+-- Splits the line at the given point and returns the two parts
+-- in a multilinestring.
+CREATE OR REPLACE FUNCTION split_line_on_node(line GEOMETRY, point GEOMETRY)
+RETURNS GEOMETRY
+ AS $$
+BEGIN
+ RETURN ST_Split(ST_Snap(line, point, 0.0005), point);
+END;
+$$
+LANGUAGE plpgsql;
+
+
CREATE OR REPLACE FUNCTION geometry_sector(partition INTEGER, place geometry) RETURNS INTEGER
AS $$
DECLARE
DECLARE
o TEXT;
BEGIN
- o := gettokenstring(transliteration(name));
+ o := public.gettokenstring(public.transliteration(name));
RETURN trim(substr(o,1,length(o)));
END;
$$
$$
LANGUAGE plpgsql;
+CREATE OR REPLACE FUNCTION getorcreate_postcode_id(postcode TEXT)
+ RETURNS INTEGER
+ AS $$
+DECLARE
+ lookup_token TEXT;
+ lookup_word TEXT;
+ return_word_id INTEGER;
+BEGIN
+ lookup_word := upper(trim(postcode));
+ lookup_token := ' ' || make_standard_name(lookup_word);
+ SELECT min(word_id) FROM word WHERE word_token = lookup_token and class='place' and type='postcode' into return_word_id;
+ IF return_word_id IS NULL THEN
+ return_word_id := nextval('seq_word');
+ INSERT INTO word VALUES (return_word_id, lookup_token, lookup_word, 'place', 'postcode', null, 0);
+ END IF;
+ RETURN return_word_id;
+END;
+$$
+LANGUAGE plpgsql;
+
CREATE OR REPLACE FUNCTION getorcreate_country(lookup_word TEXT, lookup_country_code varchar(2))
RETURNS INTEGER
AS $$
$$
LANGUAGE plpgsql;
-CREATE OR REPLACE FUNCTION getorcreate_amenity(lookup_word TEXT, lookup_class text, lookup_type text)
+CREATE OR REPLACE FUNCTION getorcreate_amenity(lookup_word TEXT, normalized_word TEXT, lookup_class text, lookup_type text)
RETURNS INTEGER
AS $$
DECLARE
return_word_id INTEGER;
BEGIN
lookup_token := ' '||trim(lookup_word);
- SELECT min(word_id) FROM word WHERE word_token = lookup_token and class=lookup_class and type = lookup_type into return_word_id;
+ SELECT min(word_id) FROM word WHERE word_token = lookup_token and word=normalized_word and class=lookup_class and type = lookup_type into return_word_id;
IF return_word_id IS NULL THEN
return_word_id := nextval('seq_word');
- INSERT INTO word VALUES (return_word_id, lookup_token, null, lookup_class, lookup_type, null, 0);
+ INSERT INTO word VALUES (return_word_id, lookup_token, normalized_word, lookup_class, lookup_type, null, 0);
END IF;
RETURN return_word_id;
END;
$$
LANGUAGE plpgsql;
-CREATE OR REPLACE FUNCTION getorcreate_amenityoperator(lookup_word TEXT, lookup_class text, lookup_type text, op text)
+CREATE OR REPLACE FUNCTION getorcreate_amenityoperator(lookup_word TEXT, normalized_word TEXT, lookup_class text, lookup_type text, op text)
RETURNS INTEGER
AS $$
DECLARE
return_word_id INTEGER;
BEGIN
lookup_token := ' '||trim(lookup_word);
- SELECT min(word_id) FROM word WHERE word_token = lookup_token and class=lookup_class and type = lookup_type and operator = op into return_word_id;
+ SELECT min(word_id) FROM word WHERE word_token = lookup_token and word=normalized_word and class=lookup_class and type = lookup_type and operator = op into return_word_id;
IF return_word_id IS NULL THEN
return_word_id := nextval('seq_word');
- INSERT INTO word VALUES (return_word_id, lookup_token, null, lookup_class, lookup_type, null, 0, op);
+ INSERT INTO word VALUES (return_word_id, lookup_token, normalized_word, lookup_class, lookup_type, null, 0, op);
END IF;
RETURN return_word_id;
END;
$$
LANGUAGE plpgsql IMMUTABLE;
+CREATE OR REPLACE FUNCTION reverse_place_diameter(rank_search SMALLINT)
+ RETURNS FLOAT
+ AS $$
+BEGIN
+ IF rank_search <= 4 THEN
+ RETURN 5.0;
+ ELSIF rank_search <= 8 THEN
+ RETURN 1.8;
+ ELSIF rank_search <= 12 THEN
+ RETURN 0.6;
+ ELSIF rank_search <= 17 THEN
+ RETURN 0.16;
+ ELSIF rank_search <= 18 THEN
+ RETURN 0.08;
+ ELSIF rank_search <= 19 THEN
+ RETURN 0.04;
+ END IF;
+
+ RETURN 0.02;
+END;
+$$
+LANGUAGE plpgsql IMMUTABLE;
+
+CREATE OR REPLACE FUNCTION get_postcode_rank(country_code VARCHAR(2), postcode TEXT,
+ OUT rank_search SMALLINT, OUT rank_address SMALLINT)
+AS $$
+DECLARE
+ part TEXT;
+BEGIN
+ rank_search := 30;
+ rank_address := 30;
+ postcode := upper(postcode);
+
+ IF country_code = 'gb' THEN
+ IF postcode ~ '^([A-Z][A-Z]?[0-9][0-9A-Z]? [0-9][A-Z][A-Z])$' THEN
+ rank_search := 25;
+ rank_address := 5;
+ ELSEIF postcode ~ '^([A-Z][A-Z]?[0-9][0-9A-Z]? [0-9])$' THEN
+ rank_search := 23;
+ rank_address := 5;
+ ELSEIF postcode ~ '^([A-Z][A-Z]?[0-9][0-9A-Z])$' THEN
+ rank_search := 21;
+ rank_address := 5;
+ END IF;
+
+ ELSEIF country_code = 'sg' THEN
+ IF postcode ~ '^([0-9]{6})$' THEN
+ rank_search := 25;
+ rank_address := 11;
+ END IF;
+
+ ELSEIF country_code = 'de' THEN
+ IF postcode ~ '^([0-9]{5})$' THEN
+ rank_search := 21;
+ rank_address := 11;
+ END IF;
+
+ ELSE
+ -- Guess at the postcode format and coverage (!)
+ IF postcode ~ '^[A-Z0-9]{1,5}$' THEN -- Probably too short to be very local
+ rank_search := 21;
+ rank_address := 11;
+ ELSE
+ -- Does it look splitable into and area and local code?
+ part := substring(postcode from '^([- :A-Z0-9]+)([- :][A-Z0-9]+)$');
+
+ IF part IS NOT NULL THEN
+ rank_search := 25;
+ rank_address := 11;
+ ELSEIF postcode ~ '^[- :A-Z0-9]{6,}$' THEN
+ rank_search := 21;
+ rank_address := 11;
+ END IF;
+ END IF;
+ END IF;
+
+END;
+$$
+LANGUAGE plpgsql IMMUTABLE;
+
+-- Find the nearest artificial postcode for the given geometry.
+-- TODO For areas there should not be more than two inside the geometry.
+CREATE OR REPLACE FUNCTION get_nearest_postcode(country VARCHAR(2), geom GEOMETRY) RETURNS TEXT
+ AS $$
+DECLARE
+ outcode TEXT;
+ cnt INTEGER;
+BEGIN
+ -- If the geometry is an area then only one postcode must be within
+ -- that area, otherwise consider the area as not having a postcode.
+ IF ST_GeometryType(geom) in ('ST_Polygon','ST_MultiPolygon') THEN
+ SELECT min(postcode), count(*) FROM
+ (SELECT postcode FROM location_postcode
+ WHERE ST_Contains(geom, location_postcode.geometry) LIMIT 2) sub
+ INTO outcode, cnt;
+
+ IF cnt = 1 THEN
+ RETURN outcode;
+ ELSE
+ RETURN null;
+ END IF;
+ END IF;
+
+ SELECT postcode FROM location_postcode
+ WHERE ST_DWithin(geom, location_postcode.geometry, 0.05)
+ AND location_postcode.country_code = country
+ ORDER BY ST_Distance(geom, location_postcode.geometry) LIMIT 1
+ INTO outcode;
+
+ RETURN outcode;
+END;
+$$
+LANGUAGE plpgsql;
+
+
CREATE OR REPLACE FUNCTION create_country(src HSTORE, lookup_country_code varchar(2)) RETURNS VOID
AS $$
DECLARE
BEGIN
place_centre := ST_PointOnSurface(place);
---DEBUG: RAISE WARNING 'get_country_code, start: %', ST_AsText(place_centre);
+-- RAISE WARNING 'get_country_code, start: %', ST_AsText(place_centre);
-- Try for a OSM polygon
FOR nearcountry IN select country_code from location_area_country where country_code is not null and not isguess and st_covers(geometry, place_centre) limit 1
RETURN nearcountry.country_code;
END LOOP;
---DEBUG: RAISE WARNING 'osm fallback: %', ST_AsText(place_centre);
+-- RAISE WARNING 'osm fallback: %', ST_AsText(place_centre);
-- Try for OSM fallback data
-- The order is to deal with places like HongKong that are 'states' within another polygon
RETURN nearcountry.country_code;
END LOOP;
---DEBUG: RAISE WARNING 'natural earth: %', ST_AsText(place_centre);
+-- RAISE WARNING 'natural earth: %', ST_AsText(place_centre);
-- Natural earth data
FOR nearcountry IN select country_code from country_naturalearthdata where st_covers(geometry, place_centre) limit 1
RETURN nearcountry.country_code;
END LOOP;
---DEBUG: RAISE WARNING 'near osm fallback: %', ST_AsText(place_centre);
+-- RAISE WARNING 'near osm fallback: %', ST_AsText(place_centre);
--
FOR nearcountry IN select country_code from country_osm_grid where st_dwithin(geometry, place_centre, 0.5) order by st_distance(geometry, place_centre) asc, area asc limit 1
RETURN nearcountry.country_code;
END LOOP;
---DEBUG: RAISE WARNING 'near natural earth: %', ST_AsText(place_centre);
+-- RAISE WARNING 'near natural earth: %', ST_AsText(place_centre);
-- Natural earth data
FOR nearcountry IN select country_code from country_naturalearthdata where st_dwithin(geometry, place_centre, 0.5) limit 1
keywords INTEGER[],
rank_search INTEGER,
rank_address INTEGER,
+ in_postcode TEXT,
geometry GEOMETRY
)
RETURNS BOOLEAN
AS $$
DECLARE
locationid INTEGER;
- isarea BOOLEAN;
centroid GEOMETRY;
diameter FLOAT;
x BOOLEAN;
splitGeom RECORD;
secgeo GEOMETRY;
+ postcode TEXT;
BEGIN
IF rank_search > 25 THEN
RAISE EXCEPTION 'Adding location with rank > 25 (% rank %)', place_id, rank_search;
END IF;
--- RAISE WARNING 'Adding location with rank > 25 (% rank %)', place_id, rank_search;
-
x := deleteLocationArea(partition, place_id, rank_search);
- isarea := false;
- IF (ST_GeometryType(geometry) in ('ST_Polygon','ST_MultiPolygon') AND ST_IsValid(geometry)) THEN
+ -- add postcode only if it contains a single entry, i.e. ignore postcode lists
+ postcode := NULL;
+ IF in_postcode is not null AND in_postcode not similar to '%(,|;)%' THEN
+ postcode := upper(trim (in_postcode));
+ END IF;
- isArea := true;
+ IF ST_GeometryType(geometry) in ('ST_Polygon','ST_MultiPolygon') THEN
centroid := ST_Centroid(geometry);
FOR secgeo IN select split_geometry(geometry) AS geom LOOP
- x := insertLocationAreaLarge(partition, place_id, country_code, keywords, rank_search, rank_address, false, centroid, secgeo);
+ x := insertLocationAreaLarge(partition, place_id, country_code, keywords, rank_search, rank_address, false, postcode, centroid, secgeo);
END LOOP;
ELSE
-- RAISE WARNING 'adding % diameter %', place_id, diameter;
secgeo := ST_Buffer(geometry, diameter);
- x := insertLocationAreaLarge(partition, place_id, country_code, keywords, rank_search, rank_address, true, ST_Centroid(geometry), secgeo);
+ x := insertLocationAreaLarge(partition, place_id, country_code, keywords, rank_search, rank_address, true, postcode, ST_Centroid(geometry), secgeo);
END IF;
LANGUAGE plpgsql;
-
--- find the parant road of an interpolation
+-- find the parent road of the cut road parts
CREATE OR REPLACE FUNCTION get_interpolation_parent(wayid BIGINT, street TEXT, place TEXT,
partition INTEGER, centroid GEOMETRY, geom GEOMETRY)
RETURNS BIGINT AS $$
IF addr_street is null and addr_place is null THEN
select nodes from planet_osm_ways where id = wayid INTO waynodes;
- FOR location IN SELECT placex.street, placex.addr_place from placex
+ FOR location IN SELECT placex.address from placex
where osm_type = 'N' and osm_id = ANY(waynodes)
- and (placex.street is not null or placex.addr_place is not null)
+ and placex.address is not null
+ and (placex.address ? 'street' or placex.address ? 'place')
and indexed_status < 100
limit 1 LOOP
- addr_street = location.street;
- addr_place = location.addr_place;
+ addr_street = location.address->'street';
+ addr_place = location.address->'place';
END LOOP;
END IF;
$$
LANGUAGE plpgsql;
-CREATE OR REPLACE FUNCTION create_interpolation(wayid BIGINT, interpolationtype TEXT,
- parent_id BIGINT, partition INTEGER,
- country_code TEXT, geometry_sector INTEGER,
- defpostalcode TEXT, geom GEOMETRY) RETURNS INTEGER
- AS $$
-DECLARE
-
- newpoints INTEGER;
- waynodes BIGINT[];
- nodeid BIGINT;
- prevnode RECORD;
- nextnode RECORD;
- startnumber INTEGER;
- endnumber INTEGER;
- stepsize INTEGER;
- orginalstartnumber INTEGER;
- originalnumberrange INTEGER;
- housenum INTEGER;
- linegeo GEOMETRY;
- splitline GEOMETRY;
- sectiongeo GEOMETRY;
- pointgeo GEOMETRY;
+CREATE OR REPLACE FUNCTION osmline_insert() RETURNS TRIGGER
+ AS $$
BEGIN
- delete from placex where osm_type = 'W' and osm_id = wayid
- and class = 'place' and type = 'address';
-
- IF interpolationtype = 'odd' OR interpolationtype = 'even' THEN
- stepsize := 2;
- ELSEIF interpolationtype = 'all' THEN
- stepsize := 1;
- ELSEIF interpolationtype ~ '^\d+$' THEN
- stepsize := interpolationtype::INTEGER;
- ELSE
- RETURN 0;
- END IF;
-
- select nodes from planet_osm_ways where id = wayid INTO waynodes;
-
- IF array_upper(waynodes, 1) IS NULL THEN
- RETURN 0;
- END IF;
-
- linegeo := geom;
- startnumber := NULL;
- newpoints := 0;
-
- FOR nodeidpos in 1..array_upper(waynodes, 1) LOOP
-
- -- 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 * from placex where osm_type = 'N' and osm_id = waynodes[nodeidpos]::BIGINT
- and indexed_status < 100 and housenumber is not NULL
- order by (type = 'address'),place_id limit 1 INTO nextnode;
- IF nextnode.place_id IS NOT NULL THEN
-
- IF nodeidpos > 1 and nodeidpos < array_upper(waynodes, 1) THEN
- -- Make sure that the point is actually on the line. That might
- -- be a bit paranoid but ensures that the algorithm still works
- -- should osm2pgsql attempt to repair geometries.
- splitline := split_line_on_node(linegeo, nextnode.geometry);
- sectiongeo := ST_GeometryN(splitline, 1);
- linegeo := ST_GeometryN(splitline, 2);
- ELSE
- sectiongeo = linegeo;
- END IF;
- endnumber := substring(nextnode.housenumber,'[0-9]+')::integer;
-
- IF startnumber IS NOT NULL AND endnumber IS NOT NULL
- AND @(startnumber - endnumber) < 1000 AND startnumber != endnumber
- AND ST_GeometryType(sectiongeo) = 'ST_LineString' THEN
-
- IF (startnumber > endnumber) THEN
- housenum := endnumber;
- endnumber := startnumber;
- startnumber := housenum;
- sectiongeo := ST_Reverse(sectiongeo);
- END IF;
- orginalstartnumber := startnumber;
- originalnumberrange := endnumber - startnumber;
-
- startnumber := startnumber + stepsize;
- -- correct for odd/even
- IF (interpolationtype = 'odd' AND startnumber%2 = 0)
- OR (interpolationtype = 'even' AND startnumber%2 = 1) THEN
- startnumber := startnumber - 1;
- END IF;
- endnumber := endnumber - 1;
-
- -- keep for compatibility with previous versions
- delete from placex where osm_type = 'N' and osm_id = prevnode.osm_id
- and place_id != prevnode.place_id and class = 'place'
- and type = 'house';
- FOR housenum IN startnumber..endnumber BY stepsize LOOP
- pointgeo := ST_LineInterpolatePoint(sectiongeo, (housenum::float-orginalstartnumber::float)/originalnumberrange::float);
- insert into placex (place_id, partition, osm_type, osm_id,
- class, type, admin_level, housenumber,
- postcode,
- country_code, parent_place_id, rank_address, rank_search,
- indexed_status, indexed_date, geometry_sector,
- calculated_country_code, centroid, geometry)
- values (nextval('seq_place'), partition, 'W', wayid,
- 'place', 'address', prevnode.admin_level, housenum,
- coalesce(prevnode.postcode, defpostalcode),
- prevnode.country_code, parent_id, 30, 30,
- 0, now(), geometry_sector, country_code,
- pointgeo, pointgeo);
- newpoints := newpoints + 1;
---RAISE WARNING 'interpolation number % % ',prevnode.place_id,housenum;
- END LOOP;
- END IF;
+ NEW.place_id := nextval('seq_place');
+ NEW.indexed_date := now();
- -- early break if we are out of line string,
- -- might happen when a line string loops back on itself
- IF ST_GeometryType(linegeo) != 'ST_LineString' THEN
- RETURN newpoints;
- END IF;
+ IF NEW.indexed_status IS NULL THEN
+ IF NEW.address is NULL OR NOT NEW.address ? 'interpolation'
+ OR NEW.address->'interpolation' NOT IN ('odd', 'even', 'all') THEN
+ -- other interpolation types than odd/even/all (e.g. numeric ones) are not supported
+ RETURN NULL;
+ END IF;
- startnumber := substring(nextnode.housenumber,'[0-9]+')::integer;
- prevnode := nextnode;
- END IF;
- END LOOP;
+ NEW.indexed_status := 1; --STATUS_NEW
+ NEW.country_code := lower(get_country_code(NEW.linegeo));
---RAISE WARNING 'interpolation points % ',newpoints;
+ NEW.partition := get_partition(NEW.country_code);
+ NEW.geometry_sector := geometry_sector(NEW.partition, NEW.linegeo);
+ END IF;
- RETURN newpoints;
+ RETURN NEW;
END;
$$
LANGUAGE plpgsql;
+
CREATE OR REPLACE FUNCTION placex_insert() RETURNS TRIGGER
AS $$
DECLARE
default_language VARCHAR(10);
diameter FLOAT;
classtable TEXT;
+ line RECORD;
BEGIN
- --DEBUG: RAISE WARNING '% %',NEW.osm_type,NEW.osm_id;
-
- -- ignore interpolated addresses
- IF NEW.class = 'place' and NEW.type = 'address' THEN
- RETURN NEW;
- END IF;
-
- 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;
- RETURN NULL;
- 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.calculated_country_code := lower(get_country_code(NEW.geometry));
+ NEW.country_code := lower(get_country_code(NEW.geometry));
- NEW.partition := get_partition(NEW.calculated_country_code);
+ NEW.partition := get_partition(NEW.country_code);
NEW.geometry_sector := geometry_sector(NEW.partition, NEW.geometry);
-- copy 'name' to or from the default language (if there is a default language)
IF NEW.name is not null AND array_upper(akeys(NEW.name),1) > 1 THEN
- default_language := get_country_language_code(NEW.calculated_country_code);
+ default_language := get_country_language_code(NEW.country_code);
IF default_language IS NOT NULL THEN
IF NEW.name ? 'name' AND NOT NEW.name ? ('name:'||default_language) THEN
NEW.name := NEW.name || hstore(('name:'||default_language), (NEW.name -> 'name'));
END IF;
END IF;
- IF NEW.admin_level > 15 THEN
- NEW.admin_level := 15;
- END IF;
-
- IF NEW.housenumber IS NOT NULL THEN
- i := getorcreate_housenumber_id(make_standard_name(NEW.housenumber));
- END IF;
-
IF NEW.osm_type = 'X' THEN
-- E'X'ternal records should already be in the right format so do nothing
ELSE
-- 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
- IF NEW.postcode IS NULL THEN
+ IF NEW.address IS NULL OR NOT NEW.address ? 'postcode' 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.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;
- NEW.rank_address := 5;
- ELSEIF NEW.postcode ~ '^([A-Z][A-Z]?[0-9][0-9A-Z]? [0-9])$' THEN
- NEW.rank_search := 23;
- NEW.rank_address := 5;
- ELSEIF NEW.postcode ~ '^([A-Z][A-Z]?[0-9][0-9A-Z])$' THEN
- NEW.rank_search := 21;
- NEW.rank_address := 5;
- END IF;
-
- ELSEIF NEW.calculated_country_code = 'sg' THEN
+ NEW.name := hstore('ref', NEW.address->'postcode');
- IF NEW.postcode ~ '^([0-9]{6})$' THEN
- NEW.rank_search := 25;
- NEW.rank_address := 11;
- END IF;
-
- ELSEIF NEW.calculated_country_code = 'de' THEN
-
- IF NEW.postcode ~ '^([0-9]{5})$' THEN
- NEW.rank_search := 21;
- NEW.rank_address := 11;
- END IF;
+ SELECT * FROM get_postcode_rank(NEW.country_code, NEW.address->'postcode')
+ INTO NEW.rank_search, NEW.rank_address;
- ELSE
- -- Guess at the postcode format and coverage (!)
- IF upper(NEW.postcode) ~ '^[A-Z0-9]{1,5}$' THEN -- Probably too short to be very local
- NEW.rank_search := 21;
- NEW.rank_address := 11;
- ELSE
- -- Does it look splitable into and area and local code?
- postcode := substring(upper(NEW.postcode) from '^([- :A-Z0-9]+)([- :][A-Z0-9]+)$');
-
- IF postcode IS NOT NULL THEN
- NEW.rank_search := 25;
- NEW.rank_address := 11;
- ELSEIF NEW.postcode ~ '^[- :A-Z0-9]{6,}$' THEN
- NEW.rank_search := 21;
- NEW.rank_address := 11;
- END IF;
- END IF;
+ IF NOT ST_GeometryType(NEW.geometry) IN ('ST_Polygon','ST_MultiPolygon') THEN
+ NEW.rank_address := 0;
END IF;
ELSEIF NEW.class = 'place' THEN
IF NEW.type in ('continent') THEN
NEW.rank_search := 2;
NEW.rank_address := NEW.rank_search;
- NEW.calculated_country_code := NULL;
+ NEW.country_code := NULL;
ELSEIF NEW.type in ('sea') THEN
NEW.rank_search := 2;
NEW.rank_address := 0;
- NEW.calculated_country_code := NULL;
+ NEW.country_code := NULL;
ELSEIF NEW.type in ('country') THEN
NEW.rank_search := 4;
NEW.rank_address := NEW.rank_search;
ELSE
NEW.rank_address := 0;
END IF;
+ ELSEIF NEW.class = 'leisure' and NEW.type in ('park') THEN
+ NEW.rank_search := 24;
+ NEW.rank_address := 0;
ELSEIF NEW.class = 'natural' and NEW.type in ('peak','volcano','mountain_range') THEN
NEW.rank_search := 18;
NEW.rank_address := 0;
-- a country code make no sense below rank 4 (country)
IF NEW.rank_search < 4 THEN
- NEW.calculated_country_code := NULL;
+ NEW.country_code := NULL;
END IF;
-- Block import below rank 22
--DEBUG: RAISE WARNING 'placex_insert:END: % % % %',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
+ RETURN NEW; -- %DIFFUPDATES% The following is not needed until doing diff updates, and slows the main index process down
IF NEW.rank_address > 0 THEN
IF (ST_GeometryType(NEW.geometry) in ('ST_Polygon','ST_MultiPolygon') AND ST_IsValid(NEW.geometry)) THEN
-- 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));
+ 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 address ? 'place'));
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));
+ 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 address ? 'place'));
END IF;
ELSE
-- mark nearby items for re-indexing, where 'nearby' depends on the features rank_search and is a complete guess :(
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);
+ -- reparenting also for OSM Interpolation Lines (and for Tiger?)
+ update location_property_osmline set indexed_status = 2 where indexed_status = 0 and ST_DWithin(location_property_osmline.linegeo, 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);
+ 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 address ? 'place');
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);
$$
LANGUAGE plpgsql;
-CREATE OR REPLACE FUNCTION placex_update() RETURNS
+CREATE OR REPLACE FUNCTION osmline_update() RETURNS
+TRIGGER
+ AS $$
+DECLARE
+ place_centroid GEOMETRY;
+ waynodes BIGINT[];
+ prevnode RECORD;
+ nextnode RECORD;
+ startnumber INTEGER;
+ endnumber INTEGER;
+ housenum INTEGER;
+ linegeo GEOMETRY;
+ splitline GEOMETRY;
+ sectiongeo GEOMETRY;
+ interpol_postcode TEXT;
+ postcode TEXT;
+BEGIN
+ -- deferred delete
+ IF OLD.indexed_status = 100 THEN
+ delete from location_property_osmline 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;
+
+ NEW.interpolationtype = NEW.address->'interpolation';
+
+ place_centroid := ST_PointOnSurface(NEW.linegeo);
+ NEW.parent_place_id = get_interpolation_parent(NEW.osm_id, NEW.address->'street',
+ NEW.address->'place',
+ NEW.partition, place_centroid, NEW.linegeo);
+
+ IF NEW.address is not NULL AND NEW.address ? 'postcode' AND NEW.address->'postcode' not similar to '%(,|;)%' THEN
+ interpol_postcode := NEW.address->'postcode';
+ housenum := getorcreate_postcode_id(NEW.address->'postcode');
+ ELSE
+ interpol_postcode := NULL;
+ END IF;
+
+ -- if the line was newly inserted, split the line as necessary
+ IF OLD.indexed_status = 1 THEN
+ select nodes from planet_osm_ways where id = NEW.osm_id INTO waynodes;
+
+ IF array_upper(waynodes, 1) IS NULL THEN
+ RETURN NEW;
+ END IF;
+
+ linegeo := NEW.linegeo;
+ startnumber := NULL;
+
+ FOR nodeidpos in 1..array_upper(waynodes, 1) LOOP
+
+ select osm_id, address, geometry
+ from place where osm_type = 'N' and osm_id = waynodes[nodeidpos]::BIGINT
+ and address is not NULL and address ? 'housenumber' limit 1 INTO nextnode;
+ --RAISE NOTICE 'Nextnode.place_id: %s', nextnode.place_id;
+ IF nextnode.osm_id IS NOT NULL THEN
+ --RAISE NOTICE 'place_id is not null';
+ IF nodeidpos > 1 and nodeidpos < array_upper(waynodes, 1) THEN
+ -- Make sure that the point is actually on the line. That might
+ -- be a bit paranoid but ensures that the algorithm still works
+ -- should osm2pgsql attempt to repair geometries.
+ splitline := split_line_on_node(linegeo, nextnode.geometry);
+ sectiongeo := ST_GeometryN(splitline, 1);
+ linegeo := ST_GeometryN(splitline, 2);
+ ELSE
+ sectiongeo = linegeo;
+ END IF;
+ endnumber := substring(nextnode.address->'housenumber','[0-9]+')::integer;
+
+ IF startnumber IS NOT NULL AND endnumber IS NOT NULL
+ AND startnumber != endnumber
+ AND ST_GeometryType(sectiongeo) = 'ST_LineString' THEN
+
+ IF (startnumber > endnumber) THEN
+ housenum := endnumber;
+ endnumber := startnumber;
+ startnumber := housenum;
+ sectiongeo := ST_Reverse(sectiongeo);
+ END IF;
+
+ -- determine postcode
+ postcode := coalesce(interpol_postcode,
+ prevnode.address->'postcode',
+ nextnode.address->'postcode',
+ postcode);
+
+ IF postcode is NULL THEN
+ SELECT placex.postcode FROM placex WHERE place_id = NEW.parent_place_id INTO postcode;
+ END IF;
+ IF postcode is NULL THEN
+ postcode := get_nearest_postcode(NEW.country_code, nextnode.geometry);
+ END IF;
+
+ IF NEW.startnumber IS NULL THEN
+ NEW.startnumber := startnumber;
+ NEW.endnumber := endnumber;
+ NEW.linegeo := sectiongeo;
+ NEW.postcode := upper(trim(postcode));
+ ELSE
+ insert into location_property_osmline
+ (linegeo, partition, osm_id, parent_place_id,
+ startnumber, endnumber, interpolationtype,
+ address, postcode, country_code,
+ geometry_sector, indexed_status)
+ values (sectiongeo, NEW.partition, NEW.osm_id, NEW.parent_place_id,
+ startnumber, endnumber, NEW.interpolationtype,
+ NEW.address, postcode,
+ NEW.country_code, NEW.geometry_sector, 0);
+ END IF;
+ END IF;
+
+ -- early break if we are out of line string,
+ -- might happen when a line string loops back on itself
+ IF ST_GeometryType(linegeo) != 'ST_LineString' THEN
+ RETURN NEW;
+ END IF;
+
+ startnumber := substring(nextnode.address->'housenumber','[0-9]+')::integer;
+ prevnode := nextnode;
+ END IF;
+ END LOOP;
+ END IF;
+
+ -- marking descendants for reparenting is not needed, because there are
+ -- actually no descendants for interpolation lines
+ RETURN NEW;
+END;
+$$
+LANGUAGE plpgsql;
+
+-- Trigger for updates of location_postcode
+--
+-- Computes the parent object the postcode most likely refers to.
+-- This will be the place that determines the address displayed when
+-- searching for this postcode.
+CREATE OR REPLACE FUNCTION postcode_update() RETURNS
+TRIGGER
+ AS $$
+DECLARE
+ partition SMALLINT;
+ location RECORD;
+BEGIN
+ IF NEW.indexed_status != 0 OR OLD.indexed_status = 0 THEN
+ RETURN NEW;
+ END IF;
+
+ NEW.indexed_date = now();
+
+ partition := get_partition(NEW.country_code);
+
+ SELECT * FROM get_postcode_rank(NEW.country_code, NEW.postcode)
+ INTO NEW.rank_search, NEW.rank_address;
+
+ NEW.parent_place_id = 0;
+ FOR location IN
+ SELECT place_id
+ FROM getNearFeatures(partition, NEW.geometry, NEW.rank_search, '{}'::int[])
+ WHERE NOT isguess ORDER BY rank_address DESC LIMIT 1
+ LOOP
+ NEW.parent_place_id = location.place_id;
+ END LOOP;
+
+ RETURN NEW;
+END;
+$$
+LANGUAGE plpgsql;
+
+CREATE OR REPLACE FUNCTION placex_update() RETURNS
TRIGGER
AS $$
DECLARE
relation_members TEXT[];
relMember RECORD;
linkedplacex RECORD;
+ addr_item RECORD;
search_diameter FLOAT;
search_prevdiameter FLOAT;
search_maxrank INTEGER;
address_street_word_id INTEGER;
address_street_word_ids INTEGER[];
parent_place_id_rank BIGINT;
-
+
+ addr_street TEXT;
+ addr_place TEXT;
+
isin TEXT[];
isin_tokens INT[];
nameaddress_vector INTEGER[];
linked_node_id BIGINT;
+ linked_importance FLOAT;
+ linked_wikipedia TEXT;
result BOOLEAN;
BEGIN
-
-- deferred delete
IF OLD.indexed_status = 100 THEN
- --DEBUG: RAISE WARNING 'placex_update_delete % %',NEW.osm_type,NEW.osm_id;
+ --DEBUG: RAISE WARNING 'placex_update delete % %',NEW.osm_type,NEW.osm_id;
delete from placex where place_id = OLD.place_id;
RETURN NULL;
END IF;
RETURN NEW;
END IF;
- -- ignore interpolated addresses
- IF NEW.class = 'place' and NEW.type = 'address' THEN
- RETURN NEW;
- END IF;
+ --DEBUG: RAISE WARNING 'placex_update % % (%)',NEW.osm_type,NEW.osm_id,NEW.place_id;
- --DEBUG: RAISE WARNING 'placex_update % %',NEW.osm_type,NEW.osm_id;
+ NEW.indexed_date = now();
---RAISE WARNING '%',NEW.place_id;
---RAISE WARNING '%', NEW;
+ result := deleteSearchName(NEW.partition, NEW.place_id);
+ DELETE FROM place_addressline 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, indexed_status = 2
+ where linked_place_id = NEW.place_id;
+ -- update not necessary for osmline, cause linked_place_id does not exist
- IF NEW.class = 'place' AND NEW.type = 'postcodearea' THEN
- -- Silently do nothing
+ IF NEW.linked_place_id is not null THEN
+ --DEBUG: RAISE WARNING 'place already linked to %', NEW.linked_place_id;
RETURN NEW;
END IF;
- -- 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;
- 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;
+ --DEBUG: RAISE WARNING 'Copy over address tags';
+ IF NEW.address is not NULL THEN
+ IF NEW.address ? 'conscriptionnumber' THEN
+ i := getorcreate_housenumber_id(make_standard_name(NEW.address->'conscriptionnumber'));
+ IF NEW.address ? 'streetnumber' THEN
+ i := getorcreate_housenumber_id(make_standard_name(NEW.address->'streetnumber'));
+ NEW.housenumber := (NEW.address->'conscriptionnumber') || '/' || (NEW.address->'streetnumber');
+ ELSE
+ NEW.housenumber := NEW.address->'conscriptionnumber';
+ END IF;
+ ELSEIF NEW.address ? 'streetnumber' THEN
+ NEW.housenumber := NEW.address->'streetnumber';
+ i := getorcreate_housenumber_id(make_standard_name(NEW.address->'streetnumber'));
+ ELSEIF NEW.address ? 'housenumber' THEN
+ NEW.housenumber := NEW.address->'housenumber';
+ i := getorcreate_housenumber_id(make_standard_name(NEW.housenumber));
+ END IF;
- IF NEW.linked_place_id is not null THEN
- RETURN NEW;
- END IF;
+ addr_street := NEW.address->'street';
+ addr_place := NEW.address->'place';
- -- Speed up searches - just use the centroid of the feature
- -- cheaper but less acurate
- place_centroid := ST_PointOnSurface(NEW.geometry);
- NEW.centroid := null;
-
- -- recalculate 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;
+ IF NEW.address ? 'postcode' and NEW.address->'postcode' not similar to '%(,|;)%' THEN
+ i := getorcreate_postcode_id(NEW.address->'postcode');
END IF;
- NEW.partition := get_partition(NEW.calculated_country_code);
- END IF;
- NEW.geometry_sector := geometry_sector(NEW.partition, place_centroid);
-
- -- interpolations
- IF NEW.class = 'place' AND NEW.type = 'houses'THEN
- IF NEW.osm_type = 'W' and ST_GeometryType(NEW.geometry) = 'ST_LineString' THEN
- NEW.parent_place_id := get_interpolation_parent(NEW.osm_id, NEW.street, NEW.addr_place,
- NEW.partition, place_centroid, NEW.geometry);
- i := create_interpolation(NEW.osm_id, NEW.housenumber, NEW.parent_place_id,
- NEW.partition, NEW.calculated_country_code,
- NEW.geometry_sector, NEW.postcode, NEW.geometry);
- END IF;
- RETURN NEW;
- END IF;
+ END IF;
- -- 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_members IN select members from planet_osm_rels r where r.id = NEW.osm_id and r.parts != array[]::bigint[]
- LOOP
- FOR i IN 1..array_upper(relation_members, 1) BY 2 LOOP
- IF relation_members[i+1] in ('', 'main_stream', 'side_stream') AND substring(relation_members[i],1,1) = 'w' THEN
- --DEBUG: RAISE WARNING 'waterway parent %, child %/%', NEW.osm_id, i, relation.members[i];
- FOR linked_node_id IN SELECT place_id FROM placex
- WHERE osm_type = 'W' and osm_id = substring(relation_members[i],2,200)::bigint
- and class = NEW.class and type = NEW.type
- and ( relation_members[i+1] != 'side_stream' or NEW.name->'name' = name->'name')
- LOOP
- UPDATE placex SET linked_place_id = NEW.place_id WHERE place_id = linked_node_id;
- END LOOP;
- END IF;
- END LOOP;
- END LOOP;
+ -- Speed up searches - just use the centroid of the feature
+ -- cheaper but less acurate
+ place_centroid := ST_PointOnSurface(NEW.geometry);
+ NEW.centroid := null;
+ NEW.postcode := null;
+ --DEBUG: RAISE WARNING 'Computing preliminary centroid at %',ST_AsText(place_centroid);
+
+ -- recalculate country and partition
+ IF NEW.rank_search = 4 AND NEW.address is not NULL AND NEW.address ? 'country' THEN
+ -- for countries, believe the mapped country code,
+ -- so that we remain in the right partition if the boundaries
+ -- suddenly expand.
+ NEW.country_code := lower(NEW.address->'country');
+ NEW.partition := get_partition(lower(NEW.country_code));
+ IF NEW.partition = 0 THEN
+ NEW.country_code := lower(get_country_code(place_centroid));
+ NEW.partition := get_partition(NEW.country_code);
END IF;
+ ELSE
+ IF NEW.rank_search >= 4 THEN
+ NEW.country_code := lower(get_country_code(place_centroid));
+ ELSE
+ NEW.country_code := NULL;
+ END IF;
+ NEW.partition := get_partition(NEW.country_code);
+ END IF;
+ --DEBUG: RAISE WARNING 'Country updated: "%"', NEW.country_code;
+
+ -- waterway ways are linked when they are part of a relation and have the same class/type
+ IF NEW.osm_type = 'R' and NEW.class = 'waterway' THEN
+ FOR relation_members IN select members from planet_osm_rels r where r.id = NEW.osm_id and r.parts != array[]::bigint[]
+ LOOP
+ FOR i IN 1..array_upper(relation_members, 1) BY 2 LOOP
+ IF relation_members[i+1] in ('', 'main_stream', 'side_stream') AND substring(relation_members[i],1,1) = 'w' THEN
+ --DEBUG: RAISE WARNING 'waterway parent %, child %/%', NEW.osm_id, i, relation_members[i];
+ FOR linked_node_id IN SELECT place_id FROM placex
+ WHERE osm_type = 'W' and osm_id = substring(relation_members[i],2,200)::bigint
+ and class = NEW.class and type in ('river', 'stream', 'canal', 'drain', 'ditch')
+ and ( relation_members[i+1] != 'side_stream' or NEW.name->'name' = name->'name')
+ LOOP
+ UPDATE placex SET linked_place_id = NEW.place_id WHERE place_id = linked_node_id;
+ END LOOP;
+ END IF;
+ END LOOP;
+ END LOOP;
+ --DEBUG: RAISE WARNING 'Waterway processed';
+ 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);
+ -- Adding ourselves to the list simplifies address calculations later
+ INSERT INTO place_addressline (place_id, address_place_id, fromarea, isaddress, distance, cached_rank_address)
+ VALUES (NEW.place_id, NEW.place_id, true, true, 0, NEW.rank_address);
- -- What level are we searching from
- search_maxrank := NEW.rank_search;
+ -- What level are we searching from
+ search_maxrank := NEW.rank_search;
- -- Thought this wasn't needed but when we add new languages to the country_name table
- -- we need to update the existing names
- IF NEW.name is not null AND array_upper(akeys(NEW.name),1) > 1 THEN
- default_language := get_country_language_code(NEW.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;
+ -- Thought this wasn't needed but when we add new languages to the country_name table
+ -- we need to update the existing names
+ IF NEW.name is not null AND array_upper(akeys(NEW.name),1) > 1 THEN
+ default_language := get_country_language_code(NEW.country_code);
+ IF default_language IS NOT NULL THEN
+ IF NEW.name ? 'name' AND NOT NEW.name ? ('name:'||default_language) THEN
+ NEW.name := NEW.name || 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;
+ --DEBUG: RAISE WARNING 'Local names updated';
- -- Initialise the name vector using our name
- name_vector := make_keywords(NEW.name);
- nameaddress_vector := '{}'::int[];
-
- FOR i IN 1..28 LOOP
- address_havelevel[i] := false;
- END LOOP;
+ -- Initialise the name vector using our name
+ name_vector := make_keywords(NEW.name);
+ nameaddress_vector := '{}'::int[];
- 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;
+ FOR i IN 1..28 LOOP
+ address_havelevel[i] := false;
+ END LOOP;
---RAISE WARNING 'before low level% %', NEW.place_id, NEW.rank_search;
+ NEW.importance := null;
+ select language||':'||title,importance from get_wikipedia_match(NEW.extratags, NEW.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;
- -- 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
+--DEBUG: RAISE WARNING 'Importance computed from wikipedia: %', NEW.importance;
---RAISE WARNING 'finding street for %', NEW;
+ -- ---------------------------------------------------------------------------
+ -- 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
- -- We won't get a better centroid, besides these places are too small to care
- NEW.centroid := place_centroid;
+ --DEBUG: RAISE WARNING 'finding street for % %', NEW.osm_type, NEW.osm_id;
- NEW.parent_place_id := null;
+ -- We won't get a better centroid, besides these places are too small to care
+ NEW.centroid := place_centroid;
- -- if we have a POI and there is no address information,
- -- see if we can get it from a surrounding building
- IF NEW.osm_type = 'N' AND 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 (housenumber is not null or street is not null or addr_place is not null)
- and rank_search > 28 AND ST_GeometryType(geometry) in ('ST_Polygon','ST_MultiPolygon')
- limit 1
- LOOP
- NEW.housenumber := location.housenumber;
- NEW.street := location.street;
- NEW.addr_place := location.addr_place;
- END LOOP;
- END IF;
+ NEW.parent_place_id := null;
- -- We have to find our parent road.
- -- Copy data from linked items (points on ways, addr:street links, relations)
+ -- if we have a POI and there is no address information,
+ -- see if we can get it from a surrounding building
+ IF NEW.osm_type = 'N' AND addr_street IS NULL AND addr_place IS NULL
+ AND NEW.housenumber IS NULL THEN
+ FOR location IN select address from placex where ST_Covers(geometry, place_centroid)
+ and address is not null
+ and (address ? 'housenumber' or address ? 'street' or address ? 'place')
+ and rank_search > 28 AND ST_GeometryType(geometry) in ('ST_Polygon','ST_MultiPolygon')
+ limit 1
+ LOOP
+ NEW.housenumber := location.address->'housenumber';
+ addr_street := location.address->'street';
+ addr_place := location.address->'place';
+ --DEBUG: RAISE WARNING 'Found surrounding building % %', location.osm_type, location.osm_id;
+ END LOOP;
+ END IF;
- -- Is this object part of a relation?
- FOR relation IN select * from planet_osm_rels where parts @> ARRAY[NEW.osm_id] and members @> ARRAY[lower(NEW.osm_type)||NEW.osm_id]
- LOOP
- -- At the moment we only process one type of relation - associatedStreet
- IF relation.tags @> ARRAY['associatedStreet'] 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
+ -- We have to find our parent road.
+ -- Copy data from linked items (points on ways, addr:street links, relations)
+
+ -- Is this object part of a relation?
+ FOR relation IN select * from planet_osm_rels where parts @> ARRAY[NEW.osm_id] and members @> ARRAY[lower(NEW.osm_type)||NEW.osm_id]
+ LOOP
+ -- At the moment we only process one type of relation - associatedStreet
+ IF relation.tags @> ARRAY['associatedStreet'] 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)::bigint
- and rank_search = 26 and name is not null INTO NEW.parent_place_id;
- END IF;
- END LOOP;
+ SELECT place_id from placex where osm_type = 'W'
+ and osm_id = substring(relation.members[i],2,200)::bigint
+ and rank_search = 26 and name is not null INTO NEW.parent_place_id;
END IF;
END LOOP;
+ END IF;
+ END LOOP;
+ --DEBUG: RAISE WARNING 'Checked for street relation (%)', NEW.parent_place_id;
-
- -- Note that addr:street links can only be indexed once the street itself is indexed
- IF NEW.parent_place_id IS NULL AND NEW.street IS NOT NULL THEN
- 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;
+ -- Note that addr:street links can only be indexed once the street itself is indexed
+ IF NEW.parent_place_id IS NULL AND addr_street IS NOT NULL THEN
+ address_street_word_ids := get_name_ids(make_standard_name(addr_street));
+ IF address_street_word_ids IS NOT NULL THEN
+ SELECT place_id from getNearestNamedRoadFeature(NEW.partition, place_centroid, address_street_word_ids) INTO NEW.parent_place_id;
END IF;
+ END IF;
+ --DEBUG: RAISE WARNING 'Checked for addr:street (%)', NEW.parent_place_id;
- 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;
+ IF NEW.parent_place_id IS NULL AND addr_place IS NOT NULL THEN
+ address_street_word_ids := get_name_ids(make_standard_name(addr_place));
+ IF address_street_word_ids IS NOT NULL THEN
+ SELECT place_id from getNearestNamedPlaceFeature(NEW.partition, place_centroid, address_street_word_ids) INTO NEW.parent_place_id;
END IF;
+ END IF;
+ --DEBUG: RAISE WARNING 'Checked for addr:place (%)', NEW.parent_place_id;
- IF NEW.parent_place_id IS NULL AND NEW.osm_type = 'N' THEN
+ -- Is this node part of an interpolation?
+ IF NEW.parent_place_id IS NULL AND NEW.osm_type = 'N' THEN
+ SELECT q.parent_place_id FROM location_property_osmline q, planet_osm_ways x
+ WHERE q.linegeo && NEW.geometry and x.id = q.osm_id and NEW.osm_id = any(x.nodes)
+ LIMIT 1 INTO NEW.parent_place_id;
+ END IF;
+ --DEBUG: RAISE WARNING 'Checked for interpolation (%)', NEW.parent_place_id;
---RAISE WARNING 'x1';
- -- Is this node part of a way?
- FOR location IN select p.* from placex p, planet_osm_ways w
- where p.osm_type = 'W' and p.rank_search >= 26
- and p.geometry && NEW.geometry and p.osm_id = w.id and NEW.osm_id = any(w.nodes)
- LOOP
---RAISE WARNING '%', location;
- -- Way IS a road then we are on it - that must be our road
- IF location.rank_search = 26 AND NEW.parent_place_id IS NULL THEN
---RAISE WARNING 'node in way that is a street %',location;
- NEW.parent_place_id := location.place_id;
- END IF;
+ -- Is this node part of a way?
+ IF NEW.parent_place_id IS NULL AND NEW.osm_type = 'N' THEN
- -- 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 location.class = 'place' and location.type='houses' THEN
- NEW.parent_place_id := location.parent_place_id;
- END IF;
+ FOR location IN
+ SELECT p.place_id, p.osm_id, p.rank_search, p.address from placex p, planet_osm_ways w
+ WHERE p.osm_type = 'W' and p.rank_search >= 26 and p.geometry && NEW.geometry and w.id = p.osm_id and NEW.osm_id = any(w.nodes)
+ LOOP
+ --DEBUG: RAISE WARNING 'Node is part of way % ', location.osm_id;
- -- Is the WAY part of a relation
- IF NEW.parent_place_id IS NULL THEN
- FOR relation IN select * from planet_osm_rels where parts @> ARRAY[location.osm_id] and members @> ARRAY['w'||location.osm_id]
- LOOP
- -- At the moment we only process one type of relation - associatedStreet
- IF relation.tags @> ARRAY['associatedStreet'] AND array_upper(relation.members, 1) IS NOT NULL THEN
- FOR i IN 1..array_upper(relation.members, 1) BY 2 LOOP
- IF NEW.parent_place_id IS NULL AND relation.members[i+1] = 'street' THEN
- --RAISE WARNING 'node in way that is in a relation %',relation;
- SELECT place_id from placex where osm_type='W' and osm_id = substring(relation.members[i],2,200)::bigint
- and rank_search = 26 and name is not null INTO NEW.parent_place_id;
- END IF;
- END LOOP;
- END IF;
- END LOOP;
- END IF;
+ -- Way IS a road then we are on it - that must be our road
+ IF location.rank_search < 28 THEN
+--RAISE WARNING 'node in way that is a street %',location;
+ NEW.parent_place_id := location.place_id;
+ EXIT;
+ END IF;
+ --DEBUG: RAISE WARNING 'Checked if way is street (%)', NEW.parent_place_id;
- -- If the way mentions a street or place address, try that for parenting.
- IF NEW.parent_place_id IS NULL AND location.street IS NOT NULL THEN
- address_street_word_ids := get_name_ids(make_standard_name(location.street));
+ -- If the way mentions a street or place address, try that for parenting.
+ IF location.address is not null THEN
+ IF location.address ? 'street' THEN
+ address_street_word_ids := get_name_ids(make_standard_name(location.address->'street'));
IF address_street_word_ids IS NOT NULL THEN
- FOR linkedplacex IN SELECT place_id from getNearestNamedRoadFeature(NEW.partition, place_centroid, address_street_word_ids) LOOP
- NEW.parent_place_id := linkedplacex.place_id;
- END LOOP;
+ SELECT place_id from getNearestNamedRoadFeature(NEW.partition, place_centroid, address_street_word_ids) INTO NEW.parent_place_id;
+ EXIT WHEN NEW.parent_place_id is not NULL;
END IF;
END IF;
+ --DEBUG: RAISE WARNING 'Checked for addr:street in way (%)', NEW.parent_place_id;
- IF NEW.parent_place_id IS NULL AND location.addr_place IS NOT NULL THEN
- address_street_word_ids := get_name_ids(make_standard_name(location.addr_place));
+ IF location.address ? 'place' THEN
+ address_street_word_ids := get_name_ids(make_standard_name(location.address->'place'));
IF address_street_word_ids IS NOT NULL THEN
- FOR linkedplacex IN SELECT place_id from getNearestNamedPlaceFeature(NEW.partition, place_centroid, address_street_word_ids) LOOP
- NEW.parent_place_id := linkedplacex.place_id;
- END LOOP;
+ SELECT place_id from getNearestNamedPlaceFeature(NEW.partition, place_centroid, address_street_word_ids) INTO NEW.parent_place_id;
+ EXIT WHEN NEW.parent_place_id is not NULL;
END IF;
END IF;
+ --DEBUG: RAISE WARNING 'Checked for addr:place in way (%)', NEW.parent_place_id;
+ END IF;
+ -- Is the WAY part of a relation
+ FOR relation IN select * from planet_osm_rels where parts @> ARRAY[location.osm_id] and members @> ARRAY['w'||location.osm_id]
+ LOOP
+ -- At the moment we only process one type of relation - associatedStreet
+ IF relation.tags @> ARRAY['associatedStreet'] AND array_upper(relation.members, 1) IS NOT NULL THEN
+ FOR i IN 1..array_upper(relation.members, 1) BY 2 LOOP
+ IF NEW.parent_place_id IS NULL AND relation.members[i+1] = 'street' THEN
+--RAISE WARNING 'node in way that is in a relation %',relation;
+ SELECT place_id from placex where osm_type='W' and osm_id = substring(relation.members[i],2,200)::bigint
+ and rank_search = 26 and name is not null INTO NEW.parent_place_id;
+ END IF;
+ END LOOP;
+ END IF;
END LOOP;
+ EXIT WHEN NEW.parent_place_id is not null;
+ --DEBUG: RAISE WARNING 'Checked for street relation in way (%)', NEW.parent_place_id;
- END IF;
+ END LOOP;
+ END IF;
---RAISE WARNING 'x4 %',NEW.parent_place_id;
- -- Still nothing, just use the nearest road
- IF NEW.parent_place_id IS NULL THEN
- FOR location IN SELECT place_id FROM getNearestRoadFeature(NEW.partition, place_centroid) LOOP
- NEW.parent_place_id := location.place_id;
- END LOOP;
- END IF;
+ -- Still nothing, just use the nearest road
+ IF NEW.parent_place_id IS NULL THEN
+ SELECT place_id FROM getNearestRoadFeature(NEW.partition, place_centroid) INTO NEW.parent_place_id;
+ END IF;
+ --DEBUG: RAISE WARNING 'Checked for nearest way (%)', NEW.parent_place_id;
---return NEW;
---RAISE WARNING 'x6 %',NEW.parent_place_id;
- -- If we didn't find any road fallback to standard method
- IF NEW.parent_place_id IS NOT NULL THEN
+ -- If we didn't find any road fallback to standard method
+ IF NEW.parent_place_id IS NOT NULL THEN
- -- Get the details of the parent road
- select * from search_name where place_id = NEW.parent_place_id INTO location;
- NEW.calculated_country_code := location.country_code;
+ -- Get the details of the parent road
+ select s.country_code, s.name_vector, s.nameaddress_vector from search_name s
+ where s.place_id = NEW.parent_place_id INTO location;
+ NEW.country_code := location.country_code;
+ --DEBUG: RAISE WARNING 'Got parent details from search name';
- -- Merge the postcode into the parent's address if necessary XXXX
- IF NEW.postcode IS NOT NULL THEN
- isin_tokens := '{}'::int[];
- address_street_word_id := getorcreate_word_id(make_standard_name(NEW.postcode));
- IF address_street_word_id is not null
- and not ARRAY[address_street_word_id] <@ location.nameaddress_vector THEN
- isin_tokens := isin_tokens || address_street_word_id;
- END IF;
- address_street_word_id := getorcreate_name_id(make_standard_name(NEW.postcode));
- IF address_street_word_id is not null
- and not ARRAY[address_street_word_id] <@ location.nameaddress_vector THEN
- isin_tokens := isin_tokens || address_street_word_id;
+ -- determine postcode
+ IF NEW.rank_search > 4 THEN
+ IF NEW.address is not null AND NEW.address ? 'postcode' THEN
+ NEW.postcode = upper(trim(NEW.address->'postcode'));
+ ELSE
+ SELECT postcode FROM placex WHERE place_id = NEW.parent_place_id INTO NEW.postcode;
END IF;
- IF isin_tokens != '{}'::int[] THEN
- UPDATE search_name
- SET nameaddress_vector = search_name.nameaddress_vector || isin_tokens
- WHERE place_id = NEW.parent_place_id;
+ IF NEW.postcode is null THEN
+ NEW.postcode := get_nearest_postcode(NEW.country_code, place_centroid);
END IF;
- END IF;
-
---RAISE WARNING '%', NEW.name;
- -- If there is no name it isn't searchable, don't bother to create a search record
- IF NEW.name is NULL THEN
- return NEW;
- END IF;
-
- -- Merge address from parent
- nameaddress_vector := array_merge(nameaddress_vector, location.nameaddress_vector);
- nameaddress_vector := array_merge(nameaddress_vector, location.name_vector);
+ END IF;
- -- 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 there is no name it isn't searchable, don't bother to create a search record
+ IF NEW.name is NULL THEN
+ --DEBUG: RAISE WARNING 'Not a searchable place % %', NEW.osm_type, NEW.osm_id;
+ return NEW;
+ END IF;
- IF NEW.rank_search <= 25 and NEW.rank_address > 0 THEN
- result := add_location(NEW.place_id, NEW.calculated_country_code, NEW.partition, name_vector, NEW.rank_search, NEW.rank_address, NEW.geometry);
- END IF;
+ -- Merge address from parent
+ nameaddress_vector := array_merge(nameaddress_vector, location.nameaddress_vector);
+ nameaddress_vector := array_merge(nameaddress_vector, location.name_vector);
- 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);
+ -- 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
- return NEW;
+ IF NEW.rank_search <= 25 and NEW.rank_address > 0 THEN
+ result := add_location(NEW.place_id, NEW.country_code, NEW.partition, name_vector, NEW.rank_search, NEW.rank_address, upper(trim(NEW.address->'postcode')), NEW.geometry);
+ --DEBUG: RAISE WARNING 'Place added to location table';
END IF;
+ result := insertSearchName(NEW.partition, NEW.place_id, NEW.country_code, name_vector, nameaddress_vector, NEW.rank_search, NEW.rank_address, NEW.importance, place_centroid, NEW.geometry);
+ --DEBUG: RAISE WARNING 'Place added to search table';
+
+ return NEW;
END IF;
--- RAISE WARNING ' INDEXING Started:';
--- RAISE WARNING ' INDEXING: %',NEW;
+ END IF;
- IF NEW.osm_type = 'R' AND NEW.rank_search < 26 THEN
+ -- ---------------------------------------------------------------------------
+ -- Full indexing
+ --DEBUG: RAISE WARNING 'Using full index mode for % %', NEW.osm_type, NEW.osm_id;
- -- see if we have any special relation members
- select members from planet_osm_rels where id = NEW.osm_id INTO relation_members;
+ IF NEW.osm_type = 'R' AND NEW.rank_search < 26 THEN
--- 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
+ -- see if we have any special relation members
+ select members from planet_osm_rels where id = NEW.osm_id INTO relation_members;
+ --DEBUG: RAISE WARNING 'Got relation members';
- 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 relation_members IS NOT NULL THEN
+ FOR relMember IN select get_osm_rel_members(relation_members,ARRAY['label']) as member LOOP
+ --DEBUG: RAISE WARNING 'Found label member %', relMember.member;
- -- 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;
+ 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
+ and class = 'place' order by rank_search desc limit 1 LOOP
- -- 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;
+ -- 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 extra tags
- NEW.extratags := hstore(linkedPlacex.class, linkedPlacex.type) || coalesce(linkedPlacex.extratags, ''::hstore) || coalesce(NEW.extratags, ''::hstore);
+ -- 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;
- -- mark the linked place (excludes from search results)
- UPDATE placex set linked_place_id = NEW.place_id where place_id = linkedPlacex.place_id;
+ -- merge in extra tags
+ NEW.extratags := hstore(linkedPlacex.class, linkedPlacex.type) || coalesce(linkedPlacex.extratags, ''::hstore) || coalesce(NEW.extratags, ''::hstore);
- -- 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;
+ -- 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;
+ select language||':'||title,importance from get_wikipedia_match(linkedPlacex.extratags, NEW.country_code) INTO linked_wikipedia,linked_importance;
+ --DEBUG: RAISE WARNING 'Linked label member';
END LOOP;
- IF NEW.centroid IS NULL THEN
+ END LOOP;
- FOR relMember IN select get_osm_rel_members(relation_members,ARRAY['admin_center','admin_centre']) as member LOOP
+ IF NEW.centroid IS NULL THEN
- 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 relMember IN select get_osm_rel_members(relation_members,ARRAY['admin_center','admin_centre']) as member LOOP
+ --DEBUG: RAISE WARNING 'Found admin_center member %', relMember.member;
- -- For 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
+ 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
+ and class = 'place' order by rank_search desc limit 1 LOOP
- -- If we don't already have one use this as the centre point of the geometry
- IF NEW.centroid IS NULL THEN
- NEW.centroid := coalesce(linkedPlacex.centroid,st_centroid(linkedPlacex.geometry));
- END IF;
+ -- 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
- -- 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;
+ -- 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 extra tags
- NEW.extratags := hstore(linkedPlacex.class, linkedPlacex.type) || coalesce(linkedPlacex.extratags, ''::hstore) || coalesce(NEW.extratags, ''::hstore);
+ -- 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;
- -- mark the linked place (excludes from search results)
- UPDATE placex set linked_place_id = NEW.place_id where place_id = linkedPlacex.place_id;
+ -- merge in extra tags
+ NEW.extratags := hstore(linkedPlacex.class, linkedPlacex.type) || coalesce(linkedPlacex.extratags, ''::hstore) || coalesce(NEW.extratags, ''::hstore);
- -- 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;
+ -- mark the linked place (excludes from search results)
+ UPDATE placex set linked_place_id = NEW.place_id where place_id = linkedPlacex.place_id;
- END LOOP;
+ -- keep a note of the node id in case we need it for wikipedia in a bit
+ linked_node_id := linkedPlacex.osm_id;
+ select language||':'||title,importance from get_wikipedia_match(linkedPlacex.extratags, NEW.country_code) INTO linked_wikipedia,linked_importance;
+ --DEBUG: RAISE WARNING 'Linked admin_center';
+ END IF;
END LOOP;
- END IF;
- END IF;
+ END LOOP;
+ 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 AND NEW.rank_address > 0 THEN
+ END IF;
- -- not found one yet? how about doing a name search
- IF NEW.centroid IS NULL AND (NEW.name->'name') is not null and make_standard_name(NEW.name->'name') != '' THEN
+ -- Name searches can be done for ways as well as relations
+ IF NEW.osm_type in ('W','R') AND NEW.rank_search < 26 AND NEW.rank_address > 0 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
+ -- 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
- -- 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;
+ --DEBUG: RAISE WARNING 'Looking for nodes with matching names';
+ FOR linkedPlacex IN select placex.* from placex WHERE
+ make_standard_name(name->'name') = make_standard_name(NEW.name->'name')
+ AND placex.rank_address = NEW.rank_address
+ 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
+ --DEBUG: RAISE WARNING 'Found matching place node %', linkedPlacex.osm_id;
+ -- If we don't already have one use this as the centre point of the geometry
+ IF NEW.centroid IS NULL THEN
+ NEW.centroid := coalesce(linkedPlacex.centroid,st_centroid(linkedPlacex.geometry));
+ END IF;
- -- merge in the name, re-init word vector
- NEW.name := linkedPlacex.name || NEW.name;
- name_vector := make_keywords(NEW.name);
+ -- 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);
+ -- 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;
+ -- 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;
+ -- keep a note of the node id in case we need it for wikipedia in a bit
+ linked_node_id := linkedPlacex.osm_id;
+ select language||':'||title,importance from get_wikipedia_match(linkedPlacex.extratags, NEW.country_code) INTO linked_wikipedia,linked_importance;
+ --DEBUG: RAISE WARNING 'Linked named place';
+ END LOOP;
+ END IF;
- 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;
+ 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.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;
-
- -- 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));
+ --DEBUG: RAISE WARNING 'Names updated from linked places';
END IF;
- NEW.parent_place_id = 0;
- parent_place_id_rank = 0;
-
- -- convert isin to array of tokenids
- isin_tokens := '{}'::int[];
- IF NEW.isin IS NOT NULL THEN
- isin := regexp_split_to_array(NEW.isin, E'[;,]');
- IF array_upper(isin, 1) IS NOT NULL THEN
- FOR i IN 1..array_upper(isin, 1) LOOP
- address_street_word_id := get_name_id(make_standard_name(isin[i]));
- IF address_street_word_id IS NOT NULL AND NOT(ARRAY[address_street_word_id] <@ isin_tokens) THEN
- nameaddress_vector := array_merge(nameaddress_vector, ARRAY[address_street_word_id]);
- isin_tokens := isin_tokens || address_street_word_id;
- END IF;
-
- -- merge 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;
+ -- Use the maximum importance if a one could be computed from the linked object.
+ IF linked_importance is not null AND
+ (NEW.importance is null or NEW.importance < linked_importance) THEN
+ NEW.importance = linked_importance;
END IF;
- IF NEW.postcode IS NOT NULL THEN
- isin := regexp_split_to_array(NEW.postcode, E'[;,]');
- IF array_upper(isin, 1) IS NOT NULL THEN
- FOR i IN 1..array_upper(isin, 1) LOOP
- address_street_word_id := get_name_id(make_standard_name(isin[i]));
- IF address_street_word_id IS NOT NULL AND NOT(ARRAY[address_street_word_id] <@ isin_tokens) THEN
- nameaddress_vector := array_merge(nameaddress_vector, ARRAY[address_street_word_id]);
- isin_tokens := isin_tokens || address_street_word_id;
- END IF;
- -- merge into address vector
- address_street_word_id := get_word_id(make_standard_name(isin[i]));
- IF address_street_word_id IS NOT NULL THEN
- nameaddress_vector := array_merge(nameaddress_vector, ARRAY[address_street_word_id]);
- END IF;
- END LOOP;
- END IF;
+ -- Still null? how about looking it up by the node id
+ IF NEW.importance IS NULL THEN
+ --DEBUG: RAISE WARNING 'Looking up importance by linked node id';
+ select language||':'||title,importance from wikipedia_article where osm_type = 'N'::char(1) and osm_id = linked_node_id order by importance desc limit 1 INTO NEW.wikipedia,NEW.importance;
END IF;
- -- 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;
+ END IF;
- -- 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;
+ -- 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 AND NEW.osm_type = 'R' THEN
+ perform create_country(NEW.name, lower(NEW.country_code));
+ --DEBUG: RAISE WARNING 'Country names updated';
+ END IF;
--- RAISE WARNING 'ISIN: %', isin_tokens;
+ NEW.parent_place_id = 0;
+ parent_place_id_rank = 0;
- -- Process area matches
- location_rank_search := 0;
- location_distance := 0;
- 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_address != location_rank_search THEN
- location_rank_search := location.rank_address;
- IF location.isguess THEN
- location_distance := location.distance * 1.5;
- ELSE
- IF location.rank_address <= 12 THEN
- -- for county and above, if we have an area consider that exact
- -- (It would be nice to relax the constraint for places close to
- -- the boundary but we'd need the exact geometry for that. Too
- -- expensive.)
- location_distance = 0;
- ELSE
- -- Below county level remain slightly fuzzy.
- location_distance := location.distance * 0.5;
- END IF;
- END IF;
- ELSE
- CONTINUE WHEN location.keywords <@ location_keywords;
- END IF;
-
- IF location.distance < location_distance OR NOT location.isguess THEN
- location_keywords := location.keywords;
- 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);
+ -- convert address store to array of tokenids
+ --DEBUG: RAISE WARNING 'Starting address search';
+ isin_tokens := '{}'::int[];
+ IF NEW.address IS NOT NULL THEN
+ FOR addr_item IN SELECT * FROM each(NEW.address)
+ LOOP
+ IF addr_item.key IN ('city', 'tiger:county', 'state', 'suburb', 'province', 'district', 'region', 'county', 'municipality', 'hamlet', 'village', 'subdistrict', 'town', 'neighbourhood', 'quarter', 'parish') THEN
+ address_street_word_id := get_name_id(make_standard_name(addr_item.value));
+ IF address_street_word_id IS NOT NULL AND NOT(ARRAY[address_street_word_id] <@ isin_tokens) THEN
+ isin_tokens := isin_tokens || address_street_word_id;
END IF;
-
- -- RAISE WARNING '% isaddress: %', location.place_id, location_isaddress;
- -- Add it to the list of search terms
- IF location.rank_search > 4 THEN
- nameaddress_vector := array_merge(nameaddress_vector, location.keywords::integer[]);
+ address_street_word_id := get_word_id(make_standard_name(addr_item.value));
+ IF address_street_word_id IS NOT NULL THEN
+ nameaddress_vector := array_merge(nameaddress_vector, ARRAY[address_street_word_id]);
END IF;
- INSERT INTO place_addressline VALUES (NEW.place_id, location.place_id, true, location_isaddress, location.distance, location.rank_address);
-
- 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;
+ END IF;
+ IF addr_item.key = 'is_in' THEN
+ -- is_in items need splitting
+ isin := regexp_split_to_array(addr_item.value, E'[;,]');
+ IF array_upper(isin, 1) IS NOT NULL THEN
+ FOR i IN 1..array_upper(isin, 1) LOOP
+ address_street_word_id := get_name_id(make_standard_name(isin[i]));
+ IF address_street_word_id IS NOT NULL AND NOT(ARRAY[address_street_word_id] <@ isin_tokens) THEN
+ isin_tokens := isin_tokens || address_street_word_id;
+ END IF;
+ -- merge 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;
-
---RAISE WARNING ' Terms: (%) %',location, nameaddress_vector;
-
END IF;
-
END LOOP;
+ END IF;
+ nameaddress_vector := array_merge(nameaddress_vector, isin_tokens);
- -- try using the isin value to find parent places
- IF array_upper(isin_tokens, 1) IS NOT NULL THEN
- FOR i IN 1..array_upper(isin_tokens, 1) LOOP
---RAISE WARNING ' getNearestNamedFeature: % % % %',NEW.partition, place_centroid, search_maxrank, isin_tokens[i];
- IF NOT ARRAY[isin_tokens[i]] <@ nameaddress_vector THEN
-
- FOR location IN SELECT * from getNearestNamedFeature(NEW.partition, place_centroid, search_maxrank, isin_tokens[i]) LOOP
-
- --RAISE WARNING ' ISIN: %',location;
-
- IF location.rank_search > 4 THEN
- nameaddress_vector := array_merge(nameaddress_vector, location.keywords::integer[]);
- INSERT INTO place_addressline VALUES (NEW.place_id, location.place_id, false, NOT address_havelevel[location.rank_address], location.distance, location.rank_address);
- address_havelevel[location.rank_address] := true;
-
- IF location.rank_address > parent_place_id_rank THEN
- NEW.parent_place_id = location.place_id;
- parent_place_id_rank = location.rank_address;
- END IF;
- END IF;
- END LOOP;
+-- RAISE WARNING 'ISIN: %', isin_tokens;
+ -- Process area matches
+ location_rank_search := 0;
+ location_distance := 0;
+ location_parent := NULL;
+ -- added ourself as address already
+ address_havelevel[NEW.rank_address] := true;
+ --DEBUG: RAISE WARNING ' getNearFeatures(%,''%'',%,''%'')',NEW.partition, place_centroid, search_maxrank, isin_tokens;
+ FOR location IN
+ SELECT * from getNearFeatures(NEW.partition,
+ CASE WHEN NEW.rank_search >= 26
+ AND NEW.rank_search < 30
+ THEN NEW.geometry
+ ELSE place_centroid END,
+ search_maxrank, isin_tokens)
+ LOOP
+ IF location.rank_address != location_rank_search THEN
+ location_rank_search := location.rank_address;
+ IF location.isguess THEN
+ location_distance := location.distance * 1.5;
+ ELSE
+ IF location.rank_address <= 12 THEN
+ -- for county and above, if we have an area consider that exact
+ -- (It would be nice to relax the constraint for places close to
+ -- the boundary but we'd need the exact geometry for that. Too
+ -- expensive.)
+ location_distance = 0;
+ ELSE
+ -- Below county level remain slightly fuzzy.
+ location_distance := location.distance * 0.5;
END IF;
-
- END LOOP;
+ END IF;
+ ELSE
+ CONTINUE WHEN location.keywords <@ location_keywords;
END IF;
- -- for long ways we should add search terms for the entire length
- IF st_length(NEW.geometry) > 0.05 THEN
-
- location_rank_search := 0;
- location_distance := 0;
+ IF location.distance < location_distance OR NOT location.isguess THEN
+ location_keywords := location.keywords;
- FOR location IN SELECT * from getNearFeatures(NEW.partition, NEW.geometry, search_maxrank, isin_tokens) LOOP
+ 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;
- IF location.rank_address != location_rank_search THEN
- location_rank_search := location.rank_address;
- location_distance := location.distance * 1.5;
+ -- RAISE WARNING '% isaddress: %', location.place_id, location_isaddress;
+ -- Add it to the list of search terms
+ IF location.rank_search > 4 THEN
+ nameaddress_vector := array_merge(nameaddress_vector, location.keywords::integer[]);
+ END IF;
+ INSERT INTO place_addressline (place_id, address_place_id, fromarea, isaddress, distance, cached_rank_address)
+ VALUES (NEW.place_id, location.place_id, true, location_isaddress, location.distance, location.rank_address);
+
+ IF location_isaddress THEN
+ -- add postcode if we have one
+ -- (If multiple postcodes are available, we end up with the highest ranking one.)
+ IF location.postcode is not null THEN
+ NEW.postcode = location.postcode;
END IF;
- IF location.rank_search > 4 AND location.distance < location_distance THEN
-
- -- Add it to the list of search terms
- nameaddress_vector := array_merge(nameaddress_vector, location.keywords::integer[]);
- INSERT INTO place_addressline VALUES (NEW.place_id, location.place_id, true, false, location.distance, location.rank_address);
+ 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;
- END LOOP;
+ END IF;
+
+ --DEBUG: RAISE WARNING ' Terms: (%) %',location, nameaddress_vector;
END IF;
- -- if we have a name add this to the name search table
- IF NEW.name IS NOT NULL THEN
+ END LOOP;
+ --DEBUG: RAISE WARNING 'address computed';
- IF NEW.rank_search <= 25 and NEW.rank_address > 0 THEN
- 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.address is not null AND NEW.address ? 'postcode'
+ AND NEW.address->'postcode' not similar to '%(,|;)%' THEN
+ NEW.postcode := upper(trim(NEW.address->'postcode'));
+ END IF;
- IF NEW.rank_search between 26 and 27 and NEW.class = 'highway' THEN
- result := insertLocationRoad(NEW.partition, NEW.place_id, NEW.calculated_country_code, NEW.geometry);
- END IF;
+ IF NEW.postcode is null AND NEW.rank_search > 8 THEN
+ NEW.postcode := get_nearest_postcode(NEW.country_code, NEW.geometry);
+ END IF;
- 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);
+ -- if we have a name add this to the name search table
+ IF NEW.name IS NOT NULL THEN
+ IF NEW.rank_search <= 25 and NEW.rank_address > 0 THEN
+ result := add_location(NEW.place_id, NEW.country_code, NEW.partition, name_vector, NEW.rank_search, NEW.rank_address, upper(trim(NEW.address->'postcode')), NEW.geometry);
+ --DEBUG: RAISE WARNING 'added to location (full)';
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;
+ IF NEW.rank_search between 26 and 27 and NEW.class = 'highway' THEN
+ result := insertLocationRoad(NEW.partition, NEW.place_id, NEW.country_code, NEW.geometry);
+ --DEBUG: RAISE WARNING 'insert into road location table (full)';
END IF;
+ result := insertSearchName(NEW.partition, NEW.place_id, NEW.country_code, name_vector, nameaddress_vector, NEW.rank_search, NEW.rank_address, NEW.importance, place_centroid, NEW.geometry);
+ --DEBUG: RAISE WARNING 'added to serach name (full)';
+
END IF;
+ -- 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;
+
+ --DEBUG: RAISE WARNING 'place update % % finsihed.', NEW.osm_type, NEW.osm_id;
+
RETURN NEW;
END;
$$
--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;
+ -- reparenting also for OSM Interpolation Lines (and for Tiger?)
+ update location_property_osmline set indexed_status = 2 where indexed_status = 0 and parent_place_id = OLD.place_id;
END IF;
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);
+ insert into import_polygon_delete (osm_type, osm_id, class, type) values (OLD.osm_type,OLD.osm_id,OLD.class,OLD.type);
RETURN NULL;
END IF;
END IF;
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;
-- interpolations are special
- IF OLD.class = 'place' and OLD.type = 'houses' THEN
- UPDATE placex set indexed_status = 100 where osm_type = OLD.osm_type and osm_id = OLD.osm_id and class = 'place' and type = 'address';
+ IF OLD.osm_type='W' and OLD.class = 'place' and OLD.type = 'houses' THEN
+ UPDATE location_property_osmline set indexed_status = 100 where osm_id = OLD.osm_id; -- osm_id = wayid (=old.osm_id)
END IF;
RETURN OLD;
i INTEGER;
existing RECORD;
existingplacex RECORD;
+ existingline RECORD;
existinggeometry GEOMETRY;
existingplace_id BIGINT;
result BOOLEAN;
--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
- 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;
- --DEBUG: RAISE WARNING '%', existingplacex;
- END IF;
-
+ -- filter wrong tupels
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
- INSERT INTO import_polygon_error values (NEW.osm_type, NEW.osm_id, NEW.class, NEW.type, NEW.name, NEW.country_code,
- now(), ST_IsValidReason(NEW.geometry), null, NEW.geometry);
+ INSERT INTO import_polygon_error (osm_type, osm_id, class, type, name, country_code, updated, errormessage, prevgeometry, newgeometry)
+ VALUES (NEW.osm_type, NEW.osm_id, NEW.class, NEW.type, NEW.name, NEW.address->'country', now(), ST_IsValidReason(NEW.geometry), null, NEW.geometry);
-- RAISE WARNING 'Invalid Geometry: % % % %',NEW.osm_type,NEW.osm_id,NEW.class,NEW.type;
RETURN null;
END IF;
- -- Patch in additional country names
- IF NEW.admin_level = 2 AND NEW.type = 'administrative' AND NEW.country_code is not null THEN
- 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?
- select * from place where osm_type = NEW.osm_type and osm_id = NEW.osm_id and class = NEW.class and type = NEW.type INTO existing;
+ -- decide, whether it is an osm interpolation line => insert intoosmline, or else just placex
+ IF NEW.class='place' and NEW.type='houses' and NEW.osm_type='W' and ST_GeometryType(NEW.geometry) = 'ST_LineString' THEN
+ -- Have we already done this place?
+ select * from place where osm_type = NEW.osm_type and osm_id = NEW.osm_id and class = NEW.class and type = NEW.type INTO existing;
- -- Get the existing place_id
- 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;
+ -- Get the existing place_id
+ select * from location_property_osmline where osm_id = NEW.osm_id INTO existingline;
- -- Handle a place changing type by removing the old data
- -- My generated 'place' types are causing havok because they overlap with real keys
- -- TODO: move them to their own special purpose key/class to avoid collisions
- IF existing.osm_type IS NULL THEN
- DELETE FROM place where osm_type = NEW.osm_type and osm_id = NEW.osm_id and class = NEW.class;
- END IF;
+ -- Handle a place changing type by removing the old data (this trigger is executed BEFORE INSERT of the NEW tupel)
+ IF existing.osm_type IS NULL THEN
+ DELETE FROM place where osm_type = NEW.osm_type and osm_id = NEW.osm_id and class = NEW.class;
+ END IF;
- --DEBUG: RAISE WARNING 'Existing: %',existing.osm_id;
- --DEBUG: RAISE WARNING 'Existing PlaceX: %',existingplacex.place_id;
+ 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;
- -- Log and discard
- IF existing.geometry is not null AND st_isvalid(existing.geometry)
- AND st_area(existing.geometry) > 0.02
- AND ST_GeometryType(NEW.geometry) in ('ST_Polygon','ST_MultiPolygon')
- AND st_area(NEW.geometry) < st_area(existing.geometry)*0.5
- THEN
- INSERT INTO import_polygon_error values (NEW.osm_type, NEW.osm_id, NEW.class, NEW.type, NEW.name, NEW.country_code, now(),
- 'Area reduced from '||st_area(existing.geometry)||' to '||st_area(NEW.geometry), existing.geometry, NEW.geometry);
- RETURN null;
- END IF;
+ -- update method for interpolation lines: delete all old interpolation lines with same osm_id (update on place) and insert the new one(s) (they can be split up, if they have > 2 nodes)
+ IF existingline.osm_id IS NOT NULL THEN
+ delete from location_property_osmline where osm_id = NEW.osm_id;
+ END IF;
- 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
- 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
- -- 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, addr_place, isin, postcode, country_code, extratags, geometry)
- values (NEW.osm_type
- ,NEW.osm_id
- ,NEW.class
- ,NEW.type
- ,NEW.name
- ,NEW.admin_level
- ,NEW.housenumber
- ,NEW.street
- ,NEW.addr_place
- ,NEW.isin
- ,NEW.postcode
- ,NEW.country_code
- ,NEW.extratags
- ,NEW.geometry
- );
-
- --DEBUG: RAISE WARNING 'insert done % % % % %',NEW.osm_type,NEW.osm_id,NEW.class,NEW.type,NEW.name;
+ -- for interpolations invalidate all nodes on the line
+ update placex p set indexed_status = 2
+ from planet_osm_ways w
+ where w.id = NEW.osm_id and p.osm_type = 'N' and p.osm_id = any(w.nodes);
- RETURN NEW;
- END IF;
- -- Various ways to do the update
+ INSERT INTO location_property_osmline (osm_id, address, linegeo)
+ VALUES (NEW.osm_id, NEW.address, NEW.geometry);
- -- Debug, what's changed?
- IF FALSE THEN
- IF coalesce(existing.name::text, '') != coalesce(NEW.name::text, '') THEN
- RAISE WARNING 'update details, name: % % % %',NEW.osm_type,NEW.osm_id,existing.name::text,NEW.name::text;
- END IF;
- IF coalesce(existing.housenumber, '') != coalesce(NEW.housenumber, '') THEN
- RAISE WARNING 'update details, housenumber: % % % %',NEW.osm_type,NEW.osm_id,existing.housenumber,NEW.housenumber;
- END IF;
- IF coalesce(existing.street, '') != coalesce(NEW.street, '') THEN
- RAISE WARNING 'update details, street: % % % %',NEW.osm_type,NEW.osm_id,existing.street,NEW.street;
+
+ IF existing.osm_type IS NULL THEN
+ return NEW;
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;
+
+ IF coalesce(existing.address, ''::hstore) != coalesce(NEW.address, ''::hstore)
+ OR (coalesce(existing.extratags, ''::hstore) != coalesce(NEW.extratags, ''::hstore))
+ OR existing.geometry::text != NEW.geometry::text
+ THEN
+
+ update place set
+ name = NEW.name,
+ address = NEW.address,
+ 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;
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;
+
+ RETURN NULL;
+
+ ELSE -- insert to placex
+
+ -- Patch in additional country names
+ IF NEW.admin_level = 2 AND NEW.type = 'administrative'
+ AND NEW.address is not NULL AND NEW.address ? 'country' THEN
+ SELECT name FROM country_name WHERE country_code = lower(NEW.address->'country') INTO existing;
+ IF existing.name IS NOT NULL THEN
+ NEW.name = existing.name || NEW.name;
+ END IF;
END IF;
- IF coalesce(existing.postcode, '') != coalesce(NEW.postcode, '') THEN
- RAISE WARNING 'update details, postcode: % % % %',NEW.osm_type,NEW.osm_id,existing.postcode,NEW.postcode;
+
+ -- Have we already done this place?
+ select * from place where osm_type = NEW.osm_type and osm_id = NEW.osm_id and class = NEW.class and type = NEW.type INTO existing;
+
+ -- Get the existing place_id
+ 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;
+
+ -- Handle a place changing type by removing the old data
+ -- My generated 'place' types are causing havok because they overlap with real keys
+ -- TODO: move them to their own special purpose key/class to avoid collisions
+ IF existing.osm_type IS NULL THEN
+ DELETE FROM place where osm_type = NEW.osm_type and osm_id = NEW.osm_id and class = NEW.class;
END IF;
- IF coalesce(existing.country_code, '') != coalesce(NEW.country_code, '') THEN
- RAISE WARNING 'update details, country_code: % % % %',NEW.osm_type,NEW.osm_id,existing.country_code,NEW.country_code;
+
+ --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)
+ AND st_area(existing.geometry) > 0.02
+ AND ST_GeometryType(NEW.geometry) in ('ST_Polygon','ST_MultiPolygon')
+ AND st_area(NEW.geometry) < st_area(existing.geometry)*0.5
+ THEN
+ INSERT INTO import_polygon_error (osm_type, osm_id, class, type, name, country_code, updated, errormessage, prevgeometry, newgeometry)
+ VALUES (NEW.osm_type, NEW.osm_id, NEW.class, NEW.type, NEW.name, NEW.address->'country', now(),
+ 'Area reduced from '||st_area(existing.geometry)||' to '||st_area(NEW.geometry), existing.geometry, NEW.geometry);
+ RETURN null;
END IF;
- END IF;
- -- Special case for polygon shape changes because they tend to be large and we can be a bit clever about how we handle them
- IF existing.geometry::text != NEW.geometry::text
- AND ST_GeometryType(existing.geometry) in ('ST_Polygon','ST_MultiPolygon')
- AND ST_GeometryType(NEW.geometry) in ('ST_Polygon','ST_MultiPolygon')
- THEN
+ 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;
- -- Get the version of the geometry actually used (in placex table)
- select geometry from placex where osm_type = NEW.osm_type and osm_id = NEW.osm_id and class = NEW.class and type = NEW.type into existinggeometry;
+ -- To paraphrase, if there isn't an existing item, OR if the admin level has changed
+ IF existingplacex.osm_type IS NULL OR
+ (existingplacex.class = 'boundary' AND
+ ((coalesce(existingplacex.admin_level, 15) != coalesce(NEW.admin_level, 15) AND existingplacex.type = 'administrative') OR
+ (existingplacex.type != NEW.type)))
+ THEN
- -- Performance limit
- IF st_area(NEW.geometry) < 0.000000001 AND st_area(existinggeometry) < 1 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;
- -- 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_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);
+ IF existing.osm_type IS NOT NULL THEN
+ -- 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;
- update placex set indexed_status = 2 where indexed_status = 0 and
- (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);
+ -- 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, address, extratags, geometry)
+ values (NEW.osm_type, NEW.osm_id, NEW.class, NEW.type, NEW.name,
+ NEW.admin_level, NEW.address, NEW.extratags, NEW.geometry);
+ --DEBUG: RAISE WARNING 'insert done % % % % %',NEW.osm_type,NEW.osm_id,NEW.class,NEW.type,NEW.name;
+
+ RETURN NEW;
END IF;
- END IF;
+ -- Special case for polygon shape changes because they tend to be large and we can be a bit clever about how we handle them
+ IF existing.geometry::text != NEW.geometry::text
+ AND ST_GeometryType(existing.geometry) in ('ST_Polygon','ST_MultiPolygon')
+ AND ST_GeometryType(NEW.geometry) in ('ST_Polygon','ST_MultiPolygon')
+ THEN
+ -- Get the version of the geometry actually used (in placex table)
+ select geometry from placex where osm_type = NEW.osm_type and osm_id = NEW.osm_id and class = NEW.class and type = NEW.type into existinggeometry;
- 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
+ -- Performance limit
+ IF st_area(NEW.geometry) < 0.000000001 AND st_area(existinggeometry) < 1 THEN
- update place 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,
- 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;
+ -- 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_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);
- 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;
+ update placex set indexed_status = 2 where indexed_status = 0 and
+ (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;
- 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;
- END IF;
+ IF coalesce(existing.name::text, '') != coalesce(NEW.name::text, '')
+ OR coalesce(existing.extratags::text, '') != coalesce(NEW.extratags::text, '')
+ OR coalesce(existing.address, ''::hstore) != coalesce(NEW.address, ''::hstore)
+ OR coalesce(existing.admin_level, 15) != coalesce(NEW.admin_level, 15)
+ OR existing.geometry::text != NEW.geometry::text
+ THEN
- -- for interpolations invalidate all nodes on the line
- IF NEW.class = 'place' and NEW.type = 'houses' and NEW.osm_type = 'W' THEN
- update placex p set indexed_status = 2 from planet_osm_ways w where w.id = NEW.osm_id and p.osm_type = 'N' and p.osm_id = any(w.nodes);
- END IF;
+ update place set
+ name = NEW.name,
+ address = NEW.address,
+ 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;
- -- Abort the add (we modified the existing place instead)
- RETURN NULL;
-END;
-$$ LANGUAGE plpgsql;
+ IF NEW.class in ('place','boundary') AND NEW.type in ('postcode','postal_code') THEN
+ IF NEW.address is NULL OR NOT NEW.address ? 'postcode' THEN
+ -- postcode was deleted, no longer retain in placex
+ DELETE FROM placex where place_id = existingplacex.place_id;
+ RETURN NULL;
+ END IF;
-CREATE OR REPLACE FUNCTION get_name_by_language(name hstore, languagepref TEXT[]) RETURNS TEXT
- AS $$
-DECLARE
- search TEXT[];
- found BOOLEAN;
-BEGIN
+ NEW.name := hstore('ref', NEW.address->'postcode');
+ END IF;
- IF name is null THEN
- RETURN null;
- END IF;
+ IF NEW.class in ('boundary')
+ AND ST_GeometryType(NEW.geometry) not in ('ST_Polygon','ST_MultiPolygon') THEN
+ DELETE FROM placex where place_id = existingplacex.place_id;
+ RETURN NULL;
+ END IF;
- search := languagepref;
+ update placex set
+ name = NEW.name,
+ address = NEW.address,
+ parent_place_id = null,
+ extratags = NEW.extratags,
+ admin_level = NEW.admin_level,
+ indexed_status = 2,
+ geometry = NEW.geometry
+ where place_id = existingplacex.place_id;
+
+ -- if a node(=>house), which is part of a interpolation line, changes (e.g. the street attribute) => mark this line for reparenting
+ -- (already here, because interpolation lines are reindexed before nodes, so in the second call it would be too late)
+ IF NEW.osm_type='N' and NEW.class='place' and NEW.type='house' THEN
+ -- Is this node part of an interpolation line? search for it in location_property_osmline and mark the interpolation line for reparenting
+ update location_property_osmline p set indexed_status = 2 from planet_osm_ways w where p.linegeo && NEW.geometry and p.osm_id = w.id and NEW.osm_id = any(w.nodes);
+ END IF;
+
+ -- linked places should get potential new naming and addresses
+ IF existingplacex.linked_place_id is not NULL THEN
+ update placex x set
+ name = p.name,
+ extratags = p.extratags,
+ indexed_status = 2
+ from place p
+ where x.place_id = existingplacex.linked_place_id
+ and x.indexed_status = 0
+ and x.osm_type = p.osm_type
+ and x.osm_id = p.osm_id
+ and x.class = p.class;
+ END IF;
- FOR j IN 1..array_upper(search, 1) LOOP
- IF name ? search[j] AND trim(name->search[j]) != '' THEN
- return trim(name->search[j]);
END IF;
- END LOOP;
- -- anything will do as a fallback - just take the first name type thing there is
- search := avals(name);
- RETURN search[1];
+ -- Abort the add (we modified the existing place instead)
+ RETURN NULL;
+ END IF;
+
END;
-$$
-LANGUAGE plpgsql IMMUTABLE;
+$$ LANGUAGE plpgsql;
-CREATE OR REPLACE FUNCTION get_address_postcode(for_place_id BIGINT) RETURNS TEXT
+CREATE OR REPLACE FUNCTION get_name_by_language(name hstore, languagepref TEXT[]) RETURNS TEXT
AS $$
DECLARE
- result TEXT[];
- search TEXT[];
- for_postcode TEXT;
- found INTEGER;
- location RECORD;
+ result TEXT;
BEGIN
+ IF name is null THEN
+ RETURN null;
+ END IF;
- found := 1000;
- search := ARRAY['ref'];
- result := '{}';
-
- select postcode from placex where place_id = for_place_id limit 1 into for_postcode;
-
- FOR location IN
- select rank_address,name,distance,length(name::text) as namelength
- from place_addressline join placex on (address_place_id = placex.place_id)
- where place_addressline.place_id = for_place_id and rank_address in (5,11)
- order by rank_address desc,rank_search desc,fromarea desc,distance asc,namelength desc
- LOOP
- IF array_upper(search, 1) IS NOT NULL AND array_upper(location.name, 1) IS NOT NULL THEN
- FOR j IN 1..array_upper(search, 1) LOOP
- FOR k IN 1..array_upper(location.name, 1) LOOP
- IF (found > location.rank_address AND location.name[k].key = search[j] AND location.name[k].value != '') AND NOT result @> ARRAY[trim(location.name[k].value)] AND (for_postcode IS NULL OR location.name[k].value ilike for_postcode||'%') THEN
- result[(100 - location.rank_address)] := trim(location.name[k].value);
- found := location.rank_address;
- END IF;
- END LOOP;
- END LOOP;
+ FOR j IN 1..array_upper(languagepref,1) LOOP
+ IF name ? languagepref[j] THEN
+ result := trim(name->languagepref[j]);
+ IF result != '' THEN
+ return result;
+ END IF;
END IF;
END LOOP;
- RETURN array_to_string(result,', ');
+ -- anything will do as a fallback - just take the first name type thing there is
+ RETURN trim((avals(name))[1]);
END;
$$
-LANGUAGE plpgsql;
+LANGUAGE plpgsql IMMUTABLE;
-CREATE OR REPLACE FUNCTION get_address_by_language(for_place_id BIGINT, languagepref TEXT[]) RETURNS TEXT
+--housenumber only needed for tiger data
+CREATE OR REPLACE FUNCTION get_address_by_language(for_place_id BIGINT, housenumber INTEGER, languagepref TEXT[]) RETURNS TEXT
AS $$
DECLARE
result TEXT[];
result := '{}';
prevresult := '';
- FOR location IN select * from get_addressdata(for_place_id) where isaddress order by rank_address desc LOOP
+ FOR location IN select * from get_addressdata(for_place_id, housenumber) where isaddress order by rank_address desc LOOP
currresult := trim(get_name_by_language(location.name, languagepref));
IF currresult != prevresult AND currresult IS NOT NULL AND result[(100 - location.rank_address)] IS NULL THEN
result[(100 - location.rank_address)] := trim(get_name_by_language(location.name, languagepref));
distance FLOAT
);
-CREATE OR REPLACE FUNCTION get_addressdata(in_place_id BIGINT) RETURNS setof addressline
+CREATE OR REPLACE FUNCTION get_addressdata(in_place_id BIGINT, in_housenumber INTEGER) RETURNS setof addressline
AS $$
DECLARE
for_place_id BIGINT;
countryname HSTORE;
hadcountry BOOLEAN;
BEGIN
+ -- first query osmline (interpolation lines)
+ select parent_place_id, country_code, 30, postcode, null, 'place', 'house' from location_property_osmline
+ WHERE place_id = in_place_id AND in_housenumber>=startnumber AND in_housenumber <= endnumber
+ INTO for_place_id,searchcountrycode, searchrankaddress, searchpostcode, searchhousename, searchclass, searchtype;
+ IF for_place_id IS NOT NULL THEN
+ searchhousenumber = in_housenumber::text;
+ END IF;
- select parent_place_id,'us', housenumber, 30, postcode, null, 'place', 'house' from location_property_tiger
- WHERE place_id = in_place_id
- INTO for_place_id,searchcountrycode, searchhousenumber, searchrankaddress, searchpostcode, searchhousename, searchclass, searchtype;
+ --then query tiger data
+ -- %NOTIGERDATA% IF 0 THEN
+ IF for_place_id IS NULL THEN
+ select parent_place_id,'us', 30, postcode, null, 'place', 'house' from location_property_tiger
+ WHERE place_id = in_place_id AND in_housenumber>=startnumber AND in_housenumber <= endnumber
+ INTO for_place_id,searchcountrycode, searchrankaddress, searchpostcode, searchhousename, searchclass, searchtype;
+ IF for_place_id IS NOT NULL THEN
+ searchhousenumber = in_housenumber::text;
+ END IF;
+ END IF;
+ -- %NOTIGERDATA% END IF;
+ -- %NOAUXDATA% IF 0 THEN
IF for_place_id IS NULL THEN
select parent_place_id,'us', housenumber, 30, postcode, null, 'place', 'house' from location_property_aux
- WHERE place_id = in_place_id
+ WHERE place_id = in_place_id
INTO for_place_id,searchcountrycode, searchhousenumber, searchrankaddress, searchpostcode, searchhousename, searchclass, searchtype;
END IF;
+ -- %NOAUXDATA% END IF;
+ -- postcode table
IF for_place_id IS NULL THEN
- select parent_place_id, calculated_country_code, housenumber, rank_search, postcode, name, class, type from placex
- WHERE place_id = in_place_id and rank_address = 30
+ select parent_place_id, country_code, rank_address, postcode, 'place', 'postcode'
+ FROM location_postcode
+ WHERE place_id = in_place_id
+ INTO for_place_id, searchcountrycode, searchrankaddress, searchpostcode, searchclass, searchtype;
+ END IF;
+
+ IF for_place_id IS NULL THEN
+ select parent_place_id, country_code, housenumber, rank_search, postcode, name, class, type from placex
+ WHERE place_id = in_place_id and rank_search > 27
INTO for_place_id, searchcountrycode, searchhousenumber, searchrankaddress, searchpostcode, searchhousename, searchclass, searchtype;
END IF;
IF for_place_id IS NULL THEN
- select coalesce(linked_place_id, place_id), calculated_country_code,
+ select coalesce(linked_place_id, place_id), country_code,
housenumber, rank_search, postcode, null
from placex where place_id = in_place_id
INTO for_place_id, searchcountrycode, searchhousenumber, searchrankaddress, searchpostcode, searchhousename;
found := 1000;
hadcountry := false;
FOR location IN
- select placex.place_id, osm_type, osm_id,
- CASE WHEN class = 'place' and type = 'postcode' THEN hstore('name', postcode) ELSE name END as name,
- class, type, admin_level, true as fromarea, true as isaddress,
+ select placex.place_id, osm_type, osm_id, name,
+ class, type, admin_level, true as isaddress,
CASE WHEN rank_address = 0 THEN 100 WHEN rank_address = 11 THEN 5 ELSE rank_address END as rank_address,
- 0 as distance, calculated_country_code, postcode
+ 0 as distance, country_code, postcode
from placex
where place_id = for_place_id
LOOP
--RAISE WARNING '%',location;
- IF searchcountrycode IS NULL AND location.calculated_country_code IS NOT NULL THEN
- searchcountrycode := location.calculated_country_code;
+ IF searchcountrycode IS NULL AND location.country_code IS NOT NULL THEN
+ searchcountrycode := location.country_code;
END IF;
- IF searchpostcode IS NOT NULL and location.type = 'postcode' THEN
+ IF location.type in ('postcode', 'postal_code') THEN
location.isaddress := FALSE;
- END IF;
- IF searchpostcode IS NULL and location.postcode IS NOT NULL THEN
- searchpostcode := location.postcode;
- END IF;
- IF location.rank_address = 4 AND location.isaddress THEN
+ ELSEIF location.rank_address = 4 THEN
hadcountry := true;
END IF;
IF location.rank_address < 4 AND NOT hadcountry THEN
END IF;
END IF;
countrylocation := ROW(location.place_id, location.osm_type, location.osm_id, location.name, location.class,
- location.type, location.admin_level, location.fromarea, location.isaddress, location.rank_address,
+ location.type, location.admin_level, true, location.isaddress, location.rank_address,
location.distance)::addressline;
RETURN NEXT countrylocation;
found := location.rank_address;
END LOOP;
FOR location IN
- select placex.place_id, osm_type, osm_id,
- CASE WHEN class = 'place' and type = 'postcode' THEN hstore('name', postcode) ELSE name END as name,
+ select placex.place_id, osm_type, osm_id, name,
CASE WHEN extratags ? 'place' THEN 'place' ELSE class END as class,
CASE WHEN extratags ? 'place' THEN extratags->'place' ELSE type END as type,
admin_level, fromarea, isaddress,
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,calculated_country_code,postcode
+ distance,country_code,postcode
from place_addressline join placex on (address_place_id = placex.place_id)
where place_addressline.place_id = for_place_id
and (cached_rank_address > 0 AND cached_rank_address < searchrankaddress)
- and address_place_id != for_place_id
- and (placex.calculated_country_code IS NULL OR searchcountrycode IS NULL OR placex.calculated_country_code = searchcountrycode)
+ and address_place_id != for_place_id and linked_place_id is null
+ and (placex.country_code IS NULL OR searchcountrycode IS NULL OR placex.country_code = searchcountrycode)
order by rank_address desc,isaddress desc,fromarea desc,distance asc,rank_search desc
LOOP
--RAISE WARNING '%',location;
- IF searchcountrycode IS NULL AND location.calculated_country_code IS NOT NULL THEN
- searchcountrycode := location.calculated_country_code;
+ IF searchcountrycode IS NULL AND location.country_code IS NOT NULL THEN
+ searchcountrycode := location.country_code;
END IF;
- IF searchpostcode IS NOT NULL and location.type = 'postcode' THEN
+ IF location.type in ('postcode', 'postal_code') THEN
location.isaddress := FALSE;
END IF;
- IF searchpostcode IS NULL and location.isaddress and location.type != 'postcode' and location.postcode IS NOT NULL THEN
- searchpostcode := location.postcode;
- END IF;
IF location.rank_address = 4 AND location.isaddress THEN
hadcountry := true;
END IF;
IF searchhousename IS NOT NULL THEN
location := ROW(in_place_id, null, null, searchhousename, searchclass, searchtype, null, true, true, 29, 0)::addressline;
--- location := ROW(in_place_id, null, null, searchhousename, 'place', 'house_name', null, true, true, 29, 0)::addressline;
RETURN NEXT location;
END IF;
LANGUAGE plpgsql;
-CREATE OR REPLACE FUNCTION update_place(search_place_id BIGINT) RETURNS BOOLEAN
- AS $$
-DECLARE
- numfeatures integer;
-BEGIN
- update placex set
- name = place.name,
- housenumber = place.housenumber,
- street = place.street,
- addr_place = place.addr_place,
- isin = place.isin,
- postcode = place.postcode,
- country_code = place.country_code,
- parent_place_id = null
- from place
- where placex.place_id = search_place_id
- and place.osm_type = placex.osm_type and place.osm_id = placex.osm_id
- and place.class = placex.class and place.type = placex.type;
- update placex set indexed_status = 2 where place_id = search_place_id;
- update placex set indexed_status = 0 where place_id = search_place_id;
- return true;
-END;
-$$
-LANGUAGE plpgsql;
-
CREATE OR REPLACE FUNCTION get_searchrank_label(rank INTEGER) RETURNS TEXT
AS $$
DECLARE
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 LOOP;
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;
END IF;
- IF out_postcode IS NULL THEN
- out_postcode := getNearestPostcode(out_partition, place_centroid);
- END IF;
+ -- XXX look into postcode table
newpoints := 0;
insert into location_property_aux (place_id, partition, parent_place_id, housenumber, postcode, centroid)
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));
+ AND rank_search > rank and indexed_status = 0 and ST_geometrytype(placex.geometry) = 'ST_Point' and (rank_search < 28 or name is not null or (rank >= 16 and address ? 'place'));
update placex set indexed_status = 2 where (st_covers(geom, placex.geometry) OR ST_Intersects(geom, placex.geometry))
- AND rank_search > rank and indexed_status = 0 and ST_geometrytype(placex.geometry) != 'ST_Point' and (rank_search < 28 or name is not null or (rank >= 16 and addr_place is not null));
+ AND rank_search > rank and indexed_status = 0 and ST_geometrytype(placex.geometry) != 'ST_Point' and (rank_search < 28 or name is not null or (rank >= 16 and address ? 'place'));
END LOOP;
ELSE
diameter := 0;
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);
+ 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 address ? 'place');
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);