Dan Baston

On the other hand, if we're only ending up in the old overlay code because the new overlay code failed, does it make sense to just ... not call the...

Would it make sense for the algorithm to check for holes outside shells and drop them in a postprocessing step?

> You're currently using the OGR bindings or Fiona to load a GeoJSON FeatureCollection to Python If the OGR bindings are available, I'm exporting the geometry as WKB https://github.com/isciences/exactextract/blob/53309e46b45437447c9e3bb85e03e5a5f97d6380/python/src/exactextract/feature.py#L15-L16 which...

Are you installing the package? It should work to run `pip install .` from the root of the repository.

I updated this so that the tests are still symlinked on Linux/Mac, so I wasn't planning more work here. It can be merged or closed if we would prefer to...

> Perhaps https://gdal.org/development/testing.html#running-a-subset-of-tests-using-pytest could get some updates to explain the subtelty on Windows of doing "python -m pytest -c autotest/pytest.ini ../autotest/gcore/vrt_read.py" Good idea, I've done this at https://github.com/OSGeo/gdal/pull/9224/files#diff-1c899043c9c79440030b3ee7fcf72b357d3333d5070807d933f03ce80b4861edR51

Remove `bIsPostGIS1` argument from `OGRPGLayer::BYTEAToGeometry` ?

> The proposed GEOSGrid_create has xmin, xmax and dx as floats, which is potentially problematic since dx may not be a multiple of (xmax - ymin). Yes, there is a...

I'd hope for it to be used by other GEOS users, otherwise there is no benefit to putting it into the library.

Updated to support multipart polygons and holes. It takes about 5 seconds to subdivide the GADM level 0 boundaries (256 polygons, 32.8 million vertices) against 250,000 grid cell polygons: ...