roll/rotate a groupby Dataframe including the index

Question

I want to rotate a Dataframe according to its groups and I need the Index to rotate with the rows as well. I want to do that, to then use a find_peaks function to get the peaks that are on the edges, because for me the data is kind a like a loop. Following is a df example with randome number values, where the df.groupby('Value a') is made.

idx	ValueA	ValueB
1	a	0.123
2	a	0.253
3	a	0.456
4	a	0.789
5	a	0.147
6	b	0.258
7	b	0.369
8	b	0.321
9	b	0.654
10	b	0.987

And this is how it should look like afterwards:

idx	ValueA	ValueB
4	a	0.789
5	a	0.147
1	a	0.123
2	a	0.253
3	a	0.456
9	b	0.654
10	b	0.987
6	b	0.258
7	b	0.369
8	b	0.321

I tried:

def roll_frame(df, shift):
    return pd.DataFrame(np.roll(df, shift, axis=0), index=df.index, columns=df.columns)

df_new = df.groupby('Value a').apply(roll_frame, 2)

The Value is shifted correctly, but not the Index. To erase the double Index, I used:

df_new.reset_index(level=0, drop=True, inplace=True)

which resulted in this:

idx	ValueA	ValueB
1	a	0.789
2	a	0.147
3	a	0.123
4	a	0.253
5	a	0.456
6	b	0.654
7	b	0.987
8	b	0.258
9	b	0.369
10	b	0.321

Answer 1

The easiest way I can think of would be to np.roll() the index as well:

def roll_frame(df, shift):
    return pd.DataFrame(
        np.roll(df, shift, axis=0), index=np.roll(df.index, shift), columns=df.columns
    )


df_new = df.groupby("ValueA").apply(roll_frame, 2)
print(df_new)

Prints:

          ValueA ValueB
ValueA                 
a      4       a  0.789
       5       a  0.147
       1       a  0.123
       2       a  0.253
       3       a  0.456
b      9       b  0.654
       10      b  0.987
       6       b  0.258
       7       b  0.369
       8       b  0.321