On Thu, May 2, 2013 at 3:46 AM, Christian Heimes <christ...@python.org>wrote:
> Am 02.05.2013 01:11, schrieb Dan Stromberg: > No wonder it's getting slow and doesn't stand a change against Python's > dict implementation. The rbtree implementation from > newcenturycomputers.net is written entirely in Python. It's good code > for academic research in order to study the properties of a rbtree. > > If you need something production ready then you have to use an > implemetation with an optimized backend like a C code, PyPy or Cython. The newcenturycomputers RBTree is similarly slow on CPython 2.x, CPython 3.x, PyPy and Jython. I imagine the tree depth is getting deeper than it should, but I've not verified that. Also, the other datastructures I'm comparing it to: B+ Tree, AVL Tree, Treap, etcetera are pure python too, but they're performing reasonably - in fact, my expectations are pretty much defined by these. (OK, the treap code has a Cython version, but I'm not using that in this project) Thanks for the response.
-- http://mail.python.org/mailman/listinfo/python-list
