Question

做的时候：

import pandas
from datetime import datetime
timestampparse = lambda t: datetime.fromtimestamp(float(t))
df = pandas.read_csv('blah.csv', delimiter=';', parse_dates=True, date_parser=timestampparse, index_col='DateTime', names=['DateTime', 'Sell'], header=None)
print df.ix['2015-12-02 12:02:21.070':'2015-12-02 12:40:21.070']

使用此blah.csv文件：

1449054136.83;1.05905
1449054139.25;1.05906
1449054139.86;1.05906
1449054140.47;1.05906

我收到此错误：

KeyError异常

如何访问按日期过滤的pandas数据帧片段？

为什么df.ix['2015-12-02 12:02:19.000':'2015-12-02 12:40:21.070']无效？

Answer 1

用零'2015-12-02 12:02:16.0859'填充第二个分数：

>>> df['2015-12-02 12:02:16.0859':'2015-12-02 12:03:20'])
                              Sell
DateTime                           
2015-12-02 12:02:16.829999  1.05905
2015-12-02 12:02:19.250000  1.05906
2015-12-02 12:02:19.859999  1.05906
2015-12-02 12:02:20.470000  1.05906

这有效：

>>> df['2015-12-02 12:02:17':'2015-12-02 12:03:20']
                               Sell
DateTime                           
2015-12-02 12:02:19.250000  1.05906
2015-12-02 12:02:19.859999  1.05906
2015-12-02 12:02:20.470000  1.05906

这适用于版本0.16.2

>>> from datetime import datetime
>>> df[datetime(2015, 12, 2, 12, 2, 16):datetime(2015, 12, 2, 12, 2, 20)]

                               Sell
DateTime                           
2015-12-02 12:02:16.829999  1.05905
2015-12-02 12:02:19.250000  1.05906
2015-12-02 12:02:19.859999  1.05906

Answer 2

我认为它不起作用，因为datetimeindex和floatindex可能是精确问题。

您可以使用partial string indexing，我在日期时间结束时省略数字 - 我只使用秒数：

print df['2015-12-02 12:02:19':'2015-12-02 12:40:20']

                            Sell
DateTime                        
2015-12-02 12:02:19.250  1.05906
2015-12-02 12:02:19.860  1.05906
2015-12-02 12:02:20.470  1.05906

Answer 3

从docs和Time/Date Components我知道你需要指定微秒的数量（与datetime对象相同）：

In [103]: df.loc["2015-12-02 14:02:10":"2015-12-02 14:02:19.899999"]
Out[103]:
                               Sell
DateTime
2015-12-02 14:02:16.829999  1.05905
2015-12-02 14:02:19.250000  1.05906
2015-12-02 14:02:19.859999  1.05906

或使用datetime指定精确的微秒数：

In [104]: df.loc["2015-12-02 14:02:10":datetime(year=2015, month=12, day=2, hour=14, minute=2, second=20, microsecond=999999)]
Out[104]:
                               Sell
DateTime
2015-12-02 14:02:16.829999  1.05905
2015-12-02 14:02:19.250000  1.05906
2015-12-02 14:02:19.859999  1.05906
2015-12-02 14:02:20.470000  1.05906

Pandas：通过按日期过滤来访问行

3 个答案: