Add a couple of overview clustering strategies

This commit is contained in:
Javier Goizueta 2016-11-03 13:31:04 +01:00
parent 5c52e7564f
commit 5992304b47

View File

@ -697,6 +697,356 @@ AS $$
END;
$$ LANGUAGE PLPGSQL;
CREATE OR REPLACE FUNCTION _CDB_GridCluster_Reduce_Strategy(reloid REGCLASS, ref_z INTEGER, overview_z INTEGER, grid_px FLOAT8 DEFAULT NULL)
RETURNS REGCLASS
AS $$
DECLARE
overview_rel TEXT;
reduction FLOAT8;
base_name TEXT;
pixel_m FLOAT8;
grid_m FLOAT8;
offset_m FLOAT8;
offset_x TEXT;
offset_y TEXT;
cell_x TEXT;
cell_y TEXT;
aggr_attributes TEXT;
attributes TEXT;
columns TEXT;
gtypes TEXT[];
schema_name TEXT;
table_name TEXT;
point_geom TEXT;
BEGIN
SELECT _CDB_GeometryTypes(reloid) INTO gtypes;
IF gtypes IS NULL OR array_upper(gtypes, 1) <> 1 OR gtypes[1] <> 'ST_Point' THEN
-- This strategy only supports datasets with point geomety
RETURN NULL;
END IF;
--TODO: check applicability: geometry type, minimum number of points...
overview_rel := _CDB_Overview_Name(reloid, ref_z, overview_z);
-- Grid size in pixels at Z level overview_z
IF grid_px IS NULL THEN
grid_px := 1.0;
END IF;
SELECT * FROM _cdb_split_table_name(reloid) INTO schema_name, table_name;
-- pixel_m: size of a pixel in webmercator units (meters)
SELECT CDB_XYZ_Resolution(overview_z) INTO pixel_m;
-- grid size in meters
grid_m = grid_px * pixel_m;
attributes := _CDB_Aggregable_Attributes_Expression(reloid);
aggr_attributes := _CDB_Aggregated_Attributes_Expression(reloid);
IF attributes <> '' THEN
attributes := ', ' || attributes;
END IF;
IF aggr_attributes <> '' THEN
aggr_attributes := aggr_attributes || ', ';
END IF;
-- Center of each cell:
cell_x := Format('gx*%1$s + %2$s', grid_m, grid_m/2);
cell_y := Format('gy*%1$s + %2$s', grid_m, grid_m/2);
-- Displacement to the nearest pixel center:
IF MOD(grid_px::numeric, 1.0::numeric) = 0 THEN
offset_m := pixel_m/2 - MOD((grid_m/2)::numeric, pixel_m::numeric)::float8;
offset_x := Format('%s', offset_m);
offset_y := Format('%s', offset_m);
ELSE
offset_x := Format('%2$s/2 - MOD((%1$s)::numeric, (%2$s)::numeric)::float8', cell_x, pixel_m);
offset_y := Format('%2$s/2 - MOD((%1$s)::numeric, (%2$s)::numeric)::float8', cell_y, pixel_m);
END IF;
point_geom := Format('ST_SetSRID(ST_MakePoint(%1$s + %3$s, %2$s + %4$s), 3857)', cell_x, cell_y, offset_x, offset_y);
-- compute the resulting columns in the same order as in the base table
WITH cols AS (
SELECT
CASE c
WHEN 'cartodb_id' THEN 'cartodb_id'
WHEN 'the_geom' THEN
Format('ST_Transform(%s, 4326) AS the_geom', point_geom)
WHEN 'the_geom_webmercator' THEN
Format('%s AS the_geom_webmercator', point_geom)
ELSE c
END AS column
FROM CDB_ColumnNames(reloid) c
)
SELECT string_agg(s.column, ',') FROM (
SELECT * FROM cols
) AS s INTO columns;
IF NOT columns LIKE '%_feature_count%' THEN
columns := columns || ', n AS _feature_count';
END IF;
EXECUTE Format('DROP TABLE IF EXISTS %I.%I CASCADE;', schema_name, overview_rel);
-- Now we cluster the data using a grid of size grid_m
-- and selecte the centroid (average coordinates) of each cluster.
-- If we had a selected numeric attribute of interest we could use it
-- as a weight for the average coordinates.
EXECUTE Format('
CREATE TABLE %7$I.%3$I AS
WITH clusters AS (
SELECT
%5$s
count(*) AS n,
Floor(ST_X(f.the_geom_webmercator)/%2$s)::int AS gx,
Floor(ST_Y(f.the_geom_webmercator)/%2$s)::int AS gy,
MIN(cartodb_id) AS cartodb_id
FROM %1$s f
GROUP BY gx, gy
)
SELECT %6$s FROM clusters
', reloid::text, grid_m, overview_rel, attributes, aggr_attributes, columns, schema_name);
RETURN Format('%I.%I', schema_name, overview_rel)::regclass;
END;
$$ LANGUAGE PLPGSQL;
-- This strategy places the aggregation of each cluster at the centroid of the cluster members.
CREATE OR REPLACE FUNCTION _CDB_GridClusterCentroid_Reduce_Strategy(reloid REGCLASS, ref_z INTEGER, overview_z INTEGER, grid_px FLOAT8 DEFAULT NULL)
RETURNS REGCLASS
AS $$
DECLARE
overview_rel TEXT;
reduction FLOAT8;
base_name TEXT;
pixel_m FLOAT8;
grid_m FLOAT8;
offset_m FLOAT8;
offset_x TEXT;
offset_y TEXT;
cell_x TEXT;
cell_y TEXT;
aggr_attributes TEXT;
attributes TEXT;
columns TEXT;
gtypes TEXT[];
schema_name TEXT;
table_name TEXT;
point_geom TEXT;
BEGIN
SELECT _CDB_GeometryTypes(reloid) INTO gtypes;
IF gtypes IS NULL OR array_upper(gtypes, 1) <> 1 OR gtypes[1] <> 'ST_Point' THEN
-- This strategy only supports datasets with point geomety
RETURN NULL;
END IF;
--TODO: check applicability: geometry type, minimum number of points...
overview_rel := _CDB_Overview_Name(reloid, ref_z, overview_z);
-- Grid size in pixels at Z level overview_z
IF grid_px IS NULL THEN
grid_px := 1.0;
END IF;
SELECT * FROM _cdb_split_table_name(reloid) INTO schema_name, table_name;
-- pixel_m: size of a pixel in webmercator units (meters)
SELECT CDB_XYZ_Resolution(overview_z) INTO pixel_m;
-- grid size in meters
grid_m = grid_px * pixel_m;
attributes := _CDB_Aggregable_Attributes_Expression(reloid);
aggr_attributes := _CDB_Aggregated_Attributes_Expression(reloid);
IF attributes <> '' THEN
attributes := ', ' || attributes;
END IF;
IF aggr_attributes <> '' THEN
aggr_attributes := aggr_attributes || ', ';
END IF;
-- Center of each cell:
cell_x := Format('gx*%1$s + %2$s', grid_m, grid_m/2);
cell_y := Format('gy*%1$s + %2$s', grid_m, grid_m/2);
-- Displacement to the nearest pixel center:
IF MOD(grid_px::numeric, 1.0::numeric) = 0 THEN
offset_m := pixel_m/2 - MOD((grid_m/2)::numeric, pixel_m::numeric)::float8;
offset_x := Format('%s', offset_m);
offset_y := Format('%s', offset_m);
ELSE
offset_x := Format('%2$s/2 - MOD((%1$s)::numeric, (%2$s)::numeric)::float8', cell_x, pixel_m);
offset_y := Format('%2$s/2 - MOD((%1$s)::numeric, (%2$s)::numeric)::float8', cell_y, pixel_m);
END IF;
point_geom := Format('ST_SetSRID(ST_MakePoint(%1$s + %3$s, %2$s + %4$s), 3857)', cell_x, cell_y, offset_x, offset_y);
-- compute the resulting columns in the same order as in the base table
WITH cols AS (
SELECT
CASE c
WHEN 'cartodb_id' THEN 'cartodb_id'
WHEN 'the_geom' THEN
'ST_Transform(ST_SetSRID(ST_MakePoint(_sum_of_x/n, _sum_of_y/n), 3857), 4326) AS the_geom'
WHEN 'the_geom_webmercator' THEN
'ST_SetSRID(ST_MakePoint(_sum_of_x/n, _sum_of_y/n), 3857) AS the_geom_webmercator'
ELSE c
END AS column
FROM CDB_ColumnNames(reloid) c
)
SELECT string_agg(s.column, ',') FROM (
SELECT * FROM cols
) AS s INTO columns;
IF NOT columns LIKE '%_feature_count%' THEN
columns := columns || ', n AS _feature_count';
END IF;
EXECUTE Format('DROP TABLE IF EXISTS %I.%I CASCADE;', schema_name, overview_rel);
-- Now we cluster the data using a grid of size grid_m
-- and selecte the centroid (average coordinates) of each cluster.
-- If we had a selected numeric attribute of interest we could use it
-- as a weight for the average coordinates.
EXECUTE Format('
CREATE TABLE %7$I.%3$I AS
WITH clusters AS (
SELECT
%5$s
count(*) AS n,
SUM(ST_X(f.the_geom_webmercator)) AS _sum_of_x,
SUM(ST_Y(f.the_geom_webmercator)) AS _sum_of_y,
Floor(ST_Y(f.the_geom_webmercator)/%2$s)::int AS gy,
Floor(ST_X(f.the_geom_webmercator)/%2$s)::int AS gx,
MIN(cartodb_id) AS cartodb_id
FROM %1$s f
GROUP BY gx, gy
)
SELECT %6$s FROM clusters
', reloid::text, grid_m, overview_rel, attributes, aggr_attributes, columns, schema_name);
RETURN Format('%I.%I', schema_name, overview_rel)::regclass;
END;
$$ LANGUAGE PLPGSQL;
-- This strategy places the aggregation of each cluster at the position of one of the cluster members.
-- TODO: the location and the cartodb_id should be chosen to correspond to the same record;
-- a more efficient alternative to the `first` aggregator should be used
CREATE OR REPLACE FUNCTION _CDB_GridClusterSample_Reduce_Strategy(reloid REGCLASS, ref_z INTEGER, overview_z INTEGER, grid_px FLOAT8 DEFAULT NULL)
RETURNS REGCLASS
AS $$
DECLARE
overview_rel TEXT;
reduction FLOAT8;
base_name TEXT;
pixel_m FLOAT8;
grid_m FLOAT8;
offset_m FLOAT8;
offset_x TEXT;
offset_y TEXT;
cell_x TEXT;
cell_y TEXT;
aggr_attributes TEXT;
attributes TEXT;
columns TEXT;
gtypes TEXT[];
schema_name TEXT;
table_name TEXT;
point_geom TEXT;
BEGIN
SELECT _CDB_GeometryTypes(reloid) INTO gtypes;
IF gtypes IS NULL OR array_upper(gtypes, 1) <> 1 OR gtypes[1] <> 'ST_Point' THEN
-- This strategy only supports datasets with point geomety
RETURN NULL;
END IF;
--TODO: check applicability: geometry type, minimum number of points...
overview_rel := _CDB_Overview_Name(reloid, ref_z, overview_z);
-- Grid size in pixels at Z level overview_z
IF grid_px IS NULL THEN
grid_px := 1.0;
END IF;
SELECT * FROM _cdb_split_table_name(reloid) INTO schema_name, table_name;
-- pixel_m: size of a pixel in webmercator units (meters)
SELECT CDB_XYZ_Resolution(overview_z) INTO pixel_m;
-- grid size in meters
grid_m = grid_px * pixel_m;
attributes := _CDB_Aggregable_Attributes_Expression(reloid);
aggr_attributes := _CDB_Aggregated_Attributes_Expression(reloid);
IF attributes <> '' THEN
attributes := ', ' || attributes;
END IF;
IF aggr_attributes <> '' THEN
aggr_attributes := aggr_attributes || ', ';
END IF;
-- Center of each cell:
cell_x := Format('gx*%1$s + %2$s', grid_m, grid_m/2);
cell_y := Format('gy*%1$s + %2$s', grid_m, grid_m/2);
-- Displacement to the nearest pixel center:
IF MOD(grid_px::numeric, 1.0::numeric) = 0 THEN
offset_m := pixel_m/2 - MOD((grid_m/2)::numeric, pixel_m::numeric)::float8;
offset_x := Format('%s', offset_m);
offset_y := Format('%s', offset_m);
ELSE
offset_x := Format('%2$s/2 - MOD((%1$s)::numeric, (%2$s)::numeric)::float8', cell_x, pixel_m);
offset_y := Format('%2$s/2 - MOD((%1$s)::numeric, (%2$s)::numeric)::float8', cell_y, pixel_m);
END IF;
point_geom := Format('ST_SetSRID(ST_MakePoint(%1$s + %3$s, %2$s + %4$s), 3857)', cell_x, cell_y, offset_x, offset_y);
-- compute the resulting columns in the same order as in the base table
WITH cols AS (
SELECT
CASE c
WHEN 'cartodb_id' THEN 'cartodb_id'
ELSE c
END AS column
FROM CDB_ColumnNames(reloid) c
)
SELECT string_agg(s.column, ',') FROM (
SELECT * FROM cols
) AS s INTO columns;
IF NOT columns LIKE '%_feature_count%' THEN
columns := columns || ', n AS _feature_count';
END IF;
EXECUTE Format('DROP TABLE IF EXISTS %I.%I CASCADE;', schema_name, overview_rel);
-- Now we cluster the data using a grid of size grid_m
-- and select the centroid (average coordinates) of each cluster.
-- If we had a selected numeric attribute of interest we could use it
-- as a weight for the average coordinates.
EXECUTE Format('
CREATE TABLE %7$I.%3$I AS
WITH clusters AS (
SELECT
%5$s
count(*) AS n,
first(the_geom) as the_geom,
first(the_geom_webmercator) as the_geom_webmercator,
Floor(ST_X(f.the_geom_webmercator)/%2$s)::int AS gx,
Floor(ST_Y(f.the_geom_webmercator)/%2$s)::int AS gy,
MIN(cartodb_id) AS cartodb_id
FROM %1$s f
GROUP BY gx, gy
)
SELECT %6$s FROM clusters
', reloid::text, grid_m, overview_rel, attributes, aggr_attributes, columns, schema_name);
RETURN Format('%I.%I', schema_name, overview_rel)::regclass;
END;
$$ LANGUAGE PLPGSQL;
-- Create overview tables for a dataset.
-- Scope: public
-- Parameters: