COMPAT: geopandas 1.0 compatibility fix in h3fy

[martinfleis]

GeoPandas now preserves the geometry column name if it has one. As a result, we end up with two columns called hex_id in the first step and a geometry assigned as an index in the second.

Fixed.

@knaaptime @jGaboardi we shall expand CI to test against nightly here. I caught it coincidentally when testing my teaching material against 1.0 alpha.

Also, this will need a patch release fairly soon.

[martinfleis]

I future, we may want to be able to pass kwargs here geopandas/geopandas#3151

[martinfleis]

@jGaboardi Should I try to include testing against nightly here or can I leave it out from this PR? There's a lot of dependencies in here and I'm not sure which we want to include as nightly.

[jGaboardi]

@jGaboardi Should I try to include testing against nightly here or can I leave it out from this PR? There's a lot of dependencies in here and I'm not sure which we want to include as nightly.

I'd say we should do that in another PR.

[knaaptime]

interesting, thanks. Wasnt aware of any of these gpd changes. Looks like we get a nice speedup with unary union? that'll be great, so probably need this fix in several other places :)

[martinfleis]

Looks like we get a nice speedup with unary union? that'll be great, so probably need this fix in several other places :)

Not by changing unary_union to union_all(), that is calling the same shapely function. But once geopandas/geopandas#3151 lands, we may be able to use coverage union which is considerably faster if you know you have a valid coverage. Which we do here.

And yes, this patch will need to happen everywhere where unary_union is used as it is deprecated.