Add comments

Author: zhuqi-lucas

datafusion/datasource-avro/src/file_format.rs

@@ -135,6 +135,10 @@ impl FileFormat for AvroFormat {
         Ok(Arc::new(merged_schema))
     }
 
+    async fn transform_schema(&self, schema: SchemaRef) -> Result<SchemaRef> {
+        Ok(schema)
+    }
+
     async fn infer_stats(
         &self,
         _state: &dyn Session,

datafusion/datasource-csv/src/file_format.rs

@@ -393,6 +393,10 @@ impl FileFormat for CsvFormat {
         Ok(Arc::new(merged_schema))
     }
 
+    async fn transform_schema(&self, schema: SchemaRef) -> Result<SchemaRef> {
+        Ok(schema)
+    }
+
     async fn infer_stats(
         &self,
         _state: &dyn Session,

datafusion/datasource-json/src/file_format.rs

@@ -233,6 +233,10 @@ impl FileFormat for JsonFormat {
         Ok(Arc::new(schema))
     }
 
+    async fn transform_schema(&self, schema: SchemaRef) -> Result<SchemaRef> {
+        Ok(schema)
+    }
+
     async fn infer_stats(
         &self,
         _state: &dyn Session,

datafusion/datasource-parquet/src/file_format.rs

@@ -359,19 +359,29 @@ impl FileFormat for ParquetFormat {
             Schema::try_merge(schemas)
         }?;
 
+        self.transform_schema(Arc::new(schema)).await
+    }
+
+    /// transform_schema for parquet format now contains two steps:
+    ///
+    /// 1. Transform a schema so that any binary types are strings
+    /// see [transform_binary_to_string]
+    ///
+    /// 2. Transform a schema to use view types for Utf8 and Binary
+    /// See [transform_schema_to_view] for details
+    async fn transform_schema(&self, schema: SchemaRef) -> Result<SchemaRef> {
         let schema = if self.binary_as_string() {
-            transform_binary_to_string(&schema)
+            Arc::new(transform_binary_to_string(schema.as_ref()))
         } else {
             schema
         };
 
         let schema = if self.force_view_types() {
-            transform_schema_to_view(&schema)
+            Arc::new(transform_schema_to_view(schema.as_ref()))
         } else {
             schema
         };
-
-        Ok(Arc::new(schema))
+        Ok(schema)
     }
 
     async fn infer_stats(
@@ -598,6 +608,8 @@ pub fn transform_schema_to_view(schema: &Schema) -> Schema {
 }
 
 /// Transform a schema so that any binary types are strings
+///
+/// See [ParquetFormat::binary_as_string] for details
 pub fn transform_binary_to_string(schema: &Schema) -> Schema {
     let transformed_fields: Vec<Arc<Field>> = schema
         .fields