Jing Zhang created FLINK-13321:
----------------------------------
Summary: Join a udf with constant arguments or without argument in
TableAPI query of Blink Planner does not work now
Key: FLINK-13321
URL: https://issues.apache.org/jira/browse/FLINK-13321
Project: Flink
Issue Type: Task
Components: Table SQL / API
Reporter: Jing Zhang
In blink planner, Join a udf with constant arguments or without argument in
TableAPI query does not work now, for example: error will be thrown if run the
following two TableAPI query in Blink planner:
{code:java}
leftT.select('c).joinLateral(func0("1", "2"))
// leftT.select('c).joinLateral(func0())
{code}
The following error will be thrown:
{code:java}
org.apache.flink.table.api.TableException: Cannot generate a valid execution
plan for the given query:
FlinkLogicalSink(name=[5771dc74-8986-4ffa-828f-8ed40602593a], fields=[c, f0])
+- FlinkLogicalCorrelate(correlation=[$cor3], joinType=[inner],
requiredColumns=[{}])
:- FlinkLogicalCalc(select=[c])
: +- FlinkLogicalDataStreamTableScan(table=[[default_catalog,
default_database, 15cbb5bf-816b-4319-9be8-6c648c868843]])
+- FlinkLogicalCorrelate(correlation=[$cor4], joinType=[inner],
requiredColumns=[{}])
:- FlinkLogicalValues(tuples=[[{ }]])
+-
FlinkLogicalTableFunctionScan(invocation=[org$apache$flink$table$util$VarArgsFunc0$2ad590150fcbadcd9e420797d27a5eb1(_UTF-16LE'1',
_UTF-16LE'2')], rowType=[RecordType(VARCHAR(2147483647) f0)],
elementType=[class [Ljava.lang.Object;])
This exception indicates that the query uses an unsupported SQL feature.
Please check the documentation for the set of currently supported SQL features.
at
org.apache.flink.table.plan.optimize.program.FlinkVolcanoProgram.optimize(FlinkVolcanoProgram.scala:72)
at
org.apache.flink.table.plan.optimize.program.FlinkChainedProgram$$anonfun$optimize$1.apply(FlinkChainedProgram.scala:63)
at
org.apache.flink.table.plan.optimize.program.FlinkChainedProgram$$anonfun$optimize$1.apply(FlinkChainedProgram.scala:58)
at
scala.collection.TraversableOnce$$anonfun$foldLeft$1.apply(TraversableOnce.scala:157)
...
....
{code}
The root cause is the `FlinkLogicalTableFunctionScan`.CONVERTER translates a
`TableFunctionScan` to a `Correlate`. Which will translate the original
`RelNode` tree to a `RelNode` with two Cascaded ·Correlate` (could be found in
the above thrown message), which could not translate to Physical `RelNode`.
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
