code sample for pandas-dev#44900

Author: simonjayhawkins

bisect/44900.py

@@ -0,0 +1,30 @@
+# BUG: Using pd.concat(axis="columns") on differently sized MultiIndexed
+# DataFrames with a datetime index level containing exclusively NaT values
+# causes the level in the returned DataFrame to be a float instead of a datetime
+# #44900
+
+import numpy as np
+
+import pandas as pd
+
+print(pd.__version__)
+df_a = pd.DataFrame({"a": range(5), "idx1": range(5), "idx2": [pd.NaT] * 5}).set_index(
+    ["idx1", "idx2"]
+)
+df_b = pd.DataFrame({"b": range(6), "idx1": range(6), "idx2": [pd.NaT] * 6}).set_index(
+    ["idx1", "idx2"]
+)
+result = pd.concat([df_a, df_b], axis="columns")
+print(result)
+
+expected = pd.DataFrame(
+    {
+        "a": list(range(5)) + [np.nan],
+        "b": range(6),
+        "idx1": range(6),
+        "idx2": [pd.NaT] * 6,
+    }
+).set_index(["idx1", "idx2"])
+expected
+
+pd.testing.assert_frame_equal(result, expected)