Refactor functions in join.py to accept GeoSeries Input
#948
Conversation
Co-authored-by: H. Thomson Comer <thomcom@gmail.com>
…into feature/geoseries_interface/join
…ature/geoseries_interface/join
…into feature/geoseries_interface/join
isVoid
added
improvement
| @cuda.jit | ||
| def binarize(in_col, out, width): |
There was a problem hiding this comment.
Whoah. Shoud we move this stuff to libcuspatial?
There was a problem hiding this comment.
Does this benefit libcuSpatial? I don't see a usage of this other than converting the result of Point in Polygon to dataframe.
There was a problem hiding this comment.
Doing it faster benefits cuSpatial users. Generally Numba and Python add a lot of overhead.
There was a problem hiding this comment.
Can probably punt to a future optimization. Another question is whether we should change the point in polygon interface, or simply leave this as a cuda kernel bind to python.
This might also have bigger question to do with the PiP interface: currently we use bits to encode the output - this is very efficient in memory but eventually we still need to convert them back to int8 arrays as shown here. If no one needs the memory saving benefit, should we keep it? Perhaps we should change the interface to return boolean array directly?
|
/merge |
Description
This PR refactors all functions in
join.pyto accept GeoSeries as input.closes #940, closes #939, closes #941, closes #942
Checklist