如何根据列值对numpy数组进行排序？

Question

我有一个numpy的阵列看起来像这样：

>>> array_text[:10] 
array([[1, 52.51, 12, 0], 
    [1, 52.52, 52, 2], 
    [1, 52.53, 1, 6], 
    [2, 52.51, 20, 0], 
    [2, 52.52, 75, 76], 
    [2, 52.53, 6, 33], 
    [3, 52.51, 84, 0], 
    [3, 52.52, 39, 68], 
    [3, 52.53, 0, 13], 
    [4, 52.51, 1, 0]], dtype=object) 

我想要做的就是按照其第二列排序，按降序排列。所以我想要的结果是：

>>> array_text[:10] 
array([[1, 52.53, 1, 6], 
    [1, 52.52, 52, 2], 
    [1, 52.51, 12, 0], 
    [2, 52.53, 6, 33], 
    [2, 52.52, 75, 76], 
    [2, 52.51, 20, 0], 
    [3, 52.53, 0, 13], 
    [3, 52.52, 39, 68], 
    [3, 52.51, 84, 0], 
    [4, 52.51, 1, 0]], dtype=object) 

我该怎么做？

Answer 1

您可以使用numpy.lexsort。颠倒列的技巧是通过乘以-1来实现的。

import numpy as np 
arr[np.lexsort((-1*arr[:,1], arr[:,0]))] 

输出继电器：

array([[1, 52.53, 1, 6], 
     [1, 52.52, 52, 2], 
     [1, 52.51, 12, 0], 
     [2, 52.53, 6, 33], 
     [2, 52.52, 75, 76], 
     [2, 52.51, 20, 0], 
     [3, 52.53, 0, 13], 
     [3, 52.52, 39, 68], 
     [3, 52.51, 84, 0], 
     [4, 52.51, 1, 0]], dtype=object) 

Answer 2

处理此问题的最简单方法是使用熊猫，因为已使用sort_values方法将功能内置到数据框中。

import pandas as pd 
import numpy as np 

x = np.array([[1, 52.51, 12, 0], 
       [1, 52.52, 52, 2], 
       [1, 52.53, 1, 6], 
       [2, 52.51, 20, 0], 
       [2, 52.52, 75, 76], 
       [2, 52.53, 6, 33], 
       [3, 52.51, 84, 0], 
       [3, 52.52, 39, 68], 
       [3, 52.53, 0, 13], 
       [4, 52.51, 1, 0]], dtype=object) 

df = pd.DataFrame(x, columns=list('ABCD')) 
df.sort_values(['A', 'B'], ascending=[True, False]).values 
# returns: 
array([[1, 52.53, 1, 6], 
     [1, 52.52, 52, 2], 
     [1, 52.51, 12, 0], 
     [2, 52.53, 6, 33], 
     [2, 52.52, 75, 76], 
     [2, 52.51, 20, 0], 
     [3, 52.53, 0, 13], 
     [3, 52.52, 39, 68], 
     [3, 52.51, 84, 0], 
     [4, 52.51, 1, 0]], dtype=object)