BUG: regression in master for DataFrame.sparse.from_spmatrix #59212

doc/source/whatsnew/v3.0.0.rst

@@ -622,7 +622,7 @@ Reshaping
 Sparse
 ^^^^^^
 - Bug in :class:`SparseDtype` for equal comparison with na fill value. (:issue:`54770`)
-- Bug in :meth:`DataFrame.sparse.from_spmatrix` which hard coded an invalid ``fill_value`` for certain subtypes. (:issue:`59063`)
+- Bug in :meth:`DataFrame.sparse.from_spmatrix` which hard coded an invalid ``fill_value`` for certain subtypes. (:issue:`59063`, :issue:`59212`)
 
 ExtensionArray
 ^^^^^^^^^^^^^^

pandas/core/dtypes/dtypes.py

@@ -1682,8 +1682,8 @@ class SparseDtype(ExtensionDtype):
         =========== ==========
         dtype       na_value
         =========== ==========
-        float       ``np.nan``
-        complex     ``np.nan``
+        float       ``0.0``
+        complex     ``0.0 + 0.0j``
         int         ``0``
         bool        ``False``
         datetime64  ``pd.NaT``
@@ -1747,6 +1747,14 @@ def __init__(self, dtype: Dtype = np.float64, fill_value: Any = None) -> None:
         if fill_value is None:
             fill_value = na_value_for_dtype(dtype)
 
+            # Default values for float type is NaN. Hence, in order to create a Sparse
+            # matrix of type float, we need to override this default value as 0.0.
+            if dtype.kind in "f":
+                fill_value = 0.0
+            # Similarly, default value needs to be overridden for complex type arrays.
+            elif dtype.kind in "c":
+                fill_value = 0.0 + 0.0j
+
         self._dtype = dtype
         self._fill_value = fill_value
         self._check_fill_value()