Add pyproject.toml as default input file format

[berislavlopac]

Added pyproject.toml as a default input file name that is searched for when pip-compile is invoked without arguments; see #1779 for details.

The default file names are searched in the following order, with the first encountered being used to compile:

1. requirements.in
2. setup.py
3. setup.cfg
4. pyproject.toml

Closes #1779.

Contributor checklist

• Provided the tests for the changes.
• Assure PR title is short, clear, and good to be included in the user-oriented changelog

Maintainer checklist

• Assure one of these labels is present: backwards incompatible, feature, enhancement, deprecation, bug, dependency, docs or skip-changelog as they determine changelog listing.
• Assign the PR to an existing or new milestone for the target version (following Semantic Versioning).

[ssbarnea]

@berislavlopac I a very happy to see your proposed that. I was planning to do it myself for quite a few weeks but failed to find the time. Now you can be sure I will do my best to get this merged.

[ssbarnea]

@berislavlopac Apparently there is a reduction in codecoverage reported by codecov, one that seems to be genuine.

[berislavlopac]

@ssbarnea I'm confused - it looks like codecov is complaining that the lines in the test file were not covered by tests? 🤔

[webknjaz]

@berislavlopac yes, from https://app.codecov.io/gh/jazzband/pip-tools/pull/1780 it's evident that the else-block code branch doesn't get executed, meaning that fname in default_file_names is always true there. At least in the CI.

[webknjaz]

Looking at the diff, it's probably related to the fact that you test against CWD which is a bad idea. Look at how other tests create temporary directory structure for all the cases they cover.

[webknjaz]

Usually, the tests should be simple and test one thing only. Once logic is added, they tend to become hard to read/understand as well as easy to be misleading/buggy.

[berislavlopac]

@webknjaz I agree with your points, and will look into the alternative to cwd. However, the issue appeared because setup.cfg was added to the list of allowed files.

I'm splitting the test into two separate cases, one for when there is an allowed file and another without it.

That said, I would still argue that we don't need to include setup.cfg in the list of automatically accepted files. This is because -- if I'm not mistaken -- on its own it doesn't do anything; it either requires setup.py or pyproject.toml to be present for setuptools to work properly.

[webknjaz]

on its own it doesn't do anything; it either requires setup.py or pyproject.toml to be present for setuptools to work properly.

It does require those, but it can still contain the deps list on its own. I don't remember exactly, but I think that pip-tools might be loading it directly.

[berislavlopac]

Okay, after further thought I am okay with keeping in the explicitly defined default list, especially if it's listed after the other two. 👍

[berislavlopac]

you test against CWD which is a bad idea

Since this test is for the functionality that depends directly on some files being present (or not) in the CWD, I'm not sure what would be the right way to test it, apart from creating those files in the CWD (which the current tests do)? 🤔

[webknjaz]

I'm not sure what would be the right way to test it,

https://docs.pytest.org/en/7.1.x/reference/reference.html#pytest.MonkeyPatch.chdir into an empty https://docs.pytest.org/en/7.1.x/how-to/tmp_path.html?

[berislavlopac]

@webknjaz Done, at the cost of a couple more fixtures added to the test. 😄

[atugushev]

Thanks for the PR. A few comments.

[atugushev]

LGTM 👍 Thank you very much for your patience!