From: stuart yarus <syarus@...>
Date: 2006-06-19T23:54:39+09:00
Subject: Re: Numeric roadmap?

On 6/19/06, M. Edward (Ed) Borasky <znmeb@cesmail.net> wrote:
>
>
> tsuraan@tsuraan.net wrote:
> >> There MUST BE ONE SINGLE AND TRUE COMPLETE NUMERIC TYPE!!!
> >> namely, class ArbitraryPrecisionHyperComplexNumber !
> >
> > You will support quaternions and octonions, right?  It had also better
> > support vector rotation by quaternion, or I'm not using it :)
> >
> >> P.S. just a joke, raised by Unicode debate :)))
> >>
> > We'll see :)
> Why not do what R did --- define "scalars" as vectors of length 1 and
> make the vector the basic numeric type?
> >

Then you have a conformability problem.  How do you multiply a vector
times a scalar?  If a scalar is represented by a vector of length one,
"doubling" a vector of length greater than one won't work or will
require special code.


> >
>
> --
> M. Edward (Ed) Borasky
>
> http://linuxcapacityplanning.com
>
>
>