How to recursively loop through parent polygons intersections to get smallest (child) polygons with no overlaps?

Question

I'm struggling with a problem for a couple of days and realized many people also get stuck when the topic is intersections in PostGIS (v2.5). That's why I decided to ask for a more detailed and generic, common question.

I have the following table:

DROP TABLE IF EXISTS tbl_foo;
CREATE TABLE tbl_foo (
    id bigint NOT NULL,
    geom public.geometry(MultiPolygon, 4326),
    att_category character varying(15),
    att_value integer
);
INSERT INTO tbl_foo (id, geom, att_category, att_value) VALUES 
    (1, ST_SetSRID('MULTIPOLYGON (((0 6, 0 12, 8 9, 0 6)))'::geometry,4326) , 'cat1', 2 );
INSERT INTO tbl_foo (id, geom, att_category, att_value) VALUES 
    (2, ST_SetSRID('MULTIPOLYGON (((5 0, 5 12, 9 12, 9 0, 5 0)))'::geometry,4326), 'cat1', 1 );
INSERT INTO tbl_foo (id, geom, att_category, att_value) VALUES 
    (3, ST_SetSRID('MULTIPOLYGON (((4 4, 3 8, 4 12, 7 14,10 12, 11 8, 10 4, 4 4)))'::geometry,4326) , 'cat2', 5 );

It looks like this:

I want to get all the child polygons based on the intersection of the parent polygons. For the outcome, it'd be expected:

• The child polygons with no overlap between them.
• An column containing the sum of the value of their parent polygons,
• An column containing the count of parent polygons of one category
• An column containing the count of another category
• An column containing the category of the child polygon, based on the following rule: -If ALL the parent polygons are from one class, the child polygon also have this class. Else, the category of the child polygon is a third category.

So it'd looks like it:

So, in the end, the output table generated (for this example) will have 7 rows (all the 7, non-overlapping, child polygons), containing columns of category, sum_value, ct_overlap_cat1, ct_overlap_cat2

The following code I started, gives me the individual intersections, comparing one parent with another.

SELECT
(ST_Dump(
    ST_SymDifference(a.geom, b.geom) 
)).geom
FROM tbl_foo a, tbl_foo b
WHERE a.ID < b.ID AND ST_INTERSECTS(a.geom, b.geom)
UNION ALL
SELECT
ST_Intersection(a.geom, b.geom) as geom
FROM tbl_foo a, tbl_foo b
WHERE a.ID < b.ID AND ST_INTERSECTS(a.geom, b.geom);

How do I recursively loop through the result of this mentioned code, that, independent of the number of overlap polygons I always get its 'smallest' (child) polygons (Fig. 2)?

Answer 1

Try this:

Download the PostGIS Addons from this link: https://github.com/pedrogit/postgisaddons

Install by running the postgis_addons.sql file to get the ST_SplitAgg() function.

Test by running the postgis_addons_test.sql file.

Here is your query:

WITH  result_table AS (
    WITH  parts AS (
      SELECT a.att_value val,
             CASE WHEN a.att_category = 'cat1' THEN 1 ELSE 0 END cat1,
             CASE WHEN a.att_category = 'cat2' THEN 1 ELSE 0 END cat2,
             unnest(ST_SplitAgg(a.geom, b.geom, 0.00001)) geom
      FROM tbl_foo a,
           tbl_foo b
      WHERE ST_Equals(a.geom, b.geom) OR
            ST_Contains(a.geom, b.geom) OR
            ST_Contains(b.geom, a.geom) OR
            ST_Overlaps(a.geom, b.geom)
      GROUP BY a.id, a.att_category , ST_AsEWKB(a.geom), val
    )
    SELECT CASE WHEN sum(cat2) = 0 THEN 'cat1'
                WHEN sum(cat1) = 0 THEN 'cat2'
                ELSE 'cat3'
           END category, 
           sum(val*1.0) sum_value, 
           sum(cat1) ct_overlap_cat1, 
           sum(cat2) ct_overlap_cat2, 
           ST_Union(geom) geom
    FROM parts
    GROUP BY ST_Area(geom)
)
SELECT category, sum_value, ct_overlap_cat1, ct_overlap_cat2,
(ST_Dump(result_table.geom)).geom as geom
FROM result_table

Answer 2

I suppose if you use polygon geometry type instead of MultiPolygon everything will fall into place:

DROP TABLE IF EXISTS tbl_foo;
CREATE TABLE tbl_foo (
    id bigint NOT NULL,
    geom public.geometry(Polygon, 4326),
    att_category character varying(15),
    att_value integer
);

INSERT INTO tbl_foo (id, geom, att_category, att_value) VALUES 
    (1, ST_SetSRID('POLYGON ((0 6, 0 12, 8 9, 0 6))'::geometry,4326) , 'cat1', 2 );
INSERT INTO tbl_foo (id, geom, att_category, att_value) VALUES 
    (2, ST_SetSRID('POLYGON ((5 0, 5 12, 9 12, 9 0, 5 0))'::geometry,4326), 'cat1', 1 );
INSERT INTO tbl_foo (id, geom, att_category, att_value) VALUES 
    (3, ST_SetSRID('POLYGON ((4 4, 3 8, 4 12, 7 14,10 12, 11 8, 10 4, 4 4))'::geometry,4326) , 'cat2', 5 );

The result is 9 entries that correspond to the different intersection options in the example given by you.