On Sun, Sep 18, 2011 at 10:03 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
Craig Ringer <ringerc@ringerc.id.au> writes: > On 18/09/2011 5:51 PM, Melaka Gunasekara wrote: >> Merge Full Join (cost=10000000074.40..10000000093.69 rows=1159 width=286)
>> Can you suggest why the merge join is being suggested when I have >> turned it off ?
> AFAIK SETting a join type to "off" really just increases the cost > estimate so high that the planner will avoid using it where it has any > alternative. In this case, it doesn't seem to think it has any other way > to execute the query, or it thinks that any other way will be so > incredibly, insanely slow that the merge join is still better.
It's the first of those --- FULL joins are only implemented in the mergejoin logic, not in hash or nestloop joins, so there is no other way to do this query. (But as of 9.1, hash joins can do them too.)