4 ms·
Those pesky pull request reviews
- auslegung 6y agoAt my previous place we did promiscuous pairing, where each person stays on a task for 3 hours, but switches roles at the halfway point from driver to navigator. http://engineering.itpro.tv/2018/12/07/our-first-experiment-with-promiscuous-pairing-is-a-success/ http://engineering.itpro.tv/2018/12/07/our-first-experiment-.... I think promiscuous pairing and mob programming result in similarly higher quality code and higher velocity. We tried mobbing a couple of times but it’s very different so it didn’t catch on easily and we didn’t take the time to become better at it. I kinda wish we had :D
- HomeDeLaPot 6y agoHow does this square with "mob mentality?" I wonder what the negative side effects of programming in a group might be. Maybe the group is more liable to jump on the first answer to a big problem rather than think it through carefully to find the best answer. Maybe some team members actually perform better on their own, or maybe it depends on the tasks at hand and the experience profiles of those involved. Also, I'm still skeptical that it makes sense in terms of code quality and velocity to have 6 people work on one feature rather than 3 pairs working on 3 features and syncing up every morning. (That's what I did in the past, we had a "tech time" every morning for a half hour or more where we'd walk through code, do reviews, and make decisions as a team.)
- NBJack 6y agoOne interesting detail in this post is the comment they made about pulling down the PR branch themselves, setting up the test harness, running it, etc. That is remarkably time consuming. Given the complexity of many of the systems I've had the privilege of working on, that would easily take an hour for a single PR before I could even begin reviewing, to say nothing of the potential disruption to my own work in the same code base. In my experience, this is where a better CI/CD pipe with automated precommit checks shines. Does the code at hand do what the author said it would? Do the tests they added to exercise it cover the right cases? Is anything amiss? Can we improve code quality? Let your automated tests handle the rest. If there's one thing I agree with the author on, trust has to start somewhere. Time a problem? Setup a small PR meeting. Let the author brief folks on what it does, encourage honest feedback, and if possible approve it in the same timeframe. Works great also for large PRs that tend to scare folks away (i.e. mass refactorings, etc.).
- zach_garwood 6y agoPR reviews are my favorite part of programming on a team! I've learned so much from both older and younger devs.
- conradludgate 6y agoPRs at my company are pretty painless. We have automated tests covering a lot of the code so we don't need to analyse the code for ages before approving it. It's just to raise some visibility on the code we're adding/removing. Sometimes we flag dodgy code or think of a better way to compute something but usually it's a simple approve