BUG: join MultiIndex with overlapping IntervalIndex level (pandas-dev#44096)

`BaseMultiIndexCodesEngine._extract_level_codes` relies on the assumption that
`get_indexer()` is always usable as any index level `is_unique`.  However, that
is not the case for overlapping `IntervalIndex` levels.  `get_indexer()` can
not be used on an `IntervalIndex` that `is_overlapping` even if it `is_unique`.

This patch uses `get_indexer_for()` instead.

Author: johannes-mueller

doc/source/whatsnew/v1.4.0.rst

@@ -702,6 +702,7 @@ Reshaping
 - Bug in :meth:`DataFrame.append` failing to retain ``index.name`` when appending a list of :class:`Series` objects (:issue:`44109`)
 - Fixed metadata propagation in :meth:`Dataframe.apply` method, consequently fixing the same issue for :meth:`Dataframe.transform`, :meth:`Dataframe.nunique` and :meth:`Dataframe.mode` (:issue:`28283`)
 - Bug in :meth:`DataFrame.stack` with ``ExtensionDtype`` columns incorrectly raising (:issue:`43561`)
+- Bug in :class:`MultiIndex` failing join operations with overlapping ``IntervalIndex`` levels (:issue:`44096`)
 -
 
 Sparse

pandas/_libs/index.pyx

@@ -649,7 +649,7 @@ cdef class BaseMultiIndexCodesEngine:
             Integers representing one combination each
         """
         zt = [target._get_level_values(i) for i in range(target.nlevels)]
-        level_codes = [lev.get_indexer(codes) + 1 for lev, codes
+        level_codes = [lev.get_indexer_for(codes) + 1 for lev, codes
                        in zip(self.levels, zt)]
         return self._codes_to_ints(np.array(level_codes, dtype='uint64').T)
 

pandas/tests/indexes/multi/test_join.py

@@ -3,6 +3,7 @@
 
 from pandas import (
     Index,
+    Interval,
     MultiIndex,
 )
 import pandas._testing as tm
@@ -115,3 +116,45 @@ def test_join_multi_return_indexers():
 
     result = midx1.join(midx2, return_indexers=False)
     tm.assert_index_equal(result, midx1)
+
+
+def test_join_overlapping_interval_level():
+    # GH 44096
+    idx_1 = MultiIndex.from_tuples(
+        [
+            (1, Interval(0.0, 1.0)),
+            (1, Interval(1.0, 2.0)),
+            (1, Interval(2.0, 5.0)),
+            (2, Interval(0.0, 1.0)),
+            (2, Interval(1.0, 3.0)),  # interval limit is here at 3.0, not at 2.0
+            (2, Interval(3.0, 5.0)),
+        ],
+        names=["num", "interval"],
+    )
+
+    idx_2 = MultiIndex.from_tuples(
+        [
+            (1, Interval(2.0, 5.0)),
+            (1, Interval(0.0, 1.0)),
+            (1, Interval(1.0, 2.0)),
+            (2, Interval(3.0, 5.0)),
+            (2, Interval(0.0, 1.0)),
+            (2, Interval(1.0, 3.0)),
+        ],
+        names=["num", "interval"],
+    )
+
+    expected = MultiIndex.from_tuples(
+        [
+            (1, Interval(0.0, 1.0)),
+            (1, Interval(1.0, 2.0)),
+            (1, Interval(2.0, 5.0)),
+            (2, Interval(0.0, 1.0)),
+            (2, Interval(1.0, 3.0)),
+            (2, Interval(3.0, 5.0)),
+        ],
+        names=["num", "interval"],
+    )
+    result = idx_1.join(idx_2, how="outer")
+
+    tm.assert_index_equal(result, expected)