[postgis-users] && operator memory leak

[Carl Anderson]

Is anyone actively working on the memory leak in the gist && operator.
Or can anyone give me a heads up start point?

I have had enough of partitioning queries to work around an out ouf  
memory condition and willing to delve into this issue.

the issue manifests most simply with a gist index on the_geom for both  
tables and

select 1 from mytab
   where
     exists ( select 1
                from othertab
                  where othertab.the_geom && mytab.the_geom
            );

with
   mytab ~ 100 thousand tuples
and
   othertab ~ 1 million tuples

much more than 3Gb of real memory is required during the query.

my datasets have much more than 1 million tuples.

C.


Carl Anderson