BUG: assign doesnt cast SparseDataFrame to DataFrame (#19178)

doc/source/whatsnew/v0.23.0.txt

@@ -822,6 +822,7 @@ Sparse
 - Bug in which creating a ``SparseDataFrame`` from a dense ``Series`` or an unsupported type raised an uncontrolled exception (:issue:`19374`)
 - Bug in :class:`SparseDataFrame.to_csv` causing exception (:issue:`19384`)
 - Bug in :class:`SparseSeries.memory_usage` which caused segfault by accessing non sparse elements (:issue:`19368`)
+- Bug in constructing a ``SparseArray``: if ``data`` is a scalar and ``index`` is defined it will coerce to ``float64`` regardless of scalar's dtype. (:issue:`19163`)
 
 Reshaping
 ^^^^^^^^^

pandas/core/sparse/array.py

@@ -26,7 +26,8 @@
     is_scalar, is_dtype_equal)
 from pandas.core.dtypes.cast import (
     maybe_convert_platform, maybe_promote,
-    astype_nansafe, find_common_type)
+    astype_nansafe, find_common_type, infer_dtype_from_scalar,
+    construct_1d_arraylike_from_scalar)
 from pandas.core.dtypes.missing import isna, notna, na_value_for_dtype
 
 import pandas._libs.sparse as splib
@@ -162,9 +163,9 @@ def __new__(cls, data, sparse_index=None, index=None, kind='integer',
                 data = np.nan
             if not is_scalar(data):
                 raise Exception("must only pass scalars with an index ")
-            values = np.empty(len(index), dtype='float64')
-            values.fill(data)
-            data = values
+            dtype = infer_dtype_from_scalar(data)[0]
+            data = construct_1d_arraylike_from_scalar(
+                data, len(index), dtype)
 
         if isinstance(data, ABCSparseSeries):
             data = data.values

pandas/tests/sparse/frame/test_frame.py

@@ -1257,3 +1257,14 @@ def test_quantile_multi(self):
 
         tm.assert_frame_equal(result, dense_expected)
         tm.assert_sp_frame_equal(result, sparse_expected)
+
+    def test_assign_with_sparse_frame(self):
+        # GH 19163
+        df = pd.DataFrame({"a": [1, 2, 3]})
+        res = df.to_sparse(fill_value=False).assign(newcol=False)
+        exp = df.assign(newcol=False).to_sparse(fill_value=False)
+
+        tm.assert_sp_frame_equal(res, exp)
+
+        for column in res.columns:
+            assert type(res[column]) is SparseSeries

pandas/tests/sparse/test_array.py

@@ -113,6 +113,21 @@ def test_constructor_spindex_dtype(self):
         assert arr.dtype == np.int64
         assert arr.fill_value == 0
 
+    @pytest.mark.parametrize('scalar,dtype', [
+        (False, bool),
+        (0.0, 'float64'),
+        (1, 'int64'),
+        ('z', 'object')])
+    def test_scalar_with_index_infer_dtype(self, scalar, dtype):
+        # GH 19163
+        arr = SparseArray(scalar, index=[1, 2, 3], fill_value=scalar)
+        exp = SparseArray([scalar, scalar, scalar], fill_value=scalar)
+
+        tm.assert_sp_array_equal(arr, exp)
+
+        assert arr.dtype == dtype
+        assert exp.dtype == dtype
+
     def test_sparseseries_roundtrip(self):
         # GH 13999
         for kind in ['integer', 'block']: