test_analytics.py

# coding=utf-8
# pylint: disable-msg=E1101,W0612

from itertools import product
from distutils.version import LooseVersion

import pytest

from numpy import nan
import numpy as np
import pandas as pd

from pandas import (Series, Categorical, DataFrame, isna, notna,
                    bdate_range, date_range, _np_version_under1p10)
from pandas.core.index import MultiIndex
from pandas.core.indexes.datetimes import Timestamp
from pandas.core.indexes.timedeltas import Timedelta
import pandas.core.config as cf

import pandas.core.nanops as nanops

from pandas.compat import lrange, range, is_platform_windows
from pandas import compat
from pandas.util.testing import (assert_series_equal, assert_almost_equal,
                                 assert_frame_equal, assert_index_equal)
import pandas.util.testing as tm

from .common import TestData


skip_if_bottleneck_on_windows = (is_platform_windows() and
                                 nanops._USE_BOTTLENECK)


class TestSeriesAnalytics(TestData):

    def test_sum_zero(self):
        arr = np.array([])
        assert nanops.nansum(arr) == 0

        arr = np.empty((10, 0))
        assert (nanops.nansum(arr, axis=1) == 0).all()

        # GH #844
        s = Series([], index=[])
        assert s.sum() == 0

        df = DataFrame(np.empty((10, 0)))
        assert (df.sum(1) == 0).all()

    def test_nansum_buglet(self):
        s = Series([1.0, np.nan], index=[0, 1])
        result = np.nansum(s)
        assert_almost_equal(result, 1)

    def test_overflow(self):
        # GH 6915
        # overflowing on the smaller int dtypes
        for dtype in ['int32', 'int64']:
            v = np.arange(5000000, dtype=dtype)
            s = Series(v)

            # no bottleneck
            result = s.sum(skipna=False)
            assert int(result) == v.sum(dtype='int64')
            result = s.min(skipna=False)
            assert int(result) == 0
            result = s.max(skipna=False)
            assert int(result) == v[-1]

        for dtype in ['float32', 'float64']:
            v = np.arange(5000000, dtype=dtype)
            s = Series(v)

            # no bottleneck
            result = s.sum(skipna=False)
            assert result == v.sum(dtype=dtype)
            result = s.min(skipna=False)
            assert np.allclose(float(result), 0.0)
            result = s.max(skipna=False)
            assert np.allclose(float(result), v[-1])

    @pytest.mark.xfail(
        skip_if_bottleneck_on_windows,
        reason="buggy bottleneck with sum overflow on windows")
    def test_overflow_with_bottleneck(self):
        # GH 6915
        # overflowing on the smaller int dtypes
        for dtype in ['int32', 'int64']:
            v = np.arange(5000000, dtype=dtype)
            s = Series(v)

            # use bottleneck if available
            result = s.sum()
            assert int(result) == v.sum(dtype='int64')
            result = s.min()
            assert int(result) == 0
            result = s.max()
            assert int(result) == v[-1]

        for dtype in ['float32', 'float64']:
            v = np.arange(5000000, dtype=dtype)
            s = Series(v)

            # use bottleneck if available
            result = s.sum()
            assert result == v.sum(dtype=dtype)
            result = s.min()
            assert np.allclose(float(result), 0.0)
            result = s.max()
            assert np.allclose(float(result), v[-1])

    @pytest.mark.xfail(
        skip_if_bottleneck_on_windows,
        reason="buggy bottleneck with sum overflow on windows")
    def test_sum(self):
        self._check_stat_op('sum', np.sum, check_allna=True)

    def test_sum_inf(self):
        import pandas.core.nanops as nanops

        s = Series(np.random.randn(10))
        s2 = s.copy()

        s[5:8] = np.inf
        s2[5:8] = np.nan

        assert np.isinf(s.sum())

        arr = np.random.randn(100, 100).astype('f4')
        arr[:, 2] = np.inf

        with cf.option_context("mode.use_inf_as_na", True):
            assert_almost_equal(s.sum(), s2.sum())

        res = nanops.nansum(arr, axis=1)
        assert np.isinf(res).all()

    def test_mean(self):
        self._check_stat_op('mean', np.mean)

    def test_median(self):
        self._check_stat_op('median', np.median)

        # test with integers, test failure
        int_ts = Series(np.ones(10, dtype=int), index=lrange(10))
        tm.assert_almost_equal(np.median(int_ts), int_ts.median())

    def test_mode(self):
        # No mode should be found.
        exp = Series([], dtype=np.float64)
        tm.assert_series_equal(Series([]).mode(), exp)

        exp = Series([1], dtype=np.int64)
        tm.assert_series_equal(Series([1]).mode(), exp)

        exp = Series(['a', 'b', 'c'], dtype=np.object)
        tm.assert_series_equal(Series(['a', 'b', 'c']).mode(), exp)

        # Test numerical data types.
        exp_single = [1]
        data_single = [1] * 5 + [2] * 3

        exp_multi = [1, 3]
        data_multi = [1] * 5 + [2] * 3 + [3] * 5

        for dt in np.typecodes['AllInteger'] + np.typecodes['Float']:
            s = Series(data_single, dtype=dt)
            exp = Series(exp_single, dtype=dt)
            tm.assert_series_equal(s.mode(), exp)

            s = Series(data_multi, dtype=dt)
            exp = Series(exp_multi, dtype=dt)
            tm.assert_series_equal(s.mode(), exp)

        # Test string and object types.
        exp = ['b']
        data = ['a'] * 2 + ['b'] * 3

        s = Series(data, dtype='c')
        exp = Series(exp, dtype='c')
        tm.assert_series_equal(s.mode(), exp)

        exp = ['bar']
        data = ['foo'] * 2 + ['bar'] * 3

        for dt in [str, object]:
            s = Series(data, dtype=dt)
            exp = Series(exp, dtype=dt)
            tm.assert_series_equal(s.mode(), exp)

        # Test datetime types.
        exp = Series(['1900-05-03', '2011-01-03',
                      '2013-01-02'], dtype='M8[ns]')
        s = Series(['2011-01-03', '2013-01-02',
                    '1900-05-03'], dtype='M8[ns]')
        tm.assert_series_equal(s.mode(), exp)

        exp = Series(['2011-01-03', '2013-01-02'], dtype='M8[ns]')
        s = Series(['2011-01-03', '2013-01-02', '1900-05-03',
                    '2011-01-03', '2013-01-02'], dtype='M8[ns]')
        tm.assert_series_equal(s.mode(), exp)

        # gh-5986: Test timedelta types.
        exp = Series(['-1 days', '0 days', '1 days'], dtype='timedelta64[ns]')
        s = Series(['1 days', '-1 days', '0 days'],
                   dtype='timedelta64[ns]')
        tm.assert_series_equal(s.mode(), exp)

        exp = Series(['2 min', '1 day'], dtype='timedelta64[ns]')
        s = Series(['1 day', '1 day', '-1 day', '-1 day 2 min',
                    '2 min', '2 min'], dtype='timedelta64[ns]')
        tm.assert_series_equal(s.mode(), exp)

        # Test mixed dtype.
        exp = Series(['foo'])
        s = Series([1, 'foo', 'foo'])
        tm.assert_series_equal(s.mode(), exp)

        # Test for uint64 overflow.
        exp = Series([2**63], dtype=np.uint64)
        s = Series([1, 2**63, 2**63], dtype=np.uint64)
        tm.assert_series_equal(s.mode(), exp)

        exp = Series([1, 2**63], dtype=np.uint64)
        s = Series([1, 2**63], dtype=np.uint64)
        tm.assert_series_equal(s.mode(), exp)

        # Test category dtype.
        c = Categorical([1, 2])
        exp = Categorical([1, 2], categories=[1, 2])
        exp = Series(exp, dtype='category')
        tm.assert_series_equal(Series(c).mode(), exp)

        c = Categorical([1, 'a', 'a'])
        exp = Categorical(['a'], categories=[1, 'a'])
        exp = Series(exp, dtype='category')
        tm.assert_series_equal(Series(c).mode(), exp)

        c = Categorical([1, 1, 2, 3, 3])
        exp = Categorical([1, 3], categories=[1, 2, 3])
        exp = Series(exp, dtype='category')
        tm.assert_series_equal(Series(c).mode(), exp)

    def test_prod(self):
        self._check_stat_op('prod', np.prod)

    def test_min(self):
        self._check_stat_op('min', np.min, check_objects=True)

    def test_max(self):
        self._check_stat_op('max', np.max, check_objects=True)

    def test_var_std(self):
        alt = lambda x: np.std(x, ddof=1)
        self._check_stat_op('std', alt)

        alt = lambda x: np.var(x, ddof=1)
        self._check_stat_op('var', alt)

        result = self.ts.std(ddof=4)
        expected = np.std(self.ts.values, ddof=4)
        assert_almost_equal(result, expected)

        result = self.ts.var(ddof=4)
        expected = np.var(self.ts.values, ddof=4)
        assert_almost_equal(result, expected)

        # 1 - element series with ddof=1
        s = self.ts.iloc[[0]]
        result = s.var(ddof=1)
        assert isna(result)

        result = s.std(ddof=1)
        assert isna(result)

    def test_sem(self):
        alt = lambda x: np.std(x, ddof=1) / np.sqrt(len(x))
        self._check_stat_op('sem', alt)

        result = self.ts.sem(ddof=4)
        expected = np.std(self.ts.values,
                          ddof=4) / np.sqrt(len(self.ts.values))
        assert_almost_equal(result, expected)

        # 1 - element series with ddof=1
        s = self.ts.iloc[[0]]
        result = s.sem(ddof=1)
        assert isna(result)

    def test_skew(self):
        tm._skip_if_no_scipy()

        from scipy.stats import skew
        alt = lambda x: skew(x, bias=False)
        self._check_stat_op('skew', alt)

        # test corner cases, skew() returns NaN unless there's at least 3
        # values
        min_N = 3
        for i in range(1, min_N + 1):
            s = Series(np.ones(i))
            df = DataFrame(np.ones((i, i)))
            if i < min_N:
                assert np.isnan(s.skew())
                assert np.isnan(df.skew()).all()
            else:
                assert 0 == s.skew()
                assert (df.skew() == 0).all()

    def test_kurt(self):
        tm._skip_if_no_scipy()

        from scipy.stats import kurtosis
        alt = lambda x: kurtosis(x, bias=False)
        self._check_stat_op('kurt', alt)

        index = MultiIndex(levels=[['bar'], ['one', 'two', 'three'], [0, 1]],
                           labels=[[0, 0, 0, 0, 0, 0], [0, 1, 2, 0, 1, 2],
                                   [0, 1, 0, 1, 0, 1]])
        s = Series(np.random.randn(6), index=index)
        tm.assert_almost_equal(s.kurt(), s.kurt(level=0)['bar'])

        # test corner cases, kurt() returns NaN unless there's at least 4
        # values
        min_N = 4
        for i in range(1, min_N + 1):
            s = Series(np.ones(i))
            df = DataFrame(np.ones((i, i)))
            if i < min_N:
                assert np.isnan(s.kurt())
                assert np.isnan(df.kurt()).all()
            else:
                assert 0 == s.kurt()
                assert (df.kurt() == 0).all()

    def test_describe(self):
        s = Series([0, 1, 2, 3, 4], name='int_data')
        result = s.describe()
        expected = Series([5, 2, s.std(), 0, 1, 2, 3, 4],
                          name='int_data',
                          index=['count', 'mean', 'std', 'min', '25%',
                                 '50%', '75%', 'max'])
        tm.assert_series_equal(result, expected)

        s = Series([True, True, False, False, False], name='bool_data')
        result = s.describe()
        expected = Series([5, 2, False, 3], name='bool_data',
                          index=['count', 'unique', 'top', 'freq'])
        tm.assert_series_equal(result, expected)

        s = Series(['a', 'a', 'b', 'c', 'd'], name='str_data')
        result = s.describe()
        expected = Series([5, 4, 'a', 2], name='str_data',
                          index=['count', 'unique', 'top', 'freq'])
        tm.assert_series_equal(result, expected)

    def test_argsort(self):
        self._check_accum_op('argsort', check_dtype=False)
        argsorted = self.ts.argsort()
        assert issubclass(argsorted.dtype.type, np.integer)

        # GH 2967 (introduced bug in 0.11-dev I think)
        s = Series([Timestamp('201301%02d' % (i + 1)) for i in range(5)])
        assert s.dtype == 'datetime64[ns]'
        shifted = s.shift(-1)
        assert shifted.dtype == 'datetime64[ns]'
        assert isna(shifted[4])

        result = s.argsort()
        expected = Series(lrange(5), dtype='int64')
        assert_series_equal(result, expected)

        result = shifted.argsort()
        expected = Series(lrange(4) + [-1], dtype='int64')
        assert_series_equal(result, expected)

    def test_argsort_stable(self):
        s = Series(np.random.randint(0, 100, size=10000))
        mindexer = s.argsort(kind='mergesort')
        qindexer = s.argsort()

        mexpected = np.argsort(s.values, kind='mergesort')
        qexpected = np.argsort(s.values, kind='quicksort')

        tm.assert_series_equal(mindexer, Series(mexpected),
                               check_dtype=False)
        tm.assert_series_equal(qindexer, Series(qexpected),
                               check_dtype=False)
        pytest.raises(AssertionError, tm.assert_numpy_array_equal,
                      qindexer, mindexer)

    def test_cumsum(self):
        self._check_accum_op('cumsum')

    def test_cumprod(self):
        self._check_accum_op('cumprod')

    def test_cummin(self):
        tm.assert_numpy_array_equal(self.ts.cummin().values,
                                    np.minimum.accumulate(np.array(self.ts)))
        ts = self.ts.copy()
        ts[::2] = np.NaN
        result = ts.cummin()[1::2]
        expected = np.minimum.accumulate(ts.valid())

        tm.assert_series_equal(result, expected)

    def test_cummax(self):
        tm.assert_numpy_array_equal(self.ts.cummax().values,
                                    np.maximum.accumulate(np.array(self.ts)))
        ts = self.ts.copy()
        ts[::2] = np.NaN
        result = ts.cummax()[1::2]
        expected = np.maximum.accumulate(ts.valid())

        tm.assert_series_equal(result, expected)

    def test_cummin_datetime64(self):
        s = pd.Series(pd.to_datetime(['NaT', '2000-1-2', 'NaT', '2000-1-1',
                                      'NaT', '2000-1-3']))

        expected = pd.Series(pd.to_datetime(['NaT', '2000-1-2', 'NaT',
                                             '2000-1-1', 'NaT', '2000-1-1']))
        result = s.cummin(skipna=True)
        tm.assert_series_equal(expected, result)

        expected = pd.Series(pd.to_datetime(
            ['NaT', '2000-1-2', '2000-1-2', '2000-1-1', '2000-1-1', '2000-1-1'
             ]))
        result = s.cummin(skipna=False)
        tm.assert_series_equal(expected, result)

    def test_cummax_datetime64(self):
        s = pd.Series(pd.to_datetime(['NaT', '2000-1-2', 'NaT', '2000-1-1',
                                      'NaT', '2000-1-3']))

        expected = pd.Series(pd.to_datetime(['NaT', '2000-1-2', 'NaT',
                                             '2000-1-2', 'NaT', '2000-1-3']))
        result = s.cummax(skipna=True)
        tm.assert_series_equal(expected, result)

        expected = pd.Series(pd.to_datetime(
            ['NaT', '2000-1-2', '2000-1-2', '2000-1-2', '2000-1-2', '2000-1-3'
             ]))
        result = s.cummax(skipna=False)
        tm.assert_series_equal(expected, result)

    def test_cummin_timedelta64(self):
        s = pd.Series(pd.to_timedelta(['NaT',
                                       '2 min',
                                       'NaT',
                                       '1 min',
                                       'NaT',
                                       '3 min', ]))

        expected = pd.Series(pd.to_timedelta(['NaT',
                                              '2 min',
                                              'NaT',
                                              '1 min',
                                              'NaT',
                                              '1 min', ]))
        result = s.cummin(skipna=True)
        tm.assert_series_equal(expected, result)

        expected = pd.Series(pd.to_timedelta(['NaT',
                                              '2 min',
                                              '2 min',
                                              '1 min',
                                              '1 min',
                                              '1 min', ]))
        result = s.cummin(skipna=False)
        tm.assert_series_equal(expected, result)

    def test_cummax_timedelta64(self):
        s = pd.Series(pd.to_timedelta(['NaT',
                                       '2 min',
                                       'NaT',
                                       '1 min',
                                       'NaT',
                                       '3 min', ]))

        expected = pd.Series(pd.to_timedelta(['NaT',
                                              '2 min',
                                              'NaT',
                                              '2 min',
                                              'NaT',
                                              '3 min', ]))
        result = s.cummax(skipna=True)
        tm.assert_series_equal(expected, result)

        expected = pd.Series(pd.to_timedelta(['NaT',
                                              '2 min',
                                              '2 min',
                                              '2 min',
                                              '2 min',
                                              '3 min', ]))
        result = s.cummax(skipna=False)
        tm.assert_series_equal(expected, result)

    def test_npdiff(self):
        pytest.skip("skipping due to Series no longer being an "
                    "ndarray")

        # no longer works as the return type of np.diff is now nd.array
        s = Series(np.arange(5))

        r = np.diff(s)
        assert_series_equal(Series([nan, 0, 0, 0, nan]), r)

    def _check_stat_op(self, name, alternate, check_objects=False,
                       check_allna=False):
        import pandas.core.nanops as nanops

        def testit():
            f = getattr(Series, name)

            # add some NaNs
            self.series[5:15] = np.NaN

            # idxmax, idxmin, min, and max are valid for dates
            if name not in ['max', 'min']:
                ds = Series(date_range('1/1/2001', periods=10))
                pytest.raises(TypeError, f, ds)

            # skipna or no
            assert notna(f(self.series))
            assert isna(f(self.series, skipna=False))

            # check the result is correct
            nona = self.series.dropna()
            assert_almost_equal(f(nona), alternate(nona.values))
            assert_almost_equal(f(self.series), alternate(nona.values))

            allna = self.series * nan

            if check_allna:
                # xref 9422
                # bottleneck >= 1.0 give 0.0 for an allna Series sum
                try:
                    assert nanops._USE_BOTTLENECK
                    import bottleneck as bn  # noqa
                    assert bn.__version__ >= LooseVersion('1.0')
                    assert f(allna) == 0.0
                except:
                    assert np.isnan(f(allna))

            # dtype=object with None, it works!
            s = Series([1, 2, 3, None, 5])
            f(s)

            # 2888
            l = [0]
            l.extend(lrange(2 ** 40, 2 ** 40 + 1000))
            s = Series(l, dtype='int64')
            assert_almost_equal(float(f(s)), float(alternate(s.values)))

            # check date range
            if check_objects:
                s = Series(bdate_range('1/1/2000', periods=10))
                res = f(s)
                exp = alternate(s)
                assert res == exp

            # check on string data
            if name not in ['sum', 'min', 'max']:
                pytest.raises(TypeError, f, Series(list('abc')))

            # Invalid axis.
            pytest.raises(ValueError, f, self.series, axis=1)

            # Unimplemented numeric_only parameter.
            if 'numeric_only' in compat.signature(f).args:
                tm.assert_raises_regex(NotImplementedError, name, f,
                                       self.series, numeric_only=True)

        testit()

        try:
            import bottleneck as bn  # noqa
            nanops._USE_BOTTLENECK = False
            testit()
            nanops._USE_BOTTLENECK = True
        except ImportError:
            pass

    def _check_accum_op(self, name, check_dtype=True):
        func = getattr(np, name)
        tm.assert_numpy_array_equal(func(self.ts).values,
                                    func(np.array(self.ts)),
                                    check_dtype=check_dtype)

        # with missing values
        ts = self.ts.copy()
        ts[::2] = np.NaN

        result = func(ts)[1::2]
        expected = func(np.array(ts.valid()))

        tm.assert_numpy_array_equal(result.values, expected,
                                    check_dtype=False)

    def test_compress(self):
        cond = [True, False, True, False, False]
        s = Series([1, -1, 5, 8, 7],
                   index=list('abcde'), name='foo')
        expected = Series(s.values.compress(cond),
                          index=list('ac'), name='foo')
        tm.assert_series_equal(s.compress(cond), expected)

    def test_numpy_compress(self):
        cond = [True, False, True, False, False]
        s = Series([1, -1, 5, 8, 7],
                   index=list('abcde'), name='foo')
        expected = Series(s.values.compress(cond),
                          index=list('ac'), name='foo')
        tm.assert_series_equal(np.compress(cond, s), expected)

        msg = "the 'axis' parameter is not supported"
        tm.assert_raises_regex(ValueError, msg, np.compress,
                               cond, s, axis=1)

        msg = "the 'out' parameter is not supported"
        tm.assert_raises_regex(ValueError, msg, np.compress,
                               cond, s, out=s)

    def test_round(self):
        self.ts.index.name = "index_name"
        result = self.ts.round(2)
        expected = Series(np.round(self.ts.values, 2),
                          index=self.ts.index, name='ts')
        assert_series_equal(result, expected)
        assert result.name == self.ts.name

    def test_numpy_round(self):
        # See gh-12600
        s = Series([1.53, 1.36, 0.06])
        out = np.round(s, decimals=0)
        expected = Series([2., 1., 0.])
        assert_series_equal(out, expected)

        msg = "the 'out' parameter is not supported"
        with tm.assert_raises_regex(ValueError, msg):
            np.round(s, decimals=0, out=s)

    def test_built_in_round(self):
        if not compat.PY3:
            pytest.skip(
                'build in round cannot be overriden prior to Python 3')

        s = Series([1.123, 2.123, 3.123], index=lrange(3))
        result = round(s)
        expected_rounded0 = Series([1., 2., 3.], index=lrange(3))
        tm.assert_series_equal(result, expected_rounded0)

        decimals = 2
        expected_rounded = Series([1.12, 2.12, 3.12], index=lrange(3))
        result = round(s, decimals)
        tm.assert_series_equal(result, expected_rounded)

    def test_prod_numpy16_bug(self):
        s = Series([1., 1., 1.], index=lrange(3))
        result = s.prod()

        assert not isinstance(result, Series)

    def test_all_any(self):
        ts = tm.makeTimeSeries()
        bool_series = ts > 0
        assert not bool_series.all()
        assert bool_series.any()

        # Alternative types, with implicit 'object' dtype.
        s = Series(['abc', True])
        assert 'abc' == s.any()  # 'abc' || True => 'abc'

    def test_all_any_params(self):
        # Check skipna, with implicit 'object' dtype.
        s1 = Series([np.nan, True])
        s2 = Series([np.nan, False])
        assert s1.all(skipna=False)  # nan && True => True
        assert s1.all(skipna=True)
        assert np.isnan(s2.any(skipna=False))  # nan || False => nan
        assert not s2.any(skipna=True)

        # Check level.
        s = pd.Series([False, False, True, True, False, True],
                      index=[0, 0, 1, 1, 2, 2])
        assert_series_equal(s.all(level=0), Series([False, True, False]))
        assert_series_equal(s.any(level=0), Series([False, True, True]))

        # bool_only is not implemented with level option.
        pytest.raises(NotImplementedError, s.any, bool_only=True, level=0)
        pytest.raises(NotImplementedError, s.all, bool_only=True, level=0)

        # bool_only is not implemented alone.
        pytest.raises(NotImplementedError, s.any, bool_only=True)
        pytest.raises(NotImplementedError, s.all, bool_only=True)

    def test_modulo(self):
        with np.errstate(all='ignore'):

            # GH3590, modulo as ints
            p = DataFrame({'first': [3, 4, 5, 8], 'second': [0, 0, 0, 3]})
            result = p['first'] % p['second']
            expected = Series(p['first'].values % p['second'].values,
                              dtype='float64')
            expected.iloc[0:3] = np.nan
            assert_series_equal(result, expected)

            result = p['first'] % 0
            expected = Series(np.nan, index=p.index, name='first')
            assert_series_equal(result, expected)

            p = p.astype('float64')
            result = p['first'] % p['second']
            expected = Series(p['first'].values % p['second'].values)
            assert_series_equal(result, expected)

            p = p.astype('float64')
            result = p['first'] % p['second']
            result2 = p['second'] % p['first']
            assert not np.array_equal(result, result2)

            # GH 9144
            s = Series([0, 1])

            result = s % 0
            expected = Series([nan, nan])
            assert_series_equal(result, expected)

            result = 0 % s
            expected = Series([nan, 0.0])
            assert_series_equal(result, expected)

    def test_ops_consistency_on_empty(self):

        # GH 7869
        # consistency on empty

        # float
        result = Series(dtype=float).sum()
        assert result == 0

        result = Series(dtype=float).mean()
        assert isna(result)

        result = Series(dtype=float).median()
        assert isna(result)

        # timedelta64[ns]
        result = Series(dtype='m8[ns]').sum()
        assert result == Timedelta(0)

        result = Series(dtype='m8[ns]').mean()
        assert result is pd.NaT

        result = Series(dtype='m8[ns]').median()
        assert result is pd.NaT

    def test_corr(self):
        tm._skip_if_no_scipy()

        import scipy.stats as stats

        # full overlap
        tm.assert_almost_equal(self.ts.corr(self.ts), 1)

        # partial overlap
        tm.assert_almost_equal(self.ts[:15].corr(self.ts[5:]), 1)

        assert isna(self.ts[:15].corr(self.ts[5:], min_periods=12))

        ts1 = self.ts[:15].reindex(self.ts.index)
        ts2 = self.ts[5:].reindex(self.ts.index)
        assert isna(ts1.corr(ts2, min_periods=12))

        # No overlap
        assert np.isnan(self.ts[::2].corr(self.ts[1::2]))

        # all NA
        cp = self.ts[:10].copy()
        cp[:] = np.nan
        assert isna(cp.corr(cp))

        A = tm.makeTimeSeries()
        B = tm.makeTimeSeries()
        result = A.corr(B)
        expected, _ = stats.pearsonr(A, B)
        tm.assert_almost_equal(result, expected)

    def test_corr_rank(self):
        tm._skip_if_no_scipy()

        import scipy
        import scipy.stats as stats

        # kendall and spearman
        A = tm.makeTimeSeries()
        B = tm.makeTimeSeries()
        A[-5:] = A[:5]
        result = A.corr(B, method='kendall')
        expected = stats.kendalltau(A, B)[0]
        tm.assert_almost_equal(result, expected)

        result = A.corr(B, method='spearman')
        expected = stats.spearmanr(A, B)[0]
        tm.assert_almost_equal(result, expected)

        # these methods got rewritten in 0.8
        if scipy.__version__ < LooseVersion('0.9'):
            pytest.skip("skipping corr rank because of scipy version "
                        "{0}".format(scipy.__version__))

        # results from R
        A = Series(
            [-0.89926396, 0.94209606, -1.03289164, -0.95445587, 0.76910310, -
             0.06430576, -2.09704447, 0.40660407, -0.89926396, 0.94209606])
        B = Series(
            [-1.01270225, -0.62210117, -1.56895827, 0.59592943, -0.01680292,
             1.17258718, -1.06009347, -0.10222060, -0.89076239, 0.89372375])
        kexp = 0.4319297
        sexp = 0.5853767
        tm.assert_almost_equal(A.corr(B, method='kendall'), kexp)
        tm.assert_almost_equal(A.corr(B, method='spearman'), sexp)

    def test_cov(self):
        # full overlap
        tm.assert_almost_equal(self.ts.cov(self.ts), self.ts.std() ** 2)

        # partial overlap
        tm.assert_almost_equal(self.ts[:15].cov(self.ts[5:]),
                               self.ts[5:15].std() ** 2)

        # No overlap
        assert np.isnan(self.ts[::2].cov(self.ts[1::2]))

        # all NA
        cp = self.ts[:10].copy()
        cp[:] = np.nan
        assert isna(cp.cov(cp))

        # min_periods
        assert isna(self.ts[:15].cov(self.ts[5:], min_periods=12))

        ts1 = self.ts[:15].reindex(self.ts.index)
        ts2 = self.ts[5:].reindex(self.ts.index)
        assert isna(ts1.cov(ts2, min_periods=12))

    def test_count(self):
        assert self.ts.count() == len(self.ts)

        self.ts[::2] = np.NaN

        assert self.ts.count() == np.isfinite(self.ts).sum()

        mi = MultiIndex.from_arrays([list('aabbcc'), [1, 2, 2, nan, 1, 2]])
        ts = Series(np.arange(len(mi)), index=mi)

        left = ts.count(level=1)
        right = Series([2, 3, 1], index=[1, 2, nan])
        assert_series_equal(left, right)

        ts.iloc[[0, 3, 5]] = nan
        assert_series_equal(ts.count(level=1), right - 1)

    def test_dot(self):
        a = Series(np.random.randn(4), index=['p', 'q', 'r', 's'])
        b = DataFrame(np.random.randn(3, 4), index=['1', '2', '3'],
                      columns=['p', 'q', 'r', 's']).T

        result = a.dot(b)
        expected = Series(np.dot(a.values, b.values), index=['1', '2', '3'])
        assert_series_equal(result, expected)

        # Check index alignment
        b2 = b.reindex(index=reversed(b.index))
        result = a.dot(b)
        assert_series_equal(result, expected)

        # Check ndarray argument
        result = a.dot(b.values)
        assert np.all(result == expected.values)
        assert_almost_equal(a.dot(b['2'].values), expected['2'])

        # Check series argument
        assert_almost_equal(a.dot(b['1']), expected['1'])
        assert_almost_equal(a.dot(b2['1']), expected['1'])

        pytest.raises(Exception, a.dot, a.values[:3])
        pytest.raises(ValueError, a.dot, b.T)

    def test_value_counts_nunique(self):

        # basics.rst doc example
        series = Series(np.random.randn(500))
        series[20:500] = np.nan
        series[10:20] = 5000
        result = series.nunique()
        assert result == 11

    def test_unique(self):

        # 714 also, dtype=float
        s = Series([1.2345] * 100)
        s[::2] = np.nan
        result = s.unique()
        assert len(result) == 2

        s = Series([1.2345] * 100, dtype='f4')
        s[::2] = np.nan
        result = s.unique()
        assert len(result) == 2

        # NAs in object arrays #714
        s = Series(['foo'] * 100, dtype='O')
        s[::2] = np.nan
        result = s.unique()
        assert len(result) == 2

        # decision about None
        s = Series([1, 2, 3, None, None, None], dtype=object)
        result = s.unique()
        expected = np.array([1, 2, 3, None], dtype=object)
        tm.assert_numpy_array_equal(result, expected)

    def test_drop_duplicates(self):
        # check both int and object
        for s in [Series([1, 2, 3, 3]), Series(['1', '2', '3', '3'])]:
            expected = Series([False, False, False, True])
            assert_series_equal(s.duplicated(), expected)
            assert_series_equal(s.drop_duplicates(), s[~expected])
            sc = s.copy()
            sc.drop_duplicates(inplace=True)
            assert_series_equal(sc, s[~expected])

            expected = Series([False, False, True, False])
            assert_series_equal(s.duplicated(keep='last'), expected)
            assert_series_equal(s.drop_duplicates(keep='last'), s[~expected])
            sc = s.copy()
            sc.drop_duplicates(keep='last', inplace=True)
            assert_series_equal(sc, s[~expected])

            expected = Series([False, False, True, True])
            assert_series_equal(s.duplicated(keep=False), expected)
            assert_series_equal(s.drop_duplicates(keep=False), s[~expected])
            sc = s.copy()
            sc.drop_duplicates(keep=False, inplace=True)
            assert_series_equal(sc, s[~expected])

        for s in [Series([1, 2, 3, 5, 3, 2, 4]),
                  Series(['1', '2', '3', '5', '3', '2', '4'])]:
            expected = Series([False, False, False, False, True, True, False])
            assert_series_equal(s.duplicated(), expected)
            assert_series_equal(s.drop_duplicates(), s[~expected])
            sc = s.copy()
            sc.drop_duplicates(inplace=True)
            assert_series_equal(sc, s[~expected])

            expected = Series([False, True, True, False, False, False, False])
            assert_series_equal(s.duplicated(keep='last'), expected)
            assert_series_equal(s.drop_duplicates(keep='last'), s[~expected])
            sc = s.copy()
            sc.drop_duplicates(keep='last', inplace=True)
            assert_series_equal(sc, s[~expected])

            expected = Series([False, True, True, False, True, True, False])
            assert_series_equal(s.duplicated(keep=False), expected)
            assert_series_equal(s.drop_duplicates(keep=False), s[~expected])
            sc = s.copy()
            sc.drop_duplicates(keep=False, inplace=True)
            assert_series_equal(sc, s[~expected])

    def test_clip(self):
        val = self.ts.median()

        assert self.ts.clip_lower(val).min() == val
        assert self.ts.clip_upper(val).max() == val

        assert self.ts.clip(lower=val).min() == val
        assert self.ts.clip(upper=val).max() == val

        result = self.ts.clip(-0.5, 0.5)
        expected = np.clip(self.ts, -0.5, 0.5)
        assert_series_equal(result, expected)
        assert isinstance(expected, Series)

    def test_clip_types_and_nulls(self):

        sers = [Series([np.nan, 1.0, 2.0, 3.0]), Series([None, 'a', 'b', 'c']),
                Series(pd.to_datetime(
                    [np.nan, 1, 2, 3], unit='D'))]

        for s in sers:
            thresh = s[2]
            l = s.clip_lower(thresh)
            u = s.clip_upper(thresh)
            assert l[notna(l)].min() == thresh
            assert u[notna(u)].max() == thresh
            assert list(isna(s)) == list(isna(l))
            assert list(isna(s)) == list(isna(u))

    def test_clip_with_na_args(self):
        """Should process np.nan argument as None """
        # GH # 17276
        s = Series([1, 2, 3])

        assert_series_equal(s.clip(np.nan), Series([1, 2, 3]))
        assert_series_equal(s.clip(upper=[1, 1, np.nan]), Series([1, 2, 3]))
        assert_series_equal(s.clip(lower=[1, np.nan, 1]), Series([1, 2, 3]))
        assert_series_equal(s.clip(upper=np.nan, lower=np.nan),
                            Series([1, 2, 3]))

    def test_clip_against_series(self):
        # GH #6966

        s = Series([1.0, 1.0, 4.0])
        threshold = Series([1.0, 2.0, 3.0])

        assert_series_equal(s.clip_lower(threshold), Series([1.0, 2.0, 4.0]))
        assert_series_equal(s.clip_upper(threshold), Series([1.0, 1.0, 3.0]))

        lower = Series([1.0, 2.0, 3.0])
        upper = Series([1.5, 2.5, 3.5])

        assert_series_equal(s.clip(lower, upper), Series([1.0, 2.0, 3.5]))
        assert_series_equal(s.clip(1.5, upper), Series([1.5, 1.5, 3.5]))

    @pytest.mark.parametrize("inplace", [True, False])
    @pytest.mark.parametrize("upper", [[1, 2, 3], np.asarray([1, 2, 3])])
    def test_clip_against_list_like(self, inplace, upper):
        # GH #15390
        original = pd.Series([5, 6, 7])
        result = original.clip(upper=upper, inplace=inplace)
        expected = pd.Series([1, 2, 3])

        if inplace:
            result = original
        tm.assert_series_equal(result, expected, check_exact=True)

    def test_clip_with_datetimes(self):

        # GH 11838
        # naive and tz-aware datetimes

        t = Timestamp('2015-12-01 09:30:30')
        s = Series([Timestamp('2015-12-01 09:30:00'),
                    Timestamp('2015-12-01 09:31:00')])
        result = s.clip(upper=t)
        expected = Series([Timestamp('2015-12-01 09:30:00'),
                           Timestamp('2015-12-01 09:30:30')])
        assert_series_equal(result, expected)

        t = Timestamp('2015-12-01 09:30:30', tz='US/Eastern')
        s = Series([Timestamp('2015-12-01 09:30:00', tz='US/Eastern'),
                    Timestamp('2015-12-01 09:31:00', tz='US/Eastern')])
        result = s.clip(upper=t)
        expected = Series([Timestamp('2015-12-01 09:30:00', tz='US/Eastern'),
                           Timestamp('2015-12-01 09:30:30', tz='US/Eastern')])
        assert_series_equal(result, expected)

    def test_cummethods_bool(self):
        # GH 6270
        # looks like a buggy np.maximum.accumulate for numpy 1.6.1, py 3.2
        def cummin(x):
            return np.minimum.accumulate(x)

        def cummax(x):
            return np.maximum.accumulate(x)

        a = pd.Series([False, False, False, True, True, False, False])
        b = ~a
        c = pd.Series([False] * len(b))
        d = ~c
        methods = {'cumsum': np.cumsum,
                   'cumprod': np.cumprod,
                   'cummin': cummin,
                   'cummax': cummax}
        args = product((a, b, c, d), methods)
        for s, method in args:
            expected = Series(methods[method](s.values))
            result = getattr(s, method)()
            assert_series_equal(result, expected)

        e = pd.Series([False, True, nan, False])
        cse = pd.Series([0, 1, nan, 1], dtype=object)
        cpe = pd.Series([False, 0, nan, 0])
        cmin = pd.Series([False, False, nan, False])
        cmax = pd.Series([False, True, nan, True])
        expecteds = {'cumsum': cse,
                     'cumprod': cpe,
                     'cummin': cmin,
                     'cummax': cmax}

        for method in methods:
            res = getattr(e, method)()
            assert_series_equal(res, expecteds[method])

    def test_isin(self):
        s = Series(['A', 'B', 'C', 'a', 'B', 'B', 'A', 'C'])

        result = s.isin(['A', 'C'])
        expected = Series([True, False, True, False, False, False, True, True])
        assert_series_equal(result, expected)

        # GH: 16012
        # This specific issue has to have a series over 1e6 in len, but the
        # comparison array (in_list) must be large enough so that numpy doesn't
        # do a manual masking trick that will avoid this issue altogether
        s = Series(list('abcdefghijk' * 10 ** 5))
        # If numpy doesn't do the manual comparison/mask, these
        # unorderable mixed types are what cause the exception in numpy
        in_list = [-1, 'a', 'b', 'G', 'Y', 'Z', 'E',
                   'K', 'E', 'S', 'I', 'R', 'R'] * 6

        assert s.isin(in_list).sum() == 200000

    def test_isin_with_string_scalar(self):
        # GH4763
        s = Series(['A', 'B', 'C', 'a', 'B', 'B', 'A', 'C'])
        with pytest.raises(TypeError):
            s.isin('a')

        with pytest.raises(TypeError):
            s = Series(['aaa', 'b', 'c'])
            s.isin('aaa')

    def test_isin_with_i8(self):
        # GH 5021

        expected = Series([True, True, False, False, False])
        expected2 = Series([False, True, False, False, False])

        # datetime64[ns]
        s = Series(date_range('jan-01-2013', 'jan-05-2013'))

        result = s.isin(s[0:2])
        assert_series_equal(result, expected)

        result = s.isin(s[0:2].values)
        assert_series_equal(result, expected)

        # fails on dtype conversion in the first place
        result = s.isin(s[0:2].values.astype('datetime64[D]'))
        assert_series_equal(result, expected)

        result = s.isin([s[1]])
        assert_series_equal(result, expected2)

        result = s.isin([np.datetime64(s[1])])
        assert_series_equal(result, expected2)

        result = s.isin(set(s[0:2]))
        assert_series_equal(result, expected)

        # timedelta64[ns]
        s = Series(pd.to_timedelta(lrange(5), unit='d'))
        result = s.isin(s[0:2])
        assert_series_equal(result, expected)

    @pytest.mark.parametrize("empty", [[], Series(), np.array([])])
    def test_isin_empty(self, empty):
        # see gh-16991
        s = Series(["a", "b"])
        expected = Series([False, False])

        result = s.isin(empty)
        tm.assert_series_equal(expected, result)

    def test_timedelta64_analytics(self):
        from pandas import date_range

        # index min/max
        td = Series(date_range('2012-1-1', periods=3, freq='D')) - \
            Timestamp('20120101')

        result = td.idxmin()
        assert result == 0

        result = td.idxmax()
        assert result == 2

        # GH 2982
        # with NaT
        td[0] = np.nan

        result = td.idxmin()
        assert result == 1

        result = td.idxmax()
        assert result == 2

        # abs
        s1 = Series(date_range('20120101', periods=3))
        s2 = Series(date_range('20120102', periods=3))
        expected = Series(s2 - s1)

        # this fails as numpy returns timedelta64[us]
        # result = np.abs(s1-s2)
        # assert_frame_equal(result,expected)

        result = (s1 - s2).abs()
        assert_series_equal(result, expected)

        # max/min
        result = td.max()
        expected = Timedelta('2 days')
        assert result == expected

        result = td.min()
        expected = Timedelta('1 days')
        assert result == expected

    def test_idxmin(self):
        # test idxmin
        # _check_stat_op approach can not be used here because of isna check.

        # add some NaNs
        self.series[5:15] = np.NaN

        # skipna or no
        assert self.series[self.series.idxmin()] == self.series.min()
        assert isna(self.series.idxmin(skipna=False))

        # no NaNs
        nona = self.series.dropna()
        assert nona[nona.idxmin()] == nona.min()
        assert (nona.index.values.tolist().index(nona.idxmin()) ==
                nona.values.argmin())

        # all NaNs
        allna = self.series * nan
        assert isna(allna.idxmin())

        # datetime64[ns]
        from pandas import date_range
        s = Series(date_range('20130102', periods=6))
        result = s.idxmin()
        assert result == 0

        s[0] = np.nan
        result = s.idxmin()
        assert result == 1

    def test_numpy_argmin_deprecated(self):
        # See gh-16830
        data = np.arange(1, 11)

        s = Series(data, index=data)
        with tm.assert_produces_warning(FutureWarning, check_stacklevel=False):
            # The deprecation of Series.argmin also causes a deprecation
            # warning when calling np.argmin. This behavior is temporary
            # until the implemention of Series.argmin is corrected.
            result = np.argmin(s)

        assert result == 1

        with tm.assert_produces_warning(FutureWarning):
            # argmin is aliased to idxmin
            result = s.argmin()

        assert result == 1

        if not _np_version_under1p10:
            with tm.assert_produces_warning(FutureWarning,
                                            check_stacklevel=False):
                msg = "the 'out' parameter is not supported"
                tm.assert_raises_regex(ValueError, msg, np.argmin,
                                       s, out=data)

    def test_idxmax(self):
        # test idxmax
        # _check_stat_op approach can not be used here because of isna check.

        # add some NaNs
        self.series[5:15] = np.NaN

        # skipna or no
        assert self.series[self.series.idxmax()] == self.series.max()
        assert isna(self.series.idxmax(skipna=False))

        # no NaNs
        nona = self.series.dropna()
        assert nona[nona.idxmax()] == nona.max()
        assert (nona.index.values.tolist().index(nona.idxmax()) ==
                nona.values.argmax())

        # all NaNs
        allna = self.series * nan
        assert isna(allna.idxmax())

        from pandas import date_range
        s = Series(date_range('20130102', periods=6))
        result = s.idxmax()
        assert result == 5

        s[5] = np.nan
        result = s.idxmax()
        assert result == 4

        # Float64Index
        # GH 5914
        s = pd.Series([1, 2, 3], [1.1, 2.1, 3.1])
        result = s.idxmax()
        assert result == 3.1
        result = s.idxmin()
        assert result == 1.1

        s = pd.Series(s.index, s.index)
        result = s.idxmax()
        assert result == 3.1
        result = s.idxmin()
        assert result == 1.1

    def test_numpy_argmax_deprecated(self):
        # See gh-16830
        data = np.arange(1, 11)

        s = Series(data, index=data)
        with tm.assert_produces_warning(FutureWarning, check_stacklevel=False):
            # The deprecation of Series.argmax also causes a deprecation
            # warning when calling np.argmax. This behavior is temporary
            # until the implemention of Series.argmax is corrected.
            result = np.argmax(s)
        assert result == 10

        with tm.assert_produces_warning(FutureWarning):
            # argmax is aliased to idxmax
            result = s.argmax()

        assert result == 10

        if not _np_version_under1p10:
            with tm.assert_produces_warning(FutureWarning,
                                            check_stacklevel=False):
                msg = "the 'out' parameter is not supported"
                tm.assert_raises_regex(ValueError, msg, np.argmax,
                                       s, out=data)

    def test_ptp(self):
        N = 1000
        arr = np.random.randn(N)
        ser = Series(arr)
        assert np.ptp(ser) == np.ptp(arr)

        # GH11163
        s = Series([3, 5, np.nan, -3, 10])
        assert s.ptp() == 13
        assert pd.isna(s.ptp(skipna=False))

        mi = pd.MultiIndex.from_product([['a', 'b'], [1, 2, 3]])
        s = pd.Series([1, np.nan, 7, 3, 5, np.nan], index=mi)

        expected = pd.Series([6, 2], index=['a', 'b'], dtype=np.float64)
        tm.assert_series_equal(s.ptp(level=0), expected)

        expected = pd.Series([np.nan, np.nan], index=['a', 'b'])
        tm.assert_series_equal(s.ptp(level=0, skipna=False), expected)

        with pytest.raises(ValueError):
            s.ptp(axis=1)

        s = pd.Series(['a', 'b', 'c', 'd', 'e'])
        with pytest.raises(TypeError):
            s.ptp()

        with pytest.raises(NotImplementedError):
            s.ptp(numeric_only=True)

    def test_empty_timeseries_redections_return_nat(self):
        # covers #11245
        for dtype in ('m8[ns]', 'm8[ns]', 'M8[ns]', 'M8[ns, UTC]'):
            assert Series([], dtype=dtype).min() is pd.NaT
            assert Series([], dtype=dtype).max() is pd.NaT

    def test_unique_data_ownership(self):
        # it works! #1807
        Series(Series(["a", "c", "b"]).unique()).sort_values()

    def test_repeat(self):
        s = Series(np.random.randn(3), index=['a', 'b', 'c'])

        reps = s.repeat(5)
        exp = Series(s.values.repeat(5), index=s.index.values.repeat(5))
        assert_series_equal(reps, exp)

        with tm.assert_produces_warning(FutureWarning):
            result = s.repeat(reps=5)
            assert_series_equal(result, exp)

        to_rep = [2, 3, 4]
        reps = s.repeat(to_rep)
        exp = Series(s.values.repeat(to_rep),
                     index=s.index.values.repeat(to_rep))
        assert_series_equal(reps, exp)

    def test_numpy_repeat(self):
        s = Series(np.arange(3), name='x')
        expected = Series(s.values.repeat(2), name='x',
                          index=s.index.values.repeat(2))
        assert_series_equal(np.repeat(s, 2), expected)

        msg = "the 'axis' parameter is not supported"
        tm.assert_raises_regex(ValueError, msg, np.repeat, s, 2, axis=0)

    def test_searchsorted(self):
        s = Series([1, 2, 3])

        idx = s.searchsorted(1, side='left')
        tm.assert_numpy_array_equal(idx, np.array([0], dtype=np.intp))

        idx = s.searchsorted(1, side='right')
        tm.assert_numpy_array_equal(idx, np.array([1], dtype=np.intp))

        with tm.assert_produces_warning(FutureWarning):
            idx = s.searchsorted(v=1, side='left')
            tm.assert_numpy_array_equal(idx, np.array([0], dtype=np.intp))

    def test_searchsorted_numeric_dtypes_scalar(self):
        s = Series([1, 2, 90, 1000, 3e9])
        r = s.searchsorted(30)
        e = 2
        assert r == e

        r = s.searchsorted([30])
        e = np.array([2], dtype=np.intp)
        tm.assert_numpy_array_equal(r, e)

    def test_searchsorted_numeric_dtypes_vector(self):
        s = Series([1, 2, 90, 1000, 3e9])
        r = s.searchsorted([91, 2e6])
        e = np.array([3, 4], dtype=np.intp)
        tm.assert_numpy_array_equal(r, e)

    def test_search_sorted_datetime64_scalar(self):
        s = Series(pd.date_range('20120101', periods=10, freq='2D'))
        v = pd.Timestamp('20120102')
        r = s.searchsorted(v)
        e = 1
        assert r == e

    def test_search_sorted_datetime64_list(self):
        s = Series(pd.date_range('20120101', periods=10, freq='2D'))
        v = [pd.Timestamp('20120102'), pd.Timestamp('20120104')]
        r = s.searchsorted(v)
        e = np.array([1, 2], dtype=np.intp)
        tm.assert_numpy_array_equal(r, e)

    def test_searchsorted_sorter(self):
        # GH8490
        s = Series([3, 1, 2])
        r = s.searchsorted([0, 3], sorter=np.argsort(s))
        e = np.array([0, 2], dtype=np.intp)
        tm.assert_numpy_array_equal(r, e)

    def test_is_unique(self):
        # GH11946
        s = Series(np.random.randint(0, 10, size=1000))
        assert not s.is_unique
        s = Series(np.arange(1000))
        assert s.is_unique

    def test_is_monotonic(self):

        s = Series(np.random.randint(0, 10, size=1000))
        assert not s.is_monotonic
        s = Series(np.arange(1000))
        assert s.is_monotonic
        assert s.is_monotonic_increasing
        s = Series(np.arange(1000, 0, -1))
        assert s.is_monotonic_decreasing

        s = Series(pd.date_range('20130101', periods=10))
        assert s.is_monotonic
        assert s.is_monotonic_increasing
        s = Series(list(reversed(s.tolist())))
        assert not s.is_monotonic
        assert s.is_monotonic_decreasing

    def test_sort_index_level(self):
        mi = MultiIndex.from_tuples([[1, 1, 3], [1, 1, 1]], names=list('ABC'))
        s = Series([1, 2], mi)
        backwards = s.iloc[[1, 0]]

        res = s.sort_index(level='A')
        assert_series_equal(backwards, res)

        res = s.sort_index(level=['A', 'B'])
        assert_series_equal(backwards, res)

        res = s.sort_index(level='A', sort_remaining=False)
        assert_series_equal(s, res)

        res = s.sort_index(level=['A', 'B'], sort_remaining=False)
        assert_series_equal(s, res)

    def test_apply_categorical(self):
        values = pd.Categorical(list('ABBABCD'), categories=list('DCBA'),
                                ordered=True)
        s = pd.Series(values, name='XX', index=list('abcdefg'))
        result = s.apply(lambda x: x.lower())

        # should be categorical dtype when the number of categories are
        # the same
        values = pd.Categorical(list('abbabcd'), categories=list('dcba'),
                                ordered=True)
        exp = pd.Series(values, name='XX', index=list('abcdefg'))
        tm.assert_series_equal(result, exp)
        tm.assert_categorical_equal(result.values, exp.values)

        result = s.apply(lambda x: 'A')
        exp = pd.Series(['A'] * 7, name='XX', index=list('abcdefg'))
        tm.assert_series_equal(result, exp)
        assert result.dtype == np.object

    def test_shift_int(self):
        ts = self.ts.astype(int)
        shifted = ts.shift(1)
        expected = ts.astype(float).shift(1)
        assert_series_equal(shifted, expected)

    def test_shift_categorical(self):
        # GH 9416
        s = pd.Series(['a', 'b', 'c', 'd'], dtype='category')

        assert_series_equal(s.iloc[:-1], s.shift(1).shift(-1).valid())

        sp1 = s.shift(1)
        assert_index_equal(s.index, sp1.index)
        assert np.all(sp1.values.codes[:1] == -1)
        assert np.all(s.values.codes[:-1] == sp1.values.codes[1:])

        sn2 = s.shift(-2)
        assert_index_equal(s.index, sn2.index)
        assert np.all(sn2.values.codes[-2:] == -1)
        assert np.all(s.values.codes[2:] == sn2.values.codes[:-2])

        assert_index_equal(s.values.categories, sp1.values.categories)
        assert_index_equal(s.values.categories, sn2.values.categories)

    def test_reshape_deprecate(self):
        x = Series(np.random.random(10), name='x')
        tm.assert_produces_warning(FutureWarning, x.reshape, x.shape)

    def test_reshape_non_2d(self):
        # see gh-4554
        with tm.assert_produces_warning(FutureWarning):
            x = Series(np.random.random(201), name='x')
            assert x.reshape(x.shape, ) is x

        # see gh-2719
        with tm.assert_produces_warning(FutureWarning):
            a = Series([1, 2, 3, 4])
            result = a.reshape(2, 2)
            expected = a.values.reshape(2, 2)
            tm.assert_numpy_array_equal(result, expected)
            assert isinstance(result, type(expected))

    def test_reshape_2d_return_array(self):
        x = Series(np.random.random(201), name='x')

        with tm.assert_produces_warning(FutureWarning):
            result = x.reshape((-1, 1))
            assert not isinstance(result, Series)

        with tm.assert_produces_warning(FutureWarning, check_stacklevel=False):
            result2 = np.reshape(x, (-1, 1))
            assert not isinstance(result2, Series)

        with tm.assert_produces_warning(FutureWarning):
            result = x[:, None]
            expected = x.reshape((-1, 1))
            tm.assert_almost_equal(result, expected)

    def test_reshape_bad_kwarg(self):
        a = Series([1, 2, 3, 4])

        with tm.assert_produces_warning(FutureWarning, check_stacklevel=False):
            msg = "'foo' is an invalid keyword argument for this function"
            tm.assert_raises_regex(
                TypeError, msg, a.reshape, (2, 2), foo=2)

        with tm.assert_produces_warning(FutureWarning, check_stacklevel=False):
            msg = r"reshape\(\) got an unexpected keyword argument 'foo'"
            tm.assert_raises_regex(
                TypeError, msg, a.reshape, a.shape, foo=2)

    def test_numpy_reshape(self):
        a = Series([1, 2, 3, 4])

        with tm.assert_produces_warning(FutureWarning, check_stacklevel=False):
            result = np.reshape(a, (2, 2))
            expected = a.values.reshape(2, 2)
            tm.assert_numpy_array_equal(result, expected)
            assert isinstance(result, type(expected))

        with tm.assert_produces_warning(FutureWarning, check_stacklevel=False):
            result = np.reshape(a, a.shape)
            tm.assert_series_equal(result, a)

    def test_unstack(self):
        from numpy import nan

        index = MultiIndex(levels=[['bar', 'foo'], ['one', 'three', 'two']],
                           labels=[[1, 1, 0, 0], [0, 1, 0, 2]])

        s = Series(np.arange(4.), index=index)
        unstacked = s.unstack()

        expected = DataFrame([[2., nan, 3.], [0., 1., nan]],
                             index=['bar', 'foo'],
                             columns=['one', 'three', 'two'])

        assert_frame_equal(unstacked, expected)

        unstacked = s.unstack(level=0)
        assert_frame_equal(unstacked, expected.T)

        index = MultiIndex(levels=[['bar'], ['one', 'two', 'three'], [0, 1]],
                           labels=[[0, 0, 0, 0, 0, 0], [0, 1, 2, 0, 1, 2],
                                   [0, 1, 0, 1, 0, 1]])
        s = Series(np.random.randn(6), index=index)
        exp_index = MultiIndex(levels=[['one', 'two', 'three'], [0, 1]],
                               labels=[[0, 1, 2, 0, 1, 2], [0, 1, 0, 1, 0, 1]])
        expected = DataFrame({'bar': s.values},
                             index=exp_index).sort_index(level=0)
        unstacked = s.unstack(0).sort_index()
        assert_frame_equal(unstacked, expected)

        # GH5873
        idx = pd.MultiIndex.from_arrays([[101, 102], [3.5, np.nan]])
        ts = pd.Series([1, 2], index=idx)
        left = ts.unstack()
        right = DataFrame([[nan, 1], [2, nan]], index=[101, 102],
                          columns=[nan, 3.5])
        assert_frame_equal(left, right)

        idx = pd.MultiIndex.from_arrays([['cat', 'cat', 'cat', 'dog', 'dog'
                                          ], ['a', 'a', 'b', 'a', 'b'],
                                         [1, 2, 1, 1, np.nan]])
        ts = pd.Series([1.0, 1.1, 1.2, 1.3, 1.4], index=idx)
        right = DataFrame([[1.0, 1.3], [1.1, nan], [nan, 1.4], [1.2, nan]],
                          columns=['cat', 'dog'])
        tpls = [('a', 1), ('a', 2), ('b', nan), ('b', 1)]
        right.index = pd.MultiIndex.from_tuples(tpls)
        assert_frame_equal(ts.unstack(level=0), right)

    def test_value_counts_datetime(self):
        # most dtypes are tested in test_base.py
        values = [pd.Timestamp('2011-01-01 09:00'),
                  pd.Timestamp('2011-01-01 10:00'),
                  pd.Timestamp('2011-01-01 11:00'),
                  pd.Timestamp('2011-01-01 09:00'),
                  pd.Timestamp('2011-01-01 09:00'),
                  pd.Timestamp('2011-01-01 11:00')]

        exp_idx = pd.DatetimeIndex(['2011-01-01 09:00', '2011-01-01 11:00',
                                    '2011-01-01 10:00'])
        exp = pd.Series([3, 2, 1], index=exp_idx, name='xxx')

        s = pd.Series(values, name='xxx')
        tm.assert_series_equal(s.value_counts(), exp)
        # check DatetimeIndex outputs the same result
        idx = pd.DatetimeIndex(values, name='xxx')
        tm.assert_series_equal(idx.value_counts(), exp)

        # normalize
        exp = pd.Series(np.array([3., 2., 1]) / 6.,
                        index=exp_idx, name='xxx')
        tm.assert_series_equal(s.value_counts(normalize=True), exp)
        tm.assert_series_equal(idx.value_counts(normalize=True), exp)

    def test_value_counts_datetime_tz(self):
        values = [pd.Timestamp('2011-01-01 09:00', tz='US/Eastern'),
                  pd.Timestamp('2011-01-01 10:00', tz='US/Eastern'),
                  pd.Timestamp('2011-01-01 11:00', tz='US/Eastern'),
                  pd.Timestamp('2011-01-01 09:00', tz='US/Eastern'),
                  pd.Timestamp('2011-01-01 09:00', tz='US/Eastern'),
                  pd.Timestamp('2011-01-01 11:00', tz='US/Eastern')]

        exp_idx = pd.DatetimeIndex(['2011-01-01 09:00', '2011-01-01 11:00',
                                    '2011-01-01 10:00'], tz='US/Eastern')
        exp = pd.Series([3, 2, 1], index=exp_idx, name='xxx')

        s = pd.Series(values, name='xxx')
        tm.assert_series_equal(s.value_counts(), exp)
        idx = pd.DatetimeIndex(values, name='xxx')
        tm.assert_series_equal(idx.value_counts(), exp)

        exp = pd.Series(np.array([3., 2., 1]) / 6.,
                        index=exp_idx, name='xxx')
        tm.assert_series_equal(s.value_counts(normalize=True), exp)
        tm.assert_series_equal(idx.value_counts(normalize=True), exp)

    def test_value_counts_period(self):
        values = [pd.Period('2011-01', freq='M'),
                  pd.Period('2011-02', freq='M'),
                  pd.Period('2011-03', freq='M'),
                  pd.Period('2011-01', freq='M'),
                  pd.Period('2011-01', freq='M'),
                  pd.Period('2011-03', freq='M')]

        exp_idx = pd.PeriodIndex(['2011-01', '2011-03', '2011-02'], freq='M')
        exp = pd.Series([3, 2, 1], index=exp_idx, name='xxx')

        s = pd.Series(values, name='xxx')
        tm.assert_series_equal(s.value_counts(), exp)
        # check DatetimeIndex outputs the same result
        idx = pd.PeriodIndex(values, name='xxx')
        tm.assert_series_equal(idx.value_counts(), exp)

        # normalize
        exp = pd.Series(np.array([3., 2., 1]) / 6.,
                        index=exp_idx, name='xxx')
        tm.assert_series_equal(s.value_counts(normalize=True), exp)
        tm.assert_series_equal(idx.value_counts(normalize=True), exp)

    def test_value_counts_categorical_ordered(self):
        # most dtypes are tested in test_base.py
        values = pd.Categorical([1, 2, 3, 1, 1, 3], ordered=True)

        exp_idx = pd.CategoricalIndex([1, 3, 2], categories=[1, 2, 3],
                                      ordered=True)
        exp = pd.Series([3, 2, 1], index=exp_idx, name='xxx')

        s = pd.Series(values, name='xxx')
        tm.assert_series_equal(s.value_counts(), exp)
        # check CategoricalIndex outputs the same result
        idx = pd.CategoricalIndex(values, name='xxx')
        tm.assert_series_equal(idx.value_counts(), exp)

        # normalize
        exp = pd.Series(np.array([3., 2., 1]) / 6.,
                        index=exp_idx, name='xxx')
        tm.assert_series_equal(s.value_counts(normalize=True), exp)
        tm.assert_series_equal(idx.value_counts(normalize=True), exp)

    def test_value_counts_categorical_not_ordered(self):
        values = pd.Categorical([1, 2, 3, 1, 1, 3], ordered=False)

        exp_idx = pd.CategoricalIndex([1, 3, 2], categories=[1, 2, 3],
                                      ordered=False)
        exp = pd.Series([3, 2, 1], index=exp_idx, name='xxx')

        s = pd.Series(values, name='xxx')
        tm.assert_series_equal(s.value_counts(), exp)
        # check CategoricalIndex outputs the same result
        idx = pd.CategoricalIndex(values, name='xxx')
        tm.assert_series_equal(idx.value_counts(), exp)

        # normalize
        exp = pd.Series(np.array([3., 2., 1]) / 6.,
                        index=exp_idx, name='xxx')
        tm.assert_series_equal(s.value_counts(normalize=True), exp)
        tm.assert_series_equal(idx.value_counts(normalize=True), exp)


@pytest.fixture
def s_main_dtypes():
    df = pd.DataFrame(
        {'datetime': pd.to_datetime(['2003', '2002',
                                     '2001', '2002',
                                     '2005']),
         'datetimetz': pd.to_datetime(
             ['2003', '2002',
              '2001', '2002',
              '2005']).tz_localize('US/Eastern'),
         'timedelta': pd.to_timedelta(['3d', '2d', '1d',
                                       '2d', '5d'])})

    for dtype in ['int8', 'int16', 'int32', 'int64',
                  'float32', 'float64',
                  'uint8', 'uint16', 'uint32', 'uint64']:
        df[dtype] = Series([3, 2, 1, 2, 5], dtype=dtype)

    return df


class TestNLargestNSmallest(object):

    @pytest.mark.parametrize(
        "r", [Series([3., 2, 1, 2, '5'], dtype='object'),
              Series([3., 2, 1, 2, 5], dtype='object'),
              # not supported on some archs
              # Series([3., 2, 1, 2, 5], dtype='complex256'),
              Series([3., 2, 1, 2, 5], dtype='complex128'),
              Series(list('abcde')),
              Series(list('abcde'), dtype='category')])
    def test_error(self, r):
        dt = r.dtype
        msg = ("Cannot use method 'n(larg|small)est' with "
               "dtype {dt}".format(dt=dt))
        args = 2, len(r), 0, -1
        methods = r.nlargest, r.nsmallest
        for method, arg in product(methods, args):
            with tm.assert_raises_regex(TypeError, msg):
                method(arg)

    @pytest.mark.parametrize(
        "s",
        [v for k, v in s_main_dtypes().iteritems()])
    def test_nsmallest_nlargest(self, s):
        # float, int, datetime64 (use i8), timedelts64 (same),
        # object that are numbers, object that are strings

        assert_series_equal(s.nsmallest(2), s.iloc[[2, 1]])
        assert_series_equal(s.nsmallest(2, keep='last'), s.iloc[[2, 3]])

        empty = s.iloc[0:0]
        assert_series_equal(s.nsmallest(0), empty)
        assert_series_equal(s.nsmallest(-1), empty)
        assert_series_equal(s.nlargest(0), empty)
        assert_series_equal(s.nlargest(-1), empty)

        assert_series_equal(s.nsmallest(len(s)), s.sort_values())
        assert_series_equal(s.nsmallest(len(s) + 1), s.sort_values())
        assert_series_equal(s.nlargest(len(s)), s.iloc[[4, 0, 1, 3, 2]])
        assert_series_equal(s.nlargest(len(s) + 1),
                            s.iloc[[4, 0, 1, 3, 2]])

    def test_misc(self):

        s = Series([3., np.nan, 1, 2, 5])
        assert_series_equal(s.nlargest(), s.iloc[[4, 0, 3, 2]])
        assert_series_equal(s.nsmallest(), s.iloc[[2, 3, 0, 4]])

        msg = 'keep must be either "first", "last"'
        with tm.assert_raises_regex(ValueError, msg):
            s.nsmallest(keep='invalid')
        with tm.assert_raises_regex(ValueError, msg):
            s.nlargest(keep='invalid')

        # GH 15297
        s = Series([1] * 5, index=[1, 2, 3, 4, 5])
        expected_first = Series([1] * 3, index=[1, 2, 3])
        expected_last = Series([1] * 3, index=[5, 4, 3])

        result = s.nsmallest(3)
        assert_series_equal(result, expected_first)

        result = s.nsmallest(3, keep='last')
        assert_series_equal(result, expected_last)

        result = s.nlargest(3)
        assert_series_equal(result, expected_first)

        result = s.nlargest(3, keep='last')
        assert_series_equal(result, expected_last)

    @pytest.mark.parametrize('n', range(1, 5))
    def test_n(self, n):

        # GH 13412
        s = Series([1, 4, 3, 2], index=[0, 0, 1, 1])
        result = s.nlargest(n)
        expected = s.sort_values(ascending=False).head(n)
        assert_series_equal(result, expected)

        result = s.nsmallest(n)
        expected = s.sort_values().head(n)
        assert_series_equal(result, expected)