On Thu, Apr 26, 2012 at 6:08 PM, Sylvain Bougerel < sylvain.bougerel.devel@gmail.com> wrote:
On Mar 9, 2012 6:14 AM, "Dan Walters" <dan683@ <dan683@gmail.com>gmail.com<dan683@gmail.com>> wrote:
Hi all,
I am proposing adding spacial containers for efficiently storing and accessing data of a spacial nature in fixed size trees.
Hi all,
This thread is long over, but I've been meaning to write something on it for so long.
I wrote a kd-tree library for generic purpose. It may not sound like much, but this library has (i believe) truly unique features : first of all, erasing really works in amortized time (on most libraries out there it does not -- and that's when they even propose erasing), secondly the kd-tree is the only one i know that is capable of self-balancing.
These 2 features combined with the fact that the library is very template based, and support any number of dimensions (that can even be determined at runtime) makes it really a general purpose library.
It as nearest neighbor iterators, region (or bisection) iterators, and a not so common but useful operation: a mapping iterator which maps all n dimensions into 1 effectively making the iterator behave as a usual multiset iterator, on 1 particular chosen dimension.
My future goal is to one day feel satisfied enough by the maturity of the library to submit the library to Boost for integration. Sadly i haven't reached that stage yet.
For those who are interested to try it out, you can pull the origin/master branch from the library (the release package is stable but the code is a bit old now) . It had burgeoning (and therefore incomplete) documentation with some obvious mistakes.
http://spatial.sourceforge.net/
And I'm still regularly modifying the interface so it is far from being mature. But one day... It will.
I think this could be really exciting to have in Boost. Unfortunately, I don't think I can offer much feedback at the moment, but definitely keep (at least) me posted on noteworthy developments. I'm particularly interested in what your interface will look like. - Jeff