BUG: pd.MultiIndex.get_loc(np.nan) (#19132)

MultiIndex.get_loc can find missing value in MultiIndex with a null
value as input

Author: proost

doc/source/whatsnew/v1.0.0.rst

@@ -229,6 +229,7 @@ Indexing
 - Bug in reindexing a :meth:`PeriodIndex` with another type of index that contained a `Period` (:issue:`28323`) (:issue:`28337`)
 - Fix assignment of column via `.loc` with numpy non-ns datetime type (:issue:`27395`)
 - Bug in :meth:`Float64Index.astype` where ``np.inf`` was not handled properly when casting to an integer dtype (:issue:`28475`)
+- When index is ``MultiIndex`, Using ``.get_loc`` can't find ``nan`` with a null value as input (:issue:`19132`)
 
 Missing
 ^^^^^^^

pandas/core/indexes/multi.py

@@ -2890,6 +2890,24 @@ def convert_indexer(start, stop, step, indexer=indexer, codes=level_codes):
 
             return m
 
+        def search_code_location(code):
+            # Base on argument ``code``, search where is ``code`` in level's codes
+
+            if level > 0 or self.lexsort_depth == 0:
+                # Desired level is not sorted
+                locs = np.array(level_codes == code, dtype=bool, copy=False)
+                if not locs.any():
+                    # The label is present in self.levels[level] but unused:
+                    raise KeyError(key)
+                return locs
+
+            i = level_codes.searchsorted(code, side="left")
+            j = level_codes.searchsorted(code, side="right")
+            if i == j:
+                # The label is present in self.levels[level] but unused:
+                raise KeyError(key)
+            return slice(i, j)
+
         if isinstance(key, slice):
             # handle a slice, returning a slice if we can
             # otherwise a boolean indexer
@@ -2933,24 +2951,14 @@ def convert_indexer(start, stop, step, indexer=indexer, codes=level_codes):
                 j = level_codes.searchsorted(stop, side="right")
                 return slice(i, j, step)
 
+        elif not is_list_like(key) and isna(key):
+            # missing data's location is denoted by -1
+            # so find missing data's location
+            code = -1
+            return search_code_location(code)
         else:
-
             code = level_index.get_loc(key)
-
-            if level > 0 or self.lexsort_depth == 0:
-                # Desired level is not sorted
-                locs = np.array(level_codes == code, dtype=bool, copy=False)
-                if not locs.any():
-                    # The label is present in self.levels[level] but unused:
-                    raise KeyError(key)
-                return locs
-
-            i = level_codes.searchsorted(code, side="left")
-            j = level_codes.searchsorted(code, side="right")
-            if i == j:
-                # The label is present in self.levels[level] but unused:
-                raise KeyError(key)
-            return slice(i, j)
+            return search_code_location(code)
 
     def get_locs(self, seq):
         """

pandas/tests/indexes/multi/test_indexing.py

@@ -439,3 +439,24 @@ def test_timestamp_multiindex_indexer():
     )
     should_be = pd.Series(data=np.arange(24, len(qidx) + 24), index=qidx, name="foo")
     tm.assert_series_equal(result, should_be)
+
+
+def test_get_loc_with_a_missing_value():
+    # issue 19132
+    idx = MultiIndex.from_product([[np.nan, 1]] * 2)
+    expected = slice(0, 2, None)
+    assert idx.get_loc(np.nan) == expected
+
+    idx = MultiIndex.from_arrays([[np.nan, 1, 2, np.nan], [3, np.nan, np.nan, 4]])
+    expected = np.array([True, False, False, True])
+    tm.assert_numpy_array_equal(idx.get_loc(np.nan), expected)
+
+
+def test_get_indexer_with_nan():
+    # issue 19132
+    idx = MultiIndex.from_arrays([[1, np.nan, 2], [3, 4, 5]])
+    expected = np.array([-1, -1, -1], dtype="int64")
+    tm.assert_numpy_array_equal(idx.get_indexer([1, np.nan, 2]), expected)
+
+    expected = np.array([1], dtype="int64")
+    tm.assert_numpy_array_equal(idx.get_indexer([(np.nan, 4)]), expected)