Propogating NaN values when using str.split (#18450)

doc/source/whatsnew/v0.21.1.txt

@@ -141,6 +141,6 @@ Categorical
 Other
 ^^^^^
 
--
+- :meth:`Series.str.split()` will now propogate ``NaN`` values across all expanded columns instead of ``None`` (:issue:`18450`)
 -
 -

pandas/core/strings.py

@@ -1423,6 +1423,10 @@ def cons_row(x):
                     return [x]
 
             result = [cons_row(x) for x in result]
+            if result:
+                # propogate nan values to match longest sequence (GH 18450)
+                max_len = max(len(x) for x in result)
+                result = [x * max_len if x[0] is np.nan else x for x in result]
 
         if not isinstance(expand, bool):
             raise ValueError("expand must be True or False")

pandas/tests/test_strings.py

@@ -2086,6 +2086,15 @@ def test_rsplit_to_multiindex_expand(self):
         tm.assert_index_equal(result, exp)
         assert result.nlevels == 2
 
+    def test_split_nan_expand(self):
+        s = Series(["foo,bar,baz", NA])
+        result = s.str.split(",", expand=True)
+        exp = DataFrame([["foo", "bar", "baz"], [NA, NA, NA]])
+        tm.assert_frame_equal(result, exp)
+
+        # extra nan check - see GH 18463
+        assert all(np.isnan(x) for x in result.iloc[1])
+
     def test_split_with_name(self):
         # GH 12617