[BUG] DataFrame columns rename. Different behaviour with regards Pandas

[miguelusque]

Describe the bug
In Pandas, when renaming a column to a column name that already exists in the DataFrame, two columns with the same column name are created.

When performing the same operation in cuDF, the column to be renamed is deleted from the DataFrame.

Steps/Code to reproduce bug

import cudf
import pandas as pd

pdf = pd.DataFrame({"a": [1, 2, 3], "b": [3, 4, 5]})
gdf = cudf.DataFrame({"a": [1, 2, 3], "b": [3, 4, 5]})
pdf.rename(columns={"a": "b"}, inplace=True)
gdf.rename(columns={"a": "b"}, inplace=True)
print ("Pandas\n", pdf, "\n")
print ("cuDF\n", gdf)

---

Pandas
b b
0 1 3
1 2 4
2 3 5

cuDF
b
0 3
1 4
2 5

Hope it helps!
Miguel
Expected behavior
Same behaviour than Pandas, or at least an error/warning highlighting that the operation could not be performed.

Environment overview (please complete the following information)
cudf.version is '21.12.02'
DGX-1 Server

[beckernick]

Thanks for raising this issue. Silently removing the column is certainly not expected behavior here, as we throw in other situations such as directly setting column names with a duplicate label (ValueError: Duplicate column names are not allowed).