hvanhovell commented on code in PR #40581:
URL: https://github.com/apache/spark/pull/40581#discussion_r1158830161
##########
connector/connect/server/src/main/scala/org/apache/spark/sql/connect/planner/SparkConnectPlanner.scala:
##########
@@ -482,27 +482,66 @@ class SparkConnectPlanner(val session: SparkSession) {
}
private def transformMapPartitions(rel: proto.MapPartitions): LogicalPlan = {
+ val baseRel = transformRelation(rel.getInput)
val commonUdf = rel.getFunc
- val pythonUdf = transformPythonUDF(commonUdf)
- val isBarrier = if (rel.hasIsBarrier) rel.getIsBarrier else false
- pythonUdf.evalType match {
- case PythonEvalType.SQL_MAP_PANDAS_ITER_UDF =>
- logical.MapInPandas(
- pythonUdf,
- pythonUdf.dataType.asInstanceOf[StructType].toAttributes,
- transformRelation(rel.getInput),
- isBarrier)
- case PythonEvalType.SQL_MAP_ARROW_ITER_UDF =>
- logical.PythonMapInArrow(
- pythonUdf,
- pythonUdf.dataType.asInstanceOf[StructType].toAttributes,
- transformRelation(rel.getInput),
- isBarrier)
+ commonUdf.getFunctionCase match {
+ case proto.CommonInlineUserDefinedFunction.FunctionCase.SCALAR_SCALA_UDF
=>
+ transformTypedMapPartitions(commonUdf, baseRel)
+ case proto.CommonInlineUserDefinedFunction.FunctionCase.PYTHON_UDF =>
+ val pythonUdf = transformPythonUDF(commonUdf)
+ val isBarrier = if (rel.hasIsBarrier) rel.getIsBarrier else false
+ pythonUdf.evalType match {
+ case PythonEvalType.SQL_MAP_PANDAS_ITER_UDF =>
+ logical.MapInPandas(
+ pythonUdf,
+ pythonUdf.dataType.asInstanceOf[StructType].toAttributes,
+ baseRel,
+ isBarrier)
+ case PythonEvalType.SQL_MAP_ARROW_ITER_UDF =>
+ logical.PythonMapInArrow(
+ pythonUdf,
+ pythonUdf.dataType.asInstanceOf[StructType].toAttributes,
+ baseRel,
+ isBarrier)
+ case _ =>
+ throw InvalidPlanInput(
+ s"Function with EvalType: ${pythonUdf.evalType} is not
supported")
+ }
case _ =>
- throw InvalidPlanInput(s"Function with EvalType: ${pythonUdf.evalType}
is not supported")
+ throw InvalidPlanInput(
+ s"Function with ID: ${commonUdf.getFunctionCase.getNumber} is not
supported")
}
}
+ private def generateObjAttr[T](enc: ExpressionEncoder[T]): Attribute = {
+ val dataType = enc.deserializer.dataType
+ val nullable = !enc.clsTag.runtimeClass.isPrimitive
+ AttributeReference("obj", dataType, nullable)()
+ }
+
+ private def transformTypedMapPartitions(
+ fun: proto.CommonInlineUserDefinedFunction,
+ child: LogicalPlan): LogicalPlan = {
+ val udf = fun.getScalarScalaUdf
+ val udfPacket =
Review Comment:
No you cannot. That would mean you cannot use case classes or java beans.
Rows would work though.
The reason we propagate types is to make the sure the engine can supply
correctly typed data.
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
