Merge pull request #6396 from jreback/index_regr

BUG: Regression in chained getitem indexing with embedded list-like from 0.12 (GH6394)

Author: jreback

doc/source/release.rst

@@ -137,6 +137,7 @@ Bug Fixes
 - Bug in Series.get, was using a buggy access method (:issue:`6383`)
 - Bug in hdfstore queries of the form ``where=[('date', '>=', datetime(2013,1,1)), ('date', '<=', datetime(2014,1,1))]`` (:issue:`6313`)
 - Bug in DataFrame.dropna with duplicate indices (:issue:`6355`)
+- Regression in chained getitem indexing with embedded list-like from 0.12 (:issue:`6394`)
 
 pandas 0.13.1
 -------------

pandas/core/generic.py

@@ -1312,7 +1312,11 @@ def xs(self, key, axis=0, level=None, copy=True, drop_level=True):
             new_values, copy = self._data.fast_xs(loc, copy=copy)
 
             # may need to box a datelike-scalar
-            if not is_list_like(new_values):
+            #
+            # if we encounter an array-like and we only have 1 dim
+            # that means that their are list/ndarrays inside the Series!
+            # so just return them (GH 6394)
+            if not is_list_like(new_values) or self.ndim == 1:
                 return _maybe_box_datetimelike(new_values)
 
             result = Series(new_values, index=self.columns,

pandas/core/series.py

@@ -477,8 +477,15 @@ def _slice(self, slobj, axis=0, raise_on_error=False, typ=None):
     def __getitem__(self, key):
         try:
             result = self.index.get_value(self, key)
-            if isinstance(result, np.ndarray):
-                return self._constructor(result,index=[key]*len(result)).__finalize__(self)
+
+            if not np.isscalar(result):
+                if is_list_like(result) and not isinstance(result, Series):
+
+                    # we need to box if we have a non-unique index here
+                    # otherwise have inline ndarray/lists
+                    if not self.index.is_unique:
+                        result = self._constructor(result,index=[key]*len(result)).__finalize__(self)
+
             return result
         except InvalidIndexError:
             pass

pandas/tests/test_indexing.py

@@ -551,6 +551,26 @@ def test_loc_setitem(self):
         expected = DataFrame({'a' : [0.5,-0.5,-1.5], 'b' : [0,1,2] })
         assert_frame_equal(df,expected)
 
+    def test_chained_getitem_with_lists(self):
+
+        # GH6394
+        # Regression in chained getitem indexing with embedded list-like from 0.12
+        def check(result, expected):
+            self.assert_numpy_array_equal(result,expected)
+            tm.assert_isinstance(result, np.ndarray)
+
+
+        df = DataFrame({'A': 5*[np.zeros(3)], 'B':5*[np.ones(3)]})
+        expected = df['A'].iloc[2]
+        result = df.loc[2,'A']
+        check(result, expected)
+        result2 = df.iloc[2]['A']
+        check(result2, expected)
+        result3 = df['A'].loc[2]
+        check(result3, expected)
+        result4 = df['A'].iloc[2]
+        check(result4, expected)
+
     def test_loc_getitem_int(self):
 
         # int label

pandas/tests/test_series.py

@@ -949,6 +949,12 @@ def test_getitem_dups_with_missing(self):
         result = s[['foo', 'bar', 'bah', 'bam']]
         assert_series_equal(result, expected)
 
+    def test_getitem_dups(self):
+        s = Series(range(5),index=['A','A','B','C','C'])
+        expected = Series([3,4],index=['C','C'])
+        result = s['C']
+        assert_series_equal(result, expected)
+
     def test_setitem_ambiguous_keyerror(self):
         s = Series(lrange(10), index=lrange(0, 20, 2))
 
@@ -4813,6 +4819,7 @@ def test_apply_args(self):
 
         result = s.apply(str.split, args=(',',))
         self.assertEqual(result[0], ['foo', 'bar'])
+        tm.assert_isinstance(result[0], list)
 
     def test_align(self):
         def _check_align(a, b, how='left', fill=None):