Re: issue with double ordering in a wrapped distinct

[Jonathan Vanasco <postgres@xxxxxxxx>]

On Nov 18, 2014, at 6:43 PM, Tom Lane wrote:

but as for why it gets a much worse plan after
flattening --- insufficient data.

Thanks.  I'll run some test cases in the morning and post the full queries matched with ANALYZE EXPLAIN.  

This is just puzzling to me.  I was hoping there might be a more general planner issue that someone would have noticed.

On Nov 18, 2014, at 6:55 PM, David G Johnston wrote:

I presume you have a reason for not simply doing away with the subquery
altogether...

When not using the subquery, the query ran in 1s -- if I didn't have a LIMIT/OFFSET

Adding a LIMIT/OFFSET to that query made it run for nearly 6 minutes.

The only way I could manage to trick it to use the better query plan, was to wrap the "good" query as a subquery, and then run a LIMIT/OFFSET in the outer query.