Python numpy.timedelta64函数代码示例

    def test_timedelta64_conversions(self):
        startdate = Series(date_range('2013-01-01', '2013-01-03'))
        enddate = Series(date_range('2013-03-01', '2013-03-03'))

        s1 = enddate - startdate
        s1[2] = np.nan

        for m in [1, 3, 10]:
            for unit in ['D', 'h', 'm', 's', 'ms', 'us', 'ns']:

                # op
                expected = s1.apply(lambda x: x / np.timedelta64(m, unit))
                result = s1 / np.timedelta64(m, unit)
                assert_series_equal(result, expected)

                if m == 1 and unit != 'ns':

                    # astype
                    result = s1.astype("timedelta64[{0}]".format(unit))
                    assert_series_equal(result, expected)

                # reverse op
                expected = s1.apply(
                    lambda x: Timedelta(np.timedelta64(m, unit)) / x)
                result = np.timedelta64(m, unit) / s1

        # astype
        s = Series(date_range('20130101', periods=3))
        result = s.astype(object)
        self.assertIsInstance(result.iloc[0], datetime)
        self.assertTrue(result.dtype == np.object_)

        result = s1.astype(object)
        self.assertIsInstance(result.iloc[0], timedelta)
        self.assertTrue(result.dtype == np.object_)

    def test_timestamp_and_series(self):
        timestamp_series = Series(date_range('2014-03-17', periods=2, freq='D', tz='US/Eastern'))
        first_timestamp = timestamp_series[0]

        delta_series = Series([np.timedelta64(0, 'D'), np.timedelta64(1, 'D')])
        assert_series_equal(timestamp_series - first_timestamp, delta_series)
        assert_series_equal(first_timestamp - timestamp_series, -delta_series)

    def test_ufunc_coercions(self):
        idx = date_range('2011-01-01', periods=3, freq='2D', name='x')

        delta = np.timedelta64(1, 'D')
        for result in [idx + delta, np.add(idx, delta)]:
            assert isinstance(result, DatetimeIndex)
            exp = date_range('2011-01-02', periods=3, freq='2D', name='x')
            tm.assert_index_equal(result, exp)
            assert result.freq == '2D'

        for result in [idx - delta, np.subtract(idx, delta)]:
            assert isinstance(result, DatetimeIndex)
            exp = date_range('2010-12-31', periods=3, freq='2D', name='x')
            tm.assert_index_equal(result, exp)
            assert result.freq == '2D'

        delta = np.array([np.timedelta64(1, 'D'), np.timedelta64(2, 'D'),
                          np.timedelta64(3, 'D')])
        for result in [idx + delta, np.add(idx, delta)]:
            assert isinstance(result, DatetimeIndex)
            exp = DatetimeIndex(['2011-01-02', '2011-01-05', '2011-01-08'],
                                freq='3D', name='x')
            tm.assert_index_equal(result, exp)
            assert result.freq == '3D'

        for result in [idx - delta, np.subtract(idx, delta)]:
            assert isinstance(result, DatetimeIndex)
            exp = DatetimeIndex(['2010-12-31', '2011-01-01', '2011-01-02'],
                                freq='D', name='x')
            tm.assert_index_equal(result, exp)
            assert result.freq == 'D'

 def test_timedelta_conversions(self):
     assert (Timedelta(timedelta(seconds=1)) ==
             np.timedelta64(1, 's').astype('m8[ns]'))
     assert (Timedelta(timedelta(microseconds=1)) ==
             np.timedelta64(1, 'us').astype('m8[ns]'))
     assert (Timedelta(timedelta(days=1)) ==
             np.timedelta64(1, 'D').astype('m8[ns]'))

def _as_timedelta64_scalar(time, unit=None):
    unit_args = [unit] if unit else []
    flt_unit = unit if unit else 's'
    # turn 'H:M:S.ms', 'M:S.ms', 'S.ms' into floating point seconds
    if isinstance(time, string_types):# and ':' in time:
        time = [float(t) for t in time.lstrip('T').split(':')][::-1]
        if len(time) > 1 and unit is not None:
            raise ValueError("When giving time as a string, units are automatic")
        if len(time) > 3:
            raise ValueError("Timedelta as string only goes up to hours")
        t_flt = 0.0
        for factor, t in zip([1, 60, 60 * 60], time):
            t_flt += factor * t
        time = t_flt
        flt_unit = 's'
    # turn floating point time into integer with the correct unit
    if is_datetime_like(time):
        time = as_datetime64(time) - as_datetime64(np.timedelta64(0, 's'))
    elif isinstance(time, (np.timedelta64, timedelta)):
        time = np.timedelta64(time).astype(_format_unit(unit, base=DELTA_BASE))
    elif isinstance(time, (int, float, np.integer, np.floating)):
        orig_time, orig_flt_unit = time, flt_unit
        unit_idx = TIME_UNITS.index(flt_unit)
        while not np.isclose(time, int(np.round(time)), rtol=1e-4, atol=1e-18):
            if unit_idx <= 0:
                raise ValueError("Floating point time {0} [{1}] is too precise "
                                 "for any time unit?".format(orig_time, orig_flt_unit))
            unit_idx -= 1
            time *= TIME_SCALE[unit_idx]
            flt_unit = TIME_UNITS[unit_idx]
        time = np.timedelta64(int(np.round(time)), flt_unit)
        unit, unit_args = flt_unit, [flt_unit]
    return np.timedelta64(time, *unit_args)

    def test_timedelta_ops_scalar(self):
        _skip_if_numpy_not_friendly()

        # GH 6808
        base = pd.to_datetime('20130101 09:01:12.123456')
        expected_add = pd.to_datetime('20130101 09:01:22.123456')
        expected_sub = pd.to_datetime('20130101 09:01:02.123456')

        for offset in [pd.to_timedelta(10,unit='s'),
                       timedelta(seconds=10),
                       np.timedelta64(10,'s'),
                       np.timedelta64(10000000000,'ns'),
                       pd.offsets.Second(10)]:
            result = base + offset
            self.assertEquals(result, expected_add)

            result = base - offset
            self.assertEquals(result, expected_sub)

        base = pd.to_datetime('20130102 09:01:12.123456')
        expected_add = pd.to_datetime('20130103 09:01:22.123456')
        expected_sub = pd.to_datetime('20130101 09:01:02.123456')

        for offset in [pd.to_timedelta('1 day, 00:00:10'),
                       pd.to_timedelta('1 days, 00:00:10'),
                       timedelta(days=1,seconds=10),
                       np.timedelta64(1,'D')+np.timedelta64(10,'s'),
                       pd.offsets.Day()+pd.offsets.Second(10)]:
            result = base + offset
            self.assertEquals(result, expected_add)

            result = base - offset
            self.assertEquals(result, expected_sub)

def rolling_correlation(data, correlant, window, min_good_ratio = 0.67, verbose = True):
    "time as here: http://docs.scipy.org/doc/numpy/reference/arrays.datetime.html#datetime-units"

    correlant = correlant.align_to(data) # I do align before merge, because it is more suffisticated!
    merged = data.copy()
    merged.data['correlant'] = correlant.data

    data_period = _np.timedelta64(int(merged._data_period), 's')
    window = _np.timedelta64(window[0], window[1])
    window = int(window/data_period)
    if verbose:
        print('Each window contains %s data points of which at least %s are not nan.'%(window, int(window * min_good_ratio)))

    min_good = window * min_good_ratio
    size = merged.data.shape[0]-window + 1
    timestamps = _pd.TimeSeries(_pd.to_datetime(_pd.Series(_np.zeros(size))))
    pear_r = _np.zeros(size)
    for i in range(size):
        secment = TimeSeries(merged.data.iloc[i:i+window,:])
        secment._data_period = merged._data_period
    #     print(secment.data.dropna().shape[0] < min_good)
        if secment.data.dropna().shape[0] < min_good:
            pear_r[i]= _np.nan
        else:
            corr = secment.correlate_to(secment, data_column=merged.data.columns[0], correlant_column=merged.data.columns[1])
            pear_r[i] = corr.pearson_r[0]
        timestamps.iloc[i] = secment.data.index[0] + ((secment.data.index[-1] - secment.data.index[0])/2.)
    #     break

    pear_r_ts = TimeSeries(_pd.DataFrame(pear_r, index = timestamps, columns = ['pearson_r']))
    pear_r_ts._data_period = merged._data_period
    pear_r_ts._y_label = 'r'
    return pear_r_ts

    def function(self, simulation, period):
        period = period.this_month
        assiette_allegement = simulation.calculate('assiette_allegement', period)
        contrat_de_travail_duree = simulation.calculate('contrat_de_travail_duree', period)  # 0: CDI, 1:CDD
        contrat_de_travail_debut = simulation.calculate('contrat_de_travail_debut', period)
        contrat_de_travail_fin = simulation.calculate('contrat_de_travail_fin', period)
        effectif_entreprise = simulation.calculate('effectif_entreprise', period)
        smic_proratise = simulation.calculate('smic_proratise', period)
        zone_revitalisation_rurale = simulation.calculate('zone_revitalisation_rurale', period)

        duree_cdd_eligible = contrat_de_travail_fin > contrat_de_travail_debut + timedelta64(365, 'D')
        # TODO: move to legislation parameters file
        contrat_de_travail_eligible = (
            contrat_de_travail_duree == 0) + (
            (contrat_de_travail_duree == 1) * (duree_cdd_eligible)
            )

        duree_validite = (
            datetime64(period.start) + timedelta64(1, 'D') - contrat_de_travail_debut
            ).astype('timedelta64[Y]') < timedelta64(1, 'Y')

        eligible = (
            contrat_de_travail_eligible *
            (effectif_entreprise <= 50) *
            zone_revitalisation_rurale *
            duree_validite
            )
        taux_max = .281 if period.start.year < 2015 else .2655  # TODO: move to legislation parameters file
        seuil_max = 2.4
        seuil_min = 1.5
        taux_exoneration = compute_taux_exoneration(assiette_allegement, smic_proratise, taux_max, seuil_max, seuil_min)
        exoneration_cotisations_zrr = taux_exoneration * assiette_allegement * eligible

        return period, exoneration_cotisations_zrr

    def test_timedelta_ops_scalar(self):
        # GH 6808
        base = pd.to_datetime('20130101 09:01:12.123456')
        expected_add = pd.to_datetime('20130101 09:01:22.123456')
        expected_sub = pd.to_datetime('20130101 09:01:02.123456')

        for offset in [pd.to_timedelta(10, unit='s'), timedelta(seconds=10),
                       np.timedelta64(10, 's'),
                       np.timedelta64(10000000000, 'ns'),
                       pd.offsets.Second(10)]:
            result = base + offset
            assert result == expected_add

            result = base - offset
            assert result == expected_sub

        base = pd.to_datetime('20130102 09:01:12.123456')
        expected_add = pd.to_datetime('20130103 09:01:22.123456')
        expected_sub = pd.to_datetime('20130101 09:01:02.123456')

        for offset in [pd.to_timedelta('1 day, 00:00:10'),
                       pd.to_timedelta('1 days, 00:00:10'),
                       timedelta(days=1, seconds=10),
                       np.timedelta64(1, 'D') + np.timedelta64(10, 's'),
                       pd.offsets.Day() + pd.offsets.Second(10)]:
            result = base + offset
            assert result == expected_add

            result = base - offset
            assert result == expected_sub

def transform_data(train, test):
    """ Transform train and test data to include new variables.
    """
    
    # Time
    initial_date = np.datetime64('2014-01-01T01:01', dtype='datetime64[m]')  # Arbitrary date chosen
    d_times = pd.DatetimeIndex(initial_date + np.timedelta64(int(mn), 'm') for mn in train['time'].values)    
    train['hour'] = d_times.hour
    train['weekday'] = d_times.weekday
    train['day_of_month'] = d_times.day
    train['month'] = d_times.month
    train['year'] = d_times.year

    d_times = pd.DatetimeIndex(initial_date + np.timedelta64(int(mn), 'm') for mn in test['time'].values)    
    test['hour'] = d_times.hour
    test['weekday'] = d_times.weekday
    test['day_of_month'] = d_times.day
    test['month'] = d_times.month
    test['year'] = d_times.year  
            
    # Accuracy 
    train['accuracy'] = np.log10(train['accuracy']) * 10.0
    test['accuracy'] = np.log10(test['accuracy']) * 10.0
      
    # Combine x and y attributes
    eps = 0.00001  
    train['x_d_y'] = train.x.values / (train.y.values + eps) 
    test['x_d_y'] = test.x.values / (test.y.values + eps)    
    
    train['x_t_y'] = train.x.values * train.y.values 
    test['x_t_y'] = test.x.values * test.y.values 
            
    # Return data
    return train, test  

def test_decode_standard_calendar_multidim_time_inside_timestamp_range(
        calendar, enable_cftimeindex):
    if enable_cftimeindex:
        pytest.importorskip('cftime')

    cftime = _import_cftime()

    units = 'days since 0001-01-01'
    times1 = pd.date_range('2001-04-01', end='2001-04-05', freq='D')
    times2 = pd.date_range('2001-05-01', end='2001-05-05', freq='D')
    noleap_time1 = cftime.date2num(times1.to_pydatetime(),
                                   units, calendar=calendar)
    noleap_time2 = cftime.date2num(times2.to_pydatetime(),
                                   units, calendar=calendar)
    mdim_time = np.empty((len(noleap_time1), 2), )
    mdim_time[:, 0] = noleap_time1
    mdim_time[:, 1] = noleap_time2

    expected1 = times1.values
    expected2 = times2.values

    actual = coding.times.decode_cf_datetime(
        mdim_time, units, calendar=calendar,
        enable_cftimeindex=enable_cftimeindex)
    assert actual.dtype == np.dtype('M8[ns]')

    abs_diff1 = abs(actual[:, 0] - expected1)
    abs_diff2 = abs(actual[:, 1] - expected2)
    # once we no longer support versions of netCDF4 older than 1.1.5,
    # we could do this check with near microsecond accuracy:
    # https://github.com/Unidata/netcdf4-python/issues/355
    assert (abs_diff1 <= np.timedelta64(1, 's')).all()
    assert (abs_diff2 <= np.timedelta64(1, 's')).all()

    def get_holidays(self, start, end, cal="FX"):
        # TODO use Pandas CustomBusinessDays to get more calendars
        holidays_list = []

        if cal == "FX":
            # filter for Christmas & New Year's Day
            for i in range(1970, 2020):
                holidays_list.append(str(i) + "-12-25")
                holidays_list.append(str(i) + "-01-01")

        if cal == "WEEKDAY":
            bday = CustomBusinessDay(weekmask="Sat Sun")

            holidays_list = pandas.date_range(start, end, freq=bday)

        holidays_list = pandas.to_datetime(holidays_list).order()

        # floor start date
        start = np.datetime64(start) - np.timedelta64(1, "D")

        # ceiling end date
        end = np.datetime64(end) + np.timedelta64(1, "D")

        holidays_list = [x for x in holidays_list if x >= start and x <= end]

        return pandas.to_datetime(holidays_list)

def PlotEwmaPredictions(daily, name):
    """
    """

    # use EWMA to estimate slopes
    filled = timeseries.FillMissing(daily)
    filled['slope'] = pandas.ewma(filled.ppg.diff(), span=180)
    filled[-1:]

    # extract the last inter and slope
    start = filled.index[-1]
    inter = filled.ewma[-1]
    slope = filled.slope[-1]

    # reindex the DataFrame, adding a year to the end
    dates = pandas.date_range(filled.index.min(), 
                              filled.index.max() + np.timedelta64(365, 'D'))
    predicted = filled.reindex(dates)

    # generate predicted values and add them to the end
    predicted['date'] = predicted.index
    one_day = np.timedelta64(1, 'D')
    predicted['days'] = (predicted.date - start) / one_day
    predict = inter + slope * predicted.days
    predicted.ewma.fillna(predict, inplace=True)

    # plot the actual values and predictions
    thinkplot.Scatter(daily.ppg, alpha=0.1, label=name)
    thinkplot.Plot(predicted.ewma)
    thinkplot.Save()

    def test_timedelta(self, freq):
        index = date_range('1/1/2000', periods=50, freq=freq)

        shifted = index + timedelta(1)
        back = shifted + timedelta(-1)
        tm.assert_index_equal(index, back)

        if freq == 'D':
            expected = pd.tseries.offsets.Day(1)
            assert index.freq == expected
            assert shifted.freq == expected
            assert back.freq == expected
        else:  # freq == 'B'
            assert index.freq == pd.tseries.offsets.BusinessDay(1)
            assert shifted.freq is None
            assert back.freq == pd.tseries.offsets.BusinessDay(1)

        result = index - timedelta(1)
        expected = index + timedelta(-1)
        tm.assert_index_equal(result, expected)

        # GH4134, buggy with timedeltas
        rng = date_range('2013', '2014')
        s = Series(rng)
        result1 = rng - pd.offsets.Hour(1)
        result2 = DatetimeIndex(s - np.timedelta64(100000000))
        result3 = rng - np.timedelta64(100000000)
        result4 = DatetimeIndex(s - pd.offsets.Hour(1))
        tm.assert_index_equal(result1, result4)
        tm.assert_index_equal(result2, result3)

    def test_nat_items(self):
        # not a datetime
        nadt_no_unit = np.datetime64("NaT")
        nadt_s = np.datetime64("NaT", "s")
        nadt_d = np.datetime64("NaT", "ns")
        # not a timedelta
        natd_no_unit = np.timedelta64("NaT")
        natd_s = np.timedelta64("NaT", "s")
        natd_d = np.timedelta64("NaT", "ns")

        dts = [nadt_no_unit, nadt_s, nadt_d]
        tds = [natd_no_unit, natd_s, natd_d]
        for a, b in itertools.product(dts, dts):
            self._assert_func(a, b)
            self._assert_func([a], [b])
            self._test_not_equal([a], b)

        for a, b in itertools.product(tds, tds):
            self._assert_func(a, b)
            self._assert_func([a], [b])
            self._test_not_equal([a], b)

        for a, b in itertools.product(tds, dts):
            self._test_not_equal(a, b)
            self._test_not_equal(a, [b])
            self._test_not_equal([a], [b])
            self._test_not_equal([a], np.datetime64("2017-01-01", "s"))
            self._test_not_equal([b], np.datetime64("2017-01-01", "s"))
            self._test_not_equal([a], np.timedelta64(123, "s"))
            self._test_not_equal([b], np.timedelta64(123, "s"))