Really? Okay, so that means that currently we don't treat a named DataArray and a single-variable Dataset as if they are the same. For example I would have expected these two operations to give the same result: python objs = [DataArray([0], dims='x', name='a'), DataArray([0], dims='x', name='b')] concat(objs, dim='x') <xarray.DataArray 'a' (x: 2)> array([0, 0]) Dimensions without coordinates: x python objs = [Dataset({'a': ('x', [0])}), Dataset({'b': ('x', [0])})] concat(objs, dim='x') ``` self = Frozen(OrderedDict([('b', <xarray.Variable (x: 1)> array([0]))])), key = 'a'

def __getitem__(self, key: K) -> V:

  return self.mapping[key]

E KeyError: 'a'

xarray/core/utils.py:385: KeyError ```

Is this what we want to do? Surely the first one should also fail, else this is counter-intuitive. I think of a named DataArray and a single-variable Dataset as being the same thing, just a single physical variable? @shoyer am I misunderstanding xarray's data model here?

Okay something has definitely gone wrong here.

My intention with that test was to check that the order of operations doesn't matter, but you're right that the test as written makes no sense. It would probably be a good idea to remove this test and check that property correctly by adding a second assert to the (poorly-named) test_auto_combine_2d: ```python

Prove it works symmetrically

datasets = [[ds(0), ds(3)], [ds(1), ds(4)], [ds(2), ds(5)]] result = combine_nested(datasets, concat_dim=["dim2", "dim1"]) assert_equal(result, expected) ``` (This passes fine)

However, that still leaves the question of why is this nonsensical test passing?

I think it's because concat is not failing when it should - that test boils down to calling concat on those DataArrays (called from _combine_1d internally). Surely concat should fail when you ask it to do this, because how can you concatenate two different variables? ```python da1 = DataArray(name="a", data=[[0]], dims=["x", "y"]) da2 = DataArray(name="b", data=[[1]], dims=["x", "y"])

result = concat([da1, da2], dim="x") However it doesn't fail, instead it gives this!: <xarray.DataArray 'a' (x: 2, y: 1)> array([[0], [1]]) Dimensions without coordinates: x, y `` Where has'b'gone?! This is the reason thattest_concat_name_symmetry` gives such a weird result.

Hmm I can look at this properly at the weekend but in the meantime the logic was motivated by discussion in #2777. If the test doesn't make sense in that context then it's not right.

On Wed, 18 Sep 2019, 18:16 Deepak Cherian, notifications@github.com wrote:

Yes/

https://github.com/pydata/xarray/blob/fddced063b7ecbea6254dc1008bb4db15a5d9304/xarray/tests/test_combine.py#L467-L478

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/pydata/xarray/issues/3315?email_source=notifications&email_token=AISNPI6OZK7YES6JUSDSWCLQKJO7VA5CNFSM4IXWOSZ2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD7AZOBI#issuecomment-532780805, or mute the thread https://github.com/notifications/unsubscribe-auth/AISNPI3UMSS4MAOZKQLNRKTQKJO7VANCNFSM4IXWOSZQ .

Sorry when you say expected result are you referring to a particular unit test?

On Wed, 18 Sep 2019, 18:07 Deepak Cherian, notifications@github.com wrote:

This honestly makes no sense to me.

da1 = xr.DataArray(name="a", data=[[0]], dims=["x", "y"]) da2 = xr.DataArray(name="b", data=[[1]], dims=["x", "y"]) da3 = xr.DataArray(name="a", data=[[2]], dims=["x", "y"]) da4 = xr.DataArray(name="b", data=[[3]], dims=["x", "y"]) xr.combine_nested([[da1, da2], [da3, da4]], concat_dim=["x", "y"])

These are dataarrays with two different names. Why is this the expected result?

<xarray.DataArray 'a' (x: 2, y: 2)> array([[0, 1], [2, 3]]) Dimensions without coordinates: x, y

That error arises because it's trying to concatenate data_vars a and b but there are datasets that don't have a. If you set those DataArrays to have the same name, this will work.

da1 = xr.DataArray(name="a", data=[[0]], dims=["x", "y"]) da2 = xr.DataArray(name="a", data=[[1]], dims=["x", "y"]) da3 = xr.DataArray(name="a", data=[[2]], dims=["x", "y"]) da4 = xr.DataArray(name="a", data=[[3]], dims=["x", "y"])

ds1 = da1.to_dataset() ds2 = da2.to_dataset() ds3 = da3.to_dataset() ds4 = da4.to_dataset() xr.combine_nested([[ds1, ds2], [ds3, ds4]], concat_dim=["x", "y"])

<xarray.Dataset> Dimensions: (x: 2, y: 2) Dimensions without coordinates: x, y Data variables: a (x, y) int64 0 1 2 3

ping @TomNicholas https://github.com/TomNicholas

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/pydata/xarray/issues/3315?email_source=notifications&email_token=AISNPI36JGXYGON2QZCX4RLQKJN6DA5CNFSM4IXWOSZ2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD7AYT7Y#issuecomment-532777471, or mute the thread https://github.com/notifications/unsubscribe-auth/AISNPIZPWVSS7SU2A3AEUMLQKJN6DANCNFSM4IXWOSZQ .