BUG: astype_nansafe with copy=False (#38835)

Author: jbrockmendel

doc/source/whatsnew/v1.3.0.rst

@@ -297,7 +297,7 @@ Sparse
 ^^^^^^
 
 - Bug in :meth:`DataFrame.sparse.to_coo` raising ``KeyError`` with columns that are a numeric :class:`Index` without a 0 (:issue:`18414`)
--
+- Bug in :meth:`SparseArray.astype` with ``copy=False`` producing incorrect results when going from integer dtype to floating dtype (:issue:`34456`)
 -
 
 ExtensionArray

pandas/core/dtypes/cast.py

@@ -1133,11 +1133,11 @@ def astype_nansafe(
         )
         raise ValueError(msg)
 
-    if copy or is_object_dtype(arr) or is_object_dtype(dtype):
+    if copy or is_object_dtype(arr.dtype) or is_object_dtype(dtype):
         # Explicit copy, or required since NumPy can't view from / to object.
         return arr.astype(dtype, copy=True)
 
-    return arr.view(dtype)
+    return arr.astype(dtype, copy=copy)
 
 
 def soft_convert_objects(

pandas/tests/arrays/sparse/test_array.py

@@ -563,6 +563,14 @@ def test_astype_nan_raises(self):
         with pytest.raises(ValueError, match="Cannot convert non-finite"):
             arr.astype(int)
 
+    def test_astype_copy_false(self):
+        # GH#34456 bug caused by using .view instead of .astype in astype_nansafe
+        arr = SparseArray([1, 2, 3])
+
+        result = arr.astype(float, copy=False)
+        expected = SparseArray([1.0, 2.0, 3.0], fill_value=0.0)
+        tm.assert_sp_array_equal(result, expected)
+
     def test_set_fill_value(self):
         arr = SparseArray([1.0, np.nan, 2.0], fill_value=np.nan)
         arr.fill_value = 2

pandas/tests/dtypes/test_common.py

@@ -724,6 +724,17 @@ def test_astype_nansafe(val, typ):
             astype_nansafe(arr, dtype=typ)
 
 
+def test_astype_nansafe_copy_false(any_int_dtype):
+    # GH#34457 use astype, not view
+    arr = np.array([1, 2, 3], dtype=any_int_dtype)
+
+    dtype = np.dtype("float64")
+    result = astype_nansafe(arr, dtype, copy=False)
+
+    expected = np.array([1.0, 2.0, 3.0], dtype=dtype)
+    tm.assert_numpy_array_equal(result, expected)
+
+
 @pytest.mark.parametrize("from_type", [np.datetime64, np.timedelta64])
 @pytest.mark.parametrize(
     "to_type",