include<bits/stdc++.h>

5 years ago, # ^ |

← Rev. 2 →

-8

One should always point out that that's not the way one spells "Phalange" ;P

→ Reply

ReginaFelangi

5 years ago, # ^ |

Dude, you lost the whole joke.
The whole point of the name is for fake identity theft. Spelling it right is not actually "right" (P.s if you are not friends fan, drop it)

→ Reply

agayush

5 years ago, # ^ |

You can use : double n; cin>>n; cout<<fixed<<setprecision(0)<<ceil(n/2);

→ Reply

heyBossCanIGetAnACPlease

4 years ago, # ^ |

1
99999999
5e+07

→ Reply

contribution

5 years ago, # |

-44

I don't like the problems which can be done without using functions or loops

→ Reply

gatecse2018

5 years ago, # ^ |

← Rev. 2 →

+58

Try adding and subtracting using loops then and wrap these loops inside a function.

→ Reply

contribution

5 years ago, # ^ |

-34

I think you understood the comment wrongly.

→ Reply

omg

5 years ago, # |

-73

Problemsetting might not be the job for you.

→ Reply

The_Dark_Knight_Rises

5 years ago, # ^ |

+14

You are green and you think you know a dime about problemsetting.

→ Reply

southball

5 years ago, # |

It shows the message that "You are not allowed to view the requested page" when trying to access the jury solutions.

→ Reply

sgshivamgupta14

5 years ago, # |

-91

Pathetic and Confusing problem statement for problem B. Shitty Contest

→ Reply

secti0n

5 years ago, # ^ |

For problem B I would have to agree the problem statement was pretty confusing, but others were fine.

→ Reply

5 years ago, # ^ |

ya the problem statement confused me too.. after getting the solution it was very easy

→ Reply

4 years ago, # ^ |

Yeah I had a hard time understanding how a shape spread in 2 rows in a calendar with x columns can be compared with the same shape spread in 2 rows in a y columns calendar

→ Reply

dnshgyl21

5 years ago, # |

← Rev. 3 →

Link for the solution codes are not working right now ... please fix them.

→ Reply

duckduckMoose

5 years ago, # |

Can someone elaborate problem D's editorial?

→ Reply

5 years ago, # ^ |

+18

The idea is to fill out as many diagonals as possible, where each diagonal could wrap around the left/right sides. Filling in a diagonal increases each $$$R_i$$$ and $$$C_j$$$ by one. For a fully complete diagonal, $$$max(R) = min(R)$$$ and $$$max(C) = min(C)$$$, hence $$$f(A)=0$$$ if $$$k \% n = 0$$$. For an incomplete diagonal, you get $$$max(R) = min(R) + 1$$$ and $$$max(C) = min(C) + 1$$$, so $$$f(A)=2$$$.

→ Reply

Abhichauhan

5 years ago, # ^ |

what does this means "each diagonal could wrap around the left/right sides??

→ Reply

5 years ago, # ^ |

+21

Wrapping around the left/right sides means that if the diagonal would go off the right side, it would continue on the left side. I think a visual example is the best explanation:

Step 0:

Step 1:

Step 2:

Step 3:

→ Reply

Abhichauhan

5 years ago, # ^ |

Thankyou so much.. got it now

→ Reply

Prajwal_

4 years ago, # ^ |

← Rev. 2 →

i am still now getting...what if k%n is not zero

→ Reply

nihal_47

5 years ago, # ^ |

Thank you! This was very helpful than the editorial.

→ Reply

amrit2000

5 years ago, # ^ |

SO much better than the editorial. Thanks :)

→ Reply

ScotchOnTheRocks

4 years ago, # ^ |

← Rev. 2 →

How did you get that intuition tho? Wrap around diagonals seems a notch above any intuition I got. The closest I got to is filling out the closest diagonal stripe in the grid —

Only later did I realise that it was wrong.

→ Reply

4 years ago, # ^ |

The diagonals aren't really the important part here, it's just easy to construct that way. These matrices are equivalent:

1000    0010
0100    1000
0010    0001
0001    0100

Think of it like filling a chessboard with rooks. If you place the first rook at $$$(r, c)$$$ , then you are left with $$$n-1$$$ rows and $$$n-1$$$ columns open (this is equivalent to 1 row and 1 column with a sum of 1 and $$$n-1$$$ rows and $$$n-1$$$ columns with a sum of 0). Since we're trying to minimize the difference in max/min sums, it's better to place the second rook on an open row and column. This will leave you with $$$n-2$$$ rows and $$$n-2$$$ columns. Keep following this pattern, and you'll end up with $$$n$$$ rooks on the board.

→ Reply

3141somethingpi

4 years ago, # ^ |

So after filling in n rooks (as per the finding open column open row method), if k is greater than n, how would the remaining rooks go into the board? Or they could just be filled randomly?

→ Reply

4 years ago, # ^ |

You would treat the $$$n$$$ spaces that the first rooks take up as unavailable spots, then place the next $$$n$$$ rooks in the same way as the first $$$n$$$ where only 1 new rook per row and 1 new rook per column. The important part is that you don't place 2 on the same row/column when they're part of the same "rook group".

→ Reply

5 years ago, # ^ |

← Rev. 3 →

The following is not the way written in the editorial, but I have done it in my solution. Firstly, lets observe that if $$$k$$$ is divisible by $$$n$$$, the answer is $$$0$$$, and when it is not, the answer is $$$2$$$.

Proof

Now, how to fit the integers?

What I did was first insert $$$1$$$s in indexes $$$[1, 1], [2, 2], ...., [n, n]$$$.

In the next iteration, I inserted $$$1$$$s in indexes $$$[1, 2], [2, 3], ...., [n, 1]$$$

Like this, we have $$$(k / n)$$$ iterations.

I had put the extra $$$(k$$$ % $$$n)$$$ 1s simply as if it was the $$$(k / n) + 1$$$ th iteration, and only the first $$$(k$$$ % $$$n)$$$ 1s need to be inserted.

My submission for reference: 85648287

→ Reply

Prajwal_

4 years ago, # ^ |

I didn't understand how you put remaining k%n one's please explain

→ Reply

4 years ago, # ^ |

We put the ones in $$$k / n$$$ iterations, and thus the remaining ones will be $$$k$$$ % $$$n$$$. Note, over here the answer is $$$2$$$. The remaining ones should be lesser than $$$n$$$, because if not, then we could have added another iteration to use the $$$n$$$ ones. As the remaining ones are lesser than $$$n$$$, it means that they are lesser than the number of rows and columns. Thus, we iterate for the first $$$k$$$ % $$$n$$$ rows and columns, in the next iteration, and initialise that point $$$a[row][col]$$$ to $$$1$$$.

→ Reply

Prajwal_

4 years ago, # ^ |

There is no need to do seperately for k/nand k%n we can fill the matrix in only one loop going from [1,k]

→ Reply

4 years ago, # ^ |

Yes, but don't you think the implementation will go for a toss? I am not telling you that my solution is the best, just telling you from my perspective, I code a solution which is the easiest to implement.

→ Reply

5 years ago, # ^ |

That formular is wrong, isn't it? $$$f(A)=2(1^2+1^2)$$$

I implemented AC that as

    int ans=0;
    if(k%n!=0 && n>1)
        ans=2;

→ Reply

123priyankarai

5 years ago, # ^ |

It says f(A)=2. Which is the result of (1^2+1^2). Correct me if I got it wrong.

→ Reply

5 years ago, # ^ |

It says f(A)=4. Which is result of 2*(1^2+1^2)

Or what is the 2 good for?

→ Reply

5 years ago, # ^ |

No, he means to say that f(A)=2, which comes from 1^2+1^2, that's why the brackets. The brackets are not indicating multiplication.

→ Reply

FinalBoss_

5 years ago, # |

+54

I think B and C should have been swapped. Thanks for fast editorial though.

→ Reply

tanmay2625

5 years ago, # |

+79

I think D was little bit similar to 1360G - A/B Matrix and thank you for a nice contest.

→ Reply

Badeline

5 years ago, # ^ |

I actually copied some code from that problem when the answer is 0

→ Reply

HelloWorld

5 years ago, # ^ |

I actually copied almost the whole code. See this submission 81245465 for problem 1360G - A/B Matrix and my submission 85681253 for problem D.

→ Reply

me17b016

5 years ago, # ^ |

That's why i was able to do this problem otherwise it would be hard to come up with such algorithm during contest.

→ Reply

towrist

5 years ago, # ^ |

← Rev. 2 →

Little bit? Dude for k%n!=0 you just add one for loop and you get AC.

→ Reply

yash_daga

5 years ago, # ^ |

+12

Wow man you have such a nice memory XD I think this is the third time I have seen you remembering similar old problems. No offence tho :)

→ Reply

Pavel61

5 years ago, # |

pattern in questions!

→ Reply

Fighter

5 years ago, # |

I know it's too much to ask from anyone , but i will be highly obliged and grateful to the one who will look at my submissions of D to see why it fails to pass test cases . My submission link : https://codeforces.net/contest/1371/submission/85694260

→ Reply

TheScrasse

5 years ago, # ^ |

+16

testcase
4 7

your answer

correct answer

→ Reply

Fighter

5 years ago, # ^ |

+19

Thanks a lot man! i understand now . Have a great luck and wonderful day/night ahead ! :)

→ Reply

rahulmysuru7

5 years ago, # ^ |

I up-vote this for your humbleness

→ Reply

bira37

5 years ago, # |

← Rev. 2 →

+30

Could E be solved, under the same constraints, if P was not prime ? I'm curious about it.

→ Reply

TheScrasse

5 years ago, # ^ |

← Rev. 7 →

-38

if $$$p$$$ is not prime, you should replace it with the minimum integer $$$k$$$ such that $$$k!$$$ is a multiple of $$$p$$$
upd: nvm, it's wrong

→ Reply

TheScrasse

5 years ago, # ^ |

← Rev. 3 →

-17

reasoning
the answers are such $$$x$$$ that some $$$P[i]$$$ can be chosen in $$$\geq k$$$ ways
you can prove that, if this condition is true, for $$$1 \leq y \leq k$$$, some $$$P[i]$$$ can be chosen in $$$y$$$ ways, so $$$f(x)$$$ is a multiple of $$$k!$$$
upd: nvm, it's wrong

→ Reply

bira37

5 years ago, # ^ |

← Rev. 2 →

→ Reply

PRUGREMMER

5 years ago, # |

+77

Div2 rounds are now tending towards Div3! who else feels same?

→ Reply

TleTerror

5 years ago, # ^ |

May be it is because it was physics0523's first round that's why..!

→ Reply

saurabhsingh1998

5 years ago, # ^ |

No Actually u r an expert!! thats why

→ Reply

PRUGREMMER

4 years ago, # ^ |

← Rev. 2 →

It's not about individual rating, just look at submissions on problems till E!, it is rare.

→ Reply

ashanksingh1997

5 years ago, # |

Question B should have been better explained

→ Reply

Abhijeet8901

5 years ago, # |

+23

To be honest, felt like a div3. except the last problem. For others, no. of submissions is almost same as on a div3.

→ Reply

HelloWorld

5 years ago, # ^ |

You can't just compare no. of submissions in a div3 contest and a div2 contest. In a div2 contest there are more high rated coders than in div3 contest. The expected rank of a 1500 rated coder in Codeforces Round 654 (Div. 2) was 4299, and his rank would be 2894 in the contest Codeforces Round 650 (Div. 3)

Problem is easy, coders are not very skilled, no of submissions are X. Problems are harder, coders are also better skilled, the problem might get a similar number of submissions.

I used this website to calculate expected ranks.

To compare the div3 rounds and div2 rounds, compare their problem difficulties, not no. of submissions.

→ Reply

rain_cf

5 years ago, # |

In C's editorial shouldn't it be m<=min(a,b) rather than m<min(a,b)?

→ Reply

karan_4336

5 years ago, # ^ |

← Rev. 2 →

yeah, i am thinking about that. it should be m <= min(a, b)

→ Reply

alon276

5 years ago, # |

+28

Why can't someone who knows English go over the problems and correct bad translations? It's so hard to read these problems with so many out of place commas, and completely messy sentences.

→ Reply

Priyansh31dec

5 years ago, # |

← Rev. 2 →

Editorial Problem E1 You can find the value of f(x) by the following simulation:

First, let v={the number of enemies they have strictly less than x candies}, ans=1. Then do the following steps for i=x,x+1,...,x+(n−1). v=v + {the number of enemies they have exactly i candies} ans=ans∗v v=v−1

Is it just me or was this part not really that intuitive?

→ Reply

4 years ago, # ^ |

← Rev. 2 →

Yeah,
I still haven't been able to fully understand the Editorial of E1.
I understood the statement by seeing some solutions but the reason it works is still a mystery to me

Found a good explanation https://codeforces.net/blog/entry/79624#comment-654544

→ Reply

himanshu_attri

4 years ago, # ^ |

for a give n x we are calculating all permutations v = all value less then x we can always put all v values in first place x could always defeat these values and will become x+1 now in second position e can select v -1 as out of v we have already put 1 value at first place but, we can also put those values which are equal to x+1 thats why for x+1 we added v += the num of enemy that has exactly x+1 candies. so we should multiply our ans with this num thatis ans = v*(v-1 + k) where k is num of enemies that have exactly same candies as x+1 and v is numbver of candies less then x, and so on.

→ Reply

aneee

5 years ago, # |

I did the same thing that's mentioned in the editorial for E1. But I am getting WA for test 5. Submission.

→ Reply

Pie-Lie-Die

5 years ago, # |

Can someone explain C with examples? And what tags do these types of problems come under?

→ Reply

souraavv

5 years ago, # ^ |

Notice Second Type of guest always consume the one which is small in quantity. So if v > c then only chocolate will be consume by Type 2 guest or if c > v then only vanilla is consume and after consuming either the inequality v > c or c > v which one is there will still hold true. So if min(v, c) >= m then we can always allocate 1 cookie to each type 2 guests.

→ Reply

Pie-Lie-Die

5 years ago, # ^ |

Thank you, that helps!

→ Reply

BNY_Op

4 years ago, # ^ |

Oh man...Thanks a lot..! After going through so many comments I was finally able get something related to problem and helpful as well. Thanks again

→ Reply

5 years ago, # |

+10

Finally understood B, thanks!

→ Reply

5 years ago, # |

+13

At least provide some explanation for problem D. We can see the other's submission if we want to see code.

→ Reply

souraavv

4 years ago, # ^ |

← Rev. 3 →

+28

You have to note the pattern. I will explain this by an example.

For k <= n it is simple : We have to fill only along the diagonal.

For k > n :

One thing that is sure : Putting 1 along diagonal is always optimal choice to minimize the final ans.

Step 1. Fill the diagonal

Step2 .

Step 3. See the pattern

Step 4. We will repeat what we have done in step 2, and step 3 till we still some entry to fill.

Repeating pattern :

Repeating pattern:

And finally :

Trick that I use to print is r-c (or c-r) is constant for a particular diagonal.

→ Reply

squirrel123

5 years ago, # |

+15

Any reason for making problem C around the same difficulty as a typical Div2 A?

→ Reply

HipHop

5 years ago, # |

In D there should be either (1^2 + 1^2) or 2 * 1^2, not 2 * (1^2 + 1^2).

→ Reply

naeem4265

5 years ago, # |

problem C was fantastic for me.I spended too much time for it.when I finded it's solution I just surprised.

→ Reply

123priyankarai

5 years ago, # |

← Rev. 2 →

First 4 problems felt more like div3 contest only maths and implementation kind. Or may be I am improving.

→ Reply

5 years ago, # ^ |

← Rev. 3 →

You are improving...

Spoiler

→ Reply

4 years ago, # ^ |

How can you judge a problem without taking part in actual contest.

→ Reply

4 years ago, # ^ |

Judging a problem without taking part in contest

→ Reply

4 years ago, # ^ |

solving a problem after the contest is much easier than during the contest. Have a look at problem C, it seemed difficult to most of the people appearing for the contest but after upsolving they found that very easy.

→ Reply

4 years ago, # ^ |

+10

@aryan12 means solving after the contest but with the below constraint

→ Reply

ayush_nishad

5 years ago, # |

+14

Felt like giving an Aptitude test!

→ Reply

4 years ago, # ^ |

Yes, At least for the first 4 problems Never thought pattern recognition and printing will help me solve my firsts Div2 D problem :D

→ Reply

5 years ago, # |

https://codeforces.net/blog/entry/79541?#comment-654173 This single comment has a better explanation for A, B, C than this editorial

→ Reply

khantil007

5 years ago, # |

Can Any One Explain me Solution of problem:E1? Because I am still not able to understand it through Editorial. Please Explain if You know .

→ Reply

khantil007

5 years ago, # ^ |

How we can find the value of f(x)???

→ Reply

https://codeforces.net/contest/1371/submission/85673827

5 years ago, # ^ |

← Rev. 4 →

+30

My reasoning was: Sort $$$a$$$. Fix $$$x$$$ $$$(1 \leq x \leq 2000)$$$. Lets see how many elements we can pick to be $$$P_0$$$ in a valid permutation. Find the last $$$i$$$ such that $$$a_i \leq x$$$. Then we can let $$$P_0 = a_0 \text{ or } a_1 \text{ or } ... \text{ or } a_i$$$. Thus there are $$$i + 1$$$ ways to pick $$$P_0$$$. To pick $$$P_1$$$ we do the same thing, i.e., find the last $$$i$$$ such that $$$a_i \leq x + 1$$$. Then we can let $$$P_1 = a_0 \text{ or } a_1 \text{ or } ... \text{ or } a_i$$$ except that we used one of these elements in choosing $$$P_0$$$. Thus there are $$$i + 1 - 1$$$ ways to pick $$$P_1$$$. Repeat this process for each $$$j < n$$$. If you get to some $$$j$$$ such that the number of ways to pick $$$P_j$$$ is divisible by $$$p$$$, then $$$x$$$ is not valid. Otherwise it is valid. The time complexity is $$$O(\max(a)n\log(n))$$$. This definitely won't pass E2 though.

Edit: My contest code (disregard the freq array, it is unused): 85694840

→ Reply

saatvik10

5 years ago, # ^ |

I have done the same thing as you. Can you tell me what is wrong with my code?

→ Reply

5 years ago, # ^ |

← Rev. 2 →

probably overflow? You never modded your ans variable, but it can be as big as $$$n!$$$

→ Reply

saatvik10

5 years ago, # ^ |

Lol you're right...was so close to solving 5 for the first time :(

→ Reply

kdjonty31

5 years ago, # ^ |

Thanks for sharing your approach mate...idk why but was struggling to understand the editorial...I found your approach more insightful. Thanks. Appreciate it! :)

→ Reply

hossainzarif

5 years ago, # ^ |

You can find the minimum x such that f(x) % p != 0, then you can do binary search on your same approach for maximum x. And print from minimum to maximum.

→ Reply

4 years ago, # ^ |

Yep that's the editorial's first solution for E2. The editorials $$$C_i(x)$$$ function is the number of ways to choose $$$P_i$$$ if the starting amount is $$$x$$$. I implemented it and it works in $$$O(\log(1e9)n\log(n))$$$ (~100ms).

→ Reply

hossainzarif

4 years ago, # ^ |

During solving this question on practice. I just guessed(weak guess) that maybe the answer would be a continuous segment. So, did a binary search. Still, not understanding why it worked?

→ Reply

4 years ago, # ^ |

Its because of these two simple observations that the editorial states:

$$$1.$$$ For each $$$x$$$, for each $$$i$$$ $$$(0 \leq i < n - 1)$$$, $$$C_i(x) \geq C_{i + 1}(x) - 1$$$, and $$$C_n(x) \leq 1$$$.

$$$2.$$$ For each $$$i$$$ ($$$0 \leq i < n$$$), for each $$$x$$$, $$$C_i(x) \leq C_i(x + 1)$$$.

From $$$(1)$$$, its clear that if $$$C_i(x) \geq p$$$ at some $$$i$$$, then there must be some $$$j \geq i$$$ such that $$$C_j(x) = p$$$.

Since $$$p$$$ is prime, $$$p \mid f(x)$$$ if and only if $$$p \mid C_i(x)$$$ for some $$$i$$$ $$$(0 \leq i < n)$$$. I think this is called Euclid's Lemma.

Thus from $$$(2)$$$ it is clear that $$$f(x) > 0$$$ and $$$p \mid f(x) \implies f(x + 1) > 0$$$ and $$$p \mid f(x + 1)$$$. Therefore there will be a smallest value $$$mn$$$ such that $$$f(mn) > 0$$$, and there will be a smallest value $$$mx \geq mn$$$ such that $$$f(mx) > 0$$$ and $$$p \mid mx$$$ and each $$$x \geq mx$$$ is not good. Thus binary search works. The answer is just the range $$$[mn, mx)$$$.

→ Reply

vaibnak7

4 years ago, # ^ |

I can`t understand the first condition.

suppose x = 3 and i be 2, then power at i=2 will be 5, so C(i,x) will give how many values are there less than or equal to 5, suppose this value is 10, out of this 2 have already been used at index 0 and 1 so value should be 8.

Now at i+1 we will have value 6 and then we will ask how many values are there less than or equal to 6, this will incude all the values less than or equal to 5 and the values equal to 6, let it be 15 (10+5(freq. of 6)), as now 3 value have been used value value here will be 12

so we have C(i,x) = 8 and C(i+1,x) = 12, then how is the above condition satisfied ?

→ Reply

4 years ago, # ^ |

Thanks a lot, for some reason I wasn't able to understand this via the official editorial. You saved me from a headache :P

→ Reply

Ram_Newton

4 years ago, # ^ |

Ha, thank you! Good explanation.

I'm trying to understand the editorial implementation now. But I can't figure out what the bk array means. Can someone give me an intuition?

→ Reply

devil_17

5 years ago, # |

← Rev. 3 →

Can someone please help me with this submission for E1. It gets WA on test 9.

EDIT : I found the mistake. I was using set data structure earlier to get the count of elements less than or equal to $$$x+i$$$ which is wrong as it only stores unique elements and thus might give me wrong count values in case of duplicates.

→ Reply

yolo86

5 years ago, # |

← Rev. 2 →

→ Reply

ruchir.mehta

5 years ago, # |

how can one arrive to the approach used in problem D. I mean is it really too intuitive? If yes, then can anyone share similar problems for practice.

→ Reply

5 years ago, # ^ |

← Rev. 2 →

My idea was to flatten the board into a line of length $$$n^2$$$ and spread the ones "evenly" across the grid. That is, put a one at positions $$$0 \bmod n^2, (n + 1) \bmod n^2, 2(n + 1) \bmod n^2, ..., (k - 1)(n + 1) \bmod n^2$$$. Then you can easily convert the linear position to an $$$(r, c)$$$ coordinate by the formula $$$coord(x) = (x / n, x \bmod n)$$$. I didn't prove it rigorously, but it got AC: 85675022

I believe a similar technique can be used to solve this problem: https://codeforces.net/problemset/problem/1360/G

→ Reply

BigO1

5 years ago, # ^ |

Thank you gentleman for this brilliant approach, now I am wondering how to prove and why this works. If possible can you please elaborate? Thanks a lot.

→ Reply

5 years ago, # ^ |

One. No square will be hit twice using this method, and all $$$n^2$$$ squares can be hit if needed (i.e. when $$$k = n^2$$$).

Proof: Since $$$\gcd(n + 1, n^2) = 1$$$ (easy to show, I hope you can figure it out), the method will hit exactly $$$\frac{n^2}{gcd(n + 1, n^2)} = n^2$$$ before returning to its initial position (0), so the claim holds. This formula follows from a theorem of group theory which states that for any positive integer $$$k$$$, for each $$$x \in \mathbb{Z}_k$$$, the smallest integer $$$d$$$ such that $$$d \cdot x \equiv 0 \pmod{k}$$$ is $$$d = \frac{n}{\gcd(x, n)}$$$. An equivalent way to say it is $$$d$$$ is equal to the number of distinct elements in $$$\mathbb{Z}_k$$$ you get by repeatedly adding $$$x$$$ to itself modulo $$$k$$$. This theorem appears surprisingly frequently on Code Forces for some reason (probably due to its relation to cycles and permutations).

Two. I'm still not sure how to rigorously prove that this method minimizes $$$f(A)$$$ ... Im guessing it has to do with the fact that also $$$\gcd(n + 1, n) = 1$$$.

→ Reply

bhavsar

5 years ago, # |

Shouldn't it be $$$m \le min(a,b)$$$ for problem $$$C$$$?

→ Reply

Wild4Stone

5 years ago, # |

In E2 for solution 1 :

"if C_i(x) exceeds p, then there will be some C_j(x) = p as C_i(x) decrements by at most 1".

Why is this necessary that some C_j(x) will be equal to p?

what if C_i(x) does not decrease but only increase? Can someone explain?

→ Reply

robostac

5 years ago, # ^ |

It has to decrease towards the end. If there was no constraints on choices at each turn and you had 5 options you end up with 5*4*3*2*1 choices. The constraints can never increase the number of choices for any turn but can reduce them so any 'good' x will ensure that you never have p or more choices on any turn.

→ Reply

vickyandvs

4 years ago, # ^ |

Cn(x) has to be one bcoz total n choices. So,if some c_i > p and each c_j can atmost decrease by one then p+a....p.....1

→ Reply

shirakia

5 years ago, # |

← Rev. 2 →

+26

The problems were wonderful. I loved them all.

However Ruby 2.7.1 (which was upgraded from 2.0.0 recently) is too slow on Codeforces. My O(1) (I mean, O(t)) solutions don't pass! https://codeforces.net/contest/1371/submission/85638755 Same code with same input passes in 150ms on Macbook Air and 75ms on AtCoder!

It seems that any Ruby2.7.1 submission takes 900ms at least. (Even if 1 line do-nothing code) So I guess Ruby2.7.1 on Codeforces has somewhat wrong settings.

And I found that previously passed code makes TLE now with Ruby2.7.1.

https://codeforces.net/contest/1266/submission/73652966 passed in 171ms previously(Ruby2.0.0)
https://codeforces.net/contest/1266/submission/85706731 TLE now (Ruby2.7.1)

Please let me know if any better place to report bugs to Codeforces. Thanks!

→ Reply

rohan_ardulous

5 years ago, # |

Shouldn't it be m <= min(a,b) in place of m < min(a,b) as written in the editorial

→ Reply

5 years ago, # ^ |

Yes

→ Reply

dgupta0812

5 years ago, # |

For all those who want to understand how the solution for D is working, have a look at the comments of my code.

→ Reply

battousaiKenshin

5 years ago, # |

I can't understand the solution of C except the base condition n + m <= a + b. And how to think solution of this kind of problem, I tried different types of conditions but couldn't unite them.

→ Reply

bayashiko

5 years ago, # ^ |

Second type guests always choose fewer remaining type of cookies out of 2 types of cookies, so after they choose, the fewer number of remaining cookie is always reduced by 1. First type guests will not get angry if at least 1 cookie left, so the best thing to do is to put them on the back burner.

→ Reply

rahulmysuru7

5 years ago, # |

The contest was mostly logic based for question A to E1 and not much related to algorithms and data structures.

→ Reply

r.zay699

5 years ago, # |

In problem E2 (Asterism), after we have calculated b_i, how are we able to find all the "good" x's in linear time?

→ Reply

5 years ago, # |

Your algorithm for comprehending D is funny, Mr physics0523 ;P

→ Reply

nexus_18

5 years ago, # |

For the whole round, I thought that in problem B, we have to count only one pattern for the row that completely divides the total number of days.The problem statement could have been written more clear.

→ Reply

dush1729

5 years ago, # |

still can't solve C :( if we call type 1 guest before type 2 guest then it won't decrease min(a,b) right? but in editorial

if there is a type 1 guest before a type 2 guest, the type 1 guest have a possibility to decrease the value of min(a,b) unnecessarily

→ Reply

Silver3

5 years ago, # ^ |

If a = b and we call two guests of type 1, they will decrease the minimum of a and b.

→ Reply

nexus_18

5 years ago, # ^ |

Yes,this statement is conflicting. You can understand it by realising that at any moment,type 1 guest tries to eat the cookie which is more in number and type 2 tries to eat cookie which is less in number.If number of cookies are same the type 1 eats chocolate cookie and type 2 eats vanilla.

→ Reply

dush1729

5 years ago, # ^ |

got it, tysm! we invite type 2 guests first because they take minimum of vanilla and choclate and in the end we invite type 1 guests and check if we have required remaining dishes.

→ Reply

redGreenCode

4 years ago, # ^ |

This statement is genius! Thanks

→ Reply

Silver3

5 years ago, # |

I'm not sure whether this was mentioned before, but there is a DP solution to E1 which generalizes to E2. I am going to present a handwavy explanation of it:

85709850 Here, dp[i] denotes the number of valid permutations of the i smallest elements. From there, we want to insert a new element in these permutations, which does not make any of the existing duels worse, as our candies are monotonically increasing, so we just find the number of positions, where we can put the new element. If we had x candies at start and our new duel is the k-th one, then we will have x + k — 1 candies, so this must be at least the new amount of candies.

Now moving on to E2:85708674

Knowing the recursion from the dp, we can now see that if we have too few candies to win a duel, f(x) is 0, and if we have at least as many candies at start than the p-th smallest enemy, we will also have trouble, because that multiplies our dp by p. So we can find the min and max candies easily, and now we just want to quickly decide whether any of our other products are zero. The parts in the dp product before the p-th smallest element cannot be divisible by p, as they are positive but smaller than p, so we just take the elements from the p-th to the n-th and see whether they give a zero residue with p, granted that we have less candies than them when starting the fights (this follows from the enemies being sorted and us having at less candies than the p-th smallest enemy). This can also be done linearly.

→ Reply

5 years ago, # |

E1, "Yuzu should have at least $$$m-n+1$$$ candies initially to win all duels."

Isn't that wrong? Having that much candis garanties that not all duels get lost, but to garantie to win all duels Yuzu must have $$$m$$$ candies. Since all a[i] can be equal $$$m$$$.

I mean, a[] is not a permutation or somehow else pairwise distinct.

→ Reply

_pastor_

5 years ago, # ^ |

-25

bro see my solution for that problem it is easy to understand.

→ Reply

5 years ago, # ^ |

I did not solve it because I was not able to come up with formular to calculate f(x) for any fixed x. And, honestly, I am still not sure about that after reading the editorial.

→ Reply

_pastor_

5 years ago, # ^ |

:D hmm editorial is not good this time

→ Reply

sahil0307

4 years ago, # ^ |

← Rev. 2 →

Agreed, they did not explain the solutions well, they just wrote the pseudo-code for the problem.

Btw, I solved E1 in contest with basic principle of counting, for a fixed x, I counted no. of ways to fill 1st place * no. of ways to fill 2nd place and so on up to n places. My submission 85691626

→ Reply

rath772k

5 years ago, # ^ |

← Rev. 2 →

If you have $$$m$$$ candies, you can definitely win all duels. What the author meant is this:

Let $$$x$$$ be the smallest number of candies you need to guarantee a win, then $$$x$$$ $$$\geq$$$ $$$m - n +1$$$. Had it been less than that, you'll have maximum $$$m-n+n-1$$$ candies against an enemy with $$$m$$$ candies.

→ Reply

5 years ago, # ^ |

Ok, what he meant is "Yuzu should have at least $$$m-n+1$$$ candies initially to not loose all duels."

→ Reply

sahil0307

4 years ago, # ^ |

Having m — n + 1 candies obviously does not guarantee winning all duels. eg — 5 5 5. Maybe, he meant that if we have candies less than this, then we are definitely going to lose.

→ Reply

Dsxv

5 years ago, # ^ |

Hi! By at least it means that any x below this and we're certain that f(x) = 0, so we don't need to check below that. It's not actually necessary to do this if formula for finding f(x) handles all cases tho.

→ Reply

indomitable_will

5 years ago, # |

In tutorial of C it should be m <= min(a,b) instead of m < min(a,b) for answer to be "Yes".

→ Reply

ash_vs_ash

5 years ago, # |

The test cases for E1 were such that I got a TLE when I solved in O(n*n*logn). Till today I thought log(n) does not make a difference. Got to learn something from the question.

→ Reply

ColorlessCoder

5 years ago, # ^ |

My O(n*n*logn) solution passed well within the time limit 85702061

→ Reply

ash_vs_ash

5 years ago, # ^ |

I think I have done the same thing. If possible could you give it a look? https://codeforces.net/contest/1371/submission/85691445

→ Reply

sahil0307

4 years ago, # ^ |

← Rev. 2 →

ash_vs_ash you are passing vector to function search() by value, therefore it is taking extra O(n) in every run and maybe that is why you're getting TLE. Try passing vector by reference to your function.

→ Reply

ash_vs_ash

4 years ago, # ^ |

It worked like a charm. Thanx man.

→ Reply

https://codeforces.net/contest/1371/submission/85673827

5 years ago, # |

← Rev. 4 →

Why am I getting WA on test case 6 of E1: 85714673

UPD: Fixed it

→ Reply

saatvik10

5 years ago, # |

Can anyone please tell me what is wrong with my code for E1?

→ Reply

chaudhary_19

5 years ago, # |

#include<bits/stdc++.h>
 
int main() {
	using namespace std;
	ios_base::sync_with_stdio(false), cin.tie(nullptr);
	int T; cin >> T;
	while (T--) {
		int N, K; cin >> N >> K;
		cout << (K % N == 0 ? 0 : 2) << '\n';
		for (int i = 0; i < N; i++) {
			for (int j = 0; j < N; j++) {
				int d = (j - i + N) % N;
				if (d * N + i < K) {
					cout << '1';
				} else {
					cout << '0';
				}
			}
			cout << '\n';
		}
	}
 
	return 0;
}

This is ecnerwala's solution for Problem D, can somebody help me in understanding how this solution works and what was the intuition behind that :)

→ Reply

ecnerwala

5 years ago, # ^ |

+34

It's the same diagonal construction as the editorial, but computed in a different order. For each cell, we compute which diagonal it's on and use that to decide whether it should be set or not.

→ Reply

SoulAdor

5 years ago, # |

+22

This editorial is so hard it needs another editorial. Seriously, it was easier for me to solve problems than to understand this, not even saying anything about jury code quality. Some standards for writing editorials are needed.

→ Reply

WitchOfTruth

5 years ago, # ^ |

← Rev. 3 →

I think it should be forbidden to have editorials in the form of pseudocode or formal algorithm description. I read editorials to understand how to arrive to the solution. If I want a formal description, I can just open anyone's submission

→ Reply

SoulAdor

5 years ago, # ^ |

Pseudocode might be fine, if it is easier to read than actual code. The actual solution must be present anyways. What you might need is a system of "tips", that describe a way of thinking that could lead you to the solution.

Unfortunately, most of the time this is not the case and we see "just do this" or "we can prove that this works" types of arguments. Even just checking that a thing works might be of little value for me, because during contests you are not checking solutions, you need to come up with one.

→ Reply

mohitsamant

5 years ago, # |

binary search solution of E2

Spoiler

→ Reply

_aadiupadhyay_

5 years ago, # |

Hello, can anyone tell me how long we have to wait to see the change in our rating, system checking is done hours ago

→ Reply

bhatKa

5 years ago, # ^ |

they have been updated

→ Reply

_aadiupadhyay_

5 years ago, # ^ |

just now I think thanks man

→ Reply

SecondThread

5 years ago, # |

+11

If you prefer video explainations, I go over everything (except a rather handwavey description of E2) at the end of my screencast of the round.

→ Reply

aadarsh026

5 years ago, # |

← Rev. 3 →

Can anyone please explain how this solution works for 1371E2 - Asterism (Hard Version). Thanks in advance.

        int n, p, a[n];
        cin >> n >> p;
        for (int i = 0; i < n; i++)
            cin >> a[i];
        sort(a, a + n);
        int l = 0, r = 1e9;

        for (int i = 0; i < n; i++)
            l = max(l, a[i] - i);
        for (int i = p - 1; i < n; i++)
            r = min(r, a[i] - i + p - 1); // need help with this part

        cout << max(0, r - l) << endl;
        for (int i = l; i < r; i++)
            cout << i << " ";
        cout << endl;

Taken from

→ Reply

golions

5 years ago, # ^ |

← Rev. 2 →

I have a proof that the solution must be a contiguous sequence of numbers here: https://codeforces.net/blog/entry/79624?#comment-654797

He found the min and max numbers of the subarray, then printed every number in between.

It seems that while I used binary search to find the maximum bound of the subarray, he found a cleverer way just using a for loop.

→ Reply

golions

5 years ago, # |

← Rev. 2 →

+22

I have an interesting solution for E2 using binary search.

It relies on the fact that the answer will always be a contiguous sequence of numbers. I have a proof of that fact below.

85713453 (find minimum value in easy O(N), binary search to find the max value that works, print every number in between).

Proof:

Claim: f(x) can be written as some factorial, lets call it X!, multiplied by some mumbers <=X.
You can see that this is true when you calculate $$$p_i$$$, the number of possible numbers to go in the $$$i$$$th spot, after you've determined the first $$$p_0$$$ to $$$p_{i-1}$$$. f(x) equals $$$p_0 \cdot p_1 \cdot \ldots \cdot p_{n-2} \cdot p_{n-1}$$$. $$$p_i$$$ is calclated by (the number of numbers in the array <= x+i) — i. $$$i$$$ increases by $$$1$$$ each iteration, and because that is the only way for $$$p_i$$$ to decrease, is impossible to decrease by more than $$$1$$$. There is $$$1$$$ possible number that can go in the last position because only $$$1$$$ number is left, so the $$$p_{n-1}$$$ must equal $$$1$$$.
Thus, say you found the position with the maximum number of possibilities had $$$X$$$ possibilities, f(x) = X! * {some numbers <= X} because all numbers between $$$1$$$ and $$$X$$$ inclusive must be represented in $$$p_0$$$ to $$$p_{n-1}$$$.
This is important because you can say that $$$f(x)$$$ is divisible by all primes $$$<= X$$$ and not divisible by any primes $$$> X$$$, where $$$X$$$ is the number of possibilities of the index with the most possibilities.
You can also show that the $$$X_a \le X_b$$$ for all $$$a < b$$$. This is because having a greater value of x in f(x) will never result in less possibilities for any position.
Thus, for all $$$f(a)$$$ and $$$f(b)$$$, where $$$a < b$$$, if $$$f(a)$$$ is divisible by $$$p$$$, then $$$f(b)$$$ must also be divisible by $$$p$$$. $$$X_a$$$ is $$$> p$$$, and since $$$X_b \ge X_a$$$, $$$X_b > p$$$ as well.
Starting from the minimum working $$$x$$$ (before of which $$$f(x)$$$ is all $$$0$$$), no value of $$$x$$$ after the first non-working $$$x$$$ can work, so the answer must be in the form of a continguous sequence of numbers.

→ Reply

limabeans

5 years ago, # ^ |

I also came up with a similar idea when upsolving, code with brief comments: 85718984

The upperbound is the largest number that never has $$$\geq p$$$ options, and every greater number will surely hit $$$p$$$ options "on the way back down" as they run out of options

→ Reply

rohan_ardulous

4 years ago, # ^ |

Thanks. That really helped.

→ Reply

5 years ago, # |

+32

My solution and explanation for 1371E2 - Asterism (Hard Version)

The number of permutations for a given $$$x$$$ can be written in:

$$$f(x)=\Pi_{i=0}^{n-1} ((\sum_{j=0}^{n-1}a_j\leq x+i)-i)$$$

Since $p$ is a prime number, in order that $$$f(x)$$$ cannot be divided by $$$p$$$, we must ensure that $$$(\sum_{j=0}^{n-1}a_j\leq x+i)-i<p$$$ holds for every $$$i$$$. And in order that there is a possible permutation, we must ensure that $$$(\sum_{j=0}^{n-1}a_j\leq x+i)-i\geq1$$$ holds for every $$$i$$$.

Proof: Let's denote $$$(\sum_{j=0}^{n-1}a_j\leq x+i)-i$$$ as $$$g(i)$$$. We can find that, since $$$\sum_{j=0}^{n-1}a_j\leq x+i$$$ never decreases, so if $$$g(i)$$$ decreases, it can decrease at most $$$1$$$. Another observation is that $$$g(n-1)=1$$$. So if there is $$$i_1$$$ such that $$$g(i_1)>p$$$, there must be $$$i_2>i_1$$$ which satisfies $$$g(i_2)=p$$$, which makes $$$f(x)$$$ divided by $$$p$$$.

So the problem is now: how many $$$x$$$ are there that satisfy

$$$\forall i, 1\leq g(i)<p$$$

Since $g(i)$ increases with x, the answer must be a continuous segment $$$[lo..hi]$$$.

Now we sort the array $$$a$$$ in the ascending order.

For $$$lo$$$, we have

$$$\forall i, lo+i\geq a_i\implies lo\leq\max(a_i-i)$$$

For $$$hi$$$, we have

$$$\forall i, hi+i<a_{i+p-1}\implies hi<\min(a_{i+p-1}-i)$$$

So the final answer is the segment

$$$[\max(a_i-i)..\min(a_{i+p-1}-i)-1]$$$

Code: 85684812

→ Reply

5 years ago, # ^ |

I can't understand the last two steps of your solution.Why Why does this formula work ? ∀i,hi+i<ai+p−1⟹hi<min(ai+p−1−i)

→ Reply

5 years ago, # ^ |

$$$\forall i, hi+i<a_{i+p-1}\implies\forall i,hi<a_{i+p-1}-i\implies hi<\min(a_{i+p-1}-i)$$$

→ Reply

5 years ago, # ^ |

I get it! Let cnt[x] be the amount of number which is equal or less than x,then we have ∀i,0<=i<n => cnt[x+i]-i<p => cnt[x+i]<p+i => x+i<a[p+i-1] => x<a[p+i-1]-i => x<min(a[p+i-1]-i). Thanks for your solution!

→ Reply

5 years ago, # ^ |

← Rev. 2 →

anodiebird How did you come up with, x+i<a[p+i-1], steps before this one are clear. Can you please elucidate this one?

→ Reply

5 years ago, # ^ |

⑴ x+i<a[p+i-1] => x<a[p+i-1]-i ⑵ a+b<c => a<c-b If you look at these two equations，you will see both of them are the same form,in the first equation I just change the position of i,in the second equation I just change the position of b.

→ Reply

5 years ago, # ^ |

anodiebird My doubt is that how did you transition from cnt[x+i]<p+i to x+i<a[p+i-1]

→ Reply

5 years ago, # ^ |

Because we have sorted the elements of a[].If index i begins from 0,then we can see cnt[a[i-1]]=i,and you can deduce that p+i=cnt[a[p+i-1]].So let's substitute that into our expression:cnt[x+i]<cnt[a[p+i-1]] => x+i<a[p+i-1] => x<a[p+i-1]-i

→ Reply

Yera_

5 years ago, # |

If author's approach for problem F is unclear, then you may check my solution. 85721976 Imho problem F is the only problem that I've found interesting in this contest.

→ Reply

KoroVishu_ACT2

5 years ago, # |

+14

It is probably because it's his/her first contest but the editorial solutions to D and E1 are the worst. There is no reasoning/intuition given whatsoever. Please take it as constructive criticism.

→ Reply

Yushen.

5 years ago, # |

My solution for problem F : 85724126

→ Reply

5 years ago, # |

← Rev. 3 →

+29

My solution to 1371F - Raging Thunder

Helping information

Obviously, this is a segment-tree problem. Then what do we need to maintain in each node?

In my implementation, the following information is maintained:

$$$ll$$$, the length of the prefix "<<<"
$$$lr$$$, the length of the prefix ">>>"
$$$rl$$$, the length of the suffix "<<<"
$$$rr$$$, the length of the suffix ">>>"
$$$lhi$$$, the length of the prefix ">>><<<" (at least 1 "<" and 1 ">")
$$$rhi$$$, the length of the suffix ">>><<<" (at least 1 "<" and 1 ">")
$$$hi$$$, the length of the longest ">>><<<" (at least 1 "<" and 1 ">")
$$$nlhi$$$, the length of the prefix "<<<>>>" (at least 1 "<" and 1 ">")
$$$nrhi$$$, the length of the suffix "<<<>>>" (at least 1 "<" and 1 ">")
$$$nhi$$$, the length of the longest "<<<>>>" (at least 1 "<" and 1 ">")
$$$mark$$$, whether the children of the current node should be flipped. It is also used as the flag for lazy propagation.

Then the answer to the query within a node becomes $$$\max(hi, ll, rr)$$$.

With such information, node combination and node flipping becomes easier and clearer.

Node flipping

Flip $$$mark$$$.
Swap $$$ll$$$ and $$$lr$$$, $$$rl$$$ and $$$rr$$$.
Swap $$$lhi$$$ and $$$nlhi$$$, $$$rhi$$$ and $$$nrhi$$$, $$$hi$$$ and $$$nhi$$$.

Node combination

New $$$ll$$$ is $$$l.ll$$$ if $$$l.ll < l.len$$$, otherwise $$$l.ll + r.ll$$$. It is similar for $$$lr$$$, $$$rl$$$ and $$$rr$$$.
New $$$lhi$$$:
- If $$$0<l.lhi<l.len$$$, just use $$$l.lhi$$$
- If $$$l.lhi=l.len$$$, use $$$l.lhi+r.ll$$$
- If $$$l.lhi=0$$$ and $$$l.lr=l.len$$$ and $$$\max(r.lhi, r.ll) > 0$$$, use $$$l.lr+\max(r.lhi, r.ll)$$$
- It is similar for $$$rhi$$$, $$$nlhi$$$ and $$$nrhi$$$
New $$$hi$$$ is the maximum of the following:
- $$$\max(lhi, rhi)$$$
- $$$\max(l.rhi+r.ll(\mathrm{if}\ l.rhi > 0), l.rr+r.ll(\mathrm{if}\ l.rr > 0, r.ll > 0), l.rr+r.lhi(\mathrm{if}\ r.lhi > 0))$$$
- $$$\max(l.hi, r.hi)$$$
New $$$nhi$$$ can be calculated similarly.

My code

85729867

→ Reply

vishad522

5 years ago, # |

why i am getting run time error, i am unable to find that?

include<bits/stdc++.h>

using namespace std; int main() { int t; cin>>t; while(t--){ int n,k; cin>>n>>k; int res[n][n]; memset(res,0,sizeof(res)); if(k%n){ cout<<"2"<<endl; } else{ cout<<"0"<<endl; } int p=0; int q=0; for(int i=0;i<k;i++){

res[p+1][q+1]=1;
        p=p+1;
        q=(q+1)%n;
        if(p==n){
            p=0;q=(q+1)%n;
        }
    }
   for(int i=0;i<n;i++){
       for(int j=0;j<n;j++){
           cout<<res[i][j]<<" ";
       }
       cout<<endl;
   }
}
return 0;

}

→ Reply

5 years ago, # ^ |

You set res[p+1][q+1]=1 while p+1 can be n.

→ Reply

r.zay699

5 years ago, # |

What obfuscation software was used for the jury solutions? It is certainly very effective.

→ Reply

harry_122

5 years ago, # |

← Rev. 2 →

→ Reply

tanmay2625

5 years ago, # ^ |

+13

Invalid testcase. 8 is not a prime.

→ Reply

cloudofpranav

5 years ago, # |

Can someone please provide a rigorous proof for Problem C as to why the order of guests where first all type 2 guests come and then type 1 guests come, is the most efficient ? I could not understand the explanation in the editorial.

→ Reply

weakestOsuPlayer_244

5 years ago, # ^ |

I don't have a rigorous proof but you can think in this manner that guests of type one can always adjust to any kind of case whether only vanilla,chocolate or both (provided the number of guests of type ones is less than equal to the total number of cookies) so our number of type of cookies is independent of guests of type one.

It's not so for the guests of type two since they always choose the lesser type of cookies so if you feed the guests of type one before guests of type two it's going to be less optimal.(more chances of failure of satisfying the condition. Our problem asked to print yes even if a single situation could allow all guests to finish them off so always looking for most optimal way)

→ Reply

cloudofpranav

4 years ago, # ^ |

Thank You I think I get it now

→ Reply

acm.iitism

5 years ago, # |

-14

Check out the video editorials:

→ Reply

4 years ago, # ^ |

Can you please make a video on E2

→ Reply

kingkong1

4 years ago, # ^ |

Are you color blind?

→ Reply

4 years ago, # ^ |

→ Reply

_reevu_lmao_ktksan_

5 years ago, # |

During the contest, I misread the question B as the total number of shapes of length n one can get if for any block, there is at least one block that shares one side with that block, thereby removing the problem constraint of n consecutive days. I still haven't been able to come up with any solution! Could anyone help with that?

→ Reply