On Jan 5, 5:12 pm, "Martin v. Löwis" <[EMAIL PROTECTED]> wrote:
> > Any other ideas?
>
> How about this:
>
> def random_pick(list, property):
> L = len(list)
> pos = start = random.randrange(L)
> while 1:
> x = list[pos]
> if property(x): return x
> pos = (pos + 1) % L
> if pos == start:
> raise ValueError, "no such item"
This might be acceptable for the OP's use, but it's strongly biased
towards values which follow a long stream of things that fail
property.
print [random_pick(range(100), lambda x:x > 90)
for i in range(999)].count(91)
929
If it were uniform, it should be around 111.
--
Paul Hankin
--
http://mail.python.org/mailman/listinfo/python-list
