Added backfill for Series & DataFrame

databricks/koalas/generic.py

@@ -2393,6 +2393,8 @@ def bfill(self, axis=None, inplace=False, limit=None):
         """
         return self.fillna(method="bfill", axis=axis, inplace=inplace, limit=limit)
 
+    backfill = bfill
+
     # TODO: add 'downcast' when value parameter exists
     def ffill(self, axis=None, inplace=False, limit=None):
         """

databricks/koalas/missing/frame.py

@@ -39,7 +39,6 @@ class _MissingPandasLikeDataFrame(object):
     asfreq = _unsupported_function("asfreq")
     asof = _unsupported_function("asof")
     at_time = _unsupported_function("at_time")
-    backfill = _unsupported_function("_backfill")
     between_time = _unsupported_function("between_time")
     boxplot = _unsupported_function("boxplot")
     combine = _unsupported_function("combine")

databricks/koalas/missing/series.py

@@ -39,7 +39,6 @@ class MissingPandasLikeSeries(object):
     asfreq = _unsupported_function("asfreq")
     at_time = _unsupported_function("at_time")
     autocorr = _unsupported_function("autocorr")
-    backfill = _unsupported_function("backfill")
     between_time = _unsupported_function("between_time")
     combine = _unsupported_function("combine")
     compare = _unsupported_function("compare")

databricks/koalas/tests/test_dataframe.py

@@ -4132,3 +4132,38 @@ def test_pad(self):
             # Test `inplace=True`
             kdf.pad(inplace=True)
             self.assert_eq(expected, kdf)
+
+    def test_backfill(self):
+        pdf = pd.DataFrame(
+            {
+                "A": [None, 3, None, None],
+                "B": [2, 4, None, 3],
+                "C": [None, None, None, 1],
+                "D": [0, 1, 5, 4],
+            },
+            columns=["A", "B", "C", "D"],
+        )
+        kdf = ks.from_pandas(pdf)
+
+        if LooseVersion(pd.__version__) >= LooseVersion("1.1"):
+            self.assert_eq(pdf.backfill(), kdf.backfill())
+
+            # Test `inplace=True`
+            pdf.backfill(inplace=True)
+            kdf.backfill(inplace=True)
+            self.assert_eq(pdf, kdf)
+        else:
+            expected = ks.DataFrame(
+                {
+                    "A": [3.0, 3.0, None, None],
+                    "B": [2.0, 4.0, 3.0, 3.0],
+                    "C": [1.0, 1.0, 1.0, 1.0],
+                    "D": [0, 1, 5, 4],
+                },
+                columns=["A", "B", "C", "D"],
+            )
+            self.assert_eq(expected, kdf.backfill())
+
+            # Test `inplace=True`
+            kdf.backfill(inplace=True)
+            self.assert_eq(expected, kdf)

databricks/koalas/tests/test_series.py

@@ -2207,3 +2207,22 @@ def test_argmin_argmax(self):
             ks.Series([]).argmin()
         with self.assertRaisesRegex(ValueError, "attempt to get argmax of an empty sequence"):
             ks.Series([]).argmax()
+
+    def test_backfill(self):
+        pser = pd.Series([np.nan, 2, 3, 4, np.nan, 6], name="x")
+        kser = ks.from_pandas(pser)
+
+        if LooseVersion(pd.__version__) >= LooseVersion("1.1"):
+            self.assert_eq(pser.backfill(), kser.backfill())
+
+            # Test `inplace=True`
+            pser.backfill(inplace=True)
+            kser.backfill(inplace=True)
+            self.assert_eq(pser, kser)
+        else:
+            expected = ks.Series([2.0, 2.0, 3.0, 4.0, 6.0, 6.0], name="x")
+            self.assert_eq(expected, kser.backfill())
+
+            # Test `inplace=True`
+            kser.backfill(inplace=True)
+            self.assert_eq(expected, kser)

docs/source/reference/frame.rst

@@ -182,6 +182,7 @@ Missing data handling
 .. autosummary::
    :toctree: api/
 
+   DataFrame.backfill
    DataFrame.dropna
    DataFrame.fillna
    DataFrame.replace

docs/source/reference/series.rst

@@ -184,6 +184,7 @@ Missing data handling
 .. autosummary::
    :toctree: api/
 
+   Series.backfill
    Series.isna
    Series.isnull
    Series.notna