> From: Edward Loper> > If we *really* want nan==nan to be false, then it seems like > we have to > say that nan is unhashable.
Hmm - I suppose we could make a subclass of float specifically for NaN, and have that be a singleton. The float constructor could check for a NaN result, and return this singleton.
I'm worried that there might be a slowdown though.
> I'm also disturbed by the fact > that cmp() > has something different to say about their equality:> > >>> cmp(float('nan'), float('nan'))> 0
Hmm - I guess it should technically throw an exception. However, I'd be happy if it returned -1. Would this affect the stability of sorting?