Decorate all operators with @apply_defaults

[arimbr]

It seems to be a common practice to decorate all operators with @appy_defaults

Check for example, BigQueryToBigQueryOperator

The apply_defaults decorator apart from allowing to pass default_args to a task. It does some argument validation like:

• Forcing to use keyword arguments when initializing operators
• Checking for required arguments

That can be useful to debug operators.

Locally I tested for example if I removed the required step argument. Error message is a bit different.

BEFORE

AFTER

Good thing is that if someone does something like:

train_model = ValohaiSubmitExecutionOperator(
    'train_model',
    'tensorflow-example',
    step='Train model (MNIST)',

It will show the error:

Broken DAG: [/home/skillup/skillup/airflow-valohai-plugin/examples/dags/example_valohai_dag.py] Use keyword arguments when initializing operators

to force using the better:

train_model = ValohaiSubmitExecutionOperator(
    task_id='train_model',
    project_name='tensorflow-example',
    step='Train model (MNIST)',
...