python - Pandas DataFrame 在索引上失败但 Series 成功

Question

鉴于美国市场时间:

In [220]: market_hours = pandas.date_range(date + ' 09:30:00', date + ' 16:00:00', freq='15min', tz='US/Eastern').tz_convert('UTC')

In [221]: market_hours
Out[221]:
<class 'pandas.tseries.index.DatetimeIndex'>
[2014-04-29 13:30:00+00:00, ..., 2014-04-29 20:00:00+00:00]
Length: 27, Freq: 15T, Timezone: UTC

我可以resample() 单个字段并限制这些市场时间:

In [222]: df.set_index('localtime')['size'].resample('15min', how='sum')[market_hours]
Out[222]:
2014-04-29 13:30:00+00:00    1093142
2014-04-29 13:45:00+00:00     556664
2014-04-29 14:00:00+00:00     467662
2014-04-29 14:15:00+00:00     460966
2014-04-29 14:30:00+00:00     275805
2014-04-29 14:45:00+00:00     192709
2014-04-29 15:00:00+00:00     226375
2014-04-29 15:15:00+00:00     175065
2014-04-29 15:30:00+00:00     181047
2014-04-29 15:45:00+00:00     129644
2014-04-29 16:00:00+00:00     193330
2014-04-29 16:15:00+00:00     170046
2014-04-29 16:30:00+00:00     130674
2014-04-29 16:45:00+00:00     107118
2014-04-29 17:00:00+00:00     156699
2014-04-29 17:15:00+00:00     153912
2014-04-29 17:30:00+00:00     180449
2014-04-29 17:45:00+00:00     223318
2014-04-29 18:00:00+00:00     211324
2014-04-29 18:15:00+00:00     152374
2014-04-29 18:30:00+00:00     121876
2014-04-29 18:45:00+00:00      90891
2014-04-29 19:00:00+00:00     138222
2014-04-29 19:15:00+00:00     167571
2014-04-29 19:30:00+00:00     264658
2014-04-29 19:45:00+00:00     492528
2014-04-29 20:00:00+00:00       8354
Freq: 15T, Name: size, dtype: int64

但是，如果我尝试 resample() 一个字段数组，我会得到一个错误:

In [223]: df.set_index('localtime')[['size']].resample('15min', how='sum')[market_hours]
...

KeyError: "['2014-04-29T09:30:00.000000000-0400' '2014-04-29T09:45:00.000000000-0400'\n '2014-04-29T10:00:00.000000000-0400' '2014-04-29T10:15:00.000000000-0400'\n '2014-04-29T10:30:00.000000000-0400' '2014-04-29T10:45:00.000000000-0400'\n '2014-04-29T11:00:00.000000000-0400' '2014-04-29T11:15:00.000000000-0400'\n '2014-04-29T11:30:00.000000000-0400' '2014-04-29T11:45:00.000000000-0400'\n '2014-04-29T12:00:00.000000000-0400' '2014-04-29T12:15:00.000000000-0400'\n '2014-04-29T12:30:00.000000000-0400' '2014-04-29T12:45:00.000000000-0400'\n '2014-04-29T13:00:00.000000000-0400' '2014-04-29T13:15:00.000000000-0400'\n '2014-04-29T13:30:00.000000000-0400' '2014-04-29T13:45:00.000000000-0400'\n '2014-04-29T14:00:00.000000000-0400' '2014-04-29T14:15:00.000000000-0400'\n '2014-04-29T14:30:00.000000000-0400' '2014-04-29T14:45:00.000000000-0400'\n '2014-04-29T15:00:00.000000000-0400' '2014-04-29T15:15:00.000000000-0400'\n '2014-04-29T15:30:00.000000000-0400' '2014-04-29T15:45:00.000000000-0400'\n '2014-04-29T16:00:00.000000000-0400'] not in index"

有没有一种方法可以按日期范围访问生成的 DataFrame？这似乎与时区无关。

Answer 1

在第一种情况下，您正在索引一个系列。在第二种情况下(使用 df[['size']].resample(..，注意双方括号)，您使用的是 DataFrame。
DataFrame 上的基本索引 (df[labels]) 将索引列，而不是行(参见 http://pandas.pydata.org/pandas-docs/stable/indexing.html#basics )。因此，您会收到标签不在(列)索引中的错误。

为了克服这个问题，您可以使用 loc(假设 result 是重采样的结果):

result.loc[market_hours, :]