Renames root Parquet message name

sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetReadSupport.scala

@@ -121,7 +121,9 @@ private[parquet] class ParquetReadSupport extends ReadSupport[InternalRow] with
             if (fileFieldNames.contains(field.name)) {
               // If the field exists in the target Parquet file, extracts the field type from the
               // full file schema and makes a single-field Parquet schema
-              new MessageType("root", fileSchema.getType(field.name))
+              new MessageType(
+                ParquetSchemaConverter.SPARK_PARQUET_SCHEMA_NAME,
+                fileSchema.getType(field.name))
             } else {
               // Otherwise, just resorts to `CatalystSchemaConverter`
               toParquet.convert(StructType(Array(field)))
@@ -131,7 +133,7 @@ private[parquet] class ParquetReadSupport extends ReadSupport[InternalRow] with
           // columns.  Note that it's possible that no columns are requested at all (e.g., count
           // some partition column of a partitioned Parquet table). That's why `fold` is used here
           // and always fallback to an empty Parquet schema.
-          .fold(new MessageType("root")) {
+          .fold(new MessageType(ParquetSchemaConverter.SPARK_PARQUET_SCHEMA_NAME)) {
             _ union _
           }
       }

sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetSchemaConverter.scala

@@ -314,7 +314,10 @@ private[parquet] class ParquetSchemaConverter(
    * Converts a Spark SQL [[StructType]] to a Parquet [[MessageType]].
    */
   def convert(catalystSchema: StructType): MessageType = {
-    Types.buildMessage().addFields(catalystSchema.map(convertField): _*).named("root")
+    Types
+      .buildMessage()
+      .addFields(catalystSchema.map(convertField): _*)
+      .named(ParquetSchemaConverter.SPARK_PARQUET_SCHEMA_NAME)
   }
 
   /**
@@ -541,6 +544,8 @@ private[parquet] class ParquetSchemaConverter(
 
 
 private[parquet] object ParquetSchemaConverter {
+  val SPARK_PARQUET_SCHEMA_NAME = "spark_schema"
+
   def checkFieldName(name: String): Unit = {
     // ,;{}()\n\t= and space are special characters in Parquet schema
     analysisRequire(