pickle broken: can't handle NaN or Infinity under win32

[Robert Kern]

Tim Peters wrote:

> OK, I looked, and it made no difference to me.  Really.  If I had an
> infinitely tall monitor, maybe I could see a difference, but I don't
> -- the sign of 0 on the nose makes no difference to the behavior of
> 1/x for any x other than 0.  On my finite monitor, I see it looks like
> the line x=0 is an asymptote, and the graph approaches minus infinity
> on that line from the left and positive infinity from the right; the
> value of 1/0 doesn't matter to that.

Well, the value of 1/0 is undefined. Occasionally, it's useful to report 
+inf as the value of 1.0/+0.0 because practically we're more concerned 
with limiting behavior from an assumed limiting process than being 
correct. By the same token, we might also be concerned with the limiting 
behavior coming from the other direction (a different limiting process), 
so we might want 1.0/-0.0 to give -inf (although it's still actually 
undefined, no different from the first expression, and inf is really the 
same thing as -inf, too).

Although I haven't read the paper you cited, it seems to me that the 
branch cut issue is the same thing. If you're on the cut itself, the 
value, practically, depends on which end of the branch you're deciding 
to approach the point from. It's arbitrary; there's no correct answer; 
but signed zeros give a way to express some of the desired, useful but 
wrong answers.

And floating point is about nothing if not being usefully wrong.

-- 
Robert Kern
rkern at ucsd.edu

"In the fields of hell where the grass grows high
  Are the graves of dreams allowed to die."
   -- Richard Harter