Re: [HACKERS] Interval aggregate regression failure (expected seems

From: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
To: Bruce Momjian <bruce(at)momjian(dot)us>
Cc: Michael Glaesemann <grzm(at)seespotcode(dot)net>, Michael Paesold <mpaesold(at)gmx(dot)at>, PostgreSQL-patches <pgsql-patches(at)postgresql(dot)org>
Subject: Re: [HACKERS] Interval aggregate regression failure (expected seems
Date: 2006-09-03 19:47:47
Message-ID: 27584.1157312867@sss.pgh.pa.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-patches

Bruce Momjian <bruce(at)momjian(dot)us> writes:
> Is this non-datetime integer only or both? I cannot reproduce the
> failure here.

On HPPA with float datetimes with today's code, Michael's case works
but it took me less than two minutes to find one that doesn't:

regression=# select interval '14 mon' * 8.2 as product_h;
product_h
---------------------------------
9 years 6 mons 23 days 24:00:00
(1 row)

I reiterate my comment that this approach will never work; any small
amount of experimentation will turn up cases that don't round correctly
on one platform or another. Float arithmetic is inherently inexact.

regards, tom lane

In response to

Responses

Browse pgsql-patches by date

  From Date Subject
Next Message Gregory Stark 2006-09-03 21:09:44 Re: Concurrent connections in psql patch
Previous Message Bruce Momjian 2006-09-03 19:45:21 Re: [HACKERS] Interval aggregate regression failure