How do you test that a Python function throws an exception?

Solution 1:

Use TestCase.assertRaises (or TestCase.failUnlessRaises) from the unittest module, for example:

import mymod

class MyTestCase(unittest.TestCase):
    def test1(self):
        self.assertRaises(SomeCoolException, mymod.myfunc)

Solution 2:

Since Python 2.7 you can use context manager to get ahold of the actual Exception object thrown:

import unittest

def broken_function():
    raise Exception('This is broken')

class MyTestCase(unittest.TestCase):
    def test(self):
        with self.assertRaises(Exception) as context:
            broken_function()

        self.assertTrue('This is broken' in context.exception)

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

http://docs.python.org/dev/library/unittest.html#unittest.TestCase.assertRaises

In Python 3.5, you have to wrap context.exception in str, otherwise you'll get a TypeError

self.assertTrue('This is broken' in str(context.exception))

Solution 3:

The code in my previous answer can be simplified to:

def test_afunction_throws_exception(self):
    self.assertRaises(ExpectedException, afunction)

And if afunction takes arguments, just pass them into assertRaises like this:

def test_afunction_throws_exception(self):
    self.assertRaises(ExpectedException, afunction, arg1, arg2)

Solution 4:

How do you test that a Python function throws an exception?

How does one write a test that fails only if a function doesn't throw an expected exception?

Short Answer:

Use the self.assertRaises method as a context manager:

    def test_1_cannot_add_int_and_str(self):
        with self.assertRaises(TypeError):
            1 + '1'

Demonstration

The best practice approach is fairly easy to demonstrate in a Python shell.

The unittest library

In Python 2.7 or 3:

import unittest

In Python 2.6, you can install a backport of 2.7's unittest library, called unittest2, and just alias that as unittest:

import unittest2 as unittest

Example tests

Now, paste into your Python shell the following test of Python's type-safety:

class MyTestCase(unittest.TestCase):
    def test_1_cannot_add_int_and_str(self):
        with self.assertRaises(TypeError):
            1 + '1'
    def test_2_cannot_add_int_and_str(self):
        import operator
        self.assertRaises(TypeError, operator.add, 1, '1')

Test one uses assertRaises as a context manager, which ensures that the error is properly caught and cleaned up, while recorded.

We could also write it without the context manager, see test two. The first argument would be the error type you expect to raise, the second argument, the function you are testing, and the remaining args and keyword args will be passed to that function.

I think it's far more simple, readable, and maintainable to just to use the context manager.

Running the tests

To run the tests:

unittest.main(exit=False)

In Python 2.6, you'll probably need the following:

unittest.TextTestRunner().run(unittest.TestLoader().loadTestsFromTestCase(MyTestCase))

And your terminal should output the following:

..
----------------------------------------------------------------------
Ran 2 tests in 0.007s

OK
<unittest2.runner.TextTestResult run=2 errors=0 failures=0>

And we see that as we expect, attempting to add a 1 and a '1' result in a TypeError.

For more verbose output, try this:

unittest.TextTestRunner(verbosity=2).run(unittest.TestLoader().loadTestsFromTestCase(MyTestCase))