Currently data types are saved in a collection of the DataFrame. IMO it would make more sense to save it in each data series.