Skip to main content

Simpler parameterization of Junit tests

Why write a custom runner/suite?

Usually Junit's default suite/runner combined with parameterized.class are great for parameterizing your tests. But the need to provide a specifically written constructor, and the need to add a specifically annotated and written method to your test class is rather annoying. The Collection syntax is also weird, though generic. To hide the details of parameterization, and provide such support simply through the use of @RunWith is what I am after. It is okay if my class has to implement another interface, but I would look to minimize impact on the user test class and coupling between test class and test infrastructure.

It is good that JUnit 4 has made it easier to write your own custom Suites and Runners.

How do suites and runners work?

In brief, a Junit Suite wraps around your test class and create a test 'runner' for your class.
A Suite can contain one or more test classes, but for design ease, I will always assume one suite contains one class.

This is how the Parameterized.class works too, the idea being that it creates as many runners for your test class as you specify in a specifically annotated method. Although quite generic, I would like to often create a runner that does not need the arcane syntax (a method with a Collection return type) as required by Parameterized.class.

In other words, I need to create my own parameterization, the details being hidden from the class using my custom suite.

Here is how I began work to achieve this (this post only covers part of the problem, I still have the need to provide a custom constructor, which I will eliminate in the future):

Step 1: Subclass Junit4 to create a custom runner, MyRunner
Step 2: Subclass Suite (create MySuite) to create as many instances of MyRunner (add each instance to a 'runners' list required by Suite)


So, MySuite extends Suite
MyRunner extends Runner

In MySuite constructor, add a method call to your additional method that creates a list of runners with your business logic.


That's it, with this, I then annotated my test class with @RunWith(MySuite.class), and had as many MyRunner per test class as I wanted. How many to create? this depends on what you want your suite to do. For example, you may have logic in your suite to create a runner per file type, thus achieving multi-file tests (of course, you wouldn't use this method for such a simple case)


What I don't like about this design

I use too much inheritance. If Suite or Junit4 classes change, my classes have to change too. My next step is too refactor and eliminate inheritance, replacing with interfaces to facilitate better dynamic dependency injection.

Hope you find this helpful. I will post a concrete example when I get a chance, and a followup post once I refactor it. hopefully Junit folks have already thought about this, or are working on it!

Comments

Popular posts from this blog

Unit testing code that uses environment variables and system properties with fakes

I did not exactly learn this today, but I am sharing it as I have found it extremely useful when unit testing code that depends on environment or system property settings. While I am using Java as an example, the general concepts apply any where. Problem : You have a piece of code you are unit testing that uses settings from env variables or system properties passed to the VM (System.getProperty), but you don't want the tests to be affected by the 'real' environment or system properties in the VM. So, your unit tests should not get different results or fail when the real environment changes. Solution : There are several. But the most straightforward is to use a mocking library to mock out the environment or fake it out, whatever your prefer. You can create a fake using a library like EasyMock, PowerMock etc. This I won't discuss in this post, since there are numerous articles for that. Or you can write a simple class that acts as a proxy, using the proxy pattern...

Sending Form data to a backend REST API using Axios

This need is incredibly common and useful, and hopefully will save you a lot of time when doing server side calls from your UI application (or even non UI clients like NodeJS applications) Example here is to send a POST request to an endoint /api/item/new (which will create a new item in the database). We will just assume tbhe backend is already setup (it's not relevant to this article). All we need to know is that we can do a POST /api/item/new and send it form data with two pieces of info     name, filter So, if you have a node.js application (I was using Vue-cli generated project, but it does not matter), install 'axios' (a most popular tool to make server calls these days) npm i axios --save OR yarn add axios (my preferred method) Now, in your service JS file (which is generally when I keep all my api calls) do something like this createNew ( name , filter ) { let formData = new FormData (); formData . append ( "name" , ...

CSS: em vs rem font sizes

 When do you use em and when do you use rem? If you have ever asked this, you are like me :) So welcome. Basically, to save you time here it is: - If you want your font-size relative to the container's font-size, use em - If you want your font-size relative to the 'root' (or html) element's font-size, use rem! If you just stop reading now that might be sufficient, but if you are more curious, go on. Example companion codepen: https://codepen.io/binodpanta/pen/RwLWRra Basically your page should ideally always have a default font-size specified for the root, such as  :root { font-size: 1em; } This typically becomes 16px default for the base font size. Now, if you use rems in your elements' styles you get a consistent scaling wrt this number! so if you do div.someclass { font-size: 0.5rem; } you are going to always get a nice scaled font size regardless of screen size. So all your fonts will scale relatively throughout the app!  If you had used 0.5em, your calculated ...