ERR: Raise ValueError when setting scalars in a dataframe with no index ( pandas-dev#16823) (pandas-dev#16968)

Author: alanbato

Diff for: doc/source/whatsnew/v0.21.0.txt

@@ -706,6 +706,8 @@ Other API Changes
 - Restricted DateOffset keyword arguments.  Previously, ``DateOffset`` subclasses allowed arbitrary keyword arguments which could lead to unexpected behavior.  Now, only valid arguments will be accepted. (:issue:`17176`).
 - Pandas no longer registers matplotlib converters on import. The converters
   will be registered and used when the first plot is draw (:issue:`17710`)
+- Setting on a column with a scalar value and 0-len index now raises a ``ValueError`` (:issue:`16823`)
+
 
 .. _whatsnew_0210.deprecations:
 

Diff for: pandas/core/frame.py

@@ -2531,13 +2531,17 @@ def _ensure_valid_index(self, value):
         passed value
         """
         # GH5632, make sure that we are a Series convertible
-        if not len(self.index) and is_list_like(value):
+        if not len(self.index):
+            if not is_list_like(value):
+                # GH16823, Raise an error due to loss of information
+                raise ValueError('If using all scalar values, you must pass'
+                                 ' an index')
             try:
                 value = Series(value)
             except:
-                raise ValueError('Cannot set a frame with no defined index '
-                                 'and a value that cannot be converted to a '
-                                 'Series')
+                raise ValueError('Cannot set a frame with no defined'
+                                 'index and a value that cannot be '
+                                 'converted to a Series')
 
             self._data = self._data.reindex_axis(value.index.copy(), axis=1,
                                                  fill_value=np.nan)

Diff for: pandas/core/reshape/pivot.py

@@ -454,6 +454,9 @@ def crosstab(index, columns, values=None, rownames=None, colnames=None,
 
     from pandas import DataFrame
     df = DataFrame(data, index=common_idx)
+    if not len(df):
+        return DataFrame(index=common_idx)
+
     if values is None:
         df['__dummy__'] = 0
         kwargs = {'aggfunc': len, 'fill_value': 0}

Diff for: pandas/tests/frame/test_indexing.py

@@ -721,6 +721,11 @@ def test_setitem_empty_frame_with_boolean(self):
                 df[df > df2] = 47
                 assert_frame_equal(df, df2)
 
+    def test_setitem_scalars_no_index(self):
+        # GH16823
+        df = DataFrame()
+        pytest.raises(ValueError, df.__setitem__, 'foo', 1)
+
     def test_getitem_empty_frame_with_boolean(self):
         # Test for issue #11859
 

Diff for: pandas/tests/indexing/test_loc.py

@@ -423,15 +423,14 @@ def test_loc_setitem_consistency(self):
 
     def test_loc_setitem_consistency_empty(self):
         # empty (essentially noops)
-        expected = DataFrame(columns=['x', 'y'])
-        expected['x'] = expected['x'].astype(np.int64)
+        # GH16823
         df = DataFrame(columns=['x', 'y'])
-        df.loc[:, 'x'] = 1
-        tm.assert_frame_equal(df, expected)
+        with tm.assert_raises_regex(ValueError, 'If using all scalar values'):
+            df.loc[:, 'x'] = 1
 
         df = DataFrame(columns=['x', 'y'])
-        df['x'] = 1
-        tm.assert_frame_equal(df, expected)
+        with tm.assert_raises_regex(ValueError, 'If using all scalar values'):
+            df['x'] = 1
 
     def test_loc_setitem_consistency_slice_column_len(self):
         # .loc[:,column] setting with slice == len of the column

Diff for: pandas/tests/indexing/test_partial.py

@@ -575,24 +575,16 @@ def f():
     def test_partial_set_empty_frame_row(self):
         # GH5720, GH5744
         # don't create rows when empty
-        expected = DataFrame(columns=['A', 'B', 'New'],
-                             index=pd.Index([], dtype='int64'))
-        expected['A'] = expected['A'].astype('int64')
-        expected['B'] = expected['B'].astype('float64')
-        expected['New'] = expected['New'].astype('float64')
-
         df = DataFrame({"A": [1, 2, 3], "B": [1.2, 4.2, 5.2]})
         y = df[df.A > 5]
-        y['New'] = np.nan
-        tm.assert_frame_equal(y, expected)
-        # tm.assert_frame_equal(y,expected)
+        # GH16823
+        # Setting a column with a scalar and no index should raise
+        with tm.assert_raises_regex(ValueError, 'If using all scalar values'):
+            y['New'] = np.nan
 
-        expected = DataFrame(columns=['a', 'b', 'c c', 'd'])
-        expected['d'] = expected['d'].astype('int64')
         df = DataFrame(columns=['a', 'b', 'c c'])
-        df['d'] = 3
-        tm.assert_frame_equal(df, expected)
-        tm.assert_series_equal(df['c c'], Series(name='c c', dtype=object))
+        with tm.assert_raises_regex(ValueError, 'If using all scalar values'):
+            df['d'] = 3
 
         # reindex columns is ok
         df = DataFrame({"A": [1, 2, 3], "B": [1.2, 4.2, 5.2]})

Diff for: pandas/tests/reshape/test_pivot.py

@@ -1226,7 +1226,8 @@ def test_crosstab_no_overlap(self):
         s2 = pd.Series([4, 5, 6], index=[4, 5, 6])
 
         actual = crosstab(s1, s2)
-        expected = pd.DataFrame()
+        expected = pd.DataFrame(
+            index=pd.Index([], dtype='int64')).astype('int64')
 
         tm.assert_frame_equal(actual, expected)