"Testing in Django"

Writing tests and test-driven development has become all the rage lately; but before declaring assertions left and right, it's important to consider where, why, and how you will write your tests.

Different frameworks have different philosophies about how to go about this but, more often than not, developers tend to create their own style or use the style of third-party test libraries. While there is nothing wrong with differing styles, you should take care to ensure that your are being efficient and effective.

This is my (by no means authoritative) take on the three important divisions of testing (unit/integration/functional) in Django.

Unit Tests

The most basic level of testing is unit testing. This is where you ensure your module uses other (already tested) modules the way it should. They are absolutely the least complex, no-brainer-type tests you will ever write.

Think about unit tests like hooking up cars on a train. When you hook up a car on the end of a train you'll double- check that it is indeed fastened to the car in front of it. You won't recheck the entire train because you assume whoever hooked up the cars in front did their part ensuring everything was in place.

For example, say I have a filter I call often so I put it into a module:

def get_authors(library):
    return Author.objects.filter(books__library=library)

We know that Django's ORM works—it is very well tested—so there is no need to re-test it. What we do need to test, though, is that our function utilizes the ORM the way we want. We accomplish this with a unit test and a wonderful library called Mock. (now included in Python 3.3 and up)

Installing Mock is easy with pip: pip install mock

Mock replaces a module with a fake module that intercepts calls and records them along with any arguments involved. Using mock in our example, we can show that our method calls Author.objects.filter without actually running it.

The basic usage of Mock is first setting it up:

my_mock = Mock(return_value="something")

And then asserting that it was called with any number of arguments:


For our Django unit tests, we'll use Mock's patch decorator which declares a Mock for the duration of the test:

def test_something(self, mymethod_mock):

Here's what a unit test looks like for our example:

import unittest

import mock

from myapp.mymodule import get_authors

class GetAuthorsTest(unittest.TestCase):

    def test_get_authors(self, AuthorMock):
        Author.objects.filter = mock.Mock(return_value="ReturnMock")

        result = get_authors("library_mock")

        self.assertEqual("ReturnMock", result)

In summary, we:

  • "Patched" our test by mocking Author. Note that we reference Author as and object within our module, not from it's original source which probably would have been something like myapp.models.Author. Also note that mock patches are passed to your function as arguments which can be any name you like. I chose AuthorMock.

  • Piggybacked another mock, objects.filter, on top of the AuthorMock. You'll find that mocks are very flexible once they've been created.

  • Assigned a return value to the objects.filter mock. We test for this later.

  • Called the function we want to test with a dummy argument.

  • Asserted that our mock was indeed called with the appropriate arguments.

  • Asserted that the function did indeed return the result of that call, which in this case should be the value we gave to return_value.

We accomplished all this without actually touching the Django ORM, but now we know for sure that when mocks are put aside, this function will operate correctly.

Why do we do it this way? Why not create some dummy data in the database in and see if it produces the right results. Well we actually will do that later, in the form of integration tests. Unit tests have two big benefits:

  • They reduce false-positives by ensuring your code is utilizing the right code rather than producing the right results (which could be right for your test, but wrong in other scenarios)

  • They are very, very fast.

Notice we don't even use Django's TestCase. We don't need it. For some unit tests you may, but often Python's unittest library and mock are all you need. These tests will run insanely fast, and for that fact, you can get into the habit of running them periodically while you're developing.

As you build your application, each building block should be unit tested. This will increase confidence in your architecture especially when you add features to existing code.

Integration Tests

These are usually the most common tests that beginners right because they tend to be pretty straight-forward, and don't require much thought. Integration tests essentially are meant to accomplish "if I put in input A, do I get expected output B?"

In Django, integration tests ensure that your modules and views are generating the content you expect them to.

The easiest way to write integration tests in Django is to use the TestClient that Django provides which allows you to fake browser requests and see what sort of responses you get back from you application.

Say we have this very basic view:

def blog_detail(request, blog_id):
    blog = get_object_or_404(Blog, id=blog_id)
    context = {'blog': blog}
    return render(request, "blog/blog_detail.html", context)
{% endhighlight %

There are really two scenarios we would want to test for here:

1. Scenario in which the blog exists
2. Scenario in which the blog does not exist

In scenario 1 we would expect the view to return a success (200) response after rendering the blog_detail template.  In
scenario 2 we would expect the view to return a not-found response (404).

To accomplish this we will use another common testing apparatus called a fixture.  The Mock library provides a
decorator to make this easy.  Essentially, you write an instance method in your test case that defines the fixture.
For example:

def user(self):
    return User.objects.create(username="bob")

Here is how we would write a test using Django's TestClient:

import mock

from django.core.urlresolvers import reverse
from django.test import TestCase

from myblogapp.models import Blog
from myblogapp.views import blog_detail

class BlogDetailTest(TestCase):

    def blog(self):
        return Blog.objects.create(
            title="BlogMock", content="BlogMock")

    def test_when_blog_exists(self):
Go Top
comments powered by Disqus