My Profile Photo

Daniel Perez


software engineer interested in programming languages and functional programming


assertRaises in Python

A friend asked me about Python unittest’s assertRaises and the answer was becoming too long for Slack, so I decided to write a short blog post about it.

In this blog post, we will cover how assertRaises in unittest.TestCase works, and implement a simplified version of it.

For the sake of example, let’s say we want to check that next(iter([])) raises a StopIteration error.

We will use a very simple Python script to try the code

import unittest


class SampleTest(unittest.TestCase):
  def test_assert_raises(self):
    pass


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

Common mistake

First, let’s think about a typical error when trying to use self.assertRaises. Let’s replace the pass with the following statement.

self.assertRaises(StopIteration, next(iter([])))

Python evaluation is strict, which means that when evaluating the above expression, it will first evaluate all the arguments, and after evaluate the method call. When evaluating the arguments we passed in, next(iter([])) will raise a StopIteration and assertRaises will not be able to do anything about it, even though we were hoping to make our assertion. We will therefore end up with the test failing because of this exception.

So, there are two ways implemented by assertRaises to go around this issue.

Function call delegation approach

The first way is to delegate the call of the function raising the exception to assertRaises directly. This means that we pass a reference to the function we would like to call and all its arguments to assertRaises, and let it take care of calling the function we passed in. assertRaises will ensure that the exception is captured when making the function call. The call looks like this

self.assertRaises(StopIteration, next, iter([]))

next is the function we want to call and iter([]) are the arguments to this function. We can try it in the above call and the test will pass, as expected. To see how this might work, here is a sample implementation of assertRaises that can be called in the same way. Note that it is not implemented exactly in this way in the unittest module.

def argsAssertRaises(self, exc_type, func, *args, **kwargs):
  raised_exc = None
  try:
    func(*args, **kwargs)
  except exc_type as e:
    raised_exc = e
  if not raised_exc:
    self.fail("{0} was not raised".format(exc_type))

There are three possible outcomes here:

  • If func raises exc_type, it will be caught, raised_exc will not be None anymore, and the function will terminate successfully
  • If nothing is raised, raised_exc will stay None and self.fail will be called
  • If another exception is raised, it will not be caught as we are only catching exc_type.

which is the expected behavior. You can try replacing self.assertRaises by self.argsAssertRaises and it should give the same result.

Context manager approach

The other way that unittest uses to assert on exceptions is by using context managers.

A sample usage would look like this

with self.assertRaises(StopIteration):
  next(iter([]))

To understand how it might work, there are a few things we need to understand about context managers. A context manager is typically used as

with MyContextManager() as m:
  do_something_with(m)

The identifier in the as clause will be assigned whatever the __enter__ method of MyContextManager returns. When the with statement exits, it will call the __exit__ method of MyContextManager, potentially passing in the exception type, exception value and exception traceback - respectively exc_type, exc_value and exc_tb. If an exception has been raised and the __exit__ method returns True, the exception is suppressed, otherwise, the exception will propagate.

Once we understand this, this gives us some idea about how we could implement assertRaises:

  • assertRaises should return a context manager instance - i.e. the instance of a class implementing __enter__ and __exit__
  • __enter__ does not seem very important here, as we are not using the as clause anywhere
  • __exit__ should check if the passed exception has been raised, in which case it should suppress it

Here is a simple implementation of a context manager doing this.

class SampleRaiseContextManager:
    def __init__(self, expected_exc, test_case):
        self.expected_exc = expected_exc
        self.test_case = test_case

    def __enter__(self):
        return self

    def __exit__(self, exc_type, exc_value, tb):
        if exc_type is None:
            self.test_case.fail("{0} was not raised".format(self.expected_exc))
        elif isinstance(exc_value, self.expected_exc):
            return True
        return False

expected_exc is the type of the exception we are expecting, and test_case is the current test case - self in our TestCase methods. In __enter__, we simply return self, although we are not doing anything useful with it here. The important part here is __exit__. If exc_type is None, it means that nothing was raised inside the with statement, so we want to fail the test case as we were expecting an exception. If we did not enter this clause, we are sure that an exception has been raised, so we want to check if it was the exception we were expecting. This is done by using isinstance on the value of the raised exception. If the exception we were expecting has been raised, we suppress it, as we want the test to succeed. Otherwise, it means that an unexpected exception was raised, so we let it propagate.

Now that we have our context manager, we simply need to write a helper method in our test case class, which will return an instance of it:

def cmAssertRaises(self, expected_exc):
  return SampleRaiseContextManager(expected_exc, self)

we can now use our helper method to test for exceptions

with self.cmAssertRaises(StopIteration):
  next(iter([]))

This is pretty much it for how assertRaises works. Here is the full Python code for the explanations above

import unittest


class SampleRaiseContextManager:
    def __init__(self, expected_exc, test_case):
        self.expected_exc = expected_exc
        self.test_case = test_case

    def __enter__(self):
        return self

    def __exit__(self, exc_type, exc_value, tb):
        if exc_type is None:
            self.test_case.fail("{0} was not raised".format(self.expected_exc))
        elif isinstance(exc_value, self.expected_exc):
            return True
        return False


class SampleTest(unittest.TestCase):
    def test_args_raise(self):
        self.argsAssertRaises(StopIteration, next, iter([]))

    def test_my_context_manager(self):
        with self.cmAssertRaises(StopIteration):
            next(iter([]))

    def argsAssertRaises(self, exc_type, func, *args, **kwargs):
        raised_exc = None
        try:
            func(*args, **kwargs)
        except exc_type as e:
            raised_exc = e
        if not raised_exc:
            self.fail("{0} was not raised".format(exc_type))

    def cmAssertRaises(self, exc_type):
        return SampleRaiseContextManager(exc_type, self)


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

The only part left is to unify the two implementations above - argsAssertRaises and cmAssertRaises. This can be done by receiving a variable number of arguments and checking the number of arguments received.

Of course, the code above is for learning purpose, so for real world use cases, use the implementation provided by the unittest module.

comments powered by Disqus