Question

I would like to unit-test the functionality of a few classes that mainly do file in- and output. Furthermore I would like to do that on multiple cores (–jobs=4).

The problem is, that the files that are created by the classes often have the same name and they get mixed up in multiple threads. What I do currently is that I run each unit-test in a separate directory like so:

def test(self):
  if os.path.exists("UniqueDir"):
    os.system("rm -rf UniqueDir")
  os.mkdir("UniqueDir")
  os.chdir("UniqueDir")
  #Do the actual testing
  os.chdir("..")
  os.rmdir("UniqueDir")

The downsides are very obvious:

1. Each test must receive a unique directory name
2. Each test has this overhead of source which really is not pleasant to look at at all

What approach could I use to 1. separate my tests from one another but 2. do it in a more elegant way?

Any help, suggestion etc. is appreciated!

Cherio Woltan

Answer 1

I would suggest to use the unittest module and build the classes like this:

import unittest
from tempfile import mkdtemp

class Test(unittest.TestCase):

    def setUp(self):
        self.tempdir = mkdtemp()
        os.chdir(self.tempdir)

    def tearDown(self):
        os.rmdir(self.tempdir)

    def testName(self):
        #Do the actual testing
        pass

if __name__ == "__main__":
    unittest.main()

Additionally you could add multiprocessing to create 4 threads.

Edit: removed the os.mkdir because mkdtemp creates a temp directory so it’s bogus. Thx Sebastian.