Add Type checking step?

[ChrisBarker-NOAA]

What should we add?

There are a fair bit (complete?, no idea) of type annotations in the code.

It's not a great idea to have the annotations and not be running a type checker on them. Bad / incorrect annotations are worse that no annotations.

I have no experience with type checkers, but it seems MyPy is the most used now (though pyright is also ...).

Here's some advice on how to get started:

https://mypy.readthedocs.io/en/latest/existing_code.html#using-mypy-with-an-existing-codebase

@ocefpaf: Do you have any experience setting this up?

Reference

No response

[ocefpaf]

@ocefpaf: Do you have any experience setting this up?

I usually set a linter, like .pre-commits, to check where type checking is missing and run mypy. I can start a PR here for this, or we can add to the GSoC heap. It kind of falls into the test improvement category we created there.

[ChrisBarker-NOAA]

add to the GSoC heap.

Sounds good to me, unless you are bored and looking for something to do :-)

[ocefpaf]

Basic type checking is done! The next steps are:

• run mypy in the realized environment (un-comment these lines.)
• enable strict mode (maybe?)

[ChrisBarker-NOAA]

I'm still a bit confused about pre-commits, and therefore what "run mypy in the realized environment" will actually do.

But let's give it a shot -- if it causes problems we can roll it back.

As for strict mode -- probably not at first, but I"m also not totally sure what that actually means in practice.

[ocefpaf]

I'm still a bit confused about pre-commits, and therefore what "run mypy in the realized environment" will actually do.

The pre-commits runs things in their own virtual environment, that means the only dependencies there at the moment are mypy and the pandas-stubs. When you have all the project dependencies installed, like cf-xarray, netcdf4, and xarray>=2024.6, mypy will likely perform a better assessment of the types and use whatever stubs are present from those libraries. It should be clearer when I send my next PR (working on it). Also, it should not cause any issues, just more accurate typing.

As for strict mode -- probably not at first, but I"m also not totally sure what that actually means in practice.

That is not super clear to me as well. It is a mypy mode that can be very pedantic. While adding the dependencies return an extra 19 type issues, with strict we get +140! I'm not looking forward to figure out if those hundreds of issues are really type problems we should be worried about.