Spatial Python, anyone?

[Magnus Lie Hetland]

From: "David Ascher" <DavidA at ActiveState.com>
>

> Magnus Lie Hetland wrote:
[...]
> 
> I don't know much about R*, but I did consider doing A* in Python, and
> gave it up because of the lousy performance of that kind of algorithm. 
> Are A* and R* relatives?

If you mean A* as in the heuristic search method -- no. (Or, at least,
they have to be extremely distant relatives.) If there is another A*,
then maybe.

R* is a version of the R-tree for spatial indexing. And I'm not planning
on implementing it in Python -- but using existing C-implementations
*in* Python, tying it to other things (that will also probably be
done in C...)

> In general, my conclusion after doing numerical work for a while is that
> the desire to look at algorithms crucial to your research as black boxes
> is futile.

Depends on how central they are :) To me, the R*-tree is just a
storage/retrieval mechanism with certain desirable properties.
Just like I don't generally care about how regular expression
matching or hash tables are implemented, as long as they work
as expected, I'm happy ;)

> In the end, I always had to dig into the details of the
> algorithms because they were either never black-boxable or
> the black-box versions didn't do a good enough job.

Well, well. Performance fine-tuning etc. can come later in my case. I'm
mainly want to experiment a bit.

> I wish you another conclusion, though. =)

Thanks :)

> -- David Ascher
>    ActiveState

--

  Magnus Lie Hetland         http://www.hetland.org

 "Reality is that which, when you stop believing in
  it, doesn't go away."           -- Philip K. Dick