Matt Richards

> > if pygeos is only being used internally by geopandas within a project, switching to shapely2 (or just removing pygeos) should not cause any breaks. > > That is...

As I said in 2060 when this was raised, I think that for users - ending up with empty geodataframes like this is unlikely (as far as I'm aware) as...

@NickLucche I don't have the same background as Joris to comment within the context of the pandas subclassing. I would also just reiterate the point about the use case to...

> So as a quick test, this indeed already works out-of-the-box (without this PR): > > ``` > df = geopandas.read_file(geopandas.datasets.get_path("naturalearth_lowres")) > df.to_file("test.shp.zip", driver="ESRI Shapefile") > ``` > > and...

> Yes, agreed. Especially now that we have better error messages for this, it might be a good time to reconsider that. From my perspective it seems like having `_geometry_column_name`...

Just a note on having the custom indexing code, it seems contrary to the consensus in https://github.com/geopandas/geopandas/issues/498 where there is an arugment to reduce complexity in sjoin by making a...

> break your code (but actually, this PR will also do that ...). That's actually why I originally referenced the other issue, but thought this is in a sense a...

Following the discussion on this at yesterdays meeting just wanted to recap that and try and articulate a direction to get this series of PR's merged (while I don't think...

An initial draft of part 2: #2329

>Should we maybe keep _geometry_column_name unset (None)? (I don't know if this actually would help in practice, it might just make it a bit easier to check if the active...