Add test to verify parameters match between operator and decorator

Body

We shouldn't have notes like:

airflow/providers/cncf/kubernetes/src/airflow/providers/cncf/kubernetes/operators/pod.py

Lines 236 to 237 in 6508f12

    
           # !!! Changes in KubernetesPodOperator's arguments should be also reflected in !!! 
        
           #  - airflow-core/src/airflow/decorators/__init__.pyi  (by a separate PR)

this is very easy to miss both by PR author and the reviewer

The task:
Add tests to all existed decorators to make sure the parameter list they accept match to the operator.

Committer

I acknowledge that I am a maintainer/committer of the Apache Airflow project.

Hi!
Can I take this issue and work on it?

@gomtangman

Hi [@gomtangman], I noticed you’re assigned to this issue. I’d love to contribute here too—are you open to collaboration? I’m happy to pair-program, tackle a subtask, or review your work. Let me know how I can assist!

@Shweta-281

@Shweta-281 Thank you for your suggestion
I am writing a test now. So can you review it when I make a pull request?

@Shweta-281

Hi @Shweta-281
I became busy doing my work to keep up with this issue.
So if you are still interested in this issue, can you work on it instead of me?
I thought it would be good to test whether the test is covering all the decorators.
So, I searched for a way to retrieve all the decorators but I failed to find one.
If you can find a way to ensure all the decorators are tested, I think that'll increase the quality of tests.

I want to take this issue and work on it. can I ?

@Debasishpradhan28

@Debasishpradhan28
Yes, please take it

@Bowrna

Looks like @Bowrna is already on it. Assigning to her.

Hi can I take this issue?

@amoghrajesh

@amoghrajesh @isharghura the work is done and it was failing in static checks when i raised PR alone. #51048
may be i can try running it again and verify it before next wednesday.

Hi, I'd like to work on this issue as my first contribution to Airflow. Is it still available ?

@wfatih

I would like to take this one, and I measured the drift first so the scope is based on numbers rather than guesses.

A static scan — ast only, nothing imported — of task-sdk/src/airflow/sdk/definitions/decorators/__init__.pyi against the operator behind each task-decorators entry registered in provider.yaml finds:

11 registered decorators with no stub entry at all: agent, analytics, llm, llm_branch, llm_file_analysis, llm_schema_compare, llm_sql, sftp_sensor, snowpark, sql, stub. @task.llm(...) and friends fall back to __getattr__ today, so they get no parameter checking at all.
Parameter drift in 7 of the 13 decorators that do have a stub. A few examples: python is missing templates_exts, which PythonOperator.__init__ accepts; virtualenv and external_python are missing string_args, templates_exts and expect_airflow; kubernetes and kubernetes_cmd are each missing around thirteen KubernetesPodOperator parameters, including cmds, callbacks, logging_interval and runtime_class_name.

Not every difference is drift. use_dill, serializer and python_command exist only on the decorator side, and bash_command, command and arguments are deliberately supplied by the decorated function rather than the caller, so the check needs an explicit exemption list instead of a strict equality assertion.

On where the check should live: the earlier attempt in #51048 needed Airflow importable to compare signatures, which pushed it into an in-container script and is where it got stuck. Comparing the stub against the operator __init__ can be done entirely statically, so it fits a plain scripts/ci/prek/ hook with no Airflow dependency and no container, runnable on a clean checkout. I have a working prototype of that comparison, which is where the numbers above come from.

Three questions before I write the real thing:

Is a static prek hook the shape you want, or would you still rather have this in the test suite?
Are the 11 missing decorators in scope here, or a follow-up? Adding them means writing stubs for provider decorators that never had one.
For the drift the check finds, should the same PR fix the stubs — the kubernetes ones are the bulk of it — or should the check land with those exempted and the fixes follow per provider?

Drafted-by: Claude Opus 5 (1M context); reviewed by @wfatih before posting

I would like to take this one and I've opened a draft PR. I did a check but not yet find an active development. Would appreciate any review feedback. Thanks!

Navigation Menu

Repository navigation

Add test to verify parameters match between operator and decorator #48448

Body

Committer

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

	# !!! Changes in KubernetesPodOperator's arguments should be also reflected in !!!
	# - airflow-core/src/airflow/decorators/__init__.pyi (by a separate PR)

Add test to verify parameters match between operator and decorator #48448

Description

Body

Committer

Activity

gomtangman commented on Mar 27, 2025

Shweta-281 commented on Mar 29, 2025

gomtangman commented on Mar 29, 2025

gomtangman commented on Apr 3, 2025

Debasishpradhan28 commented on Apr 8, 2025

gomtangman commented on Apr 8, 2025

amoghrajesh commented on Jun 13, 2025

isharghura commented on Nov 12, 2025

Bowrna commented on Nov 13, 2025

Stuaarts commented on Aug 31, 2026

wfatih commented on Sep 20, 2026

sjyangkevin commented on Oct 14, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions