feat: Add type hinting - #791
Conversation
ebb5665 to
6b258c4
Compare
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #791 +/- ##
==========================================
- Coverage 96.65% 95.49% -1.16%
==========================================
Files 27 27
Lines 3554 3865 +311
==========================================
+ Hits 3435 3691 +256
- Misses 119 174 +55 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
6b258c4 to
6c06307
Compare
2a23a5a to
3a14bc1
Compare
|
Hello, Thanks a lot for the PR and the issues you tagged along the way, it's really appreciated to have some fresh eyes on the project. This PR is quite big so I tagged the @python-zk/maintainers team to have feedback. As for me, I am not really a use-typing-in-Python guy. When I need typing, I use another language, but I know those changes are very valuable for the community. I will review it in the next days as I have some time off. Special pings to @a-ungurianu and @bringhurst: since you worked on typing features in the past, it would be really appreciated if you can chime in and add a review. |
872accb to
1a3456c
Compare
|
@bbangert @jeffwidman @ceache @a-ungurianu @StephenSorriaux It's been a while - any feedback on this? |
StephenSorriaux
left a comment
StephenSorriaux
left a comment
There was a problem hiding this comment.
After 2+ hours, I just finished going over most of the files. This one is huge. I still need to go over the "project" (constraints.txt, pyproject.toml`, etc.) files. I will do that later this week.
|
@StephenSorriaux yeah, sorry about the size. the settings that were there already were quite strict, and it was hard finding a root place to unpick from :-( Thanks for looking though. |
1a3456c to
17a48d4
Compare
|
Hopefully addressed comments. Not entirely sure why hound broke, as I can't see any messages from it. |
| zip_safe = false | ||
| include_package_data = true | ||
| packages = find: | ||
| install_requires = |
There was a problem hiding this comment.
This seems to force every user that pip install kazoo to install those 2 deps. Shouldn't those 2 only be part of typing extra (ie. user needs to pip install kazoo[typing]?
There was a problem hiding this comment.
Unfortunately, we need typing-extensions as that's where several types that are used outside of 'if TYPE_CHECKING' blocks and importing will fail. However, types-gevent was a mistake and I've removed that as it was already in the typing block.
There was a problem hiding this comment.
I think that's because some of the types used for types checking are not available in all versions of Python that kazoo supports (ParamSpec, Unpack, etc.). Bumping the min Python version we support to like 3.13 would make this dep useless. Probably something we would like to keep trace of.
|
Thank you again for this PR! I finalized my review with the files I did not check yet. Not sure what is going on with Hound too, I will merge this PR whether it's happy or not anyway. |
|
This is because this module is not intended to be used directly. See the
note at the top of
https://www.gevent.org/api/gevent.thread.html
It's fine to leave it as is, I think addressing this is another body of
work/PR.
Cheers,
On Sun, Jul 19, 2026, 05:58 Tom Tanner ***@***.***> wrote:
***@***.**** commented on this pull request.
------------------------------
In pyproject.toml
<#791 (comment)>:
> - 'kazoo.tests.test_retry',
- 'kazoo.tests.test_sasl',
- 'kazoo.tests.test_security',
- 'kazoo.tests.test_selectors_select',
- 'kazoo.tests.test_threading_handler',
- 'kazoo.tests.test_utils',
- 'kazoo.tests.test_watchers',
- 'kazoo.tests.util',
- 'kazoo.version'
-]
-ignore_errors = true
+ module = ["eventlet.*"]
+ follow_untyped_imports = true
+
+[[tool.mypy.overrides]]
+ module = ["gevent.thread"]
Oddly, no. I had to put this in because types-gevent provides types for
everything apart from from gevent.thread. I have no idea why this is the
case.
—
Reply to this email directly, view it on GitHub
<#791?email_source=notifications&email_token=AAIFTHREZA5NWJINA4PIS2T5FSSXFA5CNFSNUABKM5UWIORPF5TWS5BNNB2WEL2QOVWGYUTFOF2WK43UKJSXM2LFO4XTINZTGA3DENJVGU42M4TFMFZW63VHNVSW45DJN5XKKZLWMVXHJLDGN5XXIZLSL5RWY2LDNM#discussion_r3610416197>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAIFTHV6A3ZMJ7GPV7SFDOD5FSSXFAVCNFSNUABDKJSXA33TNF2G64TZHM2DIMRWGY4DSO2JONZXKZJ3GQ2DMMJRGE4TCMZZUF3AE>
.
Triage notifications, keep track of coding agent tasks and review pull
requests on the go with GitHub Mobile for iOS
<https://github.com/notifications/mobile/ios/AAIFTHRVJ46F6BJVJ2YHQM35FSSXFA5CNFSNUABKM5UWIORPF5TWS5BNNB2WEL2QOVWGYUTFOF2WK43UKJSXM2LFO4XTINZTGA3DENJVGU42M4TFMFZW63VHNVSW45DJN5XKKZLWMVXHJKTGN5XXIZLSL5UW64Y>
and Android
<https://github.com/notifications/mobile/android/AAIFTHRD6R52QE4POW7DD435FSSXFA5CNFSNUABKM5UWIORPF5TWS5BNNB2WEL2QOVWGYUTFOF2WK43UKJSXM2LFO4XTINZTGA3DENJVGU42M4TFMFZW63VHNVSW45DJN5XKKZLWMVXHJLTGN5XXIZLSL5QW4ZDSN5UWI>.
Download it today!
You are receiving this because you were mentioned.Message ID:
***@***.***>
Charles-Henri de Boysson
|
922b232 to
247db8c
Compare
|
@StephenSorriaux I think that's everything addressed. I've also added a fix for when none of the builds working in .github/workflows/testing.yml because the mirrors had changed (which happened to me just now). Fortunately the message when that happens is quite helpful. The py3.9/ZK3.6.4 failure seems to be one of those random failures that happens from time to time (cf #775) Somewhat weirdly, hound decided to work this time. I'm going to refrain from making comments about it being barking mad... |
StephenSorriaux
left a comment
StephenSorriaux
left a comment
There was a problem hiding this comment.
Looks good to me, thank you again for this.
|
Please rebase with |
This adds nearly full type hinting, and enables all but two of the mypy checks that were enabled in the mypy section in pyproject.toml. I hope to address the remaining checks, but it was proving tricky to address some of the 'typing.Any' uses. Also if you enable deprecation checks with --enable-error-code=deprecated You'll get warnings if you use deprecated options (at least as far as I was able to spot them!) This addresses python-zk#647 but there's not entirely strict checking as yet. Although I've tried very hard not to change any code for this, some `return None` statements have been added, which has caused the code coverage to drop as apparently those return paths were not covered by the testing. Also I've been less polite with the testing, and have changed the code in a number of places, including dropping test classes that did nothing, and adding asserts here and there to keep `mypy` happy as it sometimes doesn't spot a value could have been changed. I've added FIXMEs to address some of the more egregious type warning suppression that I did to avoid doing code changes, and intend to address those in another PR. A note: I've also had to suppress some errors in flake8 and use more general 'type: ignore' suppressions than I'd like because hound CI does something very very odd and tries to parse the text in the [] of the type: ignore[] comments as python, and generates some very peculiar messages.
247db8c to
37458f6
Compare
|
@StephenSorriaux rebased! |
|
woot! thank you |
3 participants
This adds nearly full type hinting, and enables all but two of the checks that were enabled in the
mypysection inpyproject.toml. I hope to address the remaining checks, but it was proving tricky to address some of thetyping.Anyuses without code changes.Also if you enable deprecation checks with
--enable-error-code=deprecated, you'll get warnings if you use deprecated options (at least as far as I was able to spot them!)This addresses #647 but there's not entirely strict checking as yet.
Although I've tried very hard not to change any code for this, some
return Nonestatements have been added, which has caused the code coverage to drop as apparently those return paths were not covered by the testing.Also I've been less polite with the testing, and have changed the code in a number of places, including dropping test classes that did nothing, and adding
assertshere and there to keepmypyhappy as it sometimes doesn't spot a value could have been changed.I've added
FIXMEs to indicate some of the more egregious type warning suppression that I did to avoid doing code changes, and intend to address those in another PR.I've also added tests for the
serialization.pymodule as some of it is a little unexpected and at one point I'd managed to break it but the resultant errors were very very unclear.A note: I've also had to suppress some errors in
flake8and use more general#type: ignoresuppressions than I'd like because hound CI does something very very odd and tries to parse the text in the[]of the#type: ignore[]comments as python, and generates some very peculiar messages.Why is this needed?
Type hinting makes it much easier for users to ensure they have the right parameters for functions.
Proposed Changes
Does this PR introduce any breaking change?
Well, it might break people who're already using mypy. And it's entirely possible I've got the wrong type hints (though hopefully actually using the type hinting in the tests has pulled all those out - it certainly picked up a couple of places where I'd got it wrong).
Codecov notes:
eventlet.py- this says it's missing coverage for theeventletevent_object,lock_objectandrlock_object.SequentialEventletHandlertest_eventlet_handlerwhich doesn't have enough in it to trigger lockingtest_cache.pywhich will useSequentialEventletHanderif it can't findSequentialGeventHandler- which as far as I can see it always will.connection.py- Apparently it doesn't drop out of the main loop innext_server. However, the only way I can see it doing that is if the hosts list is empty, which I don't think is a reasonable situation. Additionally, codecov on earlier code was showing the clearly unreachable code before the end of the loop as being covered.utils.py- Caused by an explicitreturn Noneat the end of a function instead of the implicit one. Cannot tell if this path was ever covered.lock.py- Caused by an explicitreturn Noneat the end of a function instead of the implicit one. Cannot tell if this path was ever covered.