integers and floating points

btanksley@hifn.com btanksley@hifn.com
Wed, 27 Oct 1999 09:47:21 -0700


> From: Laurent Martelli [mailto:martelli@iie.cnam.fr]
> Subject: Re: integers and floating points

> >>>>> "JS" == shapj  <shapj@us.ibm.com> writes:

>   JS> Let's not be so open and high level that we allow ourselves to
>   JS> ignore mathematical realities.  Integers and floating point
>   JS> numbers have deeply different mathematical properties, and in
>   JS> the end a high-level system is no good unless it generates the
>   JS> results the user wants to see.

> Yes, but the set of integers is a subset of the reals, so they are not
> so different. That's why we exepect to be able to add an integer and a
> real. 

Don't expect that anymore; it's wrong.  The set of integers is not a subset
of the floats; there are integers which cannot be represented as floats.

(x+y)-x is not always equal to x+(y-x) in floating point math.  It's always
true for integers.

> Laurent Martelli

-Billy