Add support for Python 3.10 by dropping event loop handling #180
Conversation
|
This is an issue I've also been working on, here's my implementation of the same issue fix: dc3c7ae It looks like we went about different ways of doing the same result. It's worth considering keeping the eventloop around and simply adding a deprecation warning for now I think, and instead of Pydle running and getting its own eventloop simply relying on the fact that Pydle will always be called from an existing asyncio loop. This also allows hijacking of a number of existing code fascets, requiring a less intrusive method of establishing 3.10 support. Just food for thought, I don't know if either one is "right" or "better", just another way of going about things. |
There was a problem hiding this comment.
Hey, thanks for the PR!
I am not convinced that dropping 3.6 support is a good idea for the time being, as it is still a very common python release.
Otherwise, I very much like this approach!
@Rixxan Brings up a good point, whether we should go through a deprecation cycle first, or simply do a major release as this is PR is a breaking API change.
felixonmars
force-pushed
the
python3.10
branch
from
aaa4f9d to
ddb66aa
Compare
There was a problem hiding this comment.
Thanks for the update, Looks like something went wrong resolving the conflict in the readme.
once that is fixed I will merge this.
@Rixxan Thanks for the softer deprecation suggestion, however at this point I feel it would be better to accept this PR instead. Consequently, this PR will necessitate a major release as per SEMVER since it breaks a piece of the public API.
Implements suggestions in shizmob#162 (comment) to remove Pydle's internal handling of the event loop object entirely.
felixonmars
force-pushed
the
python3.10
branch
from
ddb66aa to
e09f476
Compare
| @@ -65,9 +65,7 @@ Furthermore, since pydle is simply `asyncio`-based, you can run the client in yo | |||
| import asyncio | |||
|
|
|||
| client = MyOwnBot('MyBot') | |||
| loop = asyncio.get_event_loop() | |||
There was a problem hiding this comment.
This appears to be incorrect. Changing the entry logic for my own implementation of pydle results in a hung program, which immediately terminates uncleanly.
However, testing on Python 3.8 if I do not change my existing entry logic, the program starts properly. (Still stability-testing.)
The same behavior replicates on 3.10.
https://build.opensuse.org/request/show/1067570 by user dgarcia + dimstar_suse - Delete python-pydle-poetry-syntax.patch - Skip python 310 and python 311, pydle doesn't python 3.10 gh#shizmob/pydle#162, There's a WIP PR but looks like it's not good enough yet to be applied gh#shizmob/pydle#180 - Update to version 1.0.1 * [Docs] Document Dropping of 3.5 by @Rixxan in #179 * Adjust python version inequality string for Poetry by @txtsd in #182 * bump version to 1.0.1 by @theunkn0wn1 in #183
|
I haven't tested this yet as I'm just starting a brand new project but I think this might be the "more correct" way to handle the asyncio changes in py3.10+
If there's no event loop it will create a new event loop and set it as the main one |
|
@Rixxan you might wanna have a look :) |
|
TheHolyRoger's branch fix also seems to work cleanly. |
Implements suggestions in
#162 (comment) to
remove Pydle's internal handling of the event loop object entirely.
All tests are passing (although they are passing without these changes, too). I have been running my own project with this change and so far so good.
I am not sure about the connection pool though (I have never used it). This change should have also stopped it from having multiple event loops at all. So I have updated the
stop()method from stopping the whole event loop to justasyncio.runthedisconnect()method.