wpbonelli
wpbonelli
maybe also consider returning `pd.DataFrame` instead of `np.recarray` as pandas is now a core dependency?
> return pd.DataFrame instead of recarray. This is something of a breaking change since the return type will change, but since the syntax for accessing the columns is still sort...
Ahh. thanks. nearest cell center. I was confusing that with some notion of "nearest cell" which I recognize now is ill-defined absent a projection scheme I guess..
> Columns with indices are set to dtype int where possible. For points_to_cellids the dtype remains float to allow for NaNs (point outside grid) in the result. I don't know...
> On that note, it's probably also a good idea to suppress dataframe warnings everywhere by passing dataframe=False everywhere gridintersect is being used. Seems like a good compromise, pass `False`...
> I'm not actually sure what the policy currently is on the dependencies Neither are we, to my knowledge. I think the (de facto, little discussed) approach has been as...
I think the pattern here is great — iteratively building a dataset from select packages. A similar way to achieve the same thing with xarray could be a `to_dataset()` method...
@aestrad-intera hard to say without seeing your grid. any way you could share it? you might try intersecting shapes separately instead of the multipolygon all at once. as stated in...
do you see any performance difference without `method='structured'`?
Ah, understood. Thanks @christianlangevin