Question

我有一个庞大的Pandas Timestamps数组：

array([[Timestamp('2016-05-02 15:50:00+0000', tz='UTC', offset='5T'),
        Timestamp('2016-05-02 15:50:00+0000', tz='UTC', offset='5T'),
        Timestamp('2016-05-02 15:50:00+0000', tz='UTC', offset='5T')],
       [Timestamp('2016-05-02 17:10:00+0000', tz='UTC', offset='5T'),
        Timestamp('2016-05-02 17:10:00+0000', tz='UTC', offset='5T'),
        Timestamp('2016-05-02 17:10:00+0000', tz='UTC', offset='5T')],
       [Timestamp('2016-05-02 20:25:00+0000', tz='UTC', offset='5T'),
        Timestamp('2016-05-02 20:25:00+0000', tz='UTC', offset='5T'),
        Timestamp('2016-05-02 20:25:00+0000', tz='UTC', offset='5T')]], dtype=object)

我无法从此数组创建DataFrame，因为尝试这样做会引发以下错误：

AssertionError: Number of Block dimensions (1) must equal number of axes (2)

您可以看到数组显然是二维的，我使用ndim验证了这一点。

为什么我不能创建DataFrame？

Answer 1

我认为你可以使用list理解：

import pandas as pd
import numpy as np

a =np.array([[pd.Timestamp('2016-05-02 15:50:00+0000', tz='UTC', offset='5T'),
        pd.Timestamp('2016-05-02 15:50:00+0000', tz='UTC', offset='5T'),
        pd.Timestamp('2016-05-02 15:50:00+0000', tz='UTC', offset='5T')],
       [pd.Timestamp('2016-05-02 17:10:00+0000', tz='UTC', offset='5T'),
        pd.Timestamp('2016-05-02 17:10:00+0000', tz='UTC', offset='5T'),
        pd.Timestamp('2016-05-02 17:10:00+0000', tz='UTC', offset='5T')],
       [pd.Timestamp('2016-05-02 20:25:00+0000', tz='UTC', offset='5T'),
        pd.Timestamp('2016-05-02 20:25:00+0000', tz='UTC', offset='5T'),
        pd.Timestamp('2016-05-02 20:25:00+0000', tz='UTC', offset='5T')]], dtype=object)

df = pd.DataFrame([x for x in a], columns=['a','b','c'])
print (df)
                          a                         b  \
0 2016-05-02 15:50:00+00:00 2016-05-02 15:50:00+00:00   
1 2016-05-02 17:10:00+00:00 2016-05-02 17:10:00+00:00   
2 2016-05-02 20:25:00+00:00 2016-05-02 20:25:00+00:00   

                          c  
0 2016-05-02 15:50:00+00:00  
1 2016-05-02 17:10:00+00:00  
2 2016-05-02 20:25:00+00:00

另一个解决方案是DataFrame.from_records：

print (pd.DataFrame.from_records(a, columns=['a','b','c']))
                          a                         b  \
0 2016-05-02 15:50:00+00:00 2016-05-02 15:50:00+00:00   
1 2016-05-02 17:10:00+00:00 2016-05-02 17:10:00+00:00   
2 2016-05-02 20:25:00+00:00 2016-05-02 20:25:00+00:00   

                          c  
0 2016-05-02 15:50:00+00:00  
1 2016-05-02 17:10:00+00:00  
2 2016-05-02 20:25:00+00:00

请参阅alternate constructors of df。

Pandas无法从Numpy时间戳数组创建DataFrame

1 个答案: