Slow SDO_RELATE when using group by.. please help!!

Question

Hi All,
I'm fairly new to the spatial side of Oracle so please go easy on me.
I have 2 tables (a and b), each containing line geometries. There is a one(a) to many(b) relationship between the two.
I want to write a query which selects information from table b where the geometries in table b are completely covered by the geometries in table a, but the returned info is grouped by the rows in table a.
Here's my code:
select
a.route_id,
a.step,
max(b.gritter_status),
sum(b.salt_usage)
from
tbl_routedirs a, tbl_routelinks b
where
a.route_id = b.route_id and
sdo_covers(a.geoloc, b.geoloc) = 'TRUE'
group by
a.route_id, a.step
The query is extremely slow, taking up to 15 minutes to execute...The thing which is confusing me is without the group by clause the statement executes in under half a second.
I have checked all indexes etc - all are valid. I've tried re-ordering the tables in the from clause as ive read this can be a factor, and granted i have little knowledge as to what to swap where but this appeared to have no impact on the speed.
Any suggestions would be warmly welcomed... Thankyou.
Edited by: user8760008 on 27-Aug-2009 07:08 - i copied the script and didnt add the aggregate functions on the selected columns...

Answer 1

If that's the case, please try:
select /*+ leading(a) use_nl(a b) index(b your_index_on_b.geoloc) */
a.route_id,
a.step,
max(b.gritter_status),
sum(b.salt_usage)
from
tbl_routedirs a, tbl_routelinks b
where
a.route_id = b.route_id and
sdo_coveredby(b.geoloc, a.geoloc) = 'TRUE'
group by
a.route_id, a.step
i.e. to have a nested loops join, and take a small table as outer.
Note you may not need all three hints to get the nested loops join plan.