To improve the efficiency

Question

hi all,
I have written the following script to do near analysis on road and point data type. Road table has 36629 road and point table has 10834 points and want to identify all points which are within one km of the road.
Query is running even after 24 hours though spatial column in road table is indexed, any comment is highly appretiated to improve efficiency.
<script>
INSERT /*+APPEND*/  INTO REL161_AU_POI_ACT_NR1 ( PT_IDX , X ,Y,RD_IDX, DISTANCE, GEOM   )(
SELECT /*+ ORDERED USE_NL(p,r) */ p.idx ,p.X,p.Y, r.ID_NUMBER, SDO_NN_DISTANCE(1),
MDSYS.SDO_GEOMETRY(2001,8311,MDSYS.SDO_POINT_TYPE(P.X,P.Y,NULL),NULL,NULL)
sdo_shape FROM point P,
road R
WHERE SDO_NN( r.GEOM, MDSYS.SDO_GEOMETRY(2001,8311,MDSYS.SDO_POINT_TYPE(P.X,P.Y,NULL),NULL,NULL), 'sdo_batch_size=10 UNIT=KM',1) = 'TRUE'
AND SDO_NN_DISTANCE(1) <= 1)
</script>
Best Regards,
Vaneeza

Answer 1

Vaneeza
Have you tried using sdo_within_distance operator? You want to know ALL objects within a distance of 1KM, as opposed to the X nearest objects, for which the SDO_NN operator is more suitable. The SDO_NN operator will return the objects ordered based on their distance before applying the other where clauses.
There is no limit on numbers of nearest objects set in your query
By having set the sdo_batch_size=10, the SDO_NN will evaluate the nearest candidates against the other clauses untill they are TRUE. What I am not sure here is whether the operator stops after having the first candidate evaluated to FALSE for the SDO_NN_DISTANCE(1) < 1 (your distance criteria) or whether the operator still process all nearest candidates as there is no explicit limitations on the number of nearest neighbors. (maybe one of the Oracle guys can comment this and decide if that might cause the general long processing time).
Within 11g you can have the distance set as a parameter of the SDO_NN operator, not as an additional where clause (which might be the only distance influence to the index use within the operator, but again this is for the oracle guys to comment)
Also, you want for all points find all roads within a distance of 1 KM of the points, not one specific point. In that case it might, if feasible, be better to store the points as sdo_geometry explicitly with a spatial index. Then you can use a spatial join with the distance parameter and have the distance between the object calculated using the sdo_geom.sdo_distance function.
Luc
Edited by: lucvanlinden on May 25, 2009 10:50 PM