On Nov 25, 2017, at 3:35 PM, Matthew Rocklin <mrocklin@gmail.com> wrote:

Thoughts on basing this on a more generic Array type rather than the np.ndarray? 

This would actually be more consistent with the current python typing approach.

I can imagine other nd-array libraries (XArray, Tensorflow, Dask.array) wanting to reuse this work. 

It may be tough to come up with the right ABC though— see another recent thread on this list.

-CHB