Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!uunet!zaphod.mps.ohio-state.edu!mips!pacbell.com!ucsd!nosc!humu!pegasus!tleylan
From: tleylan@pegasus.com (Tom Leylan)
Newsgroups: comp.databases
Subject: Re: Clipper round() problem
Message-ID: <1991Jun28.221027.7939@pegasus.com>
Date: 28 Jun 91 22:10:27 GMT
References: <1991Jun27.095337.11969@vaxa.strath.ac.uk>
Organization: Pegasus,  Honolulu
Lines: 22

In article <1991Jun27.095337.11969@vaxa.strath.ac.uk> cnas09@vaxa.strath.ac.uk writes:
>
>	Can anyone explain the (to me) rather peculiar behaviour of the round()
>routine in Clipper 5? If you do something like round(37*1.175, 2) then round()
>correctly reports 43.48 but using round(39*1.175, 2) gives 45.82 when it should
>have rounded it up to 45.83! 
>
Aaron,

I can't give you _the_ reason but I believe that it is somewhat due to the
behavior of the MSC alternate floating point library that Clipper uses.

It would seem that the binary representation of 39 * 1.175 isn't 45.825 but
is more like 45.8249999999 while the output is 45.825.

I think it's annoying but I don't believe that you'll find that the ROUND()
function is "wrong" but rather that the number it is being asked to round
isn't the one that we see.

I'll see if I can track down a more exact answer.

tom