mikebridge commented on code in PR #44835:
URL: https://github.com/apache/superset/pull/44835#discussion_r4208803195
##########
superset/initialization/__init__.py:
##########
@@ -163,10 +163,23 @@ class AppContextTask(task_base): # type: ignore
# nested context here would silently hand the task a second,
# blind session unable to see the caller's uncommitted work.
def __call__(self, *args: Any, **kwargs: Any) -> Any:
- if has_app_context():
- return task_base.__call__(self, *args, **kwargs)
- with superset_app.app_context():
- return task_base.__call__(self, *args, **kwargs)
+ # Avoid circular import through superset.app during
initialization.
+ from superset.semantic_layers.metadata_binding import
metadata_operation
+
+ with (
+ contextlib.nullcontext()
+ if has_app_context()
+ else superset_app.app_context()
+ ):
+ with (
+ metadata_operation()
Review Comment:
Fixed in c56ac0a6f2. The background dashboard export task and async chart
queries now give each chart a fresh metadata acquisition scope before its query
is built, so earlier task work or a slow preceding chart can't use up its
budget. Nested metadata work, annotations and contribution totals share that
chart's deadline and captured observations. Regressions cover sequential export
charts, delayed task setup, async deserialization, nested deadlines and cleanup
on failure.
Two scope notes: the inline HTTP export still shares the request's budget
across charts (by design), and the cache warm-up task gets the same per-chart
scope in the next push.
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
