Drop selector_loop module#1587
Merged
Merged
Conversation
The rewritten ``JavaCoordinator`` (``airflow.sdk.coordinators.java``) lets the JVM connect directly to two listening sockets (``comm`` and ``logs``) and uses the accepted sockets as the supervisor's ``stdin`` / log pipes straight up. There is no bytes-bridge between two sockets, so ``make_raw_forwarder`` -- the only helper the extracted ``selector_loop`` module added beyond what was already inline in ``supervisor.py`` -- has no caller. ``_JavaActivitySubprocess`` reuses ``_register_pipe_readers`` and ``_close_unused_sockets`` via subclassing ``ActivitySubprocess``; both methods existed before the extraction and remain in ``supervisor.py`` after this revert. The inline selector dispatch loop and ``make_buffered_socket_reader`` come back into ``supervisor.py`` so the existing call sites (including the ``triggerer_job_runner`` re-export) keep working unchanged. This reverts commit 56464a8 ("Add common selector loop utilities for socket I/O handling for subprocesses") and deletes ``test_selector_loop.py`` introduced by 4b80753.
jason810496
changed the title
There was a problem hiding this comment.
Pull request overview
This PR reverts the previously introduced selector-loop modularization by removing the dedicated selector_loop module and folding the remaining needed logic back into the Task SDK supervisor, aligning with the newer JavaCoordinator direction.
Changes:
- Deleted
airflow.sdk.execution_time.selector_loopand its associated unit tests. - Inlined the selector event servicing logic into
WatchedSubprocess._service_subprocessand movedmake_buffered_socket_readerintosupervisor.py. - Updated coordinator module docs to remove references to the removed selector loop module.
Reviewed changes
Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.
| File | Description |
|---|---|
| task-sdk/tests/task_sdk/execution_time/test_selector_loop.py | Removes selector-loop focused tests (needs replacement coverage for still-used buffering helper). |
| task-sdk/src/airflow/sdk/execution_time/supervisor.py | Inlines selector loop behavior and re-homes make_buffered_socket_reader; adjusts typing/imports. |
| task-sdk/src/airflow/sdk/execution_time/selector_loop.py | Deletes the selector-loop utility module entirely. |
| task-sdk/src/airflow/sdk/execution_time/coordinator.py | Removes documentation references to selector_loop.service_selector. |
Comments suppressed due to low confidence (1)
task-sdk/src/airflow/sdk/execution_time/supervisor.py:719
- The return type annotation of
_create_log_forwarderis incorrect: it returns the(handler, on_close)tuple frommake_buffered_socket_reader(...), not a singleCallable[[socket], bool]. This mismatch will confuse callers and type checkers; change the annotation back to the appropriate tuple/type alias (e.g.SelectorCallback).
def _create_log_forwarder(self, loggers, name, log_level=logging.INFO) -> Callable[[socket], bool]:
"""Create a socket handler that forwards logs to a logger."""
There was a problem hiding this comment.
Pull request overview
Copilot reviewed 5 out of 5 changed files in this pull request and generated 2 comments.
Comments suppressed due to low confidence (2)
task-sdk/src/airflow/sdk/execution_time/supervisor.py:932
- The new inlined selector loop includes several “example”/narration comments that restate the code paths (EOF handling, BrokenPipeError behavior, etc.). These comments are likely to become stale and make the loop harder to scan; consider removing them or replacing them with a short summary of the callback contract.
# Example of handler behavior:
# If the subprocess writes "Hello, World!" to stdout:
# - `socket_handler` reads and processes the message.
# - If EOF is reached, the handler returns False to signal no more reads are expected.
# - BrokenPipeError should be caught and treated as if the handler returned false, similar
task-sdk/src/airflow/sdk/execution_time/supervisor.py:2014
make_buffered_socket_readerwas moved intosupervisor.pywhile the dedicated selector-loop test module was removed. Since this function is still used for log/request sockets and has non-trivial buffering/EOF behavior, it should retain direct unit tests (e.g. relocate the previous tests totest_supervisor.pyor a new focused test module that imports it fromsupervisor).
def make_buffered_socket_reader(
gen: Generator[None, bytes | bytearray, None],
on_close: Callable[[socket], None],
buffer_size: int = 4096,
):
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 82e4dfe277
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Member
|
Is this a direct revert? (I’ll skip reviewing the details if it is.) |
Member
Author
It's necessary to manual revert instead of direct revert as we will encounter conflict for direct revert. |
uranusjr
approved these changes
May 20, 2026
jason810496
merged commit 5abb82e
into
task-sdk/feature/coordinator-interface
4 checks passed
jason810496
deleted the
task-sdk/feature/coordinator-interface-drop-selector-loop
branch
jason810496
added a commit
that referenced
this pull request
May 22, 2026
* Drop selector_loop module; new JavaCoordinator does not need it The rewritten ``JavaCoordinator`` (``airflow.sdk.coordinators.java``) lets the JVM connect directly to two listening sockets (``comm`` and ``logs``) and uses the accepted sockets as the supervisor's ``stdin`` / log pipes straight up. There is no bytes-bridge between two sockets, so ``make_raw_forwarder`` -- the only helper the extracted ``selector_loop`` module added beyond what was already inline in ``supervisor.py`` -- has no caller. ``_JavaActivitySubprocess`` reuses ``_register_pipe_readers`` and ``_close_unused_sockets`` via subclassing ``ActivitySubprocess``; both methods existed before the extraction and remain in ``supervisor.py`` after this revert. The inline selector dispatch loop and ``make_buffered_socket_reader`` come back into ``supervisor.py`` so the existing call sites (including the ``triggerer_job_runner`` re-export) keep working unchanged. This reverts commit 56464a8 ("Add common selector loop utilities for socket I/O handling for subprocesses") and deletes ``test_selector_loop.py`` introduced by 4b80753. * Address copilot's comments
jason810496
added a commit
that referenced
this pull request
May 23, 2026
* Drop selector_loop module; new JavaCoordinator does not need it The rewritten ``JavaCoordinator`` (``airflow.sdk.coordinators.java``) lets the JVM connect directly to two listening sockets (``comm`` and ``logs``) and uses the accepted sockets as the supervisor's ``stdin`` / log pipes straight up. There is no bytes-bridge between two sockets, so ``make_raw_forwarder`` -- the only helper the extracted ``selector_loop`` module added beyond what was already inline in ``supervisor.py`` -- has no caller. ``_JavaActivitySubprocess`` reuses ``_register_pipe_readers`` and ``_close_unused_sockets`` via subclassing ``ActivitySubprocess``; both methods existed before the extraction and remain in ``supervisor.py`` after this revert. The inline selector dispatch loop and ``make_buffered_socket_reader`` come back into ``supervisor.py`` so the existing call sites (including the ``triggerer_job_runner`` re-export) keep working unchanged. This reverts commit 56464a8 ("Add common selector loop utilities for socket I/O handling for subprocesses") and deletes ``test_selector_loop.py`` introduced by 4b80753. * Address copilot's comments
jason810496
added a commit
that referenced
this pull request
May 23, 2026
* Drop selector_loop module; new JavaCoordinator does not need it The rewritten ``JavaCoordinator`` (``airflow.sdk.coordinators.java``) lets the JVM connect directly to two listening sockets (``comm`` and ``logs``) and uses the accepted sockets as the supervisor's ``stdin`` / log pipes straight up. There is no bytes-bridge between two sockets, so ``make_raw_forwarder`` -- the only helper the extracted ``selector_loop`` module added beyond what was already inline in ``supervisor.py`` -- has no caller. ``_JavaActivitySubprocess`` reuses ``_register_pipe_readers`` and ``_close_unused_sockets`` via subclassing ``ActivitySubprocess``; both methods existed before the extraction and remain in ``supervisor.py`` after this revert. The inline selector dispatch loop and ``make_buffered_socket_reader`` come back into ``supervisor.py`` so the existing call sites (including the ``triggerer_job_runner`` re-export) keep working unchanged. This reverts commit 56464a8 ("Add common selector loop utilities for socket I/O handling for subprocesses") and deletes ``test_selector_loop.py`` introduced by 4b80753. * Address copilot's comments
uranusjr
pushed a commit
that referenced
this pull request
May 25, 2026
* Drop selector_loop module; new JavaCoordinator does not need it The rewritten ``JavaCoordinator`` (``airflow.sdk.coordinators.java``) lets the JVM connect directly to two listening sockets (``comm`` and ``logs``) and uses the accepted sockets as the supervisor's ``stdin`` / log pipes straight up. There is no bytes-bridge between two sockets, so ``make_raw_forwarder`` -- the only helper the extracted ``selector_loop`` module added beyond what was already inline in ``supervisor.py`` -- has no caller. ``_JavaActivitySubprocess`` reuses ``_register_pipe_readers`` and ``_close_unused_sockets`` via subclassing ``ActivitySubprocess``; both methods existed before the extraction and remain in ``supervisor.py`` after this revert. The inline selector dispatch loop and ``make_buffered_socket_reader`` come back into ``supervisor.py`` so the existing call sites (including the ``triggerer_job_runner`` re-export) keep working unchanged. This reverts commit 56464a8 ("Add common selector loop utilities for socket I/O handling for subprocesses") and deletes ``test_selector_loop.py`` introduced by 4b80753. * Address copilot's comments
uranusjr
pushed a commit
that referenced
this pull request
May 26, 2026
* Drop selector_loop module; new JavaCoordinator does not need it The rewritten ``JavaCoordinator`` (``airflow.sdk.coordinators.java``) lets the JVM connect directly to two listening sockets (``comm`` and ``logs``) and uses the accepted sockets as the supervisor's ``stdin`` / log pipes straight up. There is no bytes-bridge between two sockets, so ``make_raw_forwarder`` -- the only helper the extracted ``selector_loop`` module added beyond what was already inline in ``supervisor.py`` -- has no caller. ``_JavaActivitySubprocess`` reuses ``_register_pipe_readers`` and ``_close_unused_sockets`` via subclassing ``ActivitySubprocess``; both methods existed before the extraction and remain in ``supervisor.py`` after this revert. The inline selector dispatch loop and ``make_buffered_socket_reader`` come back into ``supervisor.py`` so the existing call sites (including the ``triggerer_job_runner`` re-export) keep working unchanged. This reverts commit 56464a8 ("Add common selector loop utilities for socket I/O handling for subprocesses") and deletes ``test_selector_loop.py`` introduced by 4b80753. * Address copilot's comments
uranusjr
pushed a commit
that referenced
this pull request
May 26, 2026
* Drop selector_loop module; new JavaCoordinator does not need it The rewritten ``JavaCoordinator`` (``airflow.sdk.coordinators.java``) lets the JVM connect directly to two listening sockets (``comm`` and ``logs``) and uses the accepted sockets as the supervisor's ``stdin`` / log pipes straight up. There is no bytes-bridge between two sockets, so ``make_raw_forwarder`` -- the only helper the extracted ``selector_loop`` module added beyond what was already inline in ``supervisor.py`` -- has no caller. ``_JavaActivitySubprocess`` reuses ``_register_pipe_readers`` and ``_close_unused_sockets`` via subclassing ``ActivitySubprocess``; both methods existed before the extraction and remain in ``supervisor.py`` after this revert. The inline selector dispatch loop and ``make_buffered_socket_reader`` come back into ``supervisor.py`` so the existing call sites (including the ``triggerer_job_runner`` re-export) keep working unchanged. This reverts commit 56464a8 ("Add common selector loop utilities for socket I/O handling for subprocesses") and deletes ``test_selector_loop.py`` introduced by 4b80753. * Address copilot's comments
uranusjr
pushed a commit
that referenced
this pull request
May 26, 2026
* Drop selector_loop module; new JavaCoordinator does not need it The rewritten ``JavaCoordinator`` (``airflow.sdk.coordinators.java``) lets the JVM connect directly to two listening sockets (``comm`` and ``logs``) and uses the accepted sockets as the supervisor's ``stdin`` / log pipes straight up. There is no bytes-bridge between two sockets, so ``make_raw_forwarder`` -- the only helper the extracted ``selector_loop`` module added beyond what was already inline in ``supervisor.py`` -- has no caller. ``_JavaActivitySubprocess`` reuses ``_register_pipe_readers`` and ``_close_unused_sockets`` via subclassing ``ActivitySubprocess``; both methods existed before the extraction and remain in ``supervisor.py`` after this revert. The inline selector dispatch loop and ``make_buffered_socket_reader`` come back into ``supervisor.py`` so the existing call sites (including the ``triggerer_job_runner`` re-export) keep working unchanged. This reverts commit 56464a8 ("Add common selector loop utilities for socket I/O handling for subprocesses") and deletes ``test_selector_loop.py`` introduced by 4b80753. * Address copilot's comments
uranusjr
pushed a commit
that referenced
this pull request
May 27, 2026
* Drop selector_loop module; new JavaCoordinator does not need it The rewritten ``JavaCoordinator`` (``airflow.sdk.coordinators.java``) lets the JVM connect directly to two listening sockets (``comm`` and ``logs``) and uses the accepted sockets as the supervisor's ``stdin`` / log pipes straight up. There is no bytes-bridge between two sockets, so ``make_raw_forwarder`` -- the only helper the extracted ``selector_loop`` module added beyond what was already inline in ``supervisor.py`` -- has no caller. ``_JavaActivitySubprocess`` reuses ``_register_pipe_readers`` and ``_close_unused_sockets`` via subclassing ``ActivitySubprocess``; both methods existed before the extraction and remain in ``supervisor.py`` after this revert. The inline selector dispatch loop and ``make_buffered_socket_reader`` come back into ``supervisor.py`` so the existing call sites (including the ``triggerer_job_runner`` re-export) keep working unchanged. This reverts commit 56464a8 ("Add common selector loop utilities for socket I/O handling for subprocesses") and deletes ``test_selector_loop.py`` introduced by 4b80753. * Address copilot's comments
uranusjr
pushed a commit
that referenced
this pull request
May 27, 2026
* Drop selector_loop module; new JavaCoordinator does not need it The rewritten ``JavaCoordinator`` (``airflow.sdk.coordinators.java``) lets the JVM connect directly to two listening sockets (``comm`` and ``logs``) and uses the accepted sockets as the supervisor's ``stdin`` / log pipes straight up. There is no bytes-bridge between two sockets, so ``make_raw_forwarder`` -- the only helper the extracted ``selector_loop`` module added beyond what was already inline in ``supervisor.py`` -- has no caller. ``_JavaActivitySubprocess`` reuses ``_register_pipe_readers`` and ``_close_unused_sockets`` via subclassing ``ActivitySubprocess``; both methods existed before the extraction and remain in ``supervisor.py`` after this revert. The inline selector dispatch loop and ``make_buffered_socket_reader`` come back into ``supervisor.py`` so the existing call sites (including the ``triggerer_job_runner`` re-export) keep working unchanged. This reverts commit 56464a8 ("Add common selector loop utilities for socket I/O handling for subprocesses") and deletes ``test_selector_loop.py`` introduced by 4b80753. * Address copilot's comments
uranusjr
pushed a commit
that referenced
this pull request
May 27, 2026
* Drop selector_loop module; new JavaCoordinator does not need it The rewritten ``JavaCoordinator`` (``airflow.sdk.coordinators.java``) lets the JVM connect directly to two listening sockets (``comm`` and ``logs``) and uses the accepted sockets as the supervisor's ``stdin`` / log pipes straight up. There is no bytes-bridge between two sockets, so ``make_raw_forwarder`` -- the only helper the extracted ``selector_loop`` module added beyond what was already inline in ``supervisor.py`` -- has no caller. ``_JavaActivitySubprocess`` reuses ``_register_pipe_readers`` and ``_close_unused_sockets`` via subclassing ``ActivitySubprocess``; both methods existed before the extraction and remain in ``supervisor.py`` after this revert. The inline selector dispatch loop and ``make_buffered_socket_reader`` come back into ``supervisor.py`` so the existing call sites (including the ``triggerer_job_runner`` re-export) keep working unchanged. This reverts commit 56464a8 ("Add common selector loop utilities for socket I/O handling for subprocesses") and deletes ``test_selector_loop.py`` introduced by 4b80753. * Address copilot's comments
uranusjr
pushed a commit
that referenced
this pull request
May 27, 2026
* Drop selector_loop module; new JavaCoordinator does not need it The rewritten ``JavaCoordinator`` (``airflow.sdk.coordinators.java``) lets the JVM connect directly to two listening sockets (``comm`` and ``logs``) and uses the accepted sockets as the supervisor's ``stdin`` / log pipes straight up. There is no bytes-bridge between two sockets, so ``make_raw_forwarder`` -- the only helper the extracted ``selector_loop`` module added beyond what was already inline in ``supervisor.py`` -- has no caller. ``_JavaActivitySubprocess`` reuses ``_register_pipe_readers`` and ``_close_unused_sockets`` via subclassing ``ActivitySubprocess``; both methods existed before the extraction and remain in ``supervisor.py`` after this revert. The inline selector dispatch loop and ``make_buffered_socket_reader`` come back into ``supervisor.py`` so the existing call sites (including the ``triggerer_job_runner`` re-export) keep working unchanged. This reverts commit 56464a8 ("Add common selector loop utilities for socket I/O handling for subprocesses") and deletes ``test_selector_loop.py`` introduced by 4b80753. * Address copilot's comments
uranusjr
pushed a commit
that referenced
this pull request
May 27, 2026
* Drop selector_loop module; new JavaCoordinator does not need it The rewritten ``JavaCoordinator`` (``airflow.sdk.coordinators.java``) lets the JVM connect directly to two listening sockets (``comm`` and ``logs``) and uses the accepted sockets as the supervisor's ``stdin`` / log pipes straight up. There is no bytes-bridge between two sockets, so ``make_raw_forwarder`` -- the only helper the extracted ``selector_loop`` module added beyond what was already inline in ``supervisor.py`` -- has no caller. ``_JavaActivitySubprocess`` reuses ``_register_pipe_readers`` and ``_close_unused_sockets`` via subclassing ``ActivitySubprocess``; both methods existed before the extraction and remain in ``supervisor.py`` after this revert. The inline selector dispatch loop and ``make_buffered_socket_reader`` come back into ``supervisor.py`` so the existing call sites (including the ``triggerer_job_runner`` re-export) keep working unchanged. This reverts commit 56464a8 ("Add common selector loop utilities for socket I/O handling for subprocesses") and deletes ``test_selector_loop.py`` introduced by 4b80753. * Address copilot's comments
uranusjr
pushed a commit
that referenced
this pull request
May 27, 2026
* Drop selector_loop module; new JavaCoordinator does not need it The rewritten ``JavaCoordinator`` (``airflow.sdk.coordinators.java``) lets the JVM connect directly to two listening sockets (``comm`` and ``logs``) and uses the accepted sockets as the supervisor's ``stdin`` / log pipes straight up. There is no bytes-bridge between two sockets, so ``make_raw_forwarder`` -- the only helper the extracted ``selector_loop`` module added beyond what was already inline in ``supervisor.py`` -- has no caller. ``_JavaActivitySubprocess`` reuses ``_register_pipe_readers`` and ``_close_unused_sockets`` via subclassing ``ActivitySubprocess``; both methods existed before the extraction and remain in ``supervisor.py`` after this revert. The inline selector dispatch loop and ``make_buffered_socket_reader`` come back into ``supervisor.py`` so the existing call sites (including the ``triggerer_job_runner`` re-export) keep working unchanged. This reverts commit 56464a8 ("Add common selector loop utilities for socket I/O handling for subprocesses") and deletes ``test_selector_loop.py`` introduced by 4b80753. * Address copilot's comments
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
As mentioned in apache#67175 (comment) comment, we should revert the changes of selector loop modularization as the new JavaCoordinator implementation does not need it at all.