Simulating lack of a dependency when testing a python script

What's the best way to temporarily hide an installed module from a python s开发者_开发技巧cript to test how it handles environments that don't have the module installed?

I'd like to avoid having to uninstall the module just to test.

import sys
sys.modules['numpy']=None

Setting sys.modules['numpy']=None makes Python think that it has already tried and failed to import numpy. Subsequent attempts at importing numpy now raise ImportError:

try:
    import numpy
except ImportError as err:
    print(err)
    # No module named numpy

Deleting sys.modules['numpy'] allows numpy to be imported as normal:

del sys.modules['numpy']
import numpy

Change your Python Path.

The order of directories in sys.path shows the order of a search.

You can change sys.path in a test to alter the search order.