I didn't expect to see source code changes here

Please elaborate. This new ruff inspection I'm working on is a "best practices" change.

those are failing some mypy checks

I'll address those as soon as I can

I didn't expect to see source code changes here

Please elaborate. This new ruff inspection I'm working on is a "best practices" change.

Yes the best practice ruff change looks great :) I just expected updates to Dags only (to fix cases the new rule would have caught). I'm not sure why any changes happened to actual src code (in this case the normalise_json_content code). Is that because that function is called downstream inside the operator in the databricks Dag you updated?

those are failing some mypy checks

I'll address those as soon as I can

Sounds good! 😄

I'm not sure why any changes happened to actual src code (in this case the normalise_json_content code). Is that because that function is called downstream inside the operator in the databricks Dag you updated?

This is because a certain unit test was asserting an implementation detail:

    File "/opt/airflow/providers/databricks/tests/system/databricks/example_databricks.py", line 118, in <module>                                                                                             
      run_now = DatabricksRunNowOperator(                                                                                                                                                                     
    File "/opt/airflow/task-sdk/src/airflow/sdk/bases/operator.py", line 548, in apply_defaults                                                                                                               
      result = func(self, **kwargs, default_args=default_args)                                                                                                                                                
    File "/opt/airflow/providers/databricks/src/airflow/providers/databricks/operators/databricks.py", line 897, in __init__                                                                                  
      self.json = normalise_json_content(self.json)                                                                                                                                                           
    File "/opt/airflow/providers/databricks/src/airflow/providers/databricks/utils/databricks.py", line 45, in normalise_json_content                                                                         
      return {k: normalise(v, f"{json_path}[{k}]") for k, v in content.items()}                                                                                                                               
    File "/opt/airflow/providers/databricks/src/airflow/providers/databricks/utils/databricks.py", line 45, in <dictcomp>                                                                                     
      return {k: normalise(v, f"{json_path}[{k}]") for k, v in content.items()}                                                                                                                               
    File "/opt/airflow/providers/databricks/src/airflow/providers/databricks/utils/databricks.py", line 48, in normalise_json_content                                                                         
      raise AirflowException(msg)                                                                                                                                                                             
  airflow.sdk.exceptions.AirflowException: Type <class 'airflow.sdk.definitions.xcom_arg.PlainXComArg'> used for parameter json[job_id] is not a number or a string     

And Claude's analysis of this was:

The AIR004 fix changed job_id from a Jinja template string to jobs_create_named.output (a PlainXComArg). But the operator's __init__ calls normalise_json_content(self.json) at parse time,
  which doesn't handle XComArg objects — it only accepts strings, numbers, booleans, lists, and dicts.

I'm not sure why any changes happened to actual src code (in this case the normalise_json_content code). Is that because that function is called downstream inside the operator in the databricks Dag you updated?

This is because a certain unit test was asserting an implementation detail:

    File "/opt/airflow/providers/databricks/tests/system/databricks/example_databricks.py", line 118, in <module>                                                                                             
      run_now = DatabricksRunNowOperator(                                                                                                                                                                     
    File "/opt/airflow/task-sdk/src/airflow/sdk/bases/operator.py", line 548, in apply_defaults                                                                                                               
      result = func(self, **kwargs, default_args=default_args)                                                                                                                                                
    File "/opt/airflow/providers/databricks/src/airflow/providers/databricks/operators/databricks.py", line 897, in __init__                                                                                  
      self.json = normalise_json_content(self.json)                                                                                                                                                           
    File "/opt/airflow/providers/databricks/src/airflow/providers/databricks/utils/databricks.py", line 45, in normalise_json_content                                                                         
      return {k: normalise(v, f"{json_path}[{k}]") for k, v in content.items()}                                                                                                                               
    File "/opt/airflow/providers/databricks/src/airflow/providers/databricks/utils/databricks.py", line 45, in <dictcomp>                                                                                     
      return {k: normalise(v, f"{json_path}[{k}]") for k, v in content.items()}                                                                                                                               
    File "/opt/airflow/providers/databricks/src/airflow/providers/databricks/utils/databricks.py", line 48, in normalise_json_content                                                                         
      raise AirflowException(msg)                                                                                                                                                                             
  airflow.sdk.exceptions.AirflowException: Type <class 'airflow.sdk.definitions.xcom_arg.PlainXComArg'> used for parameter json[job_id] is not a number or a string     

And Claude's analysis of this was:

The AIR004 fix changed job_id from a Jinja template string to jobs_create_named.output (a PlainXComArg). But the operator's __init__ calls normalise_json_content(self.json) at parse time,
  which doesn't handle XComArg objects — it only accepts strings, numbers, booleans, lists, and dicts.

Thanks for the context and details :)