按列对二维 numpy 字符串/浮点数组进行排序

Question

I have the following numpy array (of numbers of rocket launches pr country since 1957), and I would like to sort it in ascending order on number of launches.

   ['Australia', 6.0],
   ['Brazil', 3.0],
   ['China', 269.0],
   ['France', 303.0],
   ['India', 76.0],
   ['Iran', 14.0],
   ['Israel', 11.0],
   ['Japan', 126.0],
   ['Kazakhstan', 701.0],
   ['Kenya', 9.0],
   ['New Zealand', 13.0],
   ['North Korea', 5.0],
   ['Pacific Ocean', 36.0],
   ['Russian Federation', 1398.0],
   ['South Korea', 3.0],
   ['USA', 1351.0]

Problem is, np.sort(a, axis = 0) only sort the values, but countries are not linked, so ei North Korea has launched 269 rockets (which is probably more likely than 5)

Or, if I do np.sort(a, axis = 1) then I get an error saying

TypeError: '<' not supported between instances of 'float' and 'str'

Any Ideas would be very much appreciated!

Answer 1

import numpy as np

data = [
   ['Australia', 6.0], 
   ['Brazil', 3.0],
   ['China', 269.0],
   ['France', 303.0],
   ['India', 76.0],
   ['Iran', 14.0],
   ['Israel', 11.0],
   ['Japan', 126.0],
   ['Kazakhstan', 701.0],
   ['Kenya', 9.0],
   ['New Zealand', 13.0],
   ['North Korea', 5.0],
   ['Pacific Ocean', 36.0],
   ['Russian Federation', 1398.0],
   ['South Korea', 3.0],
   ['USA', 1351.0]
]

We can create a structured array and then sort it by keys:

dtype = [
    ('name', '<U18'),    
    ('rockets', float)
]

data = np.array([tuple(x) for x in data], dtype=dtype) 
sorted_data = np.sort(data, order=['rockets'])          

print(sorted_data)

Answer 2

This is easy with python list sorting:

In [208]: alist = [   ['Australia', 6.0],
     ...:    ['Brazil', 3.0],
     ...:    ['China', 269.0],
     ...:    ['France', 303.0],
     ...:    ['India', 76.0],
     ...:    ['Iran', 14.0],
     ...:    ['Israel', 11.0],
     ...:    ['Japan', 126.0],
     ...:    ['Kazakhstan', 701.0],
     ...:    ['Kenya', 9.0],
     ...:    ['New Zealand', 13.0],
     ...:    ['North Korea', 5.0],
     ...:    ['Pacific Ocean', 36.0],
     ...:    ['Russian Federation', 1398.0],
     ...:    ['South Korea', 3.0],
     ...:    ['USA', 1351.0]]
In [209]: newlist = sorted(alist, key=lambda x: x[1])
In [210]: newlist
Out[210]: 
[['Brazil', 3.0],
 ['South Korea', 3.0],
 ['North Korea', 5.0],
 ['Australia', 6.0],
 ['Kenya', 9.0],
 ['Israel', 11.0],
 ['New Zealand', 13.0],
 ['Iran', 14.0],
 ['Pacific Ocean', 36.0],
 ['India', 76.0],
 ['Japan', 126.0],
 ['China', 269.0],
 ['France', 303.0],
 ['Kazakhstan', 701.0],
 ['USA', 1351.0],
 ['Russian Federation', 1398.0]]

With an object dtype array (to preserved string and float columns):

In [211]: arr = np.array(alist, object)
In [212]: arr
Out[212]: 
array([['Australia', 6.0],
       ['Brazil', 3.0],
       ['China', 269.0],
       ['France', 303.0],
       ...
       ['USA', 1351.0]], dtype=object)

Get a sorting index by just looking at the 2nd column:

In [213]: idx = np.argsort(arr[:,1])
In [214]: idx
Out[214]: array([ 1, 14, 11,  0,  9,  6, 10,  5, 12,  4,  7,  2,  3,  8, 15, 13])
In [215]: arr[idx]
Out[215]: 
array([['Brazil', 3.0],
       ['South Korea', 3.0],
       ['North Korea', 5.0],
       ['Australia', 6.0],
       ['Kenya', 9.0],
       ...
       ['Russian Federation', 1398.0]], dtype=object)

The structured array approach in the other answer is fine too.