[jira] [Created] (AIRFLOW-5539) default_args dictionary is mutated when passed as arugment to Airflow

["Guru Devanla (Jira)" <ji...@apache.org>]

Guru Devanla created AIRFLOW-5539:
-------------------------------------

             Summary: default_args dictionary is mutated when passed as arugment to Airflow
                 Key: AIRFLOW-5539
                 URL: https://issues.apache.org/jira/browse/AIRFLOW-5539
             Project: Apache Airflow
          Issue Type: Bug
          Components: DAG
    Affects Versions: 1.10.3
         Environment: Python 3.7.4, Linux
            Reporter: Guru Devanla


The `default_args` argument is a dictionary that can be passed into the DAG constructor while constructing a dag.  But, using the same variable to construct multiple DAGS breaks, since the `default_args` argument is mutated. Here is a minimal example:

 
{code:java}
// code placeholder

In [19]: from  functools import partial                                                                  In [20]: default_args=dict(start_date=datetime.datetime(2019, 8, 1))                                     In [21]: default_dag=partial(DAG, dag_id='test', default_args=default_args) 
                                                                               

In [23]: default_args['start_date'].tzinfo is None                                                       
Out[23]: True

In [24]: d1=default_dag()  # construct the DAG.

# Now check the value in default_args.                                                                              In [25]: default_args['start_date'].tzinfo is None                                                       
Out[25]: False

In [26]:  
{code}
 

My guess is that the deepcopy call to copy default_args is not working properly.

 

I am using Python 3.7.4.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)