diff --git a/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveSessionCatalog.scala b/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveSessionCatalog.scala
index 4a9b28a455a4..72c6f53c69ee 100644
--- a/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveSessionCatalog.scala
+++ b/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveSessionCatalog.scala
@@ -35,7 +35,7 @@ import org.apache.spark.sql.catalyst.plans.logical.{LogicalPlan, SubqueryAlias}
 import org.apache.spark.sql.catalyst.rules.Rule
 import org.apache.spark.sql.hive.HiveShim.HiveFunctionWrapper
 import org.apache.spark.sql.internal.SQLConf
-import org.apache.spark.sql.types.{DecimalType, DoubleType}
+import org.apache.spark.sql.types.{ArrayType, DecimalType, DoubleType}
 import org.apache.spark.util.Utils
 
 
@@ -166,8 +166,13 @@ private[sql] class HiveSessionCatalog(
     } catch {
       case NonFatal(_) =>
         // SPARK-16228 ExternalCatalog may recognize `double`-type only.
+        // SPARK-18527 Percentile needs explicit cast to array<double>
         val newChildren = children.map { child =>
-          if (child.dataType.isInstanceOf[DecimalType]) Cast(child, DoubleType) else child
+          child.dataType match {
+            case ArrayType(DecimalType(), nullable) => Cast(child, ArrayType(DoubleType, nullable))
+            case DecimalType() => Cast(child, DoubleType)
+            case _ => child
+          }
         }
         lookupFunction0(name, newChildren)
     }
diff --git a/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/HiveUDFSuite.scala b/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/HiveUDFSuite.scala
index 48adc833f4b2..5c8ee0bac1b6 100644
--- a/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/HiveUDFSuite.scala
+++ b/sql/hive/src/test/scala/org/apache/spark/sql/hive/execution/HiveUDFSuite.scala
@@ -149,6 +149,12 @@ class HiveUDFSuite extends QueryTest with TestHiveSingleton with SQLTestUtils {
     sql("select percentile_approx(value, 0.5) from values 1.0,2.0,3.0 T(value)")
   }
 
+  test("SPARK-18527 Percentile needs explicit cast to array<double>") {
+    sql("select percentile(value, cast(array(0.1, 0.5) as array<double>))" +
+        "from values 1,2,3 T(value)")
+    sql("select percentile(value, array(0.1, 0.5)) from values 1,2,3 T(value)")
+  }
+
   test("Generic UDAF aggregates") {
 
     checkAnswer(sql(