fruch commented on PR #36755:
URL: https://github.com/apache/airflow/pull/36755#issuecomment-1930723711

   > > nice to see more pythonista pushing the envolpe and adopting 3.12, I 
know it's teidius but it really help the community in general, thanks for those 
efforts.
   > 
   > Thanks @fruch . Much appreciated you found time to chime in here :). I 
kind-of expected 4-5 months after 3.12 was released when we will have the 
result of the work implemented by others to get the 3.12 possible to be 
supported by Airflow. And we are **nearly** there :D. 
   > 
   > Happy to get a bit more clarification if you can ?
   > 
   > ## Context: 
   > 
   > We have very little knowldege of Cassandra - not enough to be able to 
influence our user's choices (Airflow is pretty special - we have 90+ 
integrations with things out there so in case of many of integrations we have 
to rely on our users testing and verifying that the integrations work for 
them). But it's not an issue for now at least - we will simply wait until the 
"default" implementation of Cassandra for Python 3.12 will work, and then 
enable it for Python 3.12, this is usually what we do. 
   > 
   > ## Current approach
   > 
   > We have the whole mechanism in place to easily disable/enable selected 
Python versions for selected integrations, so the only effect will be that our 
"providers" for Cassandra will not have Python 3.12 support and users using 
Python 3.12 won't be able to install the integration (they can still, if they 
want get our integration, copy&paste and extend it so it's not fully blocking 
it for them, it's more they won't be able to use the default 
`apache-airflow-providers-apache-cassandra` package and get it working 
`out-of-the-box` - which in a way is what you suggest. Possibly even some of 
our Cassandra users will contribute back a fix after testing and switching to a 
different reactor.
   > 
   > ## Maybe libev by default?
   > 
   > But if you say it's a matter of installing libev (anything else? any 
configuration needed) and it will **just** work then maybe the right solution 
is to just install libev in our containers (easy) and let the test pass 
(hopefully easy) and add the docs that Python 3.12 has libev as dependency (we 
can even add a warning or error to check it when Hook gets initialized). 
   > 
   > Would that make sense? Is there high confidence tthat  it will "just work" 
when we switch ? 
   
   As I said if packages are always installed as wheel is would work out of the 
box, libev should be baked into the wheel.
   
   If not, and you can enforce libev install before install from pip, it would 
work as well, see 
https://docs.datastax.com/en/developer/python-driver/3.14/installation/#libev-support
 for more details of need thing to be able to compile from source.
   
   
   
   
   
   > 
   > Apologies for not knowing more there, but - as I say - with 90+ 
integrations out there, we are by no means experts in many of those :) 
   
   
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to