diff options
| author | Sebastian Berg <sebastian@sipsolutions.net> | 2018-02-16 21:38:23 -0800 |
|---|---|---|
| committer | Eric Wieser <wieser.eric@gmail.com> | 2018-05-25 23:06:43 -0700 |
| commit | 0addc016ba7000b27509663f4f489c6eb1838056 (patch) | |
| tree | 3b1f1cabcdee01e7a625f17ffe8970b878f4f8e3 /numpy/lib/tests/test_index_tricks.py | |
| parent | e441c291b2e10c8de85a9d950d0add552d0ebd83 (diff) | |
| download | numpy-0addc016ba7000b27509663f4f489c6eb1838056.tar.gz | |
DEP: Deprecate non-tuple multidimensional indices
Currently, arr[[None, 0]] and arr[(None, 0)] mean the same thing, yet arr[[0, 0]] and arr[(0, 0)] mean different things. This makes it super hard to make a subclass or duck array that behaves consistently with ndarray.
By deprecating this feature, we force downstream library code to stop using it, which in turn makes that library code use approaches that are easier to implement in subclasses and duck types.
Diffstat (limited to 'numpy/lib/tests/test_index_tricks.py')
0 files changed, 0 insertions, 0 deletions