soin08 commented on PR #28142:
URL: https://github.com/apache/flink/pull/28142#issuecomment-4741643742
1. > AtomicInteger callCount = new AtomicInteger();
`callCount` is serialized during job submission, so task-side execution
happens on deserialized copies, and incrementing that task-side copy does not
provide a valid synchronization path back to the original `callCount` object in
the test method. Probably there are other ways around it, but I implemented the
same idea with Flink's counter.
2. Tests show that the `Calc` node will be in the plan regardless of the
filters. For example, the query `SELECT k FROM state_table WHERE k = 999` has
```
== Optimized Execution Plan ==
Calc(select=[999 AS k])
+- TableSourceScan(table=[[default_catalog, default_database,
state_table, filter=[=(k, 999:BIGINT)]]], fields=[...])
```
So I check pushdown by looking at `TableSourceScan` with a regexp.
Queries with non pushed-down filter have empty `filter=[]`
Also moved 2 tests that were irrelevant for SQL from
`SavepointDynamicTableSourceTest` into `KeyedStateInputFormatTest`, where they
belong.
--
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]
