Horton

Re: A creepy story about dates. How to prevent it?

scott.marlowe <scott.marlowe@ihs.com>

View thread View on postgr.es
From: "scott.marlowe" <scott.marlowe@ihs.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Bruce Momjian <pgman@candle.pha.pa.us>, Peter Eisentraut <peter_e@gmx.net>, Jean-Luc Lachance <jllachan@nsd.ca>, Frank Miles <fpm@u.washington.edu>, Bruno Wolff III <bruno@wolff.to>, Ron Johnson <ron.l.johnson@cox.net>, <pgsql-general@postgresql.org>
Date: 2003-06-23T19:11:12Z
Lists: pgsql-general 
On Mon, 23 Jun 2003, Tom Lane wrote:

> "scott.marlowe" <scott.marlowe@ihs.com> writes:
> > The one thing that should absolutely be turned off is day/month swapping 
> > on dates of the form: 2003-02-22.
> 
> Agreed on that.  YYYY-DD-MM isn't used in the real world AFAIK, and it's
> reasonable to treat it as an error.
> 
> > I've seen little actual defense of the current behaviour,
> 
> Other than me, I think you mean.  dd/mm/yyyy and mm/dd/yyyy are
> inherently ambiguous in the real world, and when you can clearly
> determine what the intended meaning is, I think it's more reasonable
> to assume the datestyle isn't set correctly than to reject the data.

I thought the locale set that kind of behaviour didn't it?  If so, then 
it's better to fail loudly then quietly accept bad data.  But if the 
locale doesn't define such a thing, or it can't be set in postgresql.conf, 
the it's best to just avoid that date style altogether.