Re: [xsl] floor() and idiv

["Andrew Welch" <andrew.j.welch@xxxxxxxxx>]

2008/11/6 Colin Paul Adams <colin@xxxxxxxxxxxxxxxxxx>:
>>>>>> "Andrew" == Andrew Welch <andrew.j.welch@xxxxxxxxx> writes:
>
>    >> >> Gestalt gives 10, 10, 10.
>    >>
>    Andrew> As far as I understand, which is not very far, it's a
>    Andrew> fundamental problem with binary arithmetic.  The standard
>    Andrew> division technique is fast but incurs this sort of
>    Andrew> problem, while the BigDecimal approach is exact but much
>    Andrew> slower.
>
> You've lost me - we were talking about integer division, I thought.

1000000000000000000000000000000001 is just too big for an 32 bit
integer...  (I think that's why DC chose it)


> If you are talking about xs:double, compared with xs:decimal, then
> yes. But I don't see how that comes into the problem.

I think it's the problem of representing a decimal:

http://speleotrove.com/decimal/decifaq1.html#inexact





-- 
Andrew Welch
http://andrewjwelch.com
Kernow: http://kernowforsaxon.sf.net/