implementation categories, exact rationals Aubrey Jaffer (14 Oct 2005 18:29 UTC)
Re: implementation categories, exact rationals John.Cowan (14 Oct 2005 19:26 UTC)
Re: implementation categories, exact rationals Aubrey Jaffer (14 Oct 2005 19:38 UTC)
Re: implementation categories, exact rationals John.Cowan (14 Oct 2005 20:16 UTC)
Re: implementation categories, exact rationals bear (16 Oct 2005 18:08 UTC)
Re: implementation categories, exact rationals Michael Sperber (17 Oct 2005 07:44 UTC)
Re: implementation categories, exact rationals Aubrey Jaffer (17 Oct 2005 21:59 UTC)
Re: implementation categories, exact rationals Bradley Lucier (17 Oct 2005 22:07 UTC)

implementation categories, exact rationals Aubrey Jaffer 14 Oct 2005 18:11 UTC

 | Most implementations of Scheme fall into one of the following
 | categories:
 |
 |     * fixnums only (now rare except in toy implementations)
 |     * fixnums and flonums only
 |     * exact rationals and flonums only (no imaginary numbers)
 |     * the complete numeric tower

SCM, Guile, and SCM-Mac (and one presumes Galapagos and WinSCM) have
fixnums, bignums, and real and complex flonums.  SCM and its
derivatives have a large installed base.  In particular, Guile is part
of every Fedora GNU/Linux installation.  This category should be
included in your list.

 | This SRFI proposes to revise section 6.2 ("Numbers") of R5RS by:
 |
 |     * requiring a Scheme implementation to provide the full tower,
 |       including exact rationals of arbitrary precision, exact
 |       rectangular complex numbers with rational real and imaginary
 |       parts, and inexact real and complex arithmetic

What is the rationale for mandating exact rationals?  Over 15 years I
have written numerical Scheme code for everything from symbolic
algebra to Galois fields to linear systems to optics simulations
without needing exact rationals.

A case could be made if (expt -26. 1/3) returned -2.9624960684073702;
but I know of no Scheme implementation that does so.