Benoit Bovy

Closing. Superseded by #51.

Thanks for looking into this @jorisvandenbossche. I agree it would be nice to have an option to get coordinates in a given projection with support for tessellation.

@brendan-ward thanks for opening this issue. Nanobind has been briefly discussed in various places but it is better to gather the discussion here. Also good to know that the transition...

One thing that currently prevents us using nanobind here is that [it doesn't supports arbitrary Python objects as numpy array items](https://nanobind.readthedocs.io/en/latest/ndarray.html#limitations-related-to-dtypes).

I'm closing this issue since nanobind doesn't seem to have a plan for adding support of numpy arrays with arbitrary Python objects anytime soon. Let's re-open it if that changes.

@jorisvandenbossche I think we can close this now?

There are more things to consider here as we are dealing with Numpy `object` arrays. At runtime we need to cast Python object pointers to C++ Geography pointers (and raise...

Thanks @jorisvandenbossche, that's very helpful. The last point (vectorized functions that create new geometries as output) may be tricky to implement around `pybind11::vectorize`. I'm afraid we would need to add...

Xref https://github.com/pybind/pybind11/issues/1042 in case we would need to wrap a vector of new PyObject geometries into a new object dtype numpy array without copy. Not sure we would really need...

Tracked this down in pybind11's `type_caster_generic::cast`. The bottleneck seems that pybind11 keeps a map of all registered instances and first tries to find and return the corresponding instance from the...