[SPARK-17673][SQL] Incorrect exchange reuse with RowDataSourceScan

## What changes were proposed in this pull request?

It seems the equality check for reuse of `RowDataSourceScanExec` nodes doesn't respect the output schema. This can cause self-joins or unions over the same underlying data source to return incorrect results if they select different fields.

## How was this patch tested?

New unit test passes after the fix.

Author: Eric Liang <ekl@databricks.com>

Closes #15273 from ericl/spark-17673.

1 files changed, 8 insertions, 0 deletions

diff --git a/sql/core/src/test/scala/org/apache/spark/sql/jdbc/JDBCSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/jdbc/JDBCSuite.scala
index 10f15ca280..c94cb3b69d 100644
--- a/sql/core/src/test/scala/org/apache/spark/sql/jdbc/JDBCSuite.scala
+++ b/sql/core/src/test/scala/org/apache/spark/sql/jdbc/JDBCSuite.scala
@@ -791,4 +791,12 @@ class JDBCSuite extends SparkFunSuite
     val schema = JdbcUtils.schemaString(df, "jdbc:mysql://localhost:3306/temp")
     assert(schema.contains("`order` TEXT"))
   }
+
+  test("SPARK-17673: Exchange reuse respects differences in output schema") {
+    val df = sql("SELECT * FROM inttypes WHERE a IS NOT NULL")
+    val df1 = df.groupBy("a").agg("c" -> "min")
+    val df2 = df.groupBy("a").agg("d" -> "min")
+    val res = df1.union(df2)
+    assert(res.distinct().count() == 2)  // would be 1 if the exchange was incorrectly reused
+  }
 }