Thank you for working on this! I got slightly into this patch. I can be wrong, but my opinion is that planner/optimizer-related changes are not without dangers generally. So anyway, they should be justified by performance increase, or the previous behavior should be considered totally wrong. Patching the thing which is just a little sub-optimal seems for me seems not necessary.
So it would be very good to see measurements of a performance gain from this patch. And also I think tests with partitioned and inherited relations for demonstration of the right work in the cases discussed in the thread should be a must-do for this patch. -- Best regards, Pavel Borisov Postgres Professional: http://postgrespro.com The new status of this patch is: Waiting on Author
