Fix no raise dup index when using drop with axis=0

doc/source/whatsnew/v0.23.0.txt

@@ -408,6 +408,8 @@ Indexing
 - Bug in :func:`MultiIndex.set_labels` which would cause casting (and potentially clipping) of the new labels if the ``level`` argument is not 0 or a list like [0, 1, ... ]  (:issue:`19057`)
 - Bug in ``str.extractall`` when there were no matches empty :class:`Index` was returned instead of appropriate :class:`MultiIndex` (:issue:`19034`)
 - Bug in :class:`IntervalIndex` where set operations that returned an empty ``IntervalIndex`` had the wrong dtype (:issue:`19101`)
+- Bug in :func:`DataFrame.drop`, ``KeyError`` now raises when dropping an ``Index`` or column that has duplicates (:issue:`19186`)
+-
 
 I/O
 ^^^

pandas/core/generic.py

@@ -2909,6 +2909,9 @@ def _drop_axis(self, labels, axis, level=None, errors='raise'):
             else:
                 indexer = ~axis.isin(labels)
 
+            if errors == 'raise' and indexer.all():
+                raise KeyError('{} not found in axis'.format(labels))
+
             slicer = [slice(None)] * self.ndim
             slicer[self._get_axis_number(axis_name)] = indexer
 

pandas/core/indexes/base.py

@@ -3767,8 +3767,8 @@ def drop(self, labels, errors='raise'):
         mask = indexer == -1
         if mask.any():
             if errors != 'ignore':
-                raise ValueError('labels %s not contained in axis' %
-                                 labels[mask])
+                raise KeyError(
+                    'labels %s not contained in axis' % labels[mask])
             indexer = indexer[~mask]
         return self.delete(indexer)
 

pandas/core/reshape/pivot.py

@@ -75,7 +75,7 @@ def pivot_table(data, values=None, index=None, columns=None, aggfunc='mean',
         for key in keys:
             try:
                 values = values.drop(key)
-            except (TypeError, ValueError):
+            except (TypeError, ValueError, KeyError):
                 pass
         values = list(values)
 

pandas/tests/frame/test_axis_select_reindex.py

@@ -41,8 +41,8 @@ def test_drop_names(self):
             assert obj.columns.name == 'second'
         assert list(df.columns) == ['d', 'e', 'f']
 
-        pytest.raises(ValueError, df.drop, ['g'])
-        pytest.raises(ValueError, df.drop, ['g'], 1)
+        pytest.raises(KeyError, df.drop, ['g'])
+        pytest.raises(KeyError, df.drop, ['g'], 1)
 
         # errors = 'ignore'
         dropped = df.drop(['g'], errors='ignore')
@@ -87,10 +87,10 @@ def test_drop(self):
         assert_frame_equal(simple.drop(
             [0, 3], axis='index'), simple.loc[[1, 2], :])
 
-        pytest.raises(ValueError, simple.drop, 5)
-        pytest.raises(ValueError, simple.drop, 'C', 1)
-        pytest.raises(ValueError, simple.drop, [1, 5])
-        pytest.raises(ValueError, simple.drop, ['A', 'C'], 1)
+        pytest.raises(KeyError, simple.drop, 5)
+        pytest.raises(KeyError, simple.drop, 'C', 1)
+        pytest.raises(KeyError, simple.drop, [1, 5])
+        pytest.raises(KeyError, simple.drop, ['A', 'C'], 1)
 
         # errors = 'ignore'
         assert_frame_equal(simple.drop(5, errors='ignore'), simple)
@@ -1128,3 +1128,25 @@ def test_reindex_multi(self):
         expected = df.reindex([0, 1]).reindex(columns=['a', 'b'])
 
         assert_frame_equal(result, expected)
+
+    data = [[1, 2, 3], [1, 2, 3]]
+
+    @pytest.mark.parametrize('actual', [
+        DataFrame(data=data, index=['a', 'a']),
+        DataFrame(data=data, index=['a', 'b']),
+        DataFrame(data=data, index=['a', 'b']).set_index([0, 1]),
+        DataFrame(data=data, index=['a', 'a']).set_index([0, 1])
+    ])
+    def test_raise_on_drop_duplicate_index(self, actual):
+
+        # issue 19186
+        level = 0 if isinstance(actual.index, MultiIndex) else None
+        with pytest.raises(KeyError):
+            actual.drop('c', level=level, axis=0)
+            actual.T.drop('c', level=level, axis=1)
+        expected_no_err = actual.drop('c', axis=0, level=level,
+                                      errors='ignore')
+        assert_frame_equal(expected_no_err, actual)
+        expected_no_err = actual.T.drop('c', axis=1, level=level,
+                                        errors='ignore')
+        assert_frame_equal(expected_no_err.T, actual)

pandas/tests/indexes/test_base.py

@@ -1396,8 +1396,8 @@ def test_drop(self):
         expected = self.strIndex[lrange(5) + lrange(10, n)]
         tm.assert_index_equal(dropped, expected)
 
-        pytest.raises(ValueError, self.strIndex.drop, ['foo', 'bar'])
-        pytest.raises(ValueError, self.strIndex.drop, ['1', 'bar'])
+        pytest.raises(KeyError, self.strIndex.drop, ['foo', 'bar'])
+        pytest.raises(KeyError, self.strIndex.drop, ['1', 'bar'])
 
         # errors='ignore'
         mixed = drop.tolist() + ['foo']
@@ -1419,7 +1419,7 @@ def test_drop(self):
         tm.assert_index_equal(dropped, expected)
 
         # errors='ignore'
-        pytest.raises(ValueError, ser.drop, [3, 4])
+        pytest.raises(KeyError, ser.drop, [3, 4])
 
         dropped = ser.drop(4, errors='ignore')
         expected = Index([1, 2, 3])
@@ -1448,7 +1448,7 @@ def test_drop_tuple(self, values, to_drop):
 
         removed = index.drop(to_drop[1])
         for drop_me in to_drop[1], [to_drop[1]]:
-            pytest.raises(ValueError, removed.drop, drop_me)
+            pytest.raises(KeyError, removed.drop, drop_me)
 
     def test_tuple_union_bug(self):
         import pandas

pandas/tests/series/test_indexing.py

@@ -1838,8 +1838,8 @@ def test_drop(self):
 
         # single string/tuple-like
         s = Series(range(3), index=list('abc'))
-        pytest.raises(ValueError, s.drop, 'bc')
-        pytest.raises(ValueError, s.drop, ('a', ))
+        pytest.raises(KeyError, s.drop, 'bc')
+        pytest.raises(KeyError, s.drop, ('a', ))
 
         # errors='ignore'
         s = Series(range(3), index=list('abc'))
@@ -1861,7 +1861,7 @@ def test_drop(self):
 
         # GH 16877
         s = Series([2, 3], index=[0, 1])
-        with tm.assert_raises_regex(ValueError, 'not contained in axis'):
+        with tm.assert_raises_regex(KeyError, 'not contained in axis'):
             s.drop([False, True])
 
     def test_align(self):

pandas/tests/test_panel.py

@@ -2302,7 +2302,7 @@ def check_drop(drop_val, axis_number, aliases, expected):
             expected = Panel({"One": df})
             check_drop('Two', 0, ['items'], expected)
 
-            pytest.raises(ValueError, panel.drop, 'Three')
+            pytest.raises(KeyError, panel.drop, 'Three')
 
             # errors = 'ignore'
             dropped = panel.drop('Three', errors='ignore')