Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Replace setup.py packaging by Poetry #5266

Merged
merged 29 commits into from
Jun 12, 2024
Merged

Replace setup.py packaging by Poetry #5266

merged 29 commits into from
Jun 12, 2024

Conversation

snejus
Copy link
Member

@snejus snejus commented May 29, 2024

Migrate beets package configuration to Poetry which nowadays seems to be the gold standard.

I have been using Poetry since 2019 and I have mostly been happy a happy user: it makes local dev setup easy and has the tools I need to maintain python packages day to day, including reliable dependency resolution, versioning and publishing to Pypi.

It's a user-friendly tool, so it should make it more straightforward for contributors to setup and navigate the codebase, and ultimately, hopefully facilitate more frequent releases!

Since poetry manages local virtual environment, we do not have much need for tox any more. Therefore, it was replaced by a task runner poethepoet. Type poe in the project directory to see the available commands.

  • Documentation. (If you've added a new command-line flag, for example, find the appropriate page under docs/ to describe it.)
  • Changelog. (Add an entry to docs/changelog.rst to the bottom of one of the lists near the top of the document.)
  • Tests. (Very much encouraged but not strictly required.)

@snejus snejus self-assigned this May 29, 2024
@snejus snejus requested a review from wisp3rwind May 29, 2024 06:16
Copy link

Thank you for the PR! The changelog has not been updated, so here is a friendly reminder to check if you need to add an entry.

@Serene-Arc
Copy link
Contributor

I'm fine with this change to poetry in theory, once all the tests and linting errors have been fixed, but a couple of things:

  • What are the purposes of the changes to release.py? A lot of the things that I removed seem to have been added back in, some of which are definitely not supposed to be there?
  • What are the purposes of changing the tox file, specifically calling binaries instead of python -m black for instance?
  • There are some issues with changing version in the release workflow; does this affect any of that?

@snejus
Copy link
Member Author

snejus commented Jun 1, 2024

I'm fine with this change to poetry in theory, once all the tests and linting errors have been fixed, but a couple of things:

  • What are the purposes of the changes to release.py? A lot of the things that I removed seem to have been added back in, some of which are definitely not supposed to be there?

These commits have been made before I noticed that release.py has been adjusted on master - it needs rebasing where it will get removed :)

  • What are the purposes of changing the tox file, specifically calling binaries instead of python -m black for instance?

Some of the changes have to do with tox upgrade to 4.0.0. Regarding calling the binaries directly - that's the accepted standard, given that every command will be run within the tox environment, where it will be resolved automatically. It is slightly shorter as well.

  • There are some issues with changing version in the release workflow; does this affect any of that?

When I made changes here, the release workflow did not exist yet, so I made adjustments to release.py instead. When I rebase, I will make sure to update the release workflow instead.

@Serene-Arc
Copy link
Contributor

These commits have been made before I noticed that release.py has been adjusted on master - it needs rebasing where it will get removed :)

Fair enough!

Some of the changes have to do with tox upgrade to 4.0.0. Regarding calling the binaries directly - that's the accepted standard, given that every command will be run within the tox environment, where it will be resolved automatically. It is slightly shorter as well.

Great, was just worried with path stuff, but if it's only for the tox environment internally, makes sense.

When I made changes here, the release workflow did not exist yet, so I made adjustments to release.py instead. When I rebase, I will make sure to update the release workflow instead.

Cool.

With regards to the release.py changes, do you want to keep them here or move them to #5274? It might make more sense to get the release workflow working for the current project before doing a big tool change like poetry.

@Serene-Arc Serene-Arc self-requested a review June 3, 2024 02:56
@snejus
Copy link
Member Author

snejus commented Jun 3, 2024

do you want to keep them here or move them to #5274? It might make more sense to get the release workflow working for the current project before doing a big tool change like poetry.

I'm completely with you here @Serene-Arc - I'm moving the relevant commits over.

@snejus snejus mentioned this pull request Jun 3, 2024
@snejus snejus force-pushed the introduce-poetry branch 4 times, most recently from 5b0c665 to fa3def3 Compare June 4, 2024 14:08
snejus added a commit that referenced this pull request Jun 5, 2024
## Description

Fixes #5222. 

Drop Python 3.7. `pyupgrade` is responsible for most of the changes in
the code. I undid some of the bits it attempted to update that aren't
strictly necessary:

1. Converting `List/Dict/Tuple` -> `list/dict/tuple` in modules that
have `from __future__ import annotations` import. This should be done in
a separate PR, and for all modules
2. Converting some `.format(` calls to f-strings. It didn't do it
consistently, and it should also be done in a separate PR, I believe.

Python upgrade unblocks several other PRs, for example #5266 and #5248.
@snejus snejus force-pushed the introduce-poetry branch 15 times, most recently from aea7ad2 to 5668014 Compare June 7, 2024 15:03
This reverts commit c3b6f07.

This commit hardcoded the paths that `isort` and `black` checks. This
means that the `check-format` job will act on the entire codebase
instead of only changed files. We need to define a `path` argument with
a default value in order to achieve the above.

Regarding "." vs "beets beetsplug test", the intention behind using "."
was to also check python files like `docs/conf.py` and
`extra/release.py` which I presume we would also want to format
properly.
This reverts commit fc373f5.

See CONTRIBUTING.rst which has tools setup guidelines for users. They
are expected to install both poetry and poethepoet globally in their
system.
This reverts commit af996f4.

Since `poethepoet` is installed globally in the workflows, running it
does not require `poetry run` suffix. This is actually one of the
reasons why it's preferable to have this tool installed globally.
This reverts commit 4550d39.

I love this attempt to DRY-up the linting workflow! I remember back in
the day also initially assuming that this is how the jobs work. However,
I had to meet the harsh reality of each job needing to be set up from
zero. :(
This reverts commit 5526bd3.

Poetry must be installed before `setup-python` action, weirdly. And we
need to install poethepoet globally too!
@Serene-Arc
Copy link
Contributor

I'm not sure that checking against the changed files only is the best way to do this. Like when we did the release, it is technically possible for unformatted code to make its way into the codebase, however that happens. Currently, any code will always be found by the next PR because it'll come up with a change to the file or a format error in the CI. If we make it so only changed files are checked, this code can be missed for who-knows-how-long.

Maybe it's paranoid but I prefer the formatting tools checking the entirety of the codebase every time. It makes sure that the formatting rules are actually being enforced at every stage possible.

@snejus
Copy link
Member Author

snejus commented Jun 11, 2024

Thanks for clarifying @Serene-Arc, I now see what you mean! Given that the code base is already neat and formatted, I agree that checking everything every time makes more sense! I'm adjusting it now.

I think the only-check-what-has-changed approach makes more sense when the code base is not yet formatted, and is getting tidied up iteratively (if you adjust a file, you must format it properly), which is what we did in most cases in my previous experience. 😅

@snejus
Copy link
Member Author

snejus commented Jun 11, 2024

As you predicted @Serene-Arc, there was indeed an unformatted file docs/conf.py that would have been missed! It's fixed now, thanks for your persistence regarding this change! <3

@snejus snejus force-pushed the introduce-poetry branch 5 times, most recently from c8e8f58 to a4b32ac Compare June 11, 2024 15:28
@Serene-Arc
Copy link
Contributor

Wonderful! That's all my concerns fixed then, merge whenever you'd like.

extra/release.py Outdated
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Are there any changes here that aren't in #5274? If not, then it should be reverted in order to stop any merge conflicts when #5274 is merged.

- name: Get changed files
id: changed-files
with:
fetch-depth: 2
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If you go to the page for the tool, it says that you need either fetch depth 0 or 2. This makes the fetching 2, which is faster than the entire history (which is what fetch-depth: 0 does)

cache: 'poetry'
- name: Install PyGobject dependencies on Ubuntu
if: startsWith(matrix.platform, 'ubuntu-')
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Fair, the goal with that was because while we don't currently test it, we could add testing the LTS version of ubuntu as well, rather than just the latest.


- name: Install PyGobject dependencies
if: matrix.platform == 'ubuntu-latest'
cache: 'poetry'
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ah, okay. Seemed redundant to me because pipx and poetry perform the same role, don't they?

Why does poetry need to be installed before actions/setup-python@v5?

@@ -35,11 +35,9 @@ jobs:
files: |
**.py

format:
if: needs.changed-files.outputs.any_python_changed == 'true'
python-setup:
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'd be happy to write that, the fact that we need it and would use it means that others probably would too. Could be a good addition to the ecosystem.

- uses: actions/setup-python@v5
with:
python-version: ${{ env.PYTHON_VERSION }}
cache: poetry

- uses: abatilo/actions-poetry@v2
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I use Arch Linux, which requires a different way of managing packages. Installing global packages without using the package manager is discouraged.

pyproject.toml Outdated
@@ -152,13 +152,11 @@ poetry = ">=1.8"

[tool.poe.tasks._black]
help = "Run black"
cmd = "black $OPTS $path"
args = { path = { help = "Path to blacken", positional = true, multiple = true, default = "." } }
cmd = "black $OPTS beets beetsplug test"
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It would be, but the way it's done now changes every python file in the directory. For those that use virtual environments (like me) it means that black actually goes and formats every single python file of every installed package in the virtual environment. I think a better approach would be to add docs and extra to the list of approved formatting directories, rather than just specify the entire directory.

Doing it that way instead of trying to specify the files that have changed also removes a bit of added complexity, and means that it's less likely to go wrong. The added time cost of checking every file rather than a subset is negligible, and it'll catch files that have been changed in strange ways that haven't been caught by the system, like our recent thing with the version number being in single quotes instead of double quotes. Since that wasn't changed in subsequent commits, the tool wouldn't have caught it. I think the goal should be to keep the repository in compliant code at every step possible, rather than trying to keep individual commits compliant.

PY_COLORS: 1
IS_MAIN_PYTHON: ${{ matrix.python-version == env.MIN_PYTHON_VERSION && matrix.platform == 'ubuntu-latest' }}
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ah, damn. I'll have to do some testing. The goal with some of these changes was more of a tentative 'what if we tried it this way', so I'll need to do some testing to get it working if you approve of the methods/approach.

Comment on lines 44 to 49
run: |
if [ "${{ env.IS_MAIN_PYTHON }}" = "true" ]; then
poe test
else
poe test --no-cov
fi
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

That's very reasonable! It just seemed like the goal of both was to test, and the coverage was a dependent ancillary condition. But definitely makes sense to have it as two actions.

@@ -35,11 +35,9 @@ jobs:
files: |
**.py

format:
if: needs.changed-files.outputs.any_python_changed == 'true'
python-setup:
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Right...I admit it's been a while since I thought about GH actions, not since I did a two day deep dive for the release workflow. State isn't kept between jobs but is inside jobs, so a new action is probably the way to go. The repeated setup does seem to be a thing easily abstracted away to another job, and adding a cache might be a good thing too.

@snejus
Copy link
Member Author

snejus commented Jun 12, 2024

Ah, I'm only seeing your review comments now, but I guess they are outdated now! 😅 I'm just looking at the last bit to do with the docs linting job, and will merge once it's working as expected.

@snejus snejus merged commit d7bf28d into master Jun 12, 2024
9 checks passed
@snejus snejus deleted the introduce-poetry branch June 12, 2024 09:59
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants