4 ms·

When did code reviews become so ubiquitous? Articles complaining about the reality of code reviews seem to be commonplace. But if they're not working out, why

by matfil 8y ago

When did code reviews become so ubiquitous? Articles complaining about the reality of code reviews seem to be commonplace. But if they're not working out, why not just abandon them, maybe try something different?

marcinzm 8y ago
I'd argue that if you have a toxic team culture then no matter what you do things will escalate. Code reviews are a focal point but if they didn't exist then the focal point would move somewhere else. For example, engineer A commits some code and then engineer B rewrites it and then they get into an argument about it during a meeting. Toxicity in code reviews is at least auditable so a pro-active manager can nip it versus someone going off in a meeting they weren't part of (in which case it becomes a she said/he said situation).
- matfil 8y ago
  I don't think "toxic team culture" is the only way code reviews get frustrating. Other issues include frustrating waiting for reviews to be completed or the flip-side of people being knocked out of flow to do a review because "reviews are top priority". Relatedly, arguments about what the appropriate size is for a reviewable change. The one alternative that (some) code-review advocates seem willing to accept is XP-style full-time pair programming. But that leaves even less room for individuality and solo accomplishment.
btilly 8y ago
The issue is that code reviews are hard on people and good on code bases. Dollar for dollar, code reviews are more effective at finding and fixing bugs than any other activity we can do, including QA tests. However code reviews are very hard on people, and can easily create conflict. Which one matters more, and how careful people are, varies widely by organization.
- jacques_chester 8y ago
  My recollection is that the studies showing that code reviews have high defect yield are for extremely strict methods. Everyone in a room, prescribed roles for each person, line by line, checklists, forms, the whole nine yards. Comments on pull requests are not on the same level and those findings shouldn't be hastily generalised.
  - btilly 8y ago
    My impression is that comments on pull requests do not catch as many defects as formal review, but catch even more per dollar spent.

by matfil 8y ago

marcinzm 8y ago

I'd argue that if you have a toxic team culture then no matter what you do things will escalate. Code reviews are a focal point but if they didn't exist then the focal point would move somewhere else. For example, engineer A commits some code and then engineer B rewrites it and then they get into an argument about it during a meeting. Toxicity in code reviews is at least auditable so a pro-active manager can nip it versus someone going off in a meeting they weren't part of (in which case it becomes a she said/he said situation).

matfil 8y ago
I don't think "toxic team culture" is the only way code reviews get frustrating. Other issues include frustrating waiting for reviews to be completed or the flip-side of people being knocked out of flow to do a review because "reviews are top priority". Relatedly, arguments about what the appropriate size is for a reviewable change. The one alternative that (some) code-review advocates seem willing to accept is XP-style full-time pair programming. But that leaves even less room for individuality and solo accomplishment.

btilly 8y ago

The issue is that code reviews are hard on people and good on code bases. Dollar for dollar, code reviews are more effective at finding and fixing bugs than any other activity we can do, including QA tests. However code reviews are very hard on people, and can easily create conflict. Which one matters more, and how careful people are, varies widely by organization.

jacques_chester 8y ago
My recollection is that the studies showing that code reviews have high defect yield are for extremely strict methods. Everyone in a room, prescribed roles for each person, line by line, checklists, forms, the whole nine yards. Comments on pull requests are not on the same level and those findings shouldn't be hastily generalised.
- btilly 8y ago
  My impression is that comments on pull requests do not catch as many defects as formal review, but catch even more per dollar spent.