[BUG] Constructing a DataFrame with a multi-level column does not work

[shwina]

columns = cudf.MultiIndex.from_tuples([('A', 'one'), ('A', 'two')])
df = cudf.DataFrame(np.random.randn(2, 2), columns=columns)
df

Output:

  (A, one)  (A, two)
0 -1.004870  1.253556
1  0.176658  0.856762

Expected output:

          A
        one       two
0 -0.926630  0.718061
1 -0.920203 -0.074759

[wence-]

Maybe this?

diff --git a/python/cudf/cudf/core/dataframe.py b/python/cudf/cudf/core/dataframe.py
index 0fe8949090..cc57091539 100644
--- a/python/cudf/cudf/core/dataframe.py
+++ b/python/cudf/cudf/core/dataframe.py
@@ -722,6 +722,8 @@ class DataFrame(IndexedFrame, Serializable, GetAttrGetItemMixin):
 
         if dtype:
             self._data = self.astype(dtype)._data
+        # Fix up
+        self._data.multiindex = isinstance(columns, pd.MultiIndex)
 
     @_cudf_nvtx_annotate
     def _init_from_series_list(self, data, columns, index):