Breaking into open source

For many years I put open source software on a pedestal. I wanted very badly to contribute, but I never felt qualified to do so. I never encountered bugs in the frameworks I used, so I couldn't work on unreported issues, and all the open issues clearly required far more context and expertise than I had. All the "easy" issues seemed to be fixed by a core developer far before I had a chance to dig into them.

However, in the last year I've finally managed to find a way into the community, and it isn't nearly as impenetrable as I had made it out to be. This blog post contains the guidance I wish someone had given me four years ago.

Pick a project

The key to making valuable contributions in context. As a beginner in open source, you are going to be relatively slow at gaining context. To that end, focus your energy on one project and try to select one that sets you up for success. Here are some criteria to consider:

  • How large is the codebase? The less code, the easier it is to have contextual understanding of all of its pieces.

  • Do you use the project daily? A library that you depend upon at work or for a serious side project is a great starting place, since you will already have contextual awareness from the perspective of a user. Understanding the inner workings of the library will also pay dividends with your other work.

Gain context

Once you've selected a project, spend at least a few weeks gaining context on that project. You should be trying to gain technical context (how does the software function?) as well as social context (how does the core team work? What are their goals and norms?). Here are some ways you could go about that:

Social context

  • Does the project have any documentation about how to contribute? Read it!

  • Follow ("watch") the project on GitHub. Read all the incoming issues and pull request. I recommend getting a mobile app like iOctocat so that you can follow the discussion in near real time.

  • Do they have a Gitter or IRC channel? If so, idle in that channel and listen to the conversations. I've found the Gitter iOS app very useful.

  • If they have a mailing list, subscribe to it.

  • Follow the main contributors on Twitter.

Technical context

  • Read the source code. The smaller the project, the more of the code you will be able to internalize.

  • Read the tests. Tests are often an under-loved part of the code base, and can be a great place to jump in and help. They also help clarify, in detail, the expected behavior of the software.

  • Read the documentation. Does it accurately reflect the code you read? Is it consistent?

This step can serve multiple purposes. It will give you context but it may also help you identify possible improvements. Read linearly but keep notes of things to look into more deeply. Does a function look strangely constructed? Make a note and look into it later. Maybe there's a subtle reason it ended up that way, or maybe you could be the one to clean it up!

Run the project locally

Before you can start proposing changes to the projects, you need to be able to build and run the project locally. Download the project and learn about its build process:

  • Build the project

  • Run the tests and any linters

  • Run any other continuous integration tests: linting, code coverage, etc.

Finding your first issue

Your main goal for your first pull request should be to find something that is as small and inconsequential as possible. The likelihood that you are going to find and correctly fix a major flaw in a core code-path on your first try is low. Instead, focus on doing something small and doing it well. Larger tasks can come later. In the process of fixing a small bug or cleaning up a test or documentation, you'll quickly gain the context needed to to take on and discover larger tasks.

If you found something that was inconsistent or wrong while you were gaining context, or setting up the project, correcting that could be a great first pull request.

If you didn't see anything wrong, it's time to check out the issue tracker. Some projects on GitHub use tags to notify potential contributors (you!) where you might be able to help. Looks for tags like "needs help" or "beginner". If your project doesn't do this, just read through a few pages of the issue tracker and use your gut. With your newly gained context, pick the simplest issue you can find. If it looks like something you will actually be able to complete, comment on the issue to tell others that you are looking into it. This will help reduce the likelihood that someone else will jump in and fix the issue before you figure it out.

Working on your first issue

Working on open source software is unique in that the information you have at your disposal is quite limited. If you have a question about why something works the way it does, you cannot wander over to the author's desk. Instead you have other tools at your disposal:

  • git blame. There is a vast quantity of context in the git history. Open source projects usually have verbose commit messages which detail the intentions of the committer. If some code looks suspect, dig into the git history as deep as you need to in order to find the commit that first introduced it. You may need to go several commits deep before you find the actual commit that introduced the thing you are investigating.

  • Pull Request history. If the commit message does not answer your question, you can use the commit SHA hash to track down the GitHub pull request that included that commit. This will give you additional information like feedback that the committer was responding to which lead to her final commit.

  • The issue tracker. Get good at searching the project's issue tracker. Often things that appear "wrong" to a newbie but are actually intentional will have multiple closed issues which contain defenses of the current approach.

Writing your patch

All this work, and you're finally ready to actually write your code! Take this opportunity to slow down and write your most elegant code. The distributed (across time and geography) nature of open source software development, means that projects must lean more heavily on the code for communication. As Steve McConnell says in his book Code Complete:

Write programs for people first, computers second

This is never more true than in open source. Your code must be able to speak for itself. You won't get the opportunity to explain or defend it to the person reviewing your change. They must be able to read the code and intuitively understand your intentions.

Conform to the project's code style

Some projects document their code style as part of their contribution guidelines. Others have linting tasks you can run to check your code. Either way, make sure that your code blends in with its surroundings. The entire project should appear as if it were written by a single individual. Pay special attention to:

  • Naming conventions

  • Spacing, layout

  • Code organization

Prioritize testing

If you are fixing a bug, be sure to include a regression test that ensures the bug is not reintroduced. If you are adding functionality, write tests for as many edge cases as you can think of. This will be a benefit to the project as a whole, but it will also greatly increase the likely hood of your patch getting merged.

The person reviewing your patch will probably not manually test your code. Instead they will rely on existing tests to assure themselves you didn't break anything. Similarly, they will not give you the benefit of the doubt that you manually tested anything. Instead they will look for tests. Again, your code must speak for itself. Don't tell the reviewer that you tested something, show them, by including an automated test.

Your commit/pull request message

As I said before, the nature of open source is such that you do not have direct access to the author of the code. Similarly, future developers will not have access to you! This means artifacts like commit messages and pull request messages are doubly important. Give as much detail as possible in your commit message. Put as much care into your commit message as you do into your code, if not more. You should include:

  • Your reasoning for why you chose the approach you did.

  • Other approaches you considered, and why you rejected them.

  • Results of any research you did.

  • Links to related issues, pull requests or discussion.

Bonus points: Construct a narrative of how the project ended up in its current situation. When was the bug/inconsistency introduced? Do you have a hypothesis as to what lead to the mistake?

Opening your first pull request

Here's where all your social context will begin to pay its biggest dividends.

You only get one chance at a first impression. Expect the standards to be very high. Expect to get critical feedback or even an outright rejection. Think about opening your pull request as starting a conversation with a proposal. Expect your choices to be challenged. Be ready to defend them, but be equally ready to incorporate feedback. If you put significant effort into your pull request, you will most likely at least receive some kind of critical feedback. Even if your pull request is rejected this is not failure! In fact, critical feedback is an opportunity to learn, and grow as a developer. Take it seriously, internalize it, and try again!

Success!

If you follow this advice, after an attempt or two, you will most likely get a pull request merged into your favorite project! Awesome! I can say from personal experience that there is something illogically thrilling about knowing you were able to give back to a project you use every day.

The context you gained from this entire process will create a feedback loop. Often in the course of working on one issue, you will notice something else that you could do. If not, you are now deeply familiar with one part of the code base. Keep your eyes open for related issues, and pretty soon you'll be the expert jumping in to fix things before the other newbies have a chance to dig in and figure out what's causing the bug.