I'm not quite sure I understand, then.
On the one hand, irrational numbers can not all be represented
exactly. And, we're using IEEE-754 binary 64 bit floating point
numbers as our final precision, in current J implementations. So we
have to deal with epsilon issues.
On the other hand, J also includes =!.0 for exact comparisons without
any built-in implementation of epsilon (which can be useful for
implementing alternate concepts of epsilon).
But, also, looking at j./&(*(1,:-.&({. = +/)))&(|:&:+.) it seems
rather specialized in character. I would be hard pressed to come up
with a general statement of purpose for something like that.
Is there something I'm overlooking, here?
Thanks,
--
Raul
On Fri, Feb 19, 2021 at 3:43 PM Hauke Rehr <[email protected]> wrote:
>
> I think that’s his point. He wants b to be 0 here.
>
> Am 19.02.21 um 21:39 schrieb Raul Miller:
> > Note that b may depend on a.
> >
> > b=: 1e6
> > a=: 9e99
> > a=a+b
> > 1
> >
> > FYI,
> >
>
> --
> ----------------------
> mail written using NEO
> neo-layout.org
>
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
