Josh Berkus <josh@agliodbs.com> writes:
> On 8/15/12 1:51 PM, Tom Lane wrote:
>> Maybe you had better show us the actual query, and the table/index
>> definitions. Because it's sure making odd choices here. This seems
>> like the wrong join order altogether ...
> We'll need to do that off-list for confidentiality reasons.
If you can show us the anonymized query plan, why not the anonymized query?
It doesn't look like it could be all that complicated.
regards, tom lane