Rating changes for last rounds are temporarily rolled back. They will be returned soon. ×

### isaf27's blog

By isaf27, 5 years ago, translation,

Jury solution: 54047380

Jury solution: 54047416
Jury solution: 54047456
Jury solution: 54047487
Jury solution: 54047513
Jury solution: 54047561
Jury solution: 54047626
Jury solution: 54047665
• +62

| Write comment?
 » 5 years ago, # | ← Rev. 2 →   +140 Hard B, trivial C and D, cool E. And I completely understand that it's easy to misjudge the difficulty of B.I don't like limits in F and I want setters to take some things into consideration in the future. Here, the jury solution (attached in the editorial) takes 40% of TL what is quite tight, especially considering the fact that it does all modulo efficiently — avoiding the % operator if possible, doing if(x >= MOD) x -= MOD instead. You should do that in brute force solutions and make sure they still don't pass, but the author solution shouldn't be written very efficiently. Remember that participants won't necessarily implement everything in the same way. Do you really want to penalize them for writing a code that is twice slower? And let's remember about a bit slower Java, though maybe it wasn't an issue here — because modulo was the heaviest operation. TL 8-10s and ML 512MB look much better to me (the author solution takes 118MB and it would barely pass after replacing ints with long longs).Tight limits can be forced by a dangerously fast brute force and maybe everything was chosen as good as possible in this problem, I don't know. Even in that case, I think that my post still can be of use for future setters.Remember this: limits should be chosen in such a way that naive solutions don't pass. If this condition is satisfied, it's great to make TL to be much much more than twice the running time of your solution.And limits in C were too tight as well. Reading $500\,000$ numbers and possibly doing something with a set or a segment tree? That takes some time, especially in slower languages. It's good if easy problems are solvable with Python. Just make TL to be 2s or 3s. Why not?
•  » » 5 years ago, # ^ |   -24 I don't agree that B was difficult. If you implement brute force solution which just runs through all possible strings of length n and play around with it for a while you can clearly see the pattern.
•  » » » 5 years ago, # ^ |   +40 That is not you expect from B in six problem contest.
•  » » » 5 years ago, # ^ | ← Rev. 2 →   +8 I did B as a specialist. So I would just agree with tribute_to_Ukraine_2022
•  » » » » 4 years ago, # ^ |   0 But did you prove B as a specialist?
•  » » 5 years ago, # ^ |   -7 In problem C, segment tree/set takes $\mathcal{O}(n\log n)$ time, and author's solution is $\mathcal{O}(n)$, so such a tight TL could be the result of one of following considerations: 1) the author didn't know about $\mathcal{O}(n\log n)$, so he didn't see any difference between $500\,000$ and $200\,000$ and chose arbitrarily, or he tried to avoid participants sending unoptimal unpredicted solutions; 2) the author knew about $\mathcal{O}(n\log n)$ solutions and didn't want them to pass. Actually, in this case he didn't perform the clipping perfectly as the segment tree approach turned out to be productive for some participants, but my segtree failed the system testing (though changing the vectors to arrays and switching compiler from С++17 to C++14 let me gain an ok in upsolving in 998ms).
•  » » » 5 years ago, # ^ | ← Rev. 2 →   0 I don't know about your claim — I implemented C in $O(nlogn)$ using priority queue, and I got AC in 327 ms. Code is here
•  » » » 4 years ago, # ^ | ← Rev. 2 →   0 I implemented $O(nlogn)$ using set and passed (in 265 ms).
•  » » 5 years ago, # ^ |   +10 C can be solved in O(n)
 » 5 years ago, # |   +5 can anyone explain me div 1 C with more detail how to solve this with set/segment tree.
 » 5 years ago, # |   +5 When will the editorials of rest of the problems be translated to english?
•  » » 5 years ago, # ^ |   +48 BTW, is it only me who is thinking that after AtCoder in japanese and Codeforces in Russian, Codechef will start publishing editorials in Hindi and will refuse to translate them?
•  » » » 4 years ago, # ^ |   +1 Then only they will get our point
 » 5 years ago, # |   +53 Please translate
 » 5 years ago, # | ← Rev. 3 →   +5 UPDATE: Answered hereIn Div2E/Div1C.Does anybody know why this 54049263 gives MLE? When I change stack to set I get AC in this 54047364 My Approach: Construct a directed graph where edge (u,v) means a[u] < a[v] and then do topological sort on that graph. The way I construct it is if nxt[i] = j that means there is an edge(i, j), also there is an edge between all elements [i+1, j-1] and i since all of them must be less than i. I only keep track of the least previous element. Here's an example where ix, jx means nxt[ix] = jx:i1 i2 i3 ... j3 j2 j1. Edges are (i1, j1), (i2, j2), (i3, j3), (i2, i1), (i3, i2). I don't add (i3, i1) since it's redundant.In a case like i1 .. i2 .. j1 .. j2 In the MLE submission I exit immediately and say that it's impossible. In the AC submission with set I let the topological sort handle it.
 » 5 years ago, # |   +24 isaf27 why no english editorial ?
•  » » 5 years ago, # ^ |   +22 I'm sorry, before the round I've written only Russian editorial. Now I should translate it, but I have a very hard time at university. Before the end of the week, I will finish. You will see the update in "Recent actions" section.
•  » » » 5 years ago, # ^ | ← Rev. 3 →   +27 First of all, thank you for having prepared a CF round. It helps the community and it requires a lot of work. ... But, why couldn't you write the editorial in English from the beginning? In that case you wouldn't have to spend extra time translating.
•  » » » » 5 years ago, # ^ |   +24 Yes, I wanted to do this. But I started working on test cases shortly before the round and was busy doing test cases. Next time I will definitely begin from doing statements and editorials because it is usually hardest work for me :)
•  » » » » » 5 years ago, # ^ |   +10 Thanks
 » 5 years ago, # |   +19 Where could I find the solutions in English pls
 » 5 years ago, # |   +20 English solutions please!!
 » 5 years ago, # | ← Rev. 2 →   +5 Can anyone help, what is 'l' in the explanation of D above?
•  » » 5 years ago, # ^ |   0 It is from the statements — any position of the string t occurrence.
 » 5 years ago, # |   +27 Now editorials in English are available, sorry for waiting. If something isn't clear you can ask in comments.
 » 5 years ago, # | ← Rev. 2 →   +10 in problem c (The Party And The Sweets) in test case 1 what is wrong in this ans: b1- 1 1b2- 3 4b3- 1 1ans=11 ??
•  » » 5 years ago, # ^ |   0 It was really the most popular question during the round :) Check the formal explanation of the problem, a minimal number in the second row is wrong.
 » 5 years ago, # | ← Rev. 2 →   +5 Div2D: Editorial is not helpful at all.You explained the pattern to generate the answers.But how to come up with that kind of pattern in first place.Please share some ideas in what direction to think for problem like that to come up with patterns like that. Stating formula and why formula works is not helpful at all.
•  » » 5 years ago, # ^ | ← Rev. 3 →   +11 To solve this problem it's better to solve some cases, for example: k = n, it is an obvious case k = n — 2, I think it isn't hard to guess, that 0101010101 is the correct example k = n — 4. This is the hardest case. But if you will think for some time, checking some examples, you will guess, that periodic string is correct And after the last case, it's easy to guess the pattern. I think it is a good way to solve this problem.
 » 5 years ago, # | ← Rev. 2 →   0 About Div2-D, condition on the assumption of periodic a and minimum unique length k, I got $a >= (n - k + 1) / 2$doesn't it means the period not unique but have a lower bound? But I failed on program checking.And when the parity of n and k not same, even the lower bound of a is wrong. Why is that? Really doubtful.
•  » » 5 years ago, # ^ |   0 Hi, I guess, ( n%2 = k%2 ) ensures about parity being same. Unless you are talking about the bonus question provided at the end of solution.
•  » » » 5 years ago, # ^ |   0 Actually I have doubts to both problem. In the original problem I cannot justify the unique of a, else on bonus I completely have no idea... Anyway, Thanks for your reply.
 » 4 years ago, # |   0 I think the editorial's proof is overly complicated for Problem E. We can just prove it by induction. (construction from the right side)
 » 4 years ago, # |   0 Does the checker code of Div1B/Div2D use string hashing? I think there is some (very small) probability of failure. Is there any deterministic checker?