Question

I’ve a SDO_GEOMETRY column containing quite large multi polygons, defined like this:

INSERT INTO t1 (i, d, g)
VALUES (
  25,
  'Multipolygon - multi-touch',
  sdo_geometry (2007, null, null, sdo_elem_info_array (1,1003,1, 17,1003,1), 
  sdo_ordinate_array (50,95, 55,95, 53,96, 55,97, 53,98, 55,99, 50,99, 50,95, 55,100, 55,95, 60,95, 60,100, 55,100))
);

Instead of two polygons as in the example above, one column contains > 100 polygons.

I’d like to filter this column so it only will return a relevant subset (using a bbox?), something like:

  SELECT filter(Geometry, bbox) from Table Where Id = 1

Answer 1

A first attempt at brute force solution might look something like this:

CREATE OR REPLACE FUNCTION FILTER_MULTI_POLYGONS
(
    udtGeometry IN SDO_GEOMETRY,
    udtMask IN SDO_GEOMETRY,
    dTolerance IN NUMBER
)
RETURN SDO_GEOMETRY
AS
    iElements INTEGER;
    udtElement SDO_GEOMETRY;
    udtResult SDO_GEOMETRY := NULL;
    iCount INTEGER;
BEGIN
    IF udtGeometry IS NOT NULL THEN
        iElements := SDO_UTIL.GETNUMELEM(udtGeometry);
        FOR iElement IN 1..iElements
        LOOP
            udtElement := SDO_UTIL.EXTRACT(udtGeometry, iElement);
            IF SDO_GEOM.SDO_DISTANCE(udtElement, udtMask, dTolerance) <= dTolerance THEN
                IF udtResult IS NULL THEN
                    udtResult := udtElement;
                ELSE
                    udtResult := SDO_UTIL.APPEND(udtResult, udtElement);
                END IF;
            END IF;
        END LOOP;
    END IF;
    RETURN udtResult;
END;

I say brute force because:

1. The individual sub-polygons are not indexed, so this solution does not leverage spatial indexing. If performance is important enough, it may be worthwhile to break your multi-polygons into their constitutent sub-polygons (with one row in your source table per sub-polygon) so that you can use a different solution that takes advantage of spatial indexing. You could do this either up front (by changing the design of your source table) or behind the scenes (maybe using a materialized view based on your original table).

2. It looks like you are on Oracle XE, and so are limited to the Locator subset of the Oracle Spatial functionality.

3. Items 1 and 2 mean that your only built-in choice for determining which sub-polygons interact with the mask appears to be via SDO_GEOM.SDO_DISTANCE. This is going to be resource intensive (since it will be called for every sub-polygon), and gives you only one type of interaction vs. the many that are possible with the built-in spatial operators (that rely on spatial indexing).

4. SDO_UTIL.APPEND may not be the most performant (or correct, particularly if your multi-polygons are not disjoint) way of building up the “filtered” multi-polygon result, but it illustrates the concept.

Anyway, this is what I get running your sample geometry against some sample masks. You should make sure the function returns the expected results against your real geometries.

SQL> REM Example mask that overlaps first polygon only
SQL> SELECT 
  2      FILTER_MULTI_POLYGONS
  3      (
  4          T1.G, 
  5          SDO_GEOMETRY
  6          (
  7              2003, 
  8              NULL, 
  9              NULL, 
 10              SDO_ELEM_INFO_ARRAY(1, 1003, 3),
 11              SDO_ORDINATE_ARRAY(0, 0, 53, 96)
 12          ),
 13          0.1
 14      ) AS RESULT
 15  FROM T1;

RESULT(SDO_GTYPE, SDO_SRID, SDO_POINT(X, Y, Z), SDO_ELEM_INFO, SDO_ORDINATES)
--------------------------------------------------------------------------------
SDO_GEOMETRY(2003, NULL, NULL, SDO_ELEM_INFO_ARRAY(1, 1003, 1), SDO_ORDINATE_ARR
AY(50, 95, 55, 95, 53, 96, 55, 97, 53, 98, 55, 99, 50, 99, 50, 95))


SQL> REM Example mask that overlaps second polygon only
SQL> SELECT 
  2      FILTER_MULTI_POLYGONS
  3      (
  4          T1.G, 
  5          SDO_GEOMETRY
  6          (
  7              2003, 
  8              NULL, 
  9              NULL, 
 10              SDO_ELEM_INFO_ARRAY(1, 1003, 3),
 11              SDO_ORDINATE_ARRAY(56, 0, 60, 96)
 12          ),
 13          0.1
 14      ) AS RESULT
 15  FROM T1;

RESULT(SDO_GTYPE, SDO_SRID, SDO_POINT(X, Y, Z), SDO_ELEM_INFO, SDO_ORDINATES)
--------------------------------------------------------------------------------
SDO_GEOMETRY(2003, NULL, NULL, SDO_ELEM_INFO_ARRAY(1, 1003, 1), SDO_ORDINATE_ARR
AY(55, 100, 55, 95, 60, 95, 60, 100, 55, 100))


SQL> REM Example mask that overlaps both polygons
SQL> SELECT 
  2      FILTER_MULTI_POLYGONS
  3      (
  4          T1.G, 
  5          SDO_GEOMETRY
  6          (
  7              2003, 
  8              NULL, 
  9              NULL, 
 10              SDO_ELEM_INFO_ARRAY(1, 1003, 3),
 11              SDO_ORDINATE_ARRAY(0, 0, 100, 100)
 12          ),
 13          0.1
 14      ) AS RESULT
 15  FROM T1;

RESULT(SDO_GTYPE, SDO_SRID, SDO_POINT(X, Y, Z), SDO_ELEM_INFO, SDO_ORDINATES)
--------------------------------------------------------------------------------
SDO_GEOMETRY(2007, NULL, NULL, SDO_ELEM_INFO_ARRAY(1, 1003, 1, 17, 1003, 1), SDO
_ORDINATE_ARRAY(50, 95, 55, 95, 53, 96, 55, 97, 53, 98, 55, 99, 50, 99, 50, 95,
55, 100, 55, 95, 60, 95, 60, 100, 55, 100))


SQL> REM Example mask that overlaps neither polygon
SQL> SELECT 
  2      FILTER_MULTI_POLYGONS
  3      (
  4          T1.G, 
  5          SDO_GEOMETRY
  6          (
  7              2003, 
  8              NULL, 
  9              NULL, 
 10              SDO_ELEM_INFO_ARRAY(1, 1003, 3),
 11              SDO_ORDINATE_ARRAY(0, 0, 10, 10)
 12          ),
 13          0.1
 14      ) AS RESULT
 15  FROM T1;

RESULT(SDO_GTYPE, SDO_SRID, SDO_POINT(X, Y, Z), SDO_ELEM_INFO, SDO_ORDINATES)
--------------------------------------------------------------------------------

Hope this helps.