BUG: Series[Interval].fillna(mismatched) coerce instead of raise

doc/source/whatsnew/v1.5.0.rst

@@ -282,6 +282,7 @@ Indexing
 Missing
 ^^^^^^^
 - Bug in :meth:`Series.fillna` and :meth:`DataFrame.fillna` with ``downcast`` keyword not being respected in some cases where there are no NA values present (:issue:`45423`)
+- Bug in :meth:`Series.fillna` and :meth:`DataFrame.fillna` with :class:`IntervalDtype` and incompatible value raising instead of casting to a common (usually object) dtype (:issue:`??`)
 - Bug in :meth:`DataFrame.interpolate` with object-dtype column not returning a copy with ``inplace=False`` (:issue:`45791`)
 -
 

pandas/core/internals/blocks.py

@@ -1139,7 +1139,11 @@ def shift(self, periods: int, axis: int = 0, fill_value: Any = None) -> list[Blo
         fill_value = self._standardize_fill_value(fill_value)
 
         try:
-            casted = np_can_hold_element(self.dtype, fill_value)
+            # error: Argument 1 to "np_can_hold_element" has incompatible type
+            # "Union[dtype[Any], ExtensionDtype]"; expected "dtype[Any]"
+            casted = np_can_hold_element(
+                self.dtype, fill_value  # type: ignore[arg-type]
+            )
         except LossySetitemError:
             nb = self.coerce_to_target_dtype(fill_value)
             return nb.shift(periods, axis=axis, fill_value=fill_value)
@@ -1465,11 +1469,6 @@ def fillna(
                 # Discussion about what we want to support in the general
                 #  case GH#39584
                 blk = self.coerce_to_target_dtype(value)
-                if blk.dtype == _dtype_obj:
-                    # For now at least, only support casting e.g.
-                    #  Interval[int64]->Interval[float64],
-                    raise
-                # Never actually reached, but *could* be possible pending GH#45412
                 return blk.fillna(value, limit, inplace, downcast)
 
             elif isinstance(self, NDArrayBackedExtensionBlock):

pandas/tests/indexing/test_coercion.py

@@ -709,6 +709,30 @@ def test_fillna_datetime64tz(self, index_or_series, fill_val, fill_dtype):
         with tm.assert_produces_warning(warn, match="mismatched timezone"):
             self._assert_fillna_conversion(obj, fill_val, exp, fill_dtype)
 
+    @pytest.mark.parametrize(
+        "fill_val",
+        [
+            1,
+            1.1,
+            1 + 1j,
+            True,
+            pd.Interval(1, 2, closed="left"),
+            pd.Timestamp("2012-01-01", tz="US/Eastern"),
+            pd.Timestamp("2012-01-01"),
+            pd.Timedelta(days=1),
+            pd.Period("2016-01-01", "D"),
+        ],
+    )
+    def test_fillna_interval(self, index_or_series, fill_val):
+        ii = pd.interval_range(1.0, 5.0, closed="right").insert(1, np.nan)
+        assert isinstance(ii.dtype, pd.IntervalDtype)
+        obj = index_or_series(ii)
+
+        exp = index_or_series([ii[0], fill_val, ii[2], ii[3], ii[4]], dtype=object)
+
+        fill_dtype = object
+        self._assert_fillna_conversion(obj, fill_val, exp, fill_dtype)
+
     @pytest.mark.xfail(reason="Test not implemented")
     def test_fillna_series_int64(self):
         raise NotImplementedError