Cannot import xarray.core.rolling with xarray version 2025.3.0 and python 3.11

[YassineAbdelouadoud]

What happened?

When installing xarray version 2025.3.0 in a conda environment with the python version pinned to 3.11, the import of xarray.core.rolling fails with ModuleNotFoundError: No module named 'xarray.core.rolling'

Below is a conda env file that allows the reproduction :

name: xarray_test_env
channels:
  - conda-forge
dependencies:
  - python=3.11
  - xarray=2025.3.0

It should be noted that this issue is not present with python 3.12 or later. It is also not present for earlier versions of xarray.

What did you expect to happen?

import xarray.core.rolling should not raise an error

Minimal Complete Verifiable Example

import xarray.core.rolling

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

Traceback (most recent call last):
  File "<string>", line 1, in <module>
ModuleNotFoundError: No module named 'xarray.core.rolling'

Anything else we need to know?

#6844 could be related

Environment

INSTALLED VERSIONS ------------------ commit: None python: 3.11.11 | packaged by conda-forge | (main, Mar 3 2025, 20:43:55) [GCC 13.3.0] python-bits: 64 OS: Linux OS-release: 5.4.0-198-generic machine: x86_64 processor: x86_64 byteorder: little LC_ALL: None LANG: fr_FR.UTF-8 LOCALE: ('fr_FR', 'UTF-8') libhdf5: None libnetcdf: None

xarray: 2025.3.0
pandas: 2.2.3
numpy: 2.2.4
scipy: None
netCDF4: None
pydap: None
h5netcdf: None
h5py: None
zarr: None
cftime: None
nc_time_axis: None
iris: None
bottleneck: None
dask: None
distributed: None
matplotlib: None
cartopy: None
seaborn: None
numbagg: None
fsspec: None
cupy: None
pint: None
sparse: None
flox: None
numpy_groupies: None
setuptools: 75.8.2
pip: 25.0.1
conda: None
pytest: None
mypy: None
IPython: None
sphinx: None

[welcome]

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!

[keewis]

for reference, anything in xarray.core is private API and thus can change at any time.

What are you using xarray.core.rolling for? If this is about anything in the public API (like DatasetRolling for use as a type hint, for example), we may be able to find a way to provide a stable import path for that.

cc @max-sixty

Edit: oh, wait, I didn't notice that this only happens on python=3.11. Let me check what is actually going on here.
Edit2: I can't reproduce with

mamba create -n test python=3.12 xarray=2025.03.0; conda activate test; python -c 'import xarray.core.rolling'; conda deactivate

This always fails for me, regardless of the python version used to create the environment.

[YassineAbdelouadoud]

I am not using the private API directly, but it is used by linopy in expressions.py

But I see now that their latest version has the following code at the beginning :

try:
    # resolve breaking change in xarray 2025.03.0
    import xarray.computation.rolling
    from xarray.computation.rolling import DatasetRolling
except ImportError:
    import xarray.core.rolling
    from xarray.core.rolling import DatasetRolling  # type: ignore

So my issue comes down to using an outdated version of linopy.
Also I was mistaken on the fact it worked with later python version.

[keewis]

As suspected, that import appears to be for type hinting purposes. I guess that means we should provide a stable import path / an official way for type hinting the results of our various intermediate objects.

@pydata/xarray, what do you think?

[dcherian]

Makes sense to me.

[max-sixty]

Yes, good idea. What's the best place for that? xarray.types / xarray.api?

[dcherian]

types has a lot of internal stuff too. So, somewhere else would be good!

[keewis]

xarray.typing?