Github user icexelloss commented on a diff in the pull request:
https://github.com/apache/spark/pull/19872#discussion_r158872825
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/execution/python/ExtractPythonUDFs.scala
---
@@ -48,29 +48,46 @@ object ExtractPythonUDFFromAggregate extends
Rule[LogicalPlan] {
}.isDefined
}
+ private def isPandasGroupAggUdf(expr: Expression): Boolean = expr match {
+ case PythonUDF(_, _, _, _, PythonEvalType.SQL_PANDAS_GROUP_AGG_UDF)
=> true
+ case Alias(child, _) => isPandasGroupAggUdf(child)
+ case _ => false
+ }
+
+ private def hasPandasGroupAggUdf(agg: Aggregate): Boolean = {
+ val actualAggExpr =
agg.aggregateExpressions.drop(agg.groupingExpressions.length)
+ actualAggExpr.exists(isPandasGroupAggUdf)
+ }
+
+
private def extract(agg: Aggregate): LogicalPlan = {
val projList = new ArrayBuffer[NamedExpression]()
val aggExpr = new ArrayBuffer[NamedExpression]()
- agg.aggregateExpressions.foreach { expr =>
- if (hasPythonUdfOverAggregate(expr, agg)) {
- // Python UDF can only be evaluated after aggregate
- val newE = expr transformDown {
- case e: Expression if belongAggregate(e, agg) =>
- val alias = e match {
- case a: NamedExpression => a
- case o => Alias(e, "agg")()
- }
- aggExpr += alias
- alias.toAttribute
+
+ if (hasPandasGroupAggUdf(agg)) {
+ Aggregate(agg.groupingExpressions, agg.aggregateExpressions,
agg.child)
--- End diff --
I am not sure. But I added copy in `ExtractGroupAggPandasUDFFromAggregate`
similar to existing rules.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
