如何进一步优化这个简单的算法？

Question

Note: this is a programming challenge

---

This challenge requires usage of std::set .

• A number n
• n lines with each j and k

Sample input:

 5 1 4 2 1 1 6 3 4 1 2 

j is for the operation: 1 to insert k , 2 to delete k (if there is a k ), 3 find k

For j == 3 , output Yes or No if k is in the std::set .

---

I made different versions of the algorithm, but all are way too slow. I tried different std functions, but std::find seems the fastest one, but is still too slow. Implementing my own would be even worse, so maybe I missed a function from the library. I really have no idea how to optimize it further. Currently I have this:

 int main() { std::set<int> set{}; int Q = 0; std::cin >> Q; int type = 0; int x = 0; for (int i = 0; i < Q; ++i) { std::cin >> type; std::cin >> x; if (type == 1) set.insert(x); else if (type == 2) set.erase(x); else if (type == 3) std::cout << (std::find(std::begin(set), std::end(set), x) != std::end(set) ? "Yes" : "No") << '\\n'; //Condition may be std::find, std::count or std::binary_search } return 0; } 

---

The challenge requires it to run under 2 seconds. Here are the results of mine:

 Function Time % over std::find -> 7.410s 370.50% std::count -> 7.881s 394.05% std::binary_search -> 14.050s 702.50% 

As you can see my algorithm is 3x slower than the required algorithm. The bottlenecks are clearly those functions:

 Function % of total time std::find -> 77.61% std::count -> 80.80% std::binary_search -> 87.59% 

But currently I have no better idea than to use std::find or similar functions. Does someone have a way/idea to optimize my code further? Or are there some obvious bottlenecks that I'm missing?

Answer 1

You want to use set.find() and not std::find(set.begin(),set.end()) .

set.find() will use the set's internal structure to locate the key in O(log(n)) time. Whereas std::find is a linear search, requiring O(n) time, no matter what container type is used.