Michael Wang

cc @VibhuJawa for potential downstream influences.

Hi there, forgive my intrusion on following questions, as this is my first appearance on the repo. What criterion do you adopt to compare between algorithms? Since it's about GPU...

@thomcom do you mind writing up the python benchmark docs for #599 since this PR first introduced the benchmark suite?

Still needed.

Yes, in cpp: https://github.com/rapidsai/cuspatial/blob/8c31c2cad9cc333813848b2e33bb6d0fd884ed6b/cpp/src/spatial/polyline_bounding_box.cu#L57 https://github.com/rapidsai/cuspatial/blob/6a34c21ac13708303cda7c2fb9c49fbb42d04ed5/cpp/include/cuspatial/spatial_join.hpp#L151 in python: https://github.com/rapidsai/cuspatial/blob/92266b6cd88d3283eef45b0a38e712a250f14bdf/python/cuspatial/cuspatial/core/gis.py#L305 https://github.com/rapidsai/cuspatial/blob/92266b6cd88d3283eef45b0a38e712a250f14bdf/python/cuspatial/cuspatial/core/spatial_join.py#L160

`Linestring` isn't the first time showing up in the code base either: https://github.com/rapidsai/cuspatial/blob/92266b6cd88d3283eef45b0a38e712a250f14bdf/python/cuspatial/cuspatial/io/geopandas_adapter.py#L59

Offline summary: this algorithm is a flavor of [sinusoidal projection](https://en.wikipedia.org/wiki/Sinusoidal_projection) and is near Cartesian around equator and prime meridian. This is why all lonlats were normalized around `origin_lon` and `origin_lat`...

> The extensive discussions were mostly on the "bug" of cuspatial.point_in_polygon in Python API. It turns out that the notebook example at https://github.com/rapidsai/cuspatial/blob/branch-22.04/notebooks/nyc_taxi_years_correlation.ipynb used a subset of poly_offsets array but...

This is fixed thanks to recent refactors.

Edited PR description to address the change.