BUG: validate Index data is 1D + deprecate multi-dim indexing

doc/source/whatsnew/v1.0.0.rst

@@ -581,6 +581,7 @@ Deprecations
 - The ``pandas.util.testing`` module has been deprecated. Use the public API in ``pandas.testing`` documented at :ref:`api.general.testing` (:issue:`16232`).
 - ``pandas.SparseArray`` has been deprecated.  Use ``pandas.arrays.SparseArray`` (:class:`arrays.SparseArray`) instead. (:issue:`30642`)
 - The parameter ``is_copy`` of :meth:`DataFrame.take` has been deprecated and will be removed in a future version. (:issue:`27357`)
+- Support for indexing indexing an :class:`Index` with ``index[:, None]`` is deprecated and will be removed in a future version (:issue:`30588`)
 
 **Selecting Columns from a Grouped DataFrame**
 
@@ -1034,6 +1035,7 @@ Other
 - Bug in :class:`DataFrame` constructor when passing a 2D ``ndarray`` and an extension dtype (:issue:`12513`)
 - Bug in :meth:`DaataFrame.to_csv` when supplied a series with a ``dtype="string"`` and a ``na_rep``, the ``na_rep`` was being truncated to 2 characters. (:issue:`29975`)
 - Bug where :meth:`DataFrame.itertuples` would incorrectly determine whether or not namedtuples could be used for dataframes of 255 columns (:issue:`28282`)
+- Bug in :class:`Index` constructor incorrectly allowing 2-dimensional input arrays (:issue:`13601`, :issue:`27125`)
 
 .. _whatsnew_1000.contributors:
 

pandas/core/indexes/base.py

@@ -393,6 +393,9 @@ def __new__(
 
             if kwargs:
                 raise TypeError(f"Unexpected keyword arguments {repr(set(kwargs))}")
+            if subarr.ndim > 1:
+                # GH#13601, GH#20285, GH#27125
+                raise ValueError("Index data must be 1-dimensional")
             return cls._simple_new(subarr, name, **kwargs)
 
         elif hasattr(data, "__array__"):
@@ -608,7 +611,7 @@ def __array_wrap__(self, result, context=None):
         Gets called after a ufunc.
         """
         result = lib.item_from_zerodim(result)
-        if is_bool_dtype(result) or lib.is_scalar(result):
+        if is_bool_dtype(result) or lib.is_scalar(result) or np.ndim(result) > 1:
             return result
 
         attrs = self._get_attributes_dict()
@@ -687,11 +690,10 @@ def astype(self, dtype, copy=True):
             return Index(np.asarray(self), dtype=dtype, copy=copy)
 
         try:
-            return Index(
-                self.values.astype(dtype, copy=copy), name=self.name, dtype=dtype
-            )
+            casted = self.values.astype(dtype, copy=copy)
         except (TypeError, ValueError):
             raise TypeError(f"Cannot cast {type(self).__name__} to dtype {dtype}")
+        return Index(casted, name=self.name, dtype=dtype)
 
     _index_shared_docs[
         "take"
@@ -3931,6 +3933,17 @@ def __getitem__(self, key):
         key = com.values_from_object(key)
         result = getitem(key)
         if not is_scalar(result):
+            if np.ndim(result) > 1:
+                # GH#27125 indexer like idx[:, None] expands dim, but we
+                #  cannot do that and keep an index, so return ndarray
+                # Deprecation GH#30588
+                warnings.warn(
+                    "Support for Index[:, None] is deprecated and will be "
+                    "removed in a future version.",
+                    DeprecationWarning,
+                    stacklevel=2,
+                )
+                return result
             return promote(result)
         else:
             return result

pandas/core/indexes/numeric.py

@@ -73,6 +73,10 @@ def __new__(cls, data=None, dtype=None, copy=False, name=None):
         else:
             subarr = data
 
+        if subarr.ndim > 1:
+            # GH#13601, GH#20285, GH#27125
+            raise ValueError("Index data must be 1-dimensional")
+
         name = maybe_extract_name(name, data, cls)
         return cls._simple_new(subarr, name=name)
 

pandas/tests/indexes/test_base.py

@@ -1905,6 +1905,14 @@ def test_set_value_deprecated(self):
             idx.set_value(arr, idx[1], 80)
         assert arr[1] == 80
 
+    def test_getitem_2d_deprecated(self):
+        # GH#30588
+        idx = self.create_index()
+        with tm.assert_produces_warning(DeprecationWarning):
+            res = idx[:, None]
+
+        assert isinstance(res, np.ndarray)
+
     @pytest.mark.parametrize(
         "index", ["string", "int", "datetime", "timedelta"], indirect=True
     )
@@ -2784,9 +2792,34 @@ def test_shape_of_invalid_index():
     # about this). However, as long as this is not solved in general,this test ensures
     # that the returned shape is consistent with this underlying array for
     # compat with matplotlib (see https://github.com/pandas-dev/pandas/issues/27775)
-    a = np.arange(8).reshape(2, 2, 2)
-    idx = pd.Index(a)
-    assert idx.shape == a.shape
-
     idx = pd.Index([0, 1, 2, 3])
-    assert idx[:, None].shape == (4, 1)
+    with tm.assert_produces_warning(DeprecationWarning):
+        assert idx[:, None].shape == (4, 1)
+
+
+def test_validate_1d_input():
+    # GH#27125 check that we do not have >1-dimensional input
+    msg = "Index data must be 1-dimensional"
+
+    arr = np.arange(8).reshape(2, 2, 2)
+    with pytest.raises(ValueError, match=msg):
+        pd.Index(arr)
+
+    with pytest.raises(ValueError, match=msg):
+        pd.Float64Index(arr.astype(np.float64))
+
+    with pytest.raises(ValueError, match=msg):
+        pd.Int64Index(arr.astype(np.int64))
+
+    with pytest.raises(ValueError, match=msg):
+        pd.UInt64Index(arr.astype(np.uint64))
+
+    df = pd.DataFrame(arr.reshape(4, 2))
+    with pytest.raises(ValueError, match=msg):
+        pd.Index(df)
+
+    # GH#13601 trying to assign a multi-dimensional array to an index is not
+    #  allowed
+    ser = pd.Series(0, range(4))
+    with pytest.raises(ValueError, match=msg):
+        ser.index = np.array([[2, 3]] * 4)

pandas/tests/indexing/test_indexing.py

@@ -83,12 +83,9 @@ def test_getitem_ndarray_3d(self, index, obj, idxr, idxr_id):
         msg = (
             r"Buffer has wrong number of dimensions \(expected 1,"
             r" got 3\)|"
-            "The truth value of an array with more than one element is"
-            " ambiguous|"
             "Cannot index with multidimensional key|"
             r"Wrong number of dimensions. values.ndim != ndim \[3 != 1\]|"
-            "No matching signature found|"  # TypeError
-            "unhashable type: 'numpy.ndarray'"  # TypeError
+            "Index data must be 1-dimensional"
         )
 
         if (
@@ -106,18 +103,7 @@ def test_getitem_ndarray_3d(self, index, obj, idxr, idxr_id):
         ):
             idxr[nd3]
         else:
-            if (
-                isinstance(obj, DataFrame)
-                and idxr_id == "getitem"
-                and index.inferred_type == "boolean"
-            ):
-                error = TypeError
-            elif idxr_id == "getitem" and index.inferred_type == "interval":
-                error = TypeError
-            else:
-                error = ValueError
-
-            with pytest.raises(error, match=msg):
+            with pytest.raises(ValueError, match=msg):
                 idxr[nd3]
 
     @pytest.mark.parametrize(
@@ -148,14 +134,12 @@ def test_setitem_ndarray_3d(self, index, obj, idxr, idxr_id):
         msg = (
             r"Buffer has wrong number of dimensions \(expected 1,"
             r" got 3\)|"
-            "The truth value of an array with more than one element is"
-            " ambiguous|"
-            "Only 1-dimensional input arrays are supported|"
             "'pandas._libs.interval.IntervalTree' object has no attribute"
             " 'set_value'|"  # AttributeError
             "unhashable type: 'numpy.ndarray'|"  # TypeError
             "No matching signature found|"  # TypeError
-            r"^\[\[\["  # pandas.core.indexing.IndexingError
+            r"^\[\[\[|"  # pandas.core.indexing.IndexingError
+            "Index data must be 1-dimensional"
         )
 
         if (idxr_id == "iloc") or (
@@ -176,10 +160,8 @@ def test_setitem_ndarray_3d(self, index, obj, idxr, idxr_id):
         ):
             idxr[nd3] = 0
         else:
-            with pytest.raises(
-                (ValueError, AttributeError, TypeError, pd.core.indexing.IndexingError),
-                match=msg,
-            ):
+            err = (ValueError, AttributeError)
+            with pytest.raises(err, match=msg):
                 idxr[nd3] = 0
 
     def test_inf_upcast(self):

pandas/tests/io/test_feather.py

@@ -136,7 +136,7 @@ def test_write_with_index(self):
 
         # column multi-index
         df.index = [0, 1, 2]
-        df.columns = (pd.MultiIndex.from_tuples([("a", 1), ("a", 2), ("b", 1)]),)
+        df.columns = pd.MultiIndex.from_tuples([("a", 1)])
         self.check_error_on_write(df, ValueError)
 
     def test_path_pathlib(self):