If you’re not testing your code, well, you’ve got some nerve. For the rest of us rational people, let’s look at some basic tips for writing better Ember tests.
What Makes a Good Test
There are three core components:
1. Easy to Reason
Tests must be easy to understand. While code quality is important for developers to understand, tests often communicate a business requirement or decision which is often more important.
Tests must be very granular. Being specific is a great way of creating a manifest of all of the pathways, states and other vagaries that may not be immediately obvious.
Tests must be fast. The faster the test runs, the faster the developer can regression their code. The faster a developer can setup and write a test for their new code, the faster a developer can write more code and tests.
By default, Ember uses QUnit under the hood. There are two ways we can leverage QUnit that make our tests easier to reason about.
- Test Name
- Assertion Message
The Page Object Design pattern is a powerful one for writing re-usable, easy to understand tests. Here’s an example:
In this trivial example, the first test demonstrates a problem where as you start to add more tests, you will end up writing the same selectors multiple times. In the second test, it should be clear that the
page variable could be refactored and re-used across tests. Now all of your tests reference
page.$button; One change to the page object now updates all of your tests using it.
An even better approach would be to leverage the ember-cli-page-object addon:
In this example, we setup a
page object which has all the parts of the page we care about, similar to the examples prior. With
ember-cli-page-object, tests become much easier to reason because our assertions become almost human readable:
page.button.isVisible could be understood by non-developers.
ember-cli-page-object also makes writing tests faster for developers. Tests can be quickly read and understood without having to recall which jQuery selector to use to find your element.
Pro Tip: I have found
class selectors volatile over time. I have found using a
data-test attribute to be much more stable. For example:
In this piece of HTML, we have a
<p> tag that has two attributes,
class may change over time as styles change, or naming convention changes.
data-test is more of a universal selector for this element that denotes what the element is and signals to the developer that it’s used by tests and should be changed with hesitation.
Mock Server Requests
Chances are your application works with some kind of backend or API. Mocking your API in Ember is a great idea for several reasons:
- No backend dependency to run your app tests
- Data and state are ephemeral
- It’s faster to mock various server scenarios
There are a few ways to achieve this, but my personal favorite is ember-cli-mirage. TL;DR:
- Your application is going to make web requests, for example:
- You’re going to use Mirage to intercept web requests.
- You’re going to send a “mock” response so your tests can run without actually hitting a backend.
ember-cli-mirage does not currently support the idea of scenarios. For example, I want
GET /api/things to return a specific response most of the time, but in some different case, I want it to return something else.
Here’s what I have found to be a good solution:
- Define your Routes as if they were the “happy path” - the data that is returned most of the time.
- In the test that expects something different, specifically override that route, for example:
This is recommended by Mirage and can be found in their documentation here.
Are you using a CI tool? You should be. There are a lot of options out there. A popular option, Travis, is free for open source projects. As application complexity increases, or the number of developers increases, it becomes pretty difficult for any one person to know everything. Writing lots of good tests and having them pass before integrating new code is important for the sanity of the team and it helps build confidence in the application’s stability over time. Find a bug? No worries, write a test that reproduces it, fix the bug and never worry about it again.
Make no mistake, tests can be challenging. They often involve the same amount, or more code than your application. And no matter how easy it is to write tests, they still take time, which could be time spent writing more features or fixing bugs. The real key to writing good tests is understanding their worth. Your time is the most valuable asset you have. Don’t waste it by debugging things a simple test would have caught.
Have some tips of your own? I’d love to hear them.