cloud-fan commented on a change in pull request #34025:
URL: https://github.com/apache/spark/pull/34025#discussion_r710681748
##########
File path:
sql/core/src/test/scala/org/apache/spark/sql/DataFrameSetOperationsSuite.scala
##########
@@ -1018,6 +1018,64 @@ class DataFrameSetOperationsSuite extends QueryTest with
SharedSparkSession {
unionDF = df1.unionByName(df2)
checkAnswer(unionDF, expected)
}
+
+ test("SPARK-36673: Only merge nullability for Unions of struct") {
+ val df1 = spark.range(2).withColumn("nested", struct(expr("id * 5 AS
INNER")))
+ val df2 = spark.range(2).withColumn("nested", struct(expr("id * 5 AS
inner")))
+
+ val union1 = df1.union(df2)
+ val union2 = df1.unionByName(df2)
+
+ val schema = StructType(Seq(StructField("id", LongType, false),
+ StructField("nested", StructType(Seq(StructField("INNER", LongType,
false))), false)))
+
+ Seq(union1, union2).foreach { df =>
+ assert(df.schema == schema)
+ assert(df.queryExecution.optimizedPlan.schema == schema)
+ assert(df.queryExecution.executedPlan.schema == schema)
+
+ checkAnswer(df, Row(0, Row(0)) :: Row(1, Row(5)) :: Row(0, Row(0)) ::
Row(1, Row(5)) :: Nil)
+ checkAnswer(df.select("nested.*"), Row(0) :: Row(5) :: Row(0) :: Row(5)
:: Nil)
+ }
+ }
+
+ test("SPARK-36673: Only merge nullability for unionByName of struct") {
+ val df1 = spark.range(2).withColumn("nested", struct(expr("id * 5 AS
INNER")))
+ val df2 = spark.range(2).withColumn("nested", struct(expr("id * 5 AS
inner")))
+
+ val df = df1.unionByName(df2)
+
+ val schema = StructType(Seq(StructField("id", LongType, false),
+ StructField("nested", StructType(Seq(StructField("INNER", LongType,
false))), false)))
+
+ assert(df.schema == schema)
+ assert(df.queryExecution.optimizedPlan.schema == schema)
+ assert(df.queryExecution.executedPlan.schema == schema)
+
+ checkAnswer(df, Row(0, Row(0)) :: Row(1, Row(5)) :: Row(0, Row(0)) ::
Row(1, Row(5)) :: Nil)
+ checkAnswer(df.select("nested.*"), Row(0) :: Row(5) :: Row(0) :: Row(5) ::
Nil)
+ }
+
+ test("SPARK-36673: Union of structs with different orders") {
+ val df1 = spark.range(2).withColumn("nested",
+ struct(expr("id * 5 AS inner1"), struct(expr("id * 10 AS inner2"))))
+ val df2 = spark.range(2).withColumn("nested",
+ struct(expr("id * 5 AS inner2"), struct(expr("id * 10 AS inner1"))))
Review comment:
I think we should think more about how union by ordinal should work.
if we have 2 tables t1: [inner1, inner2] and t2: [inner2, inner1], we can
union them. But if they are inner fields, we can't union. AFAIK we are trying
to make nested columns the first-class citizen and this inner-field-only
limitation doesn't make much sense.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
