diff options
author | Wenchen Fan <wenchen@databricks.com> | 2016-06-28 10:26:01 -0700 |
---|---|---|
committer | Yin Huai <yhuai@databricks.com> | 2016-06-28 10:26:01 -0700 |
commit | 1f2776df6e87a84991537ac20e4b8829472d3462 (patch) | |
tree | 6a1bb15ed398b129c4243bb7447cf707622c90d6 | |
parent | 0923c4f5676691e28e70ecb05890e123540b91f0 (diff) | |
download | spark-1f2776df6e87a84991537ac20e4b8829472d3462.tar.gz spark-1f2776df6e87a84991537ac20e4b8829472d3462.tar.bz2 spark-1f2776df6e87a84991537ac20e4b8829472d3462.zip |
[SPARK-16181][SQL] outer join with isNull filter may return wrong result
## What changes were proposed in this pull request?
The root cause is: the output attributes of outer join are derived from its children, while they are actually different attributes(outer join can return null).
We have already added some special logic to handle it, e.g. `PushPredicateThroughJoin` won't push down predicates through outer join side, `FixNullability`.
This PR adds one more special logic in `FoldablePropagation`.
## How was this patch tested?
new test in `DataFrameSuite`
Author: Wenchen Fan <wenchen@databricks.com>
Closes #13884 from cloud-fan/bug.
-rw-r--r-- | sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala | 8 | ||||
-rw-r--r-- | sql/core/src/test/scala/org/apache/spark/sql/DataFrameSuite.scala | 9 |
2 files changed, 17 insertions, 0 deletions
diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala index 2bca31d5f1..9bc8cea377 100644 --- a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala +++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala @@ -688,6 +688,14 @@ object FoldablePropagation extends Rule[LogicalPlan] { case c: Command => stop = true c + // For outer join, although its output attributes are derived from its children, they are + // actually different attributes: the output of outer join is not always picked from its + // children, but can also be null. + // TODO(cloud-fan): It seems more reasonable to use new attributes as the output attributes + // of outer join. + case j @ Join(_, _, LeftOuter | RightOuter | FullOuter, _) => + stop = true + j case p: LogicalPlan if !stop => p.transformExpressions { case a: AttributeReference if foldableMap.contains(a) => foldableMap(a) diff --git a/sql/core/src/test/scala/org/apache/spark/sql/DataFrameSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/DataFrameSuite.scala index 6a0a7df3f4..9d53be8e2b 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/DataFrameSuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/DataFrameSuite.scala @@ -1562,4 +1562,13 @@ class DataFrameSuite extends QueryTest with SharedSQLContext { val df = Seq(1, 1, 2).toDF("column.with.dot") checkAnswer(df.distinct(), Row(1) :: Row(2) :: Nil) } + + test("SPARK-16181: outer join with isNull filter") { + val left = Seq("x").toDF("col") + val right = Seq("y").toDF("col").withColumn("new", lit(true)) + val joined = left.join(right, left("col") === right("col"), "left_outer") + + checkAnswer(joined, Row("x", null, null)) + checkAnswer(joined.filter($"new".isNull), Row("x", null, null)) + } } |