How to Write Better Unit Tests

Last Update: Dec 31, 2022

I wrote a book! Check out A Quick Guide to Coding with AI.
Become a super programmer!
Learn how to use Generative AI coding tools as a force multiplier for your career.

In working with some unit testing fairly extensively lately, I’ve gathered some guidelines I’ve tried to stick with over the years for writing better tests. Remember poorly written tests are a waste of time, and cause major problems down the road. It’s best to keep some of these guidelines in mind.

• Unit tests should not be written to pass - They should be written to fail. You can make any set of tests pass in minutes but you’re only cheating yourself.
• Tests should only test one thing - You should be testing a single method with a single function. If not you may be violating the Single Responsibility Principle
• Readability in your tests- make sure they’re commented and easy to understand, just like any other code.
• Good Naming conventions - Again tests should be just like any other code - easy for humans to understand.
• Asserts are separated from actions - Your assert should be looking for a result, and not performing logical operations
• Use concrete inputs- Don’t use any dynamic data for inputs, things like date() can introduce variance.
• Group locations of tests - from a logical standpoint this makes things easier to find when there aren’t errors pointing towards the problem.
• Good tests are isolated from everything - You should have no reliance on other tests, environment settings, etc. This creates multiple points of failure.
• Do not include private methods - They are implementation and should not be included in Unit Tests
• Don’t connect to databases or data sources - This is unreliable because you cannot be certain the data served will always be the same, and can create points of failure.
• No more than one mock per test - Again we’re trying to eliminate points of failure and inconsistenties.
• Unit tests are not integration tests - You want to test results, not implmentation with Unit Tests.
• Tests must be deterministic - You need a solid predictable result, so if it only passes sometimes, it’s not done.
• Keep your tests idempotent - you should be able to run it multiple times without changing any outcomes, and it should not change any data or increment anything. One time or a million times should have the same effect.
• Classes only test one class at a time, methods only test one method at a time. - An organizational method to pinpoint problems when they arise and help you identify dependencies in testing.
• Include exceptions in your tests- You’re going to have exceptions so don’t ignore them, use them.
• Don’t test functionality of 3rd party libraries with your own tests - Most quality libraries should have their own tests. If not consider mocks to produce consistent results
• Always limit values - When working with values be mindful of your limits and set them (min max) for maximum consistency.
• Tests should not require configuration or custom setup - Anyone should be able to jump in and make your tests work. “Works on my machine” should never apply here.

I hope this helps some of you out there learning and working with Unit Tests.

Published: Aug 24, 2014 by Jeremy Morgan. Contact me before republishing this content.