Improve MultiIndex label rename checks, docs and tests

Author: TabLand

doc/source/whatsnew/v2.3.0.rst

@@ -213,6 +213,7 @@ Styler
 Other
 ^^^^^
 - Bug in :meth:`DataFrame.where` where using a non-bool type array in the function would return a ``ValueError`` instead of a ``TypeError`` (:issue:`56330`)
+- Fixed bug in :meth:`DataFrame.rename` where checks on argument errors="raise" are not consistent with the actual transformation applied (:issue:`55169`). Logic change is accompanied with improvement to docs, a new test and a more descriptive ``KeyError`` message when a tuple label rename is attempted across :class:`MultiIndex` levels
 
 
 .. ***DO NOT USE THIS SECTION***

pandas/core/frame.py

@@ -5650,6 +5650,11 @@ def rename(
         level : int or level name, default None
             In case of a MultiIndex, only rename labels in the specified
             level.
+
+            .. note::
+               Labels are renamed individually, and not via tuples across
+               MultiIndex levels
+
         errors : {'ignore', 'raise'}, default 'ignore'
             If 'raise', raise a `KeyError` when a dict-like `mapper`, `index`,
             or `columns` contains labels that are not present in the Index

pandas/core/generic.py

@@ -1118,18 +1118,36 @@ def _rename(
 
             # GH 13473
             if not callable(replacements):
-                if ax._is_multi and level is not None:
-                    indexer = ax.get_level_values(level).get_indexer_for(replacements)
-                else:
-                    indexer = ax.get_indexer_for(replacements)
-
-                if errors == "raise" and len(indexer[indexer == -1]):
-                    missing_labels = [
-                        label
-                        for index, label in enumerate(replacements)
-                        if indexer[index] == -1
-                    ]
-                    raise KeyError(f"{missing_labels} not found in axis")
+                if errors == "raise":
+                    missing_labels = []
+                    for replacement in replacements:
+                        if ax._is_multi:
+                            indexers = [
+                                ax.get_level_values(i).get_indexer_for([replacement])
+                                for i in range(ax.nlevels)
+                                if i == level or level is None
+                            ]
+                        else:
+                            indexers = [ax.get_indexer_for([replacement])]
+
+                        found_anywhere = any(any(indexer != -1) for indexer in indexers)
+                        if not found_anywhere:
+                            missing_labels.append(replacement)
+
+                    if len(missing_labels) > 0:
+                        error = f"{missing_labels} not found in axis"
+                        if ax._is_multi:
+                            tuple_rename_tried = any(
+                                type(label) is tuple and label in ax
+                                for label in missing_labels
+                            )
+                            if tuple_rename_tried:
+                                error += (
+                                    ". Please provide individual labels for "
+                                    "replacement, and not tuples across "
+                                    "MultiIndex levels"
+                                )
+                        raise KeyError(error)
 
             new_index = ax._transform_index(f, level=level)
             result._set_axis_nocheck(new_index, axis=axis_no, inplace=True, copy=False)

pandas/tests/frame/methods/test_rename.py

@@ -164,6 +164,16 @@ def test_rename_multiindex(self):
         renamed = df.rename(index={"foo1": "foo3", "bar2": "bar3"}, level=0)
         tm.assert_index_equal(renamed.index, new_index)
 
+    def test_rename_multiindex_with_checks(self):
+        df = DataFrame({("a", "count"): [1, 2], ("a", "sum"): [3, 4]})
+        renamed = df.rename(
+            columns={"a": "b", "count": "number_of", "sum": "total"}, errors="raise"
+        )
+
+        new_columns = MultiIndex.from_tuples([("b", "number_of"), ("b", "total")])
+
+        tm.assert_index_equal(renamed.columns, new_columns)
+
     def test_rename_nocopy(self, float_frame, using_copy_on_write, warn_copy_on_write):
         renamed = float_frame.rename(columns={"C": "foo"}, copy=False)
 
@@ -225,6 +235,20 @@ def test_rename_errors_raises(self):
         with pytest.raises(KeyError, match="'E'] not found in axis"):
             df.rename(columns={"A": "a", "E": "e"}, errors="raise")
 
+    def test_rename_error_raised_for_label_across_multiindex_levels(self):
+        df = DataFrame([{"a": 1, "b": 2}, {"a": 3, "b": 4}])
+        df = df.groupby("a").agg({"b": ("count", "sum")})
+        with pytest.raises(
+            KeyError,
+            match=(
+                "\\[\\('b', 'count'\\)\\] not found "
+                "in axis\\. Please provide individual "
+                "labels for replacement, and not "
+                "tuples across MultiIndex levels"
+            ),
+        ):
+            df.rename(columns={("b", "count"): "new"}, errors="raise")
+
     @pytest.mark.parametrize(
         "mapper, errors, expected_columns",
         [