Frustrating is the only way I can describe it. The assigned task was quite interesting and I think it could be decent as an initial evaluation of C++ skill. However, in practice, it was incredibly annoying. There are a few ambiguities in the requirements, and some edge cases that aren't specified. At some point I found myself trying to guess what the person who wrote the task was thinking, which is not a hallmark of a good test. I've been going over my code, and I'm certain I've implemented it right for what I thought was the correct behaviour. Given the same task description multiple people could write code that behaves differently and still conform to the specification, which is again not a good thing for an automated test. What's more frustrating is that this could simply be solved by providing a few test cases and their correct results. This won't compromise the integrity of the task, what it will do is clarify some important details that are left unclear in the description. I'm not saying the task is unsolvable, I think I solved the most difficult part (O(logN) for one operation, O(1) for the rest). I'm saying that skill can only take you so far when you end up having to gamble on what the task creator was thinking, only for you to guess wrong and get a "sorry you wasted 6 hours, but you got 'correctness' wrong, bye!".