From: | "Anton Krokovny" <kronya(at)mail(dot)ru> |
---|---|
To: | pgsql-general(at)postgresql(dot)org |
Subject: | Join tables using the closest datetime values |
Date: | 2006-02-16 04:59:18 |
Message-ID: | dt10r6$2m5$1@news.hub.org |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-general |
Hello,
I have two tables with same structure:
CREATE TABLE k1 (
begintime timestamp with time zone NOT NULL,
rowid serial NOT NULL
);
ALTER TABLE ONLY k1
ADD CONSTRAINT k1_time_key PRIMARY KEY (begintime);
They contain a huge number of data (about 100000-200000 records).
I need to join these tables on begintime column, but in most cases there are
no equal timestamps in each other. That's why I need to choose the closest
timestamp from another table.
At this moment I'm using the following implementation (sorting and taking
first above):
CREATE OR REPLACE FUNCTION test(x timestamp with time zone, OUT y int) AS
'SELECT rowid FROM k2 WHERE begintime <= $1 ORDER BY begintime DESC
LIMIT 1'
language SQL;
SELECT rowid, test(begintime) FROM k1;
This algorithm takes about 5 seconds for executing (excluding data
fetching).
Is there any solution faster than the current one?
Anton.
From | Date | Subject | |
---|---|---|---|
Next Message | Chris | 2006-02-16 05:07:38 | Re: Oracle purchases Sleepycat - is this the "other shoe" |
Previous Message | Tom Lane | 2006-02-16 04:38:47 | Re: Oracle purchases Sleepycat - is this the "other shoe" |