BUG: handle nan values in DataFrame.update when overwrite=False (#15593) (#16430)

(cherry picked from commit 85080aa)

Author: pcluo

doc/source/whatsnew/v0.20.2.txt

@@ -37,6 +37,7 @@ Bug Fixes
 ~~~~~~~~~
 
 - Bug in using ``pathlib.Path`` or ``py.path.local`` objects with io functions (:issue:`16291`)
+- Bug in ``DataFrame.update()`` with ``overwrite=False`` and ``NaN values`` (:issue:`15593`)
 
 
 - Fixed a compatibility issue with IPython 6.0's tab completion showing deprecation warnings on Categoricals (:issue:`16409`)

pandas/core/frame.py

@@ -3852,13 +3852,13 @@ def update(self, other, join='left', overwrite=True, filter_func=None,
 
                 if overwrite:
                     mask = isnull(that)
-
-                    # don't overwrite columns unecessarily
-                    if mask.all():
-                        continue
                 else:
                     mask = notnull(this)
 
+            # don't overwrite columns unecessarily
+            if mask.all():
+                continue
+
             self[col] = expressions.where(mask, this, that,
                                           raise_on_error=True)
 

pandas/tests/frame/test_combine_concat.py

@@ -763,3 +763,25 @@ def test_concat_datetime_datetime64_frame(self):
 
         # it works!
         pd.concat([df1, df2_obj])
+
+
+class TestDataFrameUpdate(TestData):
+
+    def test_update_nan(self):
+        # #15593 #15617
+        # test 1
+        df1 = DataFrame({'A': [1.0, 2, 3], 'B': date_range('2000', periods=3)})
+        df2 = DataFrame({'A': [None, 2, 3]})
+        expected = df1.copy()
+        df1.update(df2, overwrite=False)
+
+        tm.assert_frame_equal(df1, expected)
+
+        # test 2
+        df1 = DataFrame({'A': [1.0, None, 3], 'B': date_range('2000', periods=3)})
+        df2 = DataFrame({'A': [None, 2, 3]})
+        expected = DataFrame({'A': [1.0, 2, 3], 'B': date_range('2000', periods=3)})
+        df1.update(df2, overwrite=False)
+
+        tm.assert_frame_equal(df1, expected)
+