A Spark DataFrame to be exported

Can be a URI template containing "partitionId" (e.g., "hdfs://my_data_part_partitionId.rds") where "partitionId" will be substituted with ID of each partition using 'glue', or a list of URIs to be assigned to RDS output from all partitions (e.g., "hdfs://my_data_part_0.rds", "hdfs://my_data_part_1.rds", and so on) If working with a Spark instance running locally, then all URIs should be in "file://<local file path>" form. Otherwise the scheme of the URI should reflect the underlying file system the Spark instance is working with (e.g., "hdfs://"). If the resulting list of URI(s) does not contain unique values, then it will be post-processed with 'make.unique()' to ensure uniqueness.