Enable ruff's flake8-trio rule
#2947
Conversation
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## master #2947 +/- ##
==========================================
+ Coverage 99.44% 99.64% +0.19%
==========================================
Files 117 117
Lines 17564 17565 +1
Branches 3166 3166
==========================================
+ Hits 17467 17503 +36
+ Misses 79 43 -36
- Partials 18 19 +1
|
|
At the moment the tests will fail until #2946 gets merged, and even then there will be a failure detailed in astral-sh/ruff#9855 that is a bug. |
|
Disclaimer: I'm the one that wrote 99% of the code in flake8-trio. Ruffs implementation of flake8-trio is only a small set (6 out of ~30) of the full rules: https://github.com/Zac-HD/flake8-trio Although a few of them are handled under other rules. The plugin is intended for users of trio or anyio*, so some rules might not make sense in trio itself, but can ofc then just disable those. It is also possible to run flake8-trio as a standalone*, so we don't have to bring back flake8. (because we wanted to implement autofix with libcst, which isn't possible when running with flake8). But there might be reasons to run it through flake8, I don't fully remember. * yes, both parts of the flake8-trio name are incorrect... python-trio/flake8-async#124 (comment) |
| @@ -165,7 +167,7 @@ def external_thread_fn() -> None: | |||
| thread.start() | |||
| print("waiting") | |||
| while thread.is_alive(): | |||
| await sleep(0.01) | |||
| await lowlevel.checkpoint() | |||
There was a problem hiding this comment.
I think it's probably better to use sleep here because otherwise we'll be busy looping I think?
Like this isn't a matter of "trio just needs another cycle or two of the available tasks" this is "we need to wait for the thread to start and do its stuff"
This happens 2 other places too.
There was a problem hiding this comment.
this is also a bug in ruffs implementation, flake8-trio only triggers when the literal is exactly 0
There was a problem hiding this comment.
wait no, ruff gave TRIO110, which suggests to await an Event instead of looping a sleep. Resolving that by replacing with a checkpoint is ... definitely not the way to go.
Maybe I should widen TRIO110 to also catch checkpoints to make sure people don't confuse them and thinks that because it stops warning on checkpoint that it "solves" it
There was a problem hiding this comment.
I manually changed the 0.02 to a checkpoint because an error was being raised from the sleep there, mostly because the Event part didn't make a lot of sense to me. Should have just done a noqa really.
There was a problem hiding this comment.
yeah an Event could plausibly not be possible or make sense in a context, in which case noqa would be the solution :)
|
I just want to mention that ruff's TRIO rules are significantly behind |
|
Given that ruff's version has a few issues and other points mentioned, I don't think this particular rule is fit for Trio at the moment. |
Originally posted by @TeamSpen210 in #2946 (comment)
This pull request enables the
flake8-trioruff rule (but in ruff's system it's marked asTRIO).