DOC: Add whatsnew note

Author: gfyoung

doc/source/whatsnew/v0.24.0.txt

@@ -867,3 +867,4 @@ Other
 - :meth:`DataFrame.nlargest` and :meth:`DataFrame.nsmallest` now returns the correct n values when keep != 'all' also when tied on the first columns (:issue:`22752`)
 - :meth:`~pandas.io.formats.style.Styler.bar` now also supports tablewise application (in addition to rowwise and columnwise) with ``axis=None`` and setting clipping range with ``vmin`` and ``vmax`` (:issue:`21548` and :issue:`21526`). ``NaN`` values are also handled properly.
 - Logical operations ``&, |, ^`` between :class:`Series` and :class:`Index` will no longer raise ``ValueError`` (:issue:`22092`)
+- Bug in :meth:`DataFrame.combine_first` in which column types were unexpectedly converted to float (:issue:`20699`)

pandas/core/frame.py

@@ -5077,7 +5077,7 @@ def combine(self, other, func, fill_value=None, overwrite=True):
                 # try to promote series, which is all NaN, as other_dtype.
                 new_dtype = other_dtype
                 try:
-                    series = series.astype(new_dtype)
+                    series = series.astype(new_dtype, copy=False)
                 except ValueError:
                     # e.g. new_dtype is integer types
                     pass

pandas/tests/frame/test_combine_concat.py

@@ -4,6 +4,7 @@
 
 from datetime import datetime
 
+import pytest
 import numpy as np
 from numpy import nan
 
@@ -750,13 +751,14 @@ def test_combine_first_int(self):
         tm.assert_frame_equal(res, df1)
         assert res['a'].dtype == 'int64'
 
-    def test_combine_first_with_asymmetric_other(self):
-        # GH20699
-        df1 = pd.DataFrame({'isInt': [1]})
+    @pytest.mark.parametrize("val", [1, 1.0])
+    def test_combine_first_with_asymmetric_other(self, val):
+        # see gh-20699
+        df1 = pd.DataFrame({'isNum': [val]})
         df2 = pd.DataFrame({'isBool': [True]})
 
         res = df1.combine_first(df2)
-        exp = pd.DataFrame({'isBool': [True], 'isInt': [1]})
+        exp = pd.DataFrame({'isBool': [True], 'isNum': [val]})
 
         tm.assert_frame_equal(res, exp)