diff --git a/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveStrategies.scala b/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveStrategies.scala
index 56fc85239e1c0..07d82f6fd5db5 100644
--- a/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveStrategies.scala
+++ b/sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveStrategies.scala
@@ -194,8 +194,10 @@ private[hive] trait HiveStrategies {
         // Filter out all predicates that only deal with partition keys, these are given to the
         // hive table scan operator to be used for partition pruning.
         val partitionKeyIds = AttributeSet(relation.partitionKeys)
-        val (pruningPredicates, otherPredicates) = predicates.partition {
-          _.references.subsetOf(partitionKeyIds)
+        val (pruningPredicates, otherPredicates) = predicates.partition { x =>
+          x.references.baseSet != null &&
+          !x.references.baseSet.isEmpty &&
+          x.references.subsetOf(partitionKeyIds)
         }
 
         pruneFilterProject(