cvs commit: fptools/libraries/base/GHC Float.lhs Real.lhs

[Sven Panne]

Simon wrote:
> [...] what should 'decodeFloat (read "NaN" :: Float)' do?

I'd vote for

   uncurry encodeFloat . decodeFloat
   decodeFloat . uncurry encodeFloat

to be identities for *all* RealFloats/(Integer,Int)s and don't raise an
exception for border cases like NaN or Infinity.

> Currently we return some junk values (related to the actual representation
> of NaN as a Float)

IMHO they are not junk at all.  :-)

Cheers,
   S.