Aditya-XR opened a new pull request, #76085:
URL: https://github.com/apache/airflow/pull/76085
Calling a `@task_group` function three or more times inside another
`TaskGroup` stops the Dag from importing. Two calls work, and three calls work
at the top level of a Dag.
```python
from airflow.sdk import DAG, TaskGroup, task, task_group
@task
def work():
pass
@task_group
def inner():
work()
with DAG("nested_task_group_repro", schedule=None):
with TaskGroup("outer"):
for _ in range(3):
inner()
```
```
airflow.sdk.exceptions.DuplicateTaskIdFound: Task Group id 'outer.inner__1'
has already been added to the Dag
```
This reproduces on 3.3.2 and on `main`. In a Dag file it shows up as an
import error.
`_check_for_group_id_collisions` detects the collision with the prefixed id
(`outer.inner`), but looked for taken suffixes with a regex built from the bare
id (`^inner__\d+$`). `used_group_ids` stores prefixed ids, so `outer.inner__1`
never matched and the third call got `__1` again. The lookup now uses the
prefixed id, as `get_unique_task_id` already does for tasks. It is escaped
because the dots in a prefix would otherwise match any character.
**Behaviour note:** the same mismatch also skewed numbering without
crashing. If `ingest` is called three times at the top level and twice inside
`eu`, the second nested group was `eu.ingest__3` and is now `eu.ingest__1`, so
task ids in a Dag with that exact shape change.
**Testing**
- New parametrized `test_repeated_task_group_in_parent_gets_next_suffix`:
all 4 cases fail without the fix and pass with it.
- `test_taskgroup.py`, `decorators/test_task_group.py` and
`_internal/test_loop.py` in task-sdk: 126 passed.
- `test_task_group.py` and `test_dag_serialization.py` in airflow-core: 289
passed.
- All of `task-sdk/tests`: 4241 passed, 7 failed. The same 7 fail without
this change: 3 need `deltalake`, which isn't installed here, plus 3
`test_param` format cases and `test_hook_values`.
- A Dag with two groups that each call a `@task_group` three times, loaded
through `DagBag`: import error without the fix; with it the Dag parses and its
serialization round trip matches.
- `prek` pre-commit and manual stages pass, including mypy for task-sdk.
Tests ran with `uv run`, not `breeze`, because I have no Docker available.
The `breeze testing` suites and `breeze build-docs` were not run.
related: #41596
---
##### Was generative AI tooling used to co-author this PR?
- [X] Yes — Claude Code (Opus 5.5)
Generated-by: Claude Code (Opus 5.5) following [the
guidelines](https://github.com/apache/airflow/blob/main/contributing-docs/05_pull_requests.rst#gen-ai-assisted-contributions)
--
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]
