BUG: IntervalArray.insert cast on failure (#40359)

Author: jbrockmendel

pandas/core/arrays/interval.py

@@ -1487,6 +1487,28 @@ def putmask(self, mask: np.ndarray, value) -> None:
             self._left.putmask(mask, value_left)
             self._right.putmask(mask, value_right)
 
+    def insert(self: IntervalArrayT, loc: int, item: Interval) -> IntervalArrayT:
+        """
+        Return a new IntervalArray inserting new item at location. Follows
+        Python list.append semantics for negative values.  Only Interval
+        objects and NA can be inserted into an IntervalIndex
+
+        Parameters
+        ----------
+        loc : int
+        item : Interval
+
+        Returns
+        -------
+        IntervalArray
+        """
+        left_insert, right_insert = self._validate_scalar(item)
+
+        new_left = self.left.insert(loc, left_insert)
+        new_right = self.right.insert(loc, right_insert)
+
+        return self._shallow_copy(new_left, new_right)
+
     def delete(self: IntervalArrayT, loc) -> IntervalArrayT:
         if isinstance(self._left, np.ndarray):
             new_left = np.delete(self._left, loc)

pandas/core/indexes/interval.py

@@ -873,11 +873,14 @@ def insert(self, loc, item):
         -------
         IntervalIndex
         """
-        left_insert, right_insert = self._data._validate_scalar(item)
+        try:
+            result = self._data.insert(loc, item)
+        except (ValueError, TypeError):
+            # e.g trying to insert a string
+            dtype, _ = infer_dtype_from_scalar(item, pandas_dtype=True)
+            dtype = find_common_type([self.dtype, dtype])
+            return self.astype(dtype).insert(loc, item)
 
-        new_left = self.left.insert(loc, left_insert)
-        new_right = self.right.insert(loc, right_insert)
-        result = self._data._shallow_copy(new_left, new_right)
         return type(self)._simple_new(result, name=self.name)
 
     # --------------------------------------------------------------------

pandas/tests/indexes/interval/test_interval.py

@@ -194,17 +194,24 @@ def test_insert(self, data):
         tm.assert_index_equal(result, expected)
 
         # invalid type
+        res = data.insert(1, "foo")
+        expected = data.astype(object).insert(1, "foo")
+        tm.assert_index_equal(res, expected)
+
         msg = "can only insert Interval objects and NA into an IntervalArray"
         with pytest.raises(TypeError, match=msg):
-            data.insert(1, "foo")
+            data._data.insert(1, "foo")
 
         # invalid closed
         msg = "'value.closed' is 'left', expected 'right'."
         for closed in {"left", "right", "both", "neither"} - {item.closed}:
             msg = f"'value.closed' is '{closed}', expected '{item.closed}'."
+            bad_item = Interval(item.left, item.right, closed=closed)
+            res = data.insert(1, bad_item)
+            expected = data.astype(object).insert(1, bad_item)
+            tm.assert_index_equal(res, expected)
             with pytest.raises(ValueError, match=msg):
-                bad_item = Interval(item.left, item.right, closed=closed)
-                data.insert(1, bad_item)
+                data._data.insert(1, bad_item)
 
         # GH 18295 (test missing)
         na_idx = IntervalIndex([np.nan], closed=data.closed)
@@ -214,13 +221,15 @@ def test_insert(self, data):
             tm.assert_index_equal(result, expected)
 
         if data.left.dtype.kind not in ["m", "M"]:
-            # trying to insert pd.NaT into a numeric-dtyped Index should cast/raise
+            # trying to insert pd.NaT into a numeric-dtyped Index should cast
+            expected = data.astype(object).insert(1, pd.NaT)
+
             msg = "can only insert Interval objects and NA into an IntervalArray"
             with pytest.raises(TypeError, match=msg):
-                result = data.insert(1, pd.NaT)
-        else:
-            result = data.insert(1, pd.NaT)
-            tm.assert_index_equal(result, expected)
+                data._data.insert(1, pd.NaT)
+
+        result = data.insert(1, pd.NaT)
+        tm.assert_index_equal(result, expected)
 
     def test_is_unique_interval(self, closed):
         """

pandas/tests/indexing/test_loc.py

@@ -23,7 +23,6 @@
     DatetimeIndex,
     Index,
     IndexSlice,
-    IntervalIndex,
     MultiIndex,
     Period,
     Series,
@@ -1680,9 +1679,6 @@ def test_loc_setitem_with_expansion_nonunique_index(self, index, request):
         # GH#40096
         if not len(index):
             return
-        if isinstance(index, IntervalIndex):
-            mark = pytest.mark.xfail(reason="IntervalIndex raises")
-            request.node.add_marker(mark)
 
         index = index.repeat(2)  # ensure non-unique
         N = len(index)