Sort by descending order within each group

How about sorting by every grouper key column, with "time" in descending?

df.sort_values(['code', 'date', 'tank', 'time'], ascending=[True]*3 + [False]) code date time tank 3 123 01-01-2018 13:00:00 1 2 123 01-01-2018 12:00:00 1 1 123 01-01-2018 11:00:00 1 5 123 01-01-2018 09:00:00 1 0 123 01-01-2018 08:00:00 1 4 123 01-01-2018 07:00:00 1 9 124 01-01-2018 13:00:00 2 8 124 01-01-2018 12:00:00 2 7 124 01-01-2018 11:00:00 2 11 124 01-01-2018 09:00:00 2 6 124 01-01-2018 08:00:00 2 10 124 01-01-2018 07:00:00 2 

This will achieve the same effect, but without the groupby.

If groupby is needed, you will need two reset_index calls (to remove the last level):

(df.groupby(['code', 'date', 'tank']) .time.apply(lambda x: x.sort_values(ascending=False)) .reset_index(level=-1, drop=True) .reset_index()) code date tank time 0 123 01-01-2018 1 13:00:00 1 123 01-01-2018 1 12:00:00 2 123 01-01-2018 1 11:00:00 3 123 01-01-2018 1 09:00:00 4 123 01-01-2018 1 08:00:00 5 123 01-01-2018 1 07:00:00 6 124 01-01-2018 2 13:00:00 7 124 01-01-2018 2 12:00:00 8 124 01-01-2018 2 11:00:00 9 124 01-01-2018 2 09:00:00 10 124 01-01-2018 2 08:00:00 11 124 01-01-2018 2 07:00:00