This is a fast blogpost please excuse the brevity.

Summary

  • A fast review and merge cycle is incredibly valuable
  • When reviewing, consider highlighting which comments are obligatory because they stop some regression, and which are merely commendable because they would add some enhancement
  • Don’t forget that better is the enemy of good, especially when prioritizing against all of the other work that can get done.

Example

As an OSS developer I use many projects, and when I see a simple problem in a project that I can fix quickly, I like to fix it. For a trivial example, maybe I decide to add a type annotation to a function:

-def ensure_bytes(s):
+def ensure_bytes(s) -> bytes:
     """Turn string or bytes to bytes

     >>> ensure_bytes('123')
    b"123"

Awesome. This takes me less than two minutes to make the fix and open a pull request. I can fire this off and proceed with my normal work.

Critical review

On review, someone might helpfully mention that I actually made a mistake, and that this function can also return None in some cases. They make the following suggestion:

-def ensure_bytes(s):
+def ensure_bytes(s) -> bytes | None:
     """Turn string or bytes to bytes

     >>> ensure_bytes('123')
    b"123"

This is great! They helped me to understand the problem better, and avoid doing something harmful to the project. The back-and-forth here takes a few minutes, and pushing a new commit takes a few minutes, but I’m still happy to do this, and I learned something along the way. I’m not thrilled about stopping my normal work to do this, but it makes total sense. We need to make sure that things are correct.

Non-critical review

Another reviewer then arrives and suggests that I also type the inputs, and change the docstring to reflect the ambiguous typing.

-def ensure_bytes(s):
+def ensure_bytes(s: str | bytes | None) -> bytes | None:
     """Turn string or bytes to bytes

     Except if the input is none, in which case it passes through

     >>> ensure_bytes('123')
    b"123"

This is a good idea! Someone should do this! However, I don’t think that it necessarily has to be me. Also, I’m off doing other work, and this is becoming more of a distraction.

Maybe I do this anyway (it’s not hard) but I’m probably less likely to come back here in the future.

Distinguish between critical bugs and nice-to-have enhancements

It’s great to make suggestions on PRs for enhancements or how things might be improved. However, we should always remember to mark them as optional for PR authors.

We can always come back and add the enhancements ourselves after their improvements are already in.

Prioritize ruthlessly

However, when we switch out of review mode, and into authoring mode, we might realize that the enhancements aren’t actually as important as what we were planning to work on. If they’re not something that we’re likely to prioritize ourselves, then we definitely shouldn’t have been requiring someone else to do this work for us.

This is especially true in a professional context, where time is in short supply. Many of us come from a volunteer context, where working on code and reviewing is a free thing that we ourselves donate or that we do for fun. However, in a professional context this is often not true, and the back-and-forth on a PR has very concrete costs (easily rising into the thousands of dollars). At work we prioritize with our teams to make sure that we’re using our time wisely, but we often forget about this prioritization process when we enter review mode.

Parable

Imagine that you need to travel from home into the city. It’s a long distance but a beautiful day outside, and so you decide to walk. You’re moving quickly so that you can get there and back before sunset.

You start walking swiftly towards your goal, but along the way you spot a piece of trash. You’re a good citizen and so you pick up that trash so that you can put it in the bin the next time you see one.

As you approach a trash bin to put the trash away someone stops you and points out that this bin was actually for recycling. They helpfully point you to the trash bin across the road. You thank the good citizen for stopping you from making a mistake, and bring the trash to the appropriate place.

A few folks around the trash bin notice that you’re cleaning up the streets, and point you to more trash that you can pick up. That’s fine, you’re happy to help out, and so you pick up some more trash to clean up. They get excited about this and point you down a side road that needs even more cleaning.

You’re not feeling great about this, you need to get to town and back, and daylight is running out. These kind folks are well meaning, but they’ve mistaken your intent as “clean up the streets” rather than “get to town”. You calmly explain to them that you’re not actually here to clean up the streets, explain that you’re headed to town to get medicine for your family, and that you really must be off. Now that you’ve explained your context they fully understand your situation and indeed encourage you on your way, perhaps even pointing you to a bus line that you didn’t know existed.

Summary

As we as OSS developers shift to doing professional open source maintenance we need to shift how we think a little. Open source is amazing at having strong impact, and at collaboration, but often lacks the focus or direction of professional software engineering shops. We should think about how we bring the focus of professional work into our open source development practices.

A great and ubiquitous example of this is PR review, which is a common source of scope creep, and often in directions that are not primary foci of the team.

Fast PRs also feel great, we should do more of those, if we can keep them safe.

Total write time: 21 minutes

Total edit and publish time: 2 minutes


blog comments powered by Disqus