GitHub user tgrandje added a comment to the discussion: sqlalchemy 2 set in a
virtualenv breaks variable access
I know that (I think I stated this first hand with the reference to the airflow
3 roadmap) ; that's even the reason why I had'nt posted the traceback in the
first place.
I still this needs to be referenced as :
* using a virtualenv with sqlalchemy > 2 in airflow it currently the only way
to work with recent data libraries (like polars), and that's going to stay as
it is until airflow 3 is released ;
* we have unconsistent behaviour, as variables set with the
`os.environ["AIRFLOW_VAR_..."]` **are** accessible (is this through the
SecretCache?). Again, I know this is *by design*, but I still think this info
is needed.
For instance, that dag works just fine:
```
from airflow.decorators import dag, task
@dag()
def sqlalchemy_test():
import os
os.environ["AIRFLOW_VAR_BACON"] = "sausage"
@task
def get_variable():
from airflow.models import Variable
print(Variable.get("bacon"))
@task.virtualenv(requirements=["sqlalchemy>2.0.0"])
def get_variable2():
from airflow.models import Variable
print(Variable.get("bacon"))
get_variable()
get_variable2()
sqlalchemy_test()
```
GitHub link:
https://github.com/apache/airflow/discussions/46266#discussioncomment-12003941
----
This is an automatically sent email for [email protected].
To unsubscribe, please send an email to: [email protected]
