Zak Arntson wrote:
... cut stuff about distance calculation.


This is pretty cool.
http://www.ifm.liu.se/~ulfek/projects/SCAM-current/doc/SCAM.html

ODE through pyode also supports collision detection.


Here's a couple of functions which you can quickly cut the number of 
entities to check against in two.

def get_unique_pairs(vals):
    """ finds unique combinations from a given list of values.
         Which doesn't include combinations with itself.
    """
    vals_cpy = vals
    pairs = []
    pairs_extend = pairs.extend

    while vals_cpy:
       
        vals_head = vals_cpy[0]
        vals_cpy = vals_cpy[1:]

        new_ones = map(lambda x:(vals_head,x), vals_cpy)
        pairs_extend(new_ones)

    return pairs


def get_to_collide(vals):
    """ Returns a dict keyed by one of the vals, valued by a
         list of the vals which to check collisions against.
    """
    vals_cpy = vals
    tocollide = {}

    while vals_cpy:
        vals_head = vals_cpy[0]
        vals_cpy = vals_cpy[1:]

        if vals_cpy:
            tocollide[vals_head] = vals_cpy

    return tocollide


Have fun!