test backportability of pandas-dev#36927

doc/source/whatsnew/v1.1.5.rst

@@ -23,6 +23,7 @@ Fixed regressions
 - Fixed regression in :meth:`DataFrame.groupby` aggregation with out-of-bounds datetime objects in an object-dtype column (:issue:`36003`)
 - Fixed regression in ``df.groupby(..).rolling(..)`` with the resulting :class:`MultiIndex` when grouping by a label that is in the index (:issue:`37641`)
 - Fixed regression in :meth:`DataFrame.fillna` not filling ``NaN`` after other operations such as :meth:`DataFrame.pivot` (:issue:`36495`).
+- Fixed regression in :meth:`MultiIndex.intersection` returning duplicates when at least one of the indexes had duplicates (:issue:`36915`)
 
 .. ---------------------------------------------------------------------------
 

pandas/core/indexes/base.py

@@ -2654,7 +2654,7 @@ def intersection(self, other, sort=False):
         self._assert_can_do_setop(other)
         other = ensure_index(other)
 
-        if self.equals(other):
+        if self.equals(other) and not self.has_duplicates:
             return self._get_reconciled_name_object(other)
 
         if not is_dtype_equal(self.dtype, other.dtype):
@@ -2672,7 +2672,7 @@ def intersection(self, other, sort=False):
             except TypeError:
                 pass
             else:
-                return self._wrap_setop_result(other, result)
+                return self._wrap_setop_result(other, algos.unique1d(result))
 
         try:
             indexer = Index(rvals).get_indexer(lvals)
@@ -2683,7 +2683,7 @@ def intersection(self, other, sort=False):
             indexer = algos.unique1d(Index(rvals).get_indexer_non_unique(lvals)[0])
             indexer = indexer[indexer != -1]
 
-        taken = other.take(indexer)
+        taken = other.take(indexer).unique()
         res_name = get_op_result_name(self, other)
 
         if sort is None:

pandas/core/indexes/multi.py

@@ -3398,6 +3398,8 @@ def intersection(self, other, sort=False):
         other, result_names = self._convert_can_do_setop(other)
 
         if self.equals(other):
+            if self.has_duplicates:
+                return self.unique()
             return self
 
         if not is_object_dtype(other.dtype):
@@ -3416,10 +3418,12 @@ def intersection(self, other, sort=False):
         uniq_tuples = None  # flag whether _inner_indexer was successful
         if self.is_monotonic and other.is_monotonic:
             try:
-                uniq_tuples = self._inner_indexer(lvals, rvals)[0]
-                sort = False  # uniq_tuples is already sorted
+                inner_tuples = self._inner_indexer(lvals, rvals)[0]
+                sort = False  # inner_tuples is already sorted
             except TypeError:
                 pass
+            else:
+                uniq_tuples = algos.unique(inner_tuples)
 
         if uniq_tuples is None:
             other_uniq = set(rvals)

pandas/core/ops/__init__.py

@@ -539,7 +539,9 @@ def _should_reindex_frame_op(
     if fill_value is None and level is None and axis is default_axis:
         # TODO: any other cases we should handle here?
         cols = left.columns.intersection(right.columns)
-        if not (cols.equals(left.columns) and cols.equals(right.columns)):
+        if not (
+            cols.equals(left.columns.unique()) and cols.equals(right.columns.unique())
+        ):
             return True
 
     return False

pandas/core/reshape/merge.py

@@ -1209,7 +1209,9 @@ def _validate_specification(self):
                     raise MergeError("Must pass left_on or left_index=True")
             else:
                 # use the common columns
-                common_cols = self.left.columns.intersection(self.right.columns)
+                left_cols = self.left.columns
+                right_cols = self.right.columns
+                common_cols = left_cols.intersection(right_cols)
                 if len(common_cols) == 0:
                     raise MergeError(
                         "No common columns to perform merge on. "
@@ -1218,7 +1220,10 @@ def _validate_specification(self):
                         f"left_index={self.left_index}, "
                         f"right_index={self.right_index}"
                     )
-                if not common_cols.is_unique:
+                if (
+                    not left_cols.join(common_cols, how="inner").is_unique
+                    or not right_cols.join(common_cols, how="inner").is_unique
+                ):
                     raise MergeError(f"Data columns not unique: {repr(common_cols)}")
                 self.left_on = self.right_on = common_cols
         elif self.on is not None:

pandas/tests/indexes/multi/test_setops.py

@@ -375,3 +375,26 @@ def test_setops_disallow_true(method):
 
     with pytest.raises(ValueError, match="The 'sort' keyword only takes"):
         getattr(idx1, method)(idx2, sort=True)
+
+
+@pytest.mark.parametrize(
+    ("tuples", "exp_tuples"),
+    [
+        ([("val1", "test1")], [("val1", "test1")]),
+        ([("val1", "test1"), ("val1", "test1")], [("val1", "test1")]),
+        (
+            [("val2", "test2"), ("val1", "test1")],
+            [("val2", "test2"), ("val1", "test1")],
+        ),
+    ],
+)
+def test_intersect_with_duplicates(tuples, exp_tuples):
+    # GH#36915
+    left = MultiIndex.from_tuples(tuples, names=["first", "second"])
+    right = MultiIndex.from_tuples(
+        [("val1", "test1"), ("val1", "test1"), ("val2", "test2")],
+        names=["first", "second"],
+    )
+    result = left.intersection(right)
+    expected = MultiIndex.from_tuples(exp_tuples, names=["first", "second"])
+    tm.assert_index_equal(result, expected)

pandas/tests/indexes/test_setops.py

@@ -95,3 +95,13 @@ def test_union_dtypes(left, right, expected):
     b = pd.Index([], dtype=right)
     result = (a | b).dtype
     assert result == expected
+
+
+@pytest.mark.parametrize("values", [[1, 2, 2, 3], [3, 3]])
+def test_intersection_duplicates(values):
+    # GH#31326
+    a = pd.Index(values)
+    b = pd.Index([3, 3])
+    result = a.intersection(b)
+    expected = pd.Index([3])
+    tm.assert_index_equal(result, expected)

pandas/tests/reshape/merge/test_merge.py

@@ -742,7 +742,7 @@ def test_overlapping_columns_error_message(self):
 
         # #2649, #10639
         df2.columns = ["key1", "foo", "foo"]
-        msg = r"Data columns not unique: Index\(\['foo', 'foo'\], dtype='object'\)"
+        msg = r"Data columns not unique: Index\(\['foo'\], dtype='object'\)"
         with pytest.raises(MergeError, match=msg):
             merge(df, df2)