What is the Sandbox?

This "Sandbox" is a place where Code Golf users can get feedback on prospective challenges they wish to post to the main page. This is useful because writing a clear and fully specified challenge on the first try can be difficult. There is a much better chance of your challenge being well received if you post it in the Sandbox first.

To post to the Sandbox, scroll to the bottom of this page or click on the "Add Proposal" link below, and click "Answer This Question". Click "OK" when it asks if you really want to add another answer. Write your challenge just as you would when actually posting it. You may also add some notes about specific things you would like to clarify before posting it. Other users will help you improve your challenge by rating and discussing it. When you think your challenge is ready for the public, go ahead and post it, replace the post here with a link to the challenge and delete it.

See the Sandbox FAQ for more information on how to use the Sandbox.

The Sandbox works best if you sort posts by "active".

Add Proposal

Search the Sandbox

Browse your pending proposals

Get the Sandbox Viewer to view the sandbox more easily

To add an inline tag to a proposal use shortcut link syntax with a prefix: [tag:king-of-the-hill]

  • \$\begingroup\$ How are tags added to questions? \$\endgroup\$ – guest271314 Jan 9 at 7:51
  • \$\begingroup\$ @guest271314 You can use this markup to create a tag in a draft: [tag:code-golf] \$\endgroup\$ – DJMcMayhem Aug 29 at 15:19

2518 Answers 2518


coreutils default behavior stdin/stdout

We have a lot of challenges to implement just one of these operations, but a lot more are missing. Instead of adding a challenge for each of them, I thought I'd see if I could make a multiple-holes challenge that's complex enough to inspire some code re-use. This challenge is to reproduce a small subset of what Busybox does, namely to implement the default behavior of [almost] all of the GNU coreutils that (usually) read input from stdin or a file and send output to stdout or a file.

The utilities to reproduce are as follows:

  1. cat copy stdin to stdout
  2. tac copy stdin to stdout, reversing the order of the input lines (last line first)
  3. nl copy stdin to stdout, adding a line number to the start of each line. Start at 1, use spaces to pad each number to a width of 6, and add two spaces between the line number and the original line
  4. od I can't come up with a concise unambiguous way to describe the default output from od. I might skip it.
  5. base64 for every 3 bytes of stdin, split into groups of 6 bits, look those 6-bit values up in the base64 alphabet, and output 4 such bytes to stdout. wrap output lines at 76 characters. pad missing bits with 0s, and output an all-padding 0b000000 as "="

I'll finish filling out descriptions for some subset of the following if this idea proves popular enough to proceed with.

fmt, pr, fold, head, tail, split, csplit, wc, sum, sort, shuf, uniq, ptx, tsort, cut, tr, expand, unexpand, yes

The format of an entry would be either one program or one function, which can perform all of these tasks, just like busybox can. As a program, it might read its own process name to decide which tool to run. As a program or a function, the first parameter might be which tool to run. As a program, input should come from stdin and go to stdout. As a function, input should be a single string parameter, and return a single string.

  • \$\begingroup\$ I think it would be better to say "These ones have already been implemented" with links to the appropriate questions, and to then ask for a multi-tool which implements the non-dupes. That keeps things clear and, frankly, the ones which have already been done are probably mainly the less interesting ones. \$\endgroup\$ – Peter Taylor Nov 21 '15 at 23:07
  • \$\begingroup\$ @PeterTaylor fewer holes means less likelihood of code reuse. I'm annoyed at other challenges where code reuse is possible but it's not useful. I hope to see it in winning entries here. \$\endgroup\$ – Sparr Nov 22 '15 at 2:06

Find the nested source codes

A cops and robbers challenge where the cops write between 2 and 8 programs that produce output in the same language and interweave the programs together. WLOG, let's discuss this action being performed on two programs. By interweaving, I mean adding the characters of the second program to the first program so that when the characters of either the (WLOG) first program are removed, the second program can be seen. Obviously, commenting in any program is not allowed.

The cops will post their combined codes, the number of different programs in what they post, and the language the codes are written in, and what the programs print as output. To get credit for cracking the submission, the robbers must post the split codes and what each one outputs.

As a general rule, cops cannot use a language more than once.


A cop will receive points if their submission is safe for one week from the time of posting. Their score will be the sum of the two following.

  • The first value is 256 divided by 2 to the power of the number of different programs used.
  • Round the number of bytes in the combined codes up to the nearest power of 2 and call this number x. The second value is 1024 divided by x.

A cop will lose 10 points for every code that is cracked before the one-week period is up.

Robbers will receive a number of points equal to the sum of the following for each cracked submission.

  • The first value is 2 to the power of the number of different programs used in the cop's answer.
  • Round the number of bytes in the combined codes up to the nearest power of 2 and call this number y. The second value is 1024 divided by y.


Python, 2 codes


Prints llama and h

Codes: print("llama") and print("hello"[0])

  • \$\begingroup\$ Can I see an example? \$\endgroup\$ – Conor O'Brien Nov 22 '15 at 1:46
  • \$\begingroup\$ @CᴏɴᴏʀO'Bʀɪᴇɴ Updated. \$\endgroup\$ – Arcturus Nov 22 '15 at 1:54
  • \$\begingroup\$ On interweaving: does one have to say which method they used to interweave? \$\endgroup\$ – Conor O'Brien Nov 22 '15 at 1:57
  • \$\begingroup\$ Interweaving is done in order. The example given would be prPRiIntN("T**"(HlElLamLaO")"[0]**, where bold and capitalization represent the second code. \$\endgroup\$ – Arcturus Nov 22 '15 at 2:03
  • \$\begingroup\$ @CᴏɴᴏʀO'Bʀɪᴇɴ Think this is good enough to post officially? \$\endgroup\$ – Arcturus Nov 22 '15 at 2:36
  • \$\begingroup\$ I'd wait a day or two, and ask some other people in chat. \$\endgroup\$ – Conor O'Brien Nov 22 '15 at 3:03

Complete the Digit Sequence

We have a string of digits with some elements missing and marked with .s.


We want to fill the missing parts in a way that they would form an arithmetic sequence with the previous on next elements. E.g. 8...2 becomes 8642. If this kind of filling is not possible with single digits, mark the positions with ?. E.g. 7..3 would become 7??3.

With this rules our original example becomes


You should write a program or function which receives a digit string as input and outputs or returns the filled sequence.







This is code golf so the shortest entry wins.

Sandbox note: this seems boring so ideas are welcomed.

  • 1
    \$\begingroup\$ Idea: turn the entire string into an arithmetic sequence of as many integers as possible (after inserting spaces). e.g. 7....2 would become 765432, i.e. 7 6 5 4 3 2, but 7....3 would have to become 791113, i.e. 7 9 11 13, whereas 7....4 would have to become 727374 (72 73 74) or 767574 (76 75 74) \$\endgroup\$ – Martin Ender Nov 23 '15 at 11:38
  • \$\begingroup\$ @MartinBüttner I like it although I would like to avoid the extensive brute-force approaches. \$\endgroup\$ – randomra Nov 23 '15 at 12:37
  • \$\begingroup\$ I just noticed there are a lot more possibilities for my last test case, like 72 53 44. But yeah, I'm not sure how much one could optimise solutions for this. (Also, I'd still keep the possibility of given digits within the sequence, I just used the x....y format for simplicity.) \$\endgroup\$ – Martin Ender Nov 23 '15 at 12:40

The Language Relay!

A typical relay race only has four competitors per team, but where's the fun in that? Instead, let's see how many teammates you can cram onto the bus to the stadium and still finish the race. (This analogy is starting to break down, so I'll cut to the chase.)

Your task is to write a program or function in 256 bytes or less. It will take no input, and its output will be a program or function in another language. That program or function will also take no input, and its output will be a program or function in another language... and so on and so forth, until the last program, which will output the following:

.     \O/      .
|===== |_ =====|
|    _/  |     |

The winner is whoever manages to use the most languages. If there is a tie, the winner is the one with the shortest code in bytes.

You cannot use the same language twice, and different versions of the same language do not count as different languages. The output must also be different every time (no using languages that leave the program or function untouched.)

Here's a small example:

int main(){printf("print(\".     \\O/      .\\n|===== |_ =====|\\n|    _/  |     |\")");}

This C code produces this Python code:

print(".     \O/      .\n|===== |_ =====|\n|    _/  |     |")

And the Python produces the final output. I've used two languages, so my score is 2, and I'm going to lose terribly.

  • 1
    \$\begingroup\$ I'd implement some way to incorporate byte counts into the score because otherwise I'm sure there will be plenty of ties. \$\endgroup\$ – Downgoat Nov 25 '15 at 17:45
  • \$\begingroup\$ Good idea, thanks. \$\endgroup\$ – Luke Nov 25 '15 at 17:46

This might already exist, but through my search I couldn't find anything like it, so here it goes.

Find My Number

My friends and I have made are playing a game where we have a variable N that represents a number from 0 to 10. Using an expression they give us containing + - * / for addition subtraction multiplication and division, > < = for greater than, less than, and equal to, and the integers 0-10, we must find there number or what their number can possibly be.

I’m lazy, so I want you to write a Program that takes an input from STDIN or an acceptable alternative and return the possible numbers to STDOUT or an acceptable alternative.





1, 2, 3, 4









Because I might be caught, I want the program to be a small as possible in characters to avoid me friends seeing it, so the shortest solution wins!

  • \$\begingroup\$ What it is impossible such as N<N or the result is all real numbers such as N=N? \$\endgroup\$ – Downgoat Nov 26 '15 at 16:18
  • \$\begingroup\$ Well in the first one is impossible, so nothing works, so it doesn't output anything. The second one since anything works it prints all numbers 0-10 \$\endgroup\$ – Generic User Nov 26 '15 at 20:12
  • \$\begingroup\$ I assume that the n in one of the examples is a typo for N. This is not really an interesting question. The addition of < and > makes it marginally different to existing "evaluate this expression" questions, but it's still trivial to handle with eval in languages which use infix operators. \$\endgroup\$ – Peter Taylor Nov 26 '15 at 21:14
  • \$\begingroup\$ Thanks for pointing out the typo, I think expression still works since I explained what I mean, but I will try to find a way to reword the question. Anyone have suggestions? \$\endgroup\$ – Generic User Nov 26 '15 at 21:43

I'm thinking I want to make a KOTH challenge that has bots play the classic game of Mafia. Bots will be placed in groups of seven, roles randomly assigned, and they play the game!

Explanation of Game Mechanics The game proceeds in multiple turns, each turn consisting of Night followed by Day. At night, each person completes his role as will be described below. During the daytime, players (in real life) discuss the events of the night (communicated by God, a separate person who does not participate in the game besides prompting people to their roles.) Just before night, the players have the option of voting to lynch another player, the objective being to lynch a mafia member to help the town win. Choosing not to lynch is always an option if you think it is too risky to lynch someone who might be innocent.

In real life, God would tell everyone to sleep, then ask mafia to wake up and silently agree on who to kill, tell them to sleep, and continue with the other roles in a similar manner. Town wins if all the mafia are dead; mafia wins if they at least outnumber the town.

There are four roles in this game, but I have played many games with more than seven people with more ridiculous roles that are very fun to play. Each player bot is assigned an ID number indicating his role.

  • Mafia - There will be 2 mafia in this game. The two mafia will come to a consensus on who to kill. If one is dead the other decides on his own. The two mafia can not choose one of them to be killed. Mafia has a role of either 1 or 2, which are identical in function.

  • Inspector - There will be 1 in this game. At night he visits a person, and receives a report on that person's innocence. Both Mafia members will appear as guilty and the rest will appear as innocent. He can not inspect himself nor dead people. His number is 3.

  • Doctor - There will be 1 in this game. At night he visits a person, and that person will not die that night. Doctor can not save himself nor dead people. His number is 4.

  • Townsperson - Does nothing at night. Numbers 5, 6, 7.

The Challenge

You will write a Java player class who will have (at least) three important methods: night, claim, and vote. All of these three methods will return the player to 'act' on (of course, depending on your role), given an ArrayList of information that the other players have claimed in public.

I am not very clear on how to run a KOTH challenge and what github is supposed to do and how to use it so it would be very helpful if someone could point me in the right direction. As of now I'm still working out the exact details of a Java class to run the game, as well as making a generic Player class that the answers will have to extend and use methods from.

Right now, I have a generic Player class that each answer will extend, and each Player object has two Identity objects: a public, claimed Identity and a private Identity. The private Identity will hold the players actual role (kept secret, of course), as well as a doctor arraylist of Integers and a cop arraylist of Strings. If the player happens to be a cop, the controller program will add an entry to the private cop arraylist which will serve as his way of receiving a verdict. I'll make a method called verdict() which returns the last element to make the programming easier. An example string is "4G" which means that player 4 is guilty. This player 4 may be mafia 1; player numbers are just given for discussion purposes and to identify a player based on his claims and is independent on role. Doctor is the same except integers because you only need to store who was saved that night.

The public, claimed Identity works in much the same way except that the controller will never modify it, only the player can. An identity object contains a role (an integer representing the role number), an ArrayList of strings called 'visits' (to be used if you are a cop or doctor, to store a list of people visited and any outcomes), and an ArrayList of suspicions to indicate that your bot suspects or is guessing that another player has a given role. It will also have a lynch value which can be changed during the day, and this is where a player decides on who to lynch (if at all). Finally, it will have a boolean for if the player is dead or alive. You had better not mess with this. The idea of having two is that you can claim whatever you want; so a Mafia can claim to be Cop or a townsperson and the doctor can claim to be a cop.

Initially your public claim ID is 0, indicating that you have not claimed. The program will give you your private ID, which will contain your actual role and and is not necessarily meant to be made public just yet. The program controller will only modify your private ID object, and only then if your are Doctor or Cop. It will append to the visits ArrayList in the private identity object the result of the visit; the doctor will have access to everyone he has visited, and the cop will have access to everyone he has inspected and their alignment.

The controller, each night, will ask Mafia 1 for his choice then Mafia 2 for his choice, then repeat, say, 25 times. If at any point the two mafia agree, then it stops because the Mafia have just made their choice. The number of times attempted thus far will be given to the Mafia's night method, so the code will look like M1.night(<other players info>, 1) then M1.night(<other players info>, 2), etc. is that arraylist of public identities previously mentioned. If no consensus, no kill. Similarly prompt cop and doctor once for their choice on who to visit.

In the day time, every player's lynch value will be 0, indicating that they do not vote yet; each player will be given 25 chances to act. An action can consist of adding to the list of suspected values, claiming a role, changing a role, changing the list of suspected values, voting to lynch someone, changing a lynch vote, etc., or more. Some actions are smarter than others, and your goal is to have a good action strategy. Upon each call of a players act() method, he will be given an arraylist of everyone else's public identity objects, as well as how many times before he has been prompted. This gives him access to what other players claim to be, who other players claim to have visited and outcomes, who they suspect as who, and who others are voting to lynch. If at any point there is a majority (more than half but not exactly half) of people wanting to lynch a person (No lynch, indicated by the value -1, counts too), then day time ends, that person is lynched and his role revealed. If majority voted -1 (no lynch) then nobody dies and day time ends.

Again, please guide me on how I should design my classes, controller, and what files I should put up and how people are going to test their bot at home. I would be glad to show you what I have so far, which includes a controller program and a dummy player class.

  • \$\begingroup\$ Note that we already have a Mafia KOTH here in the sandbox. \$\endgroup\$ – Alex A. Dec 5 '15 at 1:45
  • \$\begingroup\$ @AlexA. Note also that he allowed me to take over. \$\endgroup\$ – Faraz Masroor Dec 5 '15 at 2:17
  • \$\begingroup\$ Some important notes: The messages from the past post were a really good idea. Use them. Also, I think that its a good idea to have people submit a bot that plays a single role. Then, from round to round, simply switch in/out the bots for that role. \$\endgroup\$ – Nathan Merrill Dec 7 '15 at 4:40
  • \$\begingroup\$ I am not very clear on how to run a KOTH challenge and what github is supposed to do and how to use it so it would be very helpful if someone could point me in the right direction I like this KoTH, and I've got quite a few of them under my belt. If you pop into chat I'd be happy to chat about it with you (and help with the code, if you'd like) \$\endgroup\$ – Nathan Merrill Dec 7 '15 at 4:49
  • \$\begingroup\$ @FarazMasroor are you still planning on doing this? \$\endgroup\$ – Nathan Merrill Dec 27 '15 at 13:12

Generate a graphical representation of a Stern–Brocot tree of depth n

enter image description here

I am aware of this challenge. Would this be considered as a dupe?


Gilbert-Shannon-Reeds Shuffle


The Gilbert-Shannon-Reeds (GSR) shuffle is a simple model of riffle shuffling close to how real humans shuffle a deck of cards. The well-known rule of thumb to riffle-shuffle a deck of 52 cards seven times for sufficient randomness is based on the GSR shuffle.


  • Cut the deck at a position k, 0≤k≤n. If there are n cards, the probability that any given k is chosen is (n nCr k)/(2^n).

  • Put the first k cards into one pile, and the other n-k in a second pile.

  • Until all the cards are gone:

    • Where the sizes of the two piles are x and y, choose the first pile with probability x/(x+y) and the second pile with probability y/(x+y).

    • Move the first card in that pile to the new array.


Input: An array of positive integers, of length n<1000.

Output: The array shuffled once.


You may use any algorithm that gives equivalent results to the GSR shuffle.

  • \$\begingroup\$ Just to clarify: We have to shuffle once, yes? \$\endgroup\$ – Dennis Dec 4 '15 at 19:45
  • 1
    \$\begingroup\$ Closely related. This is one of the rare cases where I would actually favour closing an older question as a duplicate of a newer one. \$\endgroup\$ – Peter Taylor Dec 4 '15 at 20:27

Play Chess with a One-Move Lookahead

"I see only one move ahead, but it is always the correct one."

– Jose R. Capablanca, World Chess Champion 1921-1927

This is a chess tournament with a twist: your chess engine is only allowed to look 1 move (2 ply) ahead. In order to succeed, you must create the best board evaluation algorithm.

Additional Rules

  • En passant, castling, and under-promotion will be allowed. Of course, it is up to you if you want to bother adding those capabilities to your AI. I personally find these to be some of the best rules of chess.

I am considering making a "template" bot which implements alpha-beta pruning, and requiring users to just fill in the method for board evaluation. Otherwise, there may be ways to stretch what it means to "look ahead." The benefit for users would be that they don't have to write their own getLegalMoves() method.

  • 1
    \$\begingroup\$ How is the algorithm scored? \$\endgroup\$ – Doorknob Dec 7 '15 at 2:24
  • 1
    \$\begingroup\$ I don't think there's any way to specify "look ahead" which isn't vulnerable to stretching. Even if you just ask for the board evaluation. \$\endgroup\$ – Peter Taylor Dec 7 '15 at 9:44
  • 1
    \$\begingroup\$ Why would you need pruning for a bot that only looks 1 move ahead?? \$\endgroup\$ – feersum Dec 7 '15 at 11:50

Rearrangement Inequality: The Sequel

Read the previous version here.

Your mathematics teacher looked at your test results on inequalities. Not good.

Hence he decided to give you some homework on inequalities.

Since this is about inequalities (how unfair), he decided to give each student a differing amount of homework. The exact amounts do not matter to him.

After allocating the homework, he had a bunch of complaints. Apparently, some students have some neighbouring students who have less homework than them.

Hence, he has decided to rearrange the students such that there will be fewer complaints. As long as each student sees that at most one adjacent student has less homework than them, they will not complain. (The students are quite reasonable.)

Note that he can swap the position of two students, but he cannot move a student into an empty spot as the empty spot has no chairs and the chairs in this classroom cannot be moved around due to safety concerns.


Here are different forms of input you can consider:

Optional: You can include integers in your input for the size of the classroom for convenience. This is , so input and output does not matter too much.

Function input: Container with truthy and falsy values representing if a student is there or not. All truthy values must be the same, similarly, all falsy values must be the same.

Standard Input: A grid of 2 different characters, which can be separated by spaces. Possible input:

0 1 1 0 1
1 1 1 1 1
1 0 1 1 1
1 1 1 1 1




The teacher wants to know how to arrange the students. Each student can be represented as an integer from 1 to n from the least amount of homework to the most amount of homework. For the input:

0 1 1 0 1
1 1 1 1 1
1 0 1 1 1
1 1 1 1 1

where 1 means there is a student and 0 means there isn't any students, a possible output is:

0  12 13 0  4
10 11 16 15 3
9  0  14 17 2
8  7  6  5  1

where the students labelled 16 and 17 are not satisfied.

Your score for every test case will be the number of students which are satisfied over the total number of students in that test case. It is guaranteed that there is at least one student in each test case. Your final score will the average of your scores over all test cases.

If you want to use another form of input and output, please clearly state it in your answers.

The teacher has some concerns:

Please ensure that your code is deterministic, as the teacher may test it several times. You are allowed to code your own psuedorandom number generator if you need random numbers.

Please ensure that your code terminates within a minute for each test case, as the teacher is impatient.

The teacher has been to PPCG for a long time, so he knows all the standard loopholes.

The teacher has a very large classroom (100 by 100), and would be worried if your program fails to give an answer for such a large classroom.


This is , and the winner will be the person with the largest score.

Sandbox Meta:

I still need to generate some large test cases, so I'm leaving this in the Sandbox until I have large test cases. How should I show people large test cases?

  • \$\begingroup\$ Is it acceptable to use a built in pseudorandom number generator provided it is always run with the same seed? \$\endgroup\$ – trichoplax Dec 29 '15 at 15:37
  • \$\begingroup\$ @trichoplax I may need to be able to test the code multiple times, so that may require all compilers/interpreters to implement the same pseudorandom number generator, which I am not sure if that is the case. \$\endgroup\$ – Element118 Dec 29 '15 at 15:41
  • \$\begingroup\$ Ah, I see. That is definitely guaranteed for some languages, and definitely not for others... Banning all built in RNGs does at least seem fair to all languages. \$\endgroup\$ – trichoplax Dec 29 '15 at 15:58

Approximate an image using all colours

I know that this is basically a subset of the then very popular American Gothic in the palette of Mona Lisa: Rearrange the pixels but I think it is worth making it a new challenge.

As @MartinBüttner pointed out, there has also been the Images with all colors challenge, but in my opinion it has very little in common with the proposed challenge.


Given a image of 2^12 x 2^12 pixels as input, your program should recreate this image, but you have to use each of 8-bit RGB colour exactly once.


These are roughyl 12MP images, quite large. So one could just restrict it to 7-bit RGB then the pictures would only have to be 2^18 pixels, that means e.g. 512 x 512 pixels which would be way more suitable for the challenges here.

Or does anyone know a convenient other colour representation?

  • 1
    \$\begingroup\$ There's also this. I'm really not sure you'll get any better approaches than those used for CH's palette challenge. \$\endgroup\$ – Martin Ender Dec 22 '15 at 19:11
  • \$\begingroup\$ I was wondering that too on the other hand you have the advantage of exactly knowing the pallete and at the same time the disadvantage of having to use all the colours, even those that do not follow a "natural distribution" (a palette that is similar to "natural" images.) \$\endgroup\$ – flawr Dec 22 '15 at 19:15
  • \$\begingroup\$ I lean towards thinking this is too much of a duplicate of "in the palette of", but if it turns out to be accepted I'd recommend changing the title to "Approximate an image using all colours" to make it clear it's not "Images with all colors" again. \$\endgroup\$ – trichoplax Dec 29 '15 at 15:52
  • \$\begingroup\$ As noted in my answer to "in the palette of", it was directly adapted from a program which does this, so as far as I'm concerned it's a dupe. \$\endgroup\$ – Peter Taylor Dec 30 '15 at 10:54

Make a New Year countdown

Editor's note: This challenge is cancelled because it is the year 2016, and I'm too late. XP It is kept here for posterity.

It is New Year coming soon. As such, it would be a great idea to make a New Year countdown. That is your challenge today!

Happy New Year!

The full rules

  • Create a program taking from STDIN the current time and outputting a countdown until New Year into STDOUT.

  • The program should be flexible - after Year 2015 has concluded, the program should count down until New Year 2017, and so on. See examples for more information.

  • The program should count down until it is taken down by external means (using Ctrl-C, the Task Manager, the reset button, et cetera).

  • The countdown may be formatted however you like.

  • At the null second in the new year (YYYY-01-01 00:00:00), it should output 0 days, 0 hours, 0 minutes 0 seconds in your chosen format.

  • This is , so shortest answer wins.

  • The last rule: Have fun!


Given the input after from: the following countdown should be given, in any format you like (after to:).

2015-12-31 06:00:00
0 days 18:00:00

2015-12-31 23:59:57
0 days 00:00:03

2016-01-01 00:00:00
0 days 00:00:00

2016-01-01 00:00:01
364 days 23:59:59
  • 1
    \$\begingroup\$ Three things: 1. IMO this is close enough to this question that it's a borderline dupe. 2. 2016 is a leap year, so from 2016-01-01 00:00:01 to 2017-01-01 00:00:00 is 365 days 23:59:59 (assuming no leap seconds are added, but that can't be predicted in advance). 3. Because of points such as the previous one, for testing date questions it's usually better to take the current time on stdin rather than read it from the clock. \$\endgroup\$ – Peter Taylor Dec 31 '15 at 10:59
  • \$\begingroup\$ For 3; changed the rules to get the time from STDIN \$\endgroup\$ – user48538 Dec 31 '15 at 16:40

Sig-fig calculator

As a scientist, sig-figs are definitely one of the most important parts of measurements and calculatations (unless you are a theoretical scientist, where everything must be exact!). Sig-figs is an important way of measuring uncertainty and accuracy of a value. To calculate the number of sig-figs for a value, use the following to help you (from here):

  1. ALL non-zero numbers (1,2,3,4,5,6,7,8,9) are ALWAYS significant.

  2. ALL zeroes between non-zero numbers are ALWAYS significant.

  3. ALL zeroes which are SIMULTANEOUSLY to the right of the decimal point AND at the end of the number are ALWAYS significant.

  4. ALL zeroes which are to the left of a written decimal point and are in a number >= 10 are ALWAYS significant.

  5. Exact numbers in an equation have infinite sig-figs. For example, in the equation $A=pi*r^2$, $pi$ has infinite sig-figs and the exponent of 2 also has infinite sig-figs. (not included in the above link)

To calculate the number of sig-figs from a calculation, use the following rules:

  1. For multiplication and division, the number of sig-figs for the result is equal to the number of sig-figs for the least accurate value (i.e. the value with the least number of sig-figs). For example, 2.000*5.00 = 10.0 and 4.5*6.00 = 27

  2. For addition and subtraction, the result has as many decimal places as the one with the least decimal places. For example, 5.00-2 = 3 while 6.0-3.000 = 3.0

  3. logs (including natural logs) have as many decimal places as the number of sig-figs of the value of whose log is being taken (I am not sure if this sentence of English is correct!). For example, log(2.45) = 0.389.

  4. Other functions, such as square roots, exponents, sines, cosines, etc. can be assumed to have the same number of sig-figs as its argument. So the sin(3.14) = 0.00159


The challenge is to write a sig-fig calculator with the following functions:

  1. addition, subtraction, multiplication, division
  2. sine, cosine, tan, and their inverses
  3. sinh, cosh, tanh, and their inverses
  4. Exponents, including shortcuts for e^x and 10^x, along with sqrt. The exponents are assumed to be exact and has infinite sig-figs
  5. log of base 10 and natural log.

The result of each calculation should give you the correct result, but insert a "\" before the last sig fig. If there is infinite sig figs, there should be no "\" at all. Also, undefined and infinite values should give an error or NaN or print "infinity" or "undefined". For example,

sin(3.14) = 0.0015\9265292
log(1.01) = 0.00\432137378
5.00*2 =\10

This is a so the shortest code wins!

  • \$\begingroup\$ The positioning of the backslash in the output spec is rather confusing, but I think that one of the examples disagree with the spec. 5.00 * 2 is 10 with one sig fig according to the rule for multiplication that you take the number of sig figs of the least accurate value. Of course, this runs into the problem that your (unconventional, I believe) rule 4 does not provide any way to represent the number 10 with only 1sf, but even so I think the correct output should be \10. \$\endgroup\$ – Peter Taylor Nov 24 '15 at 22:06
  • \$\begingroup\$ @PeterTaylor thanks! It is now fixed \$\endgroup\$ – TanMath Nov 24 '15 at 22:59

I wasn't able to find a challenge like this before, but it wouldn't surprise me if it already exists.

A strand of DNA is made up of bases notated by the letters A, T, C, and G. A always pairs with T, C always pairs with G, and vice versa.

Therefore, you can find the opposite side of a DNA strand by swapping all occurrences of a base with its complement.

Here's the catch: your program cannot contain the characters A, T, C, or G at all – in string literals or in the body of the program.

The input is a string of continuous uppercase characters (you can assume that this string only contains the above bases). The input can be any length. The output should be the complementary strand of DNA.


Input:   AC
Output:  TG


Input:   GCATC
Output:  CGTAG


This is code golf, so the shortest program wins.

  • \$\begingroup\$ I think this is the closest existing question. Note that it's a more complicated question, and the winner is still only 24 bytes. I have a 16 char solution to this question in CJam. \$\endgroup\$ – Peter Taylor Jan 5 '16 at 23:18
  • \$\begingroup\$ It seems like this is basic character table translation in languages that support that, in pyth and cjam these answers seem near optimal, they're also kind of boring, idk if people will think it's too boring... \$\endgroup\$ – FryAmTheEggman Jan 6 '16 at 15:28
  • \$\begingroup\$ @FryAmTheEggman that's why I added the requirement the character restriction. I don't think it will be as easy to do a character replacement without using the actual characters. \$\endgroup\$ – erdekhayser Jan 6 '16 at 18:20
  • \$\begingroup\$ @erdekhayser in many languages you can just use base encoding of some kind to avoid that. \$\endgroup\$ – FryAmTheEggman Jan 6 '16 at 18:35
  • \$\begingroup\$ @FryAmTheEggman simple thing would be to prohibit the use of base encoding... \$\endgroup\$ – TanMath Jan 6 '16 at 19:05
  • \$\begingroup\$ @FryAmTheEggman I have a 17 byte Pyth answer using only simple ASCII characters. It is possible to solve this challenge without base encoding \$\endgroup\$ – TanMath Jan 6 '16 at 19:06
  • \$\begingroup\$ @TanMath without base encoding I think most languages would just add to character codes... \$\endgroup\$ – FryAmTheEggman Jan 6 '16 at 19:17
  • \$\begingroup\$ @FryAmTheEggman I think that is the only route to go, which is what I used. Any problem with that? \$\endgroup\$ – TanMath Jan 6 '16 at 19:20
  • \$\begingroup\$ @TanMath not necessarily, but a challenge where the only challenge is working around an obvious solution by restricting workarounds isn't one I would personally find interesting. \$\endgroup\$ – FryAmTheEggman Jan 6 '16 at 21:43
  • \$\begingroup\$ @TanMath or just specify ranges around the desired characters in the translate command. Like this 15 char Retina solution T`B-H@-U`H-BU-@ \$\endgroup\$ – Rainer P. Jan 6 '16 at 22:18
  • \$\begingroup\$ 20 bytes in Seriously using only arithmetic on character codes (no base encoding or translate): ,O`6╙(-7P/≈u6╙+c`Mεj \$\endgroup\$ – quintopia Jan 7 '16 at 16:55
  • \$\begingroup\$ Should I add a restriction on base encoding, or should I not post this challenge? \$\endgroup\$ – erdekhayser Jan 7 '16 at 19:48

Format a list of words

Your challenge is to format a list of words across multiple lines that are no longer than a given number of characters, so that each line contains as many words as possible and no words are unnecessarily cut off.


The input will be a space-separated list of words and then a number that is at least 4.


The output should be the input words grouped into lines so that none of the lines contains more characters than the input number. The lines should be output in the order they were input. The words should be comma-separated, and each line except the last should end with a comma. If a word is too long to fit on a line, it should be cut off as little as possible while following the other rules, and "..." should be added to the end.

Test cases

foo bar baz qux 12

foo, bar,
baz, qux

foo bar baz qux 5


strength dexterity constitution intelligence wisdom charisma 10


quas wex exort 4

Sandbox questions

  • Has this been done before?

"Compress" text into Zalgo

Zalgo, the Nezperdian hivemind of chaos, is a type of text that uses combining characters to make very tall and noisy text. If you don't know what I'm talking about, maybe this will jog your memory:


H̡̢̢̡̡̧̢̡͎̟͚̮͓͇̦̮̙̗̜̱̱͔̲̹̣̱̠̀̀͐̑̾̓̃́̃̍̀͆̇͆̽̔̒̚̕͘̚͜͝͠͠ͅE̢̧͓̺͉̟͙͇̳̰͉͖̺̻͕̰̱̝̳̙̰̟̠̯̘̰̲̎̑͋͂͑́͛̎͋̇̍̾̊̈́̂̽̿͆͛͑̽̒̊́͠͝͠͝͝ͅ ̨̢̟̳̥̖̺̼͎̩̘̰̣̼͇̰̫̞̜̲̰͔̗̠͔̩̻̳͇̾͌̆̑̍̄̊͗̓̃̆̊̄̽̐̂͛̏͑̒̓̆͝͝͝Ç̛̬̩͙̱̥̦̪̮̖͚͚͔̼̱̺̳̳̬̭͍̣͍̙̹̜̫̟̳͌̓͗͊̐̈̄́̏̀͂̎̃̈́̈́̎͋̀̒̊̀̈́͒̽͘̕͝Ȏ̡̡̨̡̝̬̠͚̠̯͖̹̟͓̮̻̲͙̖̪̯͇̍̅̂̌̌̒͗̈́̉͆̇̑͒̉̂̾̃̌̽͛͘͝͝ͅM̡̢̢̘͉̤͍̫̺̻͕̱̤̤̞̟̞̹͉͓̥̳͖̹̤̆̋̓͂̂͑̃̌͛͂̋̂̓̏́̀̾̋̈́̅̐̅̎̇̐̽͜͝Ȩ̨̛̭̥̹̳̫͎͖͈̳̠͍͙͉̻̼͍̞̜̺̝̻̝̗̳̏̈̓͋̐́̋͆͋̓̿͐͆̾̾̃͌͌̾̊́̚͘͘͜S̡̢̡̡̛̯̪̬̹̲̙̮̲̲̤͖͖̞̲̞̼̪͓͇̤̼͇͆͋̊̈̑̆̿̐̎͑̾̅̀̒̓̎̐̍̽̈́̋̽̓̔̍͜͝


As a new twist on the age-old Zalgo generation problem, your goal is to take a string and "compress" it into a single Zalgo character. You will write two programs, one which takes in a string of printable ASCII and outputs a single Zalgo character, and a second which takes in a single Zalgo character and outputs the original ASCII string.

  • 4
    \$\begingroup\$ I think the most golfable solution will be to encode the input in unary and use a single combining mark. Is that what you're looking for? \$\endgroup\$ – Martin Ender Jan 9 '16 at 22:11
  • 1
    \$\begingroup\$ Maybe require the zalgo output to be shorter in some sense? \$\endgroup\$ – xnor Jan 9 '16 at 22:38

Sorting trains

This problem is based off of a solitaire card game called Calculation. You can play it here (highly recommended).

You are in charge of designing a train yard. At your junction, you have an In/Out track (queues) and Storage (stack). Train cars come in a random order; your job is to arrange them into 4 different ordered trains. Land is expensive, so we need you to minimize the amount of storage tracks we need.

Your rail car mover can only handle 1 train at a time, and move a car:

  1. In track -> Out track
  2. In track -> Storage track
  3. Storage track -> Out track.

Each Out Track needs the same 25 cars, each in a different order:

Track 1: 1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,0
Track 2: 2,4,6,8,10,12,14,16,18,20,22,24,1,3,5,7,9,11,13,15,17,19,21,23,0
Track 3: 3,6,9,12,15,18,21,24,2,5,8,11,14,17,20,23,1,4,7,10,13,16,19,22,0
Track 4: 4,8,12,16,20,24,3,7,11,15,19,23,2,6,10,14,18,22,1,5,9,13,17,21,0

We can use (nx+n) mod 25 to calculate ID of the xth car of the nth track.

Important notes:

  • Storage tracks can store as many cars as you want, but you can only remove the most recently placed car
  • Cars cannot be moved from one storage track to another
  • Your algorithm must be deterministic (it must do the same thing given the same input queue and the same number of storage stacks)
  • You may get into an impossible situation, where you end up not being able to order the trains. If that happens, try again with additional storage tracks.


Your input is a list of integers, where each integer is the ID of the train car.


You need to return a list of moves that sorts the cars into the Out tracks. A move looks like I->S1 or S1->O2, where I is the input track, S# is a storage track, and O# is an output track.


Your score is the total number of tracks you need for of all test cases (you can use a different number of tracks for each test case). Lowest score wins.

Test cases:


Risk dice battle

Risk is a board game in which you attempt to conquer the world by attacking your opponents' countries. As the game progresses the number of armies involved in attack gets higher and higher, which can lead to a lot of dice rolling. I want you to make this easier for me. It's possible that my favourite answer will actually get used when playing Risk.

Rules of Risk battles

Each country in Risk must have at least one army on it. Therefore a battle can only take place if the attacker has more than one army (in case he loses.) The defender obviously has at least one army.

The attacker rolls 3 dice (only 2 if he has only 2 armies) while the defender rolls 2 dice (only 1 if he has only 1 army.) Once the dice are rolled, the highest dice of each player are compared, then the next highest. (If one player rolled more dice than the other, his lowest dice are discarded.)

For each dice comparison, the player with the lower score loses an army. If the dice are the same, the defender wins and the attacker loses an army. Note that this does not necessarily put the attacker at the disadvantage, as he frequently has more dice to roll than the defender.


               Attacker 6 3 2
               Defender 5 4
                        ^ ^
                        | |
 Defender loses 1 army -+ +-Attacker loses 1 army

For interest the probabilities are as follows:

                        Defender rolls 2 dice       Defender rolls 1 die
Attacker rolls 3 dice   Attacker loses 2 29.26%     Attacker loses 1 34.03%
                        Both lose 1      33.58%     Defender loses 1 65.97%
                        Defender loses 2 37.17%

Attacker rolls 2 dice   Attacker loses 2 44.83%     Attacker loses 1 42.13%
                        Both lose 1      32.41%     Defender loses 1 57.87%
                        Defender loses 2 22.76%


A full program is required which will accept from stdin or commandline, a number of attacking armies and a number of defending armies. There will be at least 1 army of each. Your code will display the number of armies as follows

1.If there is only 1 attacking army, your program shall immediately terminate with the message Insufficient force.

2.The code shall now accept a user input from stdin. If the user now enters anything other than an empty string, the code shall terminate. If the user enters an empty string, you must simulate the roll of the appropriate number of dice, sort each player's dice in descending order, and report the result and the updated number of armies per example below. 2 trailing newlines are required after the output.

Attacker dice: 5 3 1
Defender dice: 5 4

Attacker: p armies (where p is the number of attacking armies)
Defender: q armies (where q is the number of defending armies)
(2 trailing newlines)

3.If either player now has 0 armies, display the message Defender defeated! or Attacker defeated! as appropriate and terminate the program. Similarly, if the attacker now has only 1 army, terminate with the message Insufficient force.

The program shall now loop back to step 2 and continue until either one player's armies are depleted or the user enters a non-empty string.


The distribution of the dice throws shall be exactly as with real dice (up to the limits of uniformity both the dice and the random number generator used.) It is expected that most submissions will generate the numbers for each die and then sort them. Clever submissions that generate the output in other ways avoiding the sorting step are acceptable, but the theoretical probability distribution of the output must be identical to the real dice throws. (For example it is permissible to precalculate and presort all 216 possible throws of 3 dice and select one of these at random.)

Due to the real time nature of the application a full program is required, with input from stdin (enabling the user to run quick fire battles by reviewing the output and pressing the return key.)

Formatting of output strings and newlines shall be exactly as described above. Up to 2 additional symbols (but not alphanumerics) are acceptable between and around numbers. For example [3,2,1] is an acceptable way of displaying the roll of 3 dice.


This is code golf. Shortest code in bytes wins.

(example output to be added)

  • \$\begingroup\$ I don't see the point of requiring termination on a nonempty string. \$\endgroup\$ – lirtosiast Jan 18 '16 at 3:18
  • \$\begingroup\$ @ThomasKwa An attacker doesn't always want to fight to the death, especially if he has a run of bad luck with the dice. The user experience is designed to make attacking as easy as possible but there has to be a way to terminate. The alternative would be ctrl z or ctrl c which is ugly. If you mean you don't see the point of termination on the first iteration, well that is partly to enable implementation with a while loop but mainly to enable the attacker to change his mind right up to the last minute. \$\endgroup\$ – Level River St Jan 18 '16 at 8:21
  • \$\begingroup\$ Oh, okay. I wasn't used to seeing practical concerns on code golf questions. \$\endgroup\$ – lirtosiast Jan 18 '16 at 17:51
  • \$\begingroup\$ Your description of the number of dice rolled by the attacker doesn't seem to match the official rules. You say "The attacker rolls 3 dice (only 2 if he has only 2 armies)", but the rules say "You, the attacker, will roll 1,2 or 3 red dice: You must have at least one more army in your territory than the number of dice you roll". (There's also the subtle issue of choosing to roll fewer dice than the maximum because you want to keep more than one army behind). \$\endgroup\$ – Peter Taylor Jan 23 '16 at 19:16

Help find Mersenne primes!

There has recently been a discovery of a new prime number: 2^74207281-1. This is the biggest prime number to date and broke the previous 3-year record holder by over 4 million digits!

Your job will be to help mathematicians find some prime numbers (not really). You must take in an integer N and output the Nth Mersenne prime (OEIS A000668). You may assume that the Nth Mersenne prime is under your languages maximum integer number and/or will not cause an overflow (but your code should work for higher numbers if your language allowed it. You may not use any built ins for primality testing and cannot hard-code any values.

You can find a list of most of these numbers over here.

Test Cases

2 -> 7
5 -> 8191
8 -> 2147483647

This is , shotest code in bytes wins.

  • \$\begingroup\$ Please don't call it GIMPS XD. \$\endgroup\$ – Conor O'Brien Jan 21 '16 at 4:51
  • 1
    \$\begingroup\$ 1. Yes, that's the wrong sequence. It's actually A000668. 2. The current wording of the overflow assumption seems to allow hard-coding 4 values if your language operates purely on bytes. I also see no restriction on built-in primality testing. \$\endgroup\$ – Peter Taylor Jan 21 '16 at 9:15
  • \$\begingroup\$ I've been thinking about "code-golf or fastest-code". As code-golf it's a loop with multiplication by two and a primality test, and primality testing has been done to death. As fastest-code, it would pretty much be a case of simplifying the GIMPS client to something which could be posted. I suggest that you make it a compromise: code-golf, but with a speed constraint which rules out naïve primality testing. Maybe write a Lucas-Lehmer test in a slow language, take the highest value it can do in 6 seconds, and require answers to reach that value in 1 minute. \$\endgroup\$ – Peter Taylor Jan 21 '16 at 21:15
  • \$\begingroup\$ Somehow failed to find it when I searched before, but this question is related. \$\endgroup\$ – Peter Taylor Jan 24 '16 at 23:08

Parse an "Efficient" Encoding

Let us define a fictitious encoding "ENCWID", that follows this general form: there are three bits that denote the length of the following character, for each character. This looks something like this:

   ^^^^--- the actual character
^^^------- width bits; from 000 to 111

Perhaps this is a little vague. To understand this better, let us encode the string "Hello!" into ENCWID. Observe:

H   72   1001000
e   101  1100101
l   108  1101100
o   111  1101111
!   33   100001

This diagrams the binary ASCII codes of each character in the string. Let us put these values into an array that represents "Hello!": [1001000,1100101,1100101,1100101,1101111,100001]. The widths for each of these strings are 7, 7, 7, 7, 7, and 6 respectively. To binary, this makes 7 111 and 6 110. Now, we put the binary length of each binary ASCII code before the actual ASCII code, and combine them all into a single string, as such:

111-1001000 111-1100101 111-1101100 111-1101100 111-1101111 110-100001
(7)-(  H  ) (7)-(  e  ) (7)-(  l  ) (7)-(  l  ) (7)-(  ᴏ  ) (6)-(  !  )

(Spaces, hyphens, and parentheses added for visual clarity.)

And thus, the encoding of "Hello!" is 11110010001111100101111110110011111011001111101111110100001.

Decoding the string form is perhaps rather easy, using the following steps:

  1. Read three characters; call this N.
  2. Set N to the decimal number represented by N, from binary.
  3. Read the next N charcters; call this S.
  4. Parse S as a binary number, and append this character to the result.
  5. If there are still unread characters, go to step 1. Otherwise, continue.
  6. Return the result.

Objective Your objective is to write two programs; one that encodes and one the decodes the described encoding. Your score is the sum of the program's byte count.


Suppose that we can actually implement this in mainstream use; then, an encoding can be used that utilizes the lower-width codepoints for the most common letters in a given language. This would allow for a lesser amount of bits to convey the same message.

Say we did this for English, and that this

zI'Vq/Ux-[]jZ"  ;:QX&!

is the approximate frequency at which english letters occur, this could be our (partial) encoding:

   0  1  2  3  4  5  6  7  8  9  A  B  C  D  E  F
0     e  a  r  n  i  o  l  1  t  .  9  h  0  s  ,
1  d  C  (  )  u  2  S  c  m  y  8  g  B  J  D  W
2  7  H  v  M  6  R  b  k  A  4  3  5  w  f  L  P
3  G  p  T  K  E  F  N  Y  O  \n z  I  '  V  q  /
4  U  x  -  [  ]  j  Z  "  \t ;  :  Q  X  &  !  \

Thus, "Hello!" would be encoded as:

H   12   18   10010     (5 => 101)
e   01   01   1         (1 => 001)
l   07   07   111       (3 => 011)
o   06   06   110       (3 => 011)
!   4E   78   1001110   (7 => 111)

101-10010 001-1  011-111 011-111 011-110 111-1001110
(5)-( H ) (1)(e) (3)-(l) (3)-(l) (3)-(o) (7)-(  !  )

=> 1011001000110111110111110111101111001110
(²7ß{Î, under ISO-8859-7 encoding)

And, "Hello!" fits into a clean 5 bytes under this encoding, as opposed to 6-byte standard encoding. This would equate to a lot more bytes saved for a higher sample size.



  • \$\begingroup\$ 1. The introduction talks in terms of bits, but the description of decoding talks in terms of characters. Which is it? 2. If it's bits, how is padding to an exact multiple of 8 bits handled? (Or to an exact multiple of some other word size, if the storage/transmission model isn't based on octets). 3. With respect to the implications, see Huffman encoding and arithmetic encoding, which do the same thing better. \$\endgroup\$ – Peter Taylor Jan 23 '16 at 18:56
  • \$\begingroup\$ @PeterTaylor 1. Bits, I will revise. 2. I don't know much about this stuff; this really isn't an implementation as it is an interpretation. 3. I didn't claim that this was the best way. >_<. I know of both mentioned. \$\endgroup\$ – Conor O'Brien Jan 23 '16 at 19:02

Language Guessing

(Inspired by What's the Language?)

According to the first Google result, the most popular programming languages are:

  1. Java
  2. C++
  3. C#
  4. Python
  5. PHP

(For our purposes, C and C++ will both be treated as C++, though the test cases may include C source code and header files)

Your Task

Write a function that accepts a String, determines what programming language the String is, and returns its result.


  • The input String will be in one of the five languages listed above.
  • Your method should return a String with the name of the programming language exactly as written above.
  • Your code must be under 150 bytes.
  • No compiling, running, or otherwise evaluating the test cases to determine their language.
  • No standard loopholes.


  • The code with the highest match percentage wins
  • For every 15 bytes under the 150 byte limit, you get an extra percentage point (a 61-75 byte program would get 5% extra on top of the match percentage.
  • The method or function header does not count toward the total byte count. Return statements do count, however.
  • The code used to load the test cases and feed them into your method does not count either.
  • You get one import for "free". Any other imports count toward the byte limit.

Test Cases and Testing Code

Here is sample java code used for testing submissions. If you write in a language besides java, your tester must only accept one parameter, and cannot access any variables that are stored between iterations of the function. The entire code, including the tester, must be included in your answer.

import java.util.*;
import java.io.*;

public class Tester{

    static int right = 0;//How many are right
    static int total = 0;//How many tested in all

    public static void main(String[] args){

        //Opens up test cases
        File rootFolder = new File("./Test Cases");
        File[] subFolders = rootFolder.listFiles();

        HashMap<File, String> fileList = new HashMap<>();

        //Add all files in a subfolder to a HashMap and associate them with the language
        for(File currentFolder : subFolders){
                for(File currentFile : currentFolder.listFiles()){
                    fileList.put(currentFile, currentFolder.getName());

        //Iterate through files
        fileList.forEach((File f,String lang)->{
            System.out.print("Testing " + f.getName()+"\t");

                //Get contents of file
                BufferedInputStream input = new BufferedInputStream( new FileInputStream(f));
                Scanner sc = new Scanner(input);
                String contents = sc.next();

                //Call compute method and store result
                String result = compute(contents);

                //Increment counters

                System.out.println("Guess: " + result +"\tActual: " + lang);
            } catch(IOException e){
                System.out.println("There was an error when reading the file");

        //Print out final results
        System.out.println("Right: " + right);
        System.out.println("Total: " + total);
        System.out.println("Percent: " + (double)right * 100.0 / total + "%");

    public static String compute(String c){
        //YOUR CODE HERE

The test cases consist of 500 code samples (100 from each language) from around the internet (mostly github). Specific locations can be found in Credit.txt in the Test Cases folder.

The test cases can be downloaded here.

  • \$\begingroup\$ If you don't want to limit it to Java, don't enforce the submission tester. Most languages aren't too hard to test, with the exception of TinyMUSH. \$\endgroup\$ – Rɪᴋᴇʀ Jan 24 '16 at 1:16
  • \$\begingroup\$ @RikerW I made the wording a bit more language neutral. \$\endgroup\$ – Daniel M. Jan 24 '16 at 1:27
  • \$\begingroup\$ Why do we need a tester at all? Can't we just test it manually? \$\endgroup\$ – Rɪᴋᴇʀ Jan 24 '16 at 1:30
  • \$\begingroup\$ How else would you run through a few hundred test cases \$\endgroup\$ – Daniel M. Jan 24 '16 at 1:32
  • \$\begingroup\$ @Downgoat's was a special case, normally there aren't that many test cases. There should only be like 10 different cases, and that should cover most of it. \$\endgroup\$ – Rɪᴋᴇʀ Jan 24 '16 at 1:33
  • \$\begingroup\$ The number of test cases will vary widely with the challenge. Depending on the length of the code samples in this challenge, I can imagine a fairly large number of test cases being used to give a demanding challenge. \$\endgroup\$ – trichoplax Jan 28 '16 at 18:52
  • \$\begingroup\$ If the test cases are provided as a list of inputs and a list of outputs, would that allow for easy testing in any language? Simply automate the running of the candidate answer against every input, giving a list of outputs, and then compare that with the target list of outputs. \$\endgroup\$ – trichoplax Jan 28 '16 at 18:54
  • \$\begingroup\$ The test cases are a bunch of files of each programming language. They are sorted into subfolders with the name of the language the example is written in. The tester loads each file, gets the language based on the subfolder's name, puts the whole file into a string, and feeds it to the method. It compares the returned value to the subfolder's name. This way, to add or change test cases, it's a matter of dropping additional files into a folder, or dropping in a folder to add a new language. \$\endgroup\$ – Daniel M. Jan 28 '16 at 19:03

Pixel "Density"

In this challenge, we will be viewing each pixel in an image as a particle with two properties:

(1) vertical density (vd)

(2) horizontal density (hd)

These two properties are determined by the RGB values of the pixel in the following manner:

vd = Red - Blue

hd = 128 - Green

We call a sorted image an image that satisfies the following conditions:

For each pixel:

  1. no pixel to the right has a lower hd

  2. no pixel to the left has a higher hd

  3. no pixel above has a higher vd

  4. no pixel below has a lower hd

The challenge is to produce such sorted images.


An image in any "common, recognizable" format, via file or stdin.


A sorted image in a "common, recognizable" format (not necessarily the same format as input) containing exactly the pixels of the original , via file or stdout.

Relevant Details:

  • This is .

  • alpha values (if they exist) are ignored.

  • Libraries may be used, but size must be added to your score

  • "common, recognizable" formats will be determined by the community. I don't expect this to be an issue, but if you feel the the loose wording is being exploited, downvote the answer (or even better, help me fix it before the challenge goes live. I don't really want to enumerate the allowed formats, but maybe that is a better option?).

If there is interest in the question, I will create a reference implementation as well as test cases.

  • 1
    \$\begingroup\$ I don't think this specific task is a dupe - the closest that comes to mind involves doing the same transformation but on colours represented as hex strings rather than on the actual bytes. However, this is pretty much "Look up one value i in an array; then loop applying A[i++]^=255 until the end of the array. Adding alpha makes it slightly less boring, but not much. \$\endgroup\$ – Peter Taylor Jan 23 '16 at 13:29
  • \$\begingroup\$ @PeterTaylor You were right about it being extremely boring, I've reworked the challenge considerably. I hope this is more interesting. \$\endgroup\$ – Liam Jan 23 '16 at 23:28
  • \$\begingroup\$ There seem to be 2 separate tasks here: 1. the pixel sorting. 2. processing bmp files. Do you want the file processing to be part of the same challenge? Otherwise you could allow any recognised freely available image format. For example, I would expect code to process a ppm file to be considerably shorter than code to process a bmp file. \$\endgroup\$ – trichoplax Jan 28 '16 at 18:35
  • \$\begingroup\$ Relevant meta answer \$\endgroup\$ – trichoplax Jan 28 '16 at 18:45
  • 1
    \$\begingroup\$ @trichoplax good point. My original thinking was to discourage the use of libraries so I wanted to use the simplest format I could think of, which is bmp. I'll make changes. \$\endgroup\$ – Liam Jan 28 '16 at 18:48
  • \$\begingroup\$ changes made (for future readers of these comments) \$\endgroup\$ – Liam Jan 28 '16 at 23:52

Windows batch polyglot/hybrids challenge with any language possible /* I need some help to refine my challenge */

As the batch scripts are pretty limited, hybrid files that embed a code from another language into a batch file are used a lot lately.

Though usually there are some requirements for a good hybrid script:

  1. The embedded code should be usable as-it-is - without any ugly batch escaping sequences.
  2. There should not be redundant output. E.g. a lot of languages use /* */ for multi-line comments. If the batch script executes a line that starts with / * it will print an error message an will continue to the next line .Though it will allow you to hybridize C/C++/C#/Java/... with a batch file the error message cannot be surpassed so this will not count as a good hybrid.
  3. No temp files.It's easy to output a lot of code into a temp file that will be later executed , the IO operations will slow the script and on of the main advantages of batch scripting (the speed) will be lost. And more over is a challenging constraint. But this will be not possible for compiling languages and for extension sensitive languages.

Some examples will follow:

JScript (good for example as it comes with every windows installation) technique invented somewhere in the link by Tom Lavedas :

@if (true == false) @end /*
@echo off
cscript //nologo //e:javascript "%~dpnx0" %*
echo Output from the batch.
goto :EOF */

WScript.Echo('Output from the jscript.');

The output will be:

Output from the jscript.

Output from the batch.

The technique uses the JScript (javascript has no such thing) specific @ directives to make the code valid (or silent) both for both languages.

Another example (again with JScript) invented by Ryan Biesemeyer:

0</* :
@echo off
echo Output from the batch.
cscript /nologo /E:jscript %~f0 %*
exit /b %errorlevel%

WScript.Echo('Output from the jscript.');

This time is used the redirection priority in batch scripts and 0</* : will be parsed as 0:</* .

Here are some info that can help you:

  • every line that starts with @ will be silent - even the invalid commands.
  • every line that starts with : will be taken as label in batch and will be not executed
  • every line that starts with something like <someWord : someWord will be silent because of redirection priority
  • every line that starts with something like digit<someWord : someWord will be silent because of redirection priority (this time the output will be redirected to a stream).In this case will be best to use 0
  • you can use <space><tab> ; , = at the beginning of every line - these will be ignored as standard delimiters in batch.Can be useful if some language use some of these as comment.
  • if you start a line with %not_existing_variable% it will be replaced with nothing.Could be useful if in some language comments start with percent.
  • If you finish a line with a caret ^ the next line will be appended at the end.With the caret in batch you can escape the new line.

here's a little bit more inspiration

And here's the challenge. Your script should be with .bat or .cmd extension . It should contain a code from another language - the code should be used by the second language as it is (no escapes symbols) .REPL tools are accepted with the condition of no escape symbols - except in the line where the REPL tool is invoked. There should not be redundant output. There should be not temp files (with the exception of compiling languages , and file extension sensitive languages - then the file should copy itself in the %TEMP% folder).

Each script should accept one command line argument which will be printed 5 (five) times . Both from the batch part of the code and from the second language prefixed by the language name (for the batch case it should be BATCH: for example.If the second language is Ruby it should be RUBY:)

For sure this will be not possible with every language , but for many it is achievable.

The winner should propose the most solutions with different languages (or if the language is the same a different kind of techniques should be used like in the two examples above).In case of equal number of solutions the first one wins.


Weekly Physics Golf #{TBD}: Wavelength of a Sound Wave

META NOTE: I believe, and others in the comments, that this challenge is quite trivial. So, I will be working on making this challenge harder in the next few days.

Introduction to the Series

Every week I will be posting a physics challenge. My goal here is to design challenges that in the end, teach some people some physics. Overall, the challenges will be very basic with little information. All of these challenges will have the minimal information necessary to solve them, and the goal is for users like you to do some research, watch some videos, and understand how these concepts work to teach you how to approach these types of physics problems and explain how they work. Of course, I will also give two optional hints per challenge, which are there if you do not have the time or determination to do the research, or you cannot figure out how to do the problem after researching. The two hints will be "necessary equations for this challenge" and "process to solve the problem". The hints are completely optional to use and it is encourages to not use them, but as stated above to learn the information for yourself. The series will have one main leaderboard. Whoever has the least combined byte count for all of the challenges gets a to be determined prize. Each challenge will range in difficulty, with an upwards trend of difficulty. I wish you all luck and I hope you learn a thing or two!

Challenge #{TBD}: Wavelength of a Sound Wave (Difficulty: EASY)

Many people may not know that sound is actually a wave, which would explain why it can go through objects and bend around corners. But it has some really interesting properties. For example, the speed of the sound wave changes from a few factors, all relating to the medium it travels through. In this challenge, you will have to figure out the wavelength of a sound wave through a given medium.

But first, lets define a medium. A medium is any liquid, solid, or gas that a wave can travel through. Mediums can affect the speed of a sound wave in two ways: the density of the medium, and the bulk modulus elasticity of the medium. But what are these two things? Well lets first consider the density of an object. Density is defined to be the mass divided volume of the object, or in simple terms, amount of mass in a given space. This is important, as waves travel differently through mediums of different densities. The reason behind this is that a medium with a higher density has more inertia (the resistance to change in motion). An object with a higher inertia will be more difficult to move with a wave (which displaces particles to move). Therefore, INCREASED DENSITY corresponds to DECREASED WAVE SPEED.

Now lets define bulk modulus elasticity. This is really just a fancy name for stiffness of an object. But what does this have to do with anything? Well, when an object is stiffer, each molecule is more interconnected to other molecules. And because a wave displaces particles to "move", it will move through a stiff medium faster because it can move larger molecules, hence moving more distance in a shorter time, a.k.a. moving faster. Therefore, INCREASED BULK MODULUS ELASTICITY corresponds to INCREASED WAVE SPEED.

So in this challenge, I will give you the density and the BME (bulk modulus elasticity) of the medium, and also the frequency of the sound wave. Your job is to use this information and output the wavelength. Here are the full specs:

  • Input will be three numbers (not necessarily integers). They correspond to density in kg/m^3, bulk modulus elasticity in Pascals, and sound wave frequency in hertz.
  • Input can be in any order, in any convenient format (so 12.3 45.6 78.9, [12.3 45.6 78.9], and 78.9,[12.3,45.6] are all acceptable).
  • You may assume that input will never cause any sort of error during execution.
  • Output will be the calculated frequency of the sound wave in meters, precise to three decimal places, omitting leading and trailing zeros. For example, inputs of 6, 27, and 3 would given an output of 0.707 after rounding from 0.70710678118654752440084436210485.
  • Shortest code in bytes wins!

Test cases

META NOTE: Test cases to come, I am working on them right now.


These hints are for those who do not want to put in the time and effort of research, or those who could not find a solution. So, here are the two hints:

Hint 1: Equations

You need the following equations for this challenge:

enter image description here

Hint 2: Sample Solution Process:

Using the two equations above, we can set them equal to each other to get:

enter image description here

Solving for wavelength gets us this equation:

enter image description here

From here you can just plug in values.

This is , so shortest code in bytes wins. Good luck!


Meta note: blah blah blah, working leaderboard will eventually go here! This leaderboard will contain and combine scores for all of the weekly challenges. It will only be visible on this question, though.

  • \$\begingroup\$ Please tell me of any suggestions, I'll be happy to hear them. More importantly, please alert me of any mathematical inaccuracies and/or inaccurate explanations. \$\endgroup\$ – GamrCorps Feb 2 '16 at 2:25
  • \$\begingroup\$ +1 for the idea. I love learning new things! \$\endgroup\$ – user81655 Feb 2 '16 at 12:20
  • \$\begingroup\$ But my concern with this challenge is that it's too simple and straight-forward. The shortest answer in every language would be something like (a,b,c)=>(b/a)**.5/c (except for Mathematica, it proabably has a built-in :P ). \$\endgroup\$ – user81655 Feb 2 '16 at 12:20
  • \$\begingroup\$ I concur with @user81655, but I'm excited to see what you can come up with. \$\endgroup\$ – lirtosiast Feb 2 '16 at 18:05
  • \$\begingroup\$ @user81655 Yes, after further analysis I have realized how simple the problem is. I am currently working on an alternative challenge while I make this one more difficult. \$\endgroup\$ – GamrCorps Feb 2 '16 at 21:55

I've posted two well-received (by my standards: 10+ no downvotes) questions in the past; however, I've always missed a bit of information, so I welcome all feedback to this challenge!


A palindrome is a word or string of words that is the same forwards as it is backwards, e.g. taco cat, kayak, madam im adam. I'm sure by now you're all familiar with .

An anagram is a word or group of words that when the letters are rearranged, forms a different word or string of words. If you haven't already, Google for anagram or define anagram (not part of the puzzle, obviously).


Your task is to write a program which will take a string of lowercase characters and blank space, and return truthy/falsey if the string can be anagrammed into (or is) a palindrome (even if it doesn't form a legit word). This is so shortest answer wins. Standard loopholes apply.


  • A string of at least 1 character containing a mixture of lowercase alphanumeric characters and blank space
  • Whitespace can be ignored when determining whether a word or phrase can be anagrammed.



kayak = kayak / akyka
maam = maam / amma
her taco cat hero = hertacoocatreh
boob = obbo
boobs = bosob
ooh = oho
lol = lol (already a palindrome)
zzz = zzz (already a palindrome)
i = i
00010001 = 10000001
0xd00000d = 0d00x00d0
sevens 7even = seven7neves
glum glow worm
maam im adam
nana ant

  • 1
    \$\begingroup\$ What whitespace characters can appear in the input? And the rule that "anagrammed into a palindrome" requires the palindrome to be distinct from the original string, even if that were itself a palindrome, should be mentioned in the task description, not hidden in the examples. \$\endgroup\$ – Peter Taylor Feb 3 '16 at 14:26
  • \$\begingroup\$ Thanks for your feedback: I removed the unnecessary "palindrome must be unique from original" requirement, and clarified whitespace. \$\endgroup\$ – Tas Feb 3 '16 at 21:12
  • \$\begingroup\$ I actually thought that the requirement to be different to the original string, while needing to be stated clearly because it was unexpected, made the question more interesting. Without that requirement the test is just that there's at most one non-whitespace character which occurs an odd number of times. Although I suppose that with a permutation operator the difference isn't that big. \$\endgroup\$ – Peter Taylor Feb 3 '16 at 21:24
  • \$\begingroup\$ Not quite: na na na na can be anagrammed into the palindrome nanaanan and it only contains characters with even counts. \$\endgroup\$ – Tas Feb 3 '16 at 21:47
  • 1
    \$\begingroup\$ I said "at most one", not "exactly one". \$\endgroup\$ – Peter Taylor Feb 3 '16 at 22:27

Maximum cover time ratio

Given an undirected graph and a start node, there is an expected number of steps to reach all possible nodes if you walk at random. This expected number of steps will depend on which node you start from. For a given graph and starting node, let us call the number of steps to visit all nodes in the graph the cover time. We can estimate the cover time by just setting off 1000 walks and taking the average.

You can also set off two independent random walks from the same node at once and see how many steps it takes for every node to be visited by at least one of the walks. Clearly you can do this with more than two as well. Let us call the expected number of steps to reach each node with at least one of k random walks starting from the same node, the k-cover time. We can see that cover time = 1-cover time.


Write code that explores different undirected graphs, starting nodes and values of k and computes both the cover time and the k-cover time for each. You should use at least 1000 random walks to estimate the time to reach all nodes. The goal is to maximize the expected cover time ratio explained below.


Your code should output a single graph in any sensible format you choose, the value k, the identity of a starting node, the cover time and the k-cover time.


Your score will be cover time from your given starting node divided by k times k-cover time starting from the same node.

Languages and libraries

You can use any language or library you like (that wasn't designed for this challenge). However, for the purposes of scoring I will run your code on my machine so please provide clear instructions for how to run it on Ubuntu.

My Machine The timings will be run on my machine. This is a standard Ubuntu install on an 8GB AMD FX-8350 Eight-Core Processor. This also means I need to be able to run your code.


  • I will kill your code after 2 minutes unless it starts to run out of memory before then. Your code should therefore output something before 2 minutes is up.

Any help with finishing this question would be gratefully received. I think it just needs examples and test cases and maybe a picture or two.

  • \$\begingroup\$ Shouldn't the k-cover time be the same as the cover time by the central limit theorem? \$\endgroup\$ – Peter Taylor Feb 4 '16 at 16:26
  • \$\begingroup\$ @PeterTaylor Maybe my definition is confusing but you can explore a city more quickly with two people rather than one. Similarly, you would expect to explore a graph more quickly with two random walks rather than one. The number of steps is measured until every node has been visited by at least one random walk. \$\endgroup\$ – user9206 Feb 4 '16 at 16:39
  • \$\begingroup\$ Aha. I think the at least one belongs in the definitional statement: "expected number of steps to reach each node with at least one of k random walks starting from the same node". Have you done any testing on small cases to see whether there are easy unbounded solutions or easy bounds? \$\endgroup\$ – Peter Taylor Feb 4 '16 at 19:57
  • \$\begingroup\$ What's the input to the program going to be? \$\endgroup\$ – feersum Feb 5 '16 at 3:14
  • \$\begingroup\$ @feersum There wasn't going to be any input in a similar fashion to codegolf.stackexchange.com/questions/65876/… . \$\endgroup\$ – user9206 Feb 5 '16 at 6:17
  • \$\begingroup\$ I'm pretty sure you can actually solve this problem analytically as questions like those are known under the name Markov Chains. You can e.g. calculate the expected number of steps from A to B by making B absorbing, and calculate the expected number of steps to absorbtion. \$\endgroup\$ – flawr Feb 5 '16 at 21:45
  • \$\begingroup\$ @flawr That would be very interesting. As I can't I hope an expert can comment on this. \$\endgroup\$ – user9206 Feb 5 '16 at 21:47
  • 1
    \$\begingroup\$ I'm not very fluent anymore but I'm going to reread some of this which I found very helpful in the past. \$\endgroup\$ – flawr Feb 5 '16 at 22:22

Show me your OOP

Your task is to implement OOP that have this feature:

  1. Inheritance (Single dispatch)
  2. Class

Your submission have to translate this pseudo-code into your languages:

class Animal{
    Animal(string name){
        printLn("An animal has summoned " + name);
    public void sound();
class Sheep extends Animal{
    bool woolstate;
    Sheep(bool state){
        woolstate = state;
    public void sound(){
    public void growWool(){
        woolstate = true;
    public void getWool(){
        if(woolstate == true){
            printLn("Mbaaaa\nget " + (new Math.Random()).getInt(0, 10).toString() + " bags of wool");
            woolstate = false;
class Worm(){
    Worm tails;
    public int numbers;
    public Worm(int number, int length){
        if(length != 0){
            tails = new Worm(number, length-1);
        numbers = number;
        if(number == length){
    public Worm getTails(){
        return this.tails;
    public Worm setTails(Worm tail){
        this.tails = tail
    public Worm getNumbers(){
        return this.numbers;
    public Worm setNumbers(int number){
        this.numbers = number
    public void sound(){
        printLn("Doesn't sound");
    private void privateMethod(){
        printLn("This is a private method");
int main(){
    Sheep a = new Sheep(true);
    Animal b = new Sheep(false);
    Worm c = new Worm(5, 5);
    b.getWool(); //Error
    b.setNumber(7); //Error

If the random function is not available, you can deterministically return anything from 0 into 10.

The submission should be scored based on:

  1. Difficulty of implementation (i.e. how far is the language from OOP?)
  2. Ease of usage of implemetation
  3. Efficiency of implementation
  4. The pleasantness of syntax
  • \$\begingroup\$ This could be a potential challenge! However, it could use more objective criteria, as popularity-contests are now generally frowned upon here. For example, what prevents me from posting a 0-byte C#/Java/C++/etc. answer, and claiming that it implements OOP? (Also, some languages don't have a random feature for the Sheep.getWool() function.) Also, we apparently need an interface feature as well, for Animal.sound(), and is the Sheep constructor supposed to call the Animal constructor? If so, you forgot to call it with a name. \$\endgroup\$ – LegionMammal978 Feb 7 '16 at 12:42
  • \$\begingroup\$ @LegionMammal978 What is your suggestion \$\endgroup\$ – Akangka Feb 8 '16 at 11:58

Continue the sequence

As we all know, there are a lot of integer sequences in the world. What if we could create one program to figure out the next few items in any given sequence? That would be pretty cool, right? Well, I'm pretty sure we can't create a program that does every sequence, but we can at least do most of the simple ones.


Your challenge is to create a program or function that takes in 5 integers in any reasonable format (array, string, separate arguments, etc.), and returns/outputs the next 5 integers in the sequence.


  • Your submission must be deterministic; that is, the output should be the same every time for the same input.
  • You may not use a built-in for determining the next items. (cough Mathematica cough)


To score a submission:

  1. Take the value of each single output integer minus the expected output.
  2. Take the square root of the absolute value of each of these.
  3. Take the average of the result and add one.
  4. Multiply your byte count by this number.

For example, if the only test case were 1,2,3,4,5 => 6,7,8,9,10, and your 15-byte submission outputs 6,7,8,10,6, your score would be:

1,2. abs(6 - 6) = 0, sqrt(0) = 0
     abs(7 - 7) = 0, sqrt(0) = 0
     abs(8 - 8) = 0, sqrt(0) = 0
     abs(10 - 9) = 1, sqrt(1) = 1
     abs(7 - 10) = 4, sqrt(4) = 2
3.   0 + 0 + 0 + 1 + 2 = 3
     3 / 5 = 0.6; 0.6 + 1 = 1.6
4.   15 bytes * 1.6 = 24

Thus, your score would be 24.

Lowest score wins.

Test cases

2,2,2,2,2 => 2,2,2,2,2
1,2,3,4,5 => 6,7,8,9,10
109,117,125,133,141 => 149,157,165,173,181
1,2,4,8,16 => 32,64,128,256,512
1,10,100,1000,10000 => 100000,1000000,10000000,100000000,1000000000
1,2,3,5,8 => 13,21,34,55,89
2,1,3,4,7 => 11,18,29,47,76
1,3,6,10,15 => 21,28,36,45,55
1,4,9,16,25 => 36,49,64,81,100
8,7,6,5,4 => 3,2,1,0,-1
1,0,1,0,1 => 0,1,0,1,0
1,2,3,2,1 => 2,3,2,1,2
512,256,128,64,32 => 16,8,4,2,1
1,11,111,1111,11111 => 111111,1111111,11111111,111111111,1111111111
1,21,321,4321,54321 => 654321,7654321,87654321,987654321,10987654321
1,2,6,24,120 => 720,5040,40320,362880,3628800

I will create a GitHub Gist with all of the test cases before posting.

Sandbox questions

Currently, I'm looking for suggestions for:

  • a better name
  • more tags
  • more sequences

Feel free to post any other questions/notes you have.

1 This will be replaced with the actual number of test cases when this is posted.

  • \$\begingroup\$ Why the 100 byte limit? \$\endgroup\$ – nimi Feb 10 '16 at 20:45
  • \$\begingroup\$ @nimi The goal is to correctly predict as many sequences as possible while keeping the code fairly short. Does 100 bytes sound too short? \$\endgroup\$ – ETHproductions Feb 10 '16 at 20:47
  • 2
    \$\begingroup\$ ... well, 100 bytes of Pyth aren't 100 bytes of Java. Wordy languages will be quite handicapped. \$\endgroup\$ – nimi Feb 10 '16 at 20:51
  • \$\begingroup\$ @nimi True. How might I be able to fix this? \$\endgroup\$ – ETHproductions Feb 10 '16 at 20:51
  • \$\begingroup\$ Classical bonuses (although discouraged now) work better when relative compared to absolute. Maybe something like byte count * (2 - percentage solved), i.e. 2x bytes if 0%, 1x bytes if 100%, lowest score wins. But somehow we must prevent trivial answers like outputting the constant string 0 0 0 0 0 for 2x9 = 18 bytes. \$\endgroup\$ – nimi Feb 10 '16 at 21:00
  • 5
    \$\begingroup\$ Sequences are pretty arbitrary, so I'm having trouble seeing how this could go well. I get the feeling most submissions will just end up hardcoding in a set number of patterns, eg geometric progression, polynomial interpolation, and ignore ones which are too complex, eg Golomb. Also, standard loophole, but it might be worth mentioning explicitly that looking up things on, eg OEIS, is not okay. \$\endgroup\$ – Sp3000 Feb 10 '16 at 21:48
  • \$\begingroup\$ @Sp3000 You have a good point there. What would you think about A) choosing a select number of sequence types, B) choosing a select number of types and making it code-golf, or C) just choosing a single type to focus on? \$\endgroup\$ – ETHproductions Feb 11 '16 at 16:11
  • \$\begingroup\$ I had a think about this, and I think what this challenge might need is 1) A fairly large test-battery covering a large variety of sequence types (e.g. OEIS), and 2) A tweak to the scoring system which takes into account how far off a number is (since getting 0 points for a guess that's just off by 1 seems a bit harsh). Just a thought though. \$\endgroup\$ – Sp3000 Feb 14 '16 at 4:11
  • \$\begingroup\$ Regarding your options, I'm not a great fan of C since it changes the challenge completely, and might need dupe checking. There's no real need to limit the number of sequence types I think, as long as there's an incentive to be close with the guess rather than dead on. \$\endgroup\$ – Sp3000 Feb 14 '16 at 4:12
  • \$\begingroup\$ Perhaps make it a [test-battery] with lots of test cases. That seems like the best way to score these types of languages. You may also want to disallow built-ins as I know Mathematica has a built-in to do this \$\endgroup\$ – Downgoat Feb 15 '16 at 17:09
  • \$\begingroup\$ @Sp3000 Yes, those are some very good thoughts. This is now a test-battery challenge, and the new scoring algorithm takes into account both the byte-count and the average of the sqrt-distances of the outputs. Does this seem like a decent way to score this challenge? \$\endgroup\$ – ETHproductions Feb 15 '16 at 18:02
  • \$\begingroup\$ @Downgoat This is now a test-battery challenge, and built-ins have been banned. Thanks! \$\endgroup\$ – ETHproductions Feb 15 '16 at 18:02
  • \$\begingroup\$ @nimi The scoring system has been completely rewritten. \$\endgroup\$ – ETHproductions Feb 15 '16 at 18:06
  • \$\begingroup\$ With respect to limiting the types of sequence, you need to be careful not to turn it into a dupe of codegolf.stackexchange.com/q/3485/194 \$\endgroup\$ – Peter Taylor Feb 15 '16 at 19:09
  • \$\begingroup\$ @PeterTaylor Thanks, I'll keep that in mind. \$\endgroup\$ – ETHproductions Feb 15 '16 at 19:48

Compress by Replacement

You are to take an input string, and then turn it into another string with a list of replacements which would turn this new string into the input string. The catch is that you must output the smallest such alternate string + list of replacements you can.

The input string will be restricted to the following characters (Please note there is a space at the end and that semicolons are not allowed):


The way the output is formatted is


An example output is 111222;1;abcd;2;efgh which outputs abcdabcdabcdefghefghefgh

The replacement rules are applied in the order they are shown in the output to avoid ambiguity.

Therefore, if one were to input abcdabcdabcdefghefghefgh (which has a length of 24). Then the output would be 111222;1;abcd;2;efgh with a length of 20 which is the shortest possible output for this input

Some other notes:

  • The replacement rules are restricted to the same character set that the input has. This means that if the input contained every possible letter except for the character h, then your replacement rule would only work if the tofind part of the rule is h
  • There may be situations where the input string is the shortest possible output, this is fine
  • The program doesn't have to be computationally efficient, as long as it does calculate the correct answer when it finishes, then it is valid
  • There are also occasionally multiple correct answers, any solution which has the smallest possible count is valid and you only need to output one solution, not all the valid solutions
  • The semicolons are included in the character count for the output

Some test cases:

'abcdabcdabcdabcdefghefghefghefgh' -> '11110000;1;abcd;0;efgh' or '00001111;1;efgh;0;abcd'
'this is a test. this is a test.' -> '0 0;0;this is a test.'
'there is no shortening in this one' -> 'there is no shortening in this one'
'01001001010100101010110101010101010110101010' -> '202030321332130;3;222;2;01' or '22331131210113;3;212;2;010' or '22212233230132;3;121;2;010' or '223233133301132;3;21;2;010' or '222323133310133;3;12;2;010' or '020203031332132;3;222;2;10'

This is codegolf, so shortest answer in bytes wins

Sandbox notes:

  • This takes a long time to brute force for anything longer than 40 characters with repeated substrings, so it may take a while for brute forced solutions to verify they are correct. Because of this, I was thinking this question might be better proposed as a "best algorithm" kind of puzzle to see who can make the best algorithm. What do you think?
  • Should I allow people to choose what delimiter is used to separate the output? Ideally what would happen is that they would just say 'I am assuming that , is not used and it is my delimeter for the output'.
  • Am I being too strict on the output? The aim was that you are trying to compress it only via replacements, and as a result want to output those replacements as short as you can, which this format is the smallest you can make.
  • 1
    \$\begingroup\$ 1. I don't understand what you mean by "The replacement rules are restricted to the same character set that the input has", even with the explanation. 2. A simplified version of this problem is NP-complete, so I'm not sure what "best algorithm" winning criterion you could use. 3. It would be good to have a test case where all optimal answers have a tofind which is more than one character (ensuring that answers solve the problem stated, not just that simplified version). \$\endgroup\$ – Peter Taylor Feb 17 '16 at 16:40
  • \$\begingroup\$ Seems possible if you limited to a couple replacements. Otherwise, maybe a code challenge/fastest code version? \$\endgroup\$ – Michael Klein Feb 17 '16 at 18:44
  • \$\begingroup\$ 1. What i mean is, that the rules: 1;abcd;2;efgh for example, must only contain characters which were allowed in the input. So you can't use weird unicode characters or things to replace 2. I'd not heard of the simplified version of this problem. In fact I think that this may be more efficient, because some rules which may be valid in that problem would not be valid due to having to factor in the additional length of the semicolons, and 3. I don't believe I have such a solution available, maybe none exist like that? \$\endgroup\$ – Cameron Aavik Feb 18 '16 at 4:17
  • \$\begingroup\$ I think algorithm is a no-go from Peter's response since it does seem to be NP complete, maybe I could restrict it to be at most 3 replacements as per Michael Klein's suggestion \$\endgroup\$ – Cameron Aavik Feb 18 '16 at 4:20

Wring the changes

I want to ring Stedman Doubles, but I'm lazy so I want to get to a particular change and back as quickly as possible.

Change-ringing is essentially an enumeration of permutations of a set of small integers. To make the enumeration ringable, no bell can move more than one step in each permutation. A change therefore consists of some or all of the bells swapping places with adjacent bells.

Consider the case of five bells. Before changes start, the bells are rung in ascending numerical order, known as Rounds, which would be 12345. A typical change would then consist of leaving one bell in the same place while a double swap occurs between the remaining four bells, thus the name "Doubles". The bell that remains in the same place serves to notate the change. Note that this refers to the bell currently in that position, rather than the bell with that number.

In the case of Stedman the full notation is If you repeat those 12 changes five times you will enumerate 60 changes before returning to Rounds. The first few changes are as follows:

[12345 Rounds]
 21354 (1,2 and 4,5 swap)
 23145 (1,3 and 5,4 swap; 2 was 1st so doesn't move)
 32415 (2,3 and 1,4 swap)
 23451 (3,2 and 1,5 swap; 4 was 3rd so doesn't move)

There are of course 120 permutations of five bells, so in order to access the other 60 changes, or simply for variety, Stedman allows for every sixth change to be altered by a call known as a Single, which causes the two bells currently in position 4 and 5 to stay in the same place instead of exchanging as they would normally do.

A touch is described by a sequence of letters describing whether each sixth change is Plain or a Single. It ends when it returns to Rounds, which need not be after a multiple of six changes, but it may not repeat any change, so 120 is the longest possible length of a touch of Stedman Doubles. For instance, a complete extent of all 120 changes could be rung using the touch PPPPPPPPPSPPPPPPPPPS.

Your challenge is to write a program or function that will calculate the shortest touch that includes a given change. Your input should be the change (you can expect it to be a valid change) as a string, and your output should be (in either order) the touch description (as a string of P and S characters) and also the number of changes (including the final rounds, so up to 120).

The score for this challenge shall be the length of your program, plus the lengths of the touches it finds for the following inputs:


The Batch script @echo 120 PPPPPPPPPSPPPPPPPPPS would therefore score 630.

  • \$\begingroup\$ This could use a little background information. Ideally, the challenge should be self-contained and understandable by somebody who does not know what Stedman Doubles or change ringing is (such as myself). \$\endgroup\$ – Dennis Feb 17 '16 at 17:04
  • \$\begingroup\$ So in essence, we have a highly structured digraph of 60 vertices (corresponding to the order-60 group generated by P and S, which I assume is A5 but haven't checked); each of the 120 edges is labelled with 12 permutations (and each permutation therefore occurs on 12 edges, but symmetry), and we need to find the shortest cycle from 12345 which includes an edge labelled with the input permutation \$\endgroup\$ – Peter Taylor Feb 18 '16 at 12:34
  • \$\begingroup\$ @PeterTaylor Close. You don't need a cycle as long as you reach the input permutation before you reach an edge labelled 12345 (or 12345 is labelled after the input on the same edge), and there are 4 generators as the meaning of P alternates between 42351 and 43152 and S between 42315 and 43125, so the group is generated by P1P2, P1S2, S1P2 and S1S2 (but I don't know what the group looks like, except that (P1P2)^5==I). \$\endgroup\$ – Neil Feb 18 '16 at 13:14

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .