Hello everybody!

On July 28 and 30 will take place tours of EJOI — individual programming competition for juniors, held under the rules of the International Olympiad in Informatics.

EJOI consists of the most interesting and hard problems that are proposed by a wide community of authors, and that is why we decided to give you an opportunity to crack the complete problemset of the contest. During the second tour of Olympiad, we are going to conduct a rated Codeforces round based on problems of both days of our Olympiad.

**We kindly ask all the community members that are going to participate in the competition to show sportsmanship by not trying to cheat in any manner, in particular, by trying to figure out problem statements from the onsite participants. If you end up knowing some of the problems of EJOI (by participating in it, from some of the onsite contestants or in any other way), please do not participate in the round. We also ask onsite contestants to not discuss problems in public. Failure to comply with any of the rules above may result in a disqualification.**

The round will happen at Jul/30/2018 11:15 (Moscow time) and will last for 2.5 hours. There will be 6 problems in each division.

The tasks of the round were invented and prepared by tourist, PavelKunyavskiy, niyaznigmatul, 300iq, GlebsHP, pashka, qoo2p5, VArtem, demon1999, flyrise, ifsmirnov, isaf27, yeputons, _kun_.

Also thanks for testing grumpy_gordon, gritukan, izban, GoToCoding, Egor, dan.io, Sert, disa, alkurmtl, senek_k, BudAlNik.

And, of course, thanks to MikeMirzayanov for great systems Codeforces and Polygon.

Good luck everybody!

**UPD:** Congratulations to winners!

D1:

1) Um_nik

2) LHiC

3) dacin21

4) ksun48

5) Swistakk

D2:

2) WaldarDoppen

3) IHaveHir

4) cly_none

5) Shadi.Gh

Please translate to English

On July 28th and 30th, EJOI tours will take place — a personal programming competition for juniors, taking place according to the rules of the international schoolchildren's competition in informatics.

EJOI is made up of the most interesting and complex tasks that were proposed by a large team of our authors, therefore, not wanting to deprive you of the opportunity to break your head over the problems of the full problem of the network together with the contest participants, we are going to run a rating round of Codeforces, based on the tasks of both rounds of the Olympiad.

In this regard, we ask all members of the community who are going to participate in the competition, show respect for themselves and other participants of the competition and not try to cheat in any way, in particular, clarifying the tasks of the participants in the competition in Innopolis. If you learned any of the tasks of EJOI (by participating in it personally, from someone from the participants or in some other way), please do not write a round. We ask the participants of the Olympiad to refrain from public discussion of the tasks. Any violation of the rules above will be an excuse for disqualification.

The round will be held on Monday, July 30, 2018 at 11:05 and will last 2.5 hours. In each division, six tasks will be proposed.

at 11:05 o'clock in Moscow time.

Sorry, it is translated now.

Thanks.

Your level is increasing man.

When you click on EJOI, it writes no such blog entry.

Thanks, fixed.

Nice IQ(300).

23 coders are here.

Wish there will be no internal errors, because... well...

500...#iykwim #okthisisabadjoke #imoutisaf27 is among the names of this round while I_love_isaf27 is among the names of the last round :D

upd:my fault,isaf27 is also among the names of the last round

isaf27 is also among the names of the last round

Congratulations everyone on the 5th century of contests here on Codeforces :) :D

5th Century of Codeforces round, not Codeforces Contest. :)

As educational rounds are also Codeforces Contest but not Codeforces round. :)

Let's hope for a nice 500th anniversary!!

Happy 500th round, Codeforces!!!

Be scared, be very scared.

ez katka round

Registered.

## Young man. Are you eager for force?

(Please downvote this comment)

i am not in !!

very excited to join a contest prepared by all of these legend.

DELETED

DELETED

A not bad time for Asian users :) ?

Hey, maybe the contests based on OI have more unusual time which is not in evening or night? :P (Like Moscow Open Olympiad (Round #469))

HAHA!You are so funny.

I can see by your cf icon you too have refined taste.

Yeah,it is so tiring to take part in contests at midnight,especially when you need to go to school tomorrow

but tomorrow's div 3 will start at night again ...

For those in US wondering how the hell to wake up at 4 AM, let me give you some advice:

3 AM here in Colombia :)

Well, did you follow this advice? Somehow I actually managed not to sleep, hope I can solve more than one problem :D

haha, no, I went early to bed. Im not sure if not to sleep is a good idea haha

Is it emmm, no no nothing.

Should I skip my school to participate?

If u can, why not?

Well, I just did this in Round #469 which has the same beginning time ┑(￣ _ ￣)┍

Херня какая-то

Yes, I made some mistakes, sorry. Next meme gonna be better ;) ________________________________________________________

Our greatest glory is not in never falling, but in rising every time we fall.

Don't worry, here at codeforce we like to say "practice makes perfect", whether it is segment tree or shitpost.

another round with tourist's problems ... ಠ_ಠ

God bless me ...

Have now made my schedule free for Codeforces #500! <3

The best feeling ever (Despite it being at 1:30 pm here, on a Monday!)

I honestly expect quite a lot of people seeing EJOI Day 1's tasks from contestants, which messes with the standings by having an unfair advantage. Expecting everyone in an online competition to play fair isn't enough.

Either make this contest unrated, or change Day1's tasks with original ones.

Leaving such an obvious method for cheating unchecked certainly shouldn't happen for Round #500, which instead should be celebrated as a milestone in fair competitive programming.

I think that the main purpose of the contest is to solve interesting tasks, not to earn ratings. We will try to do everything to fight with cheaters, but I don't think that cheaters can ruin the contest. Just enjoy the tasks!

The onsite contestants were asked to don't disclose the problems.

Anyway, there are ways to cheat in codeforces rounds even without onsite competition present. So we hope, that most people on codeforces just simply don't cheat.

I really like the tone of this comment, "Do this, do that, it should be". Should we also buy you a Ferrari and a nice house?

I think this team of experienced contest authors together with Mike can figure something out, don't you think?

300iq dude you gotta chill out. You've been included in lots of rounds recently.

Are you jealous of him??

I wonder if this contest would be too difficult for Div 2 contestants.... That's because actual EJOI is for pro junior coders..

I think that first 2 or 3 problems won't be from the EJOI, because A, B and sometimes C are very easy and I think there aren't that easy problems on EJOI. Please correct me if I am wrong.

feeling good to see tourist in problem setters...i think this contest is going to be more interesting than other contest........

upvotes for tourist.

UPD1: Now downvote for starboy_jb.

UPD2: don't do anything with this comment.

downvote for sguu

keke

14 writers just for a single contest!

I will miss this contest because I have to take part in

HDU2018 Multi-University Training Contest 3contest is here...me too,hhhh!

## ?detaR tI sI

Happy round #500

What controls registration time? We can register in Educational round but not in #501 yet.

So many writers? Pfft. I hope this won't be disappointing. Stupid writers always disappointing. I wish you all minus rating changes. F*** y'all

I hate 10 minutes delay :(

Better than long submission queues during the round.

That's better than 1 hour delay,isn't it?

Not really, if you have time.

10 min delay :'(

Ten minutes for dinner:(

What's the score distribution?

As a result of 10 minutes delay, I will miss my supper.

same here, I just want to participate to contest until end of my lunch time, which means 55 minute of the contest. Now they decreased it to 45 minutes by delaying it :(

KIKI CHALLENGE !!!! #500

After getting a %l64d error, I wasn't able to submit my code on problem A, it always said that I've submitted exactly the same code before, even if I changed things in it. I also wasn't able to ask questions on the problem.

It may be caused because I registered in the 10 minutes delay time.

When you solve A,B,C then leave cuz you don't have anything to do. :')

Same in div1...

how to solve B?

Answer can only be 2, 1, 0, -1.

-1 and 0 are trivial.

It will be 1 if (a[i] & x) != a[i] and (a[i] & x) is present in the array/set.

If it's not 1, 0, or -1 then it has to be 2.

I guess I'm back to div2 again...

How to solve B?

Represent each row and column as a node, and connect a row and column if cell (x,y) exists. Answer is number of connected components minus one.

Thanks.

Make a bipartite graph (the left side of the graph is for

r_{i}and the right side is forc_{i}, edge (r_{i},c_{i}) exists if we have element (r_{i},c_{i})), then count the number of connected component.This solution works because if you will be able to make element (

r_{i},c_{i}) if you can reach nodec_{i}from noder_{i}.DSU on rows and columns. When you're given a cell, you merge the x and y coordinates. The answer is the number of components in total — 1.

szawinis Why are you doing merge(x,n+y)? A detailed explanation of your approach would be really helpful..

Well, that's just a way of separating the x and y coordinates while keeping them in the same dsu array. Since the first n vertices are taken up by the x coordinates, we can just append all the y coordinates by doing n+y.

For div1A you can prove it is always optimal to make all the x's a contiguous subarray when the

a_{i}is sorted rightYeah, so sliding window will work.

Note that when you're doing a proof you may get stuck at one point while trying to induce a contradiction, but you should also pay attention to whether the y's form a contiguous subarray as well.

If you assume that neither the x's nor the y's make a contiguous subarray, you should be able to easily induce a contradiction of the minimality.

Yes, because you can make the x's contiguous without affecting difference between the highest and lowest y (can even decrease it), resulting in difference between the highest and lowest x being decreased (or not changed at least).

How many cases are there for Div1D?

I think 2. Update: I don't know.

Div 1:

I tried C with a DP approach but got WA on test 8: 40959527.

Can anyone point out my mistake?

Please, explain what is d[i][j][0] and d[i][j][1] in your solution (we are not telepaths)

Alright. Should have some clarifications.

In this, I declared

dp[i][j] as the minimum time used to buildjhouses, using hills with 0-based-index not higher thani.dp[i][j][0] is the minimum time to buildjhouses without building a house at thei-th hill, whiledp[i][j][1] is the minimum time to buildjhouses, including one at that hill.You need to update dp[i][j][1] with minimum of dp[ [0,i-2] ][j-1][1]

I haven't found the mistake yet, but I met difficulty when I was trying to solve it in your way which uses dp[i][j][0] and dp[i][j][1]. So I tried to use dp[i][j][0] ,dp[i][j][1] and dp[i][j][2]. dp[i][j][0] is the minimum time to build j houses without building a house at the i-th hill or i-1-th hill. dp[i][j][1] is the minimum time to build j houses, including one at that hill. dp[i][j][2] is the minimum time to build j houses, including one at i-1-th hill.

Could you please explain the transitions in your solution?

I omitted a dimension of the array, maybe this cause my solution hard to understand. f[i][j][0]=min(f[i-1][j][0],f[i-1][j][2])is obivious. f[i][j][1]=min(f[i-1][j-1][0]+max(0,a[i-1]-a[i]+1),f[i-1][j-1][2]+max(0,min(a[i-1],a[i-2]-1)-a[i]+1)) Transition from f[i-1][j-1][0] means that you only have to lower the height of i-1-th hill from its origin height, which costs max(0,a[i-1]-a[i]+1) hours. Transition from f[i-1][j-1][2] means that you have to lower the height of i-1-th hill from the height which had been lowered when calculating f[i-1][j-1][2] and is already min(a[i-1],a[i-2]-1), so this costs max(0,min(a[i-1],a[i-2]-1)-a[i]+1) hours. f[i][j][2]=f[i-1][j][1]+max(0,a[i]-a[i-1]+1), because there is no other choice. I finally realised my poor English when I was trying to explain my solution... Hope you can understand it.

Thank you very much!

The first bug I see: while counting d[1][1][1] you don't consider a case when the first (0-th) hill is lowered. Can you please determine if d[i][j][1] means that you must take the i-th hill or that you can take the i-th hill?

Looks like I do have trouble clarifying stuffs... My apologies.

dp[i][j][1] meansi-th hill must be taken.I think the bug is this: when you take the-th hill you have to lower the next hill. But then you forget about it and don't put it in consideration when you take the d[i-2] value.

You have to consider the first i hills in isolation. For example see my solution (but here 0 means that we don't count the i-th hill, and 1 means that we can count or not count it).

Solution

Finally, mistake found — a stupid one I must say.

`i64 dp[5001][2500][2];`

(It should be 2501 for the second dimension).Like, I've neglected Carrays for a very long time (only using it today since nested vectors didn't suit in 1s TL), and then this happened...

Still thanks for all the support! SirRembocodina ___tutis___ ouuan

I always use dp[5010][2510][2] in this case.

Heck, I should spend sometimes revising Carrays again, just to have the flow in my head. Such issues are straight intolerable :D

I can finally see my rating change before going to sleep as a Chinese!

And how beautiful when it is going to be a great rise.

Have been monitoring your progress, and I can only say congrats! ;)

Codeforces's progress proves the revolution successful and makes itself better!^_^

Maybe I won't be able to? The rating change hasn't been shown yet.

Now I can see the rating change.

Which problems were from EJOI. The ones from Div.1?

D1 B+ were from EJOI

Cool. Thank for the quick answear.

Can tell us tasks by days ?

https://pcms.university.innopolis.ru/results/org/ejoi/2018/day2-salty.html

I thought, which tasks was A on first day, which one was B etc. I do not see that option on the link :)

Move your cursor on A,B,C,D,E,F.

Thanks, I am stupid :D

It is interesting that there is more AC on C div 1 than on B div 1.

¯\_(ツ)_/¯

Wow, had E coded for last 8 minutes but wifi went out I guess, and iPhone took too long to recharge so not submission :'(

EDIT: :'(

Cheer up T_T..

It seems that you would have won this round if this didn't happen. So sorry.

But still, congratulations on becoming nutella :)

Ehm, I spent some 20min going in wrong directions on E, then wasted a lot of time on D. About an hour after the contest, I decided to revisit E and figured out I can take Euler cycles in about a minute... and the rest is straightforward.

My Modest Solutions

A-Piles with stones

http://coolforces.blogspot.com/2018/07/1013a-codeforces-piles-with-stones-solution.html

B-And

http://coolforces.blogspot.com/2018/07/1013B-codeforces-And-solution.html

Why cannot solve AB-Strings by greedy T_T...

Was I missed something?

How was your greedy, exactly?

Also, please keep in mind that the commands with string are linear in time complexity, in case you implemented them directly on your solution.

Yeah, I also did not understand why does not it work to just break an "ab" in one string and a "ba" in another. Though I just saw the number of OKs and the number of WAs and decided not to think further =/

For a test like:

It's better to make the number of cuts in the two strings more equal. For example, we balance them on the very first move:

Oh.. I think this was it! Thanks!!

so the first time pre-processing was that I was missing

You have 6 total letters in the input data and 8 in the output, seems like you've meant:

Thanks, fixed!

Uhm, are you sure? If you consider it correct, I don't understand the format you've chosen for it.

Argh... sorry! Another try above. Yeah, now yours and mine look equal.

Okay, seems alright now, thanks for the test

Everyone solved D as DFS!!! I never thought of it

Can someone explain it to me??

Let's present a table as bipartite graph. N vertices for rows and M for columns. Then an element means an edge in this graph. If we have a path of three edges, we can connect the ends of this path with an edge. This means that if we have a longer path a_1, a_2, a_3, ..., a_k, we can at first connect a_1 and a_4, so the path becomes shorter, and so afterwards we can connect a_1 and a_k. So if two vertices are connected, then we can draw an edge between them already. Thus we only need to buy edges to make all the graph connected, so the answer is the number of components — 1.

that moment when you spend 1 Hour try solve D then you realize that you misunderstand problem because you are shit in English

How to solve Div2 C? tried a lot but could not get it :(

I approached it with greedy solution — minimize one dimension of the desired rectangle as much as possible.

To achieve that, we will sort 2 *

ncoordinates, and choosenconsecutive coordinates in the new sorted array to be coordinates of the same dimension (eitherxory, doesn't really matter). Technically, we'll keep a sliding window and try all possiblen-length consecutive subarrays.I solved Div. 2 B in 16 min. but then accidently resent it after one hour and a half. Both of them are correct Which solution will be rated?

last submission will count

Are there ways to recover smh?

sorry, i don`t know about that

your first solution will not be judged. It will get skipped

It hurts. Thanks.

18 and 43 tests in B kills all submissions

For the problem D in Div.1, Both ko_osaga (submission) and HYEA (submission) got accepted. But for the input below, ko_osaga and HYEA's outputs are different.

<< INPUT >>

<< ko_osaga's output >>

<< HYEA's output >>

hmm, I think it should be 6 for this case.

you should join codeforces team

My answer also has 6 actions.

When a solution considers a number of cases, there can be quite a few of them (mine, for example, has a total of 18 calls to the main solution function, just to be sure). So, if one of

Xsuch cases is solved incorrectly, informally, it isXtimes harder to catch it with tests. So it is understandable if something like that slipped through.I made an identical TC in contest to test my code. It's a basic one, and if I was a tester I won't tolerate it. (But still I appreciate all setters effort!)

FYI, smaller countercases :

Answer is 4.

Nice.

NO ITS NOT NICE

Very nice

I read this in Skeletor's voice.

a sudden difficulty rise between div1 C and D .

Means we need code faster again

Guys there was some cheating going on today. Guys like n00bie knew the questions beforehand and hence achieved a good rank. What a disgrace!

I guess I'll just repeat Jatin.

Get a life.

yeah,coz you can never be original.

Can anybody tell me how to solve div1D?

Consider all substrings

`ab`

and`ba`

: the switch positions where the two consecutive letters are not equal. Note that each cut can remove at most one switch position. Now, if we cut`...a|b...`

in one string and`...b|a...`

in the other, and swap the prefixes, we remove two switch positions simultaneously. However, one swap can remove no more than two of them, one fromsand one fromt.What's left to do when, for example,

shas no switch positions andthasX> 0 of them? We can remove one switch position fromtwith one swap. But we can do better: cut the stringtso that approximately half of these positions go intos. Then we can again remove two switch positions per swap, which is twice as fast.In fact, we can start by counting the number of switch positions in

sand int(let them beX>Y), and if they are not equal, make the first cut so that (X-Y) / 2 of them go fromsintot, making their count approximately equal. After that, proceed with the greedy solution which removes one switch position fromsand one fromtin each swap.What's left is to consider which string will turn into

`aaa...a`

and which into`bbb...b`

, and formalizing the "approximately equal" above. We can just solve for a few ± 1 cases and take the minimum to be sure.Thanks a lot.

Don't. Solve E instead.

Seriously, if it looks simple and there are few correct submissions, maybe the next problem will be more doable — or if it isn't, at least it won't give you a headache.

I found E pretty easy for 2500 pointer, but D was even easier, but just tedious with implementation. For me doing E was better decision than D, but you can't simply discard such problem a priori, it wasn't

thatbad.I agree that D was easier than E idea-wise, but if the implementation is tedious enough that there are only 20 correct submissions (while 180 people attempted) in a 2.5 hour contest, and if the first correct submission came after 80 minutes... that implementation is pretty brutal. I think if D and E were swapped, there would be more correct submissions on that D than on that E.

I'm not talking about discarding it a priori, but about recognising that it's not worth spending a lot of time on — either from experience, from looking at the statistics or trying it and seeing that there are a lot of cases and it's easy to miss some (so easy it happened to the authors, even).

editorial?

Rating should be updated now But it has been 3 hours since ending and rating haven't been updated even

Seems like it's not that easy this time. Like they also rolled back the rating changes for round 499.

a sudden disappear of round499 rating record?

whats wrong with cf 499 round??

see this: https://codeforces.com/blog/entry/60921

thanks.it fixed.

Is there any problem with previous round ( Round # 499 ) rating !

Problem has been fixed.

Editorial?

http://codeforces.com/blog/entry/60920

For people who came up with D in contest, how did you come up with solution to a problem like this? Did you see similar problems or did you figure it out from writing out the pairs/graphs by hand? I understand solution, but coming up with it seems like total magic to me.

Well, yes, associating rectangular grid with bipartite graph is a known thing.

congratulations man. Only your hatters will downvote this comment.

It makes sense that we need to have all (x,y) values in the same connected component, if all cells that we make for free are (x,y) where x is from set R and y is from set C and R is set of all x's and C is set of al y's in that component.

A

Suddenly noticed that both the Announcement and the Editorial aren't linked to any contest...

And the result is

no contest materials link in the contest Dashboard or problemsetUPD: Fixed.In problem B,"In one operation you can select some i (1 ≤ i ≤ n) and replace element ai with ai & x" doesn't "some i" means several i? or it just means only one i? I though that one operation can replace several i(possibly more than one i) so I got WA in test 4.

The comment is hidden because of too negative feedback, click here to view it

Where can I find the tests to the output-only problem from EJOI?

when are u going to upload editorial??

The editorial existed all this time, but the link to it wasn't posted on this post. http://codeforces.com/blog/entry/60920