Re: Datetime patch

I did a little more looking at this...

> test=> select '2003-09-31'::date;
> ERROR: Bad date external representation '2003-09-31'
>
> I did find the following change in the regression tests:
>
> INSERT INTO TIMESTAMP_TBL VALUES ('02-10-1997 17:32:01 PST');
> INSERT INTO TIMESTAMP_TBL VALUES ('19970210 173201 PST');
> INSERT INTO TIMESTAMP_TBL VALUES ('97FEB10 5:32:01PM UTC');
> + ERROR: Bad timestamp external representation '97FEB10 5:32:01PM UTC'

We still accept 10FEB97, which is a more common format. Seems this now
fails because the year/day detection is not gone, which is fine, I
think. This works as expected:

test=> select '01FEB03'::date;
date
------------
2003-02-01
(1 row)

> INSERT INTO TIMESTAMP_TBL VALUES ('97/02/10 17:32:01 UTC');
> + ERROR: Bad timestamp external representation '97/02/10 17:32:01 UTC'

Again, this one should fail.

Seems these are all related to two-digit years. This works when
specifying the year as 1997:

test=> select '97/02/10 17:32:01 UTC'::date;
ERROR: Bad date external representation '97/02/10 17:32:01 UTC'
test=> select '1997/02/10 17:32:01 UTC'::date;
date
------------
1997-02-10
(1 row)

> INSERT INTO TIMESTAMP_TBL VALUES ('97.041 17:32:01 UTC');
> + ERROR: Bad timestamp external representation '97.041 17:32:01 UTC'

I just played around with this one and got it working:

test=> select '2004.041'::date;
date
------------
2004-02-10
(1 row)

test=> select '2004.001'::date;
date
------------
2004-01-01
(1 row)

Seems you now need a 4-digit year to specify this format. Also
interesting is that the day must be three digits.

As part of the patch application, I will modify the years to have the
leading '19' so the regression tests will pass.

--
Bruce Momjian | http://candle.pha.pa.us
pgman(at)candle(dot)pha(dot)pa(dot)us | (610) 359-1001
+ If your life is a hard drive, | 13 Roberts Road
+ Christ can be your backup. | Newtown Square, Pennsylvania 19073