Type hints allows None for too many parameters

[Sevans711]

Version

2026.6.0

How did you install UXarray?

Source

What happened?

uxarray type hints allow None for too many parameters. For example, in uxarray/core/api.py:

def open_grid(
    grid_filename_or_obj: str | os.PathLike[Any] | dict | Dataset,
    chunks=None,
    use_dual: bool | None = False,
    **kwargs: dict[str, Any],
):
    """Constructs and returns a ``Grid`` from a grid file.
    ...
    use_dual : bool, optional
        Specify whether to use the primal (use_dual=False) or dual (use_dual=True) mesh if the file type is MPAS
    ...
    """
    ...

The docstring and code both expect a bool, not None; there is no special meaning of use_dual=None. So, None should not be allowed in the type hinting here.

Related: #1590

What did you expect to happen?

I expected type hinting for each kwarg to either precisely represent all accepted types, or to be excluded entirely.

Can you provide a MCVE to repoduce the bug?

N/A

[Sevans711]

Briefly clarifying a few points here for future reference:

• This applies in other cases too, not just use_dual. For example, grid.use_geodataframe() type hints allow engine=None, but the method starts by crashing with ValueError if engine not in ["spatialpandas", "geopandas"]. Other examples include destination in UxDataArray.difference(); start_index in Grid.from_topology(); and tol in Grid.from_structured().
• I'm not sure how to add regression tests for this sort of thing. Maybe no regression tests are needed? Or, maybe this would be a good opportunity to introduce mypy or pyright checks into the CI pipeline somewhere? See PR Fix use_dual type hints #1612 for more discussion on this question.

[erogluorhan]

Briefly clarifying a few points here for future reference:

• This applies in other cases too, not just use_dual. For example, grid.use_geodataframe() type hints allow engine=None, but the method starts by crashing with ValueError if engine not in ["spatialpandas", "geopandas"]. Other examples include destination in UxDataArray.difference(); start_index in Grid.from_topology(); and tol in Grid.from_structured().
• I'm not sure how to add regression tests for this sort of thing. Maybe no regression tests are needed? Or, maybe this would be a good opportunity to introduce mypy or pyright checks into the CI pipeline somewhere? See PR Fix use_dual type hints #1612 for more discussion on this question.

Answering this, and also related to #1612 (comment)

• I think taking care of all of them in a PR would be great, and @dylannelson could work on that.
• Regression tests for type hinting might not be needed since even in the codecov reports they are not shown