Github user fhueske commented on a diff in the pull request:
https://github.com/apache/flink/pull/3860#discussion_r127418874
--- Diff:
flink-libraries/flink-table/src/main/scala/org/apache/flink/table/plan/nodes/PhysicalTableSourceScan.scala
---
@@ -70,4 +73,18 @@ abstract class PhysicalTableSourceScan(
def copy(traitSet: RelTraitSet, tableSource: TableSource[_]):
PhysicalTableSourceScan
+ override def estimateRowCount(mq: RelMetadataQuery): Double = {
+ val tableSourceTable = getTable.unwrap(classOf[TableSourceTable[_]])
+
+ if (tableSourceTable.getStatistic != FlinkStatistic.UNKNOWN) {
--- End diff --
Right, but in the current implementation the statistics of the
`TableSourceTable` are preferred over the stats of the `TableSource` (`if
(tableSourceTable.getStatistic != FlinkStatistic.UNKNOWN) { return rowcount;
}`).
If I see this correctly, the stats of the `TableSourceTable` are the stats
in the catalog, but a `FilterableTableSource` knows if a filter has been pushed
down and should have more accurate stats because it can apply the selectivity
of the pushed down filter.
So in my opinion the `TableSource` statistics should be preferred over the
static `TableSourceTable` stats.
Is my reasoning correct or did I overlook something @godfreyhe?
Thanks, Fabian
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
