All, I am pleased to announce the release of Rtree 0.5.0. Rtree 0.5.0 is the first release of a complete refactoring of the library, and it requires libspatialindex 1.4.0 or greater for operation. The new code represents a significant leap forward in capabilities, flexibility, and performance. Although my opinion is that I have tested it as well as the 0.4 code was tested (including full valgrinding to clean up leaks in both Rtree and libspatialindex), you should give things a good test before putting it in production. The new code should have API parity with the old in addition to the new features and performance.
The highlights include: - ability to bulk load an index from a generator at instantiation time - multidimensional indexes (2, 3, 4, ..., kD). - ability to store pickles when inserting index items and retrieve them in queries - custom index/dat filenames! - tunable index performance with index properties (see libspatialindex documentation) - quick index bounds query - fetchable index item bounds (when using objects=True in your nearest/ intersection calls) - point storage where applicable (minx==max && miny == maxy && ... && mink == maxk) <http://pypi.python.org/pypi/Rtree/> This code represents where I wanted to see the library go when I started working on it with Sean. The index properties stuff will allow a tweaker to play with the indexing parameters to their heart's content, and the snazzy stuff like the bulk loading generator stuff will allow a Python user to take advantage of advanced capabilities of the library with ease. There's still plenty of work left to do, but this is a good hop down the road. Howard PS, of course I find an issue with libspatialindex on msvc 2003 when doing the final packaging. No windows builds for 2.4/2.5 right now, but I hope to have them up soon. 2.6 is available, however. _______________________________________________ Community mailing list [email protected] http://lists.gispython.org/mailman/listinfo/community
