[SPARK-28371][SQL] Make Parquet "StartsWith" filter null-safe

Parquet may call the filter with a null value to check whether nulls are
accepted. While it seems Spark avoids that path in Parquet with 1.10, in
1.11 that causes Spark unit tests to fail.

Tested with Parquet 1.11 (and new unit test).

Closes #25140 from vanzin/SPARK-28371.

Authored-by: Marcelo Vanzin <vanzin@cloudera.com>
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>
(cherry picked from commit 7f9da2b)
Signed-off-by: Dongjoon Hyun <dhyun@apple.com>

sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/parquet/ParquetFilters.scala

@@ -541,7 +541,7 @@ private[parquet] class ParquetFilters(
               }
 
               override def keep(value: Binary): Boolean = {
-                UTF8String.fromBytes(value.getBytes).startsWith(
+                value != null && UTF8String.fromBytes(value.getBytes).startsWith(
                   UTF8String.fromBytes(strToBinary.getBytes))
               }
             }

sql/core/src/test/scala/org/apache/spark/sql/execution/datasources/parquet/ParquetFilterSuite.scala

@@ -955,6 +955,14 @@ class ParquetFilterSuite extends QueryTest with ParquetTest with SharedSQLContex
       }
     }
 
+    // SPARK-28371: make sure filter is null-safe.
+    withParquetDataFrame(Seq(Tuple1[String](null))) { implicit df =>
+      checkFilterPredicate(
+        '_1.startsWith("blah").asInstanceOf[Predicate],
+        classOf[UserDefinedByInstance[_, _]],
+        Seq.empty[Row])
+    }
+
     import testImplicits._
     // Test canDrop() has taken effect
     testStringStartsWith(spark.range(1024).map(_.toString).toDF(), "value like 'a%'")