Not possible to run statsforecast with PYTHONOPTIMIZE=2

[moagstar]

What happened + What you expected to happen

It is not possible to run statsforecast when running with PYTHONOPTMIZE=2 (which strips __doc__) since there is some automatic formatting being applied to __doc__, for example:

__init__.__doc__ = __init__.__doc__.format(**_param_descriptions)

The result is an AttributeError because the __doc__ attribute has been stripped by running python with the optimization.

Reproducible example

__init__.__doc__ = __init__.__doc__.format(**_param_descriptions)

Issue Severity

None

[robrui]

The bug is actually in coreforecast (a dependency of statsforecast). The docstring decorators in expanding.py and rolling.py called .format() on __doc__ without checking for None.

Fixed in Nixtla/coreforecast#131 — added if function.__doc__: guards before formatting.

Once that fix is merged and released, statsforecast will work with PYTHONOPTIMIZE=2 automatically.

[jbbqqf]

Hi! Triaging older issues this week — I think this one has shipped and can be closed.

I tried to reproduce the original symptom against current main:

PYTHONOPTIMIZE=2 python -c "
import statsforecast
from statsforecast import StatsForecast
from statsforecast.models import Naive, AutoARIMA, AutoTheta, AutoETS, AutoCES, MSTL, MFLES
print('imports ok')
"
# → imports ok

grep -rn '__doc__' python/statsforecast finds no __init__.__doc__ = __init__.__doc__.format(...) patterns anymore — the codebase moved from lazydocs / nbdev-style notebook-generated docstrings to mkdocstrings (#1060), so the line you flagged no longer exists.

Would you mind closing this out? If you can still trigger an AttributeError under PYTHONOPTIMIZE=2 against main, point me at the exact failing module/symbol and I'll dig further.

(Disclosure: I drafted this comment with help from Claude Code while triaging stale issues; the references above were verified manually against the current main branch.)