Bedrock runs a suite of front-end Jasmine behavioral/unit tests, which use
Karma as a test runner. We also have a suite of functional tests using
Selenium and pytest. This allows us to emulate users interacting with a
real browser. All these test suites live in the
tests directory comprises of:
/functionalcontains pytest tests.
/pagescontains Python page objects.
/unitcontains the Jasmine tests and Karma config file.
First follow the installation instructions for bedrock, which will install the specific versions of Jasmine/Karma which are needed to run the unit tests, and guide you through installing pip and setting up a virtual environment for the functional tests. The additional requirements can then be installed by using the following commands:
$ source venv/bin/activate
$ pip install -r requirements/dev.txt
Running Jasmine tests using Karma¶
To perform a single run of the Jasmine test suite using Firefox, type the following command:
$ npm run test
Running functional tests¶
Before running the functional tests, please make sure to follow the bedrock installation docs, including the database sync that is needed to pull in external data such as event/blog feeds etc. These are required for some of the tests to pass. To run the tests using Firefox, you must also first download geckodriver and make it available in your system path. You can alternatively specify the path to geckodriver using the command line (see the pytest-selenium documentation for more information).
To run the full functional test suite against your local bedrock instance:
$ py.test --base-url http://localhost:8000 --driver Firefox --html tests/functional/results.html tests/functional/
This will run all test suites found in the
tests/functional directory and
assumes you have bedrock running at
localhost on port
8000. Results will
be reported in
If you omit the
--base-url command line option then a local instance of
bedrock will be started, however the tests are not currently able to run
against bedrock in this way.
By default, tests will run one at a time. This is the safest way to ensure
predictable results, due to
If you want to run tests in parallel (this should be safe when running against
a deployed instance), you can add
-n auto to the command line. Replace
auto with an integer if you want to set the maximum number of concurrent
There are some functional tests that do not require a browser. These can
take a long time to run, especially if they’re not running in parallel.
To skip these tests, add
-m 'not headless' to your command line.
To run a single test file you must tell py.test to execute a specific file
$ py.test --base-url http://localhost:8000 --driver Firefox --html tests/functional/results.html tests/functional/firefox/new/test_download.py
To run a single test you can filter using the
-k argument supplied with a keyword
$ py.test --base-url http://localhost:8000 --driver Firefox --html tests/functional/results.html tests/functional/firefox/new/test_download.py -k test_download_button_displayed
You can also easily run the tests against any bedrock environment by specifying the
--base-url argument. For example, to run all functional tests against dev:
$ py.test --base-url https://www-dev.allizom.org --driver Firefox --html tests/functional/results.html tests/functional/
For the above commands to work, Firefox needs to be installed in a predictable location for your operating system. For details on how to specify the location of Firefox, or running the tests against alternative browsers, refer to the pytest-selenium documentation.
For more information on command line options, see the pytest documentation.
Running tests in Sauce Labs¶
You can also run tests in Sauce Labs directly from the command line. This can be useful if you want to run tests against Internet Explorer when you’re on Mac OSX, for instance.
Sign up for an account at https://saucelabs.com/opensauce/.
Log in and obtain your Remote Access Key from user settings.
Run a test specifying
SauceLabsas your driver, and pass your credentials.
For example, to run the home page tests using Internet Explorer via Sauce Labs:
$ SAUCELABS_USERNAME=thedude SAUCELABS_API_KEY=123456789 py.test --base-url https://www-dev.allizom.org --driver SauceLabs --capability browserName 'internet explorer' -n auto --html tests/functional/results.html tests/functional/test_home.py
Writing Selenium tests¶
Tests usually consist of interactions and assertions. Selenium provides an API for opening pages, locating elements, interacting with elements, and obtaining state of pages and elements. To improve readability and maintainability of the tests, we use the Page Object model, which means each page we test has an object that represents the actions and states that are needed for testing.
Well written page objects should allow your test to contain simple interactions and assertions as shown in the following example:
def test_sign_up_for_newsletter(base_url, selenium): page = NewsletterPage(base_url, selenium).open() page.type_email('firstname.lastname@example.org') page.accept_privacy_policy() page.click_sign_me_up() assert page.sign_up_successful
It’s important to keep assertions in your tests and not your page objects, and to limit the amount of logic in your page objects. This will ensure your tests all start with a known state, and any deviations from this expected state will be highlighted as potential regressions. Ideally, when tests break due to a change in bedrock, only the page objects will need updating. This can often be due to an element needing to be located in a different way.
Please take some time to read over the Selenium documentation for details on the Python client API.
By default all tests are assumed to be destructive, which means they will be
skipped if they’re run against a sensitive environment. This prevents
accidentally running tests that create, modify, or delete data on the
application under test. If your test is nondestructive you will need to apply
nondestructive marker to it. A simple example is shown below, however
you can also read the pytest markers documentation for more options.
import pytest @pytest.mark.nondestructive def test_newsletter_default_values(base_url, selenium): page = NewsletterPage(base_url, selenium).open() assert '' == page.email assert 'United States' == page.country assert 'English' == page.language assert page.html_format_selected assert not page.text_format_selected assert not page.privacy_policy_accepted
Smoke tests are considered to be our most critical tests that must pass in a wide range of web browsers, including Internet Explorer 11. The number of smoke tests we run should be enough to cover our most critical pages where legacy browser support is important.
import pytest @pytest.mark.smoke @pytest.mark.nondestructive def test_download_button_displayed(base_url, selenium): page = DownloadPage(selenium, base_url, params='').open() assert page.is_download_button_displayed
You can run smoke tests only by adding
-m smoke when running the test suite on the
Sanity tests behave in much the same way as smoke tests, but will also run against Internet Explorer 9, which is a browser that does not receive 1st class CSS/JS support (except on certain download pages such as /firefox/new/). The number of sanity tests we run should be small and cover only a handful key of pages.
import pytest @pytest.mark.sanity @pytest.mark.smoke @pytest.mark.nondestructive def test_download_button_displayed(base_url, selenium): page = DownloadPage(selenium, base_url, params='').open() assert page.is_download_button_displayed
You can run sanity tests only by adding
-m sanity when running the test suite on the
Waits and Expected Conditions¶
Often an interaction with a page will cause a visible response. While Selenium does its best to wait for any page loads to be complete, it’s never going to be as good as you at knowing when to allow the test to continue. For this reason, you will need to write explicit waits in your page objects. These repeatedly execute code (a condition) until the condition returns true. The following example is probably the most commonly used, and will wait until an element is considered displayed:
from selenium.webdriver.support import expected_conditions as expected from selenium.webdriver.support.ui import WebDriverWait as Wait Wait(selenium, timeout=10).until( expected.visibility_of_element_located(By.ID, 'my_element'))
For convenience, the Selenium project offers some basic expected conditions, which can be used for the most common cases.
Debug information is collected on failure and added to the HTML report
referenced by the
--html argument. You can enable debug information for all
tests by setting the
SELENIUM_CAPTURE_DEBUG environment variable to
Guidelines for writing functional tests¶
Try and keep tests organized and cleanly separated. Each page should have its own page object and test file, and each test should be responsible for a specific purpose, or component of a page.
Avoid using sleeps - always use waits as mentioned above.
Don’t make tests overly specific. If a test keeps failing because of generic changes to a page such as an image filename or
hrefbeing updated, then the test is probably too specific.
Avoid string checking as tests may break if strings are updated, or could change depending on the page locale.
When writing tests, try and run them against a staging or demo environment in addition to local testing. It’s also worth running tests a few times to identify any intermittent failures that may need additional waits.
See also the Web QA style guide for Python based testing.
Testing Basket email forms¶
When writing functional tests for front-end email newsletter forms that submit to Basket, we have some special case email addresses that can be used just for testing:
Any newsletter subscription request using the email address “email@example.com” will always return success from the basket client.
Any newsletter subscription request using the email address “firstname.lastname@example.org” will always raise an exception from the basket client.
Using the above email addresses enables newsletter form testing without actually hitting the Basket instance, which reduces automated newsletter spam and improves test reliability due to any potential network flakiness.
There are targeted headless tests for the download and localized download pages. These tests and are run as part of the pipeline to ensure that download links constructed via product details are well formed and return valid 200 responses.