Use argparse instead of click

[mcflugen]

This pull request was originally intended to just replace click with argparse to reduce the number of dependencies but, since the repository hadn't been touched in a bit, I couldn't help myself from tidying things up a bit.

This addresses both #44 and #10.

Highlights:

• Dropped testing of Python 3.9; require Python 3.10+
• Added annotations everywhere
• Removed dependency on click and black
• Fixed the Windows error caused by non-ascii characters
• Test files now use a _test.py suffix—so much easier for tab-completion and they aren't included in the source dist by default.
• Converted the readme from rst to markdown

[coveralls]

coverage: 76.214% (-1.1%) from 77.32%
when pulling fd5c3b3 on mcflugen/use-argparse
into 3447c93 on master.

[mdpiper]

Would you consider removing setup.cfg and setup.py? (I'm not sure what's behind my burning need to erase them.)

[mcflugen]

Would you consider removing setup.cfg and setup.py?

Some tools still use setup.py but I don't think any that we would use do and, I'm pretty sure, the GitHub dependency graph now understands pyproject.toml. So, yeah, we should be able to remove setup.py.

As for setup.cfg, I can move the coverage and zest.releaser configuration into pyproject.toml. It'll be a while before we can migrate flake8, as they don't want to support pyproject.toml. Another option would be to move everything from pyproject.toml into setup.cfg. In that case we would only have a single config file.

Or we could use something other than flake8, or use it through ruff (but not all of the extensions we use are supported). Things are working well now, though, so I would rather not do that just to eliminate one configuration file.

[mdpiper]

This is neat! I especially like how you made _test a suffix. Most of my comments are suggestions, but the last two may need fixing.

[mdpiper]

Replace with install from URL?

pip install git+https://github.com/csdms/bmi-python.git

This is a suggestion only.

[mcflugen]

I was thinking of exactly this but thought you had written this so was going to ask you about it first.

[mcflugen]

I changed the instructions to pip-install from a url as you suggested.

[mdpiper]

Should the keywords be lowercased? A quick search didn't reveal guidance one way or another.

[mcflugen]

I have no idea but I think uppercase should be fine.

[mdpiper]

I saw that if you cap-case these table entries they get cap-cased on the PyPI project page.

[mdpiper]

In the output from bmipy-render, numpy is referred to as np in several signatures.

[mcflugen]

I've included import numpy as np to satisfy the linters but I don't think it's needed to run the code. Because we import annotations from the future, the type annotations are stringified before the code is executed to enable delayed evaluation.

In the future, we may want to annotate the numpy types with strings to remove the numpy requirement.

[mdpiper]

The type hints in the rendered output have quotes; e.g.,

def finalize(self) -> "None":

[mcflugen]

Oh, right. I originally included #46, which would have fixed this issed, in this pull request but decided to break it up first. The strings should still be fine but maybe it would be best to merge #46 first.

[mcflugen]

@mdpiper Could you check this again? I think this should be good now.

[mdpiper]

@mcflugen Yup, the quotes are gone.

Another thing, though--does NDArray get imported?

[mcflugen]

@mdpiper Thanks!

See my comment above about delayed evaluation of annotations. Although I think it's not necessary, for now, I think I'll add NDArray to the bmi template.