Setup for automated code quality #58
Conversation
| RUN_DEPS = ["geopandas", "pint", "dataretrieval"] | ||
|
|
||
| FORMAT_DEPS = ["ruff >= 0.3.5"] | ||
| TESTS_DEPS = ["pytest", "coverage"] |
There was a problem hiding this comment.
todo: Ideally, we want to keep those in sync with the deps listed in pyproject.toml.
| select = [ | ||
| "E", | ||
| "F", | ||
| "W", | ||
| "I" | ||
| ] |
There was a problem hiding this comment.
suggestion: You might want to take a look at the list of rules available. This configuration is equivalent to flake8 + isort, but we could want a different set of rules for harmonize_wq
Rules can also be ignored:
- entirely
- at the file level
- for a specifc line using a comment
# noqa: E722for instance
| known-first-party = ["harmonize_wq"] | ||
|
|
||
| [tool.ruff.format] | ||
| quote-style = "single" |
There was a problem hiding this comment.
discussion: The default format used by black (and ruff as well) is to use double quotes for strings. However, this option will avoid messing with the full code base
|
Starting to close the loop on this by copying over what I've mentioned elsewhere. The suggested settings have been working well in the pyproject.toml for running ruff locally without nox. I have pre-commits mostly working, but without the settings. I'm trying to work out if there is an easy way to do that from what is in the toml without nox. A lot of pieces I don't have a full grasp on though. What I'm aiming for is to have pre-commit make basic ruff changes on a PR, but to fail for not fixable changes. The idea being a contributor wouldn't have to get ruff/nox running locally to leverage it. |
|
Specific to the settings - I appreciated quote-style = "single", at least for now to give me time to ensure it won't matter. Outside of that the only ones I'm not sure of are: (1) adding space between some import types (I001 - isn't specific to order vs spacing so I may just have to deal with a couple extra empty lines), and (2) E501 Line too long, where I was already linting for this, but not in docs examples where those lines get compared to doctest (as long as I can find a way to fix without failing doctests no issue). |
|
TLDR - implemented w/ PR #89 . To follow-up on this: part of the discrepancy between local and the pre-commit ended up being that format wasn't being run locally, I used quote-style = "preserve" during initial updates to differentiate where is was just '' vs "". In the end since it was somewhat arbitrary the default "" is used. # noqa: worked on docstring in the end but I tried to resolve using any other means first, over many attempts. All this was merged on PR 89. This PR was very helpful in providing initial settings in the pyproject.toml - thank you! Really the only difference for the merged PR is that it uses pre-commits to run it as part of checks instead of setting up nox to run it local. |
|
whoops - meant to close w/ last comment |
As discussed in pyOpenSci/software-submission#157, here is a suggestion on how to automate code quality (but not limited to!) tasks.
The setup is simple, as we use only ruff to both format and lint the files.
I suggest using nox as the task runner.
Running the formatter + linter is just as simple as
for short, or
for the long version.
Tests are also included:
Now, for the configuration itself, I have added some entries to
pyproject.tomlwith what, I think, is a good first setup forharmonize_wq.There are things to discuss, so I'll keep this PR as a draft for now.