Re: BUG #1630: Wrong conversion in to_date() function. See example.

The following bug has been logged online:

Bug reference: 1630
Logged by: Ariel E. Carn/Elizabeth Sosa
Email address: acarna(at)tarifar(dot)com
PostgreSQL version: 7.3.8/7.4.6
Operating system: SuSE Linux
Description: Wrong conversion in to_date() function. See example.
Details:

Case PgSQL 7.3.8/SuSE Linux 8.2 (i586)
======================================
PROD=# select to_date('2005-02-32', 'YYYY-MM-DD');
to_date
------------
2005-03-04
(1 row)

PROD=# select version();
version
----------------------------------------------------------------------------
-----------------------------
PostgreSQL 7.3.8 on i686-pc-linux-gnu, compiled by GCC gcc (GCC) 3.3
20030226 (prerelease) (SuSE Linux)
(1 row)

Case PgSQL 7.4.6/SuSE Linux 9.2
(i586)======================================
PROD=> select to_date('2005-02-32', 'YYYY-MM-DD');
to_date
------------
04-03-2005
(1 row)

PROD=> select version();
version
----------------------------------------------------------------------------
-----------------
PostgreSQL 7.4.6 on i686-pc-linux-gnu, compiled by GCC gcc (GCC) 3.3.4 (pre
3.3.5 20040809)
(1 row)

On Tue, Apr 26, 2005 at 11:48:12PM +0100, Ariel E. Carná/Elizabeth Sosa wrote:
>
> Case PgSQL 7.3.8/SuSE Linux 8.2 (i586)
> ======================================
> PROD=# select to_date('2005-02-32', 'YYYY-MM-DD');
> to_date
> ------------
> 2005-03-04
> (1 row)
>
> Case PgSQL 7.4.6/SuSE Linux 9.2
> (i586)======================================
> PROD=> select to_date('2005-02-32', 'YYYY-MM-DD');
> to_date
> ------------
> 04-03-2005
> (1 row)

What bug are you reporting? These are the same date displayed in
different formats -- are you familiar with the DateStyle setting?

http://www.postgresql.org/docs/7.3/interactive/datatype-datetime.html#DATATYPE-DATETIME-OUTPUT
http://www.postgresql.org/docs/7.4/interactive/datatype-datetime.html#DATATYPE-DATETIME-OUTPUT

test=> SELECT version();
version
---------------------------------------------------------------------------
PostgreSQL 7.4.7 on sparc-sun-solaris2.9, compiled by GCC gcc (GCC) 3.4.2
(1 row)

test=> SET DateStyle TO 'ISO,MDY';
SET
test=> SELECT to_date('2005-02-32', 'YYYY-MM-DD');
to_date
------------
2005-03-04
(1 row)

test=> SET DateStyle TO 'Postgres,DMY';
SET
test=> SELECT to_date('2005-02-32', 'YYYY-MM-DD');
to_date
------------
04-03-2005
(1 row)

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.

--
Michael Fuhr
http://www.fuhr.org/~mfuhr/

"Ariel E. Carn/Elizabeth Sosa" <acarna(at)tarifar(dot)com> writes:
> PROD=# select to_date('2005-02-32', 'YYYY-MM-DD');
> to_date
> ------------
> 2005-03-04
> (1 row)

I'm not convinced that's a bug --- most implementations of the Unix
mktime function will handle out-of-range day numbers like that.

regards, tom lane

Even javascript handles those dates the same way...

On 4/27/05, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>
> "Ariel E. Carná/Elizabeth Sosa" <acarna(at)tarifar(dot)com> writes:
> > PROD=# select to_date('2005-02-32', 'YYYY-MM-DD');
> > to_date
> > ------------
> > 2005-03-04
> > (1 row)
>
> I'm not convinced that's a bug --- most implementations of the Unix
> mktime function will handle out-of-range day numbers like that.
>
> regards, tom lane
>
> ---------------------------(end of broadcast)---------------------------
> TIP 5: Have you checked our extensive FAQ?
>
> http://www.postgresql.org/docs/faq
>

Michael Fuhr <mike(at)fuhr(dot)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?

regards, tom lane

With Oracle (at least since 6.x version) this conversion is wrong.

Atentamente ,

Ariel Carná

-----Mensaje original-----
De: Tom Lane [mailto:tgl(at)sss(dot)pgh(dot)pa(dot)us]
Enviado el: Miércoles, 27 de Abril de 2005 11:54 a.m.
Para: Michael Fuhr
CC: Ariel E. Carná/Elizabeth Sosa; pgsql-bugs(at)postgresql(dot)org
Asunto: Re: [BUGS] BUG #1630: Wrong conversion in to_date() function. See
example.

Michael Fuhr <mike(at)fuhr(dot)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?

regards, tom lane

On Wed, 2005-04-27 at 10:53 -0400, Tom Lane wrote:
> Michael Fuhr <mike(at)fuhr(dot)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(at)zf(dot)jcu(dot)cz>