BUG-16807-1 SparseFrame fills with default_fill_value if data is None (#24842)

Closes gh-16807.

Author: JustinZhengBC

doc/source/whatsnew/v0.25.0.rst

@@ -247,7 +247,7 @@ Sparse
 ^^^^^^
 
 - Significant speedup in `SparseArray` initialization that benefits most operations, fixing performance regression introduced in v0.20.0 (:issue:`24985`)
--
+- Bug in :class:`SparseFrame` constructor where passing ``None`` as the data would cause ``default_fill_value`` to be ignored (:issue:`16807`)
 -
 
 

pandas/core/sparse/frame.py

@@ -124,8 +124,8 @@ def __init__(self, data=None, index=None, columns=None, default_kind=None,
                 columns = Index([])
             else:
                 for c in columns:
-                    data[c] = SparseArray(np.nan, index=index,
-                                          kind=self._default_kind,
+                    data[c] = SparseArray(self._default_fill_value,
+                                          index=index, kind=self._default_kind,
                                           fill_value=self._default_fill_value)
             mgr = to_manager(data, columns, index)
             if dtype is not None:

pandas/tests/sparse/frame/test_frame.py

@@ -270,6 +270,19 @@ def test_type_coercion_at_construction(self):
             default_fill_value=0)
         tm.assert_sp_frame_equal(result, expected)
 
+    def test_default_dtype(self):
+        result = pd.SparseDataFrame(columns=list('ab'), index=range(2))
+        expected = pd.SparseDataFrame([[np.nan, np.nan], [np.nan, np.nan]],
+                                      columns=list('ab'), index=range(2))
+        tm.assert_sp_frame_equal(result, expected)
+
+    def test_nan_data_with_int_dtype_raises_error(self):
+        sdf = pd.SparseDataFrame([[np.nan, np.nan], [np.nan, np.nan]],
+                                 columns=list('ab'), index=range(2))
+        msg = "Cannot convert non-finite values"
+        with pytest.raises(ValueError, match=msg):
+            pd.SparseDataFrame(sdf, dtype=np.int64)
+
     def test_dtypes(self):
         df = DataFrame(np.random.randn(10000, 4))
         df.loc[:9998] = np.nan
@@ -1263,6 +1276,14 @@ def test_notna(self):
                             'B': [True, False, True, True, False]})
         tm.assert_frame_equal(res.to_dense(), exp)
 
+    def test_default_fill_value_with_no_data(self):
+        # GH 16807
+        expected = pd.SparseDataFrame([[1.0, 1.0], [1.0, 1.0]],
+                                      columns=list('ab'), index=range(2))
+        result = pd.SparseDataFrame(columns=list('ab'), index=range(2),
+                                    default_fill_value=1.0)
+        tm.assert_frame_equal(expected, result)
+
 
 class TestSparseDataFrameArithmetic(object):