TST: GroupBy.ffill on a multi-index dataframe (pandas-dev#55408)

* add new automated test with ffill on a multi-index dataframe

This test checks whether the ffill method works correctly on a
multi-index dataframe. In it, we define a dataframe with indexes
[0, 1, 2, 0, 1, 2] and column "a" with values [1, 2, NaN, 3, 4, 5].
We group this dataframe by columns (level = 0) and shift it by one,
so we have:

0: [1, 3] -> [NaN, 1]
1: [2, 4] -> [NaN, 2]
2: [NaN, 5] -> [NaN, NaN]

Then, since index order remain the same, if we apply ffill method, it
should give us a dataframe with column "a" equal to [NaN, NaN, NaN, 1,
2, 2].

Co-authored-by: José Lucas Silva Mayer <[email protected]>
Co-authored-by: Willian Wang <[email protected]>

* remove shift and add cast dataframe dtype to float

Signed-off-by: José Lucas Silva Mayer <[email protected]>

* remove check of dataframe dtype on assert

Signed-off-by: José Lucas Silva Mayer <[email protected]>

---------

Signed-off-by: José Lucas Silva Mayer <[email protected]>
Co-authored-by: Willian Wang <[email protected]>

Author: josemayer

pandas/tests/groupby/test_groupby.py

@@ -3242,3 +3242,12 @@ def test_get_group_axis_1():
         }
     )
     tm.assert_frame_equal(result, expected)
+
+
+def test_groupby_ffill_with_duplicated_index():
+    # GH#43412
+    df = DataFrame({"a": [1, 2, 3, 4, np.nan, np.nan]}, index=[0, 1, 2, 0, 1, 2])
+
+    result = df.groupby(level=0).ffill()
+    expected = DataFrame({"a": [1, 2, 3, 4, 2, 3]}, index=[0, 1, 2, 0, 1, 2])
+    tm.assert_frame_equal(result, expected, check_dtype=False)