Merge pull request #8176 from jreback/expand

BUG/API: Previously an enlargement with a mixed-dtype frame would act unlike append (related GH2578)

Author: jreback

doc/source/v0.15.0.txt

@@ -296,6 +296,19 @@ API changes
 
   To insert a NaN, you must explicitly use ``np.nan``. See the :ref:`docs <missing.inserting>`.
 
+- Previously an enlargement with a mixed-dtype frame would act unlike ``.append`` which will preserve dtypes (related :issue:`2578`, :issue:`8176`):
+
+  .. ipython:: python
+
+     df = DataFrame([[True, 1],[False, 2]], columns = ["female","fitness"])
+     df
+     df.dtypes
+
+     # dtypes are now preserved
+     df.loc[2] = df.loc[1]
+     df
+     df.dtypes
+
 .. _whatsnew_0150.dt:
 
 .dt accessor

pandas/core/indexing.py

@@ -303,12 +303,27 @@ def _setitem_with_indexer(self, indexer, value):
                             "cannot set a frame with no defined columns"
                         )
 
-                    index = self.obj._get_axis(0)
-                    labels = _safe_append_to_index(index, indexer)
-                    self.obj._data = self.obj.reindex_axis(labels, 0)._data
+                    # append a Series
+                    if isinstance(value, Series):
+
+                        value = value.reindex(index=self.obj.columns,copy=True)
+                        value.name = indexer
+
+                    # a list-list
+                    else:
+
+                        # must have conforming columns
+                        if com.is_list_like(value):
+                            if len(value) != len(self.obj.columns):
+                                raise ValueError(
+                                    "cannot set a row with mismatched columns"
+                                    )
+
+                        value = Series(value,index=self.obj.columns,name=indexer)
+
+                    self.obj._data = self.obj.append(value)._data
                     self.obj._maybe_update_cacher(clear=True)
-                    return getattr(self.obj, self.name).__setitem__(indexer,
-                                                                    value)
+                    return self.obj
 
                 # set using setitem (Panel and > dims)
                 elif self.ndim >= 3:

pandas/tests/test_indexing.py

@@ -2812,7 +2812,8 @@ def f():
         df.loc[1] = df.loc[2]
         assert_frame_equal(df,expected)
 
-        expected = DataFrame(dict({ 'A' : [0,2,4,4], 'B' : [1,3,5,5] }),dtype='float64')
+        # like 2578, partial setting with dtype preservation
+        expected = DataFrame(dict({ 'A' : [0,2,4,4], 'B' : [1,3,5,5] }))
         df = df_orig.copy()
         df.loc[3] = df.loc[2]
         assert_frame_equal(df,expected)
@@ -2864,6 +2865,41 @@ def f():
         p.loc[:,:,'C'] = Series([30,32],index=p_orig.items)
         assert_panel_equal(p,expected)
 
+    def test_partial_setting_mixed_dtype(self):
+
+        # in a mixed dtype environment, try to preserve dtypes
+        # by appending
+        df = DataFrame([[True, 1],[False, 2]],
+                       columns = ["female","fitness"])
+
+        s = df.loc[1].copy()
+        s.name = 2
+        expected = df.append(s)
+
+        df.loc[2] = df.loc[1]
+        assert_frame_equal(df, expected)
+
+        # columns will align
+        df = DataFrame(columns=['A','B'])
+        df.loc[0] = Series(1,index=range(4))
+        assert_frame_equal(df,DataFrame(columns=['A','B'],index=[0]))
+
+        # columns will align
+        df = DataFrame(columns=['A','B'])
+        df.loc[0] = Series(1,index=['B'])
+        assert_frame_equal(df,DataFrame([[np.nan, 1]], columns=['A','B'],index=[0],dtype='float64'))
+
+        # list-like must conform
+        df = DataFrame(columns=['A','B'])
+        def f():
+            df.loc[0] = [1,2,3]
+        self.assertRaises(ValueError, f)
+
+        # these are coerced to float unavoidably (as its a list-like to begin)
+        df = DataFrame(columns=['A','B'])
+        df.loc[3] = [6,7]
+        assert_frame_equal(df,DataFrame([[6,7]],index=[3],columns=['A','B'],dtype='float64'))
+
     def test_series_partial_set(self):
         # partial set with new index
         # Regression from GH4825
@@ -3013,15 +3049,6 @@ def f():
         assert_frame_equal(df,DataFrame([[1]],index=['foo'],columns=[1]))
         assert_frame_equal(df,df2)
 
-        df = DataFrame(columns=['A','B'])
-        df.loc[3] = [6,7]
-        assert_frame_equal(df,DataFrame([[6,7]],index=[3],columns=['A','B']))
-
-        # no label overlap
-        df = DataFrame(columns=['A','B'])
-        df.loc[0] = Series(1,index=range(4))
-        assert_frame_equal(df,DataFrame(columns=['A','B'],index=[0]))
-
         # no index to start
         expected = DataFrame({ 0 : Series(1,index=range(4)) },columns=['A','B',0])