I'm not strictly opposed to pre-commit reviews, but I think it's incorrect to be dismissive of their cost.
My last job was perforce (no unsubmitted commit stacking) with everyone developing directly against a single branch. In theory, code reviews were supposed to happen before hand.
In practice, I submitted small changes from the hip, and complained when my coworkers gave me a week long spitball of unrelated features in a single mega-commit. Not that I wasn't guilty of that myself. Bleh.
- def f(x: X): Option[X]
+ def f(x: X): MyErrorType \/ X
Some changes simply cannot be broken into tiny atomic pieces. Often these are important changes for long term maintainability.
The Right Way (TM) to make that change, then, is to make only the change to the result type in the first commit (along with the pattern matching necessary to ignore it), followed by a commit for each place that the updated result type can be usefully used.
Change the code without changing the behavior: i.e. REFACTORING. Run the tests, verify you didn't break anything, then start writing tests for the behavior you wish to change, followed by changing that behavior.
If you think you're at atoms you probably haven't seen strings.
If you fix the behavior everyplace the compiler complains, you are guaranteed to resolve most issues (ignoring overloaded methods, e.g. getOrElse). If you use a toOption method on the union type to make a tiny atomic change, the compiler will no longer complain in all the places that mistakes are being made.
I suppose an intermediate way might be to rename f to f_DEPRECATED, replace everywhere, and build the new method returning the right type. But in my experience its better just to bite the bullet and change everything.
If the pull request is small, fixes a bug or is a blocking you, then you can spend you quick-code-review-chip for the week and ask the team to look at it at their first available opportunity. (Sometimes reviewed within 15 minutes, usually an hour tops)
It allows team members to ask the rest of the team to unblock them quickly. Limiting it to one a week prevents abuse of their team mates' time and encourages all team members to plan their work out ahead as to reduce blocking.
We also wrote a quick HipChat bot that keeps track of who's spent their quick-review-chip too.
Ordinarily this final merge is fairly painless, since you ideally won't have many code reviews that call for sweeping changes in your design. If you do, that's a sign of failure earlier in the process.
In work I currently have 3 pull requests open all touching different sections of the product. The 3 PRs are all at different stages of the review process. For example, I'll be closing one first thing tomorrow and deploying it out (don't like deploying in the evenings - its bad luck). Another I just opened as I left work today, so I doubt anyone's reviewed it yet.
I'm sure Quora has a big enough codebase and not enough developers (like most companies) that they always have a list of things a mile long that they would like to do or should do, be that new features or just refactoring.
2-3 days isn't long for a review and iteration cycle in itself. I have one PR that is currently open for 5 working days at this stage. We decided that it didn't quite fit the bill on the first draft . It is blocking progress on two other tasks in our backlog but we also have 14 things on our backlog of goals for this week. The team is working around it while we finish it off or create issues for things we think have fallen outside the scope of the initial job.
One thing we've noticed alot on my team is that we try and make our pull requests for review as small as possible. We all have a rough guide of about max 500 lines. Sometimes we do go over but its a rough guideline. Smaller PRs are just easier to review for everyone involved. Reviewers don't have walls of code to break through or need to keep tons of context in their head.
Post-commit reviews (or post-merge into master or whatever you process is) is an interesting idea. I'd be curious to see what my team think of the idea tomorrow for a small sub set of our codebase.
I've found mentoring other developers (whether it's explaining "Here's the facets of how we test All The Things", or "Here's how git branching works") to be profoundly fulfilling. I love explaining things. The best part is, fostering the idea of teaching each other means that others can mentor YOU as well on things, and you (and they) don't need to feel threatened.
Yes, but context-switches are expensive, particularly when you have several reviews in progress across several projects.
>Even if each round of review takes 2 days and there are 2-3 such iterations, it is easy for the author to stay blocked for good part of a week leading to wasted time.
How can the author of the code be blocked while waiting for a code review? Can't they use that time to review someone else's code or to work on something else in a separate branch or, heaven forbid, learn or train themselves?
I mean we never seem to find time to train developers but the time between writing code and having it pass review could be used for education not just sitting around and waiting.
>For code reviews to work well, each change should be reviewed by the people who have enough context on that change. It's even better if these code reviewers are responsible for “maintaining” the code they are reviewing so that they are incentivized to make the right long-term trade-offs
This makes sense at first but don't you also want more people looking at the code who don't have complete context so that the context must be explained to them and then they can learn more about the code base? Isn't reducing the bus factor also a priority?