BUG: DataFrame.__getitem__(number) with IntervalIndex columns pandas-dev#26490

Author: jbrockmendel

doc/source/whatsnew/v1.2.0.rst

@@ -400,6 +400,7 @@ Indexing
 - Bug in :meth:`DataFrame.sort_index` where parameter ascending passed as a list on a single level index gives wrong result. (:issue:`32334`)
 - Bug in :meth:`DataFrame.reset_index` was incorrectly raising a ``ValueError`` for input with a :class:`MultiIndex` with missing values in a level with ``Categorical`` dtype (:issue:`24206`)
 - Bug in indexing with boolean masks on datetime-like values sometimes returning a view instead of a copy (:issue:`36210`)
+- Bug in :meth:`DataFrame.__getitem__` and :meth:`DataFrame.loc.__getitem__` with :class:`IntervalIndex` columns and a numeric indexer (:issue:`26490`)
 
 Missing
 ^^^^^^^

pandas/core/frame.py

@@ -2946,7 +2946,7 @@ def __getitem__(self, key):
             # - the key itself is repeated (test on data.shape, #9519), or
             # - we have a MultiIndex on columns (test on self.columns, #21309)
             if data.shape[1] == 1 and not isinstance(self.columns, MultiIndex):
-                data = data[key]
+                data = data._ixs(0, axis=1)
 
         return data
 

pandas/tests/frame/indexing/test_indexing.py

@@ -2160,6 +2160,20 @@ def test_interval_index(self):
         result = df.loc[1, "A"]
         tm.assert_series_equal(result, expected)
 
+    def test_getitem_interval_index_partial_indexing(self):
+        # GH#36490
+        df = pd.DataFrame(
+            np.ones((3, 4)), columns=pd.IntervalIndex.from_breaks(np.arange(5))
+        )
+
+        expected = df.iloc[:, 0]
+
+        res = df[0.5]
+        tm.assert_series_equal(res, expected)
+
+        res = df.loc[:, 0.5]
+        tm.assert_series_equal(res, expected)
+
 
 class TestDataFrameIndexingUInt64:
     def test_setitem(self, uint64_frame):