Github user fhueske commented on a diff in the pull request:
https://github.com/apache/flink/pull/3829#discussion_r136847905
--- Diff:
flink-libraries/flink-table/src/main/scala/org/apache/flink/table/api/queryConfig.scala
---
@@ -19,9 +19,21 @@
package org.apache.flink.table.api
import _root_.java.io.Serializable
+
import org.apache.flink.api.common.time.Time
-class QueryConfig private[table] extends Serializable {}
+class QueryConfig private[table] extends Serializable {
+}
+
+object QueryConfig {
+ def getQueryConfigFromTableEnv(tableEnv: TableEnvironment): QueryConfig
= {
--- End diff --
The problem is that we should have the correct subtype of `QueryConfig`
(not only the returned object but also the method signature) and the separation
of the stream and batch APIs. The `StreamExecutionEnvironment.queryConfig`
method must be typed to `StreamQueryConfig` to prevent casting. However, we
don't want to have a method that is typed to `StreamQueryConfig` in the base
class `TableEnvironment` because 1) this class should be independent from the
streaming API and 2) we don't want a streaming method in the
`BatchTableEnvironment`.
So, we can add a method to `TableEnvironment` that returns `QueryConfig`
(depending on its own type a `StreamQueryConfig` or `BatchQueryConfig`) and
override this method in `StreamTableEnvironment` such that it returns a
`StreamQueryConfig` (and if later necessary in `BatchTableEnvironment` to
return a `BatchQueryConfig`).
The duplicated code are < 5 lines and we have correct types everywhere.
---
