From: | Oleg K <windspirit(at)gmail(dot)com> |
---|---|
To: | pgsql-bugs(at)postgresql(dot)org |
Subject: | 9.0_alpha4 Planner Bug |
Date: | 2010-03-21 21:25:36 |
Message-ID: | 4BA68ED0.4060906@gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-bugs |
Hello,
My OS is Gentoo Linux 2.6.32.10 x86_64
Postgre:# select version();
PostgreSQL 9.0alpha4 on x86_64-pc-linux-gnu, compiled by GCC
x86_64-pc-linux-gnu-gcc (Gentoo 4.3.4 p1.0, pie-10.1.5) 4.3.4, 64-bit
There is a regression, in some query that was working fine in Postgre
8.4.3 and became broken after upgrade
the details are here http://pastebin.ca/1848469
basically, i have 2 tables cjw.service and cjw.company_to_service
when i run query that left-joins them and produces a boolean constant
named "linked" on successful join
SELECT
service_name AS name
-- , com_serv.service::bool AS linked
, linked AS linked
FROM
cjw.service
LEFT JOIN (
SELECT
service
, true as linked
FROM
cjw.company_to_service
WHERE
company = 3
) AS com_serv USING (service)
ORDER BY
service_name
so when I select that constant by it's name - the planner completely
ignores join and returns "true" for every row, even though that is wrong
and that row could not be joined
and if i select, a real value from joined part - the planner performs as
expected and returns valid result.
please checkout pastebin url http://pastebin.ca/1848469
i included descriptions of tables and ANALYZE of queries there.
as well as partial query results
please note that first query shows WRONG result and second one is OK.
Regards.
Oleg.
From | Date | Subject | |
---|---|---|---|
Next Message | Gudala, Sridhar (GE EntSol, Intelligent Platforms) | 2010-03-22 12:01:35 | Noticed a Bug with stored procedures |
Previous Message | Tom Lane | 2010-03-20 15:30:27 | Re: BUG #5385: gd disallows agg function in subselect of update |