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

Sandbox FAQ


To post to the sandbox, scroll to the bottom of this page 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, though you can optionally add a title at the top. 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, and replace the post here with a link to the challenge and delete the sandbox post.


The purpose of the sandbox is to give and receive feedback on posts. If you want to, feel free to give feedback to any posts you see here. Important things to comment about can include:

  • Parts of the challenge you found unclear
  • Comments addressing specific points mentioned in the proposal
  • Problems that could make the challenge uninteresting or unfit for the site

You don't need any qualifications to review sandbox posts. The target audience of most of these challenges is code golfers like you, so anything you find unclear will probably be unclear to others.

If you think one of your posts requires more feedback, but it's been ignored, you can ask for feedback in The Nineteenth Byte. It's not only allowed, but highly recommended! Be patient and try not to nag people though, you might have to ask multiple times.

It is recommended to leave your posts in the sandbox for at least several days, and until it receives upvotes and any feedback has been addressed.


Search the sandbox / Browse your pending proposals

The sandbox works best if you sort posts by active.

To add an inline tag to a proposal, use shortcut link syntax with a prefix: [tag:king-of-the-hill]. To search for posts with a certain tag, include the name in quotes: "king-of-the-hill".

  • \$\begingroup\$ What if I posted on the sandbox a long time ago and get no response? \$\endgroup\$
    – None1
    Commented May 15 at 14:05
  • \$\begingroup\$ @None1 If you don't get feedback for a while you can ask in the nineteenth byte \$\endgroup\$
    – mousetail
    Commented May 29 at 13:27

4722 Answers 4722

16 17
19 20

ASKEY robbers

You are a robber in the ASCII world. ASCII lock-key work in similar fashion as the real-world: matching ridges. Your objective is to write a program which generates a duplicate key for the ASCII locks.


A Lock is given as:

   |\    |\         
 __| \___| \____             
|               | 

Its key shall be something like:

|   __     __   |
|  |  \   |  \  |  
|__|   \__|   \_|

The fit being something like:

|   __     __   |
|  ||\\   ||\\  |
|__|| \\__|| \\_|
|               |


I think the fit is weird. It didn't go as I had imagined when I finished typing in the ASCII drawing.

  • 4
    \$\begingroup\$ This has potential, but it needs a clear specification as to what a fit is, and what constitutes a lock (i.e. some constraint on the input). \$\endgroup\$ Commented Jun 14, 2017 at 17:47
  • \$\begingroup\$ Yes. This is very incomplete/unclear as of yet. \$\endgroup\$
    – 0xffcourse
    Commented Jun 15, 2017 at 3:27

When did I need to be born to celebrate a magic birthday?

I was born in 1984 and in 2016 I became 32 years old, which is 20 in base 16, what a coincidence!

Your task is, given the year of interest -say 2016- , to calculate the year I had to be born to be able to say In 2016 I have celebrated/I will celebrate my 20th birthday (in base 16).

  1. take n-digit decimal number - the year xy.
  2. Split it in half, if n is odd, the digit the middle is appended to number side.
  3. Calculate the year I had to be born to be x base ys old in the year of xy.

Your code shall return Not-a-Number or error message if the decomposition cannot be resolved.


>  foo(2016)
1: '2016' -> '20' '16'
2: 20 base 16 = 32
3: 2016-32 = 1984
>> 1984

> foo(445)
1: '445' ->'44' '5'
2: 44 base 5 = 24
3: 445-24 = 421
>> 421

1: '7' -> '7' ''
Error, base not defined
>> nan

1: '10' -> '1' '0'
Error, base 0 don't exist
>> nan

1: '1805'->'18' '5'
Syntax error
>> nan

Test cases:

 7    : nan/error
10    : nan/error
78    : 71
445   : 421 
1024  : 1000
1805  : nan/error
1936  : 1891
1984  : 1891
1999  : 1891
2016  : 1984
2015  : 1985
10002 : 9998
10912 : 10759
116015: 115769

Shortest answer in bytes wins. Standard loopholes apply.

  • 1
    \$\begingroup\$ Boo-urns to input validation! Can you add a full example? \$\endgroup\$
    – Shaggy
    Commented Jun 21, 2017 at 16:14
  • \$\begingroup\$ Do you mean full ungolfed code or full path from, say 445 to 421? \$\endgroup\$
    – Crowley
    Commented Jun 21, 2017 at 16:25
  • \$\begingroup\$ Just a walkthrough of how to get from input to output. \$\endgroup\$
    – Shaggy
    Commented Jun 21, 2017 at 16:26
  • \$\begingroup\$ @Shaggy Thanks for sugegstion. Is it better now? \$\endgroup\$
    – Crowley
    Commented Jun 21, 2017 at 17:16
  • \$\begingroup\$ Shouldn't 1999 result in 1891? \$\endgroup\$
    – Emigna
    Commented Jun 22, 2017 at 8:49
  • \$\begingroup\$ @Emigna Correct. Updated. \$\endgroup\$
    – Crowley
    Commented Jun 22, 2017 at 13:36

Is there a total ordering?


Given a set of strings, determine whether the characters expose a total ordering based on location in the strings.

In this challenge, strings are used as a predicate to determine the order in which characters should appear in this set. For example, the string



  • All instances of "N" should appear only after instances of "O"

  • All instances of "E" should appear only after instances of "N"

By this reasoning, the string "FOOOONZAi.EE" follows this ordering, but "NEEEE3#?EAO" does not (there is an "O" after an "N").

Your challenge is to take a set of strings and determine whether these strings define a total ordering without any logical flaws. This would occur as a cycle of any length, such as:

  • "N" must follow "P"

  • "P" must follow "N"

...or such as:

  • "A" must follow "B"

  • "B" must follow "C"

  • "C" must follow "A"


Rather than strings, you may take in lists of characters or integers if you wish.

Since this is a , you may output any two consistent values for yes or no, such as true and false, zero and non-zero, exception and no exception, etc. Just specify your output format in your answer.

Test Cases

["ONE", "TWO", "THREE", "FOUR", "FIVE", "SIX"] -> true (one possible ordering is "TWFOUIVXNHRE")
["SEVEN"] -> false ("E" must follow "V" which must follow "E")
["ZERO", "FOUR"] -> false ("R" must follow "O", but "O" must follow "U" which must follow "R")
["", ".", "forty", "this->~why();", " -y."] -> true
["AB", "BC", "CD", "DE", "AC", "BD", "CE", "EA"] -> false

(more can be written if needed)


This is , so the shortest answer in each language wins.


How can I make this more clear? Any suggestions?

  • \$\begingroup\$ This could be more general if you allow lists as input rather than strings. \$\endgroup\$
    – lirtosiast
    Commented Jul 13, 2017 at 22:18
  • \$\begingroup\$ @lirtosiast So a list of lists of characters? \$\endgroup\$ Commented Jul 14, 2017 at 0:52
  • \$\begingroup\$ I would recommend lists of integers instead of strings. I feel the challenge would be "cleaner" that way, but that's up to you. Also, you should have a test case where where each pair is consistent but the whole set isn't. \$\endgroup\$
    – Zgarb
    Commented Jul 15, 2017 at 7:17
  • \$\begingroup\$ Is this not effectively the same as your previous challenge of whether a directed graph has a cycle? \$\endgroup\$
    – xnor
    Commented Jul 16, 2017 at 1:19
  • \$\begingroup\$ @xnor I suppose it is, though I think the challenging part is collecting the edges by vertex. How about this: I rewrite the challenge to be "Give a valid ordering for a set of strings, guaranteed that one exists"? \$\endgroup\$ Commented Jul 16, 2017 at 3:49
  • \$\begingroup\$ @musicman523 I remember a DAG sorting challenge, I think that would be a dupe of it. \$\endgroup\$
    – xnor
    Commented Jul 16, 2017 at 3:52

Make a Minecraft Crafting Table

  • \$\begingroup\$ +1 This seems like a really fun challenge! I think you should clarify what counts as a recipe. For instance, are gold chestplates and iron chestplates counted as 2 working recipes, or do you need to implement all chestplates to count for 1 recipe? Similarly, for unshaped crafting, does each valid combination (ie ----G---- and -G------- for gold nuggets) count seperately? \$\endgroup\$ Commented Jul 15, 2017 at 0:03
  • \$\begingroup\$ Unfortunately, you were ninja'd by Minecraft 1.12, which adds recipe hints ;) \$\endgroup\$
    – hyper-neutrino Mod
    Commented Jul 15, 2017 at 20:24
  • 1
    \$\begingroup\$ @HyperNeutrino I posted this first (MC 1.12 came out June 7th) so I ninja'd Minecraft \$\endgroup\$ Commented Jul 15, 2017 at 20:25


Your Task

You must create a self-mutable program that, when run, outputs a non-zero integer and also overwrites the file with a program that outputs double the number.

For example, if I run the program self-improvement and it outputs 10, it must output 20 when I run it the second time, output 40 the next time, and so on.

Additional Notes

  • You must not rely on any file on the computer other than your program.
  • Said program must consist of only one file.
  • Of course, no loopholes that are banned from the entire site.
  • You can assume that your program won't go tested beyond the range -2^16 to 2^16-1.
  • 1
    \$\begingroup\$ no loopholes banned -> no loopholes that are banned, overwrite -> overwrites. Also, something on how far this needs to go might be important - does it have to work infinitely, or only to INT_MAX for the language in question? If you choose the second option, keep in mind language with very small numeric caps. Otherwise, welcome to PPCG, I really like this question, and I'm glad you Sandbox'd it :) \$\endgroup\$
    – Stephen
    Commented Jul 17, 2017 at 21:33
  • \$\begingroup\$ @StepHen Thanks; I decided that they can assume that the parameters won't go beyond INT_MAX. \$\endgroup\$ Commented Jul 17, 2017 at 21:38
  • 2
    \$\begingroup\$ In case anyone needs to find it at some point, don't forget about this standard loophole. \$\endgroup\$
    – Stephen
    Commented Jul 17, 2017 at 22:24
  • \$\begingroup\$ @Artyer I stated "any nonzero integer". That excludes 0. \$\endgroup\$ Commented Jul 17, 2017 at 22:55
  • \$\begingroup\$ @LawfulLazy I didn't read that too well. \$\endgroup\$
    – Artyer
    Commented Jul 17, 2017 at 22:58
  • \$\begingroup\$ @Artyer I'll make sure to embolden it. I've also set a minimum integer range for StepHen. \$\endgroup\$ Commented Jul 17, 2017 at 22:59

How high can you count in English?


In 500 bytes (or fewer) write a program that outputs a list of the English word forms of as many consecutive integers greater than zero as you can.

For example, score 6:

one two three four five six

Example submission (hopefully you can do better than this):

Python 3, score 43 (488 bytes)

print(["one","two","three","four","five","six","seven","eight","nine","ten","eleven","twelve","thirteen","fourteen","fifteen", "sixteen","seventeen","eighteen","nineteen","twenty","twenty-one","twenty-two","twenty-three","twenty-four","twenty-five","twenty-six","twenty-seven","twenty-eight","twenty-nine","thirty","thirty-one","thirty-two","thirty-three","thirty-four","thirty-five","thirty-six","thirty-seven","thirty-eight","thirty-nine","forty","forty-one","forty-two","forty-three"])

Try it online!

Scoring and rules

For each language, the person whose code counts the highest wins. In case of a tie, the person who submitted first wins.

  • No modules, libraries, builtins that convert from numeric to word form are allowed.
  • You must output all integers from 1 (one) to n (your score) without missing any. If you want to output 0 (zero) as well, that's fine.
  • You are allowed up to 500 bytes of code. Your code may be a full program or function.
  • Number format: for consistency, all numbers must match the output of this site.
  • Standard loopholes apply (of course)
  • Standard output rules apply
  • 2
    \$\begingroup\$ Come on, you could at least make your example better by just using a single space-separated string ;) but nice challenge! :) \$\endgroup\$
    – hyper-neutrino Mod
    Commented Jul 18, 2017 at 18:00
  • 1
    \$\begingroup\$ Alternate title: How high can you count in English? \$\endgroup\$
    – Stephen
    Commented Jul 18, 2017 at 18:06
  • \$\begingroup\$ You may, however, omit "and" if you'd like. do we have to omit and or not? Personally, I'd say you wouldn't as it 'sounds more correct'. \$\endgroup\$
    – Okx
    Commented Jul 18, 2017 at 18:11
  • \$\begingroup\$ @StepHen changed, thanks. \$\endgroup\$
    – wrymug
    Commented Jul 18, 2017 at 18:15
  • \$\begingroup\$ Your edit doesn't answer my question. Do we have to omit and or not? \$\endgroup\$
    – Okx
    Commented Jul 18, 2017 at 18:15
  • \$\begingroup\$ @Okx I take your point. Removed. \$\endgroup\$
    – wrymug
    Commented Jul 18, 2017 at 18:28
  • 1
    \$\begingroup\$ I'm not sure this adds to existing number-to-english challenges like this or this. Past getting the same basic pattern of digits down, the question seems to be how many prefixes for powers of 1000 one can compress into the remaining bytes. \$\endgroup\$
    – xnor
    Commented Jul 18, 2017 at 18:33
  • 1
    \$\begingroup\$ Off the jokes, what if I can count to infinity? \$\endgroup\$ Commented Jul 20, 2017 at 13:40
  • \$\begingroup\$ @courtois I don't see how that would be possible, but I guess you'd win \$\endgroup\$
    – wrymug
    Commented Jul 20, 2017 at 13:42
  • \$\begingroup\$ @rosslh yeah you're right, I just saw the sentence Number format: for consistency, all numbers must match the output of [this](https://lingojam.com/NumbersToWords) site. And for my method, it would have been with million of billion of billion of ... though the site stops at one hundred novenonagintanongentillion, something like 3002 zeros. \$\endgroup\$ Commented Jul 20, 2017 at 13:43

ASCII addition


Given two strings, your job is to:

  1. Convert each character to their respective ASCII decimal value
  2. Concatenate the numbers into one large number
  3. Add these values together
  4. Get the ASCII characters represented of each pair of numbers starting from the right (or if there are not enough numbers, take a number alone)
  5. Leave unprintables (ie not in the range 32 - 126), and output the rest

An example for HELLO and WORLD

"HELLO"    + "WORLD"
H E L L O  + W O R L D
7269767679 + 8779827668
1 60 49 59 53 47    (separated to show ASCII conversion easily)
   <  1  ;  5  /    convert to ASCII by converting pairs to their respective characters (note: you start from the last pair)
<1;5/               output (note there is no 0x01) 


  • Each string will be a maximum of 6 characters long
  • Input will always contain readable ASCII characters
  • You have to take pairs of numbers from the end of the sum and convert each one of them to ASCII
  • You must not print unreadable characters if their values appear and instead skip them


ABC + XYZ  //input
656667 + 888990
1 54 56 57
   6  8  9       //again 0x01 is left out
689      //output


  • Your submission can be either a program or a function

This is so the program with the shortest bytecount wins!

Sandbox Questions

  • Are the specs clear enough?
  • Will this question give me hats?
  • Any better title suggestions?
  • \$\begingroup\$ What range of ASCII characters do we need to be able to handle in the input? Only printable characters or any? \$\endgroup\$ Commented Dec 20, 2016 at 20:07
  • 1
    \$\begingroup\$ I swear we've had this challenge, but I can't seem to find it ... \$\endgroup\$ Commented Dec 20, 2016 at 20:21
  • \$\begingroup\$ Edited that into the specs \$\endgroup\$
    – user41805
    Commented Dec 20, 2016 at 20:22
  • \$\begingroup\$ @TimmyD How's this now? \$\endgroup\$
    – user41805
    Commented Dec 21, 2016 at 7:28
  • \$\begingroup\$ what if a character is unprintable? do we still need to output it then? \$\endgroup\$ Commented Dec 21, 2016 at 7:46
  • \$\begingroup\$ @DestructibleWatermelon Yes \$\endgroup\$
    – user41805
    Commented Dec 21, 2016 at 7:56
  • \$\begingroup\$ Better wording. Given that a bunch of control codes are possible, some test cases that demonstrate expected behavior when (e.g.) 11 or 13 occur in the output would be of good value. \$\endgroup\$ Commented Dec 21, 2016 at 13:49
  • \$\begingroup\$ @TimmyD Do you think it would be a good idea if I restrict the output to displaying only characters if their values are between 32 and 126 because outputting other values might be difficult in some languages? \$\endgroup\$
    – user41805
    Commented Dec 21, 2016 at 13:54
  • \$\begingroup\$ That may be a better way to go, but you'll need to be very careful with the wording. For example, suppose that the output number is ...226... and the 26 is slated to be the pair of digits that get converted to ASCII. Obviously, that's outside the printable range, so let's look at the next digit, but now 226 is also outside. Does that mean just the 6 is skipped? The 26 is skipped? The 226 is skipped? \$\endgroup\$ Commented Dec 21, 2016 at 13:58
  • \$\begingroup\$ @Cowsquack You don't say precisely (but we see it in the test case) how many digits we need to parse at a time, before outputting. \$\endgroup\$ Commented Jul 18, 2017 at 14:06
  • \$\begingroup\$ @V.Courtois Is this clearer now? \$\endgroup\$
    – user41805
    Commented Jul 19, 2017 at 12:48
  • \$\begingroup\$ @Cowsquack This is! In fact you could see, in the comments some were wondering too if we had to print things like 125 -> }. \$\endgroup\$ Commented Jul 19, 2017 at 12:55

Deduplicate equivalent expressions

Suppose we wanted to generate all expressions containing at most 2 of + and −. We might have a list like this:

a + b + c         b + c + a
a + b - c         b + c - a
a - b + c         b - c + a
a - b - c         b - c - a
a + c + b         c + a + b
a + c - b         c + a - b
a - c + b         c - a + b
a - c - b         c - a - b
b + a + c         c + b + a
b + a - c         c + b - a
b - a + c         c - b + a
b - a - c         c - b - a

There is a lot of repetition here. It surely isn't necessary to include all of a + b + c, a + c + b, b + a + c, b + c + a, c + a + b, and c + b + a, since they all mean the same thing. This can be deduced from knowing that, for any x and y, x + y is the same as y + x.

Similarly, b + a - c and a - c + b are equivalent. To deduce this, one must know that, for any x and y, x - y is the same as x + (-y).

Let's assume the following:

[1]: a + b == b + a
[2]: a - b == a + -b

Then, we can deduce that b + a - c and a - c + b are equivalent:

start:  b + a - c
        b + a + -c        by 2
        a + b + -c        by 1
        a + -c + b        by 1
end:    a - c + b         by 2

Therefore, they are the same. After performing similar proofs, we are left with the list:

a + b + c
a + b - c
a - b + c
a - b - c
b - a + c
b - c - a
c - b - a

Definition of an expression

An expression can be described as:

variable   = "a" | "b" | "c" | ... | "y" | "z";
digit      = "0" | "1" | "2" | ... | "8" | "9";
number     = digit . digit*;
operator   = "!" | "#" | "$" | "%" | "&" | "*" | "+"
           | "~" | "-" | "." | "/" | ":" | ";" | "<"
           | "=" | ">" | "?" | "@" | "^" | "_" | "`"
           | "|";
data       = number | variable;
subexpr    = data | operator* . data;
expression = subexpr
           | subexpr . operator . expression;

Where | suggests alternatives, . suggests concatenation (with potential whitespace around each operand), * suggests "0 or more times", and " is a string literal.

x + y, j * i - 3, u & 4 * 2 < ~4, q % ~*^t and r are all expressions.

You should assume all operators are left-associative.

Definition of assumptions

An assumption is a pair of expressions said to be equivalent. This means one can be transformed into the other. When performing a transformation using an assumption, one replaces all the appropriate variables and maintains the numbers as they are. (These "variables" can also be sub-expressions, which is any expression not using an operator in the assumption.)

For example, if the assumption is !a == a + 5, then one can transform t + 5 into !t and 3 + 5 into !3.

Another example: if the assumption is a + b == a * b @ b, then 5 + 2 can become 5 * 2 @ 2 and z * 3 @ 3 can become z + 3, but z * 4 @ a cannot be reduced further using this rule.

One last example: if the assumption is a < b == a, then 1 + 3 & 5 < 2 * 3 + 6 would become 1 + 3 & 5, and 1 + 2 < x + y < 7 $ q would become 1 + 2, since it would be equivalent to (1 + 2) < (x + y) < (7 $ q), which is thus 1 + 2.

If either side of the assumption is a single variable, numbers are excluded from this assumption. E.g., the assumption a == 3 would only apply to variables.

Expression equality

Two expressions are equal if they can be proven to be the same. Variables must be the same for each expression; for example, a + b is not by default the same as b + c.


Your task is to remove "duplicate" expressions given some assumptions. You can use any unambiguous symbol or method, including taking a pair of strings, to represent an expression. The expressions remaining in the result do not necessarily have to be in the set, but must be equivalent by the given assumptions. E.g., if you have a + b - c and b - c + a in the input, you can have -c + b + a represent these in the resultant set. You should try each equation in the order that it's given to you (to simulate "precedence").

The input consists of a list of assumptions and a list of input expressions. The input expressions can be an array or container of strings or string pointers, or in any way standard to your language. (E.g., for C, one should expect null-terminated strings.) The input format must be consistent for all runs.

The output can be a list representation (as is standard to your language), can be separated by newlines (\r, \n, and \r\n are acceptable), or separated by commas. The output format must be consistent between runs.

This is a , so the shortest program in bytes wins.

Test cases

Every output is merely an example, and is not the only valid output.

Assumptions: { e1 == e2, e3 == e4, ... eN-1 == eN }
Input: { expr1, expr2, ... exprN }
Output: { expr1, expr2, ..., exprK }

Assumptions: { "a + b" == "b + a" }
Input: { "3 + 4", "4 + 3", "5 * a", "a + 2", "2 * a", "a * 5", "a + b", "2 + a" }
Output: { "3 + 4", "5 * a", "a + 2", "2 * a", "a * 5", "a + b" }

Assumptions: { "a + 0" == "a", "a * 1" == "a", "a * b" == "b * a", "a + b" == "b + a" }
Input: { "1 * 2 * 3", "3 * 2 + 0", "1 + 2 + 3" }
Output: { "1 * 2 * 3", "1 + 2 + 3" }
    OR: { "2 * 3", "1 + 2 + 3" }

Assumptions: { "~a" == "a ~ a" }
Input: { "~z", "z ~ z", "~a", "~~a", "a ~ a ~ a ~ a" }
Output: { "~a", "~z", "~~a" }

Assumptions: { "a + b" == "0" }
Input: { "x + y", "0", "3 + y + a + v + k", "75", "4 + 2" }
Output: { "0", "75" }

Assumptions: { }
Input: { "x + y", "x + y", "y + x", "3", "3 ! 3" }
Output: { "x + y", "y + x", "3", "3 ! 3" }

Assumptions: { "j" == "3" }
Input: { "v + t", "z", "q", "q + r + t", "4 + 2" }
Output: { "v + t", "z", "q + r + t", "4 + 2" }

Assumptions: { "a" == "b" }
Input: { "a", "b + c", "e % t", "q & t", "!3", "z" }
Output: { "a", "b + c", "e % t", "q & t", "!3" }

Assumptions: { "1 & 0" == "0", "1 & 1" == "1", "0 & 0" == "0", "a & b" == "b & a", "0 ? a : b" == "b", "1 ? a : b" == "a" }
Input: { "1 & 1 & 0", "j & k", "y & z", "z & y", "1 & 0 ? k & j : 0" }
Output: { "0", "j & k", "y & z" }
  • \$\begingroup\$ Several of your expressions are inconsistently quoted. I also think you should explain what should happen when the assumption is something like 3 == a. Would I remove one of: 5, 7? It seems if you had a == b as the assumption you would, but the other one is counter-intuitive to me. Also, a == b is rather odd on its own, perhaps that is also a good test case. Also, the empty assumption if you intend to allow it. \$\endgroup\$ Commented Jul 17, 2017 at 3:18
  • \$\begingroup\$ @FryAmTheEggman Could you clarify your first question? Are you asking what should happen if 5, 7 is the input given the assumption a == 3? \$\endgroup\$ Commented Jul 17, 2017 at 4:00
  • \$\begingroup\$ Yes, that's what I meant. \$\endgroup\$ Commented Jul 17, 2017 at 4:08
  • \$\begingroup\$ 1. subexpr = data | operator* . data; is surely equivalent to just subexpr = operator* . data;? 2. For the second test case it would be more illustrative to include an example output like { "0 * 2 + 3", "1 + 2 + 3" }. 3. I'm not sure what you mean by "If either side of the assumption is a single variable, numbers are excluded from this assumption". Is it that an assumption a == 3 with input {"b == 2", "c == 2", "4 == 2"} should give output e.g. {"3 == 2", "4 == 2"} rather than {"3 == 2"}? \$\endgroup\$ Commented Jul 17, 2017 at 7:43
  • \$\begingroup\$ @PeterTaylor 1. yeah, I original had * mean "1 or more"; will fix. 2. good idea. 3. Well, the way that input would be parsed is b = = 2, with a binary = followed by a unary =. Could you perhaps use a different symbol? I don't quite understand your confusion. \$\endgroup\$ Commented Jul 17, 2017 at 16:53
  • \$\begingroup\$ Sure: assumptions: { "a" == "3" }; input: {"b + 2", "c + 2", "4 + 2"}. Is {"3 + 2", "4 + 2"} the correct output? \$\endgroup\$ Commented Jul 17, 2017 at 18:03
  • \$\begingroup\$ @PeterTaylor Yes, it is. \$\endgroup\$ Commented Jul 17, 2017 at 18:34
  • \$\begingroup\$ The definition of assumptions includes an assumption with more than one operator, but none of the test cases do. I would think that's an important thing to test. \$\endgroup\$ Commented Jul 26, 2017 at 11:06
  • \$\begingroup\$ @PeterTaylor Added. \$\endgroup\$ Commented Jul 26, 2017 at 19:32

Challenge: Count integers 1 to 10, but slowly on a time delay.

Your challenge is to print the integers 1 to 10 to the screen with each integer separated by a newline, but on a time delay, such that it should take “n” seconds before the integer “n” is printed to the screen. For example, the program will wait 1 second before printing 1 and a newline, and then the program will wait 2 seconds before print 2 and a newline, and then the program will wait 3 seconds before print 3 and a newline.

So the expected standard output to the screen is this:


But the program will patiently wait “n” seconds before printing the integer “n” to the screen, so as your program may interpret it as …

waiting 1 second …


waiting 2 seconds …


waiting 3 seconds …


, and so on …


1) Since some programming languages may not have a sense of system time, you are not allowed to use any modules/libraries/functions which can measure the time of your program within your program. Therefore all programming languages can be used. This restriction puts all the programming languages on an equal level.

2) Restriction 1 makes it such that you are required to write a function which takes about 1 second to process, and then you can rerun that function “n” integer of times before printing the next integer “n” in 1 to 10 to the screen. You can call your function whatever you want or if you can get away without naming the function then you can do that, too. So, your program would see this behind the scenes:

performing function fx 1 time. #which the arbitrary function fx takes about 1 second to process.


performing function fx 2 times


performing function fx 3 times


… and so on

3) Therefore, your program should take about 1 + 2 + 3 + 4 … + 10 = 55 seconds to finish printing all the integers from 1 to 10 to the screen. Since you need to write a function that takes about 1 second to process, acceptable solutions can be off by plus or minus 2 seconds from 55 seconds.

4) You must time your script which counts the integers 1 to 10 on a delay, so you can use any external program to time your script. I recommend using the bash function time, and giving me the “real” time or actual elapsed time of your script in seconds. If you time your script another way, tell me how you did it. Give me three digits after the decimal place for the real time in seconds. Do not round or truncate.

winning condition:

The winning condition is the fastest program in terms of actual elapsed real time in seconds which is closest to 55 seconds. If, for example, one submission is 54.9 seconds and another is 55.1 seconds, the 55.1 second submission wins because 54.9 seconds is too fast. Again, the point of this program is to slowly print integers to the screen.

Again, it should take approximately 55 +/- 2 seconds to print the integers 1 to 10 to the screen.

  • \$\begingroup\$ I modified my problem in such a way that no programming language should have an advantage. \$\endgroup\$
    – xyz123
    Commented Jul 28, 2017 at 6:17
  • \$\begingroup\$ you are not allowed to use any modules/libraries/functions which can measure the time of your program Does this mean things like Stopwatch, getting the DateTime, Thread.Sleep, etc? \$\endgroup\$ Commented Jul 28, 2017 at 9:56
  • \$\begingroup\$ Nice challenge by the way but I'm not sure I like the winning condition. What's wrong with a code-golf? Also if I was you I'd trim some of the fluff down, at the moment there is a lot of text to information. \$\endgroup\$ Commented Jul 28, 2017 at 10:00
  • 2
    \$\begingroup\$ Will this not also vary dependent on what machine is running it? I really like the idea but I can imagine it being quite hard to test - especially with that winning condition, as I can imagine it coming down to hundredths of seconds, which would doubtlessly vary depending on the computer running it. Really like the idea nonetheless \$\endgroup\$
    – space junk
    Commented Jul 28, 2017 at 11:01
  • 2
    \$\begingroup\$ I suppose you could do it by only running them on TIO? TIO's time limit is 60 seconds so you'll have ample time, & it should be a bit more uniform in terms of processing time. \$\endgroup\$
    – space junk
    Commented Jul 28, 2017 at 11:02
  • \$\begingroup\$ Yes, TheLethalCoder you would not be allowed to do that. After I thought about, I realize that different computers may have different specifications for runtimes, although I suppose the winning condition could be the least number of bytes, but I was thinking about setting the winning condition to the fastest program which is closest to 55 seconds because if somebody wrote a program which had less bytes, but it was 1.9 seconds over 55 seconds, then it might be too slow for a lot of the other programs which had more bytes, but were faster. \$\endgroup\$
    – xyz123
    Commented Jul 28, 2017 at 15:08
  • \$\begingroup\$ Although if I set the winning condition to the least number of bytes, then I believe that I would have to increase the restraint to acceptable programs cannot be off by +/- 1 second from 55 seconds. \$\endgroup\$
    – xyz123
    Commented Jul 28, 2017 at 15:09

Random number from 0 to n


Write a program/function that, given a positive integer n, outputs a uniformly random integer from 0 to n.


  • Input will be a positive (non-zero) integer.
  • It will be in your language's number handling capabilities. // reword


  • Output must be a uniformly (pseudo)random integer.
  • Every integer in the range [0, n) must have an equal chance of being outputted.
  • You may assume that your chosen language's built-in RNG is uniform.
  • Must be in 0 to n, in [0, n), ≥ 0 and < n


  • Standard I/O rules apply.
  • Standard loopholes are forbidden.
  • This challenge is not about finding the shortest approach in all languages, rather, it is about finding the shortest approach in each language.
  • Your code will be scored in bytes, usually in the encoding UTF-8, unless specified otherwise.
  • Built-in functions that perform this task are allowed but including a solution that doesn't rely on a built-in is encouraged.
  • Explanations, even for "practical" languages, are encouraged.

Test cases

Note that this challenge is tagged and hence will have non-deterministic outputs.



  • We seriously don't have this already?!
  • \$\begingroup\$ Interesting idea, but how will the test cases work? Also I imagine builtins will be the shortest for many languages \$\endgroup\$ Commented Aug 9, 2017 at 8:39
  • \$\begingroup\$ @Chris_Rands not all challenges need test cases. \$\endgroup\$ Commented Aug 9, 2017 at 10:27
  • \$\begingroup\$ Definition of random? We don't have it because it's a bad idea. \$\endgroup\$
    – feersum
    Commented Aug 9, 2017 at 11:32
  • \$\begingroup\$ @Chris_Rands They're going to be random test cases. :P And yes, built-ins, but I'm afraid I don't like banning built-ins. \$\endgroup\$ Commented Aug 9, 2017 at 12:39
  • \$\begingroup\$ @feersum Yeah, I have to iterate over that better. Basically, it has to be uniformly pseudo-random. Any suggestions? \$\endgroup\$ Commented Aug 9, 2017 at 12:41
  • \$\begingroup\$ @AdmBorkBork If you do find a dupe, please let me know! I've searched everywhere and asked in chat, but I couldn't find one. \$\endgroup\$ Commented Aug 9, 2017 at 12:42
  • \$\begingroup\$ Extremely closely related \$\endgroup\$ Commented Aug 9, 2017 at 12:54
  • \$\begingroup\$ @AdmBorkBork Aha, I didn't find that (probably because it isn't tagged with number :P). It's very close but it has some arbitrary restrictions... Good enough to have a regular one? \$\endgroup\$ Commented Aug 9, 2017 at 13:19
  • 1
    \$\begingroup\$ The only way an answer can be valid here and fail the "arbitrary restrictions" there is if it's a trivial call to a built-in, so this question doesn't add anything to the site. \$\endgroup\$ Commented Aug 9, 2017 at 14:12
  • \$\begingroup\$ Now define pseudorandom. Also, using a hardware random generator is not allowed? \$\endgroup\$
    – feersum
    Commented Aug 9, 2017 at 14:16
  • \$\begingroup\$ @feersum I am not sure how much more I can define "random". I have edited in the fact that you can assume that your language's RNG is uniform. Where have I disallowed hardware random generators? \$\endgroup\$ Commented Aug 10, 2017 at 2:07
  • \$\begingroup\$ it has to be uniformly pseudo-random Pseudorandom is not random. \$\endgroup\$
    – feersum
    Commented Aug 10, 2017 at 12:23
  • \$\begingroup\$ So now the only definition of "random" we have is that built-ins that are designated as random number generators are considered "random". So the challenge can only be answered by builtins, since there is no other definition provided. \$\endgroup\$
    – feersum
    Commented Aug 10, 2017 at 12:24
  • \$\begingroup\$ @totallyhuman What if the language has a built-in RNG which is not a uniform RNG? \$\endgroup\$
    – isaacg
    Commented Aug 12, 2017 at 13:09

Case Matching Find Replace

Take three inputs, a string of text, T; a string of characters to replace, F; and a string of characters to replace them with, R. For each substring of T with the same (case insensitive) characters as F, replace them with the characters in R. However, keep the same case as the original text.

If there are more characters in R than F, the extra characters should be the same case as they are in R. If there are numbers or symbols in F, then the corresponding characters in R should keep the case they have in R. F will not necessarily appear in T.

You can assume all text will be in the printable ASCII range.


"Text input", "text", "test" -> "Test input"

"tHiS Is a PiEcE oF tExT", "is", "abcde" -> "tHaBcde Abcde a PiEcE oF tExT"

"The birch canoe slid on the smooth planks", "o", " OH MY " -> "The birch can OH MY e slid  OH MY n the sm OH MY  OH MY th planks"

"The score was 10 to 5", "10", "tEn" -> "The score was tEn to 5"

"I wrote my code in Brain$#@!", "$#@!", "Friend" -> "I wrote my code in BrainFriend"

"This challenge was created by Andrew Piliser", "Andrew Piliser", "Martin Ender" -> "This challenge was created by Martin Ender"

// Has a match, but does not match case 
"John does not know", "John Doe", "Jane Doe" -> "Jane does not know"

// No match
"Glue the sheet to the dark blue background", "Glue the sheet to the dark-blue background", "foo" -> "Glue the sheet to the dark blue background"

// Only take full matches
"aaa", "aa", "b" -> "ba"

// Apply matching once across the string as a whole, do not iterate on replaced text
"aaaa", "aa", "a" -> "aa"
  • \$\begingroup\$ Example(s) where the replacement string narrowly does not appear in the text might help. \$\endgroup\$
    – isaacg
    Commented Aug 17, 2017 at 19:55
  • \$\begingroup\$ @isaacg Added one, let me know if it's not what you were thinking \$\endgroup\$
    – Andrew
    Commented Aug 17, 2017 at 21:11
  • \$\begingroup\$ What is the expected output for "aaa","aa","b" and "aaaa","aa","a"? \$\endgroup\$ Commented Aug 17, 2017 at 21:33
  • \$\begingroup\$ Will we guaranteed that F will appear at least once in T? \$\endgroup\$
    – Shaggy
    Commented Aug 17, 2017 at 21:58
  • \$\begingroup\$ @Shaggy F will not necessarily appear in T. \$\endgroup\$
    – Andrew
    Commented Aug 17, 2017 at 22:04
  • \$\begingroup\$ @fireflame241 Added your two examples, good catch. \$\endgroup\$
    – Andrew
    Commented Aug 17, 2017 at 22:07
  • \$\begingroup\$ I would personally appreciate some kind of note on test case 3. It took me quite a while to realize that John Doe actually appears in full in the input string, rather than requiring we be able to replace all instances of John with Jane in that situation. \$\endgroup\$ Commented Aug 18, 2017 at 17:45
  • 1
    \$\begingroup\$ @KamilDrakari Added explanation for that case and a few others. Thanks for the feedback! \$\endgroup\$
    – Andrew
    Commented Aug 18, 2017 at 19:03

n-gon in m-gon

What is the greatest equilateral triangle you can fit into a regular pentagon? This is what this challenge is about, but with regular n-gons/m-gons.


Given two integers m,n greater or equal to 3, find the maximal ratio of the areas of the two polygons such that the m-gon is completely contained in the n-gon.


We are always talking about regular polygons, this means that all sides have the same length and all vertices are on a circle. The output can be a floating point-, fixed point- or rational number and must be correct to three decimal places.


 m  n  ratio (area m-gon / area n-gon)
 x  x  1 (for all x)
 3  4  1/4*sqrt(3) = 0.4330
 4  3  12/(7*sqrt(3)+12) = 0.4974
 3  6  1/2 = 0.5
 6  3  2/3 = 0.6667

Would be nice to have larger examples.

  • \$\begingroup\$ You may consider changing the language of the initial problem specification and input/output chart to specify that you are comparing the areas of the m- and n- gons \$\endgroup\$ Commented Sep 26, 2017 at 21:37
  • \$\begingroup\$ I see what you mean but right now I have difficulty coming up with a better wording (english is not my native language), if you have a better suggestion feel free to directly edit it! \$\endgroup\$
    – flawr
    Commented Sep 26, 2017 at 22:39

Got Your PIN!

I bet I can write down your PIN! 0000 0001 0002 0003 ... 9999.

Task: Write a program that outputs a string containing all possible four digit PINs.

The string should not contain any whitespace (trailing newline is okay).

Score: Length of program (in bytes) + length of output.


The output string only needs to contain each PIN somewhere in it; digits can be reused. For instance the string “123456789” contains the PINs 1234, 2345,3456,4567,5678,and 6789. By reusing digits, it's possible to save a significant amount of space from the naive implementation (0000000100020003...9999).

At best, this could be written as a 10,003 digit string. 4 digit for the first PIN, then one more digit for the other 9999 PINs.

Testing your code

I've written a basic Python script that can check your solution and indicate any PINs you are missing.

import sys
s = sys.stdin.readline()
passes = True
for ix in range(10000):
    pin = "%04d" % (ix)
    if s.find(pin) == -1:
        print("Missing PIN %s" % pin)
        passes = False

if passes:
    print('Result passes!')
print('String length is %d characters.' % len(s))
  • 1
    \$\begingroup\$ Nice, but I think you need to replace code-golf with code-challenge as this isn't just scored by shortest code. \$\endgroup\$
    – Adám
    Commented Sep 28, 2017 at 22:25
  • \$\begingroup\$ Very related \$\endgroup\$
    – H.PWiz
    Commented Sep 28, 2017 at 22:29
  • \$\begingroup\$ This is just codegolf.stackexchange.com/q/42728/194 with hard-coded input, and hence qualifies as a dupe. \$\endgroup\$ Commented Sep 29, 2017 at 11:08
  • \$\begingroup\$ Too bad it's a duplicate... but now I know about De Bruijn sequences! \$\endgroup\$
    – Dominic A.
    Commented Sep 29, 2017 at 23:19
  • \$\begingroup\$ @PeterTaylor: I think there's one even more dupey than that; I've been trying to find it all day. The "background" to it was a keypad on a building door that had a 4-digit code. \$\endgroup\$
    – Shaggy
    Commented Sep 30, 2017 at 0:59

Roll identical boson dice

When you roll two dice, the chance of a 5 and 6 in either order is a 2 in 36 (or 1/18), since it could happen as (5,6) or (6,5). But (6,6) can only happen one way and has a 1 in 36 chance.

Our dice will instead work as identical bosons: (5,6) and (6,5) are a single outcome {5,6} that is equally likely to {6,6}. So, each of these 21 unordered pairs is equally likely, with chance 1/21.

{1,1}, {1,2}, {1,3}, {1,4}, {1,5}, {1,6}, {2,2}, {2,3}, {2,4}, {2,5}, {2,6}, {3,3}, {3,4}, {3,5}, {3,6}, {4,4}, {4,5}, {4,6}, {5,5}, {5,6}, {6,6}

Similarly, for 3 dice, [5,5,6], [5,6,5], and [6,5,5] all count as a single outcome {5,5,6}, but [6,5,6] is different.

Task: Output a random roll of n boson dice, so that each possible result is equally likely as an unordered multiset, i.e. when sorted.

It's OK if your output is ordered as long as the overall probabilities are right. You may sometimes output [5,6] and sometimes [6,5] as long as their total chance is 1/21. They don't have to each be 1/42 chance. You could output [5,6] with 1/21 chance and never [6,5].

Input: A positive integer n

Output: A random list of n numbers from 1 to 6, so that each outcome is equally like when taken as an unordered multiset.

Time restriction: Your code has work up to n=50 within 1 minute.

  • \$\begingroup\$ I think it's unlikely that there'll be an algorithm for this terser than simply constructing the list of all possibilities, then selecting a random element from it. As such, I'm not sure the randomness adds to the task here. It's possible I'm wrong. (I toyed with some methods of restricting complexity to ban this solution, but I couldn't find one I was satisfied with.) \$\endgroup\$
    – user62131
    Commented Jun 22, 2017 at 22:48
  • \$\begingroup\$ @ais523 I know some fairly nice methods to generate these directly, but I don't know if they'll beat out the boring one you suggest. Rolling repeatedly until sorted is also boring and direct. Maybe a complexity restriction is needed. I'll see what I can golf up. \$\endgroup\$
    – xnor
    Commented Jun 23, 2017 at 7:24
  • \$\begingroup\$ with the (new) time restriction this is more like math-golf than code-golf. Can you give away any of your "nice methods"? Is it the weighted probability of every digit of the 50-digit number ,one of them? \$\endgroup\$
    – user72269
    Commented Oct 2, 2017 at 17:29

Limited-Information Maze-Solving Bot

(Still in draft/beta format. Feedback welcome.)

Based on a challenge idea I posted in chat and the ensuing conversation. Thanks to NathanMerrill, zgarb, and Jonathan Frech for assistance in fleshing this out.

The challenge

You're writing two separate programs/functions/routines/etc. The first, which we'll call the helper program, takes the input maze and calls the second, which we'll call the solving bot. The solving bot must solve the maze based on its interaction with the helper program.

The maze is 51x51 characters in size. For clarity in this challenge description, it is composed of # walls and corridors, but you can use any two distinct, consistent ASCII characters of your choice. E.g., use ! for walls and x for corridors, use 1 for walls and 0 for corridors, etc. This maze is one of many that will be the input to your helper program/function.

The maze start is always guaranteed to be somewhere on the left-most column, and the exit is always guaranteed to be somewhere on the right-most column. The maze is guaranteed to have at least one path from the start to the exit. As a result of this construction, the very top row and very bottom row are all #, and the very left and right columns are all # except for the start and exit.

The solving bot that you're creating needs to find a solution to the maze (not necessarily the shortest), but is limited in that it can only "see" a new 5x5 section of the maze at a time. The bot is scored by how many times it needs to request a new 5x5 section from the helper program/function.

The upper bound is obviously to simply request every possible 5x5 section, for a score of around 100. The lower bound is where someone with perfect knowledge of the maze can request only those 5x5 sections containing the exact route of the shortest solution, possibly as low as 10. Your bot will be run through (1000?) different mazes, and the bot with the fewest total requests will be the winner.

The solving bot is placed on the left-hand side where the start is, and the first 5x5 section is provided for free. However, the bot doesn't know where, vertically, it is on the left-hand side of the maze. It could be in the top corner (as in the example below), in the bottom corner, or anywhere in between.


The code you're writing be required to take the maze as an input (STDIN, a file to read, etc.) and call a subroutine of some sort for the solution bot.

Input: (1000?) 51x51 mazes Output: How many total requests your solving bot took

Yes, this is a non-observable requirement to be on the honor system and ensure the two "halves" of your program (i.e., the I/O half and the solving half) talk to each other correctly and accurately. I trust the community enough to believe that this is OK.

Further rules

Your solving bot should be deterministic. That is, when presented with the same maze two or more times, it should request the same number of 5x5 sections.

Example Maze

(generated from http://www.delorie.com/game-room/mazes/genmaze.cgi )

            #     # #   # #   #   #         #   # #
# ######### ##### # # # # # # # # # ####### ### # #
#   #       #   #   # # #   # # #         #   # # #
### # ####### # # ### # ### # # ############# # # #
#   #   #   # # #   # #     # #   #   #   # #   # #
# ####### # ### ### ####### # ### # # # # # # ### #
#     #   #       #       # # #   # # # # # # #   #
##### # ################# # # # ### ### # # # # ###
#   #       #     #         # #         # # # #   #
### ####### ##### # ######### ##### ##### # ##### #
#   #           #   #       #     # #   # #   #   #
# ##### ### ### # ### ########### # # # # # # ### #
#       # # #   # # #   #         # # # #   # #   #
######### # ### # # ### # ######### # ##### # # ###
#       # #   #   # # #   #       # # #     # # # #
# ##### # ### ##### # ##### ##### # # # ##### # # #
#     #   #       # # #   #     # #   # #     # # #
### # ####### # ### # # # ##### # ### # ####### # #
#   #     #   #   # #   #   #   #   # # #         #
# ####### # ##### # ####### # ### # ### # ### # ###
# #     # #     # #       # # #   #   #   #   #   #
# # ### # ##### # # ####### # # ##### ####### ### #
#   #   #       # #     #   # #     #   #   # # # #
##### ########### ##### # ########### # # # # # # #
# #   #       #   #   # #       #   # # # # #   # #
# # ### # ##### ### # # ### ### # # # ### # ##### #
#   # # #       #   #     # # #   # #   # #   #   #
##### # # ##### # ##### # # # ##### ### # ### # ###
#   # # #     # #   #   # #   #         #   # #   #
# # # # ##### # ### # ### ##### ### ##### # # ### #
# # #   #     #     # # #         #       # # #   #
# ####### ########### # ############### ##### # ###
#     #   #   #       #             # #   #   # # #
##### # ### # # ####### ######### # # ### # ### # #
#   #     # # # #       #       # #   #   # #   # #
# ##### # ### # # ######### ### ##### # ### # ### #
# #     #   #   #     #   # #   #     # #   #     #
# # ####### ######### # # # ##### ### # # ##### ###
# # # #     #     #   # # #     # # #   #     #   #
# # # # ######### # ##### ##### # # ##### ### # # #
#   #   # #     # # #     #   #     #   # #   # # #
### ##### # ### # # # ### # # ####### ### ### # # #
# #     # # #   # # # #   # #       #       #   # #
# ##### # # # ### # # ##### ####### ####### ##### #
#       #   # #   # #       # #   # #       #   # #
### ######### # # # ######### # # # # ####### # # #
#   #   #   # # # #   #         #   #     #   #   #
# ### # # # # # ##### # ####### ########### ##### #
#     #   #   #       #       #             #      

Example starting block:


# ###
#   #
### #
  • \$\begingroup\$ Does the solving bot also have to find the maze's start by scanning the left side or is it placed at the start position? \$\endgroup\$ Commented Sep 17, 2017 at 19:18
  • \$\begingroup\$ related \$\endgroup\$
    – Liam
    Commented Sep 17, 2017 at 21:47
  • \$\begingroup\$ @JonathanFrech Good point. We'll say that it's placed at the start position and the first 5x5 area is free. \$\endgroup\$ Commented Sep 18, 2017 at 16:39
  • \$\begingroup\$ You could maybe only use one space wide corridors, as the second space does not add any information and thus you do not really get a true 5x5 section of the maze. Just a thought, though. \$\endgroup\$ Commented Sep 25, 2017 at 11:50
  • \$\begingroup\$ @JonathanFrech No, that's a good point. I went with double-wide corridors since it looks more even with the character height, but looks aren't important when doing this challenge. I'll update the sizing. \$\endgroup\$ Commented Sep 25, 2017 at 15:54
  • \$\begingroup\$ Are this maze's dimensions not 51x51? \$\endgroup\$ Commented Sep 25, 2017 at 16:29
  • \$\begingroup\$ @JonathanFrech Apparently, with that generator, a size 50 doesn't yield a size 50. I'll either find a different generator or make my own for the actual challenge. Thanks! \$\endgroup\$ Commented Sep 25, 2017 at 17:28
  • \$\begingroup\$ I think the problem does not lie in the generator; it lies in the maze's nature. If your corridors are always one character wide and you have walls on the far left and far right, your maze size has to be odd. \$\endgroup\$ Commented Sep 25, 2017 at 19:13
  • \$\begingroup\$ Does this allow for "relative" vision? As in, after we get the initial block we would request "The next block to the right" instead of "the block at (2,7)"? \$\endgroup\$ Commented Sep 26, 2017 at 20:22
  • \$\begingroup\$ @KamilDrakari Sure, that would be allowed. It's dependent upon how that half of your submission works. \$\endgroup\$ Commented Sep 26, 2017 at 20:32
  • \$\begingroup\$ @JonathanFrech Good point. Maybe I'll compromise and make it a 51x51 maze. Thanks for the insight. \$\endgroup\$ Commented Sep 26, 2017 at 20:32

Type the alphabet as fast as you can

Your task is to make a program that measures how fast you can type the letters of the English alphabet.

  • The program shall only accept lowercase letters a to z in alphabetical order.
  • Each letter is echoed as typed on the same line (without new line or any other separators between letters).
  • If you type an invalid character the program shall output Fail and exit.
  • If you type all 26 letters the program shall output the time in milliseconds it took from the first to the last letter and exit.
  • The timer starts when you type the first letter, a.

Example outputs:




This is , so shortest answer in bytes wins.

  • \$\begingroup\$ Related. I would also recommend to put "time" in some way into the title. \$\endgroup\$
    – Laikoni
    Commented Oct 27, 2017 at 13:49
  • \$\begingroup\$ Thanks @Laikoni \$\endgroup\$ Commented Oct 27, 2017 at 14:53
  • \$\begingroup\$ Does the output have to be strictly in milliseconds, or at least in milliseconds, e.g. nanoseconds. Many builtin timing functions are in ns these days. \$\endgroup\$
    – nimi
    Commented Oct 27, 2017 at 16:39

Stable modular exponents

It is well-known that the final three digits of Graham's number are 387. This is because Graham's number is a ridiculously tall exponent tower of threes: 3^(3^(3^...))), and it can be shown that any such tower of height at least 5 has 387 as its final three digits.

This generalises: given any base n and any starting number a, the exponent tower a^(a^(a^...))) will eventually stabilize modulo n. After that point, whatever you put in the topmost exponent, be it just a last a, or a continuation of the exponent tower (i.e. more than one a), or any other number, its congruency class modulo n will not change. That is the challenge that I set before you here today.

Problem statement

Write a program or a function that takes two numbers a and n (within your language's standard signed or unsigned integer range) as input, and outputs the limit l of the sequence a%n, (a^a)%n, (a^(a^a))%n,..., which can be mathematically proven to be eventually constant (and therefore have a well-defined limit).

Your program should be able to handle a > n (note that a and a+n doesn't necessarily give the same result), and we require that 0 <= l < n.

Warning: Reducing the exponents mudulo n, i.e. calculating this sequence recursively using b[0] = a%n, b[i] = (a^b[i-1])%n will yield the wrong result, and might not stabilize.

Test cases

If we call the function f(a, n), it should give the following:

> f(3, 1000)
> f(6, 10)
> f(5, 9)
> f(14, 9)
> f(3, 81)

Scoring criteria

Standard code golf rules, use as few bytes as possible.


Voronoi Iteration

Given a finite set of points in the plane, output the set vertices of the corresponding Voronoi diagram.


A vertex of the voronoi diagram is a point of the plane that has the same distance to the three or more closest input points. As usual, you don't have to worry about the rounding issues of limited precision floating point numbers.


[(0,0),(2,0),(0,2),(2,2)] -> [(1,1)]
[(0,0),(1,0)] -> []
[(0,0),(2,0),(0,2)] -> [(1,1)]

Inspired by this question on MO.

  • 1
    \$\begingroup\$ I suggest having a test case with floating point coordinates. \$\endgroup\$
    Commented Jan 22, 2018 at 14:10
  • \$\begingroup\$ (although I already understood the test cases) it would be helpful to have an image. \$\endgroup\$
    Commented Jan 22, 2018 at 14:16
  • \$\begingroup\$ @user202729 It definitely needs more examples, but you can obviously interpret these coordinates as floating points as well. \$\endgroup\$
    – flawr
    Commented Jan 22, 2018 at 14:17
  • \$\begingroup\$ Having a test case with floating point coordinates will help testing if programs handle floating point input correctly. \$\endgroup\$
    Commented Jan 28, 2018 at 7:32

Invisible target - probability KotH


In short

Walls are gradually added and the player nearest to the stationary invisible target at the end of the game wins.


Players are all present on a 32 by 32 grid of square cells, which wraps toroidally. One randomly chosen cell is the target, which is not indicated to any of the players (regardless of whether they are on that cell or not). The target does not move.

Players all take their turn simultaneously. After each turn there is a small chance of a wall being added.

Wall rules

  • The wall will never be placed on a player.
  • The wall will never be placed in a cell that does not have a route to the target.
  • Of the possible positions for the wall to be placed, one will be chosen uniformly pseudorandomly.
  • The probability of a wall being placed each turn is 1/7.
  • The wall will be placed such that every player still has a route to the target (this includes never placing a wall on the target).

Note that a player having a route to the target means that there exists a path that does not include a wall. If another player blocks the path it still counts as a path.

Movement rules

  • A player can move to any of the 4 orthogonally adjacent cells (or stay still).
  • A player cannot share a cell with another player.
  • A player cannot move onto a wall.
  • A player can move onto the target, but will have no way of knowing that this has happened.

Starting position

At the start of the game the arena will have no walls and the players will be randomly positioned with the guarantee that there are no other players within each player's 5 by 5 neighbourhood.


Play will continue until no wall can be placed for 10 consecutive attempts (note that attempts only occur with probability 1/7 each turn so this will take more than 10 turns). When play stops the player closest to the target (by Manhattan distance) is the winner. Although this makes it possible to have an arbitrary number of joint winners, the density of walls by this point makes it unlikely there will be many, and in most cases there will be a player on the target cell, meaning only a single winner.

Each of the (one or several) joint winners scores one point. Games will be played until one player is the clear winner, or until it is clear there should be joint winners overall.

Input and output


During an N player game the input will be a space separated string of N+1 integers received on STDIN:

  • The player's position (an integer).
  • The position of any wall added since the player's last turn (an integer).
  • The position of every enemy player (N-1 integers).

Positions will be single integers from 0 to 1023, representing the distance in English reading order from the top left cell.

For a 4 by 4 arena this would give the following numbering:

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

If no wall was added the wall location will be 1024.

During a particular game the order of enemy players will be consistent - the nth location will always refer to the same enemy player.


The player must send an integer from 0 to 4 to STDOUT representing a move in English reading order:

1 2 3

(2 being no move).

A move to an unoccupied cell will not necessarily succeed - it will fail if another player is also trying to move to the same cell.

A move to an occupied cell will not necessarily fail - it will succeed if that player is also moving away from that cell (provided that player succeeds in moving away from that cell, and no other player is also trying to move to that cell).

This means two players can swap cells if they both decide to on the same turn.

A player taking longer than 50 milliseconds to respond will not move.

Sandbox questions

  • If someone can demonstrate that there can exist no better strategy than moving uniformly randomly, then I will not post this challenge. I'm hoping that the knowledge of the rules behind wall placement and the ability to block the movement of other players will make probability estimating competitive strategies non-trivial. This is answered - Nathan Merrill's strategy of moving to the reachable cell whose maximum distance to any other reachable cell is the shortest will beat the strategy of moving uniformly randomly (although in a crowded arena I don't believe this will be the best strategy so I still consider the question worth posting).

  • Should this be tagged ? I am expecting answers to make use of probability theory, but I can't know in advance what all the strategies will be. Is this close enough to use the tag?

  • I'm aiming for this to be a language agnostic challenge communicating with STDIN/STDOUT. Is there a language that is overdue to have its own language specific KotH contest, but that would still allow most users to participate? If not, I'll stick with language agnostic and include at least one example answer so that the processing of STDIN and STDOUT is provided in at least one language.

  • Method for deciding which attempted moves succeed. Is there any problem with this: Make a list of every intended destination (including own current cell for non-movers). For any destination that appears more than once, make all players aiming for that destination aim for their own current cell instead. Repeat (as this may have created more clashes) until no change is made. Move all the players to the resulting destination. Guaranteed to finish in N steps per turn for an N player game (worst case being a chain of players each moving to the next player's current cell, with the last player in the chain attempting to move onto a wall).

  • Pseudo random number source: Does anyone have a preferred/recommended random number generator? Is there any reason to consider a true random number source?

  • Alternative adversarial 2 player version: One player is the target, and the other player is seeking the target. Each player can move one square orthogonally or stay still. Walls are added as in the multiplayer game, and the game ends when the seeker moves onto the target's cell. The score of each player is the number of moves the game lasted. Lower score is better for the seeker player, higher score is better for the target player. The target can always see the location of the seeker. The seeker can never see the location of the target. Might also be interesting to allow both players to choose where to place a wall on their turn (in addition to moving). This might open up the possibility of double bluff. Walls would still be prevented from being placed on a cell that doesn't leave a path from seeker to target. Would this be more/less interesting than the multiplayer version? Are they sufficiently distinct to post as separate challenges, or should one be chosen as the one to be posted? Would this adversarial version work best as two KotHs that use each other's answers to judge their own answers (like a cops and robbers challenge) or should all the seeker answers and target answers be posted to one challenge? Alternatively each answer could be required to deal with being either a seeker or a target, but I like the idea of people being able to specialise and build just one or other, without being obliged to write both.

  • \$\begingroup\$ I believe the best strategy is to take all of the connected squares, and find the one that has the shortest walking distance to all other squares. Also, I think that walls should be placed every turn, as it appears to only slow down the game. \$\endgroup\$ Commented Jun 20, 2015 at 0:18
  • \$\begingroup\$ Also, I believe an interesting variant would be to have it more like the cats/mice KotH, where several mice compete to stand on the square first, and the cat tries to delay them as long as possible. \$\endgroup\$ Commented Jun 20, 2015 at 0:21
  • \$\begingroup\$ @NathanMerrill thanks for the feedback. When you say walls should be placed every turn, do you mean each time any player moves, or each time all N players have moved? \$\endgroup\$ Commented Jun 20, 2015 at 0:23
  • \$\begingroup\$ Each time all N players have moved, although I wouldn't be against the other either. \$\endgroup\$ Commented Jun 20, 2015 at 0:24
  • \$\begingroup\$ Would the variant involve a cat that knows the position of the target? \$\endgroup\$ Commented Jun 20, 2015 at 0:24
  • \$\begingroup\$ Yes, it would (wouldn't be interesting otherwise, I think) \$\endgroup\$ Commented Jun 20, 2015 at 0:25
  • \$\begingroup\$ @NathanMerrill my problem with adding the wall at the same point each time, after N players have moved, is that this means the players at the start of the cycle get first choice about where to move, which becomes more relevant in the later stages of the game. \$\endgroup\$ Commented Jun 20, 2015 at 0:26
  • 1
    \$\begingroup\$ Then I would add a wall after N+1 players have taken a turn, or rotate the player's turns \$\endgroup\$ Commented Jun 20, 2015 at 0:26
  • \$\begingroup\$ That's an interesting idea - I'll consider changing it to that (N + 1 turns between walls) \$\endgroup\$ Commented Jun 20, 2015 at 0:27
  • \$\begingroup\$ I'm not against several moves per player between walls though. I've chosen the arena small enough to allow a large number of moves in total, and the cell a player wishes to be on may be several cells away from the current cell, so I don't feel a strong need to add walls at high frequency. \$\endgroup\$ Commented Jun 20, 2015 at 0:30
  • \$\begingroup\$ Moving to the center of the arena may well be worse than a uniformly random strategy ;) There's definitely one better than random though. \$\endgroup\$
    – feersum
    Commented Jun 20, 2015 at 10:01
  • \$\begingroup\$ Question: will player bots have information storage? That is, will a player be able to remember every wall that has been placed? \$\endgroup\$ Commented Jul 29, 2017 at 1:30
  • \$\begingroup\$ @Draco18s yes the players will persist between moves, and can store information. I may place an upper limit on the amount of storage, but if I do it will be very generous. \$\endgroup\$ Commented Jul 29, 2017 at 1:36
  • \$\begingroup\$ In which case, moving randomly is not the best strategy :) \$\endgroup\$ Commented Jul 29, 2017 at 1:38
  • \$\begingroup\$ No, I'm thinking not, especially as a player can track what all of the other players are doing (they are distinguishable from each other, not just generic "enemy", but "player 1", "player 2", ...) \$\endgroup\$ Commented Jul 29, 2017 at 1:39

Number rewinder

Inspired by this SO question and little bit expanded.

Your task is to rotate left a given integer by one digit in a given base and return integer (move the MSD to the LSD position).

Input: Two integers Number > Base > 1
Output: Result

Test cases:

Number Base  Result  String representations
61453   10    14536  61453 -> 14536
61453   16      223  F00D -> DF (00DF)
61453    8   229481  170015 -> 700151
60429   16    49374  EC0D -> C0DE
62977   16    24607  F601 -> 601F

This is Code-golf, standard loopholes are forbidden and shortest answer wins.

  • 3
    \$\begingroup\$ I don't think your restriction about strings is going to work out well, there's not really a way to verify strings aren't being used. In any case, string rotation isn't that much different from the arithmetic required to do this operation. Personally, I'd recommend just allowing them. Thanks for using the sandbox! \$\endgroup\$ Commented Feb 5, 2018 at 22:09
  • \$\begingroup\$ (i.e., we don't like do X without Y or unobservable behavior) \$\endgroup\$
    Commented Feb 6, 2018 at 5:45
  • \$\begingroup\$ @FryAmTheEggman If there isn't some ToString-like function that opperates on any base using strings may be more difficult than dealing with it numerically. \$\endgroup\$
    – Crowley
    Commented Feb 6, 2018 at 8:12
  • 3
    \$\begingroup\$ I've never heard "rewind" used with this meaning before. I would call that operation "rotation left". \$\endgroup\$ Commented Feb 6, 2018 at 11:09
  • \$\begingroup\$ @PeterTaylor Thanks for suggestion. I've updated bounds for the inputs to avoid rotating 1 digit "number" \$\endgroup\$
    – Crowley
    Commented Feb 6, 2018 at 16:04

The Programming Language Quiz, Mark II - Cops

  • \$\begingroup\$ I would say posting the challenge separately for cops and for robbers with cross-referencing links will be better, because posting robbers' answers to the cops' thread is a bit obfuscating I think, and it seems to be the general method to post a C&R. \$\endgroup\$ Commented Dec 14, 2017 at 0:56
  • 5
    \$\begingroup\$ For the languages on Try it Online!, what's to stop me from trying to run a cops submission on every language there? Doesn't that put all those languages at a disadvantage? \$\endgroup\$
    – KSmarts
    Commented Dec 14, 2017 at 18:46
  • 1
    \$\begingroup\$ Same concern as @KSmarts. With the C&R challenges where it is instead the code is hidden it's hard to crack answers because there are a large amount of possible programs they could have written (naively) but there are a relatively small number of languages so people could just try them all. I feel like there could be some added rules to get around this but I don't know what they'd be \$\endgroup\$
    – dylnan
    Commented Dec 15, 2017 at 1:17
  • \$\begingroup\$ So basically it is possible but impractical to try all of the languages on TIO. \$\endgroup\$
    Commented Dec 15, 2017 at 11:21
  • \$\begingroup\$ @KSmarts Good point. As DLosc pointed out, changing the time restriction to, say, 5 minutes would most likely handle it. \$\endgroup\$ Commented Dec 15, 2017 at 15:20
  • \$\begingroup\$ @DLosc 1) Unary output seems perfectly acceptable, I'll edit it in. 2) I think allowing arbitrary starting and ending delimiters is a good idea 3) non-standard flags are flags that aren't required to run the program. Although, in hindsight, a better idea would be to allow flags but the answer has to state which flags are used. 4) Thanks for pointing that out, editing now. \$\endgroup\$ Commented Dec 15, 2017 at 15:27
  • \$\begingroup\$ @DLosc Edited in \$\endgroup\$ Commented Dec 15, 2017 at 22:53
  • \$\begingroup\$ The character limit for PPCG answers has been bumped to 65,536. Also, not that printable ASCII excludes newlines. \$\endgroup\$
    – Dennis
    Commented Dec 16, 2017 at 4:33

How acceptable is it to base challenges off of pre-existing challenges? I saw the challenge for Your Own Pet Ascii Snake and had a thought about making the output look more 'snakelike' by printing the characters |,\,/,(,),_ instead of always using the + character.

Here's how it would work. You would get some positive, negative, and 0 numbers as input, and based on those numbers, the snake moves one row down and that many characters in that direction. So, for a snake like the ones in the previous problem, your array would be restricted to the numbers 0, 1, and -1.

Here are the rules to draw the snake, the characters you print are dependent on the spacing of the lines before and after it.

So, say your snake is at position n (in the previous problem, n=30 to start, in this one you need to figure out a number for n that will keep your entire snake on the screen),

if the input is 0 you print n-1 spaces and a |
if the input is +1 you print n spaces and a \
if the input is +2 you print n spaces and \_, +10 would be \_________ (9 _ and a backslash)
if the input is -1 you print n-2 spaces and /
if the input is -2 you print n-3 spaces and _/, -10 would be _________/

Here's an example snake based on this array [+4, -3, +1, -4, +2, -3, +2, +5, -4, -1, -2, 0, +4, -4]


I could also add optional 'curvy' rules that would include the '(' and ')' characters on direction changes to produce a snake like this, based on the same array above:


note that for the curvy snake, 0s are handled differently depending on if there is a direction change in the rows above and below them, here there is a negative, 0, positive pattern, so we use a '(', if there were no direction change we'd use a |, and in the opposite pattern, a ')'

There are spaces to the left of my example snake because I didn't want to count out how many spaces I should leave exactly, I don't know whether it should be mandatory to cut out extra whitespace to the left, or whether to let people have as much or as little whitespace as they want, provided that their snake doesn't 'go off screen'

  • \$\begingroup\$ I don't think there's a problem determining the next line randomly. For example, my program could generate the numbers for all 30 lines and then go back and format the characters appropriately. However, I'm not sure that's different enough to warrant its own challenge. Your second idea, though, regarding taking +1/-1/0 as input, that has some merit as it's pretty radically different than the existing snake challenge. \$\endgroup\$ Commented Feb 23, 2018 at 19:29
  • \$\begingroup\$ I'm far, far too tired for you to rely on my opinion alone but this looks sufficiently different enough from the 2 existing challenges to not be a dupe, once it's fleshed out a bit more. I'd suggest waiting a while before posting it, though, as people do get jaded of a barrage of similarly themed challenges. \$\endgroup\$
    – Shaggy
    Commented Feb 24, 2018 at 21:26

The TI series of calculators. I've put a lot of hours in writing TI-BASIC programs. The single most tedious part was either scrolling through the program, or switching between alpha and numeric input.

For those unfamiliar, the entire Latin alphabet is overlaid on the existing keys alphabetically, and to type any of them (For variable names or assembly programming), you had to first press the ALPHA key, then your desired letter. Alternatively, you could press 2ND, A-LOCK and type in any number of letters, before pressing ALPHA again for numeric input.

Why am I saying all this? I want to write a code golf like challenge where scoring is done with this tediousness in mind, like switching between numbers/punctuation and letters carrying an extra penalty.

My original idea is that the source code of the submission would be converted to hexadecimal, and the number of transitions between alphas and numerics would be the "score," with a lower score being better.

How can/should I better refine this scoring system, and what sort of challenge should accompany it?

The Challenge

Implement a program or function, in the language of your choice, that takes a string of characters as input and outputs that string's "tediousness score."

How tediousness is scored:

  1. The string is converted into its hexadecimal representation. This is Unicode for languages that don't specify, but for a language like, say, Jelly, the Jelly codepage is used.
    "The quick brown fox jumps, over the lazy dog's back." in Unicode == 0x54686520717569636b2062726f776e20666f78206a756d70732c206f76657220746865206c617a7920646f672773206261636b2e
  2. Each switch between numbers and letters in the hexadecimal representation is counted.
    ..0 12 34 56 78 9...
  3. This is your tediousness score. For this example, it is 25.

Standard loopholes apply.

Should I change the tediousness score? How might I implement golf? I thought about adding/multiplying the two together, but it still seemed to be in favor of just plain golfing. I wanted to make golfing languages be a bit harder for a golfing challenge, since they really pack their character sets and there would be a high "tediousness score".

Is the challenge, as written, clear enough? Should I restrict or relax the I/O requirements?

  • 5
    \$\begingroup\$ I have found that challenges with novel scoring mechanisms work nicely if the task is to implement the score calculator itself, see e.g. this challenge. \$\endgroup\$
    – Laikoni
    Commented Mar 12, 2018 at 8:24
  • 1
    \$\begingroup\$ Probably tedious×log(nbytes)? \$\endgroup\$
    Commented Mar 13, 2018 at 14:16
  • \$\begingroup\$ What about giving a list of integers (byte values), or restricting the input to printable ASCII? \$\endgroup\$
    Commented Mar 13, 2018 at 14:29
  • \$\begingroup\$ @user202729 Because I wanted entries to be scored by this method, and I didn't want to exclude esoteric golfing languages. I specifically want to include them, and make them hard to win with. \$\endgroup\$
    – Orion
    Commented Mar 14, 2018 at 6:06
  • \$\begingroup\$ What about "the input is a byte string"? That way you don't need to worry about the encoding. \$\endgroup\$
    Commented Mar 14, 2018 at 8:19

Signed exponentiation

Let us define signed exponentiation of a base x to a power n as this procedure:

  1. Take the absolute value of x.
  2. Raise to the power of n.
  3. Re-apply the original sign of x.

Or, for a more mathematical (albeit slightly flawed) definition:

the absolute value of x, to the power of n - 1, times x

For this challenge, we will denote the signed exponentiation of x to a as x ' a. Some notes:

  • Unlike f(x) = xa, f(x) = x ' a is defined for negative x no matter the value of a.
  • The negative portion of the graph of x ' n is the positive portion rotated halfway about the origin; hence, f(x) = x ' a is an odd function for all values of a.
  • If a is odd, x ' a = xa for all values of x.
  • x ' 1 = x, while x ' -1 = 1 / x. x ' 0 is a sign function (+1 for positive x, -1 for negative x).


Given a base x and a power n, compute x ' n.


  • You may assume that -9 ≤ x, n ≤ +9.
  • You may assume that x ≠ 0, for the sake of avoiding 0 ' 0, 0 ' -1, etc.
  • You may assume that 1e-4 ≤ |x ' n| ≤ 1e9.
  • The output must be precise to at least 3 significant digits for the given test cases.
  • Input/output may be taken/given in any standard format.

Test cases

x, n -> output
1, 0 -> 1
3.14159, 0 -> 1
-9, 0 -> -1
7, 1 -> 7
-1.23456, 1 -> -1.23456
5, 2 -> 25
-4, 3 -> -64
0.1, 4 -> 0.0001
-9, 9 -> -387420489
1, -1 -> 1
-3, -1 -> -0.33333
5, -2 -> 0.04
2.71828, -9 -> 0.0001234
4, 0.5 -> 2
-2, 0.5 -> -1.41421
2, -0.5 -> 0.7071
3.8236, -1.6702 -> 0.10645
0.1, -9 -> 1000000000


This is , so the shortest code in bytes in each langauge wins.

Sandbox questions

  • Could the definition be improved?
  • Too many test cases? Are there any important ones missing?
  • Any issues with the rules?
  • \$\begingroup\$ This isn't a huge deal, but I find the "your answer must be accurate to x significant figures" to be a largely unobservable property. The only ways to know that a submission meets this criterium is to check each and every possible combination, or to give an incredibly complex proof. I've always thought it was better to say floating point errors don't matter and to use loopholes otherwise. Maybe just a small thing about having to be able to represent any of the numbers from 1e-4 to 1e9? \$\endgroup\$ Commented Oct 2, 2017 at 21:23
  • \$\begingroup\$ Thanks @FryAmTheEggman; I like your suggestion, but I'm afraid if it's not objective enough it'll cause any issues. Any suggestions on how to phrase it? \$\endgroup\$ Commented Oct 3, 2017 at 21:38
  • \$\begingroup\$ Perhaps have a few special test cases where they have to match the first three sig figs, and put in a line that those can't be hard coded. Then otherwise floating point errors don't matter? I'm honestly unsure. I've been getting to the point where I want to ask on meta because it always feels like fp questions are unclear or unobservable. \$\endgroup\$ Commented Oct 4, 2017 at 16:38
  • \$\begingroup\$ @FryAmTheEggman Hopefully a little more objective now. \$\endgroup\$ Commented Feb 27, 2018 at 1:54
  • \$\begingroup\$ I think this looks much better, but there are perhaps too many "mandatory" test cases? Maybe making a split will make it easier for people to check the validity of answers. \$\endgroup\$ Commented Feb 27, 2018 at 6:04
  • \$\begingroup\$ Am I allowed to print more than 3 sigfigs? \$\endgroup\$
    – Nissa
    Commented Apr 6, 2018 at 22:02
  • \$\begingroup\$ @StephenLeppik Of course. Should be slightly clarified now \$\endgroup\$ Commented Apr 7, 2018 at 1:37
  • \$\begingroup\$ The way to prevent hardcoding is to have hidden test cases. \$\endgroup\$
    Commented Apr 7, 2018 at 11:15
  • \$\begingroup\$ @user202729 Not sure I understand. How could hard coding possibly be useful for this challenge? \$\endgroup\$ Commented Apr 7, 2018 at 14:09
  • \$\begingroup\$ "for the given test cases". Although, yes, normally it is not useful. Whether you think it's necessary is up to you, but I doubt anyone would do it. \$\endgroup\$
    Commented Apr 7, 2018 at 14:32

Inversion languages


For this challenge you will design two languages, A and B. Both A and B should be Turing equivalent1. When run in your two languages a program has four options:

  1. It halts in both A and B.

  2. It halts in A but not B.

  3. It halts in B but not A.

  4. It doesn't halt in either A or B.

Your goal is to try to design A and B such that as many programs as possible halt in exactly one language (options 2 and 3). Doing such for all programs is impossible2, thus there will always be programs that meet either criterion 1 or criterion 4. The robbers will attempt to find these programs.

Further rules on languages

Since we are only considering whether programs halt or not we don't care about I/O. Because of this programs should not take any input. You may produce output but really doesn't matter because it can neither hurt nor help you.

All programs should be deterministic in both languages, this means that any program that can halt must always halt and any program that can run indefinitely must do so.


Your answer should contain the following information.

  1. A concrete description of both languages

  2. Proofs that each language as described is Turing complete.

Your definitions should be rigorous and unambiguous. Meaning that a reasonable person should (given enough time) be able to workout the result of any computation. This means all edge cases should be covered and there should be no undefined behavior.


Your score will be the time between your post and the first crack with a higher score being better.


Robbers will score 1 point for every answer that they crack, with a higher score being better.

Your answers should include the program which cracks the cop's answer and a proof that it does.


This is kind of just an idea right now. I have to iron out a lot of details and I will. I'll flesh it out later I just want to get it down so I don't forget it. Right now the most useful feedback would be in broad strokes. Don't worry about details, that's my job.

1: Turing equivalent is very similar to Turing completeness, but also stipulates that the language can be simulated by a Turing machine. Turing complete languages are not Turing equivalent if they are incomputable.

2: Suppose that we did have two Turing complete languages A and B such that every program halted in exactly one of the languages. We could solve the halting problem in language A by running a program in both A and B on separate threads until one of them halted, then outputting whether it was A or not.

  • 1
    \$\begingroup\$ +1, very nice idea. But "to maintain fairness, this challenge will only be open to submissions for one week after the first cop's answer is posted" doesn't make sense to me - it seems to make it less fair, by preventing anyone from competing who comes across the challenge late or just spends time on their submission. \$\endgroup\$
    – N. Virgo
    Commented Apr 3, 2018 at 4:05
  • \$\begingroup\$ @Nathaniel The problem is that answers are scored based on how long they last answers that are late to the party have a huge advantage, since people won't be monitoring the challenge as much the older it gets. I would love to have a challenge that is open indefinitely, I strive to make that for all my challenges, so if anyone can think of a objective winning criterion that allows for this I'm all ears. \$\endgroup\$
    – Wheat Wizard Mod
    Commented Apr 3, 2018 at 4:08
  • \$\begingroup\$ I see, hmm, that makes sense even if it's unsatisfying. Nothing comes to mind as an alternative scoring mechanism for this challenge for now. \$\endgroup\$
    – N. Virgo
    Commented Apr 3, 2018 at 4:10
  • \$\begingroup\$ A separate point: I wonder if this might be a bit hard for the cops. Off the top of my head, I find it hard to think of a way to design two languages such that it wouldn't be obvious how to crack it. This might just be a failure of my imagination, however! \$\endgroup\$
    – N. Virgo
    Commented Apr 3, 2018 at 4:12
  • \$\begingroup\$ What do you mean with input [a program] into your two languages? Is this about polyglots? If so, how would you consider programs which are syntax errors in one or both of the languages? \$\endgroup\$
    – Leo
    Commented Apr 3, 2018 at 4:25
  • \$\begingroup\$ @Leo Input should probably be replaced with the verb "run". This is about polyglots. This is one of the things I plan on expanding on later on, but I'll either disallow syntax errors or consider them to be halting. \$\endgroup\$
    – Wheat Wizard Mod
    Commented Apr 3, 2018 at 4:34
  • 2
    \$\begingroup\$ Maybe avoid encrypting? (If key doesn't match, A always halt and B never halt; you hve no key lol \$\endgroup\$
    – l4m2
    Commented Apr 3, 2018 at 5:23
  • \$\begingroup\$ but once known turing complete it's usually easily converted to an ans \$\endgroup\$
    – l4m2
    Commented Apr 3, 2018 at 5:25
  • \$\begingroup\$ Although languages A and B such that all programs halt in exactly one language is impossible, A and B such that can't find one is possible \$\endgroup\$
    – l4m2
    Commented Apr 3, 2018 at 7:24
  • 2
    \$\begingroup\$ With respect to @Nathaniel's observation about "only open for a week", it's certainly true that one of the problems of C'n'R is that the robbers get bored or don't notice late cop submissions, but it's not clear what "open to submissions" means. You're surely not thinking of asking the mods to lock the question, so you can't stop people posting new submissions. Perhaps that would be better rephrased in terms of when you will award the "accepted answer". \$\endgroup\$ Commented Apr 3, 2018 at 8:47
  • \$\begingroup\$ The issue of testability is also a big one. If B is something along the lines of "Run the program through A for N steps. If it terminates, enter an infinite loop. Otherwise take the parts which are comments in language A and run them." Tune N so that it takes about a week to get any program to terminate in B. \$\endgroup\$ Commented Apr 3, 2018 at 8:52
  • \$\begingroup\$ @l4m2 I would rather not ban encrypting because I consider it an unenforcable restriction. As you pointed out proving an answer to be Turing complete becomes a large hurdle for cops to overcome if they choose to cryptography. For example it is unknown whether Malbodge Unshackled is TC. \$\endgroup\$
    – Wheat Wizard Mod
    Commented Apr 3, 2018 at 17:00
  • 1
    \$\begingroup\$ @tsh The problem with bubblegum or similar languages is that crypto functions are designed to be really hard to predict. This makes it hard to say anything concrete about their behavior, which in turn makes any language that uses them in a non-trivial manner, really hard to prove TC. TCness means it is possible to write a program for every Turing machine, and since it is almost impossible to write a program in the language it is hard to show that every program can be written. I would recommend trying to come up with a language pair that has these properties. You might run into some issues. \$\endgroup\$
    – Wheat Wizard Mod
    Commented Apr 8, 2018 at 16:05
  • 1
    \$\begingroup\$ @tsh There is no such thing as undefined behavior. \$\endgroup\$
    – Wheat Wizard Mod
    Commented Apr 9, 2018 at 2:30
  • 1
    \$\begingroup\$ Another couple of related questions: (1) to what extent does the interpreter have to meet the spec? Obviously it can't meet it completely, since the interpreter runs on a finite machine with finite addressing, and therefore can't technically be Turing complete. What about cases like integer overflow? (2) How would you handle bugs in the interpreter, and/or ambiguities in the spec? \$\endgroup\$
    – N. Virgo
    Commented Apr 9, 2018 at 3:57

Really skirting bit of sandboxing here (I haven't taken the time to develop the question):

I'd like to do a 2-d variant of this one.

Two lasers between two mirrors

The input would be an ascii maze, with a marker for the laser starting point, e.g.:

|*  |      |
|   |      |
|   |  ----+
|   |      |
|   |      |
|   |      |
|   +---   |
|          |
|          |
|          |
  • Astrisk(*): laser starting point.
  • Whitespace: passable.
  • Any other character is a reflective obstacle.


|\  |   /\ |
| \ |  /  \|
|  \| /----+
|  /|/   /\|
| / |\  / /|
|/  | \/ / |
|\  +---/  |
| \    /   |
|  \  /    |
|   \/     |


  • When you can't move horizontally, reverse horizontal direction.
  • When you can't move vertically, reverse vertical direction.
  • Stop going when you meet the same laser coming the other way (e.g. when you hit a corner or reach your starting position again).I'm thinking of a golfing challenge.
  • If the laser revisits a space in the other direction, draw an upper case X.

Simplifying factors:

  • Let's just say the laser always goes the same way.
  • The map is always a rectangle.
  • Just the one laser.

It'll probably just be a golfing challenge.

  • 1
    \$\begingroup\$ What's the goal of the program, just to draw the path? Food for thought, if you wanted a related but different challenge you could include a "start" and "goal" marker, and make the program determine if the laser ever hits the goal, returning "true" or "false" instead of a drawn maze. Or perhaps it should return how many steps it takes to hit the goal, returning either an integer, or a falsy value if it never hits it? Up to you, but some ideas. \$\endgroup\$
    – BradC
    Commented Apr 11, 2018 at 16:34
  • 3
    \$\begingroup\$ Consider removing the outside wall. You say that the laser always goes in the same direction, but I think that you should explicitly say that it always starts heading south-east (or down-right). Important test cases: All of the walls aren't connected to the border. Laser hits a 1-length wall. Laser crosses itself. Laser hits a + (on the inside). Laser barely misses a +. Laser starts next a right wall. Laser starts next to bottom wall. Laser starts at a bottom-right corner. \$\endgroup\$ Commented Apr 11, 2018 at 17:08

X1M4L got in before me and posted the very similar challenge Print the previous answer!, which I don't mind at all. However, that one seems to have turned out quite a lot more open-ended than this version would be, with answers tending to score in the tens of thousands. So after a sufficiently long delay this will probably not be considered a duplicate - I intend to leave it in the sandbox for at least a few months before posting.

Compress the previous answer

This is an challenge. The first answer must output the empty string. The second answer must output the source code of the first answer, and so on, with each answer outputting the code of the previous one.

To make this challenging, no answer may be longer than 100 bytes in length. Once the answers start getting close to this limit, it will become necessary to compress the text of the previous answers, which itself will become harder over time.

Each answer should output the exact string of bytes that forms the previous program. Because of this, if your program is in any format other than plain ASCII, you should post a hexdump as well as the source, so that the next person knows exactly what to output.

An answer's score is its position in the chain, i.e. the Nth answer scores N points. Your score is the score of your highest-scoring answer. You are encouraged to treat this challenge cooperatively, keeping the chain going as long as possible by not adding extra unnecessary entropy to your code.

The code of your answer cannot be identical to the code of any previous answer. The first answer must be at least one byte in length.

You may not post two consecutive answers. There are no restrictions on language.


  • \$\begingroup\$ that's so weird, I was thinking of this the other day, after seeing all those quine/answer chaining challenges. \$\endgroup\$
    – geokavel
    Commented Mar 21, 2018 at 16:13
  • 1
    \$\begingroup\$ You'd probably need to provide some non-empty seed, to prevent the answers from all being blank themselves. Also, would there be any restrictions on language choice - e.g. would each answer need to be in a different language than all those that came before? \$\endgroup\$
    – Sok
    Commented Mar 21, 2018 at 16:29
  • \$\begingroup\$ @user56656 just position in the chain - I've added it \$\endgroup\$
    – N. Virgo
    Commented Mar 22, 2018 at 0:30
  • \$\begingroup\$ @Sok I've fixed the "all programs empty" issue by requiring each entry to be different from the previous one. I tend towards not putting restrictions on language choice (I've added that for now) but if there's a good argument for including such restrictions I'm open minded. \$\endgroup\$
    – N. Virgo
    Commented Mar 22, 2018 at 0:31
  • \$\begingroup\$ You should probably prevent source reading. A pretty good strategy will to be to make a valid submission in language A that is also a period 2 quine in language B. Alternatively, you might want to impose the restriction that new answers are different from every previous answer instead of just the last one. \$\endgroup\$
    – Wheat Wizard Mod
    Commented Mar 22, 2018 at 0:33
  • \$\begingroup\$ Another consideration is whether consecutive answers should be permitted. It is pretty common to put restrictions on individual answerers, to prevent schemes designed to game the system. \$\endgroup\$
    – Wheat Wizard Mod
    Commented Mar 22, 2018 at 0:35
  • \$\begingroup\$ @user56656 I've implemented your idea about making them distinct from all previous answers, and I've banned consecutive answers by the same person. Does that cover all the bases? \$\endgroup\$
    – N. Virgo
    Commented Mar 22, 2018 at 0:38
  • \$\begingroup\$ I've also changed the length restriction to 100 (makes a shorter challenge, but the restriction will kick in sooner) \$\endgroup\$
    – N. Virgo
    Commented Mar 22, 2018 at 0:39
  • 1
    \$\begingroup\$ It might be prudent to try a test drive of the challenge (in chat) to test if the 100 byte limit is good. I'd be willing to help with that. Other than that I have no further thoughts at this time. \$\endgroup\$
    – Wheat Wizard Mod
    Commented Mar 22, 2018 at 0:43
  • \$\begingroup\$ @user56656 that makes sense - I'll try that when I have a chance (might be in a week or two). \$\endgroup\$
    – N. Virgo
    Commented Mar 22, 2018 at 5:40
  • \$\begingroup\$ I think you probably gain more from saying languages cannot be repeated (and including a stack snipped of used languages) than you do by adding a lot of workarounds for quines. I think it will cause the answers to become more interesting more quickly. It will also result in the challenge ending sooner, but given how the hello world answer chaining question went, I don't think you have too much to worry about with respect to longevity (though increasing the 100 may be a good idea if you decide to do this). \$\endgroup\$ Commented Mar 22, 2018 at 18:00
  • 1
    \$\begingroup\$ @FryAmTheEggman I don't think quines will be a problem in any case, because the Nth answer has to output a program that outputs a program that ... that outputs the empty string, so by definition it can't be a quine. That could be got around with polyglots, but I'd be very surprised if it would be an issue in practice - the criterion of not being identical to a previous answer is there to prevent quines not only in practice but also in theory! \$\endgroup\$
    – N. Virgo
    Commented Mar 23, 2018 at 0:59
  • \$\begingroup\$ The problem that I see is that if languages are repeatable you might get a really boring and repetitive sequence of 3-4 languages that somewhat trivialise the problem appearing quickly after each more interesting answer. I perhaps should have said "trivial modifications of near-quines" instead? Anyway, this may not happen, but I think you should focus on this as well when doing the short test run (i.e. check what portion of answers can be trivially followed by a /// or Bubblegum program). \$\endgroup\$ Commented Mar 23, 2018 at 1:31
  • 1
    \$\begingroup\$ How about making the maximum length of answer n a slowly growing function over n? \$\endgroup\$
    – Laikoni
    Commented Mar 24, 2018 at 10:43
  • \$\begingroup\$ @Laikoni that's an interesting idea. I guess probably logarithmic would be good choice. I'll give it some thought. \$\endgroup\$
    – N. Virgo
    Commented Mar 24, 2018 at 14:11

Rebuild My Scrabble Board

My daughter and I have an ongoing Scrabble tournament. We enjoy admiring the board at the end of a game. But recently we knocked over the board and didn't get a chance to get a good look or take a picture. Fortunately, we kept track of all our moves. Your challenge is to take the list of moves and use it to rebuild the Scrabble board.


Each move includes 4 pieces of information

  1. Column - A - O, like the column label of a typical spreadsheet. Upper or lower case is OK, but you only need to support one or the other.
  2. Row - 1 - 15, like the row label of a typical spreadsheet. 0-indexed 0 - 14 is acceptable.
  3. Orientation - H=Horizontal, V=Vertical - Upper or lower case is OK, or you may choose any other 2 printable ASCII characters.
  4. Tiles - A - Z plus @ for blank. Upper or lower case is OK, but you only need to support one or the other. You can use a different printable ASCII character for blank if you prefer, but not the space character or a period. The tiles for a move may be a string or a list of characters - e.g., CAT or (C,A,T).

Input format may be a tuple, list, separate lines (4 per move) or any other reasonable format. Column + Row may be combined, spreadsheet-style (e.g., A1, O15), but the orientation and tiles must be separate fields. Column can't be numeric - that is easy for the computer but in the middle of a game we have enough to keep track of, so remembering which # is Row and which # is Column would be too confusing - spreadsheet notation is easy to remember.


The completed Scrabble board is to be printed in a format similar to Draw an Empty Scrabble Board. However, the double/triple squares are not printed and the moves are, obviously, included. Specifically:

  • 15x15 matrix
  • Unused spaces display as a period .
  • No leading spaces allowed.
  • Trailing spaces allowed.
  • One line feed at the end of each row.
  • Up to one leading and one trailing line feed permitted.
  • Blank tiles are to be output using the same character as input (default @).


Each move is placed starting with the first tile in the specified location. Each additional tile is to be placed in the next available location. In other words, your program must keep track of filled locations and skip them, just like a player placing tiles on the board.

What you don't need to worry about:

  1. Invalid words - if my daughter and I decided it was OK, you don't have to check it.
  2. Invalid locations - these are real moves, so there will of course be enough space to place all the tiles, and you don't have to double-check that a move starts on an empty space (but you can do that if it makes your algorithm shorter since you do have to check for all the other tiles in each move).
  3. Proper use of the standard Scrabble set of tiles - e.g., you don't have to check that there are only 2 blanks, 1 Z, etc.


[Sandbox note: Plan to include a couple of complete games plus a few shorter examples to highlight particular issues]






  • The input and output can be given by any convenient method.
  • Either a full program or a function are acceptable. If a function, you can return the output rather than printing it.
  • Standard loopholes are forbidden.
  • This is so all usual golfing rules apply, and the shortest code (in bytes) wins.
  • 3
    \$\begingroup\$ This is a very nice challenge! :-) A few suggestions/answers: yes, allow zero indexed values. I suggest that the input may optionally be numeric for both row and column, but that's up to you. You don't need to mention the next challenge. \$\endgroup\$ Commented Apr 22, 2018 at 7:01
  • \$\begingroup\$ Should you include scrabble game rules too? \$\endgroup\$
    Commented Apr 22, 2018 at 13:12
  • \$\begingroup\$ @user202729 I am limiting the rules to those required for the challenge. There are plenty of rules (especially scoring, but also passing/trading letters and a lot of other stuff) that isn't relevant to this challenge. If this goes well and I make another challenge for scoring then I will include the relevant rules there. I believe I have enough information here that someone could produce functional code without understanding anything else about Scrabble. \$\endgroup\$ Commented Apr 22, 2018 at 13:47
  • \$\begingroup\$ Suggest to allow other characters than . for empty board tiles. \$\endgroup\$
    – JayCe
    Commented Jun 11, 2018 at 3:45
  • \$\begingroup\$ Are trailing new lines allowed? \$\endgroup\$
    – JayCe
    Commented Jun 11, 2018 at 3:47
  • \$\begingroup\$ @JayCe ok for both. I need to finish this and turn it into an actual challenge \$\endgroup\$ Commented Jun 11, 2018 at 3:53
  • 1
    \$\begingroup\$ It is a really nice challenge indeed :) \$\endgroup\$
    – JayCe
    Commented Jun 11, 2018 at 3:55

Find the minimum set of letters to buy

Imagine that you own a board like this:

Black letter board

which lets you write any message that you want, given that you have the letters for the message. Given that you have a list of messages that you want to switch between, find the minimum number of each letter that you have to order with your board.

As an example, if your messages are Hello world and Hello aliens, you would need the letters Helloworldaiens, as both messages share the Hello part and the letter l in the second word.

For this challenge, input messages will only consist of letters a-zA-Z and space, however note that spaces are not included in the output as you don't need to buy them.


Input consists of a list of strings, which all match [a-zA-Z ]+. There is no limit on the amount of messages that can be provided, or their length.


Output consists of a string of a list of characters, where the letters in the output can be rearranged to create all the input messages, and is of minimal length. The output does not need to be sorted.

['Hello world', 'Hello aliens'] => Hadeeilllnoorsw
['foo', 'bar', 'baz'] => abfoorz
['Eat more tacos', 'Drink more tequila'] => DEaaceeiiklmnooqrrsttu
['Golfing is a fun activity', 'Code should be readable'] => CGaabbcdddeeeeffghiiiillnnoorsttuvy
['a', 'b', 'c', 'd', 'e', 'f', 'g'] => abcdefg

This would of course be a code-golf challenge

  • \$\begingroup\$ Hmm ... this is ringing a bell; I think we might have had it before but asking how many of each letter was needed. \$\endgroup\$
    – Shaggy
    Commented May 9, 2018 at 21:50
  • \$\begingroup\$ I tried searching for similar questions, but I couldn't find anything. This question would also ask for how many of each letter is needed, but I guess the output format might be different. \$\endgroup\$
    – maxb
    Commented May 10, 2018 at 6:14
  • 1
    \$\begingroup\$ It's this, isn't it? \$\endgroup\$ Commented May 10, 2018 at 10:53
  • \$\begingroup\$ Darn it, here I thought I had a good challenge going, but I guess the topic was too common. Thanks for the help! \$\endgroup\$
    – maxb
    Commented May 10, 2018 at 10:55

Note: I'm putting this challenge on the back burner for a indefinite time in favor of the Hierarchies challenge. Go check that or the Formic Forest out if you're interested in a Formic sequel.

Formic Functions 3: Memory

This is a preliminary write-up of a new challenge heavily inspired by Formic Functions. The spec is based on the original challenge's spec - credits for most of what you'll read here to trichoplax.

Each player starts with one ant - a queen, who collects food. Each piece of food can be held or used to produce a worker. Workers also collect food to be brought back to the queen.

All players compete in one arena. The winner is the queen holding the most food after she has taken 8,000 10,000 [Thanks @Draco18s] turns. Ants can communicate by changing the colors of the arena squares (which can also be modified by rivals), as well as by storing messages for their peers.

The arena

The arena is a toroidal (edge wrapping) grid of hexagonal cells arranged in a rhombus of side length 1000. All cells start as color 1.

Initially exactly 1% 0.5% of cells will contain food. The 5000 pieces of food will be scattered uniformly randomly. No new food will be introduced during the game.

The queens will be placed randomly on empty cells, with no guarantee that they will not be adjacent to each other (although this is very unlikely).

Ant abilities

  • Sight: Each ant sees the 7 cells in its neighborhood. It has no knowledge of any other ants outside this neighborhood. It sees the contents of each of the 7 cells (other ants and food), and also each cell's color.
  • Memory: Each ant has access to a string as its memory. It is initially empty for the queen, and must be initialized by the queen when spawning a worker. For ways to change the memory after initialization, see Output below.
  • No orientation: An ant does not know where it is or which way it faces - it has no concept of North. The neighborhood will be presented to it at a randomly rotated orientation that changes each turn so it cannot even walk in a straight line unless it has colors to guide it. (Making the same move every turn will result in a random walk rather than a straight line.)
  • Moving, color marking, producing workers and transferring food: See Output below.
  • Immortality: These are highland ants that cannot die. You can confuse rival ants by changing the colors around them, or constrain them from moving by surrounding them with 6 ants of your own, but they cannot be harmed apart from this. [Should ants be able to die? If so, how?]
  • Carrying food: A worker can carry up to 1 piece of food. A queen can carry an arbitrary amount of food.


Provide a function body

Each ant is controlled by an ant function. Each turn the player's ant function is called separately for each ant (not just once per player, but once for the queen and once for each worker that player controls). Each turn, the ant function will receive its input and return a move for that particular ant.

Post an answer containing a code block showing the body of a JavaScript function, and it will be automatically included in the controller. The name of the player forms the title of the answer, in the form # PlayerName.

No access to outside data

Functions must be fully deterministic. When called with a given input, they must return the same output every time. A function must not access global variables and must not store state between turns in other ways than through the provided memory string. It may use built in functions that do not involve storing state or accessing data from the outside. For example, the use of Math.abs() is fine, but Date.getTime(), Math.random() must not be used.

An ant function may only use a pseudo random number generator that it supplies itself, that utilizes data provided through input. For example, it may implement its own pseudo RNG via its memory string, seeded by the environment (or statically).

A simple random strategy is still possible due to the random orientation of the input - an ant that always chooses the same direction will perform a random walk rather than a straight line path.

An ant function is permitted to contain further functions within its body.

Input and output


The orientation of the input will be chosen at random for each ant and for each turn. The input will be rotated by 0, 60, 120, 180, 240 or 300 degrees, but will never be reflected.

Cells are numbered in this order:

 0 1
5 6 2
 4 3

The ant function will receive an array called view, containing an object for each of the 7 visible cells. Each object will have the following:

color: a number from 1 to 8
food: 0 or 1
ant: null if there is no ant on that cell, or otherwise an ant object

[Is 8 colors perhaps too many?]

If a cell contains an ant, the ant object will have the following:

food: 0 or more (maximum 1 for a worker)
queen: true or false
friend: true or false
memory: memory string when friendly, otherwise undefined

[Should ants be able to read rivals' memories? This would cause rampant edit wars. Rejected.]

The ant can determine its own details by looking at the ant in the central cell, view[6].ant. For example, view[6].ant.memory contains the memory of the executing ant.


Output is returned as an object representing the action to take. This can have any of the following:

cell: a number from 0 to 6 (mandatory)
color: a number from 1 to 8 (optional)
spawn: a string (optional)
memory: a string (optional)

If color and spawn are omitted or non-truthy, then cell indicates the cell to move to.

If color is a number, the indicated cell is set to that number.

If spawn is a string, a worker ant is created on the indicated cell. The new worker will have its memory initialized to that string. The string cannot be longer than 256 characters. Only a queen can create a new worker, and only if she has food, as this costs one piece of food per worker.

If memory is a string, the executing ant will have its memory immediately changed to that string. The string cannot be longer than 65,536 characters for the queen, and 256 characters for workers. An ant may change its memory while also performing a different action - changing memory does not take a turn.

[Should changing own memory take a turn?]

[Should ants be able to send a message directly to another ant's inbox? For example, a message could look like this: {title:"help", content:view_array_of_sender}. An ant should also be able to perform an action while sending a message, otherwise the described behavior could be emulated.]

[Should ants be able to see the age of an ant? This behavior will often be emulated with memory. Is there a reason not to do that?]

[Is 65,536 characters a good number to pick for the max length of memory? Thanks to @Draco18s's and @dzaima's advice, workers now have significantly less memory than a queen.]

Example outputs:

{cell:0}: move to cell 0
{cell:5, memory:"abc"}: move to cell 5 and set own memory to "abc"
{cell:6}: move to cell 6 (that is, do nothing, as 6 is the central cell)
{cell:6, color:8}: set own cell to color 8
{cell:2, color:1, memory:"hey"}: set cell 2 to color 1 and set own memory to "hey"
{cell:1, spawn:"def", memory:"5252"}: create a worker with its memory initialized to "def" on cell 1 and set own memory to "5252"
{cell:3, color:0}: equivalent to just `{cell:3}` - move rather than set color
{cell:1, spawn:0}: equivalent to just `{cell:1}` - move rather than create worker
{cell:4, color:0, spawn:0}: move to cell 4 - color 0 and type 0 are ignored

Invalid outputs:

{cell:7}: cell must be from 0 to 6
{cell:0, color:9}: color must be from 1 to 8
{cell:0, spawn:true}: spawn must be a string
{cell:6, spawn:"254"}: cannot create a worker on a non-empty cell
{cell:0, color:1, spawn:"77"}: cannot set color and create worker in the same turn
{cell:3, memory:true}: cannot set memory to non-string
{cell:2, spawn:true}: cannot create a worker with a non-string memory
{cell:0, memory:long_string}: (if long_string is a string of length > 65536) cannot set memory to a string of length greater than 65536

[Missed any?]

An ant moving onto a cell containing food will automatically pick up the piece of food. If that ant is a laden worker, it will move onto the cell without picking up the piece of food.

An unladen worker trying to move onto an enemy queen with food will steal one piece of food from her instead. A laden worker trying to move onto an unladen friendly worker or a friendly queen will give its food to their target instead.

Turn order

Ants take turns in a set order. At the start of a game the queens are assigned a random order which does not change for the rest of the game. When a queen creates a worker, that worker is inserted into the turn order at the position before its queen. This means that all other ants belonging to all players will move exactly once before the new worker takes its first turn.

Limit on number of players

Obviously an unlimited number of players cannot fit into the arena. If there are more than 8 answers, only 8 of them will play in any one game.

[Good max number of players?]

Time limit per turn

Each time the ant function is called, it should return within 20 milliseconds. Since the time limit may be exceeded due to fluctuations outside the ant function's control, an average will be calculated. If at any point the average is above 20 milliseconds and the total time taken by that particular ant function across all calls so far is more than 10 seconds, the relevant player will be disqualified.

[Enough time?]


This means the player will not be eligible to win and their ant function will not be called again during that game. They will also not be included in any further games. If a player is disqualified on the tournament machine during a leaderboard game then it will be excluded from all future leaderboard games until edited.

A player will be disqualified for any of the following for any of its ants (queen or worker):

  • Exceeding the time limit as described.
  • Returning an invalid move as described under Output.
  • The cell to move to contains an ant and the case isn't defined under Output.
  • The cell to produce a worker on contains an ant.
  • A worker is trying to produce a worker.

[Did I miss any?]

It may seem harsh to disqualify for invalid moves, rather than simply interpreting this as no move. However, I believe that enforcing correct implementations will lead to more interesting strategies over time. This is not intended to be an additional challenge, so a clear reason will be displayed when a player is disqualified, with the specific input and output alongside to aid in fixing the code.

Multiple answers and editing

You may provide multiple answers, provided that they do not team up against the others. Provided each answer is working solely towards its own victory, you are permitted to tailor your strategy to take advantage of weaknesses in specific other strategies, including changing the color of the cells to confuse or manipulate them. Bear in mind that as more answers come in, the likelihood of meeting any particular player in a given game will diminish.

You may also edit your answers whenever you choose. It is up to you whether you post a new answer or edit an existing one. Provided the game is not flooded with many near-identical variations, there should be no problem.

If you make a variation of another person's answer, please remember to give them credit by linking to their answer from yours.


At the end of each game, a player's score is the number of other players who have less food carried by their queen. Food carried by workers is not counted. This score is added to the leaderboard, which is displayed in order of average score per game.

Joint places indicate that the order of players is not yet consistent between 6 subsets of the games played so far. The list of games is split into 6 subsets because this is the minimum number that will give a probability of less than 5% that a given pair of players will be assigned distinct places in the wrong order.

[Scoring mechanism subject to change.]

[New feature/modification recommendations are very welcome! I want this challenge to be as different from the original Formic as possible, while keeping its spirit.]

  • 2
    \$\begingroup\$ 8000 turns is too few, 65k of memory is way too much. \$\endgroup\$ Commented Mar 7, 2018 at 22:12
  • \$\begingroup\$ I feel like the food density being so high is compensation enough for 8k moves. This was also chosen out of speed concerns - no caching like the previous Formic. 65k memory was designed as a "use all you want". I wanted to allow doing some crazy stuff, like storing large portions of the map. After some consideration, however, limiting this to for example 1k could be a good idea. \$\endgroup\$
    – Alion
    Commented Mar 7, 2018 at 22:21
  • 1
    \$\begingroup\$ One thing that might make this challenge stand out would be if food "respawns" over time. That is, every N turns Y food is added to the board. Also, I still don't like the hex layout for the reasons dzaima pointed out in chat. \$\endgroup\$ Commented Mar 8, 2018 at 0:38
  • \$\begingroup\$ That would encourage periodically re-searching parts of the board for more complex entries... It would be simple to do, considering you have a lot of memory, but... Isn't this just complexity tacked on for the sake of complexity? Hmm, I'm not sure how I feel about this. \$\endgroup\$
    – Alion
    Commented Mar 8, 2018 at 9:17
  • 1
    \$\begingroup\$ I'd suggest that the queen has the 65k memory but the workers - maybe ~500. Also do consider that each character is (IIRC) 2 bytes as JS uses UTF-16 \$\endgroup\$
    – dzaima
    Commented Mar 8, 2018 at 9:21
  • \$\begingroup\$ I was just thinking about more memory for the queen. Great idea! \$\endgroup\$
    – Alion
    Commented Mar 8, 2018 at 9:22
  • 3
    \$\begingroup\$ Saying that a "grid of hexagonal cells" is "toroidal" and has a specific "height" and "width" is somewhat ambiguous. I think I understand now, but it bears some more detailed explanation. \$\endgroup\$ Commented Mar 8, 2018 at 15:42
  • \$\begingroup\$ Oh, good point. I thought it was obvious enough. I'll find a way to explain it better. \$\endgroup\$
    – Alion
    Commented Mar 8, 2018 at 16:41
  • \$\begingroup\$ @KamilDrakari Is it clear enough now? \$\endgroup\$
    – Alion
    Commented Mar 21, 2018 at 16:48
16 17
19 20

You must log in to answer this question.

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