The mission of RBush is to provide a performant, thread-safe RTree implementation for Python which
does not have libspatialindex or geos as a dependencies.
conda install -c parietal.io rbushPython port of JS rbush library
To run the (unit) tests, type:
$ pytest rbushor
>>> import rbush
>>> rbush.test()As in JS rbush,
the following sample performance test was done by generating
random uniformly distributed rectangles of ~0.01% area and setting maxEntries to 16
(see benchmark.py script).
Performed with Python 3.6.1 on a Intel-i7 1.9GHz.
| Test | RBush(naive) | RBush(numba,v0.0.1) | RBush(numpy,v0.0.2) |
|---|---|---|---|
| insert 1M items one by one | 224.9s | 834.8s | 251.4s |
| 1000 searches of 0.01% area | 5.30s | 79.1s | 7.5s |
| 1000 searches of 1% area | 13.5s | 277.6s | 8.1s |
| 1000 searches of 10% area | 47.6s | 935.7s | 16.5s |
| remove 1000 items one by one | 2.61s | 9.56s | 7.3s |
| bulk-insert 1M items | 63.7s | 101.1s | 25.1s |
insert accepts arrays for xmin, xmax, ymin, ymaxload receives a numpy array