Hi Stephen,
If we have to install any of SciPy to get access to spatial, I am in
complete agreement about avoiding it.
The pure python implementation is fine with me, but keep in mind it
won't scale to a huge number of points -- it might only be useful for
large spheres, and we won't be able to use it for things like data
points or particles unless the number considered is very low. Our
long term strategy for such a thing would be to convert the code to
optimized Cython and include that. Alternatively, the kD trees in HOP
and FOF could be wrapped, which all-in-all would not be such a bad
thing.
Including this code is also okay with me. It will get shuffled around
a bit after the reorganization, but for now tossing it into its own
file in yt/lagos/ is fine. I would really appreciate it, however, if
you could clean it up so that it's more in compliance with PEP-8.
Thanks very much, Stephen, keeping us looking forward with geometric
analysis would be great -- and distance measures for halos will be
super useful.
-Matt
On Mon, Jun 15, 2009 at 10:49 PM, Stephen Skory
Hi guys,
I want to add a couple routines to the HaloFinder having to do with halo nearest neighbor stuff. In particular, I'd like to add something like:
halo_list.nearest_neighbor_3D(haloID,N) where it finds the Nth nearest neighbor to halo 'haloID' in three dimensions.
halo_list.nearest_neighbor_2D(haloID,N,[x,y,z]) where it finds the Nth nearest neighbor in two dimensions, projected along vector [x,y,z].
The obvious way to go about this is a kdTree. I know this is something that Matt has put a lot of thought into. I've been playing around with this one http://sites.google.com/site/mikescoderama/Home/kd-tree-knn simply because it's in pure Python, not for any other reason. scipy comes with the spatial package, but it appears that some recent changes have caused its install to become tied to the rabbit hole that is the scipy package. Ideally, the scipy spatial would be better because it's surely faster, but scipy is a hassle.
Before I put time into this, what thoughts do people have? The python kdTree works, and it can probably be included in yt. A kdTree will probably have uses elsewhere in yt, so if I write routines that use it, I'd like the kdTree I use to be useful elsewhere.
Thanks!
_______________________________________________________ sskory@physics.ucsd.edu o__ Stephen Skory http://physics.ucsd.edu/~sskory/ _.>/ _Graduate Student ________________________________(_)_\(_)_______________
_______________________________________________ Yt-dev mailing list Yt-dev@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org