Re:The Code Review Pyramid
by hinkley on Tuesday March 22, 2022 @11:57PM
(#30774773)
There's a flip between when the feedback is super, super helpful (consensus building, as you say), to a moderate inconvenience to screwing up your estimates because you seem to always post a PR right after the reviewers have started some new thing.
Then farther down that continuum someone is busting your chops for some stupid reason that, after you get back from lunch or coffee, you realize is completely valid (or even gives you an idea for an enhancement) and okay you change how a conditional works or add some more tests.
It makes me wonder if in a corporate setting PRs would be better treated as pseudorandom audits, based on a random factor and a weight for automated linting (so people don't learn to game the linter, which can be some of the worst code you ever see.)
A related issue is that most tools make it pretty difficult to look at and interact with old PRs. Feedback loops have some slop, and while it's better that you get the feedback as you are writing the code, fast follow often works as well, and I think we'd be better as a group if you could request revisions on a recently landed PR after the fact. This is mostly okay but you really need to fix your variable names, conditional block, and you missed a boundary condition over here that might not break today, but there are already stories in the backlog that will break it for sure.
Then farther down that continuum someone is busting your chops for some stupid reason that, after you get back from lunch or coffee, you realize is completely valid (or even gives you an idea for an enhancement) and okay you change how a conditional works or add some more tests.
It makes me wonder if in a corporate setting PRs would be better treated as pseudorandom audits, based on a random factor and a weight for automated linting (so people don't learn to game the linter, which can be some of the worst code you ever see.)
A related issue is that most tools make it pretty difficult to look at and interact with old PRs. Feedback loops have some slop, and while it's better that you get the feedback as you are writing the code, fast follow often works as well, and I think we'd be better as a group if you could request revisions on a recently landed PR after the fact. This is mostly okay but you really need to fix your variable names, conditional block, and you missed a boundary condition over here that might not break today, but there are already stories in the backlog that will break it for sure.