On 08/03/2011 06:29 AM, Robert Haas wrote:
b. the Merge Join cost estimator did a poor job with the data it was given:
In function eqjoinsel_inner there are two cases (1) ANALYZE data is
available for both sides of the join and (2) ANALYZE data is missing for one
or both sides. Due to the GROUP
On Fri, Jul 8, 2011 at 9:33 PM, Clem Dickey dicke...@us.ibm.com wrote:
a. The Join cost estimators could have been given more information
The functions which estimate JOIN selectivity (e.g. the chance that tuples
will match in an equijoin, for instance) use data produced by ANALYZE. But
the
On 07/06/2011 05:59 PM, Clem Dickey wrote:
On 07/05/2011 07:26 PM, Clem Dickey wrote:
Column | Type | Modifiers
+-+---
y | integer | not null
x | integer | not null
k | integer | not null
j | integer | not null
z | integer | not null
Indexes:
t_pkey PRIMARY KEY, btree
On 07/05/2011 07:26 PM, Clem Dickey wrote:
Updates after belatedly reading the slow queries guidelines:
Version: PostgreSQL 8.4.8 on x86_64-redhat-linux-gnu, compiled by GCC
gcc (GCC) 4.4.5 20101112 (Red Hat 4.4.5-2), 64-bit
The query has always been slow; the table for this test case is