Pandas:按日期时间(可能不存在)和返回视图对数据帧进行切片

Question

我有一个大的DataFrame，我想对它进行切片，这样我就可以对切片的数据帧执行一些计算，以便在原始数据帧中更新值。此外，我正在按索引中可能不存在的开始和结束时间对数据帧进行切片。下面是一个简化的例子，但我实际上想根据不同的计算更新一些列。

In [1]: df
Out[1]:
                         A        B         C
TIME
2014-01-02 14:00:00 -1.172285  1.706200    NaN
2014-01-02 14:05:00  0.039511 -0.320798    NaN
2014-01-02 14:10:00 -0.192179 -0.539397    NaN
2014-01-02 14:15:00 -0.475917 -0.280055    NaN
2014-01-02 14:20:00  0.163376  1.124602    NaN
2014-01-02 14:25:00 -2.477812  0.656750    NaN

我已经尝试了以下所有语句来创建sdf作为我的时间范围的视图：

start = datetime.strptime('2014-01-02 14:07:00', '%Y-%m-%d %H:%M:%S')
end = datetime.strptime('2014-01-02 14:22:00', '%Y-%m-%d %H:%M:%S')
sdf = df[start:end]
sdf = df[start < df.index < end]
sdf = df.ix[start:end]
sdf = df.loc[start:end]
sdf = df.truncate(before=start, after=end, copy=False)
sdf[C] == 100

大多数都会返回一份副本，我会收到SettingWithCopyWarning警告。loc函数表示索引与日期时间不兼容。这是我应该能做的事情吗？更新切片后我想要的结果是：

In [1]: df
Out[1]:
                         A        B         C
TIME
2014-01-02 14:00:00 -1.172285  1.706200    NaN
2014-01-02 14:05:00  0.039511 -0.320798    NaN
2014-01-02 14:10:00 -0.192179 -0.539397    100
2014-01-02 14:15:00 -0.475917 -0.280055    100
2014-01-02 14:20:00  0.163376  1.124602    100
2014-01-02 14:25:00 -2.477812  0.656750    NaN

有人能建议一种方法吗？我是不是走错路了？

感谢

Answer 1

一种方法是使用loc并将条件包装在括号中，然后使用逐位运算符&，当您比较值数组而不是单个值时，需要逐位运算符，由于运算符优先级，括号是必需的。然后，我们可以使用它来使用loc执行标签选择，并设置"C"列，如下所示：

In [15]:
import datetime as dt
start = dt.datetime.strptime('2014-01-02 14:07:00', '%Y-%m-%d %H:%M:%S')
end = dt.datetime.strptime('2014-01-02 14:22:00', '%Y-%m-%d %H:%M:%S')
df.loc[(df.index > start) & (df.index < end), 'C'] = 100
df
Out[15]:
                            A         B    C
TIME                                        
2014-01-02 14:00:00 -1.172285  1.706200  NaN
2014-01-02 14:05:00  0.039511 -0.320798  NaN
2014-01-02 14:10:00 -0.192179 -0.539397  100
2014-01-02 14:15:00 -0.475917 -0.280055  100
2014-01-02 14:20:00  0.163376  1.124602  100
2014-01-02 14:25:00 -2.477812  0.656750  NaN

如果我们看看你尝试的每种方法，以及它们为什么不起作用：

sdf = df[start:end] #  will raise KeyError if start and end are not present in index
sdf = df[start < df.index < end] #  will raise ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all(), this is because you are comparing arrays of values not a single scalar value
sdf = df.ix[start:end] # raises KeyError same as first example
sdf = df.loc[start:end] #  raises KeyError same as first example
sdf = df.truncate(before=start, after=end, copy=False) # generates correct result but operations on this will raise SettingWithCopyWarning as you've found

编辑

您可以将sdf设置为掩码，并将其与loc一起使用来设置"C"列：

In [7]:
import datetime as dt
start = dt.datetime.strptime('2014-01-02 14:07:00', '%Y-%m-%d %H:%M:%S')
end = dt.datetime.strptime('2014-01-02 14:22:00', '%Y-%m-%d %H:%M:%S')
sdf = (df.index > start) & (df.index < end)
df.loc[sdf,'C'] = 100
df
Out[7]:
                            A         B    C
TIME                                        
2014-01-02 14:00:00 -1.172285  1.706200  NaN
2014-01-02 14:05:00  0.039511 -0.320798  NaN
2014-01-02 14:10:00 -0.192179 -0.539397  100
2014-01-02 14:15:00 -0.475917 -0.280055  100
2014-01-02 14:20:00  0.163376  1.124602  100
2014-01-02 14:25:00 -2.477812  0.656750  NaN