### awoo's blog

By awoo, history, 16 months ago, translation, 1303A - Erasing Zeroes

Idea: Roms

Tutorial
Solution (Roms)

1303B - National Project

Idea: adedalic

Tutorial
Solution (adedalic)

1303C - Perfect Keyboard

Idea: Roms

Tutorial
Solution (Ne0n25)

1303D - Fill The Bag

Idea: Roms

Tutorial
Solution (Roms)

1303E - Erase Subsequences

Idea: adedalic

Tutorial
Solution (adedalic)

1303F - Number of Components

Idea: Neon

Tutorial
Solution (Ne0n25)

1303G - Sum of Prefix Sums

Idea: Neon

Tutorial
Solution (BledDest)  Comments (62)
 » Why this post doesn't have any comments?
 » can someone please explain a little bit about the proof of algorithm of problem F? Why considering all the delete queries in the last doesn't affect the final results?
•  » » because the delete queries are after the add queries.when considering color $t$,all the add queries is the queries with $c_i=t$ ,then for sure,they are consquent,and so for sure all the delete queries are after the add queries
•  » » » But this is true for a particular color right? Like for a color c what ideally should have been done is to perform add query and then immediately after that perform it's delete queries but in code above they are doing all add first and then all delete query.
•  » » » » No...for a certain color c and all the queries with ci=c,the queries are consquent,so they changed some of the cells into color c.There isn't any delete query between them.After that,some queries may change the color of some cells which were color c,these are the delete queries.Like this example: 2 2 4 1 1 1 1 2 1 2 1 2 1 1 2 For color 1,the add queries are the 1st and the 2nd while the delete query is the 4th.
•  » » » » » 16 months ago, # ^ | ← Rev. 2 →   Sorry but I think I didn't convey my doubt properly. I am talking about this part of code .. forn(i, clrs) recalc(add[i], +1); forn(i, clrs) recalc(del[i], -1); Here we are processing all the delete queries at the end (even after the add queries of same cell i.e. even after add query of (x, y) from c1 to c2).
 » In problem DIf in binary representation of n the i-th bit is equal to 1 and we have at most one box of size 2^i,I think this should be at least?
•  » » yes it should be at least
 » can someone please explain me about problem B why TotalG = ceil(needG / g) * (g + b) , i don't understand that clearly =)) ????
•  » » every (g + b) days have exactly (g) days good, we call it a block. So if we need (needG) days good, we must have ceil(needG / g) blocks, thus TotalG = ceil(needG / g) * (g + b).
•  » » » ok thank you guy =)) i have understood
•  » » Firstly you should know needG is the Title requirements(make sure that at least half of the highway will have high-quality pavement), then y = ceil(needG / g) means the units of good days you can meet the requirements,next the days is again g good days, b bad days and so on, so y*(g+b) is the least days you need to meet the requirements,at the end, if some highway is still not repaired, answer should add it.Now are you clearly? My english is not good, so hope you can understand my words.
•  » » » ok thank you i understood it obviously =))
 » What is the time complexity for G? nlgnlgn?
•  » » 16 months ago, # ^ | ← Rev. 2 →   Yes. You process each node O(lg n) times using centroid decomposition and each time you do it, you add 2 lines and you make 2 queries for the best path in O(lg n). Total is O(n lg^2 n).
 » 16 months ago, # | ← Rev. 2 →   1330-D HELPWhy in this test case answer is 4 ?182 61 1 1 2 2 256We can divide the box of 256 into two boxes of size 128(less than 182) in One division. Then why the answer is 4 divisions ?
•  » » Binary 182 = 10110110 So u need 2^7 + 2^5 + 2^4 + 2^2 + 2^1 Which is 128 + 32 + 16 + 4 + 2 So given = 1 1 1 2 2 256 Division 1 = 1 1 1 2 2 (128 128) Division 2 = 1 1 1 2 2 (64 64) 128 Division 3 = 1 1 1 2 2 (32 32) 64 128 Division 4 = 1 1 1 2 2 (16 16) 32 64 128 Now you can regroup as 1 (1 1) (2 2) (16) (32) (128) 16 64 All these terms in brackets give you the above values that will fill the box
•  » » » Thank You Very Much, I got it now.
 » 16 months ago, # | ← Rev. 3 →   I have a doubt, in B, does it mean at least half of the highway (consecutive units) must be done on good days or does it mean it can be done any way but at least half of it must be done on good days?Example,Should it be: g skip skip skip g skip skip skip g b b borshould it be: g b skip skip g b skip skip g b skip skipor, even bettershould it be: g b b b g skip skip skip gIf it's latter, then the minimum number of days will be different.I got this doubt after looking sample test case when n = 1000000, g = 1, b = 1000000. If done with first approach, the answer will be same as the output given (499999500000), but if done with second approach, then the answer will less than the output given (I think). ceil(n / 2) = 1000000 / 2 = 500000 Good days appear on: 1st, 1000001st, 2000001st,... days It's basically an AP where a0 = 1, d = g + b - 1 and an = a0 + (n - 1) * d. So 500000th good day appears on a500000 = 1 + (500000 - 1) * (1 + 1000000 - 1) = 499999000001 Then if we add remaining bad days (500000), then the answer is 499999500001.
•  » » yes I had the same doubt. What are the chances that the solution is wrong? In the editorial they have given in case of (n % (2 * g)) == 0 we can remove the bad days as we have counted them extra. They have subtracted b from the result in that case.But we should subtract bad days that have not been used for construction of the road aka b — g. If we do this as g is 1 in the example considered by you above the answer will be 499999500001.So it seems like there has been some negligence while formulating the solution, or I'm just a noob.
 » if(g>n)cout<
 » I am getting Wrong Answer in C. Here is the link.My approach is first check for cycle: if parent[x] exists and adj(y, x) and parent[y] != x, then error. Here parent means immediate parent.Example,input: abcdathen, a->b->c->d->a, here parent['a'] = 'a', parent['d'] = 'c' and at the end of input, we are trying to make parent['a'] as 'd' which is not possible since 'a' already has a parent assigned, thus error.If no error occurs, add first input symbol, add first child to the beginning and repeat the process add second child to the end and repeat the process add remaining characters. Is there some test case which I missed?
•  » » For my dfs approach, I checked for cycles and that degree of each vertex is less than 3.
•  » » » 16 months ago, # ^ | ← Rev. 2 →   Thanks. Yeah, that was my initial approach and I am sure it gives AC. But, I thought my current approach automatically handled the check for degree. I will have a look at it again and see if the check is required to be explicitly handled.EDIT: Yeah the condition wasn't handled, now it gave AC :-)
 » Has anyone used adjacency list to solve C ? If so could you share your idea/implementation thanks
•  » » Here is my submission — 70873445 Just store the letter as an undirected graph and make sure that every vertex has degree 1 or 2. Then loop through all the vertex and find the vertex which has degree 1 because that would be the start/end of the graph.Now go for dfs, if you find a cycle in the graph, that means it's not possible, otherwise dfs will give you the starting letters of the keyboard.Then just add the remaining alphabets back in the string.
•  » » » Thank you as usual. Mentor ;)
•  » » » your code is very messy didnt understand a thing ... or may be I m big big noob
•  » » » » I would suggest you read my explanation and try to implement it yourself if the idea is clear.
 » 16 months ago, # | ← Rev. 2 →   The problem statement of F says $1 \le c_i \le max(1000,\lceil \frac{2⋅10^6}{nm} \rceil)$.Should not that be $min$ instead of $max$?
•  » » 16 months ago, # ^ | ← Rev. 2 →   [Deleted]
 » Can someone help me understand why my code for G is TLE on test 33? I've read the editorial and I think I am doing the same thing.https://codeforces.com/contest/1303/submission/71252477
•  » » 16 months ago, # ^ | ← Rev. 2 →   I had also struggled from test 33, and finally made it accepted.I don't know what exactly is test 33, but this test generator code would be helpful. It takes about 20 sec on N=150000 on your code. #include #include #include #include #include #include #include #include using namespace std; inline int randint(int a, int b) { return rand()%(b-a+1) + a; } int main() { int n; //cin >> n; n = 150000; cout << n << endl; srand(time(0)); vector perm(n+1, 0); for (int i=1; i<=n; i++) perm[i] = i; for (int it=n*10; it--; ) { int a = randint(1, n); int b = randint(1, n); swap(perm[a], perm[b]); } // sqrt tree, binary tree int sq = sqrt(n); // int sq = 2; // binary queue qu; qu.push(1); int p = 2; while (p <= n) { int a = qu.front(); qu.pop(); for (int i=0; i
•  » » » For me, it was a problem from cache. I had used so many random-accessing on arrays, and it became x8 faster when I fixed into a cache-friendly code.
•  » » » » 16 months ago, # ^ | ← Rev. 2 →   Thanks I may have a chance to fix it now. All my own generated samples ran in <2 seconds
 » can someone explain the solution for problem bi think it gives right answer for 5 1 5 but wrong for 10 1 10
•  » » doubt cleared!
 » Typo in D, case 2, "most" should be "least".
 » How to proof that D solution is optimal(finds min number of divides)?
 » Getting WA on Test 2: "Jury has found answer but participant has not (test case 34)" https://codeforces.com/contest/1303/submission/71554557Help pls.
 » Hi awoo! Why the scheduled codeforces educational round 83 was deleted?
 » 16 months ago, # | ← Rev. 2 →   I wanna ask that in problem D, how can we make sure that when sum of all the ai is >= n, the answer is yes. I don't really get it :(P/s: Now i got it, we don't have to put all the boxes into the bag. Right
 » 13 months ago, # | ← Rev. 2 →   For problem G, how can we get all the "first parts" and "second parts" efficiently?
 » I think there is a mistake in the tests in the B problem. MY solution as well as the tutorial both fail at the same place in the second test. Here is my code: (https://codeforces.com/problemset/submission/1303/79669445)(https://codeforces.com/problemset/submission/1303/79668860). Could someone tell me if there is a mistake
•  » » When you calculate minN = round(n / 2), n is truncated before round is called because it's an integer. So n always rounds down instead of rounding up. You can fix this by doing minN = (n+1) / 2 instead.Also, it's a little arrogant to assume that the tests are wrong if your code doesn't get AC. Especially if nearly 6,000 others have solved without any issues. So test your code a little on some cases first, before assuming the writers messed up.
•  » » » Oh thanks so much! I just thought that it was wrong because i followed the tutorial and did not think about the rounding error. I'll keep this in mind next time!
 » 80808954 Which case am i missing? (Problem B)
•  » » 13 months ago, # ^ | ← Rev. 3 →   you have to take the case when req is divisible by g than you dont need to add the last b days because you have already finished your work and dont and instead of subtracting g from req and than doing other operations better try to do it like thisans -> the number of days that we need to calculate int n; // length of road int nn = (n+1)/2; long long req= nn/g; if(nn%g == 0) { ans = req* g + (req- 1)*b; } else { ans = req* g + req* b + nn % g; } ans = max(ans,n); cout << ans << "\n"; 
•  » » » You mean a case like this: 1 12 3 8 It outputs : 14 
•  » » » » No just take the example of 16 3 8 as input Your code will give 16 as output But the correct answer will be 24
•  » » » » » Oh ok.. Got it. Thanks
 » 12 months ago, # | ← Rev. 3 →   In problem F, can someone please explain the intuition behind deleting cells. How is it similar to adding in reverse order? awoo
 » Can someone explain the problem C Perfect Keyboard?
•  » » U basically want to reorganize the given string so that all the neighbeiring characters come together(if possible).To solve this,assume each character as node of a graph and trace it ,if it do not have cycle its possible to organize it.Now,just run a dfs to get the reorganised string and the remaining characters(a-z) not present in the given string( eg: ababa).
•  » » » Do you have an AC code for refernce purposes?
•  » » » » I implemented it but was having some bugs ,but idea is same u can get the code from other submission.
 » in D how do we go about the optimality?
 » Alternate dp for E. Let $dp[len_{s}][len_{a}]$ be the max $len_{b}$ such that the prefix of $s$ of length $len_{s}$ contains non-intersecting subsequences of $a$ and $b$ of $len_{a}$ and $len_{b}$. Set it to -1 if the prefix does not contain a subsequence of length $len_{a}$. With this approach we don't have to calculate the array $nxt$ or make the observation in the editorial. Code 97250299
 » How is solution for D optimal?
 » The case n=1 in problem C is so silly. Constraints should start from n=2.
 » Is it possible to use Binary search on answer in B ?
 » Anybody please help!! In C- Perfect Keyboard. I am using deque and iterators to solve with same strategy as described in the editorial. I am getting WA!! Here is what the checkers log say- wrong answer Token "psqacdefghijklmnortuvwxyz" doesn't correspond to pattern "[a-z]{26}" (test case 966)Here is my submission-Your text to link here...