARNING 'placex_insert:END: % % % %',NEW.osm_type,NEW.osm_id,NEW.class,NEW.type; - IF (ST_GeometryType(NEW.geometry) in ('ST_Polygon','ST_MultiPolygon') AND ST_IsValid(NEW.geometry)) THEN - -- Performance: We just can't handle re-indexing for country level changes - 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; + RETURN NEW; -- @DIFFUPDATES@ The following is not needed until doing diff updates, and slows the main index process down - -- 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'; - 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'; - END IF; - ELSE - -- mark nearby items for re-indexing, where 'nearby' depends on the features rank_search and is a complete guess :( - diameter := 0; - -- 16 = city, anything higher than city is effectively ignored (polygon required!) - IF NEW.type='postcode' THEN - diameter := 0.05; - ELSEIF NEW.rank_search < 16 THEN + IF NEW.rank_address > 0 THEN + IF (ST_GeometryType(NEW.geometry) in ('ST_Polygon','ST_MultiPolygon') AND ST_IsValid(NEW.geometry)) THEN + -- Performance: We just can't handle re-indexing for country level changes + 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, 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_covers(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 < 28 or name is not null or (NEW.rank_search >= 16 and addr_place is not null)); + update placex set indexed_status = 2 where (st_covers(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 < 28 or name is not null or (NEW.rank_search >= 16 and addr_place 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 :( diameter := 0; - ELSEIF NEW.rank_search < 18 THEN - diameter := 0.1; - ELSEIF NEW.rank_search < 20 THEN - diameter := 0.05; - ELSEIF NEW.rank_search = 21 THEN - diameter := 0.001; - ELSEIF NEW.rank_search < 24 THEN - diameter := 0.02; - ELSEIF NEW.rank_search < 26 THEN - diameter := 0.002; -- 100 to 200 meters - ELSEIF NEW.rank_search < 28 THEN - diameter := 0.001; -- 50 to 100 meters - 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); + -- 16 = city, anything higher than city is effectively ignored (polygon required!) + IF NEW.type='postcode' THEN + diameter := 0.05; + ELSEIF NEW.rank_search < 16 THEN + diameter := 0; + ELSEIF NEW.rank_search < 18 THEN + diameter := 0.1; + ELSEIF NEW.rank_search < 20 THEN + diameter := 0.05; + ELSEIF NEW.rank_search = 21 THEN + diameter := 0.001; + ELSEIF NEW.rank_search < 24 THEN + diameter := 0.02; + ELSEIF NEW.rank_search < 26 THEN + diameter := 0.002; -- 100 to 200 meters + ELSEIF NEW.rank_search < 28 THEN + diameter := 0.001; -- 50 to 100 meters + END IF; + IF diameter > 0 THEN + -- RAISE WARNING 'placex point insert: % % % % %',NEW.osm_type,NEW.osm_id,NEW.class,NEW.type,diameter; + IF NEW.rank_search >= 26 THEN + -- roads may cause reparenting for >27 rank places + update placex set indexed_status = 2 where indexed_status = 0 and rank_search > NEW.rank_search and ST_DWithin(placex.geometry, NEW.geometry, diameter); + ELSEIF NEW.rank_search >= 16 THEN + -- up to rank 16, street-less addresses may need reparenting + 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 or addr_place is not null); + ELSE + -- for all other places the search terms may change as well + 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; 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; + SELECT count(*)>0 FROM pg_tables WHERE tablename = classtable and schemaname = current_schema() INTO result; IF result THEN EXECUTE 'INSERT INTO ' || classtable::regclass || ' (place_id, centroid) VALUES ($1,$2)' USING NEW.place_id, ST_Centroid(NEW.geometry); @@ -1189,11 +1275,15 @@ DECLARE location RECORD; way RECORD; relation RECORD; + relation_members TEXT[]; + relMember RECORD; + linkedplacex RECORD; search_diameter FLOAT; search_prevdiameter FLOAT; search_maxrank INTEGER; address_maxrank INTEGER; address_street_word_id INTEGER; + address_street_word_ids INTEGER[]; parent_place_id_rank BIGINT; isin TEXT[]; @@ -1201,6 +1291,8 @@ DECLARE location_rank_search INTEGER; location_distance FLOAT; + location_parent GEOMETRY; + location_isaddress BOOLEAN; tagpairid INTEGER; @@ -1208,9 +1300,24 @@ DECLARE name_vector INTEGER[]; nameaddress_vector INTEGER[]; + linked_node_id BIGINT; + result BOOLEAN; BEGIN + -- deferred delete + IF OLD.indexed_status = 100 THEN + --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; + + IF NEW.indexed_status != 0 OR OLD.indexed_status = 0 THEN + RETURN NEW; + END IF; + + --DEBUG: RAISE WARNING 'placex_update % %',NEW.osm_type,NEW.osm_id; + --RAISE WARNING '%',NEW.place_id; --RAISE WARNING '%', NEW; @@ -1219,34 +1326,70 @@ BEGIN RETURN NEW; END IF; - -- deferred delete - IF OLD.indexed_status = 100 THEN - delete from placex where place_id = OLD.place_id; - RETURN NULL; - END IF; - - IF NEW.indexed_status = 0 and OLD.indexed_status != 0 THEN + -- TODO: this test is now redundant? + IF OLD.indexed_status != 0 THEN NEW.indexed_date = now(); + 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, NEW.rank_search); + UPDATE placex set linked_place_id = null where linked_place_id = NEW.place_id; + + IF NEW.linked_place_id is not null THEN + RETURN NEW; + END IF; + IF NEW.class = 'place' AND NEW.type = 'houses' THEN i := create_interpolation(NEW.osm_id, NEW.housenumber); RETURN NEW; END IF; - 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); + -- Speed up searches - just use the centroid of the feature + -- cheaper but less acurate + place_centroid := ST_PointOnSurface(NEW.geometry); + NEW.centroid := null; + + -- reclaculate country and partition + IF NEW.rank_search = 4 THEN + -- for countries, believe the mapped country code, + -- so that we remain in the right partition if the boundaries + -- suddenly expand. + NEW.partition := get_partition(lower(NEW.country_code)); + IF NEW.partition = 0 THEN + NEW.calculated_country_code := lower(get_country_code(place_centroid)); + NEW.partition := get_partition(NEW.calculated_country_code); + ELSE + NEW.calculated_country_code := lower(NEW.country_code); + END IF; + ELSE + IF NEW.rank_search > 4 THEN + --NEW.calculated_country_code := lower(get_country_code(NEW.geometry, NEW.country_code)); + NEW.calculated_country_code := lower(get_country_code(place_centroid)); + ELSE + NEW.calculated_country_code := NULL; + END IF; + NEW.partition := get_partition(NEW.calculated_country_code); END IF; + NEW.geometry_sector := geometry_sector(NEW.partition, place_centroid); - -- 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)); - NEW.partition := get_partition(NEW.geometry, NEW.country_code); - NEW.geometry_sector := geometry_sector(NEW.partition, NEW.geometry); + -- 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 + FOR relation IN select * from planet_osm_rels r where r.id = NEW.osm_id + LOOP + FOR i IN 1..array_upper(relation.members, 1) BY 2 LOOP + IF relation.members[i+1] in ('', 'main_stream') AND substring(relation.members[i],1,1) = 'w' THEN + --DEBUG: RAISE WARNING 'waterway parent %, child %/%', NEW.osm_id, i, relation.parts[i]; + FOR location IN SELECT * FROM placex WHERE osm_type = 'W' and osm_id = substring(relation.members[i],2,200)::bigint and class = NEW.class and type = NEW.type + LOOP + UPDATE placex SET linked_place_id = NEW.place_id WHERE place_id = location.place_id; + END LOOP; + END IF; + END LOOP; + END LOOP; + END IF; -- Adding ourselves to the list simplifies address calculations later INSERT INTO place_addressline VALUES (NEW.place_id, NEW.place_id, true, true, 0, NEW.rank_address); @@ -1254,19 +1397,15 @@ BEGIN -- What level are we searching from search_maxrank := NEW.rank_search; - -- Speed up searches - just use the centroid of the feature - -- cheaper but less acurate - place_centroid := ST_Centroid(NEW.geometry); - -- 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(%#NEW.name,1) > 1 THEN - default_language := get_country_language_code(NEW.country_code); + IF NEW.name is not null AND array_upper(akeys(NEW.name),1) > 1 THEN + default_language := get_country_language_code(NEW.calculated_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')); + NEW.name := NEW.name || hstore(('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 || hstore('name', (NEW.name -> ('name:'||default_language))); END IF; END IF; END IF; @@ -1285,13 +1424,22 @@ BEGIN address_havelevel[i] := false; END LOOP; ---RAISE WARNING '% %', NEW.place_id, NEW.rank_search; + NEW.importance := null; + select language||':'||title,importance from get_wikipedia_match(NEW.extratags, NEW.calculated_country_code) 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 'before low level% %', NEW.place_id, NEW.rank_search; -- 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; + -- We won't get a better centroid, besides these places are too small to care + NEW.centroid := place_centroid; + NEW.parent_place_id := null; -- to do that we have to find our parent road @@ -1299,16 +1447,26 @@ BEGIN -- Note that addr:street links can only be indexed once the street itself is indexed IF NEW.parent_place_id IS NULL AND NEW.osm_type = 'N' THEN + -- if there is no address information, see if we can get it from a surrounding building + IF NEW.street IS NULL AND NEW.addr_place IS NULL AND NEW.housenumber IS NULL THEN + FOR location IN select * from placex where ST_Covers(geometry, place_centroid) and rank_search > 28 and (housenumber is not null or street is not null or addr_place is not null) AND ST_GeometryType(geometry) in ('ST_Polygon','ST_MultiPolygon') + LOOP + NEW.housenumber := location.housenumber; + NEW.street := location.street; + NEW.addr_place := location.addr_place; + END LOOP; + END IF; + -- 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 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 relation %',relation; - SELECT place_id from placex where osm_type='W' and osm_id = substring(relation.members[i],2,200)::integer - and rank_search = 26 INTO NEW.parent_place_id; + 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; @@ -1316,7 +1474,7 @@ 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 @@ -1329,15 +1487,15 @@ 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 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)::integer - and rank_search = 26 INTO NEW.parent_place_id; + 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; @@ -1345,20 +1503,22 @@ BEGIN END IF; -- If the way contains an explicit name of a street copy it - IF NEW.street IS NULL AND location.street IS NOT NULL THEN + -- Slightly less strict then above because data is copied from any object. + IF NEW.street IS NULL AND NEW.addr_place IS NULL THEN --RAISE WARNING 'node in way that has a streetname %',location; NEW.street := location.street; + NEW.addr_place := location.addr_place; END IF; -- If this way is a street interpolation line then it is probably as good as we are going to get - IF NEW.parent_place_id IS NULL AND NEW.street IS NULL AND location.class = 'place' and location.type='houses' THEN + IF NEW.parent_place_id IS NULL AND NEW.street IS NULL AND NEW.addr_place IS NULL AND location.class = 'place' and location.type='houses' THEN -- Try and find a way that is close roughly parellel to this line FOR relation IN SELECT place_id FROM placex WHERE ST_DWithin(location.geometry, placex.geometry, 0.001) and placex.rank_search = 26 and st_geometrytype(location.geometry) in ('ST_LineString') - ORDER BY (ST_distance(placex.geometry, ST_Line_Interpolate_Point(location.geometry,0))+ - ST_distance(placex.geometry, ST_Line_Interpolate_Point(location.geometry,0.5))+ - ST_distance(placex.geometry, ST_Line_Interpolate_Point(location.geometry,1))) ASC limit 1 + ORDER BY (ST_distance(placex.geometry, ST_LineInterpolatePoint(location.geometry,0))+ + ST_distance(placex.geometry, ST_LineInterpolatePoint(location.geometry,0.5))+ + ST_distance(placex.geometry, ST_LineInterpolatePoint(location.geometry,1))) ASC limit 1 LOOP --RAISE WARNING 'using nearest street to address interpolation line,0.001 %',relation; NEW.parent_place_id := relation.place_id; @@ -1374,15 +1534,15 @@ 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 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 '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)::integer - and rank_search = 26 INTO NEW.parent_place_id; + 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; @@ -1392,9 +1552,18 @@ BEGIN --RAISE WARNING 'x3 %',NEW.parent_place_id; IF NEW.parent_place_id IS NULL AND NEW.street IS NOT NULL THEN - address_street_word_id := get_name_id(make_standard_name(NEW.street)); - IF address_street_word_id IS NOT NULL THEN - FOR location IN SELECT * from getNearestNamedRoadFeature(NEW.partition, place_centroid, address_street_word_id) LOOP + address_street_word_ids := get_name_ids(make_standard_name(NEW.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; + END IF; + END IF; + + IF NEW.parent_place_id IS NULL AND NEW.addr_place IS NOT NULL THEN + address_street_word_ids := get_name_ids(make_standard_name(NEW.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; END IF; @@ -1425,7 +1594,7 @@ 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; + NEW.calculated_country_code := location.country_code; --RAISE WARNING '%', NEW.name; -- If there is no name it isn't searchable, don't bother to create a search record @@ -1435,23 +1604,169 @@ BEGIN -- Merge address from parent nameaddress_vector := array_merge(nameaddress_vector, location.nameaddress_vector); + nameaddress_vector := array_merge(nameaddress_vector, location.name_vector); --return NEW; -- Performance, it would be more acurate to do all the rest of the import process but it takes too long -- Just be happy with inheriting from parent road only IF NEW.rank_search <= 25 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.calculated_country_code, NEW.partition, name_vector, NEW.rank_search, NEW.rank_address, NEW.geometry); 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); + result := insertSearchName(NEW.partition, NEW.place_id, NEW.calculated_country_code, name_vector, nameaddress_vector, NEW.rank_search, NEW.rank_address, NEW.importance, place_centroid, NEW.geometry); return NEW; END IF; END IF; +-- RAISE WARNING ' INDEXING Started:'; -- 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; + +-- 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 + + 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 + + -- 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 + IF NOT linkedPlacex.name IS NULL THEN + NEW.name := linkedPlacex.name || NEW.name; + name_vector := array_merge(name_vector, make_keywords(linkedPlacex.name)); + END IF; + + -- merge in extra tags + NEW.extratags := hstore(linkedPlacex.class, linkedPlacex.type) || coalesce(linkedPlacex.extratags, ''::hstore) || coalesce(NEW.extratags, ''::hstore); + + -- 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 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 + + 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 + + -- 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_address = linkedPlacex.rank_address 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 + IF NOT linkedPlacex.name IS NULL THEN + NEW.name := linkedPlacex.name || NEW.name; + name_vector := make_keywords(NEW.name); + END IF; + + -- merge in extra tags + NEW.extratags := hstore(linkedPlacex.class, linkedPlacex.type) || coalesce(linkedPlacex.extratags, ''::hstore) || coalesce(NEW.extratags, ''::hstore); + + -- 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 LOOP; + + END IF; + END IF; + + END IF; + + -- Name searches can be done for ways as well as relations + IF NEW.osm_type in ('W','R') AND NEW.rank_search < 26 THEN + + -- 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_address = NEW.rank_address + AND placex.place_id != NEW.place_id + AND placex.osm_type = 'N'::char(1) AND placex.rank_search < 26 + AND st_covers(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 := hstore(linkedPlacex.class, linkedPlacex.type) || coalesce(linkedPlacex.extratags, ''::hstore) || coalesce(NEW.extratags, ''::hstore); + + -- 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; + -- Place might have had only a name tag before but has now received translations + -- from the linked place. Make sure a name tag for the default language exists in + -- this case. + IF NEW.name is not null AND array_upper(akeys(NEW.name),1) > 1 THEN + default_language := get_country_language_code(NEW.calculated_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 || hstore(('name:'||default_language), (NEW.name -> 'name')); + ELSEIF NEW.name ? ('name:'||default_language) AND NOT NEW.name ? 'name' THEN + NEW.name := NEW.name || hstore('name', (NEW.name -> ('name:'||default_language))); + END IF; + END IF; + END IF; + 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.calculated_country_code) 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'::char(1) and osm_id = linked_node_id order by importance desc limit 1 INTO NEW.wikipedia,NEW.importance; + END IF; + + 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 + perform create_country(NEW.name, lower(NEW.country_code)); + END IF; + NEW.parent_place_id = 0; parent_place_id_rank = 0; @@ -1463,8 +1778,15 @@ BEGIN 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 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 IF; @@ -1474,40 +1796,80 @@ BEGIN 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; ---RAISE WARNING 'ISIN: %', isin_tokens; + + -- for the USA we have an additional address table. Merge in zip codes from there too + IF NEW.rank_search = 26 AND NEW.calculated_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 LOOP; + END IF; + +-- RAISE WARNING 'ISIN: %', isin_tokens; -- Process area matches - location_rank_search := 100; + location_rank_search := 0; location_distance := 0; ---RAISE WARNING ' getNearFeatures(%,''%'',%,''%'')',NEW.partition, place_centroid, search_maxrank, isin_tokens; + 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; FOR location IN SELECT * from getNearFeatures(NEW.partition, place_centroid, search_maxrank, isin_tokens) LOOP --RAISE WARNING ' AREA: %',location; - IF location.rank_search < location_rank_search THEN - location_rank_search := location.rank_search; + IF location.rank_address != location_rank_search THEN + location_rank_search := location.rank_address; location_distance := location.distance * 1.5; END IF; IF location.distance < location_distance OR NOT location.isguess THEN + location_isaddress := NOT address_havelevel[location.rank_address]; + IF location_isaddress AND location.isguess AND location_parent IS NOT NULL THEN + location_isaddress := ST_Contains(location_parent,location.centroid); + END IF; + + -- RAISE WARNING '% isaddress: %', location.place_id, location_isaddress; -- Add it to the list of search terms - nameaddress_vector := array_merge(nameaddress_vector, location.keywords::integer[]); - INSERT INTO place_addressline VALUES (NEW.place_id, location.place_id, true, NOT address_havelevel[location.rank_address], location.distance, location.rank_address); - address_havelevel[location.rank_address] := true; + IF location.rank_search > 4 THEN + nameaddress_vector := array_merge(nameaddress_vector, location.keywords::integer[]); + END IF; + INSERT INTO place_addressline VALUES (NEW.place_id, location.place_id, true, location_isaddress, location.distance, location.rank_address); ---RAISE WARNING ' Terms: (%) %',location, nameaddress_vector; + IF location_isaddress THEN + + address_havelevel[location.rank_address] := true; + IF NOT location.isguess THEN + SELECT geometry FROM placex WHERE place_id = location.place_id INTO location_parent; + END IF; + + 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; - 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; +--RAISE WARNING ' Terms: (%) %',location, nameaddress_vector; + END IF; END LOOP; @@ -1516,21 +1878,25 @@ BEGIN 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 + FOR location IN SELECT * from getNearestNamedFeature(NEW.partition, place_centroid, search_maxrank, isin_tokens[i]) LOOP ---RAISE WARNING ' ISIN: %',location; + --RAISE WARNING ' ISIN: %',location; - nameaddress_vector := array_merge(nameaddress_vector, location.keywords::integer[]); - INSERT INTO place_addressline 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_search > 4 THEN + nameaddress_vector := array_merge(nameaddress_vector, location.keywords::integer[]); + INSERT INTO place_addressline 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; + 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 LOOP; + END IF; END LOOP; END IF; @@ -1538,17 +1904,17 @@ BEGIN -- for long ways we should add search terms for the entire length IF st_length(NEW.geometry) > 0.05 THEN - location_rank_search := 100; + 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_search < location_rank_search THEN - location_rank_search := location.rank_search; + IF location.rank_address != location_rank_search THEN + location_rank_search := location.rank_address; location_distance := location.distance * 1.5; END IF; - IF location.distance < location_distance THEN + 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[]); @@ -1564,16 +1930,20 @@ BEGIN IF NEW.name IS NOT NULL THEN IF NEW.rank_search <= 25 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.calculated_country_code, NEW.partition, name_vector, NEW.rank_search, NEW.rank_address, NEW.geometry); 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); + result := insertLocationRoad(NEW.partition, NEW.place_id, NEW.calculated_country_code, NEW.geometry); 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); + result := insertSearchName(NEW.partition, NEW.place_id, NEW.calculated_country_code, name_vector, nameaddress_vector, NEW.rank_search, NEW.rank_address, NEW.importance, place_centroid, NEW.geometry); --- 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; @@ -1589,38 +1959,59 @@ DECLARE b BOOLEAN; classtable TEXT; BEGIN + -- RAISE WARNING 'placex_delete % %',OLD.osm_type,OLD.osm_id; + + update placex set linked_place_id = null, indexed_status = 2 where linked_place_id = OLD.place_id and indexed_status = 0; + --DEBUG: RAISE WARNING 'placex_delete:01 % %',OLD.osm_type,OLD.osm_id; + update placex set linked_place_id = null where linked_place_id = OLD.place_id; + --DEBUG: RAISE WARNING 'placex_delete:02 % %',OLD.osm_type,OLD.osm_id; IF OLD.rank_address < 30 THEN -- mark everything linked to this place for re-indexing + --DEBUG: RAISE WARNING 'placex_delete:03 % %',OLD.osm_type,OLD.osm_id; UPDATE placex set indexed_status = 2 from place_addressline where address_place_id = OLD.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 place_addressline.isaddress; + --DEBUG: RAISE WARNING 'placex_delete:04 % %',OLD.osm_type,OLD.osm_id; DELETE FROM place_addressline where address_place_id = OLD.place_id; + --DEBUG: RAISE WARNING 'placex_delete:05 % %',OLD.osm_type,OLD.osm_id; b := deleteRoad(OLD.partition, OLD.place_id); + --DEBUG: RAISE WARNING 'placex_delete:06 % %',OLD.osm_type,OLD.osm_id; update placex set indexed_status = 2 where parent_place_id = OLD.place_id and indexed_status = 0; + --DEBUG: RAISE WARNING 'placex_delete:07 % %',OLD.osm_type,OLD.osm_id; END IF; + --DEBUG: RAISE WARNING 'placex_delete:08 % %',OLD.osm_type,OLD.osm_id; + IF OLD.rank_address < 26 THEN - b := deleteLocationArea(OLD.partition, OLD.place_id); + b := deleteLocationArea(OLD.partition, OLD.place_id, OLD.rank_search); END IF; + --DEBUG: RAISE WARNING 'placex_delete:09 % %',OLD.osm_type,OLD.osm_id; + IF OLD.name is not null THEN b := deleteSearchName(OLD.partition, OLD.place_id); END IF; + --DEBUG: RAISE WARNING 'placex_delete:10 % %',OLD.osm_type,OLD.osm_id; + DELETE FROM place_addressline where place_id = OLD.place_id; + --DEBUG: RAISE WARNING 'placex_delete:11 % %',OLD.osm_type,OLD.osm_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; + SELECT count(*)>0 FROM pg_tables WHERE tablename = classtable and schemaname = current_schema() INTO b; IF b THEN EXECUTE 'DELETE FROM ' || classtable::regclass || ' WHERE place_id = $1' USING OLD.place_id; END IF; + --DEBUG: RAISE WARNING 'placex_delete:12 % %',OLD.osm_type,OLD.osm_id; + RETURN OLD; END; @@ -1630,21 +2021,23 @@ LANGUAGE plpgsql; CREATE OR REPLACE FUNCTION place_delete() RETURNS TRIGGER AS $$ DECLARE - placeid BIGINT; + has_rank BOOLEAN; BEGIN --- RAISE WARNING 'delete: % % % %',OLD.osm_type,OLD.osm_id,OLD.class,OLD.type; + --DEBUG: RAISE WARNING 'delete: % % % %',OLD.osm_type,OLD.osm_id,OLD.class,OLD.type; - -- deleting large polygons can have a massive effect ont he system - require manual intervention to let them through - IF st_area(OLD.geometry) > 2 THEN - insert into import_polygon_delete values (OLD.osm_type,OLD.osm_id,OLD.class,OLD.type); - RETURN NULL; + -- deleting large polygons can have a massive effect on the system - require manual intervention to let them through + IF st_area(OLD.geometry) > 2 and st_isvalid(OLD.geometry) THEN + SELECT bool_or(not (rank_address = 0 or rank_address > 26)) as ranked FROM placex WHERE osm_type = OLD.osm_type and osm_id = OLD.osm_id and class = OLD.class and type = OLD.type INTO has_rank; + IF has_rank THEN + insert into import_polygon_delete values (OLD.osm_type,OLD.osm_id,OLD.class,OLD.type); + RETURN NULL; + END IF; END IF; -- mark for delete UPDATE placex set indexed_status = 100 where osm_type = OLD.osm_type and osm_id = OLD.osm_id and class = OLD.class and type = OLD.type; --- delete from placex where osm_type = OLD.osm_type and osm_id = OLD.osm_id and class = OLD.class and type = OLD.type; RETURN OLD; END; @@ -1663,18 +2056,23 @@ DECLARE partition INTEGER; BEGIN + --DEBUG: RAISE WARNING '-----------------------------------------------------------------------------------'; + --DEBUG: RAISE WARNING 'place_insert: % % % % %',NEW.osm_type,NEW.osm_id,NEW.class,NEW.type,st_area(NEW.geometry); + IF FALSE and NEW.osm_type = 'R' THEN - RAISE WARNING '-----------------------------------------------------------------------------------'; - RAISE WARNING 'place_insert: % % % % %',NEW.osm_type,NEW.osm_id,NEW.class,NEW.type,st_area(NEW.geometry); select * from placex where osm_type = NEW.osm_type and osm_id = NEW.osm_id and class = NEW.class and type = NEW.type INTO existingplacex; - RAISE WARNING '%', existingplacex; + --DEBUG: RAISE WARNING '%', existingplacex; END IF; - -- Just block these - lots and pointless - IF NEW.class = 'highway' and NEW.type in ('turning_circle','traffic_signals','mini_roundabout','noexit','crossing') THEN - RETURN null; + -- remove operator tag for most places, messes too much with search_name indexes + IF NEW.class not in ('amenity', 'shop') THEN + NEW.name := delete(NEW.name, 'operator'); END IF; + + -- Just block these - lots and pointless IF NEW.class in ('landuse','natural') and NEW.name is null THEN + -- if the name tag was removed, older versions might still be lurking in the place table + DELETE FROM place where osm_type = NEW.osm_type and osm_id = NEW.osm_id and class = NEW.class and type = NEW.type; RETURN null; END IF; @@ -1687,7 +2085,7 @@ BEGIN -- Patch in additional country names IF NEW.admin_level = 2 AND NEW.type = 'administrative' AND NEW.country_code is not null THEN - select country_name.name || NEW.name from country_name where country_name.country_code = lower(NEW.country_code) INTO NEW.name; + select coalesce(country_name.name || NEW.name,NEW.name) from country_name where country_name.country_code = lower(NEW.country_code) INTO NEW.name; END IF; -- Have we already done this place? @@ -1703,7 +2101,8 @@ BEGIN DELETE FROM place where osm_type = NEW.osm_type and osm_id = NEW.osm_id and class = NEW.class and type not in ('postcode','house','houses'); END IF; --- RAISE WARNING 'Existing: %',existing.place_id; + --DEBUG: RAISE WARNING 'Existing: %',existing.osm_id; + --DEBUG: RAISE WARNING 'Existing PlaceX: %',existingplacex.place_id; -- Log and discard IF existing.geometry is not null AND st_isvalid(existing.geometry) @@ -1719,34 +2118,30 @@ BEGIN DELETE from import_polygon_error where osm_type = NEW.osm_type and osm_id = NEW.osm_id; DELETE from import_polygon_delete where osm_type = NEW.osm_type and osm_id = NEW.osm_id; - -- To paraphrase, if there isn't an existing item, OR if the admin level has changed, OR if it is a major change in geometry - IF existing.osm_type IS NULL - OR existingplacex.osm_type IS NULL - OR coalesce(existing.admin_level, 100) != coalesce(NEW.admin_level, 100) - OR coalesce(existing.country_code, '') != coalesce(NEW.country_code, '') - OR (existing.geometry::text != NEW.geometry::text AND ST_Distance(ST_Centroid(existing.geometry),ST_Centroid(NEW.geometry)) > 0.01 AND NOT - (ST_GeometryType(existing.geometry) in ('ST_Polygon','ST_MultiPolygon') AND ST_GeometryType(NEW.geometry) in ('ST_Polygon','ST_MultiPolygon'))) - THEN - --- IF existing.osm_type IS NULL THEN --- RAISE WARNING 'no existing place'; --- END IF; --- IF existingplacex.osm_type IS NULL THEN --- RAISE WARNING 'no existing placex %', existingplacex; --- END IF; - --- RAISE WARNING 'delete and replace'; + -- 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') + THEN + + IF existingplacex.osm_type IS NOT NULL THEN + -- sanity check: ignore admin_level changes on places with too many active children + -- or we end up reindexing entire countries because somebody accidentally deleted admin_level + --LIMIT INDEXING: SELECT count(*) FROM (SELECT 'a' FROM placex , place_addressline where address_place_id = existingplacex.place_id and placex.place_id = place_addressline.place_id and indexed_status = 0 and place_addressline.isaddress LIMIT 100001) sub INTO i; + --LIMIT INDEXING: IF i > 100000 THEN + --LIMIT INDEXING: RETURN null; + --LIMIT INDEXING: END IF; + END IF; IF existing.osm_type IS NOT NULL THEN --- RAISE WARNING 'insert delete % % % % %',NEW.osm_type,NEW.osm_id,NEW.class,NEW.type,ST_Distance(ST_Centroid(existing.geometry),ST_Centroid(NEW.geometry)),existing; - DELETE FROM place where osm_type = NEW.osm_type and osm_id = NEW.osm_id and class = NEW.class and type = NEW.type; - END IF; - --- RAISE WARNING 'delete and replace2'; + -- pathological case caused by the triggerless copy into place during initial import + -- force delete even for large areas, it will be reinserted later + UPDATE place set geometry = ST_SetSRID(ST_Point(0,0), 4326) where osm_type = NEW.osm_type and osm_id = NEW.osm_id and class = NEW.class and type = NEW.type; + DELETE from place where osm_type = NEW.osm_type and osm_id = NEW.osm_id and class = NEW.class and type = NEW.type; + END IF; -- No - process it as a new insertion (hopefully of low rank or it will be slow) insert into placex (osm_type, osm_id, class, type, name, admin_level, housenumber, - street, isin, postcode, country_code, extratags, geometry) + street, addr_place, isin, postcode, country_code, extratags, geometry) values (NEW.osm_type ,NEW.osm_id ,NEW.class @@ -1755,6 +2150,7 @@ BEGIN ,NEW.admin_level ,NEW.housenumber ,NEW.street + ,NEW.addr_place ,NEW.isin ,NEW.postcode ,NEW.country_code @@ -1762,7 +2158,7 @@ BEGIN ,NEW.geometry ); --- RAISE WARNING 'insert done % % % %',NEW.osm_type,NEW.osm_id,NEW.class,NEW.type; + --DEBUG: RAISE WARNING 'insert done % % % % %',NEW.osm_type,NEW.osm_id,NEW.class,NEW.type,NEW.name; RETURN NEW; END IF; @@ -1780,6 +2176,9 @@ BEGIN IF coalesce(existing.street, '') != coalesce(NEW.street, '') THEN RAISE WARNING 'update details, street: % % % %',NEW.osm_type,NEW.osm_id,existing.street,NEW.street; END IF; + IF coalesce(existing.addr_place, '') != coalesce(NEW.addr_place, '') THEN + RAISE WARNING 'update details, street: % % % %',NEW.osm_type,NEW.osm_id,existing.addr_place,NEW.addr_place; + END IF; IF coalesce(existing.isin, '') != coalesce(NEW.isin, '') THEN RAISE WARNING 'update details, isin: % % % %',NEW.osm_type,NEW.osm_id,existing.isin,NEW.isin; END IF; @@ -1805,14 +2204,14 @@ BEGIN -- re-index points that have moved in / out of the polygon, could be done as a single query but postgres gets the index usage wrong 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; + (st_covers(NEW.geometry, placex.geometry) OR ST_Intersects(NEW.geometry, placex.geometry)) + AND NOT (st_covers(existinggeometry, placex.geometry) OR ST_Intersects(existinggeometry, placex.geometry)) + 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; + (st_covers(existinggeometry, placex.geometry) OR ST_Intersects(existinggeometry, placex.geometry)) + AND NOT (st_covers(NEW.geometry, placex.geometry) OR ST_Intersects(NEW.geometry, placex.geometry)) + AND rank_search > existingplacex.rank_search AND (rank_search < 28 or name is not null); END IF; @@ -1822,6 +2221,7 @@ BEGIN IF FALSE AND existingplacex.rank_search < 26 AND coalesce(existing.housenumber, '') = coalesce(NEW.housenumber, '') AND coalesce(existing.street, '') = coalesce(NEW.street, '') + AND coalesce(existing.addr_place, '') = coalesce(NEW.addr_place, '') AND coalesce(existing.isin, '') = coalesce(NEW.isin, '') AND coalesce(existing.postcode, '') = coalesce(NEW.postcode, '') AND coalesce(existing.country_code, '') = coalesce(NEW.country_code, '') @@ -1832,7 +2232,8 @@ 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; @@ -1843,6 +2244,7 @@ BEGIN IF coalesce(existing.name::text, '') != coalesce(NEW.name::text, '') OR coalesce(existing.housenumber, '') != coalesce(NEW.housenumber, '') OR coalesce(existing.street, '') != coalesce(NEW.street, '') + OR coalesce(existing.addr_place, '') != coalesce(NEW.addr_place, '') OR coalesce(existing.isin, '') != coalesce(NEW.isin, '') OR coalesce(existing.postcode, '') != coalesce(NEW.postcode, '') OR coalesce(existing.country_code, '') != coalesce(NEW.country_code, '') THEN @@ -1857,13 +2259,21 @@ BEGIN END IF; + -- refuse to update multiplpoygons with too many objects, too much of a performance hit + IF ST_NumGeometries(NEW.geometry) > 2000 THEN + RAISE WARNING 'Dropping update of % % because of geometry complexity.', NEW.osm_type, NEW.osm_id; + RETURN NULL; + END IF; + IF coalesce(existing.name::text, '') != coalesce(NEW.name::text, '') OR coalesce(existing.extratags::text, '') != coalesce(NEW.extratags::text, '') OR coalesce(existing.housenumber, '') != coalesce(NEW.housenumber, '') OR coalesce(existing.street, '') != coalesce(NEW.street, '') + OR coalesce(existing.addr_place, '') != coalesce(NEW.addr_place, '') OR coalesce(existing.isin, '') != coalesce(NEW.isin, '') OR coalesce(existing.postcode, '') != coalesce(NEW.postcode, '') OR coalesce(existing.country_code, '') != coalesce(NEW.country_code, '') + OR coalesce(existing.admin_level, 15) != coalesce(NEW.admin_level, 15) OR existing.geometry::text != NEW.geometry::text THEN @@ -1871,30 +2281,40 @@ BEGIN name = NEW.name, housenumber = NEW.housenumber, street = NEW.street, + addr_place = NEW.addr_place, isin = NEW.isin, postcode = NEW.postcode, country_code = NEW.country_code, extratags = NEW.extratags, + admin_level = NEW.admin_level, geometry = NEW.geometry where osm_type = NEW.osm_type and osm_id = NEW.osm_id and class = NEW.class and type = NEW.type; + IF NEW.class in ('place','boundary') AND NEW.type in ('postcode','postal_code') THEN + IF NEW.postcode IS NULL THEN + -- postcode was deleted, no longer retain in placex + DELETE FROM placex where place_id = existingplacex.place_id; + RETURN NULL; + END IF; + + NEW.name := hstore('ref', NEW.postcode); + END IF; + update placex set name = NEW.name, housenumber = NEW.housenumber, street = NEW.street, + addr_place = NEW.addr_place, isin = NEW.isin, postcode = NEW.postcode, country_code = NEW.country_code, parent_place_id = null, extratags = NEW.extratags, + admin_level = CASE WHEN NEW.admin_level > 15 THEN 15 ELSE NEW.admin_level END, indexed_status = 2, geometry = NEW.geometry where place_id = existingplacex.place_id; --- now done as part of insert --- partition := get_partition(NEW.geometry, existingplacex.country_code); --- result := update_location(partition, existingplacex.place_id, existingplacex.country_code, NEW.name, existingplacex.rank_search, existingplacex.rank_address, NEW.geometry); - END IF; -- Abort the add (we modified the existing place instead) @@ -1922,7 +2342,9 @@ BEGIN END IF; END LOOP; - RETURN null; + -- anything will do as a fallback - just take the first name type thing there is + search := avals(name); + RETURN search[1]; END; $$ LANGUAGE plpgsql IMMUTABLE; @@ -2004,7 +2426,7 @@ BEGIN FOR location IN select * from get_addressdata(for_place_id) where isaddress order by rank_address desc LOOP currresult := trim(get_name_by_language(location.name, languagepref)); - IF currresult != prevresult AND currresult IS NOT NULL THEN + IF currresult != prevresult AND currresult IS NOT NULL AND result[(100 - location.rank_address)] IS NULL THEN result[(100 - location.rank_address)] := trim(get_name_by_language(location.name, languagepref)); prevresult := currresult; END IF; @@ -2015,11 +2437,11 @@ END; $$ LANGUAGE plpgsql; -DROP TYPE addressline CASCADE; +DROP TYPE IF EXISTS addressline CASCADE; create type addressline as ( place_id BIGINT, osm_type CHAR(1), - osm_id INTEGER, + osm_id BIGINT, name HSTORE, class TEXT, type TEXT, @@ -2061,14 +2483,14 @@ BEGIN END IF; IF for_place_id IS NULL THEN - select parent_place_id, country_code, housenumber, rank_address, postcode, name, class, type from placex + select parent_place_id, calculated_country_code, housenumber, rank_search, postcode, name, class, type from placex WHERE place_id = in_place_id and rank_address = 30 INTO for_place_id, searchcountrycode, searchhousenumber, searchrankaddress, searchpostcode, searchhousename, searchclass, searchtype; END IF; IF for_place_id IS NULL THEN for_place_id := in_place_id; - select country_code, housenumber, rank_address, postcode, null from placex where place_id = for_place_id + select calculated_country_code, housenumber, rank_search, postcode, null from placex where place_id = for_place_id INTO searchcountrycode, searchhousenumber, searchrankaddress, searchpostcode, searchhousename; END IF; @@ -2078,20 +2500,23 @@ BEGIN hadcountry := false; FOR location IN select placex.place_id, osm_type, osm_id, - CASE WHEN class = 'place' and type = 'postcode' THEN 'name' => postcode ELSE name END as name, + 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, 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 + 0 as distance, calculated_country_code, postcode from placex where place_id = for_place_id LOOP --RAISE WARNING '%',location; - IF searchcountrycode IS NULL AND location.country_code IS NOT NULL THEN - searchcountrycode := location.country_code; + IF searchcountrycode IS NULL AND location.calculated_country_code IS NOT NULL THEN + searchcountrycode := location.calculated_country_code; END IF; IF searchpostcode IS NOT NULL and location.type = 'postcode' 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 hadcountry := true; END IF; @@ -2111,24 +2536,29 @@ BEGIN FOR location IN select placex.place_id, osm_type, osm_id, - CASE WHEN class = 'place' and type = 'postcode' THEN 'name' => postcode ELSE name END as name, - class, type, admin_level, fromarea, isaddress, + CASE WHEN class = 'place' and type = 'postcode' THEN hstore('name', postcode) ELSE name END as 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, 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 + distance,calculated_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 (placex.country_code IS NULL OR searchcountrycode IS NULL OR placex.country_code = searchcountrycode OR rank_address < 4) + and (placex.calculated_country_code IS NULL OR searchcountrycode IS NULL OR placex.calculated_country_code = searchcountrycode) order by rank_address desc,isaddress desc,fromarea desc,distance asc,rank_search desc LOOP --RAISE WARNING '%',location; - IF searchcountrycode IS NULL AND location.country_code IS NOT NULL THEN - searchcountrycode := location.country_code; + IF searchcountrycode IS NULL AND location.calculated_country_code IS NOT NULL THEN + searchcountrycode := location.calculated_country_code; END IF; IF searchpostcode IS NOT NULL and location.type = 'postcode' 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; @@ -2156,7 +2586,7 @@ BEGIN END IF; IF searchcountrycode IS NOT NULL THEN - location := ROW(null, null, null, 'ref'=>searchcountrycode, 'place', 'country_code', null, true, false, 4, 0)::addressline; + location := ROW(null, null, null, hstore('ref', searchcountrycode), 'place', 'country_code', null, true, false, 4, 0)::addressline; RETURN NEXT location; END IF; @@ -2167,12 +2597,12 @@ BEGIN END IF; IF searchhousenumber IS NOT NULL THEN - location := ROW(in_place_id, null, null, 'ref'=>searchhousenumber, 'place', 'house_number', null, true, true, 28, 0)::addressline; + location := ROW(in_place_id, null, null, hstore('ref', searchhousenumber), 'place', 'house_number', null, true, true, 28, 0)::addressline; RETURN NEXT location; END IF; IF searchpostcode IS NOT NULL THEN - location := ROW(null, null, null, 'ref'=>searchpostcode, 'place', 'postcode', null, true, true, 5, 0)::addressline; + location := ROW(null, null, null, hstore('ref', searchpostcode), 'place', 'postcode', null, true, true, 5, 0)::addressline; RETURN NEXT location; END IF; @@ -2267,6 +2697,7 @@ BEGIN name = place.name, housenumber = place.housenumber, street = place.street, + addr_place = place.addr_place, isin = place.isin, postcode = place.postcode, country_code = place.country_code, @@ -2408,146 +2839,340 @@ END; $$ LANGUAGE plpgsql; -CREATE AGGREGATE array_agg(INT[]) -( - sfunc = array_cat, - stype = INT[], - initcond = '{}' -); - -CREATE OR REPLACE FUNCTION tigger_create_interpolation(linegeo GEOMETRY, in_startnumber INTEGER, - in_endnumber INTEGER, interpolationtype TEXT, - in_street TEXT, in_isin TEXT, in_postcode TEXT) RETURNS INTEGER +CREATE OR REPLACE FUNCTION aux_create_property(pointgeo GEOMETRY, in_housenumber TEXT, + in_street TEXT, in_isin TEXT, in_postcode TEXT, in_countrycode char(2)) RETURNS INTEGER AS $$ DECLARE - - startnumber INTEGER; - endnumber INTEGER; - stepsize INTEGER; - housenum INTEGER; + newpoints INTEGER; - 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; + out_postcode TEXT; BEGIN - IF in_endnumber > in_startnumber THEN - startnumber = in_startnumber; - endnumber = in_endnumber; - ELSE - startnumber = in_endnumber; - endnumber = in_startnumber; - END IF; - - numberrange := endnumber - startnumber; - rangestartnumber := startnumber; - - IF (interpolationtype = 'odd' AND startnumber%2 = 0) OR (interpolationtype = 'even' AND startnumber%2 = 1) THEN - startnumber := startnumber + 1; - stepsize := 2; - ELSE - IF (interpolationtype = 'odd' OR interpolationtype = 'even') THEN - stepsize := 2; - ELSE -- everything else assumed to be 'all' - stepsize := 1; - END IF; - END IF; - - -- Filter out really broken tiger data - IF numberrange > 0 AND (numberrange::float/stepsize::float > 500) - AND ST_length(linegeo)/(numberrange::float/stepsize::float) < 0.000001 THEN - RAISE WARNING 'Road too short for number range % to % on %, % (%)',startnumber,endnumber,in_street,in_isin, - ST_length(linegeo)/(numberrange::float/stepsize::float); - RETURN 0; - END IF; - - place_centroid := ST_Centroid(linegeo); - partition := get_partition(place_centroid, 'us'); - parent_place_id := null; + place_centroid := ST_Centroid(pointgeo); + out_partition := get_partition(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 - 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 getNearestRoadFeature(out_partition, place_centroid) 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; - END LOOP; + out_postcode := in_postcode; + 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; 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, - ST_Line_Interpolate_Point(linegeo, (housenum::float-rangestartnumber::float)/numberrange::float)); - newpoints := newpoints + 1; - END LOOP; + insert into location_property_aux (place_id, partition, parent_place_id, housenumber, postcode, centroid) + values (nextval('seq_place'), out_partition, out_parent_place_id, in_housenumber, out_postcode, place_centroid); + newpoints := newpoints + 1; RETURN newpoints; END; $$ LANGUAGE plpgsql; -CREATE OR REPLACE FUNCTION aux_create_property(pointgeo GEOMETRY, in_housenumber TEXT, - in_street TEXT, in_isin TEXT, in_postcode TEXT, in_countrycode char(2)) RETURNS INTEGER +CREATE OR REPLACE FUNCTION get_osm_rel_members(members TEXT[], member TEXT) RETURNS TEXT[] AS $$ DECLARE + result TEXT[]; + i INTEGER; +BEGIN - newpoints INTEGER; - place_centroid GEOMETRY; - partition INTEGER; - out_parent_place_id BIGINT; - location RECORD; - address_street_word_id INTEGER; - out_postcode TEXT; + 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 - place_centroid := ST_Centroid(pointgeo); - partition := get_partition(place_centroid, in_countrycode); - out_parent_place_id := null; + 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; - 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 - out_parent_place_id := location.place_id; - END LOOP; - END IF; + RETURN; +END; +$$ +LANGUAGE plpgsql; - IF out_parent_place_id IS NULL THEN - FOR location IN SELECT place_id FROM getNearestRoadFeature(partition, place_centroid) LOOP - out_parent_place_id := location.place_id; - END LOOP; +-- See: http://stackoverflow.com/questions/6410088/how-can-i-mimic-the-php-urldecode-function-in-postgresql +CREATE OR REPLACE FUNCTION decode_url_part(p varchar) RETURNS varchar + AS $$ +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; +$$ +LANGUAGE plpgsql IMMUTABLE; + +DROP TYPE wikipedia_article_match CASCADE; +create type wikipedia_article_match as ( + language TEXT, + title TEXT, + importance FLOAT +); + +CREATE OR REPLACE FUNCTION get_wikipedia_match(extratags HSTORE, country_code varchar(2)) RETURNS wikipedia_article_match + AS $$ +DECLARE + langs TEXT[]; + i INT; + wiki_article TEXT; + wiki_article_title TEXT; + wiki_article_language TEXT; + result wikipedia_article_match; +BEGIN + langs := ARRAY['english','country','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] in ('english','country') THEN 'wikipedia' ELSE 'wikipedia:'||langs[i] END); + IF wiki_article is not null THEN + wiki_article := regexp_replace(wiki_article,E'^(.*?)([a-z]{2,3}).wikipedia.org/wiki/',E'\\2:'); + wiki_article := regexp_replace(wiki_article,E'^(.*?)([a-z]{2,3}).wikipedia.org/w/index.php\\?title=',E'\\2:'); + wiki_article := regexp_replace(wiki_article,E'^(.*?)/([a-z]{2,3})/wiki/',E'\\2:'); + wiki_article := regexp_replace(wiki_article,E'^(.*?)([a-z]{2,3})[=:]',E'\\2:'); + wiki_article := replace(wiki_article,' ','_'); + wiki_article_title := trim(split_part(wiki_article, ':', 2)); + IF wiki_article_title IS NULL OR wiki_article_title = '' THEN + wiki_article_title := trim(wiki_article); + wiki_article_language := CASE WHEN langs[i] = 'english' THEN 'en' WHEN langs[i] = 'country' THEN get_country_language_code(country_code) ELSE langs[i] END; + ELSE + wiki_article_language := lower(trim(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; +$$ +LANGUAGE plpgsql; + +CREATE OR REPLACE FUNCTION quad_split_geometry(geometry GEOMETRY, maxarea FLOAT, maxdepth INTEGER) + RETURNS SETOF GEOMETRY + AS $$ +DECLARE + xmin FLOAT; + ymin FLOAT; + xmax FLOAT; + ymax FLOAT; + xmid FLOAT; + ymid FLOAT; + secgeo GEOMETRY; + secbox GEOMETRY; + seg INTEGER; + geo RECORD; + area FLOAT; + remainingdepth INTEGER; + added INTEGER; + +BEGIN + +-- RAISE WARNING 'quad_split_geometry: maxarea=%, depth=%',maxarea,maxdepth; + + IF (ST_GeometryType(geometry) not in ('ST_Polygon','ST_MultiPolygon') OR NOT ST_IsValid(geometry)) THEN + RETURN NEXT geometry; + RETURN; END IF; - out_postcode := in_postcode; - IF out_postcode IS NULL THEN - SELECT postcode from placex where place_id = out_parent_place_id INTO out_postcode; + remainingdepth := maxdepth - 1; + area := ST_AREA(geometry); + IF remainingdepth < 1 OR area < maxarea THEN + RETURN NEXT geometry; + RETURN; END IF; - IF out_postcode IS NULL THEN - out_postcode := getNearestPostcode(partition, place_centroid); + + xmin := st_xmin(geometry); + xmax := st_xmax(geometry); + ymin := st_ymin(geometry); + ymax := st_ymax(geometry); + secbox := ST_SetSRID(ST_MakeBox2D(ST_Point(ymin,xmin),ST_Point(ymax,xmax)),4326); + + -- if the geometry completely covers the box don't bother to slice any more + IF ST_AREA(secbox) = area THEN + RETURN NEXT geometry; + RETURN; 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); - newpoints := newpoints + 1; + xmid := (xmin+xmax)/2; + ymid := (ymin+ymax)/2; - RETURN newpoints; + added := 0; + FOR seg IN 1..4 LOOP + + IF seg = 1 THEN + secbox := ST_SetSRID(ST_MakeBox2D(ST_Point(xmin,ymin),ST_Point(xmid,ymid)),4326); + END IF; + IF seg = 2 THEN + secbox := ST_SetSRID(ST_MakeBox2D(ST_Point(xmin,ymid),ST_Point(xmid,ymax)),4326); + END IF; + IF seg = 3 THEN + secbox := ST_SetSRID(ST_MakeBox2D(ST_Point(xmid,ymin),ST_Point(xmax,ymid)),4326); + END IF; + IF seg = 4 THEN + secbox := ST_SetSRID(ST_MakeBox2D(ST_Point(xmid,ymid),ST_Point(xmax,ymax)),4326); + END IF; + + IF st_intersects(geometry, secbox) THEN + secgeo := st_intersection(geometry, secbox); + IF NOT ST_IsEmpty(secgeo) AND ST_GeometryType(secgeo) in ('ST_Polygon','ST_MultiPolygon') THEN + FOR geo IN select quad_split_geometry(secgeo, maxarea, remainingdepth) as geom LOOP + IF NOT ST_IsEmpty(geo.geom) AND ST_GeometryType(geo.geom) in ('ST_Polygon','ST_MultiPolygon') THEN + added := added + 1; + RETURN NEXT geo.geom; + END IF; + END LOOP; + END IF; + END IF; + END LOOP; + + RETURN; +END; +$$ +LANGUAGE plpgsql; + +CREATE OR REPLACE FUNCTION split_geometry(geometry GEOMETRY) + RETURNS SETOF GEOMETRY + AS $$ +DECLARE + geo RECORD; +BEGIN + -- 10000000000 is ~~ 1x1 degree + FOR geo IN select quad_split_geometry(geometry, 0.25, 20) as geom LOOP + RETURN NEXT geo.geom; + END LOOP; + RETURN; +END; +$$ +LANGUAGE plpgsql; + + +CREATE OR REPLACE FUNCTION place_force_delete(placeid BIGINT) RETURNS BOOLEAN + AS $$ +DECLARE + osmid BIGINT; + osmtype character(1); + pclass text; + ptype text; +BEGIN + SELECT osm_type, osm_id, class, type FROM placex WHERE place_id = placeid INTO osmtype, osmid, pclass, ptype; + DELETE FROM import_polygon_delete where osm_type = osmtype and osm_id = osmid and class = pclass and type = ptype; + DELETE FROM import_polygon_error where osm_type = osmtype and osm_id = osmid and class = pclass and type = ptype; + -- force delete from place/placex by making it a very small geometry + UPDATE place set geometry = ST_SetSRID(ST_Point(0,0), 4326) where osm_type = osmtype and osm_id = osmid and class = pclass and type = ptype; + DELETE FROM place where osm_type = osmtype and osm_id = osmid and class = pclass and type = ptype; + + RETURN TRUE; +END; +$$ +LANGUAGE plpgsql; + +CREATE OR REPLACE FUNCTION place_force_update(placeid BIGINT) RETURNS BOOLEAN + AS $$ +DECLARE + placegeom GEOMETRY; + geom GEOMETRY; + diameter FLOAT; + rank INTEGER; +BEGIN + UPDATE placex SET indexed_status = 2 WHERE place_id = placeid; + SELECT geometry, rank_search FROM placex WHERE place_id = placeid INTO placegeom, rank; + IF placegeom IS NOT NULL AND ST_IsValid(placegeom) THEN + 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 addr_place is not null)); + 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 addr_place is not null)); + END LOOP; + ELSE + diameter := 0; + IF rank = 11 THEN + diameter := 0.05; + ELSEIF rank < 18 THEN + diameter := 0.1; + ELSEIF rank < 20 THEN + diameter := 0.05; + ELSEIF rank = 21 THEN + diameter := 0.001; + ELSEIF rank < 24 THEN + diameter := 0.02; + ELSEIF rank < 26 THEN + diameter := 0.002; -- 100 to 200 meters + ELSEIF rank < 28 THEN + diameter := 0.001; -- 50 to 100 meters + END IF; + IF diameter > 0 THEN + IF rank >= 26 THEN + -- roads may cause reparenting for >27 rank places + update placex set indexed_status = 2 where indexed_status = 0 and rank_search > rank and ST_DWithin(placex.geometry, placegeom, diameter); + ELSEIF rank >= 16 THEN + -- up to rank 16, street-less addresses may need reparenting + update placex set indexed_status = 2 where indexed_status = 0 and rank_search > rank and ST_DWithin(placex.geometry, placegeom, diameter) and (rank_search < 28 or name is not null or addr_place is not null); + ELSE + -- for all other places the search terms may change as well + update placex set indexed_status = 2 where indexed_status = 0 and rank_search > rank and ST_DWithin(placex.geometry, placegeom, diameter) and (rank_search < 28 or name is not null); + END IF; + END IF; + END IF; + RETURN TRUE; + END IF; + + RETURN FALSE; END; $$ LANGUAGE plpgsql;