ImportError+cannot+import+name+Appender+from+statsmodels+compat+pandas

Not specified by the original author.

TL;DR: statsmodels removed its Appender shim. Import it from pandas.util._decorators instead, or upgrade the package that does the old import.

The error

ImportError: cannot import name 'Appender' from 'statsmodels.compat.pandas'

Fix it

1. Replace the import in your code:

   from pandas.util._decorators import Appender

   Expected: no import error; Appender behaves the same.

2. If the bad import is inside a third-party package, upgrade that package:

   pip install -U [the-package]

   Expected: new release uses the new path.

3. Verify the import that failed now works.

When this applies

Use this when code does from statsmodels.compat.pandas import Appender on a recent statsmodels.

When it does not apply

If the error names a different statsmodels.compat member, check whether the shim still exists. If statsmodels itself will not import, reinstall it.

Compatibility

Recent statsmodels versions where the compat shim was removed. pandas.util._decorators.Appender exists across pandas 1.x and 2.x.

Root cause

statsmodels.compat.pandas held shims for pandas API drift, including a re-export of pandas' Appender docstring decorator. statsmodels cleaned the module up and dropped the re-export; anything importing it breaks.

Edge cases

Appender is itself semi-private in pandas (pandas.util._decorators), so prefer upgrading the dependent package over relying on the new path long-term. Grep for other statsmodels.compat imports when upgrading.

GitHub issue statsmodels/statsmodels#6760 reports ImportError: cannot import name 'Appender' from 'statsmodels.compat.pandas' with a modern statsmodels installed. The verified answer: statsmodels removed the Appender compatibility shim from statsmodels.compat.pandas, so code importing it must use pandas.util._decorators.Appender directly (or upgrade the package that does the old import to a fixed release). Full thread: https://github.com/statsmodels/statsmodels/issues/6760

Not specified by the original author.

TL;DR: your statsmodels is too old for your scipy. scipy removed factorial from scipy.misc, so upgrade statsmodels to a release that no longer uses it, or pin an older scipy.

The error

ImportError: cannot import name 'factorial' from 'scipy.misc'

Fix it

1. Upgrade statsmodels:

   pip install -U statsmodels

   Expected: installs a release fixed for modern scipy.

2. Verify:

   python -c "import statsmodels; print(statsmodels.__version__)"

   Expected: imports cleanly.

3. Workaround if you cannot upgrade statsmodels: pin scipy below the removal:

   pip install "scipy<1.13"

   Expected: the old name exists again.

When this applies

Use this when import statsmodels raises this error with a recent scipy.

When it does not apply

If your own code imports factorial from scipy.misc, use math.factorial or scipy.special.factorial instead. If statsmodels is current, reinstall it.

Compatibility

statsmodels releases before the scipy-compat fix; scipy 1.13+ where scipy.misc.factorial is gone.

Root cause

scipy.misc was a grab-bag module that scipy emptied out over several releases; factorial moved to scipy.special long ago and the scipy.misc alias was finally deleted. statsmodels imported the old path, so the deletion broke it at import.

Edge cases

The same scipy.misc cleanup also removed imread, imsave, and comb; old statsmodels hit factorial first, but grep for scipy.misc imports when upgrading. Pinning old scipy can break packages that need new scipy.

GitHub issue statsmodels/statsmodels#5619 reports ImportError: cannot import name 'factorial' from 'scipy.misc' when importing statsmodels with a modern scipy. The verified resolution: scipy removed factorial from scipy.misc, so old statsmodels broke at import; the fix is upgrading statsmodels to a release that imports factorial from its new location (or pinning an older scipy as a workaround). Full thread: https://github.com/statsmodels/statsmodels/issues/5619

Not specified by the original author.

TL;DR: your ray version imports a private pandas helper that pandas 2.x removed. Upgrade ray to a release that supports pandas 2, or pin pandas below 2.0 as a workaround.

The error

ImportError: cannot import name 'ExtensionArrayFormatter' from 'pandas.io.formats.format'

Fix it

1. Check the versions in play:

   pip show ray pandas | grep -E "Name|Version"

   Expected: shows ray and pandas versions; pandas is 2.x.

2. Upgrade ray to a pandas-2-compatible release:

   pip install -U ray

   Expected: installs a newer ray.

3. Verify:

   python -c "import ray; print(ray.__version__)"

   Expected: imports cleanly.

4. If you cannot upgrade ray, pin pandas instead:

   pip install "pandas<2.0"

   Expected: the old import path exists again.

When this applies

Use this when import ray (or a ray submodule) raises this error with pandas 2.x installed.

When it does not apply

If your own code imports pandas.io.formats.format internals, stop using the private API instead. If ray is already current, reinstall it; the install may be corrupt.

Compatibility

ray releases before pandas 2 support landed; pandas 2.0+.

Root cause

ExtensionArrayFormatter was a private helper inside pandas.io.formats.format. pandas 2.0 refactored the formatting internals and removed it. ray imported the private name, so the removal broke ray at import time.

Edge cases

Pinning pandas below 2.0 can conflict with other packages that require pandas 2; upgrading ray is the cleaner path. In conda envs, update both from the same channel to avoid a mixed solve.

GitHub issue ray-project/ray#42572 reports ImportError: cannot import name 'ExtensionArrayFormatter' from 'pandas.io.formats.format' when importing ray with pandas 2.x installed. The verified resolution: ray was importing a private pandas formatting helper that pandas 2.0 removed, so the fix is upgrading ray to a release that supports pandas 2 (or pinning pandas below 2.0 as a workaround). Full thread: https://github.com/ray-project/ray/issues/42572