From: Chris Pine <nemo@...>
Date: 2003-04-19T07:57:21+09:00
Subject: Re: Biased weighted random?

----- Original Message -----
From: "Hal E. Fulton" <hal9000@hypermetrics.com>

I've also noticed that for sufficiently strange
weights, Chris's method blows up with some
ridiculuous Avogadro-style numbers.

Could be partly a roundoff issue...
----------------------------

Hmmm... could be.  You could use Rationals instead of floats.  I did at
first, but it was *much* slower.  Plus, I was iterating, anyway, so I had no
sense of mathematical pride about it.  :)

I noticed it giving strange answers when I asked it to do something
impossible (like weights 3 and 1), but otherwise it seemed ok... has this
not been your experience?

If we solved this the *right* way, it would probably tell us we were
dividing by zero (or something) in these impossible cases.

Chris