BUG: Join behaved like an inner join when only one side had MultiIndex

doc/source/whatsnew/v1.3.0.rst

@@ -295,6 +295,7 @@ Reshaping
 ^^^^^^^^^
 - Bug in :meth:`DataFrame.unstack` with missing levels led to incorrect index names (:issue:`37510`)
 - Bug in :func:`concat` incorrectly casting to ``object`` dtype in some cases when one or more of the operands is empty (:issue:`38843`)
+- Bug in :func:`join` behaved like an inner join, when one side had a regular :class:`Index` while the other side had a :class:`MultiIndex` (:issue:`34292`)
 -
 
 

pandas/core/indexes/base.py

@@ -3863,66 +3863,44 @@ def _join_multi(self, other, how, return_indexers=True):
         if not overlap:
             raise ValueError("cannot join with no overlapping index names")
 
-        if isinstance(self, MultiIndex) and isinstance(other, MultiIndex):
-
-            # Drop the non-matching levels from left and right respectively
-            ldrop_names = sorted(self_names - overlap, key=self_names_order)
-            rdrop_names = sorted(other_names - overlap, key=other_names_order)
-
-            # if only the order differs
-            if not len(ldrop_names + rdrop_names):
-                self_jnlevels = self
-                other_jnlevels = other.reorder_levels(self.names)
-            else:
-                self_jnlevels = self.droplevel(ldrop_names)
-                other_jnlevels = other.droplevel(rdrop_names)
-
-            # Join left and right
-            # Join on same leveled multi-index frames is supported
-            join_idx, lidx, ridx = self_jnlevels.join(
-                other_jnlevels, how, return_indexers=True
-            )
-
-            # Restore the dropped levels
-            # Returned index level order is
-            # common levels, ldrop_names, rdrop_names
-            dropped_names = ldrop_names + rdrop_names
-
-            levels, codes, names = restore_dropped_levels_multijoin(
-                self, other, dropped_names, join_idx, lidx, ridx
-            )
-
-            # Re-create the multi-index
-            multi_join_idx = MultiIndex(
-                levels=levels, codes=codes, names=names, verify_integrity=False
-            )
-
-            multi_join_idx = multi_join_idx.remove_unused_levels()
+        # Drop the non-matching levels from left and right respectively
+        ldrop_names = sorted(self_names - overlap, key=self_names_order)
+        rdrop_names = sorted(other_names - overlap, key=other_names_order)
+
+        # if only the order differs
+        if not len(ldrop_names + rdrop_names):
+            self_jnlevels = self
+            other_jnlevels = other.reorder_levels(self.names)
+        else:
+            self_jnlevels = self.droplevel(ldrop_names)
+            other_jnlevels = other.droplevel(rdrop_names)
 
-            if return_indexers:
-                return multi_join_idx, lidx, ridx
-            else:
-                return multi_join_idx
+        # Join left and right
+        # Join on same leveled multi-index frames is supported
+        join_idx, lidx, ridx = self_jnlevels.join(
+            other_jnlevels, how, return_indexers=True
+        )
 
-        jl = list(overlap)[0]
+        # Restore the dropped levels
+        # Returned index level order is
+        # common levels, ldrop_names, rdrop_names
+        dropped_names = ldrop_names + rdrop_names
 
-        # Case where only one index is multi
-        # make the indices into mi's that match
-        flip_order = False
-        if isinstance(self, MultiIndex):
-            self, other = other, self
-            flip_order = True
-            # flip if join method is right or left
-            how = {"right": "left", "left": "right"}.get(how, how)
+        levels, codes, names = restore_dropped_levels_multijoin(
+            self, other, dropped_names, join_idx, lidx, ridx
+        )
 
-        level = other.names.index(jl)
-        result = self._join_level(
-            other, level, how=how, return_indexers=return_indexers
+        # Re-create the multi-index
+        multi_join_idx = MultiIndex(
+            levels=levels, codes=codes, names=names, verify_integrity=False
         )
 
-        if flip_order and isinstance(result, tuple):
-            return result[0], result[2], result[1]
-        return result
+        multi_join_idx = multi_join_idx.remove_unused_levels()
+
+        if return_indexers:
+            return multi_join_idx, lidx, ridx
+        else:
+            return multi_join_idx
 
     @final
     def _join_non_unique(self, other, how="left", return_indexers=False):

pandas/tests/indexes/multi/test_join.py

@@ -56,9 +56,9 @@ def test_join_multi():
 
     # inner
     jidx, lidx, ridx = midx.join(idx, how="inner", return_indexers=True)
-    exp_idx = pd.MultiIndex.from_product([np.arange(4), [1, 2]], names=["a", "b"])
-    exp_lidx = np.array([1, 2, 5, 6, 9, 10, 13, 14], dtype=np.intp)
-    exp_ridx = np.array([0, 1, 0, 1, 0, 1, 0, 1], dtype=np.intp)
+    exp_idx = pd.MultiIndex.from_product([[1, 2], np.arange(4)], names=["b", "a"])
+    exp_lidx = np.array([1, 5, 9, 13, 2, 6, 10, 14], dtype=np.intp)
+    exp_ridx = np.array([0, 0, 0, 0, 1, 1, 1, 1], dtype=np.intp)
     tm.assert_index_equal(jidx, exp_idx)
     tm.assert_numpy_array_equal(lidx, exp_lidx)
     tm.assert_numpy_array_equal(ridx, exp_ridx)
@@ -71,15 +71,17 @@ def test_join_multi():
     # keep MultiIndex
     jidx, lidx, ridx = midx.join(idx, how="left", return_indexers=True)
     exp_ridx = np.array(
-        [-1, 0, 1, -1, -1, 0, 1, -1, -1, 0, 1, -1, -1, 0, 1, -1], dtype=np.intp
+        [-1, -1, -1, -1, 0, 0, 0, 0, 1, 1, 1, 1, -1, -1, -1, -1], dtype=np.intp
     )
-    tm.assert_index_equal(jidx, midx)
-    assert lidx is None
+    exp_idx = midx.copy()
+    exp_idx.names = ["b", "a"]
+    tm.assert_index_equal(jidx, exp_idx)
+    # assert lidx is None
     tm.assert_numpy_array_equal(ridx, exp_ridx)
     # flip
     jidx, ridx, lidx = idx.join(midx, how="right", return_indexers=True)
-    tm.assert_index_equal(jidx, midx)
-    assert lidx is None
+    tm.assert_index_equal(jidx, exp_idx)
+    # assert lidx is None
     tm.assert_numpy_array_equal(ridx, exp_ridx)
 
 
@@ -113,3 +115,14 @@ def test_join_multi_return_indexers():
 
     result = midx1.join(midx2, return_indexers=False)
     tm.assert_index_equal(result, midx1)
+
+
+def test_join_multi_and_index():
+    # GH#34292
+    idx = Index([1, 2], name="a")
+    midx = pd.MultiIndex.from_tuples([(1, 4), (3, 0), (1, 5)], names=["a", "b"])
+    result = idx.join(midx, how="left")
+    expected = pd.MultiIndex.from_tuples(
+        [(1, 4), (1, 5), (2, np.nan)], names=["a", "b"]
+    )
+    tm.assert_index_equal(result, expected)