potiuk commented on PR #44249:
URL: https://github.com/apache/airflow/pull/44249#issuecomment-2493571703

   > If we do get more reports, depending on their validity, we will then 
consider tweaking the current logic.
   
   Sure. It is indeed up to you. We are pretty well protected - we have 
"constraint mechanism" and we can restric airlfow 2.10.4 - and we are rather 
happy to adjust future versions of airflow, I was just merely suggesting that 
this change might have more consequences for you - so It was more of a friendly 
advice than complaint.
   
   But yes - I agree my assesment might be wrong.
   
   > Your usage of type annotations is inconsistent here, as you are mixing the 
new union syntax with Optional. Note that if this class were to be defined in 
the same module as 
[TaskInstancePydantic](https://github.com/apache/airflow/blob/1307e37bc6338c43aa5099752d012e244f8494ba/airflow/serialization/pydantic/taskinstance.py#L84),
 the issue would have appeared already.
   
   Yes -  but to be perfectly honest it's because of Pydantic. We have a rule 
in our pre-commit that enforces using new style type annotations. Because of 
inability of pydantic to handle it, we specifically converted our Pydantic 
classes to not use `from future import __annotations__` and we even had to add 
exeption to our rule: 
   
   https://github.com/apache/airflow/blob/main/pyproject.toml#L326
   
   So our "inconistent use" is exclusively because of Pydantic. And our use is 
perfectly fine according to current state of type annotations - both style of 
annotations can be inter-mixed and `from __future__ import annoutattion` as 
defined in https://peps.python.org/pep-0563/ . And while I know there are a lot 
of controversies around that PEP and the way how it works will likely change in 
the future, this is the current state of affairs, and we are following the 
accepted PEPs (while Pydantic does not really handle them properly and we had 
to introduce inconsistency in order to workaround it).
   
   So technically speaking it's because of Pydantic we are inconsistent. 
   
   And I am not complaining and not trying to heat it up, I am just stating the 
fact - explaining that well, it's Pydantic who is the root cause of that 
inconsistency, not Airflow.
   
   So really - @Viicos - please treat it as a friendly advice, that maybe it's 
a good idea to consider supporting that case. While we will handle it ourselves 
(luckily our constraint mechanism we introduce many years ago and educate our 
users to use them will prevent a lot of our users from being affected). It's 
really a maintainer-to-maintainer friendly suggestion that this might hit you 
back.
   


-- 
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