Re: [Scheme-reports] Reformulated numeric-tower ballot
Bear 30 Apr 2014 20:52 UTC
On Mon, 2014-04-28 at 23:48 -0400, John Cowan wrote:
> 1) Should R7RS-large require arbitrarily large (up to implementation
> restrictions like memory) exact integers?
It should require exact integers. It should encourage range
limits much higher than typical programming languages. It
should not require an implementation to exhaust all memory
and crash in an attempt to represent a very large exact integer.
> 2) Should R7RS-large require support for exact rational numbers?
It should require exact ratios. It should not require an
implementation to exhaust all memory and crash in an attempt
to represent a very precise ratio.
> 3) Should R7RS-large require support for exact complex numbers?
Previously I voted 'yes' subject to the same representation limits
as other exact numbers. But as I consider this there's an issue.
I would vote 'yes' for consistency with the rest of the numeric
tower, but it is hard to say exactly what a 'yes' here means in
the absence of any constraint on whether the exact numbers
represented are, eg, stored in polar or cartesian format, or in
some union that could be either, and if in polar format whether
the angle measurement is given in radians (either directly or as
some product of pi like degrees), or as a slope ratio.
Working with a set of exact numbers means to me that you should
be able to know what set of operations preserve or do not preserve
that property to the limits of representation, but depending on
what schema is chosen to represent complex numbers, the sets of
exactness preserving operations are different and the sets of
exactly representable numbers are not related to each other in
any simple way.
> 4) Should R7RS-large require inexact complex numbers?
Yes.
_______________________________________________
Scheme-reports mailing list
Scheme-reports@scheme-reports.org
http://lists.scheme-reports.org/cgi-bin/mailman/listinfo/scheme-reports