[ https://issues.apache.org/jira/browse/FLINK-7439?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16143297#comment-16143297 ]
ASF GitHub Bot commented on FLINK-7439: --------------------------------------- Github user wuchong commented on a diff in the pull request: https://github.com/apache/flink/pull/4536#discussion_r135430161 --- Diff: flink-libraries/flink-table/src/main/scala/org/apache/flink/table/plan/schema/FlinkTableFunctionImpl.scala --- @@ -36,10 +36,8 @@ import org.apache.flink.table.calcite.FlinkTypeFactory class FlinkTableFunctionImpl[T]( val typeInfo: TypeInformation[T], val fieldIndexes: Array[Int], - val fieldNames: Array[String], - val evalMethod: Method) - extends ReflectiveFunctionBase(evalMethod) --- End diff -- In the first implementation of UDTF, we use `ReflectiveFunctionBase` to infer the operand types. But with the new approach in this PR, we customize a `SqlOperandTypeChecker` to check operands which means we don't need the `ReflectiveFunctionBase` anymore. If we extends `ReflectiveFunctionBase` we have to register every `eval` method as a UDTF. But with the new approach in this PR, we only need to register once. > Support variable arguments for UDTF in SQL > ------------------------------------------ > > Key: FLINK-7439 > URL: https://issues.apache.org/jira/browse/FLINK-7439 > Project: Flink > Issue Type: Improvement > Components: Table API & SQL > Reporter: Jark Wu > Assignee: Jark Wu > > Currently, both UDF and UDAF support variable parameters, but UDTF not. > FLINK-5882 supports variable UDTF for Table API only, but missed SQL. -- This message was sent by Atlassian JIRA (v6.4.14#64029)