C Complex Spatial Queries: Examples

C.1 Tables Used in the Examples

The examples in this appendix refer to tables named GEOD_CITIES, GEOD_COUNTIES, and GEOD_INTERSTATES, which are defined as follows:

CREATE TABLE GEOD_CITIES(
  LOCATION     SDO_GEOMETRY,
  CITY         VARCHAR2(42),
  STATE_ABRV   VARCHAR2(2),
  POP90        NUMBER,
  RANK90       NUMBER);

CREATE TABLE GEOD_COUNTIES(
  COUNTY_NAME   VARCHAR2(40),
  STATE_ABRV    VARCHAR2(2),
  GEOM          SDO_GEOMETRY);

CREATE TABLE GEOD_INTERSTATES(
  HIGHWAY   VARCHAR2(35),
  GEOM      SDO_GEOMETRY);

C.2 SDO_WITHIN_DISTANCE Examples

The SDO_WITHIN_DISTANCE operator identifies the set of spatial objects that are within some specified distance of a given object. You can indicate that the distance is approximate or exact. If you specify querytype=FILTER, the distance is approximate because only a primary filter operation is performed; otherwise, the distance is exact because both primary and secondary filtering operations are performed.

Example C-1 finds all cities within 15 miles of the interstate highway I170.

SELECT /*+ ORDERED */ c.city
FROM geod_interstates i, geod_cities c
WHERE i.highway = 'I170'
   AND sdo_within_distance (
        c.location, i.geom,
       'distance=15 unit=mile') = 'TRUE';

Example C-1 finds all cities within 15 miles ('distance=15 unit=mile') of the specified highway (i.highway = 'I170'), and by default the result is exact (because the querytype parameter was not used to limit the query to a primary filter operation). In the WHERE clause of this example:

• i.highway refers to the HIGHWAY column of the INTERSTATES table, and I170 is a value from the HIGHWAY column.

• c.location specifies the search column (geometry1): the LOCATION column of the GEOD_CITIES table.

• i.geom specifies the query window (aGeom): the spatial geometry in the GEOM column of the GEOD_INTERSTATES table, in the row whose HIGHWAY column contains the value I170.

Example C-2 finds all interstate highways within 15 miles of the city of Tampa.

SELECT /*+ ORDERED */ i.highway
FROM geod_cities c, geod_interstates i
WHERE c.city = 'Tampa'
   AND sdo_within_distance (
        i.geom, c.location,
        'distance=15 unit=mile') = 'TRUE';

Example C-2 finds all highways within 15 miles ('distance=15 unit=mile') of the specified city (c.city = 'Tampa'), and by default the result is exact (because the querytype parameter was not used to limit the query to a primary filter operation). In the WHERE clause of this example:

• c.city refers to the CITY column of the GEOD_CITIES table, and Tampa is a value from the CITY column.

• i.geom specifies the search column (geometry1): the GEOM column of the GEOD_INTERSTATES table.

• c.location specifies the query window (aGeom): the spatial geometry in the LOCATION column of the GEOD_CITIES table, in the row whose CITY column contains the value Tampa.

C.3 SDO_NN Examples

The SDO_NN operator determines the nearest neighbor geometries to a geometry. No assumptions should be made about the order of the returned results. If you specify no optional parameters, one nearest neighbor geometry is returned.

If you specify the optional sdo_num_res keyword, you can request how many nearest neighbors you want, but no other conditions in the WHERE clause are evaluated. For example, assume that you want the five closest banks from an intersection, but only where the bank name is CHASE. If the five closest banks are not named CHASE, SDO_NN with sdo_batch_size=5 will return no rows because the sdo_num_res keyword only takes proximity into account, and not any conditions in the WHERE clause.

If you specify the optional sdo_batch_size keyword, SDO_NN keeps returning neighbor geometries in distance order to the WHERE clause. If the WHERE clause specifies bank_name = 'CHASE' AND rownum < 6, you can return the five closest banks with bank_name = 'CHASE'.

SDO_NN_DISTANCE is an ancillary operator to the SDO_NN operator. It returns the distance of an object returned by the SDO_NN operator and is valid only within a call to the SDO_NN operator.

Example C-3 finds the five cities nearest to the interstate highway I170 and the distance in miles for each city, ordered by distance in miles.

SELECT /*+ ORDERED */
       c.city,
       sdo_nn_distance (1) distance_in_miles
FROM geod_interstates i, 
     geod_cities c 
WHERE i.highway = 'I170'
  AND sdo_nn(c.location, i.geom,
             'sdo_num_res=5 unit=mile', 1) = 'TRUE'
ORDER by distance_in_miles;

In Example C-3, because the /*+ ORDERED*/ optimizer hint is used, it is important to have an index on the GEOD_INTERSTATES.HIGHWAY column. In this example, the hint forces the query to locate highway I170 before it tries to find nearest neighbor geometries. In the WHERE clause of this example:

• i.highway refers to the HIGHWAY column of the INTERSTATES table, and I170 is a value from the HIGHWAY column.

• c.location specifies the search column (geometry1): the LOCATION column of the GEOD_CITIES table.

• i.geom specifies the query window (geometry2): the spatial geometry in the GEOM column of the GEOD_INTERSTATES table, in the row whose HIGHWAY column contains the value I170.

• sdo_num_res=5 specifies how many nearest neighbor geometries to find.

• unit=mile specifies the unit of measurement to associate with distances returned by the SDO_NN_DISTANCE ancillary operator.

• 1 (in sdo_nn_distance (1) and 'sdo_num_res=5 unit=mile', 1) is the number parameter value that associates the call to SDO_NN to the call to SDO_NN_DISTANCE.

In Example C-3, ORDER BY distance_in_miles orders the results from the WHERE clause by distance in miles.

The statement in Example C-3 produces the following output (slightly reformatted for readability):

CITY                    DISTANCE_IN_MILES
----------------------  ------------------------------
St Louis                  5.36297295
Springfield              78.7997464
Peoria                  141.478022
Evansville              158.22422
Springfield             188.508631

Example C-4 extends Example C-3 by limiting the results to cities with a 1990 population over a certain number. It finds the five cities nearest to the interstate highway I170 that have a population greater than 300,000, the 1990 population, and the distance in miles for each city, ordered by distance in miles.

SELECT /*+ ORDERED  NO_INDEX(c pop90_idx) */
       c.city, pop90,
       sdo_nn_distance (1) distance_in_miles
FROM geod_interstates i,
     geod_cities c
WHERE i.highway = 'I170'
  AND sdo_nn(c.location, i.geom,
            'sdo_batch_size=10 unit=mile', 1) = 'TRUE'
  AND c.pop90 > 300000
  AND rownum < 6
ORDER BY distance_in_miles;

In Example C-4, because the ORDERED optimizer hint is used, it is important to have an index on the GEOD_INTERSTATES.HIGHWAY column. In this example, the hint forces the query to locate highway I170 before it tries to find nearest neighbor geometries.

To ensure correct results, disable all nonspatial indexes on columns that come from the same table as the SDO_NN search column (geometry1). In this example, the NO_INDEX(c pop90_idx) optimizer hint disables the nonspatial index on the POP90 column.

In the WHERE clause of this example:

• sdo_batch_size=10 causes geometries to be returned continually (in distance order, in batches of 10 geometries), to be checked to see if they satisfy the other conditions in the WHERE clause.

• c.pop90 > 300000 restricts the results to rows where the POP90 column value is greater than 300000.

• rownum < 6 limits the number of results returned to five.

In Example C-4, ORDER BY distance_in_miles orders the results from the WHERE clause by distance in miles.

The statement in Example C-4 produces the following output (slightly reformatted for readability):

CITY               POP90    DISTANCE_IN_MILES
-----------------  -------  ---------------------
St Louis            396685     5.36297295
Kansas City         435146   227.404883
Indianapolis        741952   234.708666
Memphis             610337   244.202072
Chicago            2783726   253.547961

C.4 SDO_AGGR_UNION Example

When you use the SDO_AGGR_UNION aggregate function, very large geometries can result. When geometries have many coordinates, spatial operations (such as union) can be time-consuming. It may be better to divide a single spatial aggregate union operation function into multiple nested aggregate functions in the same SQL statement.

Example C-5 aggregates all the counties in Texas, producing the boundary for the state of Texas.

select sdo_aggr_union(mdsys.sdoaggrtype(aggr_geom,0.5)) aggr_geom
from (select sdo_aggr_union(mdsys.sdoaggrtype(aggr_geom,0.5)) aggr_geom
  from (select sdo_aggr_union(mdsys.sdoaggrtype(aggr_geom,0.5)) aggr_geom
      from (select sdo_aggr_union(mdsys.sdoaggrtype(aggr_geom,0.5)) aggr_geom
            from (select sdo_aggr_union(mdsys.sdoaggrtype(geom,0.5)) aggr_geom
                  from geod_counties where state_abrv='TX'
                  group by mod(rownum,16)
                 )
            group by mod (rownum, 8)
           )
      group by mod (rownum, 4)
     )
  group by mod (rownum, 2)
 );