Should ExprName allow Function as a type for the parent parameter?

[echoix]

I'm almost done working on a feature to enhance sphinx docstring handling of parameter types when defined with :param a_type foo: or :param foo:+:type foo: a_type pair, by passing the annotation through parse_docstring_annotation() like the google docstrings.

I'm writing some new unit tests for this, and the type annotations for ExprName doesn't allow the parent to be a Function object. However, when running mkdocs build on a real project (for example astroid) through a debugger, with an editable installation of griffe, with breakpoints in the sphinx.py file, the parent passed in the Docstring object is really a Function object, and in the parsed annotation as expressions (like parse_docstring_annotation() returns), the parts where ExprName are used, they have their parent attribute set to the function that the docstring is parsed from.

The type annotation for parent of ExprName are defined here:

griffe/src/_griffe/expressions.py

Lines 595 to 603 in b18b63d

	# YORE: EOL 3.9: Replace `**_dataclass_opts` with `slots=True` within line.
	@dataclass(eq=False, **_dataclass_opts)
	class ExprName(Expr): # noqa: PLW1641
	"""This class represents a Python object identified by a name in a given scope."""
	name: str
	"""Actual name."""
	parent: str | ExprName | Module | Class | None = None
	"""Parent (for resolution in its scope)."""

[pawamoy]

The investigation is sound. It looks like we should allow ExprName.parent to be a Function indeed 🙂 Thanks! Would you like to send a PR?

[echoix]

I'll try next time I have time