Re: Insert increment based on a group by?

From: Bruno Wolff III <bruno(at)wolff(dot)to>
To: mike <mike(at)bristolreccc(dot)co(dot)uk>
Cc: pgsql-general(at)postgresql(dot)org
Subject: Re: Insert increment based on a group by?
Date: 2004-06-04 15:34:37
Message-ID: 20040604153437.GB12744@wolff.to
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

On Fri, Jun 04, 2004 at 11:39:53 +0100,
mike <mike(at)bristolreccc(dot)co(dot)uk> wrote:
>
> So far I am getting an overall max or no insert.
>
> this is my query
>
> SELECT subj,area, CASE WHEN max(file_no) IS NULL THEN '1' ELSE max
> (file_no)+1 END FROM tb_files GROUP BY (subj||area),subj,area;

You should do a lock table in exclusive mode before doing this, unless
you know nothing else is going on.

An insert would like like:
INSERT INTO table_name (sub, area, file_no) select 'sub_value', 'area_value',
(SELECT max(file_no)+1 FROM table_name where sub = 'sub_value' AND
area = 'area_value')

If you have an index on either area, sub, file_no or sub, area, file_no,
you can rewrite the subselect to use ORDER BY and LIMIT.

It might be better to just assign unique values of file_no for the whole
table. You can use a sequence to support that. That would allow you
to support more concurrency.

In response to

Browse pgsql-general by date

  From Date Subject
Next Message zhicheng wang 2004-06-04 15:40:11 Re: after using pg_resetxlog, db lost
Previous Message Tom Lane 2004-06-04 15:20:56 Re: pg_class could not be found