August 8, 2020
By: Stefan

Continuous Integration: What It Is and Four Reasons Why I Like It

Software Development: Easy To Learn, Hard To Master

continuous integration rapid feedback software quality ci/cd continuous delivery


  1. The second thing for new code
  2. What is continuous integration
  3. The value of continuous integration
    1. Safety net / quality baseline
    2. Makes it easier for others to contribute
    3. Executable documentation of build & test (& deployment)
    4. Ensure that code keeps working in the future
  4. Some encouraging words

In my daily work, I have the good fortune to work with many very talented people. Some are professional software developers, like myself. Others are for example researchers, scientist, project leaders, and so on. Is that you? Do you also write software, but not as your main activity? Do you have some experience but maybe no formal training? In this series of blog posts I will explain the best practices that I use, so that you may be able to benefit from them. So that writing software hopefully takes you less effort and gives you more pleasure.

The second thing for new code

The very first thing I always do when starting something new, is git init. Potential new hobby project? git init. Sample code for a blog post? git init. Doing programming exercises on something like git init. And then I start using git the way I described in the first post of this series. There are virtually no exceptions to this; it hardly costs me anything and it has helped me plenty of times.

The second thing that I do, I do just slightly less often, but still very often: set up continuous integration. Obviously I won't do that for things that I never push to a remote. And some things that I push to a remote don't need it either, for example those Exercism coding exercises. But for anything that I suspect may have a longer life than a few weeks, including for that sample code for a blog post, I'll set it up.

What is continuous integration

To get rid of potential confusion, let me explain what I mean with continuous integration, or CI for short. In its most basic form, it's actually quite simple: every single time you push something to your git remote (e.g. GitHub, BitBucket, GitLab, etc), an automated process will start that builds your code and runs all your tests. (You do have unit tests, right? Maybe a nice topic for a future post in this series.) That's all, if you do that you can claim you're doing continuous integration.

The other day, I was talking to one of my researcher colleagues about coding practices. When I asked him whether he was using continuous integration, he thankfully asked for clarification. For me the term has become so commonplace that I hadn't realize that of course for him it wasn't. He is working on a software library that his (internal) customers integrate into their product. So quite logically he thought that was what I meant: continuously automatically integrating his releases into the customer's product code. While doing that is of course fantastic, that's going quite a few steps further than just basic CI. In fact, I would say that that is the combination of continuous delivery (the CD in CI/CD) on his part, plus continuous integration of his delivery on his customer's end.

So if you have a Travis job, a CircleCI pipeline, a GitLab pipeline, a GitHub Actions workflow, a Bamboo build, anything that automatically build and tests your code when you push it to your remote, you're doing continuous integration.


The value of continuous integration

Ok, so now that we're hopefully on the same page regarding the term continuous integration, you may be curious why you should bother. Here's a few reasons that I have.

Safety net / quality baseline

I'm human. I make mistakes. There, I've said it, now you know. I have learnt that mistakes that escape the development process and end up at the customer, are way more expensive than mistakes that are caught earlier. So that's why I like pair programming, because your pair can catch your mistakes while your making them. It doesn't get any cheaper than that. That's also why I like unit tests, because they run many times an hour so they also catch mistakes very quickly. And that's also why I like continuous integration, because it catches even more mistakes before they escape the development process. For instance, when I forget to run my unit tests before pushing my code to the remote: CI will run and immediately notify me. And my reaction to that notification is: cool, thanks for catching that!


Makes it easier for others to contribute

CI runs for any code that is pushed to the remote, not just by me/my team. So if it's source code that others can also see and use, having CI will lower the threshold for them to contribute. Even when you have a new team member it will help. They don't know the source code as well as you do, so they are a lot more likely to make mistakes. When I contribute code to somebody else's project, it feels very comforting to see "all green" as part of the automated tests in the pull request!

Executable documentation of build & test (& deployment)

Just like unit tests are a form of executable code documentation, CI is a form of executable documentation of how your project can be built and tested. And, if you're doing CI/CD, deployed. And I don't think I need to explain the benefits of having executable documentation over plain only-human-readable documentation, or do I?1

Ensure that code keeps working in the future

The final reason I'll mention here, is the future. Because there are two possible sources of problems: you as part of the development team pushing code, but also the environment changes. Libraries that you depend on may turn out to contain security issues2, or cease to exist altogether. Operating systems are updated, as well as libraries in there that you may directly or indirectly depend on. If you have a CI script that periodically runs, even when you don't push anything, you will know that your code not only works as specified when you push it, but also in the future!


Some encouraging words

When I'm setting up CI, which I have done countless times, it's always a challenge. Every single time I have to brace myself and mentally prepare for doing quite a few tries before getting it right. It's just such a complex black box! You write a script that will run on a not-quite-known configuration on a unknown server somewhere, and if it fails you get limited feedback. For example, if you want to automatically test an iOS app, you'll have to specify which version of the iOS simulator should be used. But before trying the build, you may not know which are available and what the simulator device identifiers are that you can use. So you have to try, look at the output, and try again. And then something else isn't quite right, and you try again. And again. And once more. Just have a look at the eighteen (18!) commits I did for sample code for a blog post. And you can't even really clean up those commits because you have to push them to the remote. I guess if it's important to you, you can do them on a separate branch and squash-merge the branch when you are done, but I often don't bother. But bottom line: if it doesn't work immediately and takes a lot of attempts, it's not you. You're not losing your touch, it's not because you're getting older. You're fine, it's just complicated.


And when you finally have that working CI build and it passes, go ahead and slap that badge on your readme!

CI passing


Since this is a privacy-friendly static web site, I'm not including the ability to post comments directly here. I do love feedback though, so I created a ticket on GitHub that you can use to leave your comments. Tell me if it's bad, tell me if it's good, but please don't forget to tell me why. So please head over there and leave your comments!

  1. Please let me know if I should explain by leaving a comment!

  2. A quick search on GitHub shows that there are currently over 6.5 million commits containing the word "dependabot", which is GitHub's automatic security-issue bot.