issues

What happened: When packaging for a Linux distribution (openSUSE in my case) there is no network access to the build machines for obvious reasons. Therefore we are very happy to preloaded cache and xr.tutorial.open_dataset() function. Unfortunately, with 0.18.2 (not sure about the previous versions) it doesn’t work, because pooch.retrieve() without fname parameter (even when known_hash is None) generates unique name for the data file, which is not available, therefore it fails.

With the data directory expanded to ~/.cache/xarray_tutorial_data call of xr.tutorial.open_dataset('air_temperature') should return the appropriate data even without the network access.

This patch fixes that. When the datafile could be found in the local cache, pooch.retrieve is called with fname parameter.

What happened: When running the test suite while packaging the package for openSUSE I’ve got these failed tests on i586: [ 107s] ______________________ TestDataset.test_resample_loffset _______________________ [ 107s] [gw4] linux -- Python 3.9.5 /usr/bin/python3.9 [ 107s] [ 107s] self = <xarray.tests.test_dataset.TestDataset object at 0xf01f3ec8> [ 107s] [ 107s] def test_resample_loffset(self): [ 107s] times = pd.date_range("2000-01-01", freq="6H", periods=10) [ 107s] ds = Dataset( [ 107s] { [ 107s] "foo": (["time", "x", "y"], np.random.randn(10, 5, 3)), [ 107s] "bar": ("time", np.random.randn(10), {"meta": "data"}), [ 107s] "time": times, [ 107s] } [ 107s] ) [ 107s] ds.attrs["dsmeta"] = "dsdata" [ 107s] [ 107s] # Our use of `loffset` may change if we align our API with pandas' changes. [ 107s] # ref https://github.com/pydata/xarray/pull/4537 [ 107s] actual = ds.resample(time="24H", loffset="-12H").mean().bar [ 107s] expected_ = ds.bar.to_series().resample("24H").mean() [ 107s] expected_.index += to_offset("-12H") [ 107s] expected = DataArray.from_series(expected_) [ 107s] > assert_identical(actual, expected) [ 107s] E AssertionError: Left and right DataArray objects are not identical [ 107s] E [ 107s] E Differing values: [ 107s] E L [ 107s] E array([ 0.018477, -0.178807, -0.792644]) [ 107s] E R [ 107s] E array([ 0.018477, -0.178807, -0.792644]) [ 107s] [ 107s] /home/abuild/rpmbuild/BUILD/xarray-0.18.2/xarray/tests/test_dataset.py:3991: AssertionError [ 107s] _______________________ test_interpolate_1d[1-x-linear] ________________________ [ 107s] [gw2] linux -- Python 3.9.5 /usr/bin/python3.9 [ 107s] [ 107s] method = 'linear', dim = 'x', case = 1 [ 107s] [ 107s] @pytest.mark.parametrize("method", ["linear", "cubic"]) [ 107s] @pytest.mark.parametrize("dim", ["x", "y"]) [ 107s] @pytest.mark.parametrize("case", [0, 1]) [ 107s] def test_interpolate_1d(method, dim, case): [ 107s] if not has_scipy: [ 107s] pytest.skip("scipy is not installed.") [ 107s] [ 107s] if not has_dask and case in [1]: [ 107s] pytest.skip("dask is not installed in the environment.") [ 107s] [ 107s] da = get_example_data(case) [ 107s] xdest = np.linspace(0.0, 0.9, 80) [ 107s] actual = da.interp(method=method, **{dim: xdest}) [ 107s] [ 107s] # scipy interpolation for the reference [ 107s] def func(obj, new_x): [ 107s] return scipy.interpolate.interp1d( [ 107s] da[dim], [ 107s] obj.data, [ 107s] axis=obj.get_axis_num(dim), [ 107s] bounds_error=False, [ 107s] fill_value=np.nan, [ 107s] kind=method, [ 107s] )(new_x) [ 107s] [ 107s] if dim == "x": [ 107s] > coords = {"x": xdest, "y": da["y"], "x2": ("x", func(da["x2"], xdest))} [ 107s] [ 107s] /home/abuild/rpmbuild/BUILD/xarray-0.18.2/xarray/tests/test_interp.py:89: [ 107s] _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ [ 107s] [ 107s] obj = <xarray.DataArray 'x2' (x: 100)> [ 107s] dask.array<xarray-x2, shape=(100,), dtype=float64, chunksize=(100,), chunktype=numpy.....0101 0.0202 0.0303 ... 0.9697 0.9798 0.9899 1.0 [ 107s] x2 (x) float64 dask.array<chunksize=(100,), meta=np.ndarray> [ 107s] new_x = array([0. , 0.01139241, 0.02278481, 0.03417722, 0.04556962, [ 107s] 0.05696203, 0.06835443, 0.07974684, 0.091139...35, 0.80886076, 0.82025316, 0.83164557, 0.84303797, [ 107s] 0.85443038, 0.86582278, 0.87721519, 0.88860759, 0.9 ]) [ 107s] [ 107s] def func(obj, new_x): [ 107s] > return scipy.interpolate.interp1d( [ 107s] da[dim], [ 107s] obj.data, [ 107s] axis=obj.get_axis_num(dim), [ 107s] bounds_error=False, [ 107s] fill_value=np.nan, [ 107s] kind=method, [ 107s] )(new_x) [ 107s] E AttributeError: module 'scipy' has no attribute 'interpolate' [ 107s] [ 107s] /home/abuild/rpmbuild/BUILD/xarray-0.18.2/xarray/tests/test_interp.py:79: AttributeError This is the build log showing all used packages and steps leading to the failures.

Anything else we need to know?: The first failure is easily fixed by scipy-interpolate.patch

Environment: see the above linked log