美国大选日的熊猫假期日历规则

Question

在使用 Pandas Holiday 类创建假期日历时，我无法确定为美国选举日创建日历的正确规则。美国选举日定义为 11 月第一个星期一之后的星期二，每偶数年举行一次。使用定义的 Holiday 类：

class USElectionCalendar(AbstractHolidayCalendar):
"""        
Federal Presidential  and Congressional election day.
Tuesday following the first Monday, 2 to 8 November every two even numbered years.
Election Days can only occur from November 2nd through 8th inclusive.
"""

rules = [
    Holiday("Election Day",month=11, day=2, offset=pd.DateOffset(weekday=TU(1))),
]

与

start_date = '20160108'
end_date   = '20261231'

进入函数

def holidays_between_dates(calendar, start_date, end_date):

    cal = calendar
    dates = cal.holidays(start_date, end_date, return_name=True)

    return dates

返回

2016-11-08    Election Day
2017-11-07    Election Day
2018-11-06    Election Day
2019-11-05    Election Day
2020-11-03    Election Day
2021-11-02    Election Day
2022-11-08    Election Day
2023-11-07    Election Day
2024-11-05    Election Day
2025-11-04    Election Day
2026-11-03    Election Day

一切都很好，除了奇数年份。我已经尝试合并两个偏移量，如issue 中所述。为规则添加 2 年的偏移量

Holiday("Election Day", month=11, day=2, offset=[ pd.DateOffset(weekday=TU(1)), pd.DateOffset(years=2) ])

只需将第一个事件移到未来 2 年。我不确定所需的时间序列是否可行。那么问题来了：

是否可以直接构建这个日历，还是需要使用第二个函数从 Pandas 日历对象中删除奇数年份？

Answer 1

您可以在制作假期时使用遵守而不是偏移，并让它在奇数年返回 None：

def election_observance(dt):
    if dt.year % 2 == 1:
        return None
    else:
        return dt + pd.DateOffset(weekday=TU(1))


class USElectionCalendar(AbstractHolidayCalendar):
    """
    Federal Presidential  and Congressional election day.
    Tuesday following the first Monday, 2 to 8 November every two even numbered years.
    Election Days can only occur from November 2nd through 8th inclusive.
    """

    rules = [
        Holiday('Election Day', month=11, day=2, observance=election_observance)
    ]


cal = USElectionCalendar()

start_date = '20160108'
end_date = '20261231'

print cal.holidays(start_date, end_date, return_name=True)

输出：

2016-11-08    Election Day
2018-11-06    Election Day
2020-11-03    Election Day
2022-11-08    Election Day
2024-11-05    Election Day
2026-11-03    Election Day
dtype: object

请注意，您不想在构建 Holiday 时同时使用偏移量和遵守量，并且尝试这样做会在更新的 pandas 版本中引发异常。