udsy19 opened a new pull request, #76181:
URL: https://github.com/apache/airflow/pull/76181
## Why
The Task SDK `EventsTimetable.__init__`
(`task-sdk/src/airflow/sdk/definitions/timetables/events.py`) has the
`presorted` flag inverted:
```python
sorted(event_dates) if presorted else list(event_dates),
```
So with the default `presorted=False`, the dates are kept in the given
order, contrary to the docstring ("if True, event_dates will be assumed to be
in ascending order") and to the core
`airflow.timetables.events.EventsTimetable`, which sorts `if not presorted`, as
Airflow 2 did.
The order matters downstream: the encoder serializes `event_dates` as-is,
and the core `EventsTimetable.deserialize` rebuilds with `presorted=True`.
`next_dagrun_info` walks the list and takes the first event after the last run,
so with `event_dates=[Mar 1, Jan 1, Feb 1]` it schedules Mar 1 first and the
earlier events are silently skipped; the `"N events between X and Y"` summary
is also wrong, since it reads the first and last items.
(Today this path is masked by the serialization `KeyError: '_summary'` that
#69401 fixes; once that lands, unsorted event dates start dropping runs. In an
end-to-end encode → core decode → `next_dagrun_info` loop with that fix applied
by hand, `[Mar 1, Jan 1, Feb 1]` schedules only `2025-03-01` before this change
and all three, in order, after.)
## Fix
Swap the branches: `list(event_dates) if presorted else sorted(event_dates)`.
## Tests
New `task-sdk/tests/task_sdk/definitions/timetables/test_events.py`: dates
are sorted by default, and kept as given (list or iterator) with
`presorted=True`. All three fail on `main` and pass with the fix;
`tests/task_sdk/definitions/timetables/` passes.
---
##### Was generative AI tooling used to co-author this PR?
- [X] Yes. I reviewed the change and ran the tests above.
--
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]
