Github user viirya commented on a diff in the pull request:
https://github.com/apache/spark/pull/15523#discussion_r85488370
--- Diff: sql/core/src/test/scala/org/apache/spark/sql/DataFrameSuite.scala
---
@@ -1617,6 +1617,72 @@ class DataFrameSuite extends QueryTest with
SharedSQLContext {
}
}
+ private def verifyNullabilityInFilterExec(
+ df: DataFrame,
+ expr: String,
+ expectedNonNullableColumns: Seq[String]): Unit = {
+ val dfWithFilter = df.where(s"isnotnull($expr)").selectExpr(expr)
+ // In the logical plan, all the output columns of input dataframe are
nullable
+ dfWithFilter.queryExecution.optimizedPlan.collect {
+ case e: Filter => assert(e.output.forall(_.nullable))
+ }
+
+ dfWithFilter.queryExecution.executedPlan.collect {
+ // When the child expression in isnotnull is null-intolerant (i.e.
any null input will
+ // result in null output), the involved columns are converted to not
nullable;
+ // otherwise, no change should be made.
+ case e: FilterExec =>
+ assert(e.output.forall { o =>
+ if (expectedNonNullableColumns.contains(o.name)) !o.nullable
else o.nullable
+ })
+ }
+ }
+
+ test("SPARK-17957: no change on nullability in FilterExec output") {
+ val df = sparkContext.parallelize(Seq(
+ null.asInstanceOf[java.lang.Integer] -> new java.lang.Integer(3),
+ new java.lang.Integer(1) -> null.asInstanceOf[java.lang.Integer],
+ new java.lang.Integer(2) -> new java.lang.Integer(4))).toDF()
+
+ verifyNullabilityInFilterExec(df,
+ expr = "Rand()", expectedNonNullableColumns = Seq.empty[String])
+ verifyNullabilityInFilterExec(df,
+ expr = "coalesce(_1, _2)", expectedNonNullableColumns =
Seq.empty[String])
+ verifyNullabilityInFilterExec(df,
+ expr = "coalesce(_1, 0) + Rand()", expectedNonNullableColumns =
Seq.empty[String])
+ verifyNullabilityInFilterExec(df,
+ expr = "cast(coalesce(cast(coalesce(_1, _2) as double), 0.0) as
int)",
+ expectedNonNullableColumns = Seq.empty[String])
+ }
+
+ test("SPARK-17957: set nullability to false in FilterExec output") {
+ val df = sparkContext.parallelize(Seq(
+ null.asInstanceOf[java.lang.Integer] -> new java.lang.Integer(3),
+ new java.lang.Integer(1) -> null.asInstanceOf[java.lang.Integer],
+ new java.lang.Integer(2) -> new java.lang.Integer(4))).toDF()
+
+ verifyNullabilityInFilterExec(df,
+ expr = "_1 + _2 * 3", expectedNonNullableColumns = Seq("_1", "_2"))
+ verifyNullabilityInFilterExec(df,
+ expr = "_1 + _2", expectedNonNullableColumns = Seq("_1", "_2"))
+ verifyNullabilityInFilterExec(df,
+ expr = "_1", expectedNonNullableColumns = Seq("_1"))
+ verifyNullabilityInFilterExec(df,
+ expr = "_2 + Rand()", expectedNonNullableColumns = Seq("_2"))
--- End diff --
Yeah, so I said I will left that to @cloud-fan or others to decide...
---
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
