Add support for polars dataframes and series

[fonnesbeck]

Description

Mostly superficial changes to recognize polars data structures.

Related Issue

• Closes ENH: Replace pandas dependence/use with narwhals #7462

Checklist

• Checked that the pre-commit linting/style checks pass
• Included tests that prove the fix is effective or that the new feature works
• Added necessary documentation (docstrings and/or example notebooks)
• If you are a pro: each commit corresponds to a relevant logical change

Type of change

• New feature / enhancement
• Bug fix
• Documentation
• Maintenance
• Other (please specify):

---

📚 Documentation preview 📚: https://pymc--7463.org.readthedocs.build/en/7463/

[ricardoV94]

polars should be an optional dependency. For the dispatch it can be done with a try except import

[ricardoV94]

This is more succinct. Also type hint of df was wrong, so I just removed it.

Suggested change

	if pl is not None:
	@_as_tensor_variable.register(pd.Series)
	@_as_tensor_variable.register(pd.DataFrame)
	@_as_tensor_variable.register(pl.DataFrame)
	@_as_tensor_variable.register(pl.Series)
	def dataframe_to_tensor_variable(df: pd.DataFrame | pl.DataFrame, *args, **kwargs) -> TensorVariable:
	return pt.as_tensor_variable(df.to_numpy(), *args, **kwargs)
	else:
	@_as_tensor_variable.register(pd.Series)
	@_as_tensor_variable.register(pd.DataFrame)
	def dataframe_to_tensor_variable(df: pd.DataFrame, *args, **kwargs) -> TensorVariable:
	return pt.as_tensor_variable(df.to_numpy(), *args, **kwargs)
	@_as_tensor_variable.register(pd.Series)
	@_as_tensor_variable.register(pd.DataFrame)
	def dataframe_to_tensor_variable(df, *args, **kwargs) -> TensorVariable:
	return pt.as_tensor_variable(df.to_numpy(), *args, **kwargs)
	if pl is not None:
	@_as_tensor_variable.register(pl.DataFrame)
	@_as_tensor_variable.register(pl.Series)
	def polars_dataframe_to_tensor_variable(df, *args, **kwargs) -> TensorVariable:
	return pt.as_tensor_variable(df.to_numpy(), *args, **kwargs)

[ricardoV94]

Suggested change

	elif hasattr(data, "to_numpy") and hasattr(data, "is_null"):
	elif hasattr(data, "to_numpy") and hasattr(data, "is_null"):
	# Probably polars object

[lorentzenchr]

Why not a bit more explicit:

Suggested change

	elif hasattr(data, "to_numpy") and hasattr(data, "is_null"):
	elif pl is not None and isinstance(data, (pl.DataFrame, pl.Series)):

The polars namespace is used anyway (in the except clause).

[codecov]

Codecov Report

Attention: Patch coverage is 42.30769% with 15 lines in your changes missing coverage. Please review.

Project coverage is 92.10%. Comparing base (8cdc9ee) to head (f304035).
Report is 91 commits behind head on main.

Files with missing lines	Patch %	Lines
pymc/pytensorf.py	40.00%	15 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #7463      +/-   ##
==========================================
- Coverage   92.17%   92.10%   -0.08%     
==========================================
  Files         103      103              
  Lines       17258    17279      +21     
==========================================
+ Hits        15908    15914       +6     
- Misses       1350     1365      +15     

Files with missing lines	Coverage Δ
pymc/data.py	89.44% <100.00%> (ø)
pymc/pytensorf.py	87.50% <40.00%> (-3.02%)	⬇️