tfdv.WriteStatisticsToRecordsAndBinaryFile

API for writing statistics to both sharded records and binary pb.

tfdv.WriteStatisticsToRecordsAndBinaryFile(
    binary_proto_path: str,
    records_path_prefix: str,
    columnar_path_prefix: Optional[str] = None
) -> None

This PTransform assumes that input represents sharded statistics, which are written directly. These statistics are also merged and written to a binary proto.

Currently Experimental.

binary_proto_path Output path for writing statistics as a binary proto.
records_path_prefix File pattern for writing statistics to sharded records.
columnar_path_prefix Optional file pattern for writing statistics to columnar outputs. If provided, columnar outputs will be written when supported.

pipeline None