How to sort date by descending order and time by ascending order using Pandas

Question

My df looks like this. It is an hourly dataset.

time                    Open
2017-01-03 09:00:00     5.2475
2017-01-03 08:00:00     5.2180
2017-01-03 07:00:00     5.2128
2017-01-02 09:00:00     5.4122
2017-01-02 08:00:00     5.2123
2017-01-02 07:00:00     5.2475
2017-01-01 09:00:00     5.2180
2017-01-01 08:00:00     5.2128
2017-01-01 07:00:00     5.4122

I want to sort the hourly only data by ascending order.

What did I do?

I did:

df.sort_values(by='time', ascending=True)

but it sort entire value of time however I want to only sort the time section.

My new df should look like this:

time                    Open
2017-01-03 07:00:00     5.2475
2017-01-03 08:00:00     5.2180
2017-01-03 09:00:00     5.2128
2017-01-02 07:00:00     5.4122
2017-01-02 08:00:00     5.2123
2017-01-02 09:00:00     5.2475
2017-01-01 07:00:00     5.2180
2017-01-01 08:00:00     5.2128
2017-01-01 09:00:00     5.4122

Here date stays the same but the time is in ascending order.

Answer 1

If need sorting by dates and times create new columns for sorting by DataFrame.assign, then sort by both columns with DataFrame.sort_values and ascending parameter, because sorting by dates is descending and by times is ascending and last remove helper columns with DataFrame.drop:

df1 = (df.assign(d=df['time'].dt.date,
                 t=df['time'].dt.time)
         .sort_values(['d','t'], ascending=[False, True])
         .drop(['d','t'], axis=1))
print (df1)
                 time    Open
2 2017-01-03 07:00:00  5.2128
1 2017-01-03 08:00:00  5.2180
0 2017-01-03 09:00:00  5.2475
5 2017-01-02 07:00:00  5.2475
4 2017-01-02 08:00:00  5.2123
3 2017-01-02 09:00:00  5.4122
8 2017-01-01 07:00:00  5.4122
7 2017-01-01 08:00:00  5.2128
6 2017-01-01 09:00:00  5.2180

Or if dates cannot be changed and need sorting only by times use DataFrame.groupby with lambda function - groupby not sorting, because sort=False parameter and group_keys=False is for avoid MultiIndex:

df1 = (df.groupby(df['time'].dt.date, sort=False, group_keys=False)
         .apply(lambda x: x.sort_values('time')))
print (df1)
                 time    Open
2 2017-01-03 07:00:00  5.2128
1 2017-01-03 08:00:00  5.2180
0 2017-01-03 09:00:00  5.2475
5 2017-01-02 07:00:00  5.2475
4 2017-01-02 08:00:00  5.2123
3 2017-01-02 09:00:00  5.4122
8 2017-01-01 07:00:00  5.4122
7 2017-01-01 08:00:00  5.2128
6 2017-01-01 09:00:00  5.2180