From c22c01b2e6884b0da2bab86d6c3aed4735414fdf Mon Sep 17 00:00:00 2001
From: makbigc <makszechun@gmail.com>
Date: Sat, 11 May 2019 22:33:06 +0800
Subject: [PATCH 01/22] Add mask as return value of _values_for_argsort

---
 pandas/core/arrays/base.py                    | 29 ++++++++++++++++---
 pandas/core/arrays/categorical.py             |  2 +-
 pandas/core/arrays/datetimelike.py            |  2 +-
 pandas/core/arrays/integer.py                 |  6 ++--
 pandas/core/arrays/numpy_.py                  |  2 +-
 pandas/core/arrays/period.py                  |  2 +-
 pandas/tests/extension/base/methods.py        |  5 ++++
 .../tests/extension/decimal/test_decimal.py   |  6 ++++
 pandas/tests/extension/json/array.py          |  4 +--
 pandas/tests/extension/json/test_json.py      |  6 ++++
 pandas/tests/extension/test_categorical.py    |  6 ++++
 pandas/tests/extension/test_datetime.py       |  9 ++++++
 pandas/tests/extension/test_integer.py        |  6 ++++
 pandas/tests/extension/test_interval.py       |  7 +++++
 pandas/tests/extension/test_numpy.py          | 17 +++++++++++
 pandas/tests/extension/test_period.py         |  6 ++++
 pandas/tests/extension/test_sparse.py         |  6 ++++
 17 files changed, 109 insertions(+), 12 deletions(-)

diff --git a/pandas/core/arrays/base.py b/pandas/core/arrays/base.py
index c709cd9e9f0b2..c0e88bf6cfc1c 100644
--- a/pandas/core/arrays/base.py
+++ b/pandas/core/arrays/base.py
@@ -361,7 +361,7 @@ def isna(self) -> ArrayLike:
         """
         raise AbstractMethodError(self)
 
-    def _values_for_argsort(self) -> np.ndarray:
+    def _values_for_argsort(self) -> Tuple[np.ndarray, np.ndarray]:
         """
         Return values for sorting.
 
@@ -376,7 +376,7 @@ def _values_for_argsort(self) -> np.ndarray:
         ExtensionArray.argsort
         """
         # Note: this is used in `ExtensionArray.argsort`.
-        return np.array(self)
+        return np.array(self), self.isna()
 
     def argsort(self, ascending=True, kind='quicksort', *args, **kwargs):
         """
@@ -406,8 +406,29 @@ def argsort(self, ascending=True, kind='quicksort', *args, **kwargs):
         # 1. _values_for_argsort : construct the values passed to np.argsort
         # 2. argsort : total control over sorting.
         ascending = nv.validate_argsort_with_ascending(ascending, args, kwargs)
-        values = self._values_for_argsort()
-        result = np.argsort(values, kind=kind, **kwargs)
+        values, mask = self._values_for_argsort()
+
+        def permutation(mask):
+            # Return a permutation which maps the indices of the
+            # subarray without nan to the indices of the original array.
+            permu = np.arange(len(mask))
+            nan_loc = np.arange(len(mask))[mask]
+            offset = 0
+            for x in nan_loc:
+                permu[x - offset:] += 1
+                offset += 1
+            return permu
+
+        if mask.any():
+            notmask = ~mask
+            notnull = np.argsort(values[notmask], kind=kind, **kwargs)
+            permu = permutation(mask)
+            notnull = permu[notnull]
+            allnan = np.arange(len(self))[mask]
+            result = np.append(notnull, allnan)
+        else:
+            result = np.argsort(values, kind=kind, **kwargs)
+
         if not ascending:
             result = result[::-1]
         return result
diff --git a/pandas/core/arrays/categorical.py b/pandas/core/arrays/categorical.py
index 49dd0041854bc..c4ce43b55741b 100644
--- a/pandas/core/arrays/categorical.py
+++ b/pandas/core/arrays/categorical.py
@@ -1521,7 +1521,7 @@ def check_for_ordered(self, op):
                             "Categorical to an ordered one\n".format(op=op))
 
     def _values_for_argsort(self):
-        return self._codes.copy()
+        return self._codes.copy(), self.isna()
 
     def argsort(self, *args, **kwargs):
         # TODO(PY2): use correct signature
diff --git a/pandas/core/arrays/datetimelike.py b/pandas/core/arrays/datetimelike.py
index c32f8642dc2ed..a1696c3fd52da 100644
--- a/pandas/core/arrays/datetimelike.py
+++ b/pandas/core/arrays/datetimelike.py
@@ -621,7 +621,7 @@ def _from_factorized(cls, values, original):
         return cls(values, dtype=original.dtype)
 
     def _values_for_argsort(self):
-        return self._data
+        return self._data, self._isnan
 
     # ------------------------------------------------------------------
     # Additional array methods
diff --git a/pandas/core/arrays/integer.py b/pandas/core/arrays/integer.py
index 42aa6a055acca..78dda4790a067 100644
--- a/pandas/core/arrays/integer.py
+++ b/pandas/core/arrays/integer.py
@@ -512,7 +512,8 @@ def value_counts(self, dropna=True):
 
         return Series(array, index=index)
 
-    def _values_for_argsort(self) -> np.ndarray:
+    #def _values_for_argsort(self) -> np.ndarray:
+    def _values_for_argsort(self):
         """Return values for sorting.
 
         Returns
@@ -526,8 +527,9 @@ def _values_for_argsort(self) -> np.ndarray:
         ExtensionArray.argsort
         """
         data = self._data.copy()
+        mask = self._mask
         data[self._mask] = data.min() - 1
-        return data
+        return data, mask
 
     @classmethod
     def _create_comparison_method(cls, op):
diff --git a/pandas/core/arrays/numpy_.py b/pandas/core/arrays/numpy_.py
index 0deefd2b10b6e..ab71d45bdbaf0 100644
--- a/pandas/core/arrays/numpy_.py
+++ b/pandas/core/arrays/numpy_.py
@@ -281,7 +281,7 @@ def copy(self, deep=False):
         return type(self)(self._ndarray.copy())
 
     def _values_for_argsort(self):
-        return self._ndarray
+        return self._ndarray, self.isna()
 
     def _values_for_factorize(self):
         return self._ndarray, -1
diff --git a/pandas/core/arrays/period.py b/pandas/core/arrays/period.py
index 8a6640e11ad74..8f1da6dbe5c4a 100644
--- a/pandas/core/arrays/period.py
+++ b/pandas/core/arrays/period.py
@@ -674,7 +674,7 @@ def _check_timedeltalike_freq_compat(self, other):
         _raise_on_incompatible(self, other)
 
     def _values_for_argsort(self):
-        return self._data
+        return self._data, self._isnan
 
 
 PeriodArray._add_comparison_ops()
diff --git a/pandas/tests/extension/base/methods.py b/pandas/tests/extension/base/methods.py
index 1852edaa9e748..be7346e66ed4c 100644
--- a/pandas/tests/extension/base/methods.py
+++ b/pandas/tests/extension/base/methods.py
@@ -44,6 +44,11 @@ def test_argsort_missing(self, data_missing_for_sorting):
         expected = pd.Series(np.array([1, -1, 0], dtype=np.int64))
         self.assert_series_equal(result, expected)
 
+    def test_argsort_nan_loc(self, data_multiple_nan):
+        result = data_multiple_nan.argsort()
+        expected = np.array([3, 9, 7, 1, 0, 6, 2, 4, 5, 8])
+        tm.assert_numpy_array_equal(result, expected)
+
     @pytest.mark.parametrize('ascending', [True, False])
     def test_sort_values(self, data_for_sorting, ascending):
         ser = pd.Series(data_for_sorting)
diff --git a/pandas/tests/extension/decimal/test_decimal.py b/pandas/tests/extension/decimal/test_decimal.py
index 97fae41bcc720..ba68a8edc947c 100644
--- a/pandas/tests/extension/decimal/test_decimal.py
+++ b/pandas/tests/extension/decimal/test_decimal.py
@@ -46,6 +46,12 @@ def data_missing_for_sorting():
                          decimal.Decimal('0')])
 
 
+@pytest.fixture
+def data_multiple_nan():
+    return DecimalArray([decimal.Decimal(x) for x in
+                        [5, 4, np.nan, 1, np.nan, np.nan, 6, 3, np.nan, 2]])
+
+
 @pytest.fixture
 def na_cmp():
     return lambda x, y: x.is_nan() and y.is_nan()
diff --git a/pandas/tests/extension/json/array.py b/pandas/tests/extension/json/array.py
index 4b93f0e12e32a..dde88a4838e6a 100644
--- a/pandas/tests/extension/json/array.py
+++ b/pandas/tests/extension/json/array.py
@@ -171,7 +171,7 @@ def _concat_same_type(cls, to_concat):
         return cls(data)
 
     def _values_for_factorize(self):
-        frozen = self._values_for_argsort()
+        frozen, _ = self._values_for_argsort()
         if len(frozen) == 0:
             # _factorize_array expects 1-d array, this is a len-0 2-d array.
             frozen = frozen.ravel()
@@ -182,7 +182,7 @@ def _values_for_argsort(self):
         # If all the elemnts of self are the same size P, NumPy will
         # cast them to an (N, P) array, instead of an (N,) array of tuples.
         frozen = [()] + [tuple(x.items()) for x in self]
-        return np.array(frozen, dtype=object)[1:]
+        return np.array(frozen, dtype=object)[1:], self.isna()
 
 
 def make_data():
diff --git a/pandas/tests/extension/json/test_json.py b/pandas/tests/extension/json/test_json.py
index 97c329e0a5c92..47b7fd670b1da 100644
--- a/pandas/tests/extension/json/test_json.py
+++ b/pandas/tests/extension/json/test_json.py
@@ -50,6 +50,12 @@ def data_missing_for_sorting():
     return JSONArray([{'b': 1}, {}, {'a': 4}])
 
 
+@pytest.fixture
+def data_multiple_nan():
+    return JSONArray([{'e': 5}, {'d': 4}, {}, {'a': 1}, {},
+                      {}, {'f': 6}, {'c': 3}, {}, {'b': 2}])
+
+
 @pytest.fixture
 def na_value(dtype):
     return dtype.na_value
diff --git a/pandas/tests/extension/test_categorical.py b/pandas/tests/extension/test_categorical.py
index 4cf9f78e1531d..6105db64d8f19 100644
--- a/pandas/tests/extension/test_categorical.py
+++ b/pandas/tests/extension/test_categorical.py
@@ -68,6 +68,12 @@ def data_missing_for_sorting():
                        ordered=True)
 
 
+@pytest.fixture
+def data_multiple_nan():
+    return Categorical(['E', 'D', None, 'A', None,
+                        None, 'F', 'C', None, 'B'], ordered=True)
+
+
 @pytest.fixture
 def na_value():
     return np.nan
diff --git a/pandas/tests/extension/test_datetime.py b/pandas/tests/extension/test_datetime.py
index baee04c3b79eb..600410e8864cb 100644
--- a/pandas/tests/extension/test_datetime.py
+++ b/pandas/tests/extension/test_datetime.py
@@ -45,6 +45,15 @@ def data_missing_for_sorting(dtype):
                          dtype=dtype)
 
 
+@pytest.fixture
+def data_multiple_nan(dtype):
+    idx = pd.date_range(start='2000-01-01', end='2000-01-06')
+    return DatetimeArray(np.array([idx[4], idx[3], 'NaT', idx[0], 'NaT',
+                                   'NaT', idx[5], idx[2], 'NaT', idx[1]],
+                                  dtype='datetime64[ns]'),
+                         dtype=dtype)
+
+
 @pytest.fixture
 def data_for_grouping(dtype):
     """
diff --git a/pandas/tests/extension/test_integer.py b/pandas/tests/extension/test_integer.py
index 22bb086a919ca..09295138c74ca 100644
--- a/pandas/tests/extension/test_integer.py
+++ b/pandas/tests/extension/test_integer.py
@@ -57,6 +57,12 @@ def data_for_sorting(dtype):
     return integer_array([1, 2, 0], dtype=dtype)
 
 
+@pytest.fixture
+def data_multiple_nan(dtype):
+    return integer_array([5, 4, np.nan, 1, np.nan,
+                          np.nan, 6, 3, np.nan, 2], dtype=dtype)
+
+
 @pytest.fixture
 def data_missing_for_sorting(dtype):
     return integer_array([1, np.nan, 0], dtype=dtype)
diff --git a/pandas/tests/extension/test_interval.py b/pandas/tests/extension/test_interval.py
index f1f90b298ffe2..ff322c7db0ed3 100644
--- a/pandas/tests/extension/test_interval.py
+++ b/pandas/tests/extension/test_interval.py
@@ -57,6 +57,13 @@ def data_missing_for_sorting():
     return IntervalArray.from_tuples([(1, 2), None, (0, 1)])
 
 
+@pytest.fixture
+def data_multiple_nan():
+    return IntervalArray.from_tuples([(5, 6), (4, 5), None, (1, 2),
+                                      None, None, (6, 7), (3, 4),
+                                      None, (2, 3)])
+
+
 @pytest.fixture
 def na_value():
     return np.nan
diff --git a/pandas/tests/extension/test_numpy.py b/pandas/tests/extension/test_numpy.py
index f31fa5b87cfe5..8b632f0c1d184 100644
--- a/pandas/tests/extension/test_numpy.py
+++ b/pandas/tests/extension/test_numpy.py
@@ -101,6 +101,23 @@ def data_missing_for_sorting(allow_in_pandas, dtype):
     )
 
 
+@pytest.fixture
+def data_multiple_nan(allow_in_pandas, dtype):
+    """Length-10 array with a known sort order.
+
+    This should be three items [B, NA, A] with
+    A < B and NA missing.
+    """
+    if dtype.numpy_dtype == 'object':
+        return PandasArray(
+            np.array([(5,), (4,), np.nan, (1,), np.nan,
+                      np.nan, (6,), (3,), np.nan, (2,)])
+        )
+    return PandasArray(
+        np.array([5, 4, np.nan, 1, np.nan, np.nan, 6, 3, np.nan, 2])
+    )
+
+
 @pytest.fixture
 def data_for_grouping(allow_in_pandas, dtype):
     """Data for factorization, grouping, and unique tests.
diff --git a/pandas/tests/extension/test_period.py b/pandas/tests/extension/test_period.py
index b988dcb211dd0..b289e318b9050 100644
--- a/pandas/tests/extension/test_period.py
+++ b/pandas/tests/extension/test_period.py
@@ -40,6 +40,12 @@ def data_missing_for_sorting(dtype):
     return PeriodArray([2018, iNaT, 2017], freq=dtype.freq)
 
 
+@pytest.fixture
+def data_multiple_nan(dtype):
+    return PeriodArray([2005, 2004, iNaT, 2001, iNaT, iNaT,
+                        2006, 2003, iNaT, 2002], freq=dtype.freq)
+
+
 @pytest.fixture
 def data_for_grouping(dtype):
     B = 2018
diff --git a/pandas/tests/extension/test_sparse.py b/pandas/tests/extension/test_sparse.py
index faf1905ea1763..df1b87868c28d 100644
--- a/pandas/tests/extension/test_sparse.py
+++ b/pandas/tests/extension/test_sparse.py
@@ -65,6 +65,12 @@ def data_missing_for_sorting(request):
     return SparseArray([2, np.nan, 1], fill_value=request.param)
 
 
+@pytest.fixture(params=[0, np.nan])
+def data_multiple_nan(request):
+    return SparseArray([5, 4, np.nan, 1, np.nan,
+                        np.nan, 6, 3, np.nan, 2], fill_value=request.param)
+
+
 @pytest.fixture
 def na_value():
     return np.nan

From 93ac5899cad7c96b1db419dbd3c55d7b55f49c11 Mon Sep 17 00:00:00 2001
From: makbigc <makszechun@gmail.com>
Date: Sun, 12 May 2019 13:44:47 +0800
Subject: [PATCH 02/22] Fix mypy typing and add annotation to the test

---
 pandas/core/arrays/integer.py          | 5 ++---
 pandas/tests/extension/base/methods.py | 1 +
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/pandas/core/arrays/integer.py b/pandas/core/arrays/integer.py
index 78dda4790a067..3ad4e5670cb0c 100644
--- a/pandas/core/arrays/integer.py
+++ b/pandas/core/arrays/integer.py
@@ -1,6 +1,6 @@
 import copy
 import sys
-from typing import Type
+from typing import Tuple, Type
 import warnings
 
 import numpy as np
@@ -512,8 +512,7 @@ def value_counts(self, dropna=True):
 
         return Series(array, index=index)
 
-    #def _values_for_argsort(self) -> np.ndarray:
-    def _values_for_argsort(self):
+    def _values_for_argsort(self) -> Tuple[np.ndarray, np.ndarray]:
         """Return values for sorting.
 
         Returns
diff --git a/pandas/tests/extension/base/methods.py b/pandas/tests/extension/base/methods.py
index be7346e66ed4c..60aa8b9aa18e1 100644
--- a/pandas/tests/extension/base/methods.py
+++ b/pandas/tests/extension/base/methods.py
@@ -45,6 +45,7 @@ def test_argsort_missing(self, data_missing_for_sorting):
         self.assert_series_equal(result, expected)
 
     def test_argsort_nan_loc(self, data_multiple_nan):
+        # GH 21801
         result = data_multiple_nan.argsort()
         expected = np.array([3, 9, 7, 1, 0, 6, 2, 4, 5, 8])
         tm.assert_numpy_array_equal(result, expected)

From b401cba0e9a9448816681f58625e8eecd152677f Mon Sep 17 00:00:00 2001
From: makbigc <makszechun@gmail.com>
Date: Sun, 12 May 2019 13:59:47 +0800
Subject: [PATCH 03/22] Add whatsnew note

---
 doc/source/whatsnew/v0.25.0.rst | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/doc/source/whatsnew/v0.25.0.rst b/doc/source/whatsnew/v0.25.0.rst
index 1fb9b5ae695a0..969eef201910f 100644
--- a/doc/source/whatsnew/v0.25.0.rst
+++ b/doc/source/whatsnew/v0.25.0.rst
@@ -82,7 +82,8 @@ Other Enhancements
 - :meth:`DataFrame.query` and :meth:`DataFrame.eval` now supports quoting column names with backticks to refer to names with spaces (:issue:`6508`)
 - :func:`merge_asof` now gives a more clear error message when merge keys are categoricals that are not equal (:issue:`26136`)
 - :meth:`pandas.core.window.Rolling` supports exponential (or Poisson) window type (:issue:`21303`)
--
+- :meth:`ExtensionArray.argsort` places NA values at the end of the sorted array (:issue:`21801`)
+>>>>>>> Add whatsnew note
 
 .. _whatsnew_0250.api_breaking:
 

From aa50eed89ca3a178bbdda7eb42ae2652d451073a Mon Sep 17 00:00:00 2001
From: makbigc <makszechun@gmail.com>
Date: Sun, 12 May 2019 16:06:04 +0800
Subject: [PATCH 04/22] Modify nargsort

---
 pandas/core/sorting.py | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

diff --git a/pandas/core/sorting.py b/pandas/core/sorting.py
index 21c0c8f747b10..a3631576978a1 100644
--- a/pandas/core/sorting.py
+++ b/pandas/core/sorting.py
@@ -247,12 +247,11 @@ def nargsort(items, kind='quicksort', ascending=True, na_position='last'):
         mask = isna(items)
         cnt_null = mask.sum()
         sorted_idx = items.argsort(ascending=ascending, kind=kind)
-        if ascending and na_position == 'last':
-            # NaN is coded as -1 and is listed in front after sorting
-            sorted_idx = np.roll(sorted_idx, -cnt_null)
-        elif not ascending and na_position == 'first':
-            # NaN is coded as -1 and is listed in the end after sorting
+
+        if ascending and na_position == 'first':
             sorted_idx = np.roll(sorted_idx, cnt_null)
+        elif not ascending and na_position == 'last':
+            sorted_idx = np.roll(sorted_idx, -cnt_null)
         return sorted_idx
 
     with warnings.catch_warnings():

From d38a5e1ecf8273854380a922e6988ecef84d2e21 Mon Sep 17 00:00:00 2001
From: makbigc <makszechun@gmail.com>
Date: Thu, 16 May 2019 19:42:51 +0800
Subject: [PATCH 05/22] Simplify the subfunction permutation

---
 pandas/core/arrays/base.py | 6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)

diff --git a/pandas/core/arrays/base.py b/pandas/core/arrays/base.py
index c0e88bf6cfc1c..a1385f74adc61 100644
--- a/pandas/core/arrays/base.py
+++ b/pandas/core/arrays/base.py
@@ -412,11 +412,7 @@ def permutation(mask):
             # Return a permutation which maps the indices of the
             # subarray without nan to the indices of the original array.
             permu = np.arange(len(mask))
-            nan_loc = np.arange(len(mask))[mask]
-            offset = 0
-            for x in nan_loc:
-                permu[x - offset:] += 1
-                offset += 1
+            permu = permu[~mask]
             return permu
 
         if mask.any():

From bae30fd7e8ba2363996b5823269d39e9b6443c0a Mon Sep 17 00:00:00 2001
From: makbigc <makszechun@gmail.com>
Date: Thu, 16 May 2019 20:01:06 +0800
Subject: [PATCH 06/22] Remove test_argsort_nan_loc

---
 pandas/tests/extension/test_categorical.py |  6 ------
 pandas/tests/extension/test_datetime.py    |  9 ---------
 pandas/tests/extension/test_integer.py     |  6 ------
 pandas/tests/extension/test_interval.py    |  7 -------
 pandas/tests/extension/test_numpy.py       | 17 -----------------
 pandas/tests/extension/test_period.py      |  6 ------
 pandas/tests/extension/test_sparse.py      |  6 ------
 7 files changed, 57 deletions(-)

diff --git a/pandas/tests/extension/test_categorical.py b/pandas/tests/extension/test_categorical.py
index 6105db64d8f19..4cf9f78e1531d 100644
--- a/pandas/tests/extension/test_categorical.py
+++ b/pandas/tests/extension/test_categorical.py
@@ -68,12 +68,6 @@ def data_missing_for_sorting():
                        ordered=True)
 
 
-@pytest.fixture
-def data_multiple_nan():
-    return Categorical(['E', 'D', None, 'A', None,
-                        None, 'F', 'C', None, 'B'], ordered=True)
-
-
 @pytest.fixture
 def na_value():
     return np.nan
diff --git a/pandas/tests/extension/test_datetime.py b/pandas/tests/extension/test_datetime.py
index 600410e8864cb..baee04c3b79eb 100644
--- a/pandas/tests/extension/test_datetime.py
+++ b/pandas/tests/extension/test_datetime.py
@@ -45,15 +45,6 @@ def data_missing_for_sorting(dtype):
                          dtype=dtype)
 
 
-@pytest.fixture
-def data_multiple_nan(dtype):
-    idx = pd.date_range(start='2000-01-01', end='2000-01-06')
-    return DatetimeArray(np.array([idx[4], idx[3], 'NaT', idx[0], 'NaT',
-                                   'NaT', idx[5], idx[2], 'NaT', idx[1]],
-                                  dtype='datetime64[ns]'),
-                         dtype=dtype)
-
-
 @pytest.fixture
 def data_for_grouping(dtype):
     """
diff --git a/pandas/tests/extension/test_integer.py b/pandas/tests/extension/test_integer.py
index 09295138c74ca..22bb086a919ca 100644
--- a/pandas/tests/extension/test_integer.py
+++ b/pandas/tests/extension/test_integer.py
@@ -57,12 +57,6 @@ def data_for_sorting(dtype):
     return integer_array([1, 2, 0], dtype=dtype)
 
 
-@pytest.fixture
-def data_multiple_nan(dtype):
-    return integer_array([5, 4, np.nan, 1, np.nan,
-                          np.nan, 6, 3, np.nan, 2], dtype=dtype)
-
-
 @pytest.fixture
 def data_missing_for_sorting(dtype):
     return integer_array([1, np.nan, 0], dtype=dtype)
diff --git a/pandas/tests/extension/test_interval.py b/pandas/tests/extension/test_interval.py
index ff322c7db0ed3..f1f90b298ffe2 100644
--- a/pandas/tests/extension/test_interval.py
+++ b/pandas/tests/extension/test_interval.py
@@ -57,13 +57,6 @@ def data_missing_for_sorting():
     return IntervalArray.from_tuples([(1, 2), None, (0, 1)])
 
 
-@pytest.fixture
-def data_multiple_nan():
-    return IntervalArray.from_tuples([(5, 6), (4, 5), None, (1, 2),
-                                      None, None, (6, 7), (3, 4),
-                                      None, (2, 3)])
-
-
 @pytest.fixture
 def na_value():
     return np.nan
diff --git a/pandas/tests/extension/test_numpy.py b/pandas/tests/extension/test_numpy.py
index 8b632f0c1d184..f31fa5b87cfe5 100644
--- a/pandas/tests/extension/test_numpy.py
+++ b/pandas/tests/extension/test_numpy.py
@@ -101,23 +101,6 @@ def data_missing_for_sorting(allow_in_pandas, dtype):
     )
 
 
-@pytest.fixture
-def data_multiple_nan(allow_in_pandas, dtype):
-    """Length-10 array with a known sort order.
-
-    This should be three items [B, NA, A] with
-    A < B and NA missing.
-    """
-    if dtype.numpy_dtype == 'object':
-        return PandasArray(
-            np.array([(5,), (4,), np.nan, (1,), np.nan,
-                      np.nan, (6,), (3,), np.nan, (2,)])
-        )
-    return PandasArray(
-        np.array([5, 4, np.nan, 1, np.nan, np.nan, 6, 3, np.nan, 2])
-    )
-
-
 @pytest.fixture
 def data_for_grouping(allow_in_pandas, dtype):
     """Data for factorization, grouping, and unique tests.
diff --git a/pandas/tests/extension/test_period.py b/pandas/tests/extension/test_period.py
index b289e318b9050..b988dcb211dd0 100644
--- a/pandas/tests/extension/test_period.py
+++ b/pandas/tests/extension/test_period.py
@@ -40,12 +40,6 @@ def data_missing_for_sorting(dtype):
     return PeriodArray([2018, iNaT, 2017], freq=dtype.freq)
 
 
-@pytest.fixture
-def data_multiple_nan(dtype):
-    return PeriodArray([2005, 2004, iNaT, 2001, iNaT, iNaT,
-                        2006, 2003, iNaT, 2002], freq=dtype.freq)
-
-
 @pytest.fixture
 def data_for_grouping(dtype):
     B = 2018
diff --git a/pandas/tests/extension/test_sparse.py b/pandas/tests/extension/test_sparse.py
index df1b87868c28d..faf1905ea1763 100644
--- a/pandas/tests/extension/test_sparse.py
+++ b/pandas/tests/extension/test_sparse.py
@@ -65,12 +65,6 @@ def data_missing_for_sorting(request):
     return SparseArray([2, np.nan, 1], fill_value=request.param)
 
 
-@pytest.fixture(params=[0, np.nan])
-def data_multiple_nan(request):
-    return SparseArray([5, 4, np.nan, 1, np.nan,
-                        np.nan, 6, 3, np.nan, 2], fill_value=request.param)
-
-
 @pytest.fixture
 def na_value():
     return np.nan

From effe18c4df62e36710ba23b3282f4fbc89d9d98a Mon Sep 17 00:00:00 2001
From: makbigc <makszechun@gmail.com>
Date: Thu, 16 May 2019 20:09:51 +0800
Subject: [PATCH 07/22] Change the docstring of _values_for_argsort

---
 pandas/core/arrays/base.py    | 2 ++
 pandas/core/arrays/integer.py | 2 ++
 2 files changed, 4 insertions(+)

diff --git a/pandas/core/arrays/base.py b/pandas/core/arrays/base.py
index a1385f74adc61..2a884a65d0fe6 100644
--- a/pandas/core/arrays/base.py
+++ b/pandas/core/arrays/base.py
@@ -370,6 +370,8 @@ def _values_for_argsort(self) -> Tuple[np.ndarray, np.ndarray]:
         ndarray
             The transformed values should maintain the ordering between values
             within the array.
+        ndarray
+            The mask which indicates the NaN values.
 
         See Also
         --------
diff --git a/pandas/core/arrays/integer.py b/pandas/core/arrays/integer.py
index 3ad4e5670cb0c..8395b800ed0fa 100644
--- a/pandas/core/arrays/integer.py
+++ b/pandas/core/arrays/integer.py
@@ -520,6 +520,8 @@ def _values_for_argsort(self) -> Tuple[np.ndarray, np.ndarray]:
         ndarray
             The transformed values should maintain the ordering between values
             within the array.
+        ndarray
+            The mask which indicates the NaN values.
 
         See Also
         --------

From d1dd5fa272ea05a7520b5be9fe6b52ec8827614d Mon Sep 17 00:00:00 2001
From: makbigc <makszechun@gmail.com>
Date: Thu, 16 May 2019 20:17:44 +0800
Subject: [PATCH 08/22] Remove the code related to test_argsort_nan_loc

---
 pandas/tests/extension/base/methods.py         | 6 ------
 pandas/tests/extension/decimal/test_decimal.py | 6 ------
 pandas/tests/extension/json/test_json.py       | 6 ------
 3 files changed, 18 deletions(-)

diff --git a/pandas/tests/extension/base/methods.py b/pandas/tests/extension/base/methods.py
index 60aa8b9aa18e1..1852edaa9e748 100644
--- a/pandas/tests/extension/base/methods.py
+++ b/pandas/tests/extension/base/methods.py
@@ -44,12 +44,6 @@ def test_argsort_missing(self, data_missing_for_sorting):
         expected = pd.Series(np.array([1, -1, 0], dtype=np.int64))
         self.assert_series_equal(result, expected)
 
-    def test_argsort_nan_loc(self, data_multiple_nan):
-        # GH 21801
-        result = data_multiple_nan.argsort()
-        expected = np.array([3, 9, 7, 1, 0, 6, 2, 4, 5, 8])
-        tm.assert_numpy_array_equal(result, expected)
-
     @pytest.mark.parametrize('ascending', [True, False])
     def test_sort_values(self, data_for_sorting, ascending):
         ser = pd.Series(data_for_sorting)
diff --git a/pandas/tests/extension/decimal/test_decimal.py b/pandas/tests/extension/decimal/test_decimal.py
index ba68a8edc947c..97fae41bcc720 100644
--- a/pandas/tests/extension/decimal/test_decimal.py
+++ b/pandas/tests/extension/decimal/test_decimal.py
@@ -46,12 +46,6 @@ def data_missing_for_sorting():
                          decimal.Decimal('0')])
 
 
-@pytest.fixture
-def data_multiple_nan():
-    return DecimalArray([decimal.Decimal(x) for x in
-                        [5, 4, np.nan, 1, np.nan, np.nan, 6, 3, np.nan, 2]])
-
-
 @pytest.fixture
 def na_cmp():
     return lambda x, y: x.is_nan() and y.is_nan()
diff --git a/pandas/tests/extension/json/test_json.py b/pandas/tests/extension/json/test_json.py
index 47b7fd670b1da..97c329e0a5c92 100644
--- a/pandas/tests/extension/json/test_json.py
+++ b/pandas/tests/extension/json/test_json.py
@@ -50,12 +50,6 @@ def data_missing_for_sorting():
     return JSONArray([{'b': 1}, {}, {'a': 4}])
 
 
-@pytest.fixture
-def data_multiple_nan():
-    return JSONArray([{'e': 5}, {'d': 4}, {}, {'a': 1}, {},
-                      {}, {'f': 6}, {'c': 3}, {}, {'b': 2}])
-
-
 @pytest.fixture
 def na_value(dtype):
     return dtype.na_value

From bf1e927cf318d8af10e0c6e30794a163e585d872 Mon Sep 17 00:00:00 2001
From: makbigc <makszechun@gmail.com>
Date: Tue, 21 May 2019 12:46:02 +0800
Subject: [PATCH 09/22] Change docstring and move the entry in whatsnew

---
 doc/source/whatsnew/v0.25.0.rst | 6 +++++-
 pandas/core/arrays/base.py      | 2 +-
 pandas/core/arrays/integer.py   | 2 +-
 3 files changed, 7 insertions(+), 3 deletions(-)

diff --git a/doc/source/whatsnew/v0.25.0.rst b/doc/source/whatsnew/v0.25.0.rst
index 969eef201910f..8f5072a2db3ea 100644
--- a/doc/source/whatsnew/v0.25.0.rst
+++ b/doc/source/whatsnew/v0.25.0.rst
@@ -82,8 +82,8 @@ Other Enhancements
 - :meth:`DataFrame.query` and :meth:`DataFrame.eval` now supports quoting column names with backticks to refer to names with spaces (:issue:`6508`)
 - :func:`merge_asof` now gives a more clear error message when merge keys are categoricals that are not equal (:issue:`26136`)
 - :meth:`pandas.core.window.Rolling` supports exponential (or Poisson) window type (:issue:`21303`)
+<<<<<<< HEAD
 - :meth:`ExtensionArray.argsort` places NA values at the end of the sorted array (:issue:`21801`)
->>>>>>> Add whatsnew note
 
 .. _whatsnew_0250.api_breaking:
 
@@ -435,10 +435,14 @@ Other API Changes
 - Comparing :class:`Timestamp` with unsupported objects now returns :py:obj:`NotImplemented` instead of raising ``TypeError``. This implies that unsupported rich comparisons are delegated to the other object, and are now consistent with Python 3 behavior for ``datetime`` objects (:issue:`24011`)
 - Bug in :meth:`DatetimeIndex.snap` which didn't preserving the ``name`` of the input :class:`Index` (:issue:`25575`)
 - The ``arg`` argument in :meth:`pandas.core.groupby.DataFrameGroupBy.agg` has been renamed to ``func`` (:issue:`26089`)
+<<<<<<< HEAD
 - The ``arg`` argument in :meth:`pandas.core.window._Window.aggregate` has been renamed to ``func`` (:issue:`26372`)
 - Most Pandas classes had a ``__bytes__`` method, which was used for getting a python2-style bytestring representation of the object. This method has been removed as a part of dropping Python2 (:issue:`26447`)
 - The ``.str``-accessor has been disabled for 1-level :class:`MultiIndex`, use :meth:`MultiIndex.to_flat_index` if necessary (:issue:`23679`)
 - Removed support of gtk package for clipboards (:issue:`26563`)
+=======
+- :meth:`ExtensionArray.argsort` places the nan at the end of the sorted array (:issue:`21801`)
+>>>>>>> Change docstring and move the entry in whatsnew
 
 .. _whatsnew_0250.deprecations:
 
diff --git a/pandas/core/arrays/base.py b/pandas/core/arrays/base.py
index 2a884a65d0fe6..25626985e9b83 100644
--- a/pandas/core/arrays/base.py
+++ b/pandas/core/arrays/base.py
@@ -371,7 +371,7 @@ def _values_for_argsort(self) -> Tuple[np.ndarray, np.ndarray]:
             The transformed values should maintain the ordering between values
             within the array.
         ndarray
-            The mask which indicates the NaN values.
+            The mask which indicates the NA values.
 
         See Also
         --------
diff --git a/pandas/core/arrays/integer.py b/pandas/core/arrays/integer.py
index 8395b800ed0fa..ac1c99b05efcc 100644
--- a/pandas/core/arrays/integer.py
+++ b/pandas/core/arrays/integer.py
@@ -521,7 +521,7 @@ def _values_for_argsort(self) -> Tuple[np.ndarray, np.ndarray]:
             The transformed values should maintain the ordering between values
             within the array.
         ndarray
-            The mask which indicates the NaN values.
+            The mask which indicates the NA values.
 
         See Also
         --------

From 19254233f3ae6d8d3f2e3cffa9658b4724a8d446 Mon Sep 17 00:00:00 2001
From: makbigc <makszechun@gmail.com>
Date: Sun, 26 May 2019 21:36:51 +0800
Subject: [PATCH 10/22] amend after review

---
 doc/source/whatsnew/v0.25.0.rst |  4 ++++
 pandas/core/arrays/base.py      | 16 ++++++++--------
 pandas/core/arrays/integer.py   |  2 ++
 3 files changed, 14 insertions(+), 8 deletions(-)

diff --git a/doc/source/whatsnew/v0.25.0.rst b/doc/source/whatsnew/v0.25.0.rst
index 8f5072a2db3ea..47150e3e1d6a2 100644
--- a/doc/source/whatsnew/v0.25.0.rst
+++ b/doc/source/whatsnew/v0.25.0.rst
@@ -438,11 +438,15 @@ Other API Changes
 <<<<<<< HEAD
 - The ``arg`` argument in :meth:`pandas.core.window._Window.aggregate` has been renamed to ``func`` (:issue:`26372`)
 - Most Pandas classes had a ``__bytes__`` method, which was used for getting a python2-style bytestring representation of the object. This method has been removed as a part of dropping Python2 (:issue:`26447`)
+<<<<<<< HEAD
 - The ``.str``-accessor has been disabled for 1-level :class:`MultiIndex`, use :meth:`MultiIndex.to_flat_index` if necessary (:issue:`23679`)
 - Removed support of gtk package for clipboards (:issue:`26563`)
 =======
 - :meth:`ExtensionArray.argsort` places the nan at the end of the sorted array (:issue:`21801`)
 >>>>>>> Change docstring and move the entry in whatsnew
+=======
+- :meth:`ExtensionArray.argsort` places the ``nan`` at the end of the sorted array (:issue:`21801`)
+>>>>>>> amend after review
 
 .. _whatsnew_0250.deprecations:
 
diff --git a/pandas/core/arrays/base.py b/pandas/core/arrays/base.py
index 25626985e9b83..66affb187a13a 100644
--- a/pandas/core/arrays/base.py
+++ b/pandas/core/arrays/base.py
@@ -373,6 +373,8 @@ def _values_for_argsort(self) -> Tuple[np.ndarray, np.ndarray]:
         ndarray
             The mask which indicates the NA values.
 
+            .. versionadded:: 0.25.0
+
         See Also
         --------
         ExtensionArray.argsort
@@ -410,17 +412,15 @@ def argsort(self, ascending=True, kind='quicksort', *args, **kwargs):
         ascending = nv.validate_argsort_with_ascending(ascending, args, kwargs)
         values, mask = self._values_for_argsort()
 
-        def permutation(mask):
-            # Return a permutation which maps the indices of the
-            # subarray without nan to the indices of the original array.
-            permu = np.arange(len(mask))
-            permu = permu[~mask]
-            return permu
-
         if mask.any():
             notmask = ~mask
             notnull = np.argsort(values[notmask], kind=kind, **kwargs)
-            permu = permutation(mask)
+
+            # permu maps the indices of the subarray
+            # without nan to the indices of the original array.
+            permu = np.arange(len(mask))
+            permu = permu[~mask]
+
             notnull = permu[notnull]
             allnan = np.arange(len(self))[mask]
             result = np.append(notnull, allnan)
diff --git a/pandas/core/arrays/integer.py b/pandas/core/arrays/integer.py
index ac1c99b05efcc..02e16840b95bf 100644
--- a/pandas/core/arrays/integer.py
+++ b/pandas/core/arrays/integer.py
@@ -523,6 +523,8 @@ def _values_for_argsort(self) -> Tuple[np.ndarray, np.ndarray]:
         ndarray
             The mask which indicates the NA values.
 
+            .. versionadded:: 0.25.0
+
         See Also
         --------
         ExtensionArray.argsort

From c57dff7d3063a5a3db877871621b6cc824d27514 Mon Sep 17 00:00:00 2001
From: makbigc <makszechun@gmail.com>
Date: Thu, 6 Jun 2019 14:51:59 +0800
Subject: [PATCH 11/22] Use sorting.nargsort and remove some
 _values_for_argsort

---
 pandas/core/arrays/base.py           | 45 +++++-----------------------
 pandas/core/arrays/categorical.py    | 12 +++++---
 pandas/core/arrays/datetimelike.py   |  3 --
 pandas/core/arrays/integer.py        | 22 --------------
 pandas/core/arrays/numpy_.py         |  3 --
 pandas/core/arrays/period.py         |  3 --
 pandas/core/sorting.py               |  9 +++---
 pandas/tests/extension/json/array.py |  4 +--
 8 files changed, 22 insertions(+), 79 deletions(-)

diff --git a/pandas/core/arrays/base.py b/pandas/core/arrays/base.py
index 66affb187a13a..61ed936ace2a2 100644
--- a/pandas/core/arrays/base.py
+++ b/pandas/core/arrays/base.py
@@ -23,6 +23,7 @@
 
 from pandas._typing import ArrayLike
 from pandas.core import ops
+from pandas.core.sorting import nargsort
 
 _not_implemented_message = "{} does not implement {}."
 
@@ -361,27 +362,6 @@ def isna(self) -> ArrayLike:
         """
         raise AbstractMethodError(self)
 
-    def _values_for_argsort(self) -> Tuple[np.ndarray, np.ndarray]:
-        """
-        Return values for sorting.
-
-        Returns
-        -------
-        ndarray
-            The transformed values should maintain the ordering between values
-            within the array.
-        ndarray
-            The mask which indicates the NA values.
-
-            .. versionadded:: 0.25.0
-
-        See Also
-        --------
-        ExtensionArray.argsort
-        """
-        # Note: this is used in `ExtensionArray.argsort`.
-        return np.array(self), self.isna()
-
     def argsort(self, ascending=True, kind='quicksort', *args, **kwargs):
         """
         Return the indices that would sort this array.
@@ -410,25 +390,14 @@ def argsort(self, ascending=True, kind='quicksort', *args, **kwargs):
         # 1. _values_for_argsort : construct the values passed to np.argsort
         # 2. argsort : total control over sorting.
         ascending = nv.validate_argsort_with_ascending(ascending, args, kwargs)
-        values, mask = self._values_for_argsort()
 
-        if mask.any():
-            notmask = ~mask
-            notnull = np.argsort(values[notmask], kind=kind, **kwargs)
-
-            # permu maps the indices of the subarray
-            # without nan to the indices of the original array.
-            permu = np.arange(len(mask))
-            permu = permu[~mask]
-
-            notnull = permu[notnull]
-            allnan = np.arange(len(self))[mask]
-            result = np.append(notnull, allnan)
+        if hasattr(self, '_values_for_argsort'):
+            values = self._values_for_argsort()
         else:
-            result = np.argsort(values, kind=kind, **kwargs)
-
-        if not ascending:
-            result = result[::-1]
+            values = self
+        na_position = 'last' if ascending else 'first'
+        result = nargsort(values, kind=kind, ascending=ascending,
+                          na_position=na_position)
         return result
 
     def fillna(self, value=None, method=None, limit=None):
diff --git a/pandas/core/arrays/categorical.py b/pandas/core/arrays/categorical.py
index c4ce43b55741b..4785be000bcb6 100644
--- a/pandas/core/arrays/categorical.py
+++ b/pandas/core/arrays/categorical.py
@@ -1521,9 +1521,9 @@ def check_for_ordered(self, op):
                             "Categorical to an ordered one\n".format(op=op))
 
     def _values_for_argsort(self):
-        return self._codes.copy(), self.isna()
+        return self._codes.copy()
 
-    def argsort(self, *args, **kwargs):
+    def argsort(self, ascending=True, kind='quicksort', *args, **kwargs):
         # TODO(PY2): use correct signature
         # We have to do *args, **kwargs to avoid a a py2-only signature
         # issue since np.argsort differs from argsort.
@@ -1567,8 +1567,12 @@ def argsort(self, *args, **kwargs):
         >>> cat.argsort()
         array([3, 0, 1, 2])
         """
-        # Keep the implementation here just for the docstring.
-        return super().argsort(*args, **kwargs)
+        ascending = nv.validate_argsort_with_ascending(ascending, args, kwargs)
+        values = self._values_for_argsort()
+        result = np.argsort(values, kind=kind, **kwargs)
+        if not ascending:
+            result = result[::-1]
+        return result
 
     def sort_values(self, inplace=False, ascending=True, na_position='last'):
         """
diff --git a/pandas/core/arrays/datetimelike.py b/pandas/core/arrays/datetimelike.py
index a1696c3fd52da..e1a7eb625a5c7 100644
--- a/pandas/core/arrays/datetimelike.py
+++ b/pandas/core/arrays/datetimelike.py
@@ -620,9 +620,6 @@ def _values_for_factorize(self):
     def _from_factorized(cls, values, original):
         return cls(values, dtype=original.dtype)
 
-    def _values_for_argsort(self):
-        return self._data, self._isnan
-
     # ------------------------------------------------------------------
     # Additional array methods
     #  These are not part of the EA API, but we implement them because
diff --git a/pandas/core/arrays/integer.py b/pandas/core/arrays/integer.py
index 02e16840b95bf..3b856323348e6 100644
--- a/pandas/core/arrays/integer.py
+++ b/pandas/core/arrays/integer.py
@@ -512,28 +512,6 @@ def value_counts(self, dropna=True):
 
         return Series(array, index=index)
 
-    def _values_for_argsort(self) -> Tuple[np.ndarray, np.ndarray]:
-        """Return values for sorting.
-
-        Returns
-        -------
-        ndarray
-            The transformed values should maintain the ordering between values
-            within the array.
-        ndarray
-            The mask which indicates the NA values.
-
-            .. versionadded:: 0.25.0
-
-        See Also
-        --------
-        ExtensionArray.argsort
-        """
-        data = self._data.copy()
-        mask = self._mask
-        data[self._mask] = data.min() - 1
-        return data, mask
-
     @classmethod
     def _create_comparison_method(cls, op):
         def cmp_method(self, other):
diff --git a/pandas/core/arrays/numpy_.py b/pandas/core/arrays/numpy_.py
index ab71d45bdbaf0..92cf7e9425cd3 100644
--- a/pandas/core/arrays/numpy_.py
+++ b/pandas/core/arrays/numpy_.py
@@ -280,9 +280,6 @@ def take(self, indices, allow_fill=False, fill_value=None):
     def copy(self, deep=False):
         return type(self)(self._ndarray.copy())
 
-    def _values_for_argsort(self):
-        return self._ndarray, self.isna()
-
     def _values_for_factorize(self):
         return self._ndarray, -1
 
diff --git a/pandas/core/arrays/period.py b/pandas/core/arrays/period.py
index 8f1da6dbe5c4a..0d775bdec5a7b 100644
--- a/pandas/core/arrays/period.py
+++ b/pandas/core/arrays/period.py
@@ -673,9 +673,6 @@ def _check_timedeltalike_freq_compat(self, other):
 
         _raise_on_incompatible(self, other)
 
-    def _values_for_argsort(self):
-        return self._data, self._isnan
-
 
 PeriodArray._add_comparison_ops()
 
diff --git a/pandas/core/sorting.py b/pandas/core/sorting.py
index a3631576978a1..21c0c8f747b10 100644
--- a/pandas/core/sorting.py
+++ b/pandas/core/sorting.py
@@ -247,11 +247,12 @@ def nargsort(items, kind='quicksort', ascending=True, na_position='last'):
         mask = isna(items)
         cnt_null = mask.sum()
         sorted_idx = items.argsort(ascending=ascending, kind=kind)
-
-        if ascending and na_position == 'first':
-            sorted_idx = np.roll(sorted_idx, cnt_null)
-        elif not ascending and na_position == 'last':
+        if ascending and na_position == 'last':
+            # NaN is coded as -1 and is listed in front after sorting
             sorted_idx = np.roll(sorted_idx, -cnt_null)
+        elif not ascending and na_position == 'first':
+            # NaN is coded as -1 and is listed in the end after sorting
+            sorted_idx = np.roll(sorted_idx, cnt_null)
         return sorted_idx
 
     with warnings.catch_warnings():
diff --git a/pandas/tests/extension/json/array.py b/pandas/tests/extension/json/array.py
index dde88a4838e6a..4b93f0e12e32a 100644
--- a/pandas/tests/extension/json/array.py
+++ b/pandas/tests/extension/json/array.py
@@ -171,7 +171,7 @@ def _concat_same_type(cls, to_concat):
         return cls(data)
 
     def _values_for_factorize(self):
-        frozen, _ = self._values_for_argsort()
+        frozen = self._values_for_argsort()
         if len(frozen) == 0:
             # _factorize_array expects 1-d array, this is a len-0 2-d array.
             frozen = frozen.ravel()
@@ -182,7 +182,7 @@ def _values_for_argsort(self):
         # If all the elemnts of self are the same size P, NumPy will
         # cast them to an (N, P) array, instead of an (N,) array of tuples.
         frozen = [()] + [tuple(x.items()) for x in self]
-        return np.array(frozen, dtype=object)[1:], self.isna()
+        return np.array(frozen, dtype=object)[1:]
 
 
 def make_data():

From 16cff4e0cdf95c68d82425b4bb0dd78b23d99a68 Mon Sep 17 00:00:00 2001
From: makbigc <makszechun@gmail.com>
Date: Thu, 6 Jun 2019 14:56:10 +0800
Subject: [PATCH 12/22] Fix lint error

---
 pandas/core/arrays/integer.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/pandas/core/arrays/integer.py b/pandas/core/arrays/integer.py
index 3b856323348e6..3c4ae83b31bdc 100644
--- a/pandas/core/arrays/integer.py
+++ b/pandas/core/arrays/integer.py
@@ -1,6 +1,6 @@
 import copy
 import sys
-from typing import Tuple, Type
+from typing import Type
 import warnings
 
 import numpy as np

From bc713cd4f1efd69a7ca20b263be40db261aaa511 Mon Sep 17 00:00:00 2001
From: makbigc <makszechun@gmail.com>
Date: Thu, 6 Jun 2019 15:09:41 +0800
Subject: [PATCH 13/22] remove relundent lines added in rebase

---
 doc/source/whatsnew/v0.25.0.rst | 11 +----------
 1 file changed, 1 insertion(+), 10 deletions(-)

diff --git a/doc/source/whatsnew/v0.25.0.rst b/doc/source/whatsnew/v0.25.0.rst
index 47150e3e1d6a2..4f45e4554487a 100644
--- a/doc/source/whatsnew/v0.25.0.rst
+++ b/doc/source/whatsnew/v0.25.0.rst
@@ -82,8 +82,6 @@ Other Enhancements
 - :meth:`DataFrame.query` and :meth:`DataFrame.eval` now supports quoting column names with backticks to refer to names with spaces (:issue:`6508`)
 - :func:`merge_asof` now gives a more clear error message when merge keys are categoricals that are not equal (:issue:`26136`)
 - :meth:`pandas.core.window.Rolling` supports exponential (or Poisson) window type (:issue:`21303`)
-<<<<<<< HEAD
-- :meth:`ExtensionArray.argsort` places NA values at the end of the sorted array (:issue:`21801`)
 
 .. _whatsnew_0250.api_breaking:
 
@@ -435,18 +433,11 @@ Other API Changes
 - Comparing :class:`Timestamp` with unsupported objects now returns :py:obj:`NotImplemented` instead of raising ``TypeError``. This implies that unsupported rich comparisons are delegated to the other object, and are now consistent with Python 3 behavior for ``datetime`` objects (:issue:`24011`)
 - Bug in :meth:`DatetimeIndex.snap` which didn't preserving the ``name`` of the input :class:`Index` (:issue:`25575`)
 - The ``arg`` argument in :meth:`pandas.core.groupby.DataFrameGroupBy.agg` has been renamed to ``func`` (:issue:`26089`)
-<<<<<<< HEAD
 - The ``arg`` argument in :meth:`pandas.core.window._Window.aggregate` has been renamed to ``func`` (:issue:`26372`)
 - Most Pandas classes had a ``__bytes__`` method, which was used for getting a python2-style bytestring representation of the object. This method has been removed as a part of dropping Python2 (:issue:`26447`)
-<<<<<<< HEAD
 - The ``.str``-accessor has been disabled for 1-level :class:`MultiIndex`, use :meth:`MultiIndex.to_flat_index` if necessary (:issue:`23679`)
 - Removed support of gtk package for clipboards (:issue:`26563`)
-=======
-- :meth:`ExtensionArray.argsort` places the nan at the end of the sorted array (:issue:`21801`)
->>>>>>> Change docstring and move the entry in whatsnew
-=======
-- :meth:`ExtensionArray.argsort` places the ``nan`` at the end of the sorted array (:issue:`21801`)
->>>>>>> amend after review
+- :meth:`ExtensionArray.argsort` places NA values at the end of the sorted array (:issue:`21801`)
 
 .. _whatsnew_0250.deprecations:
 

From b10e5f9e9ae7971f813c3926ed153bb2d34ee1f3 Mon Sep 17 00:00:00 2001
From: makbigc <makszechun@gmail.com>
Date: Sat, 8 Jun 2019 17:18:22 +0800
Subject: [PATCH 14/22] Add test for NA-last behaviour

---
 pandas/core/arrays/base.py                 | 5 ++++-
 pandas/tests/extension/base/methods.py     | 5 +++++
 pandas/tests/extension/decimal/array.py    | 6 ++++++
 pandas/tests/extension/json/array.py       | 5 ++++-
 pandas/tests/extension/test_categorical.py | 5 +++++
 5 files changed, 24 insertions(+), 2 deletions(-)

diff --git a/pandas/core/arrays/base.py b/pandas/core/arrays/base.py
index 61ed936ace2a2..5610300f6f690 100644
--- a/pandas/core/arrays/base.py
+++ b/pandas/core/arrays/base.py
@@ -387,7 +387,10 @@ def argsort(self, ascending=True, kind='quicksort', *args, **kwargs):
         """
         # Implementor note: You have two places to override the behavior of
         # argsort.
-        # 1. _values_for_argsort : construct the values passed to np.argsort
+        # 1. _values_for_argsort : construct the values passed to np.argsort.
+        # The transformed values should maintain the ordering between values
+        # within the array. Preferably, the largest value is assiged to NA
+        # values in transformation.
         # 2. argsort : total control over sorting.
         ascending = nv.validate_argsort_with_ascending(ascending, args, kwargs)
 
diff --git a/pandas/tests/extension/base/methods.py b/pandas/tests/extension/base/methods.py
index 1852edaa9e748..50fd4cd292641 100644
--- a/pandas/tests/extension/base/methods.py
+++ b/pandas/tests/extension/base/methods.py
@@ -44,6 +44,11 @@ def test_argsort_missing(self, data_missing_for_sorting):
         expected = pd.Series(np.array([1, -1, 0], dtype=np.int64))
         self.assert_series_equal(result, expected)
 
+    def test_argsort_nan_last(self, data_missing_for_sorting):
+        # GH 21801
+        result = data_missing_for_sorting.argsort()
+        assert result[-1] == 1
+
     @pytest.mark.parametrize('ascending', [True, False])
     def test_sort_values(self, data_for_sorting, ascending):
         ser = pd.Series(data_for_sorting)
diff --git a/pandas/tests/extension/decimal/array.py b/pandas/tests/extension/decimal/array.py
index 1823eeb4d7fc0..ef0e44ec3c01c 100644
--- a/pandas/tests/extension/decimal/array.py
+++ b/pandas/tests/extension/decimal/array.py
@@ -153,6 +153,12 @@ def _reduce(self, name, skipna=True, **kwargs):
                                       "the {} operation".format(name))
         return op(axis=0)
 
+    def _values_for_argsort(self):
+        data = self._data
+        mask = self.isna()
+        data[mask] = decimal.Decimal('Infinity')
+        return data
+
 
 def to_decimal(values, context=None):
     return DecimalArray([decimal.Decimal(x) for x in values], context=context)
diff --git a/pandas/tests/extension/json/array.py b/pandas/tests/extension/json/array.py
index 4b93f0e12e32a..12713a8565228 100644
--- a/pandas/tests/extension/json/array.py
+++ b/pandas/tests/extension/json/array.py
@@ -182,7 +182,10 @@ def _values_for_argsort(self):
         # If all the elemnts of self are the same size P, NumPy will
         # cast them to an (N, P) array, instead of an (N,) array of tuples.
         frozen = [()] + [tuple(x.items()) for x in self]
-        return np.array(frozen, dtype=object)[1:]
+        data = np.array(frozen, dtype=object)[1:]
+        mask = self.isna()
+        data[mask] = np.nan
+        return data
 
 
 def make_data():
diff --git a/pandas/tests/extension/test_categorical.py b/pandas/tests/extension/test_categorical.py
index 4cf9f78e1531d..a169a08fa0d51 100644
--- a/pandas/tests/extension/test_categorical.py
+++ b/pandas/tests/extension/test_categorical.py
@@ -193,6 +193,11 @@ def test_searchsorted(self, data_for_sorting):
         if not data_for_sorting.ordered:
             raise pytest.skip(reason="searchsorted requires ordered data.")
 
+    def test_argsort_nan_last(self, data_missing_for_sorting):
+        # GH 21801
+        # TODO: Categorical.argsort places NA values at the end
+        pass
+
 
 class TestCasting(base.BaseCastingTests):
     pass

From 7811a081ae162ab4ecb94390a7d6ce55b6b6fcd5 Mon Sep 17 00:00:00 2001
From: makbigc <makszechun@gmail.com>
Date: Sun, 9 Jun 2019 12:32:13 +0800
Subject: [PATCH 15/22] Add _values_for_argsort in core/arrays/base.py

---
 pandas/core/arrays/base.py | 23 +++++++++++++++++++----
 1 file changed, 19 insertions(+), 4 deletions(-)

diff --git a/pandas/core/arrays/base.py b/pandas/core/arrays/base.py
index 5610300f6f690..6d0c907c22198 100644
--- a/pandas/core/arrays/base.py
+++ b/pandas/core/arrays/base.py
@@ -362,6 +362,21 @@ def isna(self) -> ArrayLike:
         """
         raise AbstractMethodError(self)
 
+    def _values_for_argsort(self) -> np.ndarray:
+        """
+        Return values for sorting.
+        Returns
+        -------
+        ndarray
+            The transformed values should maintain the ordering between values
+            within the array.
+        See Also
+        --------
+        ExtensionArray.argsort
+        """
+        # Note: this is used in `ExtensionArray.argsort`.
+        return np.array(self)
+
     def argsort(self, ascending=True, kind='quicksort', *args, **kwargs):
         """
         Return the indices that would sort this array.
@@ -394,10 +409,10 @@ def argsort(self, ascending=True, kind='quicksort', *args, **kwargs):
         # 2. argsort : total control over sorting.
         ascending = nv.validate_argsort_with_ascending(ascending, args, kwargs)
 
-        if hasattr(self, '_values_for_argsort'):
-            values = self._values_for_argsort()
-        else:
-            values = self
+        #if hasattr(self, '_values_for_argsort'):
+        values = self._values_for_argsort()
+        #else:
+            #values = self
         na_position = 'last' if ascending else 'first'
         result = nargsort(values, kind=kind, ascending=ascending,
                           na_position=na_position)

From c172ba1f586a5ff644ebcc965c11f1b3fd8d74ba Mon Sep 17 00:00:00 2001
From: makbigc <makszechun@gmail.com>
Date: Sun, 9 Jun 2019 12:38:43 +0800
Subject: [PATCH 16/22] Fix PEP8 error

---
 pandas/core/arrays/base.py | 6 ------
 1 file changed, 6 deletions(-)

diff --git a/pandas/core/arrays/base.py b/pandas/core/arrays/base.py
index 6d0c907c22198..2063dfdfe1917 100644
--- a/pandas/core/arrays/base.py
+++ b/pandas/core/arrays/base.py
@@ -403,16 +403,10 @@ def argsort(self, ascending=True, kind='quicksort', *args, **kwargs):
         # Implementor note: You have two places to override the behavior of
         # argsort.
         # 1. _values_for_argsort : construct the values passed to np.argsort.
-        # The transformed values should maintain the ordering between values
-        # within the array. Preferably, the largest value is assiged to NA
-        # values in transformation.
         # 2. argsort : total control over sorting.
         ascending = nv.validate_argsort_with_ascending(ascending, args, kwargs)
 
-        #if hasattr(self, '_values_for_argsort'):
         values = self._values_for_argsort()
-        #else:
-            #values = self
         na_position = 'last' if ascending else 'first'
         result = nargsort(values, kind=kind, ascending=ascending,
                           na_position=na_position)

From d9660b11f2acbb57552c92e4feca0c48e2daba12 Mon Sep 17 00:00:00 2001
From: makbigc <makszechun@gmail.com>
Date: Sun, 9 Jun 2019 12:59:34 +0800
Subject: [PATCH 17/22] Add one more test

---
 pandas/tests/test_sorting.py | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/pandas/tests/test_sorting.py b/pandas/tests/test_sorting.py
index 75fa37eb9af09..fb23c8822a000 100644
--- a/pandas/tests/test_sorting.py
+++ b/pandas/tests/test_sorting.py
@@ -7,7 +7,8 @@
 import pytest
 
 from pandas import (
-    DataFrame, MultiIndex, Series, array, concat, merge, to_datetime)
+    Categorical, DataFrame, MultiIndex, Series, array, concat, merge,
+    to_datetime)
 from pandas.core import common as com
 from pandas.core.sorting import (
     decons_group_index, get_group_index, is_int64_overflow_possible,
@@ -188,6 +189,14 @@ def test_nargsort_datetimearray_warning(self):
         with tm.assert_produces_warning(None):
             nargsort(data)
 
+    @pytest.mark.parametrize('data', [
+        Categorical(['a', 'c', 'a', 'b']),
+        to_datetime([0, 2, 0, 1]).tz_localize('Europe/Brussels')])
+    def test_nargsort_extension_array(self, data):
+        result = nargsort(data)
+        expected = np.array([0, 2, 3, 1], dtype=np.intp)
+        tm.assert_numpy_array_equal(result, expected)
+
 
 class TestMerge:
 

From c333e8c2a133f440f014caefd48de702b712254b Mon Sep 17 00:00:00 2001
From: makbigc <makszechun@gmail.com>
Date: Sun, 9 Jun 2019 22:17:51 +0800
Subject: [PATCH 18/22] Add _values_for_argsort back to numpy_.py for
 efficiency

---
 pandas/core/arrays/numpy_.py | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/pandas/core/arrays/numpy_.py b/pandas/core/arrays/numpy_.py
index 92cf7e9425cd3..6ae464f169a13 100644
--- a/pandas/core/arrays/numpy_.py
+++ b/pandas/core/arrays/numpy_.py
@@ -283,6 +283,9 @@ def copy(self, deep=False):
     def _values_for_factorize(self):
         return self._ndarray, -1
 
+    def _values_for_argsort(self):
+        return self._ndarray
+
     def unique(self):
         from pandas import unique
 

From 315e6f2a49d8dc14ad572a96dcb82727eeba0eff Mon Sep 17 00:00:00 2001
From: makbigc <makszechun@gmail.com>
Date: Mon, 10 Jun 2019 10:53:40 +0800
Subject: [PATCH 19/22] Set na_position last and revert _values_for_argsort in
 datetimelike.py

---
 pandas/core/arrays/base.py         | 3 +--
 pandas/core/arrays/datetimelike.py | 3 +++
 pandas/core/arrays/period.py       | 3 +++
 pandas/core/sorting.py             | 2 +-
 4 files changed, 8 insertions(+), 3 deletions(-)

diff --git a/pandas/core/arrays/base.py b/pandas/core/arrays/base.py
index 2063dfdfe1917..38c4fe9b2c007 100644
--- a/pandas/core/arrays/base.py
+++ b/pandas/core/arrays/base.py
@@ -407,9 +407,8 @@ def argsort(self, ascending=True, kind='quicksort', *args, **kwargs):
         ascending = nv.validate_argsort_with_ascending(ascending, args, kwargs)
 
         values = self._values_for_argsort()
-        na_position = 'last' if ascending else 'first'
         result = nargsort(values, kind=kind, ascending=ascending,
-                          na_position=na_position)
+                          na_position='last')
         return result
 
     def fillna(self, value=None, method=None, limit=None):
diff --git a/pandas/core/arrays/datetimelike.py b/pandas/core/arrays/datetimelike.py
index e1a7eb625a5c7..c32f8642dc2ed 100644
--- a/pandas/core/arrays/datetimelike.py
+++ b/pandas/core/arrays/datetimelike.py
@@ -620,6 +620,9 @@ def _values_for_factorize(self):
     def _from_factorized(cls, values, original):
         return cls(values, dtype=original.dtype)
 
+    def _values_for_argsort(self):
+        return self._data
+
     # ------------------------------------------------------------------
     # Additional array methods
     #  These are not part of the EA API, but we implement them because
diff --git a/pandas/core/arrays/period.py b/pandas/core/arrays/period.py
index 0d775bdec5a7b..dea8f3038ebd8 100644
--- a/pandas/core/arrays/period.py
+++ b/pandas/core/arrays/period.py
@@ -673,6 +673,9 @@ def _check_timedeltalike_freq_compat(self, other):
 
         _raise_on_incompatible(self, other)
 
+    def _values_for_argsort(self):
+        return self
+
 
 PeriodArray._add_comparison_ops()
 
diff --git a/pandas/core/sorting.py b/pandas/core/sorting.py
index 21c0c8f747b10..f4b71f69f481b 100644
--- a/pandas/core/sorting.py
+++ b/pandas/core/sorting.py
@@ -255,6 +255,7 @@ def nargsort(items, kind='quicksort', ascending=True, na_position='last'):
             sorted_idx = np.roll(sorted_idx, cnt_null)
         return sorted_idx
 
+    mask = isna(items)
     with warnings.catch_warnings():
         # https://github.com/pandas-dev/pandas/issues/25439
         # can be removed once ExtensionArrays are properly handled by nargsort
@@ -263,7 +264,6 @@ def nargsort(items, kind='quicksort', ascending=True, na_position='last'):
             message="Converting timezone-aware DatetimeArray to")
         items = np.asanyarray(items)
     idx = np.arange(len(items))
-    mask = isna(items)
     non_nans = items[~mask]
     non_nan_idx = idx[~mask]
     nan_idx = np.nonzero(mask)[0]

From 0dc48a454e1445a4215cd5bcca483b6530140d77 Mon Sep 17 00:00:00 2001
From: makbigc <makszechun@gmail.com>
Date: Fri, 14 Jun 2019 14:42:13 +0800
Subject: [PATCH 20/22] Add np.ndarray check to aviod double conversion

---
 pandas/core/arrays/base.py    |  2 ++
 pandas/core/arrays/integer.py | 13 +++++++++++++
 pandas/core/arrays/numpy_.py  |  6 +++---
 pandas/core/sorting.py        | 16 +++++++++-------
 4 files changed, 27 insertions(+), 10 deletions(-)

diff --git a/pandas/core/arrays/base.py b/pandas/core/arrays/base.py
index 38c4fe9b2c007..088ce9e380174 100644
--- a/pandas/core/arrays/base.py
+++ b/pandas/core/arrays/base.py
@@ -365,11 +365,13 @@ def isna(self) -> ArrayLike:
     def _values_for_argsort(self) -> np.ndarray:
         """
         Return values for sorting.
+
         Returns
         -------
         ndarray
             The transformed values should maintain the ordering between values
             within the array.
+
         See Also
         --------
         ExtensionArray.argsort
diff --git a/pandas/core/arrays/integer.py b/pandas/core/arrays/integer.py
index 1dbc8cde595e5..d74975b6d29fd 100644
--- a/pandas/core/arrays/integer.py
+++ b/pandas/core/arrays/integer.py
@@ -501,6 +501,19 @@ def value_counts(self, dropna=True):
 
         return Series(array, index=index)
 
+    def _values_for_argsort(self) -> np.ndarray:
+        """Return values for sorting.
+        Returns
+        -------
+        ndarray
+            The transformed values should maintain the ordering between values
+            within the array.
+        See Also
+        --------
+        ExtensionArray.argsort
+        """
+        return self
+
     @classmethod
     def _create_comparison_method(cls, op):
         def cmp_method(self, other):
diff --git a/pandas/core/arrays/numpy_.py b/pandas/core/arrays/numpy_.py
index 6ae464f169a13..0deefd2b10b6e 100644
--- a/pandas/core/arrays/numpy_.py
+++ b/pandas/core/arrays/numpy_.py
@@ -280,12 +280,12 @@ def take(self, indices, allow_fill=False, fill_value=None):
     def copy(self, deep=False):
         return type(self)(self._ndarray.copy())
 
-    def _values_for_factorize(self):
-        return self._ndarray, -1
-
     def _values_for_argsort(self):
         return self._ndarray
 
+    def _values_for_factorize(self):
+        return self._ndarray, -1
+
     def unique(self):
         from pandas import unique
 
diff --git a/pandas/core/sorting.py b/pandas/core/sorting.py
index f4b71f69f481b..4478166626661 100644
--- a/pandas/core/sorting.py
+++ b/pandas/core/sorting.py
@@ -256,13 +256,15 @@ def nargsort(items, kind='quicksort', ascending=True, na_position='last'):
         return sorted_idx
 
     mask = isna(items)
-    with warnings.catch_warnings():
-        # https://github.com/pandas-dev/pandas/issues/25439
-        # can be removed once ExtensionArrays are properly handled by nargsort
-        warnings.filterwarnings(
-            "ignore", category=FutureWarning,
-            message="Converting timezone-aware DatetimeArray to")
-        items = np.asanyarray(items)
+    if not isinstance(items, np.ndarray):
+        with warnings.catch_warnings():
+            # https://github.com/pandas-dev/pandas/issues/25439
+            # can be removed once ExtensionArrays are properly handled by
+            # nargsort
+            warnings.filterwarnings(
+                "ignore", category=FutureWarning,
+                message="Converting timezone-aware DatetimeArray to")
+            items = np.asanyarray(items)
     idx = np.arange(len(items))
     non_nans = items[~mask]
     non_nan_idx = idx[~mask]

From 7186d9dc378a15dddc1d1ddbdfdb5a3139f23568 Mon Sep 17 00:00:00 2001
From: makbigc <makszechun@gmail.com>
Date: Fri, 14 Jun 2019 15:55:58 +0800
Subject: [PATCH 21/22] change whatsnew

---
 doc/source/whatsnew/v0.25.0.rst | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/doc/source/whatsnew/v0.25.0.rst b/doc/source/whatsnew/v0.25.0.rst
index 273cab7003746..18dcaf1359fe3 100644
--- a/doc/source/whatsnew/v0.25.0.rst
+++ b/doc/source/whatsnew/v0.25.0.rst
@@ -459,7 +459,7 @@ Other API Changes
 - Most Pandas classes had a ``__bytes__`` method, which was used for getting a python2-style bytestring representation of the object. This method has been removed as a part of dropping Python2 (:issue:`26447`)
 - The ``.str``-accessor has been disabled for 1-level :class:`MultiIndex`, use :meth:`MultiIndex.to_flat_index` if necessary (:issue:`23679`)
 - Removed support of gtk package for clipboards (:issue:`26563`)
-- :meth:`ExtensionArray.argsort` places NA values at the end of the sorted array (:issue:`21801`)
+- :meth:`ExtensionArray.argsort` places NA values at the end of the sorted array, except :class:`Categorical` (:issue:`21801`)
 
 .. _whatsnew_0250.deprecations:
 

From 84f87a84b1295b6221302d465942e5b265c62f4a Mon Sep 17 00:00:00 2001
From: makbigc <makszechun@gmail.com>
Date: Fri, 14 Jun 2019 16:09:48 +0800
Subject: [PATCH 22/22] Undo the addition

---
 pandas/core/arrays/integer.py | 13 -------------
 1 file changed, 13 deletions(-)

diff --git a/pandas/core/arrays/integer.py b/pandas/core/arrays/integer.py
index d74975b6d29fd..1dbc8cde595e5 100644
--- a/pandas/core/arrays/integer.py
+++ b/pandas/core/arrays/integer.py
@@ -501,19 +501,6 @@ def value_counts(self, dropna=True):
 
         return Series(array, index=index)
 
-    def _values_for_argsort(self) -> np.ndarray:
-        """Return values for sorting.
-        Returns
-        -------
-        ndarray
-            The transformed values should maintain the ordering between values
-            within the array.
-        See Also
-        --------
-        ExtensionArray.argsort
-        """
-        return self
-
     @classmethod
     def _create_comparison_method(cls, op):
         def cmp_method(self, other):