Bug in loc not raising KeyError with MultiIndex containing no longer used levels (pandas-dev#41358)

Author: phofl

doc/source/whatsnew/v1.3.0.rst

@@ -788,6 +788,7 @@ Indexing
 - Bug in setting ``numpy.timedelta64`` values into an object-dtype :class:`Series` using a boolean indexer (:issue:`39488`)
 - Bug in setting numeric values into a into a boolean-dtypes :class:`Series` using ``at`` or ``iat`` failing to cast to object-dtype (:issue:`39582`)
 - Bug in :meth:`DataFrame.__setitem__` and :meth:`DataFrame.iloc.__setitem__` raising ``ValueError`` when trying to index with a row-slice and setting a list as values (:issue:`40440`)
+- Bug in :meth:`DataFrame.loc` not raising ``KeyError`` when key was not found in :class:`MultiIndex` when levels contain more values than used (:issue:`41170`)
 - Bug in :meth:`DataFrame.loc.__setitem__` when setting-with-expansion incorrectly raising when the index in the expanding axis contains duplicates (:issue:`40096`)
 - Bug in :meth:`DataFrame.loc` incorrectly matching non-boolean index elements (:issue:`20432`)
 - Bug in :meth:`Series.__delitem__` with ``ExtensionDtype`` incorrectly casting to ``ndarray`` (:issue:`40386`)
@@ -808,6 +809,7 @@ MultiIndex
 - Bug in :meth:`MultiIndex.intersection` duplicating ``NaN`` in result (:issue:`38623`)
 - Bug in :meth:`MultiIndex.equals` incorrectly returning ``True`` when :class:`MultiIndex` containing ``NaN`` even when they are differently ordered (:issue:`38439`)
 - Bug in :meth:`MultiIndex.intersection` always returning empty when intersecting with :class:`CategoricalIndex` (:issue:`38653`)
+- Bug in :meth:`MultiIndex.reindex` raising ``ValueError`` with empty MultiIndex and indexing only a specific level (:issue:`41170`)
 
 I/O
 ^^^

pandas/core/indexes/base.py

@@ -4224,7 +4224,8 @@ def _get_leaf_sorter(labels: list[np.ndarray]) -> np.ndarray:
 
             else:  # tie out the order with other
                 if level == 0:  # outer most level, take the fast route
-                    ngroups = 1 + new_lev_codes.max()
+                    max_new_lev = 0 if len(new_lev_codes) == 0 else new_lev_codes.max()
+                    ngroups = 1 + max_new_lev
                     left_indexer, counts = libalgos.groupsort_indexer(
                         new_lev_codes, ngroups
                     )

pandas/core/indexes/multi.py

@@ -72,6 +72,7 @@
 from pandas.core.arrays import Categorical
 from pandas.core.arrays.categorical import factorize_from_iterables
 import pandas.core.common as com
+from pandas.core.indexers import is_empty_indexer
 import pandas.core.indexes.base as ibase
 from pandas.core.indexes.base import (
     Index,
@@ -2634,6 +2635,10 @@ def _convert_listlike_indexer(self, keyarr):
             mask = check == -1
             if mask.any():
                 raise KeyError(f"{keyarr[mask]} not in index")
+            elif is_empty_indexer(indexer, keyarr):
+                # We get here when levels still contain values which are not
+                # actually in Index anymore
+                raise KeyError(f"{keyarr} not in index")
 
         return indexer, keyarr
 

pandas/tests/indexes/multi/test_reindex.py

@@ -104,3 +104,14 @@ def test_reindex_non_unique():
     msg = "cannot handle a non-unique multi-index!"
     with pytest.raises(ValueError, match=msg):
         a.reindex(new_idx)
+
+
+@pytest.mark.parametrize("values", [[["a"], ["x"]], [[], []]])
+def test_reindex_empty_with_level(values):
+    # GH41170
+    idx = MultiIndex.from_arrays(values)
+    result, result_indexer = idx.reindex(np.array(["b"]), level=0)
+    expected = MultiIndex(levels=[["b"], values[1]], codes=[[], []])
+    expected_indexer = np.array([], dtype=result_indexer.dtype)
+    tm.assert_index_equal(result, expected)
+    tm.assert_numpy_array_equal(result_indexer, expected_indexer)

pandas/tests/indexing/test_loc.py

@@ -1624,6 +1624,13 @@ def test_loc_getitem_preserves_index_level_category_dtype(self):
         result = df.loc[["a"]].index.levels[0]
         tm.assert_index_equal(result, expected)
 
+    @pytest.mark.parametrize("lt_value", [30, 10])
+    def test_loc_multiindex_levels_contain_values_not_in_index_anymore(self, lt_value):
+        # GH#41170
+        df = DataFrame({"a": [12, 23, 34, 45]}, index=[list("aabb"), [0, 1, 2, 3]])
+        with pytest.raises(KeyError, match=r"\['b'\] not in index"):
+            df.loc[df["a"] < lt_value, :].loc[["b"], :]
+
 
 class TestLocSetitemWithExpansion:
     @pytest.mark.slow

pandas/tests/series/methods/test_reindex.py

@@ -4,6 +4,7 @@
 from pandas import (
     Categorical,
     Index,
+    MultiIndex,
     NaT,
     Period,
     PeriodIndex,
@@ -345,3 +346,16 @@ def test_reindex_periodindex_with_object(p_values, o_values, values, expected_va
     result = ser.reindex(object_index)
     expected = Series(expected_values, index=object_index)
     tm.assert_series_equal(result, expected)
+
+
+@pytest.mark.parametrize("values", [[["a"], ["x"]], [[], []]])
+def test_reindex_empty_with_level(values):
+    # GH41170
+    ser = Series(
+        range(len(values[0])), index=MultiIndex.from_arrays(values), dtype="object"
+    )
+    result = ser.reindex(np.array(["b"]), level=0)
+    expected = Series(
+        index=MultiIndex(levels=[["b"], values[1]], codes=[[], []]), dtype="object"
+    )
+    tm.assert_series_equal(result, expected)