用于在Cloud Dataflow中读取Parquet文件的自定义源

I have a requirement to read parquet file in my dataflow written in java and upload on bigquery. As there is no out of the box functionality given yet I know I have to write a custom source with hadoopFileFormat but I am not able to find any documentation regarding the same. Can somebody help me with some code or documentation on how to write a custom source or any other approach available to read a parquet file in Cloud dataflow. I have a requirement to read parquet file in my