如何将日期时间移动到日，周或小时结束

Question

以下是将所有时间都移动到月末的代码：

import numpy as np
import pandas as pd

times = np.array([
       '2013-07-22T02:10:32.000000000+0900',
       '2013-07-22T01:11:13.000000000+0900',
       '2013-07-21T23:23:32.000000000+0900',
       '2013-07-21T05:59:21.000000000+0900',
       '2013-07-21T05:57:30.000000000+0900',
       '2013-07-21T05:44:27.000000000+0900',
       '2013-07-20T10:45:17.000000000+0900',
       '2013-07-20T10:36:53.000000000+0900',
       '2013-07-20T09:57:46.000000000+0900',
       '2013-07-20T09:57:06.000000000+0900',
       '2013-07-20T09:30:57.000000000+0900',
       '2013-07-20T08:20:27.000000000+0900',], dtype='datetime64[ns]')

dti = pd.DatetimeIndex(times)
dti.shift(1, "M").values

结果是：

array(['2013-07-31T09:00:00.000000000+0900',
       '2013-07-31T09:00:00.000000000+0900',
       '2013-07-31T09:00:00.000000000+0900',
       '2013-07-31T09:00:00.000000000+0900',
       '2013-07-31T09:00:00.000000000+0900',
       '2013-07-31T09:00:00.000000000+0900',
       '2013-07-31T09:00:00.000000000+0900',
       '2013-07-31T09:00:00.000000000+0900',
       '2013-07-31T09:00:00.000000000+0900',
       '2013-07-31T09:00:00.000000000+0900',
       '2013-07-31T09:00:00.000000000+0900',
       '2013-07-31T09:00:00.000000000+0900'], dtype='datetime64[ns]')

但是如何将所有时间一直移动到小时，一天或一周结束？

Answer 1

我找到的最好的方法是to_period和to_timestamp ：

In [39]:

dti.to_period("W-SAT").to_timestamp(how="end").values

Out[39]:

array(['2013-07-27T09:00:00.000000000+0900',
       '2013-07-27T09:00:00.000000000+0900',
       '2013-07-27T09:00:00.000000000+0900',
       '2013-07-20T09:00:00.000000000+0900',
       '2013-07-20T09:00:00.000000000+0900',
       '2013-07-20T09:00:00.000000000+0900',
       '2013-07-20T09:00:00.000000000+0900',
       '2013-07-20T09:00:00.000000000+0900',
       '2013-07-20T09:00:00.000000000+0900',
       '2013-07-20T09:00:00.000000000+0900',
       '2013-07-20T09:00:00.000000000+0900',
       '2013-07-20T09:00:00.000000000+0900'], dtype='datetime64[ns]')

In [40]:

dti.to_period("H").to_timestamp(how="end").values

Out[40]:

array(['2013-07-22T02:59:59.000000000+0900',
       '2013-07-22T01:59:59.000000000+0900',
       '2013-07-21T23:59:59.000000000+0900',
       '2013-07-21T05:59:59.000000000+0900',
       '2013-07-21T05:59:59.000000000+0900',
       '2013-07-21T05:59:59.000000000+0900',
       '2013-07-20T10:59:59.000000000+0900',
       '2013-07-20T10:59:59.000000000+0900',
       '2013-07-20T09:59:59.000000000+0900',
       '2013-07-20T09:59:59.000000000+0900',
       '2013-07-20T09:59:59.000000000+0900',
       '2013-07-20T08:59:59.000000000+0900'], dtype='datetime64[ns]')

Answer 2

我同意安迪; 这不可能是shift的预期行为。 将时间转移到月末的更简洁方法是：

from pandas.tseries.offsets import MonthEnd
times = Series(times)
times.map(lambda x: x + MonthEnd())

但是没有HourEnd，DayEnd或WeekEnd这样的东西。 对于那些情况，如何遵循这种模式？

from pandas.tseries.offsets import Second, Minute, Hour, Day

times.map(lambda x: x + Minute(59-x.minute) + Second(59-x.second))

times.map(lambda x: x + Hour(23-x.hour) + Minute(59-x.minute) + Second(59-x.second))

times.map(lambda x: x + Day(6-x.weekday()) + Hour(23-x.hour) + \
          Minute(59-x.minute) + Second(59-x.second))

如果你想在一周的最后一天，但不一定是那一天的最后一秒 ，然后表达显然是简单。