Hi @DWesl, @benbovy

I got a bug very similar to what was initially report with the following minimal example (xarray version 2022.11.0) with the error:

ValueError: cannot unstack dimensions that do not have exactly one multi-index: ('z',)

It is working fine with my previous version (xarray 0.16.0) python import xarray as xr ds = xr.Dataset( data_vars=dict( mydata = ( ['across', 'along'], [[0,1],[2,3]]) ) ) stacked = ds.stack(z=("across", "along")) newlist = [None] * stacked.z.size for ii, zindex in enumerate(stacked.z.data): newlist[ii] = stacked.mydata.sel(z=zindex) newds = xr.concat(newlist, dim='z') newds['z'] = stacked.z print('xarray version:' + xr.__version__) newds.unstack(dim='z')

If I do newds = newds.reset_index('z') just before the last line, it is solved for my minimal example. But it still doesn't work for my real problem and I don't think it is the best way to proceed.

Many thanks for your help, Adrien

Hi @benbovy, my code works for 2 dimensions, but it doesn't with 1D. I think the error comes from .set_index(z=['across']) which delivers a "PandasIndex(Int64Index(" and not the expected "PandasIndex(MultiIndex(" for unstack.

I created a Pandas-MultiIndex "MultiIndex([((0,), (1,))], names=['z', 'across'])" (see code just below), but it is not exactly the same one as the one created by .stack() "PandasIndex(MultiIndex([(0,),(1,)], name='z'))" and I didn't find out how to pass the 'index' to xarray. python import pandas as pd tuples = [(0,),(1,)], index = pd.MultiIndex.from_tuples(tuples, names=["z","across"]) index

My modified minimal example is as below: python import xarray as xr ds = xr.Dataset( data_vars=dict( mydata = ( ['across', 'along'], [[0,1],[2,3]]) ), coords=dict( # <- changed across=[0,1], # <- changed along=[0,1], # <- changed ), # <- changed ) stacked = ds.isel(along=0).stack(z=["across"]) # <- changed newlist = [None] * stacked.z.size for ii, zindex in enumerate(stacked.z.data): newlist[ii] = stacked.mydata.sel(z=zindex) newds = xr.concat(newlist, dim='z') newds = newds.set_index(z=['across']) # <- changed newds.unstack(dim='z') # Fail with ValueError: cannot unstack dimensions that do not have exactly one multi-index: ('z',)

Many thanks for the explanation and the rapid answer. I am not very familiar with multi-index.

My real problem works with a DataSet with different sort of variables:

ipdb> lmmap <xarray.Dataset> Dimensions: (x_variables: 4, Observables: 2, Antenna: 4, z: 4, Ambiguities: 4, fun_variables: 8, extrema: 2) Coordinates: * x_variables (x_variables) <U3 'u' 'v' 'c_u' 'c_v' * Observables (Observables) <U6 'sigma0' 'RVL' * Antenna (Antenna) <U4 'Fore' 'MidV' 'MidH' 'Aft' fun_variables (Observables, Antenna) int64 0 1 2 3 4 5 6 7 Dimensions without coordinates: z, Ambiguities, extrema Data variables: (12/17) x (z, Ambiguities, x_variables) float64 -4.503 7.8 ... -0.4265 active_mask (z, Ambiguities, x_variables) int64 0 0 0 0 0 0 ... 0 0 0 0 0 grad (z, Ambiguities, x_variables) float64 7.816e-11 ... -1.873... fun (z, Ambiguities, Observables, Antenna) float64 0.001137 ..... jac (z, Ambiguities, fun_variables, x_variables) float64 0.359... x0 (z, Ambiguities, x_variables) float64 -7.002 10.03 ... 1.219 ... ... message (z, Ambiguities) <U42 '`xtol` termination condition is sat... success (z, Ambiguities) bool True True True True ... True True True method (z) <U3 'trf' 'trf' 'trf' 'trf' xtol (z) float64 0.001 0.001 0.001 0.001 x_scale (z, x_variables) float64 7.0 7.0 0.5 0.5 ... 7.0 7.0 0.5 0.5 bounds (z, extrema, x_variables) int64 -30 -30 -5 -5 ... 30 30 5 5

I don't know where I should apply set_index(z=["across", "along"]) I tried on the full DataSet or a single DataArray but without success. I get the following error:

ValueError: across, along variable(s) do not exist

Which is obvious as across and along variables disappeared. Thank you, I will dig in this direction why it happens.

However, it works perfectly with the minimal example. Many thanks

Thanks @benbovy, my issue is now resolved, once I have enabled the coords to follow.

I didn't see any warning for this

the issue is when you do newds['z'] = stacked.z

Are we raising a warning here?

Many thanks for the explanation and the rapid answer. I am not very familiar with multi-index.

My real problem works with a DataSet with different sort of variables:

ipdb> lmmap <xarray.Dataset> Dimensions: (x_variables: 4, Observables: 2, Antenna: 4, z: 4, Ambiguities: 4, fun_variables: 8, extrema: 2) Coordinates: * x_variables (x_variables) <U3 'u' 'v' 'c_u' 'c_v' * Observables (Observables) <U6 'sigma0' 'RVL' * Antenna (Antenna) <U4 'Fore' 'MidV' 'MidH' 'Aft' fun_variables (Observables, Antenna) int64 0 1 2 3 4 5 6 7 Dimensions without coordinates: z, Ambiguities, extrema Data variables: (12/17) x (z, Ambiguities, x_variables) float64 -4.503 7.8 ... -0.4265 active_mask (z, Ambiguities, x_variables) int64 0 0 0 0 0 0 ... 0 0 0 0 0 grad (z, Ambiguities, x_variables) float64 7.816e-11 ... -1.873... fun (z, Ambiguities, Observables, Antenna) float64 0.001137 ..... jac (z, Ambiguities, fun_variables, x_variables) float64 0.359... x0 (z, Ambiguities, x_variables) float64 -7.002 10.03 ... 1.219 ... ... message (z, Ambiguities) <U42 '`xtol` termination condition is sat... success (z, Ambiguities) bool True True True True ... True True True method (z) <U3 'trf' 'trf' 'trf' 'trf' xtol (z) float64 0.001 0.001 0.001 0.001 x_scale (z, x_variables) float64 7.0 7.0 0.5 0.5 ... 7.0 7.0 0.5 0.5 bounds (z, extrema, x_variables) int64 -30 -30 -5 -5 ... 30 30 5 5

I don't know where I should apply set_index(z=["across", "along"]) I tried on the full DataSet or a single DataArray but without success. I get the following error:

ValueError: across, along variable(s) do not exist

Which is obvious as across and along variables disappeared. Thank you, I will dig in this direction why it happens.

However, it works perfectly with the minimal example. Many thanks

@ACHMartin the issue is when you do newds['z'] = stacked.z. In the last versions of Xarray multi-index levels have each their own (real) coordinates, for consistency and clarity we soon won't support assigning a multi-index to a single coordinate of a Dataset / DataArray like that.

I think that in other places we still do support it with a deprecation notice, but apparently in your example this is not the case. unstack doesn't work because the multi-index(es) and the coordinates of newds are not consistent.

I don't know exactly what is your real problem, but from now on you should avoid implicitly assign a multi-index with xr_obj["my_coord"] = ... or xr_obj.assign(my_coord=...). Instead you should re-create the multi-index, e.g., in your minimal example newds = newds.set_index(z=["across", "along"]).

Fix confirmed, thank you.

Hi @DWesl,

Your example is working with the main branch, it has been fixed in #6889.

```python

after #6889

arr2.unstack("index0")

<xarray.DataArray (index2: 2, level_0: 3, level_1: 5)>

array([[[ 0, 1, 2, 3, 4],

[ 5, 6, 7, 8, 9],

[10, 11, 12, 13, 14]],

[[ 0, 1, 2, 3, 4],

[ 5, 6, 7, 8, 9],

[10, 11, 12, 13, 14]]])

Coordinates:

* level_0 (level_0) int64 0 1 2

* level_1 (level_1) int64 0 1 2 3 4

Dimensions without coordinates: index2

```