CODEWITHSUNDEEP
pythonlistsortingnumpy
Mark Kennedy
Mark Kennedy

Reputation: 1781

Python numpy argsort

Suppose I have the following lists

['foo', '333', 32.3]
['bar', 4.0]
['baz', '555', '2232',  -1.9]

I want to be able to sort this via the last element (float)

['baz', '555', '2232',  -1.9]
['bar', 4.0]
['foo', '333', 32.3]

In ascending order

count_array = np.array([('foo', '333', 32.3),('bar', 4.0),('baz', '555', '2232', -1.9)], dtype = np.object)

idx = np.argsort(count_array[:, 1])

print(count_array[idx])

I want to be able to sort a 2-D list by comparing its last element. This code ONLY works if the sublists are the same length.

How can I get it to work for variable length sublists?

The problem is this line idx = np.argsort(count_array[:, 1])

Upvotes: 2

Views: 854

Answers (1)

Akavall
Akavall

Reputation: 86188

I would just use plain lists here, when you create a numpy.array with sub-lists of different lengths (or different data types in your array) you get an array of type object, which lacks many useful numpy features and is rarely a good idea to implement.

x = [['foo', '333', 32.3],
     ['bar', 4.0],
     ['baz', '555', '2232',  -1.9]]

result = sorted(x, key=lambda k : k[-1], reverse=True)

Result:

>>> result
[['foo', '333', 32.3], ['bar', 4.0], ['baz', '555', '2232', -1.9]]

Upvotes: 2

Related Questions