Whether you’re working on an existing or new application, you’ll often find yourself playing catch up when it comes to tests. Soon deploying code changes feels like poking at some ugly, sleeping code monster — you aren’t sure what’s going to happen, but you know it won’t be good.
Here are the 4 things you should do first to tame the beast and improve test coverage:
1. Add the Right Tests
Start by adding tests in the areas where it is easiest. It’s important to consider the order in which you do this to make sure you get the most out of your scarce resources. You want to start adding tests in the following order:
- Create tests as you fix bugs
Add tests to prove that your specific fix is working and keep them running to show that this does not break again. This prevents from being somewhat targeted — you are creating tests in your weakest areas first. The weaker it is (i.e. more bugs) the faster you will build up tests.
- Create tests with all new features
All new features will need to include tests created to prove that the feature works as expected. If you’re covering the new aspects of your application, then at least things aren’t getting worse.
- Create tests as you add to old features
When updating old features, add tests as you go to show that the older functionality is not breaking in unexpected ways.
- Create tests in identified risk areas
Talk to the developers and testers on your team and ask them to point out any weak spots or issues they have experienced. Also, you talk to your support team — they are an excellent resource with a direct line to the customer. They’ll know the features of your product that frequently causes issues.
2. Turn on Code Coverage
Code coverage is a tool included in most continuous integration systems (or one that can be added with a plugin). This tool will instrument and monitor your code as you run the tests to determine how much of your code used by the tests. For this to be useful, follow these steps:
- Start running code coverage against all your code
- Get a baseline
Find out what the tool can see, where you are currently at etc.
- Determine areas that you want to exclude.
There are likely areas of your code that you don’t want to cover — third-party tools, ancient code untouched for years etc.
- Determine coverage goals
Sit down with your team and discuss what your current coverage is and what your ideal can realistically be (usually 90% or above).
- Work-out steps to improve your coverage
You aren’t going to fix this problem overnight. Put in place some specific tasks which are going to help you achieve your goals over time.
- Determine your pass/fail criteria
Is staying the same OK, or should it always go up? Do you define any drop as a fail?
- Run Code Coverage constantly
Use automation to run your coverage test and use the criteria you determined as a team to report a pass/fail and do this constantly. It is a lot easier to add tests when the code is still front and center in your mind than later on.
3. Run your Tests on a Scheduled Basis
You should run your tests regularly, on several schedules:
- Run them on every check-in
Use CI tools like Jenkins to run (at least) your unit tests on every check-in. Run them simultaneously if they are taking too long to run at this frequency.
- Run them on every build (package)
Depending on how your systems work, your CI infrastructure can help you with this. This could be on every check-in if you are on Continuous Deployment, or every day/week/month that you use. This should be a clean install on a test environment and a full run of all your automated tests.
- Run them on every deploy
You should run all your automated tests against your environments immediately after a deploy.
- Run them every X days/hours/minutes
Run your automation suite against your constant environments as often as you can (Production, Staging etc). This should be at least a “once a day task” and take place during ‘off-peak’ times when it does not interrupt others too much. You can increase the frequency further if your tests are short, just be mindful not to overload the system.
4. Provide a Button to Run the Tests
Again, use a tool like Jenkins to turn test runs into a self-service operation. A developer should not be delayed by asking a QA person to run a test for them. Get a system in place where your tests will run and just give them a button to press. Remove as many barriers for everyone to run the tests as possible.
If you follow these steps, over time, you’ll see that you are able to turn an unwieldy application into something more manageable. First, by adding tests to the key areas, then making things as easier as possible, you can build confidence around your code changes and deploys.
Comments