QST: is the new behavior of GroupByRolling in v1.1.0 intended?

[itholic]

Question about pandas

Let's say we have a DataFrame like the below.

>>> pdf = pd.DataFrame({"a": [1, 2, 3, 2], "b": [4.0, 2.0, 3.0, 1.0], "c": [10, 20, 30, 20]})
>>> pdf
   a    b   c
0  1  4.0  10
1  2  2.0  20
2  3  3.0  30
3  2  1.0  20

Then, when I use GroupByRolling, In the version of pandas <= 1.0.5 shows result as below.

>>> pdf.groupby('a')[['b']].rolling(2).max()
       b
a
1 0  NaN
3 2  NaN
2 1  NaN
  3  2.0

However, In the pandas 1.1.0, the result seems different from the previous version as below.

>>> pdf.groupby('a')[['b']].rolling(2).max()
       a    b     c
a
1 0  NaN  NaN   NaN
2 1  NaN  NaN   NaN
  3  2.0  2.0  20.0
3 2  NaN  NaN   NaN

Could someone let me know Is it intended? or unexpected behavior (maybe kind of bug) ?

Thanks :)

[simonjayhawkins]

the change in behaviour is from #34052 cc @mroeschke

bad52a9 is the first bad commit
commit bad52a9
Author: Matthew Roeschke emailformattr@gmail.com
Date: Mon May 25 10:26:19 2020 -0700

PERF: Use Indexers to implement groupby rolling (#34052)

[itholic]

Thanks, @simonjayhawkins :)