How to group by and get 1 max value and 2 max value rows in pandas?

Question

Raw Data...

RefNo	rest1	Pair	rest2	rest3
A	11	E	55	7
B	22	F	66	6
C	33	G	77	5
C	33	H	88	4
D	44	I	99	3
D	44	J	111	2
D	44	K	222	1

1. Groupby(RefNo) and get MaxValue(rest2) row...

RefNo	rest1	Pair	rest2	rest3
A	11	E	55	7
B	22	F	66	6
C	33	H	88	4
D	44	K	222	1

2. Groupby(RefNo) and get 2 MaxValue(rest2) rows...

RefNo	rest1	Pair	rest2	rest3
A	11	E	55	7
B	22	F	66	6
C	33	H	88	4
C	33	G	77	5
D	44	K	222	1
D	44	J	111	2

Appreciate any help.

Answer 1

Using groupby.rank and boolean indexing:

n = 2

out = df[df.groupby('RefNo')['rest2'].rank('dense').le(n)]

Output:

  RefNo  rest1 Pair  rest2  rest3
0     A     11    E     55      7
1     B     22    F     66      6
2     C     33    G     77      5
3     C     33    H     88      4
4     D     44    I     99      3
5     D     44    J    111      2

Intermediates:

  RefNo  rest1 Pair  rest2  rest3  rank   mask
0     A     11    E     55      7   1.0   True
1     B     22    F     66      6   1.0   True
2     C     33    G     77      5   1.0   True
3     C     33    H     88      4   2.0   True
4     D     44    I     99      3   1.0   True
5     D     44    J    111      2   2.0   True
6     D     44    K    222      1   3.0  False