[plt-scheme] number system

From: Paul Schlie (schlie at comcast.net)
Date: Fri Oct 22 11:53:43 EDT 2004

One of the suggestions which I really liked and thought it would help
eliminate a lot of confusion, is the notion of you-see-what-you-get,
with respect to the conversion between inexact and exact representations;
as it would seem to be nice to expect something like:

> (exact 0.3)
3/10
> (inexact (exact 0.3))
0.3
> (inexact 1/3)
0.33333333333
> (exact (inexact 1/3))
33333333333/100000000000

Where both exact and inexact could accept an optional number system radix
parameter which defines the rounding mode of the conversion (defaulting to
the environment setting), such that:

> (exact (inexact 3/10 2) 2)
5404319552844595/18014398509481984

As we get now, likely often creating confusion at times, as opposed to:

> (exact (inexact 3/10))
3/10

Which most would logically expect.

> Zhu Chongkai wrote:
> I just read the "Cleaning up the tower: Numbers in Scheme" from Scheme
> WorkShop 2004. Although I do not agree with some of the points from the
> paper, the current number system of Scheme implements is not so
> satisfying. Will PLT scheme support arbitrary big real number (and
> more)? Would it be a better solution to add a better number system into
> SRFI? 




Posted on the users mailing list.