Re: [Chicken-users] Odd floating-point representations

[felix]

On Wed, 15 Oct 2003 12:04:35 +0100, Category 5 <address@hidden> 
wrote:

> When I evaluate most flonums the result is as expected:
>
> > > > 0.39
> 0.39
>
> But sometimes it isn't:
>
> > > > (list 0.07 0.56 0.69 0.81 0.94)
> (0.07000000000000001 0.5600000000000001 0.6899999999999999
> 0.8100000000000001 0.9399999999999999)
>
> This is problematic in some situations.  Can anyone explain what's going
> on here?

The C floating-point to string conversion routines used by the
runtime system produce this when the precision is set too high.
Currently it uses 16, so if you change the definition of
FLONUM_PRINT_PRECISION in runtime.c from 16 to 15, you should get
decent output (at least on Linux, where I've tested it).

(Is it me, or is the whole number-to-string-and-back conversion stuff
provided by the C runtime library kludgy hack?)


cheers,
felix