BUG: #12815 Always use np.nan for missing values of object dtypes

doc/source/whatsnew/v0.22.0.txt

@@ -48,6 +48,7 @@ Other API Changes
 - :class:`CacheableOffset` and :class:`WeekDay` are no longer available in the ``pandas.tseries.offsets`` module (:issue:`17830`)
 - `tseries.frequencies.get_freq_group()` and `tseries.frequencies.DAYS` are removed from the public API (:issue:`18034`)
 - :func:`Series.truncate` and :func:`DataFrame.truncate` will raise a ``ValueError`` if the index is not sorted instead of an unhelpful ``KeyError`` (:issue:`17935`)
+- :func:`Dataframe.unstack` will now default to filling with ``np.nan`` for ``object`` columns. (:issue:`12815`)
 
 
 .. _whatsnew_0220.deprecations:

pandas/core/dtypes/cast.py

@@ -320,6 +320,7 @@ def maybe_promote(dtype, fill_value=np.nan):
             fill_value = iNaT
         else:
             dtype = np.object_
+            fill_value = np.nan
     else:
         dtype = np.object_
 

pandas/tests/frame/test_reshape.py

@@ -783,3 +783,26 @@ def test_stack_preserve_categorical_dtype(self):
                 expected = Series([10, 11, 12], index=midx)
 
                 tm.assert_series_equal(result, expected)
+
+
+def test_unstack_fill_frame_object():
+    # GH12815 Test unstacking with object.
+    data = pd.Series(['a', 'b', 'c', 'a'], dtype='object')
+    data.index = pd.MultiIndex.from_tuples(
+        [('x', 'a'), ('x', 'b'), ('y', 'b'), ('z', 'a')])
+
+    # By default missing values will be NaN
+    result = data.unstack()
+    expected = pd.DataFrame(
+        {'a': ['a', np.nan, 'a'], 'b': ['b', 'c', np.nan]},
+        index=list('xyz')
+    )
+    assert_frame_equal(result, expected)
+
+    # Fill with any value replaces missing values as expected
+    result = data.unstack(fill_value='d')
+    expected = pd.DataFrame(
+        {'a': ['a', 'd', 'a'], 'b': ['b', 'c', 'd']},
+        index=list('xyz')
+    )
+    assert_frame_equal(result, expected)