Repository navigation
DataArray.groupby drops empty coordinates #11188
Description
What happened?
When I run groupby on an empty DataArray, empty coordinates are dropped.
This is is xarray version 2024.10.0.
import numpy as np
import xarray as xr
data = xr.DataArray(np.empty((0, 2)), dims=['x', 'y'], coords={'x': [], 'y': [1, 1]})
print(data.groupby('y').sum())
<xarray.DataArray (x: 0, y: 1)> Size: 0B
array([], shape=(0, 1), dtype=float64)
Coordinates:
* y (y) int64 8B 1
Dimensions without coordinates: x
As you see, x coordinate is not longer present.
What did you expect to happen?
I expect x coordinate to be preserved.
E.g., here is the optuput of the same snippet in xarray==2024.2.0:
<xarray.DataArray (x: 0, y: 1)> Size: 0B
array([], shape=(0, 1), dtype=float64)
Coordinates:
* x (x) float64 0B
* y (y) int64 8B 1
Minimal Complete Verifiable Example
import xarray as xr
xr.show_versions()
# your reproducer code ...
import numpy as np
import xarray as xr
data = xr.DataArray(np.empty((0, 2)), dims=['x', 'y'], coords={'x': [], 'y': [1, 1]})
assert 'x' in data.groupby('y').sum().coordsSteps to reproduce
No response
MVCE confirmation
- Minimal example — the example is as focused as reasonably possible to demonstrate the underlying issue in xarray.
- Complete example — the example is self-contained, including all data and the text of any traceback.
- Verifiable example — the example copy & pastes into an IPython prompt or Binder notebook, returning the result.
- New issue — a search of GitHub Issues suggests this is not a duplicate.
- Recent environment — the issue occurs with the latest version of xarray and its dependencies.
Relevant log output
Anything else we need to know?
This issue is caused by the change introduced to DataArray._replace_maybe_drop_dims function in scope of #5361.
In this case self.dims == ('x', 'y') and variable.dims == ('y', 'x') so the condition set(self.dims) == set(variable.dims) is True.
However, the next line new_sizes = dict(zip(self.dims, variable.shape, strict=True)) assumes that the dims order is the same, otherwise it does not have any sense.
Environment
Details
INSTALLED VERSIONS
commit: None
python: 3.12.7 (main, Oct 1 2024, 02:05:46) [GCC 13.3.0]
python-bits: 64
OS: Linux
OS-release: 4.18.0-553.34.1.el8_10.x86_64
machine: x86_64
processor:
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8
LOCALE: ('en_US', 'UTF-8')
libhdf5: 1.14.5
libnetcdf: 4.9.2
xarray: 2024.10.0
pandas: 2.2.3
numpy: 1.26.4
scipy: 1.14.1
netCDF4: 1.7.1.post2
pydap: None
h5netcdf: 1.4.0
h5py: 3.12.1
zarr: 3.1.1
cftime: 1.6.4
nc_time_axis: 1.4.1
iris: None
bottleneck: 1.4.0
dask: 2025.11.0
distributed: 2025.11.0
matplotlib: 3.9.2
cartopy: None
seaborn: 0.13.2
numbagg: 0.8.2
fsspec: 2024.3.0
cupy: 13.3.0
pint: None
sparse: 0.15.4
flox: None
numpy_groupies: None
setuptools: 75.1.1.post0
pip: 24.0
conda: None
pytest: 8.3.3
mypy: 1.11.2
IPython: 8.29.0
sphinx: 7.4.7
Activity
welcome commentedon Feb 20, 2026
Thanks for opening your first issue here at xarray! Be sure to follow the issue template!
If you have an idea for a solution, we would really welcome a Pull Request with proposed changes.
See the Contributing Guide for more.
It may take us a while to respond here, but we really value your contribution. Contributors like you help make xarray better.
Thank you!
ianhi commentedon Feb 25, 2026
Hi @eugene57 thanks for the report and debugging pass!
a quick workaround for you is to install flox into your environment which xarray will then automatically use. This fixes your bug and should also be faster (or as fast) as the default. xarray groupby.
note for any one else trying to repro this means that to reproduce from a dev env you must set:
xr.set_options(use_flox=False)nevertheless there is a bug here as you have identified. Its also beyond just empty dims. I think its any time you group by a non-leading dim (i.e. reordering hapapens) and the dims don't have the same lengths
kkollsga commentedon Mar 8, 2026
@ianhi confirmed — it's any non-leading dim groupby where dims have different lengths.
The root cause is in
_replace_maybe_drop_dims(dataarray.py:529-531). PR #9517 relaxed theelifcondition toset(self.dims) == set(variable.dims), but the next line still assumes matching order:When dims are reordered, this maps wrong shapes to wrong dims, so the coordinate filter drops coords it shouldn't. Fix: replace with
dict(variable.sizes).