amoghrajesh commented on code in PR #66859:
URL: https://github.com/apache/airflow/pull/66859#discussion_r3271801305
##########
shared/state/tests/state/test_state.py:
##########
@@ -70,3 +70,71 @@ def test_abstract_methods_cover_full_interface(self):
"""BaseStateBackend enforces all 8 sync+async methods as abstract."""
expected = {"get", "set", "delete", "clear", "aget", "aset",
"adelete", "aclear"}
assert BaseStateBackend.__abstractmethods__ == expected
+
+ def test_task_state_serialize_deserialize_round_trip(self, backend):
+ original = "app_1234"
+ serialized = backend.serialize_task_state_to_ref(value=original,
key="job_id", ti_id="abc-123")
+ deserialized = backend.deserialize_task_state_from_ref(serialized)
+ assert deserialized == original
+
+ def test_custom_backend_overrides_task_state_ser_deser(self):
+ class MyBackend(BaseStateBackend):
+ def get(self, scope, key): ...
+ def set(self, scope, key, value): ...
+ def delete(self, scope, key): ...
+ def clear(self, scope, *, all_map_indices=False): ...
+ async def aget(self, scope, key): ...
+ async def aset(self, scope, key, value): ...
+ async def adelete(self, scope, key): ...
+ async def aclear(self, scope, *, all_map_indices=False): ...
+
+ def serialize_task_state_value(self, *, value, key, ti_id):
Review Comment:
Handled in 59e33f7b39
##########
shared/state/tests/state/test_state.py:
##########
@@ -70,3 +70,71 @@ def test_abstract_methods_cover_full_interface(self):
"""BaseStateBackend enforces all 8 sync+async methods as abstract."""
expected = {"get", "set", "delete", "clear", "aget", "aset",
"adelete", "aclear"}
assert BaseStateBackend.__abstractmethods__ == expected
+
+ def test_task_state_serialize_deserialize_round_trip(self, backend):
+ original = "app_1234"
+ serialized = backend.serialize_task_state_to_ref(value=original,
key="job_id", ti_id="abc-123")
+ deserialized = backend.deserialize_task_state_from_ref(serialized)
+ assert deserialized == original
+
+ def test_custom_backend_overrides_task_state_ser_deser(self):
+ class MyBackend(BaseStateBackend):
+ def get(self, scope, key): ...
+ def set(self, scope, key, value): ...
+ def delete(self, scope, key): ...
+ def clear(self, scope, *, all_map_indices=False): ...
+ async def aget(self, scope, key): ...
+ async def aset(self, scope, key, value): ...
+ async def adelete(self, scope, key): ...
+ async def aclear(self, scope, *, all_map_indices=False): ...
+
+ def serialize_task_state_value(self, *, value, key, ti_id):
+ return f"s3://bucket/{ti_id}/{key}"
+
+ def deserialize_task_state_value(self, stored):
+ return f"fetched:{stored}"
+
+ b = MyBackend()
+ assert (
+ b.serialize_task_state_value(value="app_1234", key="job_id",
ti_id="abc-123")
+ == "s3://bucket/abc-123/job_id"
+ )
+ assert (
+ b.deserialize_task_state_value("s3://bucket/abc-123/job_id")
+ == "fetched:s3://bucket/abc-123/job_id"
+ )
+
+ def test_asset_state_serialize_deserialize_round_trip(self, backend):
+ original = "2026-05-01"
+ serialized = backend.serialize_asset_state_to_ref(
+ value="2026-05-01", key="watermark", asset_ref="my_asset"
+ )
+ deserialized = backend.deserialize_asset_state_from_ref(serialized)
+ assert deserialized == original
+
+ def test_custom_backend_overrides_asset_state_ser_deser(self):
+ class MyBackend(BaseStateBackend):
+ def get(self, scope, key): ...
+ def set(self, scope, key, value): ...
+ def delete(self, scope, key): ...
+ def clear(self, scope, *, all_map_indices=False): ...
+ async def aget(self, scope, key): ...
+ async def aset(self, scope, key, value): ...
+ async def adelete(self, scope, key): ...
+ async def aclear(self, scope, *, all_map_indices=False): ...
+
+ def serialize_asset_state_value(self, *, value, key, asset_ref):
+ return f"s3://bucket/assets/{asset_ref}/{key}"
Review Comment:
Handled in 59e33f7b39
##########
task-sdk/src/airflow/sdk/execution_time/task_runner.py:
##########
@@ -1455,6 +1465,10 @@ def _handle_current_task_success(
if conf.getboolean("state_store", "clear_on_success"):
log.info("Task state will be cleared by the server because
clear_on_success is enabled.")
+ if _get_worker_state_backend() is not None:
+ # clear the task state keys for custom state backends configured
on worker side
+ context["task_state"].clear()
Review Comment:
Handled in 59e33f7b39
--
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]
