Re: [BUGS] planner regression in 8.4 (from 8.1)

2010-02-21 Thread Robert Haas
On Wed, Feb 17, 2010 at 1:06 PM, Ben Chobot be...@silentmedia.com wrote: - Hash (cost=153.63..153.63 rows=2178408 width=4) (actual time=0.207..0.207 rows=1 loops=1) - Nested Loop (cost=4.58..153.63 rows=2178408 width=4) (actual time=0.203..0.204 rows=1 loops=1)

Re: [BUGS] planner regression in 8.4 (from 8.1)

2010-02-21 Thread Ben Chobot
On Feb 21, 2010, at 4:28 AM, Robert Haas wrote: On Wed, Feb 17, 2010 at 1:06 PM, Ben Chobot be...@silentmedia.com wrote: - Hash (cost=153.63..153.63 rows=2178408 width=4) (actual time=0.207..0.207 rows=1 loops=1) - Nested Loop (cost=4.58..153.63 rows=2178408

[BUGS] planner regression in 8.4 (from 8.1)

2010-02-17 Thread Ben Chobot
I have a query which performs much better on 8.1.19 than on 8.4.2, unless I add offset 0 to the subqueries. I believe this is due to miscalculating the expected row count of nest loop joins. I cannot give you my data, but I can give you the query and the plans. Let me know if anything else