Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Edge case conversions, exceptions to IEEE FPA

On Friday 2008-11-14 14:12 UTC I wrote:
> Radius is clear:  +Inf for unbounded (incl. Entire),
> 0 for Empty, and finite non-zero positive otherwise.

I goofed; 0 is for a Singleton.

I suppose the choice for rad(Empty) is NaN or, as John Pryce
does in his implementation, -Inf.  If mid(Empty) returns NaN,
shouldn't rad(Empty) do so too?

Michel.

P.S.  Thanks, John, for setting me straight.
Sent: 2008-11-14 17:57:34 UTC