Python Pandas.Series.asof : Cannot compare type 'Timestamp' with type 'struct_time'

Question

我在传递 datetime 时遇到问题进入Pandas.Series.asof :

def valueAsOf(self, date):
    if type(date) is str:
        return time.strptime(date, '%Y%m%d')
    return self.__series.index.asof(date)

然后出现如下错误

Traceback (most recent call last):
  File "/Users/x/test.py", line 106, in <module>
    print someTs.series.asof('20150101')
  File "/Users/x/anaconda/envs/test/lib/python2.7/site-packages/pandas/core/series.py", line 2460, in asof
    if where < start:
  File "pandas/tslib.pyx", line 836, in pandas.tslib._Timestamp.__richcmp__ (pandas/tslib.c:15612)
TypeError: Cannot compare type 'Timestamp' with type 'struct_time'

目前pandas.Series的索引是<type 'time.struct_time'>


转换date会不会解决问题？来自 struct_time至 Timestamp ？如果可以，如何实现？

---

我尝试转换 date到 datetime对象，但仍然收到相同的错误消息，即使 type(dt)显示其 <type 'datetime.datetime'>

dt = datetime.fromtimestamp(mktime(date))
return self.__series.index.asof(dt)

Answer 1

Pandas 时间序列应该总是有一个 DatetimeIndex 类型的索引。您发布的 TypeError，如果系列索引是 dtype object 的普通 Index 包含 time.struct_time，则可以重现。

例如，这会重现 TypeError:

import datetime as DT
import time
import numpy as np
import pandas as pd

x = pd.date_range('2015-1-1', '2015-1-5', freq='D')
index = [date.timetuple() for date in x.to_pydatetime()]
ts = pd.Series(np.arange(len(index)), index=index)
try:
    print(ts.asof('20150101'))
except TypeError as err:
    print(err)

产量

Cannot compare type 'Timestamp' with type 'struct_time'

可以通过将 ts.index 转换为 DatetimeIndex 来解决此问题:

ts.index = pd.to_datetime([DT.datetime.fromtimestamp(time.mktime(item)) 
                           for item in ts.index])

然后

print(ts.asof('20150101'))

打印与日期 20150101 关联的 ts 的值:

0

更好的是，不要使用时间元组。将日期序列转换为DatetimeIndex 越早越好:

In [59]: pd.to_datetime(['20150101'])
Out[59]: DatetimeIndex(['2015-01-01'], dtype='datetime64[ns]', freq=None, tz=None)