BUG: Series[listlike_of_ints] incorrect on MultiIndex (#33539)

Author: jbrockmendel

doc/source/whatsnew/v1.1.0.rst

@@ -325,6 +325,36 @@ key and type of :class:`Index`.  These now consistently raise ``KeyError`` (:iss
     ...
     KeyError: Timestamp('1970-01-01 00:00:00')
 
+.. _whatsnew_110.api_breaking.indexing_int_multiindex_raises_key_errors:
+
+Failed Integer Lookups on MultiIndex Raise KeyError
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+Indexing with integers with a :class:`MultiIndex` that has a integer-dtype
+first level incorrectly failed to raise ``KeyError`` when one or more of
+those integer keys is not present in the first level of the index (:issue:`33539`)
+
+.. ipython:: python
+
+    idx = pd.Index(range(4))
+    dti = pd.date_range("2000-01-03", periods=3)
+    mi = pd.MultiIndex.from_product([idx, dti])
+    ser = pd.Series(range(len(mi)), index=mi)
+
+*Previous behavior*:
+
+.. code-block:: ipython
+
+    In [5]: ser[[5]]
+    Out[5]: Series([], dtype: int64)
+
+*New behavior*:
+
+.. code-block:: ipython
+
+    In [5]: ser[[5]]
+    ...
+    KeyError: '[5] not in index'
+
 :meth:`DataFrame.merge` preserves right frame's row order
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 :meth:`DataFrame.merge` now preserves right frame's row order when executing a right merge (:issue:`27453`)

pandas/core/indexing.py

@@ -1082,37 +1082,6 @@ def _getitem_axis(self, key, axis: int):
             return self._getbool_axis(key, axis=axis)
         elif is_list_like_indexer(key):
 
-            # convert various list-like indexers
-            # to a list of keys
-            # we will use the *values* of the object
-            # and NOT the index if its a PandasObject
-            if isinstance(labels, ABCMultiIndex):
-
-                if isinstance(key, (ABCSeries, np.ndarray)) and key.ndim <= 1:
-                    # Series, or 0,1 ndim ndarray
-                    # GH 14730
-                    key = list(key)
-                elif isinstance(key, ABCDataFrame):
-                    # GH 15438
-                    raise NotImplementedError(
-                        "Indexing a MultiIndex with a "
-                        "DataFrame key is not "
-                        "implemented"
-                    )
-                elif hasattr(key, "ndim") and key.ndim > 1:
-                    raise NotImplementedError(
-                        "Indexing a MultiIndex with a "
-                        "multidimensional key is not "
-                        "implemented"
-                    )
-
-                if (
-                    not isinstance(key, tuple)
-                    and len(key)
-                    and not isinstance(key[0], tuple)
-                ):
-                    key = tuple([key])
-
             # an iterable multi-selection
             if not (isinstance(key, tuple) and isinstance(labels, ABCMultiIndex)):
 

pandas/tests/indexing/multiindex/test_loc.py

@@ -134,16 +134,15 @@ def test_loc_multiindex_missing_label_raises(self):
 
     @pytest.mark.parametrize("key, pos", [([2, 4], [0, 1]), ([2], []), ([2, 3], [])])
     def test_loc_multiindex_list_missing_label(self, key, pos):
-        # GH 27148 - lists with missing labels do not raise:
+        # GH 27148 - lists with missing labels _do_ raise
         df = DataFrame(
             np.random.randn(3, 3),
             columns=[[2, 2, 4], [6, 8, 10]],
             index=[[4, 4, 8], [8, 10, 12]],
         )
 
-        expected = df.iloc[pos]
-        result = df.loc[key]
-        tm.assert_frame_equal(result, expected)
+        with pytest.raises(KeyError, match="not in index"):
+            df.loc[key]
 
     def test_loc_multiindex_too_many_dims_raises(self):
         # GH 14885
@@ -295,8 +294,8 @@ def convert_nested_indexer(indexer_type, keys):
     [
         ([], []),  # empty ok
         (["A"], slice(3)),
-        (["A", "D"], slice(3)),
-        (["D", "E"], []),  # no values found - fine
+        (["A", "D"], []),  # "D" isnt present -> raise
+        (["D", "E"], []),  # no values found -> raise
         (["D"], []),  # same, with single item list: GH 27148
         (pd.IndexSlice[:, ["foo"]], slice(2, None, 3)),
         (pd.IndexSlice[:, ["foo", "bah"]], slice(2, None, 3)),
@@ -310,8 +309,13 @@ def test_loc_getitem_duplicates_multiindex_missing_indexers(indexer, pos):
     )
     s = Series(np.arange(9, dtype="int64"), index=idx).sort_index()
     expected = s.iloc[pos]
-    result = s.loc[indexer]
-    tm.assert_series_equal(result, expected)
+
+    if expected.size == 0 and indexer != []:
+        with pytest.raises(KeyError, match=str(indexer)):
+            s.loc[indexer]
+    else:
+        result = s.loc[indexer]
+        tm.assert_series_equal(result, expected)
 
 
 def test_series_loc_getitem_fancy(multiindex_year_month_day_dataframe_random_data):

pandas/tests/indexing/multiindex/test_slice.py

@@ -118,11 +118,11 @@ def test_per_axis_per_level_getitem(self):
         with pytest.raises(ValueError, match=msg):
             df.loc[(slice(None), np.array([True, False])), :]
 
-        # ambiguous notation
-        # this is interpreted as slicing on both axes (GH #16396)
-        result = df.loc[slice(None), [1]]
-        expected = df.iloc[:, []]
-        tm.assert_frame_equal(result, expected)
+        with pytest.raises(KeyError, match=r"\[1\] not in index"):
+            # slice(None) is on the index, [1] is on the columns, but 1 is
+            #  not in the columns, so we raise
+            #  This used to treat [1] as positional GH#16396
+            df.loc[slice(None), [1]]
 
         result = df.loc[(slice(None), [1]), :]
         expected = df.iloc[[0, 3]]

pandas/tests/series/indexing/test_getitem.py

@@ -114,6 +114,19 @@ def test_getitem_intlist_intervalindex_non_int(self, box):
         result = ser[key]
         tm.assert_series_equal(result, expected)
 
+    @pytest.mark.parametrize("box", [list, np.array, pd.Index])
+    @pytest.mark.parametrize("dtype", [np.int64, np.float64, np.uint64])
+    def test_getitem_intlist_multiindex_numeric_level(self, dtype, box):
+        # GH#33404 do _not_ fall back to positional since ints are ambiguous
+        idx = pd.Index(range(4)).astype(dtype)
+        dti = date_range("2000-01-03", periods=3)
+        mi = pd.MultiIndex.from_product([idx, dti])
+        ser = Series(range(len(mi))[::-1], index=mi)
+
+        key = box([5])
+        with pytest.raises(KeyError, match="5"):
+            ser[key]
+
 
 def test_getitem_generator(string_series):
     gen = (x > 0 for x in string_series)