shahar1 opened a new pull request, #73040:
URL: https://github.com/apache/airflow/pull/73040
`validate-operators-init` flags every non-sanctioned read of a template
field in `__init__`, on the premise that such a read acts on the un-rendered
Jinja expression and belongs in `execute()`. `DataprocSubmitJobOperator` is
flagged for six reads that only copy template fields verbatim into
`start_trigger_args.trigger_kwargs`. With `start_from_trigger`, the scheduler
hands those kwargs straight to the triggerer, which renders the operator
template fields itself (`BaseTrigger.task_instance` setter in
`airflow-core/src/airflow/triggers/base.py`), so copying them un-rendered at
construct time is the intended data flow and there is nothing to move. The
entry cannot be cleared by an operator change without hiding the reads from the
hook (see the discussion on #70296 from 30 July and the approach in #72083).
This teaches the hook that verbatim reads inside `StartTriggerArgs(...)` or
`dataclasses.replace(self.start_trigger_args, ...)` are sanctioned (one level
into `{...}` / `dict(...)`; a transformation inside them is still flagged), and
removes the `DataprocSubmitJobOperator` exemption. The operator itself is
unchanged. The same narrowing approach was used for argument-provision checks
in #70505.
Verified locally: `prek run validate-operators-init --all-files` passes with
the change and fails on the six Dataproc reads with the hook change stashed and
the exemption removed; `uv run --project scripts pytest
scripts/tests/ci/prek/test_validate_operators_init.py` passes (53 tests, 4 new
cases).
related: #70296
---
##### Was generative AI tooling used to co-author this PR?
- [X] Yes — Claude Code (Fable 5.1)
Generated-by: Claude Code (Fable 5.1) following [the
guidelines](https://github.com/apache/airflow/blob/main/contributing-docs/05_pull_requests.rst#gen-ai-assisted-contributions)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
