But... demonstrating the ability to write subtle code is a useful question. "Describing" or "discussing" algorithms isn't the same thing at all.
I mean, I can't speak to the thinking of the original interviewer, but this is what I'm looking for with that sort of question. And I don't see how a binary tree is a bad choice. Again, it's something that everyone sees in school, so it doesn't require a ton of description in the interview.
I guess I put the question back to you: if you won't write a binary tree traversal in an interview, what subtle code would you be willing to demonstrate? And why is that better than a binary tree?
I really can’t more strenuously disagree. At Giant Search writing “subtle” code is very strongly discouraged. The very last thing I want is candidates with a penchant toward subtlety.
The point is to demonstrate capability, not "penchant". I mean, look, subtle code happens. Maybe it shouldn't. But it does, and it has to be fixed. There was a story here just a few days ago about some Project Zero work to find a bug in Chrome that involved a state machine with something like 50+ states! And realistically finding people who can do that requires that they be able to also do things like traversing a binary tree, right?
So I'm going to ask the question one more time: if you won't traverse a tree in an interview, how else do you propose to select for people able to reason about that kind of problem in practical code?
I mean, I can't speak to the thinking of the original interviewer, but this is what I'm looking for with that sort of question. And I don't see how a binary tree is a bad choice. Again, it's something that everyone sees in school, so it doesn't require a ton of description in the interview.
I guess I put the question back to you: if you won't write a binary tree traversal in an interview, what subtle code would you be willing to demonstrate? And why is that better than a binary tree?