Github user sddyljsx commented on a diff in the pull request: https://github.com/apache/spark/pull/21859#discussion_r209417551 --- Diff: sql/core/src/main/scala/org/apache/spark/sql/execution/exchange/ShuffleExchangeExec.scala --- @@ -294,7 +296,12 @@ object ShuffleExchangeExec { sorter.sort(iter.asInstanceOf[Iterator[UnsafeRow]]) } } else { - rdd + part match { + case partitioner: RangePartitioner[InternalRow @unchecked, _] + if partitioner.getSampledArray != null => + sparkContext.parallelize(partitioner.getSampledArray.toSeq, rdd.getNumPartitions) --- End diff -- ``` part match { case partitioner: RangePartitioner[InternalRow @unchecked, _] if partitioner.getSampledArray != null => sparkContext.parallelize(partitioner.getSampledArray.toSeq, rdd.getNumPartitions) case _ => rdd } ``` When the optimization works, It will return the parallelized sampled data instead of the rdd. So I keep the number of the partitions same as the rdd's here
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org