Date: Thu, 18 May 2006 11:58:29 -0400
From: Andrew Sullivan <ajs@crankycanuck.ca>
To: pgsql-sql@postgresql.org
Subject: Re: Add column and specify the column position in a table
Message-ID: <20060518155829.GE29472@phlogiston.dyndns.org>
References: <446B9418.3080700@encs.concordia.ca> <446C0BC0.7050805@gmail.com>
	<20060518140014.GA3122@surnet.cl>
	<d854adec0605180843y646f953cwb490a03f3c1d0602@mail.gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <d854adec0605180843y646f953cwb490a03f3c1d0602@mail.gmail.com>
User-Agent: Mutt/1.5.9i

On Thu, May 18, 2006 at 05:43:19PM +0200, Guillaume Lelarge wrote:
> Do you mean that, using "alter table test add column" with the "after"
> option, MySQL creates a new table, populates it with the old table
> data and finally drops the old table ? I mean, there's the same
> performance problem with big tables ?

MySQL does that for a great deal of its DDL.  Yes, the performance is
awful for this on big tables.

The reason I didn't answer the OP's question, though, is that I can't
think of a legitimate reason to do this anyway.  The physical layout
of the columns should not be of concern to the developer, who should
be naming the columns anyway.

A

-- 
Andrew Sullivan  | ajs@crankycanuck.ca
The whole tendency of modern prose is away from concreteness.
		--George Orwell