diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/Expression.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/Expression.scala
index fa1a2ad56ccb..9edc1ceff26a 100644
--- a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/Expression.scala
+++ b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/Expression.scala
@@ -511,7 +511,7 @@ abstract class BinaryOperator extends BinaryExpression with ExpectsInputTypes {
 
   override def checkInputDataTypes(): TypeCheckResult = {
     // First check whether left and right have the same type, then check if the type is acceptable.
-    if (left.dataType != right.dataType) {
+    if (!left.dataType.sameType(right.dataType)) {
       TypeCheckResult.TypeCheckFailure(s"differing types in '$sql' " +
         s"(${left.dataType.simpleString} and ${right.dataType.simpleString}).")
     } else if (!inputType.acceptsType(left.dataType)) {
diff --git a/sql/core/src/test/scala/org/apache/spark/sql/DataFrameSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/DataFrameSuite.scala
index e87baa454c8b..efd64165d60c 100644
--- a/sql/core/src/test/scala/org/apache/spark/sql/DataFrameSuite.scala
+++ b/sql/core/src/test/scala/org/apache/spark/sql/DataFrameSuite.scala
@@ -1631,4 +1631,13 @@ class DataFrameSuite extends QueryTest with SharedSQLContext {
     dates.except(widenTypedRows).collect()
     dates.intersect(widenTypedRows).collect()
   }
+
+  test("SPARK-18070 binary operator should not consider nullability when comparing input types") {
+    val rows = Seq(Row(Seq(1), Seq(1)))
+    val schema = new StructType()
+      .add("array1", ArrayType(IntegerType))
+      .add("array2", ArrayType(IntegerType, containsNull = false))
+    val df = spark.createDataFrame(spark.sparkContext.makeRDD(rows), schema)
+    assert(df.filter($"array1" === $"array2").count() == 1)
+  }
 }