The principle of not reinventing the wheel has nothing to do with interview questions. Sure, if you're actually working on solving an actual problem at your job, then most of the time you'll be better off using a standard library function instead of rolling your own.
In the context of the job interview, however, it doesn't matter if the solution to the problem is well-known or whether it can be found in a standard library. The problem of the interviewer is not to find out how to compare strings, it's determining if the candidate will be able to write proper code if hired. And the string-compare question might be a good starting point for evaluating the candidate's level. In any case, the fact that a solution is in a standard library is irrelevant.
If I was the interviewer I'd create a new data structure Foo and then ask the candidate to implement Foo.compare(). The question remains essentially the same, so I wonder what the candidate would reply.
The trouble with the question is that proper code to compare strings is almost certainly going to be a call to some existing library function. There are only rare cases (e.g. when you're one of the 50 people in the world who implement libc) that it makes sense to write it yourself. It's not clear from "code a string compare" exactly which set of wheels the interviewer wants reinvented: if strcmp is out of bounds, can you use strlen and memcmp? Because if strcmp was somehow buggy, that might be a reasonable thing to do. If the problem is that strcmp is too slow, should we maybe drop to assembly? Or change to a counted-string representation to avoid byte-by-byte operations? Or calculate hashes when strings are mutated, or intern them?
(Maybe in C strcmp is a bad example, since
while (*s && *s == *t) {
s++;
t++;
}
return *t - *s;
> If I was the interviewer I'd create a new data structure Foo and then ask the candidate to implement Foo.compare().
I think that's a better approach.
I love the GGP's solution for the same reason. If the regex is compiled only once, it may be only marginally slower (if at all), and it significantly improves readability and maintainability, and tremendously reduces the chances for bugs.
I'd hire the him. I have the impression most Google-interviewers (and similar dudes/dudettes) wouldn't. I wonder why.
Perhaps that was rhetorical, but I'll answer anyway:
Being able to wire up existing libraries to accomplish a goal is a pretty low bar to set as far as proficiency goes. Google doesn't want code monkeys. The solution above is perfectly good from a software engineering perspective, but it doesn't show the depth of the candidates knowledge nor how strong their grasp of CS techniques is.
Google's interviews are more like IQ tests than software engineering tests, using CS as the measuring tool. When you're Google you can afford to be that selective.
I suspect it's because the match group is being replaced with the last match rather than added as another group, but it will work as a state machine, and is pretty much equivalent to the backtracking example in the article (although with much less code, and no memoization).
That said, I think that the reason interviewers ask about functions for which we have well-known implementations is to see whether or not you know how they work and/or could implement them yourself. Nobody will reasonably expect you to implement your own string comparison routine, but you could score points if, for example, you said Boyer-Moore for string searching rather than the naive iterative version.
Standard string comparisons exit on the first mismatched character, which is insecure.