[BUG] Index of the RHS is being ignored in assignments

[magnatelee]

Describe the bug
cuDF seems to be ignoring the index of the RHS when performing an assignment.

Steps/Code to reproduce bug

In the following example, assignment (1) treats y as if it were a series with an index [0, 1, 2], which deviates from the semantics of Pandas:

>>> import cudf
>>> x = cudf.Series([1, 2, 3])
>>> x
0    1
1    2
2    3
dtype: int64
>>> y = cudf.Series([3, 2, 1], index=[2, 1, 0])
>>> x.iloc[:] = y  # (1)                                                
>>> x
0    3
1    2
2    1
dtype: int64

Here is the result of the same program in Pandas:

>>> import pandas as pd
>>> x = pd.Series([1, 2, 3])
>>> x
0    1
1    2
2    3
dtype: int64
>>> y = pd.Series([3, 2, 1], index=[2, 1, 0])
>>> x.iloc[:] = y
>>> x
0    1
1    2
2    3
dtype: int64

Expected behavior
cuDF should yield the same result as Pandas. I believe internally Pandas performs a left outer join between indices of the LHS and the RHS to find the values of the RHS that correspond to the LHS.

[isVoid]

As of Pandas 1.2.0, they seem to have reintroduced the behavior:

>>> x = pd.Series([1, 2, 3])
>>> y = pd.Series([3, 2, 1], index=[2, 1, 0])
>>> x.iloc[:] = y
>>> x
0    3
1    2
2    1
dtype: int64

[isVoid]

Reported: pandas-dev/pandas#40184

[isVoid]

According to the Pandas thread, it seems the reported behavior is expected. Closing.