The process of matching points to polygons is computationally expensive and can be slow. H3 has a snappy C backend, but my other constructors use geopandas/shapely's within function that can be slow. For the Chicago traffic accident dataset, matching 200K records to 80 census tracts takes 10 minutes.
The process of matching points to polygons is computationally expensive and can be slow. H3 has a snappy C backend, but my other constructors use geopandas/shapely's
withinfunction that can be slow. For the Chicago traffic accident dataset, matching 200K records to 80 census tracts takes 10 minutes.This is an open issue in geopandas: https://github.com/geopandas/geopandas/issues/430 They are currently (as of October 2019) working to integrate the optimized pygeos library into geopandas as per https://github.com/geopandas/geopandas/issues/1155.
At this time I do not plan to directly integrating pygeos into porygon, instead hoping that the integration into geopandas goes smoothly.