iloc and loc return new object when slice is a tuple

Author: margaret

pandas/core/indexing.py

@@ -845,7 +845,7 @@ def _getitem_tuple(self, tup):
             return self._multi_take(tup)
 
         # no shortcut needed
-        retval = self.obj
+        retval = self.obj.copy()
         for i, key in enumerate(tup):
             if i >= self.obj.ndim:
                 raise IndexingError('Too many indexers')
@@ -854,7 +854,6 @@ def _getitem_tuple(self, tup):
                 continue
 
             retval = getattr(retval, self.name)._getitem_axis(key, axis=i)
-
         return retval
 
     def _multi_take_opportunity(self, tup):
@@ -1665,7 +1664,7 @@ def _getitem_tuple(self, tup):
         except:
             pass
 
-        retval = self.obj
+        retval = self.obj.copy()
         axis = 0
         for i, key in enumerate(tup):
             if i >= self.obj.ndim:

pandas/tests/indexing/test_iloc.py

@@ -596,3 +596,5 @@ def test_loc_identity_slice_returns_new_object(self):
         # GH13873
         df = DataFrame({'a': [1, 3, 5], 'b': [2, 4, 6]})
         assert not df.iloc[:] is df
+        assert not df.iloc[:,:] is df
+        assert not df.iloc[pd.IndexSlice[:, :]] is df

pandas/tests/indexing/test_loc.py

@@ -635,3 +635,5 @@ def test_loc_identity_slice_returns_new_object(self):
         # GH13873
         df = DataFrame({'a': [1, 3, 5], 'b': [2, 4, 6]})
         assert not df.loc[:] is df
+        assert not df.loc[:,:] is df
+        assert not df.loc[pd.IndexSlice[:, :]] is df