Re: [HACKERS] Interval aggregate regression failure

Michael Glaesemann wrote:
>
> On Aug 29, 2006, at 15:38 , Michael Glaesemann wrote:
>
> > I think I've got it. I plan to update the regression tests this
> > evening, but I wanted to post what I believe is a solution.
>
> I've cleaned up the patch a bit in terms of whitespace, comments, and
> parens. I've also updated the interval and horology regression tests.
> The horology tests needed updating because I added 5 rows to
> INTERVAL_TBL. I didn't check the math for every row of time(tz |
> stamp | stamptz)/interval arithmetic in the horology tests as I think
> problems in this area would have shown up before. Does that make
> sense or it just rationalization on my part?
>
> Both with and without --enable-integer-datetimes pass the regression
> tests.

Uh, I came up with a cleaner one, I think. I didn't test
--enable-integer-datetimes yet.

I tested a few of your examples:

test=> select '41 mon 10:00:00'::interval / 10 as "pos";
pos
------------------------
4 mons 3 days 01:00:00
(1 row)

It basically rounds the remainders to full values if they are close to
full (+/- 0.000001).

--
Bruce Momjian bruce(at)momjian(dot)us
EnterpriseDB http://www.enterprisedb.com

+ If your life is a hard drive, Christ can be your backup. +