I don't mind using the newest compiler to get optimized more on the average case, however, I'm quite more familiar with this compiler than other newer ones. What do you guys think about this ? Should it be kept around for a few more years ?

Ok, I know it's a dumb question as I'm a high-rated expert now, but I have encountered lots of times when I was stuck in greedy problems. Even if the pattern is quite "easy" to see for most people, I could literally spend an entire hour just to recognize it. Now I wonder if I can seek advice from you guys resolve this issue as it affects me a lot on classic problems. Thank you.

I registered an USACO account several months ago. However, I did forget my username and my password so maybe that's why I cannot use "Forgot password?" feature. I wonder if there is another way to recover the account and I can come back to solve USACO problems ? Or is there any way to contact with the staffs to work on this ? Thank you.

Recently, I've been dealing with this problem Here

I have the same code submitted in 3 different compiler (C++ 14, C++ 17, C++ 20), you can check out all those below:

264203324

264204158

264204181

As you can see, C++ 14 is the balanced one and that's why I usually use this compiler for my submissions. What do you guys think about this ?

I have been stuck in the blue color for 10 months lol. A long way to GM to go but I'll make it possible :D

Hi guys, I've been introduced to dynamic connectivity for a long time, especially dsu rollback which can be operated in O(log n) for updating and asking the root. Now, I'm facing with a problem with the edges stricting to only attach i and i + 1 and since the time limit is tight, I may have to do in O(1) (I'd tried the above one but it didn't work).

I was introduced to using left right for this at the same time, however it comes to a problem that the range that I get when updating some edges nearby isn't just what I've expected.

For example, I have 5 nodes here:

If I use edge (3, 4), (4, 5), (1, 2) and (2, 3) accordingly then there exists a position u which left[u] != 1 or left[u] != 5.

Please let me know where I implement wrong or prove that it's impossible to optimize it into O(1) for both updating and getting the longest range that u is covered.

    vector <int> lef_m(n), rig_m(n);
    vector <pair <int, int>> rollback;
    auto Rollback = [&] {
        while (not rollback.empty()) {
            int u, v; tie(u, v) = rollback.back();

            rig_m[u] = u; lef_m[v] = v;

            rollback.pop_back();
        }  
    };
    auto Join = [&] (int u, int v, int temp) {
        if (rig_m[u] >= v) return;
        
        if (temp) rollback.emplace_back(u, v);

        rig_m[u] = rig_m[v];
        lef_m[v] = lef_m[u];
    };

Before this happened, I had been training nonstop for like 200 days (as you can see here)

and after more than half a year, I became an expert from a pupil

I always want to become CM or M before VOI happens and even though it may be hard to achieve, I'm gonna try as best as I can. It was an accident when I forgot to maintain my streak, however, I realised that not every time training gives benefits so this may be a short break for me lol. Anyways, wish me luck on my near future.

Let's see if I can do that in one day soon (I'll come back when I finish this achievement).