MAINT, TST: NumPy stride deprecation shim #23199
Conversation
| @@ -1295,7 +1295,7 @@ def test_odd_strides(self, xp): | |||
| # us into wrong memory if used (but it does not need to be used) | |||
| dummy = xp.arange(10, dtype=xp.float64) | |||
| a = dummy[5:6] | |||
| a.strides = 16 | |||
| a = xp.lib.stride_tricks.as_strided(a) | |||
There was a problem hiding this comment.
I don't know if using xp here contributes that much--pretty sure the stride trick stuff can't be part of the array API standard anyway.
There was a problem hiding this comment.
I should probably set the value to 16 though...
There was a problem hiding this comment.
xp.asarray(np.stride_tricks.as_strided(np.asarray(a),...)))
There was a problem hiding this comment.
Or make the test numpy-only via @skip_xp_baxkends(np_only=True)
There was a problem hiding this comment.
it is also NumPy-only in the decorator I think, unless you're thinking of farther into the future (although some array API backends still don't propagate the strides do they?)
* This patch allows the SciPy testsuite to pass against NumPy `main` branch, after numpy/numpy#28925 was merged in the last day or so.
1242ade to
3e40b21
Compare
|
note that this patch causes a failure on the maintenance branch when backported: https://github.com/scipy/scipy/actions/runs/15788333092/job/44509424585?pr=23188 The array API stuff and its deps are pretty tricky to handle portably on both branches.. |
|
thanks Tyler |
mainbranch, after DEP: Deprecate setting the strides attribute of a numpy array numpy/numpy#28925 was merged in the last day or so.We shouldn't really have to backport this, since it is slated for NumPy
2.4.xseries, but if it becomes annoying on the maintenance branch we could backport it since it is a test-only shim (though not running dev version tests on maintenance branches is also something we discussed in the past).