On Wed, 2005-04-27 at 10:53 -0400, Tom Lane wrote:
> Michael Fuhr <mike@fuhr.org> writes:
> > If anything I'd expect 2005-02-32 to be rejected as invalid, but I
> > don't know the history or rationale behind to_date's behavior.
>
> It is rejected by the standard date input converter:
>
> regression=# select '2005-02-32'::date;
> ERROR: date/time field value out of range: "2005-02-32"
> HINT: Perhaps you need a different "datestyle" setting.
>
> However we consider that to_date() exists to be Oracle compatible,
> and so I would regard this as a bug if and only if Oracle does
> something different with the same input. Anyone know?
It's "almost" bug. And it's in TODO (but it's really long todo..:-( I
think to_date/timestamp() should be more pedantic.
Karel
PS. for volunteers for the work on new generation of to_char/date():
http://people.redhat.com/kzak/libfmt/libfmt-0.2-03262005.tar.gz
--
Karel Zak <zakkr@zf.jcu.cz>