sha,message,author_date,committer_date,raw_author,raw_committer,repo,author,committer
bbd25c65a874dfd505ab9df43f6109f6343e5f4d,"Support for __array_function__ implementers (sparse arrays) [WIP] (#3117)

* Support for __array_function__ implementers

* Pep8

* Consistent naming

* Check for NEP18 enabled and nep18 non-numpy arrays

* Replace .values with .data

* Add initial test for nep18

* Fix linting issues

* Add parameterized tests

* Internal clean-up of isnull() to avoid relying on pandas

This version should be much more compatible out of the box with duck typing.

* Add sparse to ci requirements

* Moar tests

* Two more patches for __array_function__ duck-arrays

* Don't use coords attribute from duck-arrays that aren't derived from DataWithCoords

* Improve checking for coords, and autopep8

* Skip tests if NEP-18 envvar is not set

* flake8

* Update xarray/core/dataarray.py

Co-Authored-By: Stephan Hoyer <shoyer@gmail.com>

* Fix coords parsing

* More tests

* Add align tests

* Replace nep18 tests with more extensive tests on pydata/sparse

* Add xfails for missing np.result_type (fixed by pydata/sparse/pull/261)

* Fix xpasses

* Revert isnull/notnull

* Fix as_like_arrays by coercing dense arrays to COO if any sparse

* Make Variable.load a no-op for non-dask duck arrays

* Add additional method tests

* Fix utils.as_scalar to handle duck arrays with ndim>0",2019-08-05T18:44:43Z,2019-08-05T18:44:43Z,84e620c8a0c57a3b676a20d0134cbc5f75cefe5b,f10b21bed2846b879806f87039b77245b18e7671,13221727,1270651,1217238