issue_comments

Ah, thanks @pletchm , and @shoyer for spotting this

I should have caught this - I didn't separately check the windows tests; I expected the same unrelated failures as on Travis and didn't consider this would be an area that would be different between platforms. Mea culpa. It's always the avenues you least suspect...

@shoyer, I spent sometime looking into it and it looks like the test fails for the shallow copy, and apparently only on Windows for some reason. In Windows coords seem to be immutable unless it's one dataarray deep copied from another (which is why only the deep=False test fails). So I decided to just mark the tests as xfail for now (but I'd be happy to create an issue and look into it more in the future).

I'll open a new PR marking the tests with xfail shortly.

Don't worry about it, this happens. If you can't figure out an immediate fix, please just mark the tests as "xfail" so they don't block other PRs. I'm pretty sure the issue here is in the tests, not your fix.

On Thu, May 9, 2019 at 9:08 AM Martin Pletcher notifications@github.com wrote:

Sure. I'm looking into it now.

Sorry about that.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/pydata/xarray/pull/2936#issuecomment-490967152, or mute the thread https://github.com/notifications/unsubscribe-auth/AAJJFVWYEUWLGSDKNURDUWTPURD6RANCNFSM4HKLQTTQ .

Sure. I'm looking into it now.

Sorry about that.

It seems xarray (mistakenly?) relies on the immutability of index data somewhere else, and hence reuses the same pandas.Index object for data in both data and coords.

Can you take a look at the test failures on Windows? These seem to be due to your charge: ``` ================================== FAILURES =================================== __ TestDataArray.test_copy_coords[False-expected_orig1] _____ self = <xarray.tests.test_dataarray.TestDataArray object at 0x000000AAE0637F28> deep = False expected_orig = <xarray.DataArray 'a' (a: 2)> array([999, 2]) Coordinates: * a (a) int32 999 2 @pytest.mark.parametrize('deep, expected_orig', [ [True, xr.DataArray(xr.IndexVariable('a', np.array([1, 2])), coords={'a': [1, 2]}, dims=['a'])], [False, xr.DataArray(xr.IndexVariable('a', np.array([999, 2])), coords={'a': [999, 2]}, dims=['a'])]]) def test_copy_coords(self, deep, expected_orig): da = xr.DataArray( np.ones([2, 2, 2]), coords={'a': [1, 2], 'b': ['x', 'y'], 'c': [0, 1]}, dims=['a', 'b', 'c']) da_cp = da.copy(deep) da_cp['a'].data[0] = 999

    expected_cp = xr.DataArray(
        xr.IndexVariable('a', np.array([999, 2])),
        coords={'a': [999, 2]}, dims=['a'])

  assert_identical(da_cp['a'], expected_cp)

xarray\tests\test_dataarray.py:3342:

a = <xarray.DataArray 'a' (a: 2)> array([1, 2]) Coordinates: * a (a) int32 1 2 b = <xarray.DataArray 'a' (a: 2)> array([999, 2]) Coordinates: * a (a) int32 999 2 def assert_identical(a, b):

  xarray.testing.assert_identical(a, b)

E AssertionError: Left and right DataArray objects are not identical E
E Differing values: E L E array([1, 2]) E R E array([999, 2]) E Differing coordinates: E L * a (a) int32 1 2 E R * a (a) int32 999 2 xarray\tests__init__.py:195: AssertionError ___ TestDataset.test_copy_coords[False-expected_orig1] ______ self = <xarray.tests.test_dataset.TestDataset object at 0x000000AAE05256A0> deep = False expected_orig = <xarray.DataArray 'a' (a: 2)> array([999, 2]) Coordinates: * a (a) int32 999 2 @pytest.mark.parametrize('deep, expected_orig', [ [True, xr.DataArray(xr.IndexVariable('a', np.array([1, 2])), coords={'a': [1, 2]}, dims=['a'])], [False, xr.DataArray(xr.IndexVariable('a', np.array([999, 2])), coords={'a': [999, 2]}, dims=['a'])]]) def test_copy_coords(self, deep, expected_orig): ds = xr.DataArray( np.ones([2, 2, 2]), coords={'a': [1, 2], 'b': ['x', 'y'], 'c': [0, 1]}, dims=['a', 'b', 'c'], name='value').to_dataset() ds_cp = ds.copy(deep=deep) ds_cp.coords['a'].data[0] = 999

    expected_cp = xr.DataArray(
        xr.IndexVariable('a', np.array([999, 2])),
        coords={'a': [999, 2]}, dims=['a'])

  assert_identical(ds_cp.coords['a'], expected_cp)

xarray\tests\test_dataset.py:1993:

a = <xarray.DataArray 'a' (a: 2)> array([1, 2]) Coordinates: * a (a) int32 1 2 b = <xarray.DataArray 'a' (a: 2)> array([999, 2]) Coordinates: * a (a) int32 999 2 def assert_identical(a, b):

  xarray.testing.assert_identical(a, b)

E AssertionError: Left and right DataArray objects are not identical E
E Differing values: E L E array([1, 2]) E R E array([999, 2]) E Differing coordinates: E L * a (a) int32 1 2 E R * a (a) int32 999 2 xarray\tests__init__.py:195: AssertionError ============================== warnings summary =============================== ```

Thanks @pletchm !

Perfect @pletchm !

Sorry for the test failures on our end - we had some dependency issues.

I'll merge tomorrow unless anyone has other comments.

Yep, the tests are back now.

Then it's fine! They'll be fixed separately.

But did we lose the tests on this branch? Can we add those back?

@pletchm thanks for the PR!

I think the test failuers are dependencies rather than your changes! I kicked off a build of master here: https://travis-ci.org/pydata/xarray/builds/524754660 to confirm - if it's the same failures then we'll know it's not this code

Thanks for looking into this, @max-sixty! It looks like the errors are the same, so the problem is dependencies, right? How do we go about fixing that?

@pletchm thanks for the PR!

I think the test failuers are dependencies rather than your changes! I kicked off a build of master here: https://travis-ci.org/pydata/xarray/builds/524754660 to confirm - if it's the same failures then we'll know it's not this code

Tests and documentation forthcoming. I had already opened a PR, but was perplexed when the build and tests weren't passing -- suspected it was related to my commits.