ENH: stats.skew: add array-API support #20541
Conversation
There was a problem hiding this comment.
agreed, beat me to it @tupui :) just a small typo and one thought about the testing
| @@ -1121,6 +1121,8 @@ def _var(x, axis=0, ddof=0, mean=None): | |||
| @_axis_nan_policy_factory( | |||
| lambda x: x, result_to_tuple=lambda x: (x,), n_outputs=1 | |||
| ) | |||
| # nan_policy handled by `_axis_nan_policy, but needs to be left | |||
There was a problem hiding this comment.
| # nan_policy handled by `_axis_nan_policy, but needs to be left | |
| # nan_policy handled by `_axis_nan_policy`, but needs to be left |
There was a problem hiding this comment.
Oops I'll fix this next time.
| def test_empty_1d(self): | ||
| # This is not essential behavior to maintain w/ array API |
There was a problem hiding this comment.
perhaps more clear to @skip_xp_backends(np_only=True), although I don't have a strong opinion. Right now, not parameterising with xp very clearly means 'this is just for NumPy'. In the future, when we potentially have the majority of tests converted, that might not be so clear. But happy with this for now, can be left for the future.
There was a problem hiding this comment.
I will consider tacking this on to my next PR. But honestly this test can probably just be removed. We've talked about silencing this warning throughout stats.
Reference issue
Follow-up to gh-20292
What does this implement/fix?
This adds array-API support to
scipy.stats.skew. It might be useful to review one commit at a time:nan_policysupport is handled by the_axis_nan_policydecorator)_axis_nan_policydecorator raise when it gets arguments that won't currently work with the array API. It only works when these parameters are passed as keyword arguments (which they should be anyway), but I am not too concerned because a) array-API support is still experimental and behind an environment variable and b) these features should be available with non-NumPy backends by the time the environment variable is removed, at which point this error will be removed. Nothing like this commit will be needed with other array-API conversions.Additional information
After this, I will open an issue with a list of stats functions I think are ripe for array API support. We can point to this PR as an example and hopefully crowdsource the effort. I'd be willing to review PRs for functions on that list.