How to make better Pull Requests: Adding Steps to Test
Do your Pull Requests (PRs) seem unfocused? Do they languish for days before someone finally gets around to looking at it? Is the feedback that you get unhelpful? Are you finding nasty bugs in your code only when someone looks at your PR's?
I've been there. And I know a better way - follow along!
At my current position we have adapted our process such that now our Pull Request stage is one of the most crucial parts of our workflow. This is the stage when we not only look at the code, but also evaluate the functionality (i.e. QA) and make sure that everything works, and nothing breaks unexpectedly.
In our workflow, the next stop after a PR gets merged is production. So, we have to take them pretty seriously.
What helps most when opening a Pull Request is writing out, clearly, a few pieces of information so that potential reviewers are much more quicly orientated.
- The objective of the work / the bug you're trying to fix
- Steps to test your solution
Adding the objective / bug
You have to keep in mind the mentality of the other reviewers on your team who you're asking to look at your Pull Request. They're busy with their own work - they would probably much rather be solving their problems.
Spell out simply and clearly the point of your Pull Request. I can't tell you how many PR's I've seen with no 'description' at all. The titles of your Pull Request, 98 times out of 100, are not enough information to go on.
Remember that your teammates aren't looking at your ticket - they don't necessarily know what it's about. And even if they are familiar with the work needed, they might not have that information top of mind when they get around to looking at your work.
Within your Pull Request's description - either add some semblance of the acceptance criteria provided, or the steps to reproduce a bug.
Adding Steps to Test
The biggest thing that I've seen help PR's along is adding clear steps to test your work. They should be easy to follow, short, steps in order.
Something like this, if you're fixing a bug (in this case, a weird layout overlap):
- Start from the
- Visit the /about/ page
- Note that the author names are overlapping the author photos
- Pull down this branch
- Refresh the /about/ page
- Note that the overlap is fixed
Or if you're adding a feature:
- Pull down this branch
- Visit the /store/ page
- Note the new promotional component at the top of the page
- It should pull the featured image from our CMS's API
- The headline and supporting text are pulled from our CMS
- The component layout works across breakpoints
- Clicking the Call to Action button takes you to the catalog page for the featured item
Why spell all this out?
First and foremost, what you are now doing is making your PR actionable for your reviewers. They don't have to think about how to evaluate your PR. Sure, they can nit pick coding decisions, but you're done the hard work in terms of laying out what they should be looking at.
You want to make this as easy as falling off a log for your reviewers. The less that they have to think about your PR, the faster they'll actually look at it. Your objective should be to remove as much friction as possible for reviewers of your PR.
There is one big drawback to this approach - opening PR's takes a lot longer. Sometimes it takes me close to half an hour just to fully fill out my PR description.
But I think this time will more than be made up for by the speed with which your PR's get approved. Also, if you are the one opening a PR, you're asking of someone else's time - isn't it up to you to make it as easy as possible for the other parties?
Clarity of thought
Another major benefit that I'll point out to adding steps to test to your PR is that it causes you to stop and think about what you're trying to acheive. Because you're trying to articulate these steps, it forces you to crystalize what is happening. You may find yourself considering important edge cases at this step that you weren't originally.
Or, maybe there are multiple permutations of what you're trying to make happen, and only some work. I have found also that when you're doing work that touches many places in your project, it's important to list different views individually and mention what to look for.
A blueprint for automated tests
The last benefit that I'll mention is that when you've done a good job listing out all of the ways your work should be evaluated - you now have in your posession a very good test plan.
If the PR that you just authored doesn't contain automated tests already, you at least already have a very good guide of how to go about authoring them.
I hope that with some care and some forethought, you can open Pull Requests that are actionable for your reviewers. By adding steps to test your work you remove friction for your reviewers, which makes them more likely to get your work reviewed faster. Along the way you'll be forced to think very carefully about your work so that you can articulate these steps, which will hopefully lead to better, more complete work.