Joris Van den Bossche

What version of geopandas and fiona are you using?

Hmm, with exactly the same versions (on linux), I see: ``` In [4]: geopandas.read_file("scipy2018-geospatial-data/data/berlin-districts.geojson").to_crs(epsg=3857).head() Out[4]: district district_group median_price geometry 0 Blankenfelde/Niederschönhausen Pankow 37.5 (POLYGON ((1493006.880445722 6912074.798336806... 1 Helmholtzplatz Pankow 58.0...

Personally I am fine with adding it to `requirements-dev.txt` (I don't use that file myself) I am bit curious though, do we actually need those test requirements of pandas? For...

@brendan-ward only seeing your response now. In the meantime I just did some benchmarks on the total_bounds / tree creation in case of no overlap, and opened https://github.com/geopandas/geopandas/issues/2116 with that....

@gjoseph92 thanks a lot for chiming in and for the insights! I have been rerunning a modified version of the notebooks today (with the latest versions of dask / distributed...

> I'd expect an `sjoin` to be significantly faster after running `ddf.calculate_spatial_partitions()` This can _potentially_ speed up the spatial join, but that that depends on the characteristics of the data....

That sounds as a good idea. It could be something like a simple class like, or using dataclasses: ``` from dataclasses import dataclass @dataclass class DriverInfo: name: str ext: str...

@awa5114 sorry for the late reply In general, subclassing GeoDataFrame is not really supported: we (as geopandas project) never really considered this use case or made effort in making this...

Yes, indeed, for non-point geometries that will need to be a list of values. That doesn't make those results very easy to work with (since eg pandas doesn't provide much...

> However, it does save the file, seemingly correctly. It's only failing when writing the common `_metadata` file at the end, so that's the reason it appears (and also does)...