Use __geo_interface__ on GeoJson

[ocefpaf]

We can check if the object has __geo_interface__ and just use that. This would make folium more friendly to many objects that follow: https://gist.github.com/sgillies/2217756

[BibMartin]

That's a nice idea ; and nice to have that common attribute.
But this protocol does not guarantee that data are in the good EPSG projection. We need to know the projection to plot data in folium. For example, see how we handle geopandas in https://github.com/python-visualization/folium/blob/master/folium/features.py#L243-L245

I don't know how we can handle it. We can use __geo_interface__ and let the user be responsible for projection,or have a source_projection kwarg and embed pyproj...

[ocefpaf]

For example, see how we handle geopandas in https://github.com/python-visualization/folium/blob/master/folium/features.py#L243-L245

It was your geopandas implementation that inspired this idea 😉

let the user be responsible for projection

I'd rather leave it for the user. (We could add a kw and convert if an optional dependency, pyproj, is available... But I think that would add an extra complexity layer.)

[pelson]

Just to add a little more info: GeoJSON is epsg:3857 only. This suggests that the __geo_interface__ is epsg:3857 only. From that, I'd say that __geo_interface__ is a very reasonable interface for the GeoJson object to handle.

It is plausible that one-day, that will change and that __geo_interface__ will gain a crs attribute, but that is something that can be handled if/when that occurs IMO.

[ocefpaf]

This suggests that the __geo_interface__ is epsg:3857

My guess is that most __geo_interface__ implementations just outputs whatever crs the data is in, no? That would still force us to convert it before sending to leaflet (or trust the users will do that on their own).