Python子模块导入差异

Question

I am dynamically importing the numpy package into a python environment of another proprietary system. The top level numpy package gets imported from the right place, but the numpy.random package is pointing to the standard library. Why is this happening?

Code

import sys

LIB_PATH = 'T:\\Some\\Path\\'

if LIB_PATH not in sys.path:
    sys.path.insert(0, LIB_PATH)

import numpy

print numpy
print numpy.random

Output

<module 'numpy' from 'T:\Some\Path\numpy\__init__.pyc'>
<module 'random' from 'C:\Python26x64\Lib\random.pyc'>

Why is numpy.random pointing to C:\\Python26x64\\Lib\\random.pyc . When I run this from my standard python interpreter at C:\\Python26x64\\ , then random is indeed the one in the numpy package

Answer 1

Here is the solution that worked for me. It don't think its a good solution cause it requires changing one line of code in the numpy package. Nonetheless, it allow us to use pandas and numpy from an embedded Python interpreter in a proprietary software by simply adding a folder to the sys.path .

File Changed

T:\\Some\\Path\\numpy\\__init__.py

Line Before

Line: 171 - import random

This line assumes that the random will be loaded from T:\\Some\\Path\\numpy\\random\\

Line After

Line: 171 - from numpy import random

This forces it to use the random package from numpy instead of the standard package

I thought I'd just leave there here and one of the un-accepted answer until someone can come up with a better solution.