Many data stores already have sorted datetime columns. Dask does can do nice partitioning on datetimes. If I could tell dd.read_csv to set the index on a datetime column and not to worry about sorting the columns before partitioning that could be helpful.