[SPARK-23207][SQL][FOLLOW-UP] Don't perform local sort for DataFrame.…

…repartition(1)

## What changes were proposed in this pull request?

In `ShuffleExchangeExec`, we don't need to insert extra local sort before round-robin partitioning, if the new partitioning has only 1 partition, because under that case all output rows go to the same partition.

## How was this patch tested?

The existing test cases.

Author: Xingbo Jiang <xingbo.jiang@databricks.com>

Closes #20426 from jiangxb1987/repartition1.

sql/core/src/main/scala/org/apache/spark/sql/execution/exchange/ShuffleExchangeExec.scala

@@ -257,7 +257,11 @@ object ShuffleExchangeExec {
       //
       // Currently we following the most straight-forward way that perform a local sort before
       // partitioning.
+      //
+      // Note that we don't perform local sort if the new partitioning has only 1 partition, under
+      // that case all output rows go to the same partition.
       val newRdd = if (SQLConf.get.sortBeforeRepartition &&
+          newPartitioning.numPartitions > 1 &&
           newPartitioning.isInstanceOf[RoundRobinPartitioning]) {
         rdd.mapPartitionsInternal { iter =>
           val recordComparatorSupplier = new Supplier[RecordComparator] {

sql/core/src/test/scala/org/apache/spark/sql/execution/streaming/ForeachSinkSuite.scala

@@ -162,7 +162,7 @@ class ForeachSinkSuite extends StreamTest with SharedSQLContext with BeforeAndAf
       val allEvents = ForeachSinkSuite.allEvents()
       assert(allEvents.size === 1)
       assert(allEvents(0)(0) === ForeachSinkSuite.Open(partition = 0, version = 0))
-      assert(allEvents(0)(1) === ForeachSinkSuite.Process(value = 2))
+      assert(allEvents(0)(1) === ForeachSinkSuite.Process(value = 1))
 
       // `close` should be called with the error
       val errorEvent = allEvents(0)(2).asInstanceOf[ForeachSinkSuite.Close]