BUG: Preserve dtype in DataFrame.fillna when filling from another DataFrame

Author: iabhi4

doc/source/whatsnew/v3.0.0.rst

@@ -748,7 +748,7 @@ Indexing
 Missing
 ^^^^^^^
 - Bug in :meth:`DataFrame.fillna` and :meth:`Series.fillna` that would ignore the ``limit`` argument on :class:`.ExtensionArray` dtypes (:issue:`58001`)
--
+- Bug in :meth:`DataFrame.fillna` where filling from another ``DataFrame`` of the same dtype could incorrectly cast the result to ``object`` dtype. (:issue:`61568`)
 
 MultiIndex
 ^^^^^^^^^^

pandas/core/generic.py

@@ -7145,7 +7145,24 @@ def fillna(
             else:
                 new_data = self._mgr.fillna(value=value, limit=limit, inplace=inplace)
         elif isinstance(value, ABCDataFrame) and self.ndim == 2:
-            new_data = self.where(self.notna(), value)._mgr
+            filled_columns = {}
+            for col in self.columns:
+                lhs = self[col]
+                if col in value.columns:
+                    rhs = value[col]
+                    filled = lhs.where(notna(lhs), rhs)
+
+                    # restore original dtype if fallback to object occurred
+                    if lhs.dtype == rhs.dtype and filled.dtype == object:
+                        try:
+                            filled = filled.astype(lhs.dtype)
+                        except Exception:
+                            pass
+                else:
+                    filled = lhs
+                filled_columns[col] = filled
+
+            new_data = type(self)(filled_columns, index=self.index)._mgr
         else:
             raise ValueError(f"invalid fill value with a {type(value)}")
 

pandas/tests/frame/methods/test_fillna.py

@@ -795,3 +795,22 @@ def test_fillna_out_of_bounds_datetime():
     msg = "Cannot cast 0001-01-01 00:00:00 to unit='ns' without overflow"
     with pytest.raises(OutOfBoundsDatetime, match=msg):
         df.fillna(Timestamp("0001-01-01"))
+
+
+def test_fillna_dataframe_preserves_dtypes_mixed_columns():
+    # GH#61568
+    empty = DataFrame([[None] * 4] * 4, columns=list("ABCD"), dtype=np.float64)
+    full = DataFrame(
+        [
+            [1.0, 2.0, "3.0", 4.0],
+            [5.0, 6.0, "7.0", 8.0],
+            [9.0, 10.0, "11.0", 12.0],
+            [13.0, 14.0, "15.0", 16.0],
+        ],
+        columns=list("ABCD"),
+    )
+    result = empty.fillna(full)
+    expected_dtypes = Series(
+        {"A": "float64", "B": "float64", "C": "object", "D": "float64"}
+    )
+    tm.assert_series_equal(result.dtypes, expected_dtypes)