Re: Red Black Tree implementation?
On 12/05/13 03:02, duncan smith wrote: On 12/05/13 02:29, Dan Stromberg wrote: On Sat, May 11, 2013 at 4:24 PM, Dan Stromberg drsali...@gmail.com mailto:drsali...@gmail.com wrote: [snip] What should BinaryTree.find() do if it finds a data.node that is None? A call to find(data) should find and return either a node containing data; or the sentinel node where data should be added. It should not get as far as the left or right child of a sentinel node (which would equal None). I'll look at this tomorrow. I did have the truth value of a node depending on it's data value (None implying False). Then I considered the possibility of actually wanting None as a value in the tree and changed it, so I could have introduced a bug here. It's a Python3 thing. The initial sentinel node was evaluating to True. __nonzero__ needs to be changed to __bool__. Thanks! PS: Is it about time we moved this discussion off python-list? Let's do that from now. Duncan -- http://mail.python.org/mailman/listinfo/python-list
Re: Red Black Tree implementation?
I'm afraid I'm having some trouble with the module. I've checked it into my SVN at http://stromberg.dnsalias.org/svn/red-black-tree-mod/trunk/duncan I have two versions of your tests in there now - t is minimally changed, and test-red_black_tree_mod is pretty restructured to facilitate adding more tests later. I get the same problem with either version of the tests. The problem I'm seeing is that the tree, when built from items, isn't looking quite right. I inserted a print(tree) into the for loop, and I'm getting the following, where I expected the tree to grow by one element on each iteration: $ python t 6 False None None 6 False 3 None 6 False 3 15 6 False 3 15 6 False 3 11 6 False 3 11 6 False 3 11 11 False 6 15 11 False 6 15 11 False 6 15 11 False 6 15 11 False 6 15 11 False 6 15 11 False 6 15 11 False 6 15 11 False 6 15 11 False 6 15 11 False 6 15 11 False 6 15 11 False 6 15 Thoughts? BTW, printing an empty tree seems to say sentinel. 'not sure if that was intended. Thanks! On Thu, May 9, 2013 at 6:52 AM, duncan smith buzzard@invalid.invalidwrote: On 09/05/13 02:40, Dan Stromberg wrote: OK, I've got one copy of trees.py with md5 211f80c0fe7fb9cb42feb9645b4b3f**fe. You seem to be saying I should have two though, but I don't know that I do... [snip] Yes, 211f80c0fe7fb9cb42feb9645b4b3f**fe is the correct checksum for the latest version. The previous version had an issue when adding non-distinct items (items that compare equal to items already in the tree). Cheers. Duncan -- http://mail.python.org/**mailman/listinfo/python-listhttp://mail.python.org/mailman/listinfo/python-list -- http://mail.python.org/mailman/listinfo/python-list
Re: Red Black Tree implementation?
On Sat, May 11, 2013 at 4:24 PM, Dan Stromberg drsali...@gmail.com wrote: I'm afraid I'm having some trouble with the module. I've checked it into my SVN at http://stromberg.dnsalias.org/svn/red-black-tree-mod/trunk/duncan I have two versions of your tests in there now - t is minimally changed, and test-red_black_tree_mod is pretty restructured to facilitate adding more tests later. I get the same problem with either version of the tests. The problem I'm seeing is that the tree, when built from items, isn't looking quite right. I inserted a print(tree) into the for loop, and I'm getting the following, where I expected the tree to grow by one element on each iteration: $ python t 6 False None None 6 False 3 None 6 False 3 15 6 False 3 15 I figured out that this was printing a single node and some of its attributes, not an entire tree. I changed it to print an entire tree using self.in_order(). I've also changed around the comparisons a bit, to use a __cmp__ method but still provide __eq__, __neq__ and a new __lt__. I'm up against a new problem now that it'd be nice if you could look at: In BinaryTree.find(), it sometimes compares the item being searched for against None. In 2.x, this gives strange results, but may be benign in this code. In 3.x, this raises an exception. I've added a comment about this in the SVN repo I mentioned above. You can see the traceback yourself with python3 test-red_black_tree_mod . What should BinaryTree.find() do if it finds a data.node that is None? Thanks! PS: Is it about time we moved this discussion off python-list? -- http://mail.python.org/mailman/listinfo/python-list
Re: Red Black Tree implementation?
On 12/05/13 00:24, Dan Stromberg wrote: I'm afraid I'm having some trouble with the module. I've checked it into my SVN at http://stromberg.dnsalias.org/svn/red-black-tree-mod/trunk/duncan I have two versions of your tests in there now - t is minimally changed, and test-red_black_tree_mod is pretty restructured to facilitate adding more tests later. I get the same problem with either version of the tests. The problem I'm seeing is that the tree, when built from items, isn't looking quite right. I inserted a print(tree) into the for loop, and I'm getting the following, where I expected the tree to grow by one element on each iteration: $ python t 6 False None None 6 False 3 None 6 False 3 15 6 False 3 15 6 False 3 11 6 False 3 11 6 False 3 11 11 False 6 15 11 False 6 15 11 False 6 15 11 False 6 15 11 False 6 15 11 False 6 15 11 False 6 15 11 False 6 15 11 False 6 15 11 False 6 15 11 False 6 15 11 False 6 15 11 False 6 15 Thoughts? BTW, printing an empty tree seems to say sentinel. 'not sure if that was intended. Thanks! The leaf node has parent equal to None. All tree nodes have two children. One or both children may be sentinels, and a sentinel is signified by having both left and right (children) equal to None. So an empty tree is a sentinel node that is also root. So the string sentinel is expected (although possibly not the most sensible option). For non-sentinel nodes the string is generated by, return '%s %s %s' % (self.data, self.left.data, self.right.data) for the BinaryTree class, and by return '%s %s %s %s' % (self.data, self.is_red, self.left.data, self.right.data) for the RedBlackTree class. So what is being printed above is (in each case) the value contained in the root node, followed by its colour (True if red), and the values contained in the root node's left and right children. The root node remains root, although it's value and its children (and their values) might change due to tree rotations. It looks OK to me. The empty tree would print sentinel. After adding the value 6 there is one tree node with sentinels as children (values equal to None). Adding 3 results in 3 being the value of the root's left child. It's right child is still a sentinel. Adding 15 results in that value being assigned to the right child. Adding 9 results in no change to the values in the root or its children. Adding 11 results in a tree rotation and 11 becomes the value in the right child of the root. At a later point a tree rotation results in the value of the root node being changed. I haven't implemented a way of representing the structure of the whole red black tree. I would probably write some code to generate a dot file and use that to generate a png. But you could add something like, print tree.height, tree.size, list(tree) and get output like, 0 1 [6] 1 2 [3, 6] 1 3 [3, 6, 15] 2 4 [3, 6, 9, 15] 3 5 [3, 6, 9, 11, 15] 4 6 [3, 6, 9, 11, 12, 15] 4 7 [3, 6, 9, 11, 12, 15, 16] 5 8 [3, 6, 9, 11, 12, 14, 15, 16] 5 9 [3, 6, 9, 11, 12, 14, 15, 16, 17] 5 10 [3, 6, 7, 9, 11, 12, 14, 15, 16, 17] 5 11 [3, 6, 7, 9, 11, 12, 14, 15, 16, 17, 18] 5 12 [3, 5, 6, 7, 9, 11, 12, 14, 15, 16, 17, 18] 5 13 [3, 5, 6, 7, 8, 9, 11, 12, 14, 15, 16, 17, 18] 6 14 [3, 5, 6, 7, 8, 9, 11, 12, 13, 14, 15, 16, 17, 18] 6 15 [0, 3, 5, 6, 7, 8, 9, 11, 12, 13, 14, 15, 16, 17, 18] 6 16 [0, 2, 3, 5, 6, 7, 8, 9, 11, 12, 13, 14, 15, 16, 17, 18] 6 17 [0, 2, 3, 4, 5, 6, 7, 8, 9, 11, 12, 13, 14, 15, 16, 17, 18] 6 18 [-1, 0, 2, 3, 4, 5, 6, 7, 8, 9, 11, 12, 13, 14, 15, 16, 17, 18] 6 19 [-1, 0, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18] It doesn't give you the structure, but it does show that it seems to be growing reasonably. Cheers. Duncan -- http://mail.python.org/mailman/listinfo/python-list
Re: Red Black Tree implementation?
On 12/05/13 02:29, Dan Stromberg wrote: On Sat, May 11, 2013 at 4:24 PM, Dan Stromberg drsali...@gmail.com mailto:drsali...@gmail.com wrote: I'm afraid I'm having some trouble with the module. I've checked it into my SVN at http://stromberg.dnsalias.org/svn/red-black-tree-mod/trunk/duncan I have two versions of your tests in there now - t is minimally changed, and test-red_black_tree_mod is pretty restructured to facilitate adding more tests later. I get the same problem with either version of the tests. The problem I'm seeing is that the tree, when built from items, isn't looking quite right. I inserted a print(tree) into the for loop, and I'm getting the following, where I expected the tree to grow by one element on each iteration: $ python t 6 False None None 6 False 3 None 6 False 3 15 6 False 3 15 I figured out that this was printing a single node and some of its attributes, not an entire tree. I changed it to print an entire tree using self.in_order(). Yes, I've just posted regarding that. I've also changed around the comparisons a bit, to use a __cmp__ method but still provide __eq__, __neq__ and a new __lt__. I have implemented a lot (maybe all?) of the set methods in a subclass. I should probably root that out and have a think about what should be in the RedBlackTree class and what subclasses might look like. I'm up against a new problem now that it'd be nice if you could look at: In BinaryTree.find(), it sometimes compares the item being searched for against None. In 2.x, this gives strange results, but may be benign in this code. In 3.x, this raises an exception. I've added a comment about this in the SVN repo I mentioned above. You can see the traceback yourself with python3 test-red_black_tree_mod . What should BinaryTree.find() do if it finds a data.node that is None? A call to find(data) should find and return either a node containing data; or the sentinel node where data should be added. It should not get as far as the left or right child of a sentinel node (which would equal None). I'll look at this tomorrow. I did have the truth value of a node depending on it's data value (None implying False). Then I considered the possibility of actually wanting None as a value in the tree and changed it, so I could have introduced a bug here. Thanks! PS: Is it about time we moved this discussion off python-list? Maybe. You have my official e-mail address. Cheers. Duncan -- http://mail.python.org/mailman/listinfo/python-list
Re: Red Black Tree implementation?
On 09/05/13 02:40, Dan Stromberg wrote: OK, I've got one copy of trees.py with md5 211f80c0fe7fb9cb42feb9645b4b3ffe. You seem to be saying I should have two though, but I don't know that I do... [snip] Yes, 211f80c0fe7fb9cb42feb9645b4b3ffe is the correct checksum for the latest version. The previous version had an issue when adding non-distinct items (items that compare equal to items already in the tree). Cheers. Duncan -- http://mail.python.org/mailman/listinfo/python-list
Re: Red Black Tree implementation?
On 07/05/13 02:20, Dan Stromberg wrote: On Mon, May 6, 2013 at 5:55 PM, duncan smith buzzard@invalid.invalid mailto:buzzard@invalid.invalid wrote: [snip] I'd prefer Apache or MIT or BSD 3-clause, but I could probably work with this. http://joinup.ec.europa.eu/community/eupl/news/licence-proliferation-way-out I'm eager to see the code, and would love it if you sorted out the deletion rebalance issue. I just plunked some time into https://github.com/headius/redblack/blob/master/red_black_tree.py , only to find that it didn't appear to be doing deletions correctly - the tree would become unprintable after deleting one element. It's possible I introduced the bug, but right now I don't particularly suspect so, having not changed the __del__ method. I'm starting to think Red Black Trees are pretty complex. Mine is fixed now (sent to your gmail address). Restoring the tree properties after deletion is awkward to get right, and doesn't affect the performance noticeably for smallish trees if you get it wrong. I realised my code was buggy when I tried adding, then removing a million items and ran into the recursion limit. It now passes a test where I check the tree properties after each addition / deletion. Duncan -- http://mail.python.org/mailman/listinfo/python-list
Re: Red Black Tree implementation?
OK, I've got one copy of trees.py with md5 211f80c0fe7fb9cb42feb9645b4b3ffe. You seem to be saying I should have two though, but I don't know that I do... On 5/8/13, duncan smith buzzard@invalid.invalid wrote: On 07/05/13 02:20, Dan Stromberg wrote: On Mon, May 6, 2013 at 5:55 PM, duncan smith buzzard@invalid.invalid mailto:buzzard@invalid.invalid wrote: [snip] I'd prefer Apache or MIT or BSD 3-clause, but I could probably work with this. http://joinup.ec.europa.eu/community/eupl/news/licence-proliferation-way-out I'm eager to see the code, and would love it if you sorted out the deletion rebalance issue. I just plunked some time into https://github.com/headius/redblack/blob/master/red_black_tree.py , only to find that it didn't appear to be doing deletions correctly - the tree would become unprintable after deleting one element. It's possible I introduced the bug, but right now I don't particularly suspect so, having not changed the __del__ method. I'm starting to think Red Black Trees are pretty complex. Mine is fixed now (sent to your gmail address). Restoring the tree properties after deletion is awkward to get right, and doesn't affect the performance noticeably for smallish trees if you get it wrong. I realised my code was buggy when I tried adding, then removing a million items and ran into the recursion limit. It now passes a test where I check the tree properties after each addition / deletion. Duncan -- http://mail.python.org/mailman/listinfo/python-list -- http://mail.python.org/mailman/listinfo/python-list
Re: Red Black Tree implementation?
On 09/05/13 02:40, Dan Stromberg wrote: OK, I've got one copy of trees.py with md5 211f80c0fe7fb9cb42feb9645b4b3ffe. You seem to be saying I should have two though, but I don't know that I do... I've just re-sent it. Duncan -- http://mail.python.org/mailman/listinfo/python-list
Re: Red Black Tree implementation?
On 07/05/13 02:20, Dan Stromberg wrote: [snip] I'm starting to think Red Black Trees are pretty complex. A while ago I looked at a few different types of self-balancing binary tree. Most look much easier to implement. BTW, the licence might be MIT - I just copied it from someone else's code. Duncan -- http://mail.python.org/mailman/listinfo/python-list
Re: Red Black Tree implementation?
On 03/05/13 03:00, Dan Stromberg wrote: On Wed, May 1, 2013 at 7:06 PM, duncan smith buzzard@invalid.invalid mailto:buzzard@invalid.invalid wrote: I have an implementation that you can try out. It's not based on any other implementation, so my bugs will be independent of any bugs in the code you're currently using. It looks more like a set - add, remove, discard. Not tried on Python 3 or run through pylint. I just tried adding a million items to a tree, and it takes about 25% longer to add items at the end compared to those at the beginning. Timing removals uncovered a bug. So if you want the code I'll fix the bug and send it (to your gmail e-mail address?). Cheers. Duncan -- http://mail.python.org/__mailman/listinfo/python-list http://mail.python.org/mailman/listinfo/python-list What license? Thanks! Here's the text I usually prepend. ##Copyright (c) 2013 duncan g. smith ## ##Permission is hereby granted, free of charge, to any person obtaining a ##copy of this software and associated documentation files (the Software), ##to deal in the Software without restriction, including without limitation ##the rights to use, copy, modify, merge, publish, distribute, sublicense, ##and/or sell copies of the Software, and to permit persons to whom the ##Software is furnished to do so, subject to the following conditions: ## ##The above copyright notice and this permission notice shall be included ##in all copies or substantial portions of the Software. ## ##THE SOFTWARE IS PROVIDED AS IS, WITHOUT WARRANTY OF ANY KIND, EXPRESS ##OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, ##FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL ##THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR ##OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ##ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR ##OTHER DEALINGS IN THE SOFTWARE. Basically, do what you want with it but don't blame me if it goes tits up. I'm happy to consider tidying it up a bit and using a more recognized form of licence. Just had a bank holiday here, so bug not yet squashed. But it is the sort of bug that might account for what you've seen (if a similar bug exists in the code you've been using). The tree doesn't always get properly rebalanced on node removals. I'll attack the problem later tomorrow (technically, later today). Cheers. Duncan -- http://mail.python.org/mailman/listinfo/python-list
Re: Red Black Tree implementation?
On Mon, May 6, 2013 at 5:55 PM, duncan smith buzzard@invalid.invalidwrote: What license? Thanks! Here's the text I usually prepend. ##Copyright (c) 2013 duncan g. smith ## ##Permission is hereby granted, free of charge, to any person obtaining a ##copy of this software and associated documentation files (the Software), ##to deal in the Software without restriction, including without limitation ##the rights to use, copy, modify, merge, publish, distribute, sublicense, ##and/or sell copies of the Software, and to permit persons to whom the ##Software is furnished to do so, subject to the following conditions: ## ##The above copyright notice and this permission notice shall be included ##in all copies or substantial portions of the Software. ## ##THE SOFTWARE IS PROVIDED AS IS, WITHOUT WARRANTY OF ANY KIND, EXPRESS ##OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, ##FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL ##THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR ##OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ##ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR ##OTHER DEALINGS IN THE SOFTWARE. Basically, do what you want with it but don't blame me if it goes tits up. I'm happy to consider tidying it up a bit and using a more recognized form of licence. Just had a bank holiday here, so bug not yet squashed. But it is the sort of bug that might account for what you've seen (if a similar bug exists in the code you've been using). The tree doesn't always get properly rebalanced on node removals. I'll attack the problem later tomorrow (technically, later today). Cheers. I'd prefer Apache or MIT or BSD 3-clause, but I could probably work with this. http://joinup.ec.europa.eu/community/eupl/news/licence-proliferation-way-out I'm eager to see the code, and would love it if you sorted out the deletion rebalance issue. I just plunked some time into https://github.com/headius/redblack/blob/master/red_black_tree.py , only to find that it didn't appear to be doing deletions correctly - the tree would become unprintable after deleting one element. It's possible I introduced the bug, but right now I don't particularly suspect so, having not changed the __del__ method. I'm starting to think Red Black Trees are pretty complex. -- http://mail.python.org/mailman/listinfo/python-list
Re: Red Black Tree implementation?
On Tue, May 7, 2013 at 10:55 AM, duncan smith buzzard@invalid.invalid wrote: Here's the text I usually prepend. ##Copyright (c) 2013 duncan g. smith ## ##Permission is hereby granted, free of charge, to any person obtaining a ##copy of this software and associated documentation files (the Software), ##to deal in the Software without restriction, including without limitation ##the rights to use, copy, modify, merge, publish, distribute, sublicense, ##and/or sell copies of the Software, and to permit persons to whom the ##Software is furnished to do so, subject to the following conditions: ## ##The above copyright notice and this permission notice shall be included ##in all copies or substantial portions of the Software. ## ##THE SOFTWARE IS PROVIDED AS IS, WITHOUT WARRANTY OF ANY KIND, EXPRESS ##OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, ##FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL ##THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR ##OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ##ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR ##OTHER DEALINGS IN THE SOFTWARE. Basically, do what you want with it but don't blame me if it goes tits up. I'm happy to consider tidying it up a bit and using a more recognized form of licence. Is that the MIT license? If not, consider using it; it's well known and trusted. I haven't eyeballed yours closely but it looks extremely similar, at least. ChrisA -- http://mail.python.org/mailman/listinfo/python-list
Re: Red Black Tree implementation?
Am 02.05.2013 01:11, schrieb Dan Stromberg: What's the best Red Black Tree implementation for Python with an opensource license? I started out looking at http://newcenturycomputers.net/projects/rbtree.html http://newcenturycomputers.net/projects/rbtree..html because it was pretty high in Google and had the operators I wanted, but it gets very slow at about half a million elements. I've been discussing this with a C programmer who believes that Red Black Trees should perform very similarly to an AVL tree, but that's not at all what I'm getting with the newcenturycomputers implementation. 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. Christian -- http://mail.python.org/mailman/listinfo/python-list
Re: Red Black Tree implementation?
On Thu, May 2, 2013 at 3:46 AM, Christian Heimes christ...@python.orgwrote: 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
Re: Red Black Tree implementation?
On Wed, May 1, 2013 at 7:06 PM, duncan smith buzzard@invalid.invalidwrote: I have an implementation that you can try out. It's not based on any other implementation, so my bugs will be independent of any bugs in the code you're currently using. It looks more like a set - add, remove, discard. Not tried on Python 3 or run through pylint. I just tried adding a million items to a tree, and it takes about 25% longer to add items at the end compared to those at the beginning. Timing removals uncovered a bug. So if you want the code I'll fix the bug and send it (to your gmail e-mail address?). Cheers. Duncan -- http://mail.python.org/**mailman/listinfo/python-listhttp://mail.python.org/mailman/listinfo/python-list What license? Thanks! -- http://mail.python.org/mailman/listinfo/python-list
Red Black Tree implementation?
What's the best Red Black Tree implementation for Python with an opensource license? I started out looking at http://newcenturycomputers.net/projects/rbtree.htmlbecause it was pretty high in Google and had the operators I wanted, but it gets very slow at about half a million elements. I've been discussing this with a C programmer who believes that Red Black Trees should perform very similarly to an AVL tree, but that's not at all what I'm getting with the newcenturycomputers implementation. I'd prefer something that looks like a dictionary, runs on 2.x and 3.x, and passes pylint, but if that's not yet available I might make it so. This is part of a comparison of Python tree types I did a while back... I've been thinking that I've given Red Black Trees short shrift by using a poor implementation. The comparison so far is at http://stromberg.dnsalias.org/~strombrg/python-tree-and-heap-comparison/ Thanks! -- http://mail.python.org/mailman/listinfo/python-list
Re: Red Black Tree implementation?
On 02/05/13 00:11, Dan Stromberg wrote: What's the best Red Black Tree implementation for Python with an opensource license? I started out looking at http://newcenturycomputers.net/projects/rbtree.html because it was pretty high in Google and had the operators I wanted, but it gets very slow at about half a million elements. I've been discussing this with a C programmer who believes that Red Black Trees should perform very similarly to an AVL tree, but that's not at all what I'm getting with the newcenturycomputers implementation. I'd prefer something that looks like a dictionary, runs on 2.x and 3.x, and passes pylint, but if that's not yet available I might make it so. This is part of a comparison of Python tree types I did a while back... I've been thinking that I've given Red Black Trees short shrift by using a poor implementation. The comparison so far is at http://stromberg.dnsalias.org/~strombrg/python-tree-and-heap-comparison/ Thanks! I have an implementation that you can try out. It's not based on any other implementation, so my bugs will be independent of any bugs in the code you're currently using. It looks more like a set - add, remove, discard. Not tried on Python 3 or run through pylint. I just tried adding a million items to a tree, and it takes about 25% longer to add items at the end compared to those at the beginning. Timing removals uncovered a bug. So if you want the code I'll fix the bug and send it (to your gmail e-mail address?). Cheers. Duncan -- http://mail.python.org/mailman/listinfo/python-list