Allow melt(var_name=) to be a falsy label (#16981)

closes #16972

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

Approvers:
  - Vyas Ramasubramani (https://github.com/vyasr)

URL: #16981

python/cudf/cudf/core/reshape.py

@@ -681,7 +681,7 @@ def _tile(A, reps):
     nval = len(value_vars)
     dtype = min_unsigned_type(nval)
 
-    if not var_name:
+    if var_name is None:
         var_name = "variable"
 
     if not value_vars:

python/cudf/cudf/tests/test_reshape.py

@@ -119,6 +119,15 @@ def test_melt_str_scalar_id_var():
     assert_eq(result, expected)
 
 
+def test_melt_falsy_var_name():
+    df = cudf.DataFrame({"A": ["a", "b", "c"], "B": [1, 3, 5], "C": [2, 4, 6]})
+    result = cudf.melt(df, id_vars=["A"], value_vars=["B"], var_name="")
+    expected = pd.melt(
+        df.to_pandas(), id_vars=["A"], value_vars=["B"], var_name=""
+    )
+    assert_eq(result, expected)
+
+
 @pytest.mark.parametrize("num_cols", [1, 2, 10])
 @pytest.mark.parametrize("num_rows", [1, 2, 1000])
 @pytest.mark.parametrize(