chore(typing): Ignore another EagerDataFrame intermittent [False Negative]

[dangotbanned]

What type of PR is this? (check all applicable)

• 💾 Refactor
• ✨ Feature
• 🐛 Bug Fix
• 🔧 Optimization
• 📝 Documentation
• ✅ Test
• 🐳 Other

Related issues

• refactor: Organizing Compliant* APIs #3045

Checklist

• Code follows style guide (ruff)
• Tests added
• Documented the changes

If you have comments or can explain your changes, please do so below

Been seeing this locally for a few days, but never in CI

[FBruzzesi]

Out of curiosity, do you know why this is the case?

[dangotbanned]

@FBruzzesi

Out of curiosity, do you know why this is the case?

Maybe related to https://discuss.python.org/t/unsoundness-of-contravariant-self-type/86338

But I guess to zoom in on it a bit more ...

I think the issue is EagerExprT gets bound in EagerDataFrame.aggregate - but then occasionally fails to be re-used for EagerDataFrame.select.

My intention for the typing is to say:

EagerDataFrame is generic over EagerDataFrame._expr: type[EagerExprT]
The type of EagerExprT should bind to the class

This is problematic because here that type is contravariant:

narwhals/narwhals/_compliant/dataframe.py

Line 148 in 76dbde2

def select(self, *exprs: CompliantExprT_contra) -> Self: ...

But here that type is invariant:

narwhals/narwhals/_compliant/dataframe.py

Line 324 in 76dbde2

def aggregate(self, *exprs: EagerExprT) -> Self:

... and then in EagerExpr itself, it depends on an invariant EagerDataFrame 😳:

narwhals/narwhals/_compliant/expr.py

Lines 213 to 226 in 76dbde2

	class EagerExpr(
	DepthTrackingExpr[EagerDataFrameT, EagerSeriesT],
	Protocol[EagerDataFrameT, EagerSeriesT],
	):
	_call: EvalSeries[EagerDataFrameT, EagerSeriesT]
	_scalar_kwargs: ScalarKwargs
	def __init__(
	self,
	call: EvalSeries[EagerDataFrameT, EagerSeriesT],
	*,
	depth: int,
	function_name: str,
	evaluate_output_names: EvalNames[EagerDataFrameT],