Github user rxin commented on a diff in the pull request:
https://github.com/apache/spark/pull/9364#discussion_r43823876
--- Diff: sql/core/src/test/scala/org/apache/spark/sql/DataFrameSuite.scala
---
@@ -997,4 +1001,116 @@ class DataFrameSuite extends QueryTest with
SharedSQLContext {
}
}
}
+
+ /**
+ * Verifies that there is no Exchange between the Aggregations for `df`
+ */
+ private def verifyNonExchangingAgg(df: DataFrame) = {
+ var atFirstAgg: Boolean = false
+ df.queryExecution.executedPlan.foreach {
+ case agg: TungstenAggregate => {
+ atFirstAgg = !atFirstAgg
+ }
+ case _ => {
+ if (atFirstAgg) {
+ fail("Should not have operators between the two aggregations")
+ }
+ }
+ }
+ }
+
+ /**
+ * Verifies that there is an Exchange between the Aggregations for `df`
+ */
+ private def verifyExchangingAgg(df: DataFrame) = {
+ var atFirstAgg: Boolean = false
+ df.queryExecution.executedPlan.foreach {
+ case agg: TungstenAggregate => {
+ if (atFirstAgg) {
+ fail("Should not have back to back Aggregates")
+ }
+ atFirstAgg = true
+ }
+ case e: Exchange => atFirstAgg = false
+ case _ =>
+ }
+ }
+
+ test("distributeBy and localSort") {
+ val original = testData.repartition(1)
+ assert(original.rdd.partitions.length == 1)
+ val df = original.distributeBy(Column("key") :: Nil, 5)
+ assert(df.rdd.partitions.length == 5)
+ checkAnswer(original.select(), df.select())
+
+ val df2 = original.distributeBy(Column("key") :: Nil, 10)
+ assert(df2.rdd.partitions.length == 10)
+ checkAnswer(original.select(), df2.select())
+
+ // Group by the column we are distributed by. This should generate a
plan with no exchange
+ // between the aggregates
+ val df3 = testData.distributeBy(Column("key") ::
Nil).groupBy("key").count()
+ verifyNonExchangingAgg(df3)
+ verifyNonExchangingAgg(testData.distributeBy(Column("key") ::
Column("value") :: Nil)
+ .groupBy("key", "value").count())
+
+ // Grouping by just the first distributeBy expr, need to exchange.
+ verifyExchangingAgg(testData.distributeBy(Column("key") ::
Column("value") :: Nil)
+ .groupBy("key").count())
+
+ val data = sqlContext.sparkContext.parallelize(
+ (1 to 100).map(i => TestData2(i % 10, i))).toDF()
+
+ // Distribute and order by.
+ val df4 = data.distributeBy(Column("a") :: Nil).localSort($"b".desc)
+ // Walk each partition and verify that it is sorted descending and
does not contain all
+ // the values.
+ df4.rdd.foreachPartition(p => {
--- End diff --
for future reference, we usually do
```scala
df4.rdd.foreachPartition { p =>
...
}
```
rather than `(p => {`
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
