I have a simple understanding that information about files in a parquet dataset can be placed into the _metadata file and used to give a more efficient creation of
_metadata
