diff --git a/connector/connect/server/src/main/scala/org/apache/spark/sql/connect/planner/SparkConnectPlanner.scala b/connector/connect/server/src/main/scala/org/apache/spark/sql/connect/planner/SparkConnectPlanner.scala
index 8ca004d520c5..d7b3c057d927 100644
--- a/connector/connect/server/src/main/scala/org/apache/spark/sql/connect/planner/SparkConnectPlanner.scala
+++ b/connector/connect/server/src/main/scala/org/apache/spark/sql/connect/planner/SparkConnectPlanner.scala
@@ -1508,8 +1508,10 @@ class SparkConnectPlanner(val session: SparkSession) {
         maxRecordsPerBatch,
         maxBatchSize,
         timeZoneId)
-      assert(batches.size == 1)
-      batches.next()
+      assert(batches.hasNext)
+      val bytes = batches.next()
+      assert(!batches.hasNext, s"remaining batches: ${batches.size}")
+      bytes
     }
 
     // To avoid explicit handling of the result on the client, we build the expected input
diff --git a/python/pyspark/sql/tests/connect/test_connect_basic.py b/python/pyspark/sql/tests/connect/test_connect_basic.py
index 806fe6e2329b..dad303d3463e 100644
--- a/python/pyspark/sql/tests/connect/test_connect_basic.py
+++ b/python/pyspark/sql/tests/connect/test_connect_basic.py
@@ -2876,6 +2876,12 @@ def test_unsupported_io_functions(self):
             with self.assertRaises(NotImplementedError):
                 getattr(df.write, f)()
 
+    def test_sql_with_command(self):
+        # SPARK-42705: spark.sql should return values from the command.
+        self.assertEqual(
+            self.connect.sql("show functions").collect(), self.spark.sql("show functions").collect()
+        )
+
 
 @unittest.skipIf(not should_test_connect, connect_requirement_message)
 class ClientTests(unittest.TestCase):