ALTER TABLE - ALTER COLUMN question

Lists: pgsql-admin
From: Keaton Adams <kadams(at)mxlogic(dot)com>
To: pgsql-admin(at)postgresql(dot)org
Subject: ALTER TABLE - ALTER COLUMN question
Date: 2007-05-29 20:42:13
Message-ID: 1180471333.19986.13.camel@MXLRMT-208.corp.mxlogic.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-admin

When I issue an alter table ..... alter column command, the table
contents are copied from the old structure to the new, correct? So I
need 2x the table space available so the old and new tables can exist at
the same time? PostgreSQL doesn't use an in-place alter-table for these
types of changes, correct?

In this example the created and msg_datetime columns are currently
defined as TIMESTAMP WITHOUT TIME ZONE and no default. If the alter
table for these changes is not in-place then I would want to run it as a
single alter with all of the changes, such as this, instead of
individual alters for each change, correct? (Because each individual
alter would require another table copy?)

ALTER TABLE my_table ALTER COLUMN created TYPE TIMESTAMP WITH TIME
ZONE,
ALTER COLUMN created SET DEFAULT ('now'::text)::TIMESTAMP(6) WITH
TIME ZONE,
ALTER COLUMN msg_datetime TYPE TIMESTAMP WITH TIME ZONE,
ALTER COLUMN msg_datetime SET DEFAULT ('now'::text)::TIMESTAMP(6) WITH
TIME ZONE;

Thanks,

Keaton


From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: kadams(at)mxlogic(dot)com
Cc: pgsql-admin(at)postgresql(dot)org
Subject: Re: ALTER TABLE - ALTER COLUMN question
Date: 2007-05-29 21:16:59
Message-ID: 17811.1180473419@sss.pgh.pa.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-admin

Keaton Adams <kadams(at)mxlogic(dot)com> writes:
> In this example the created and msg_datetime columns are currently
> defined as TIMESTAMP WITHOUT TIME ZONE and no default. If the alter
> table for these changes is not in-place then I would want to run it as a
> single alter with all of the changes, such as this, instead of
> individual alters for each change, correct? (Because each individual
> alter would require another table copy?)

That's why ALTER TABLE has the ability to do more than one thing in
parallel ...

regards, tom lane


From: Keaton Adams <kadams(at)mxlogic(dot)com>
To: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
Cc: pgsql-admin(at)postgresql(dot)org
Subject: Re: ALTER TABLE - ALTER COLUMN question
Date: 2007-05-29 21:19:59
Message-ID: 1180473599.19986.15.camel@MXLRMT-208.corp.mxlogic.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-admin

After I sent my request I found the notes related to ALTER TABLE on page
1,331 of the 8.1 manual, so that helped to explain it. Your reply to my
post confirmed what I thought I was reading.

Thanks for the reply,

Keaton

On Tue, 2007-05-29 at 17:16 -0400, Tom Lane wrote:

> Keaton Adams <kadams(at)mxlogic(dot)com> writes:
> > In this example the created and msg_datetime columns are currently
> > defined as TIMESTAMP WITHOUT TIME ZONE and no default. If the alter
> > table for these changes is not in-place then I would want to run it as a
> > single alter with all of the changes, such as this, instead of
> > individual alters for each change, correct? (Because each individual
> > alter would require another table copy?)
>
> That's why ALTER TABLE has the ability to do more than one thing in
> parallel ...
>
> regards, tom lane