AttributeError：“模块”对象没有属性“选择”

Question

I am using python3. First I use random.choice in Terminal, and it works.

Python 3.2.3 (default, Feb 27 2014, 21:31:18) 
[GCC 4.6.3] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import random
>>> x = [1, 2, 3]
>>> random.choice(x)
3

But when I run it in my script, I get the message:

 AttributeError: 'module' object has no attribute 'choice' 

Here is the code

import random
from scipy import *
from numpy import linalg as LA
import pickle
import operator


def new_pagerank_step(current_page, N, d, links):
    print(links[current_page])
    if random.rand() > 1 - d:
        next_page = random.choice(links[current_page])
    else:
        next_page = random.randint(0, N)
    return next_page


def pagerank_wikipedia_demo():
    with open("wikilinks.pickle", "rb") as f:
        titles, links = pickle.load(f)
    current_page = 0
    T = 1000
    N = len(titles)
    d = 0.4
    Result = {}
    result = []
    for i in range(T):
        result.append(current_page)
        current_page = new_pagerank_step(current_page, N, d, links)
    for i in range(N):
        result.count(i)
        Result[i] = result.count(i) / T
    Sorted_Result = sorted(Result.items(), key=operator.itemgetter(1))

pagerank_wikipedia_demo()

Here, links[i] ( i is an integer) is a list. When I run this script, it fails with the message mentioned above.

I have also checked that the name of the script is not random . And there is only one file named random.py in /usr/lib/python3.2/random.py

Why this happens?

Answer 1

You masked the module with the numpy.random object here:

import random
from scipy import *

The from scipy import * import brings a whole lot of names , including random :

>>> from scipy import *
>>> random
<module 'numpy.random' from '/Users/mj/Development/venvs/stackoverflow-2.7/lib/python2.7/site-packages/numpy/random/__init__.pyc'>

This replaced the random module.

Either don't use a wildcard import, or import the random module after importing everything from scipy .

You could also import choice from the random module and refer to it directly, or use a different name for the import to bind to:

from random import choice
from scipy import *

# use choice, not random.choice

or

import random as stdlib_random
from scipy import *

# use stdlib_random.choice, not random.choice

Answer 2

In addition to Martijin Pieters answer I want to add that you could also import random module with an alias:

import random as rdm
from scipy import *       

# Then you can 
rdm.choice(some_sequence)