DataFrame with namedtuples uses ._field as column names (#13824)

Allow namedtuple's `_field` attribute to be mapped to DataFrame column labels like pandas

closes #13823

Authors:
  - Matthew Roeschke (https://github.com/mroeschke)

Approvers:
  - GALI PREM SAGAR (https://github.com/galipremsagar)
  - Lawrence Mitchell (https://github.com/wence-)

URL: #13824

python/cudf/cudf/core/dataframe.py

@@ -848,6 +848,15 @@ def _init_from_list_like(self, data, index=None, columns=None):
                 for col in data
             ):
                 raise TypeError("Inputs should be an iterable or sequence.")
+            if (
+                len(data) > 0
+                and columns is None
+                and isinstance(data[0], tuple)
+                and hasattr(data[0], "_fields")
+            ):
+                # pandas behavior is to use the fields from the first
+                # namedtuple as the column names
+                columns = data[0]._fields
 
             data = list(itertools.zip_longest(*data))
 

python/cudf/cudf/tests/test_dataframe.py

@@ -10,7 +10,7 @@
 import string
 import textwrap
 import warnings
-from collections import OrderedDict, defaultdict
+from collections import OrderedDict, defaultdict, namedtuple
 from copy import copy
 
 import cupy
@@ -10261,3 +10261,21 @@ def __getitem__(self, key):
 
     with pytest.raises(TypeError):
         cudf.DataFrame({"a": A()})
+
+
+def test_dataframe_constructor_from_namedtuple():
+    Point1 = namedtuple("Point1", ["a", "b", "c"])
+    Point2 = namedtuple("Point1", ["x", "y"])
+
+    data = [Point1(1, 2, 3), Point2(4, 5)]
+    idx = ["a", "b"]
+    gdf = cudf.DataFrame(data, index=idx)
+    pdf = pd.DataFrame(data, index=idx)
+
+    assert_eq(gdf, pdf)
+
+    data = [Point2(4, 5), Point1(1, 2, 3)]
+    with pytest.raises(ValueError):
+        cudf.DataFrame(data, index=idx)
+    with pytest.raises(ValueError):
+        pd.DataFrame(data, index=idx)