# Nest a string inside an array n times

You must produce a function which nests a string s inside an array, n times

>>> N("stackoverflow",2)
[['stackoverflow']]


# Parameters:

1. s - An ascii string
2. n - An integer >= 0

## Rules

• Shortest code wins.
• The output will be a nested array, list or tuple (or similar type based off an array)

## Test Cases

>>> N("stackoverflow",0)
'stackoverflow'
>>> N("stackoverflow",1)
['stackoverflow']
>>> N("stackoverflow",5)
[[[[['stackoverflow']]]]]


Inspired by: Nesting a string inside a list n times ie list of a list of a list

• Does the output have to be a list, or can it be a string representing that list? – clismique Oct 17 '16 at 10:53
• Can we take the parameters in any order? – Socratic Phoenix Oct 17 '16 at 10:56
• @SocraticPhoenix I think unless explicitly forbidden, yes - you can take the input in any reasonable format (which would include taking the two as a list too I believe). Maybe someone more experienced can point to a relevant meta post. – Jonathan Allan Oct 17 '16 at 11:37
• Will the string ever include an escaped "? E.g. N("stack\"overflow",5) – Riley Oct 17 '16 at 14:06
• @Riley It could contain any ascii character – jamylak Oct 17 '16 at 23:48

# Jelly, 2 bytes

W¡


Slightly confusing, since: (1) Jelly has no strings, only lists of characters; and (2); the output wont show the nesting. To see that this actually is doing what is asked look at a Python string representation of the result with:

W¡ŒṘ


An extra pair of [] will be present since the string itself will be a list of characters. For example

### How?

W¡ - Main link: s, n
W  - wrap left, initially s, in a list
¡ - repeat previous link n times


W¡ŒṘ - Main link: s, n
ŒṘ - Python string representation


# Java and C#, 62 bytes

Object f(String s,int n){return n<1?s:new Object[]{f(s,n-1)};}


Should work without modification in both Java and C#.

• Smart! +1 I was trying to get it to work in Java by nesting a String-array, which didn't really work out. Using an Object as return type and nest it in an Object[] is just the solution required for this challenge, since Object[] (or any array) is also an Object themselves. Nice one. – Kevin Cruijssen Oct 17 '16 at 13:19

# 05AB1E, 3 bytes

Code

F)


Explanation

   # Flatten the input array on the stack.
F  # Element_2 times do:
) # Wrap the total stack into a single array.


This means that this also works for the 0-testcase, since the string is already on the stack.

Try it online!

## JavaScript (ES6), 20 bytes

d=>g=n=>n--?[g(n)]:d


Although people normally nag me to curry my functions for the 1-byte saving, this is a case where it actually contributes to the solution.

• Great use of currying. I think you can make it slightly more readable: d=>g=n=>n?[g(n-1)]:d – ETHproductions Oct 17 '16 at 22:59

## Mathematica, 13 bytes

List~Nest~##&


# CJam, 7 6 bytes

{{a}*}


Online interpreter

This is an unnamed function that takes its arguments from the stack as S N, S being the string and N being the wraps. You can execute it with the ~ operator, meaning eval.

Explanation:

{{a}*}
{      Open block    [A B]
{     Open block    [A]
a    Wrap in array [[A]]
}   Close block   [A B λwrap]
*  Repeat        [A:wrap(*B)]
} Close block   ["S" N λ(λwrap)repeat]

• Just use an unnamed block to avoid the awkward input format {{a}*} or {'a*~}. – Martin Ender Oct 17 '16 at 13:06
• @MartinEnder I'm afraid it would take bytes, though, and I think that input format is 100% acceptable. It's just a list, and I think there is no restriction as to how those two parameters are entered. Also, I never named the block. – Erik the Outgolfer Oct 17 '16 at 13:13
• I don't know what you mean about the bytes? Both of those solutions are only 6 bytes. – Martin Ender Oct 17 '16 at 13:14
• @MartinEnder Oh, were these whole solutions? I thought you were talking about extending my program, but you just converted it to a function? Well, that changes the whole point. I'm a newbie at CJam/GolfScript/Pyth. I prefer the first one because it's more comprehensible (repeat {a} n times) instead of the second one (produce a string of n as and execute it). – Erik the Outgolfer Oct 17 '16 at 13:16

# Javascript ES6, 23 bytes

Recursive function

f=(a,i)=>i?f([a],--i):a

console.log(f("stackoverflow",0))
console.log(f("stackoverflow",1))
console.log(f("stackoverflow",2))
console.log(f("stackoverflow",5))

Currying results in the same length

f=a=>i=>i?f([a])(--i):a


# Brachylog, 10 bytes

tT,?h:T:gi


Try it online!

### Explanation

tT,            T is the integer (second element of the Input)
?h:T:g      The list [String, T, built-in_group]
i     Iterate: Apply built-in_group T times to String


This would be 3 bytes if it wasn't bugged. Here we need all this to get the list [String, T, built-in_group] even though [String, T] is already our input.

Unfortunately :g directly results in [[String, T], built-in_group], which is not recognized properly by i because the integer T is inside the first list.

# MATL, 6 bytes

ji:"Xh


This produces a nested cell array as the output. With MATL's default display, however, you can't necessary see that's what it is since it won't show all of the curly braces. The demo below is a slightly modified version which shows the string representation of the output.

ji:"Xh]&D


Try it Online

Explanation

j       % Explicitly grab the first input as a string
i       % Explicitly grab the second input as an integer (n)
:"      % Create an array [1...n] and loop through it
Xh  % Each time through the loop place the entire stack into a cell array
% Implicit end of for loop and display


## Pyke, 3 bytes

V]1


Try it here!

# Pyth, 3 bytes

]Fw


This will output something like ...[[[[['string']]]]].... It will not quote for zero depth: string.

Explanation:

]Fw
Q Implicit: Eval first input line
]    Function: Wrap in array
w  Input line
F   Apply multiple times


If you want quoting on zero depth, use this 4-byte solution instead (explanation):

]Fw
Q Implicit: Eval first input line
]    Function: Wrap in array
w  Input line
F   Apply multiple times
Representation


# PHP, 60 Bytes

for($r=$argv[1];$i++<$argv[2];)$r=[$r];echo json_encode($r);  48 Bytes if it looks only like the task for($r=$argv[1];$i++<$argv[2];)$r="[$r]";echo$r;

• I think a direct rewrite of the question owner's own Python answer is still the shortest in PHP too: function f($s,$n){return$n?[f($s,$n-1)]:$s;}. – manatwork Oct 17 '16 at 11:41
• print_r() and, if you don't like that option, serialize() are both shorter than json_encode()ing it while differentiating the output. – user59178 Oct 17 '16 at 11:56
• BTW, that lonely ') at the end of code looks strange. – manatwork Oct 17 '16 at 12:46
• @manatwork Copy and Paste error Thank You – Jörg Hülsermann Oct 17 '16 at 13:09

# Ruby: 23 bytes

->n,s{n.times{s=[s]};s}


This is updated to make it a callable Proc rather than the original snippet. I'd be interested to know whether there's a way to have s implicitly returned rather than having to explicitly return it.

• Generally your "a few more words" should be an explanation of how your code works. But it's a good answer nonetheless. – wizzwizz4 Oct 17 '16 at 17:10
• “You must produce a function” This is a code snippet. Unless explicitly specified otherwise, input and output has to be handled explicitly by the code or implicitly by the interpreter if it has such feature. You can not expect some global variables to be set and you can not just leave the result in some global variables. – manatwork Oct 17 '16 at 17:14
• Welcome to PPCG! All answers should be callable functions or full programs, though. In your case, the shortest fix would be to use an unnamed function like ->s,n{...}. – Martin Ender Oct 17 '16 at 17:14
• @wizzwizz4, and Martin, thank you for your encouragement and helpful input, I have learned something and will update. manatwork, I've got thick skin and have plenty of points on SO but you know that blunt statements like that scare newbies away from Stack sites and intimidate them. Seems a shame no? – Peter Nixey Oct 18 '16 at 14:13

## C, 44 bytes, 41 bytes

int*n(int*s,int a){return a?n(&s,a-1):s;}


You can test it by doing the following:

int main(void) {
char* s = "stackoverflow";

/* Test Case 0 */
int* a = n(s,0);
printf("'%s'\n", a);

/* Test Case 1 */
int* b = n(s,1);
printf("['%s']\n", *b);

/* Test Case 2 */
int** c = n(s,2);
printf("[['%s']]\n", **c);

/* Test Case 3 */
int*** d = n(s,3);
printf("[[['%s']]]\n", ***d);

/* Test Case 4 */
int********** e = n(s,10);
printf("[[[[[[[[[['%s']]]]]]]]]]\n", **********e);

return 0;
}


The output:

'stackoverflow'
['stackoverflow']
[['stackoverflow']]
[[['stackoverflow']]]
[[[[[[[[[['stackoverflow']]]]]]]]]]


Of course, you'll get warnings. This works on gcc on bash on my Windows machine (gcc version 4.8.4 (Ubuntu 4.8.4-2ubuntu1~14.04.3), as well as on a true Linux machine (gcc version 4.6.3 (Ubuntu/Linaro 4.6.3-1ubuntu5)).

• Not sure about other compilers, but int*n(s,a)int*s;{return!a?s:n(&s,a-1);} works with gcc. – Dennis Oct 17 '16 at 22:38
• It segfaults for cc -v -> Apple LLVM version 8.0.0 (clang-800.0.38). – nimi Oct 17 '16 at 22:51
• Can you drop the ! from the ternary condition and switch the order of s and n(&s,a-1) to save a byte? – Riley Oct 18 '16 at 14:53
• @VolAnd When you call n(s,6), you have to change *** to ****** in the variable declaration and use. This is needed exactly because the function does what it is expected to do: nest the string into an array several (here: 6) times. Of course you would still get three levels of [] because they are hardcoded. I think the program shouldn't output them at all. This challenge is not about brackets, it is about nesting. Some languages print arrays with brackets, C hasn't any builtin function to print them at all. So what? It is not needed here. – Christian Sievers Oct 19 '16 at 2:58
• Can you drop the spaces after the * in the function signature? – Fund Monica's Lawsuit Oct 19 '16 at 4:04

# Python, 32 bytes

N=lambda s,n:n and[N(s,n-1)]or s


# Ruby, 25 characters

Rewrite of jamylak's Python solution.

f=->s,n{n>0?[f[s,n-1]]:s}


Sample run:

irb(main):001:0> f=->s,n{n>0?[f[s,n-1]]:s}
=> #<Proc:0x00000002006e80@(irb):1 (lambda)>

irb(main):002:0> f["stackoverflow",0]
=> "stackoverflow"

irb(main):003:0> f["stackoverflow",1]
=> ["stackoverflow"]

irb(main):004:0> f["stackoverflow",5]
=> [[[[["stackoverflow"]]]]]


## C# 6, 50 bytes

dynamic a(dynamic s,int n)=>n<2?s:a(new[]{s},n-1);

• Shouldn't it be n<1? Also -2 bytes if you use object instead of dynamic. – milk Oct 17 '16 at 20:26

# Ruby, 24 bytes

f=->*s,n{s[n]||f[s,n-1]}


Called the same as in manatwork's answer, but a weirder implementation. *s wraps the input (a possibly-nested string) in an array. Then if n is zero, s[n] returns the first element of s, turning the function into a no-op. Otherwise, it returns nil since s will only ever have one element, so we pass through to the recursive call.

# V, 6 bytes

Àñys$]  Try it online! Explanation: À "Arg1 times ñ "repeat: ys$  "surround this line
] "with square brackets


# Perl 6, 23 bytes

{($^a,{[$_]}...*)[$^b]}  ## Expanded: { # bare block lambda with two placeholder parameters ｢$a｣ and ｢$b｣ ( # generate Sequence$^a,       # declare first input
{ [ $_ ] } # lambda that adds one array layer ... # do that until * # Whatever )[$^b ]     # index into the sequence
}

• Perl never ceases to amaze me with its syntax – Fund Monica's Lawsuit Oct 19 '16 at 4:06

## Agda, 173 bytes

Since the return type of the function depends on the number given as argument, this is clearly a case where a dependently typed language should be used. Unfortunately, golfing isn't easy in a language where you have to import naturals and lists to use them. On the plus side, they use suc where I would have expected the verbose succ. So here is my code:

module l where
open import Data.List
open import Data.Nat
L : ℕ -> Set -> Set
L 0 a = a
L(suc n)a = List(L n a)
f : ∀ n{a}-> a -> L n a
f 0 x = x
f(suc n)x = [ f n x ]


(I hope I found all places where spaces can be omitted.) L is a type function that given a natural n and a type a returns the type of n times nested lists of a, so L 3 Bool would be the type of lists of lists of lists of Bool (if we had imported Bool). This allows us to express the type of our function as (n : ℕ) -> {a : Set} -> a -> L n a, where the curly braces make that argument implicit. The code uses a shorter way to write this type. The function can now be defined in an obvious way by pattern matching on the first argument.

Loading this file with an .agda extension into emacs allows to use C-c C-n (evaluate term to normal form), input for example f 2 3 and get the correct answer in an awkward form: (3 ∷ []) ∷ []. Now of course if you want to do that with strings you have to import them...

• Just remembered that I could write → instead of ->, but of course that increases the size of an UTF-8 encoded file. – Christian Sievers Oct 18 '16 at 12:11
• My ugly translation of this into Haskell is somewhat shorter. I have to stick to manual unary to keep it short. – dfeuer Jun 17 '19 at 23:15

# k, 3 bytes

,:/


Taken as a dyadic function, / will iteratively apply the left-hand function ,: (enlist) n times to the second argument.

Example:

k),:/[3;"hello"]
,,,"hello"


# PHP, 44 bytes

function n($s,$n){return$n?n([$s],--$n):$s;}


nothing sophisticated, just a recursive function

## Python 2, 32 bytes

lambda s,n:eval('['*n+s+']'*n)


Puts n open brackets before the string and n close brackets before it, then evals the result. If a string output is allowed, the eval can be removed.

# Actually, 4 bytes

Input is string then n. Golfing suggestions welcome. Try it online!

kn


Ungolfing

          Implicit input string, then n.
...n    Run the function n times.
k         Wrap the stack in a list.
Implicit return.


## R, 39 40 bytes

EDIT: Fixed the n=0 issue thanks to @rturnbull.

Function that takes two inputs s (string) and n (nestedness) and outputs the nested list. Note that R-class list natively prints output differently than most other languages, however, is functionally similar to a key/value map (with possibly unnamed keys) or a list in python.

f=function(s,n)if(n)list(f(s,n-1))else s


Example

> f=function(s,n)if(n)list(f(s,n-1))else s
> f("hello",3)
[[1]]
[[1]][[1]]
[[1]][[1]][[1]]
[1] "hello"

> # to access the string nested 5 times in the "list-object" named "list" we can do the following
> list = f("nested string",5)
> list[[1]][[1]][[1]][[1]][[1]]
[1] "nested string"

• Very nice! It doesn't give the desired output for n=0, though. Before I saw your answer, I came up with a recursive solution which can deal with n=0, but it's 1 byte longer than your solution (40 bytes): f=function(s,n)if(n)list(f(s,n-1))else s – rturnbull Oct 18 '16 at 11:07
• @rturnbull You're of course right. Your solution is much more elegant in my opinion and I totally forgot about the n=0 case. However, your solution is actually 38 bytes excluding the naming of the function and hence shorter. Great catch – Billywob Oct 18 '16 at 11:13
• Since it's a recursive function it must be named, unfortunately! (Otherwise it can't interpret the f(s,n-1) call inside of it.) Recursive anonymous functions are not possible in R, as far as I know. – rturnbull Oct 18 '16 at 11:33
• @rturnbull You're again right. Updating the answer. – Billywob Oct 18 '16 at 11:36
• A year later, I've golfed off another byte: f=function(s,n)'if'(n,list(f(s,n-1)),s). – rturnbull Nov 15 '17 at 20:01

## Racket 83 bytes

(for((c n))(set! s(apply string-append(if(= c 0)(list"[\'"s"\']")(list"["s"]")))))s


Ungolfed:

(define (f s n)
(for ((c n))
(set! s (apply string-append
(if (= c 0)
(list "[\'" s "\']")
(list "[" s "]"))
)))
s)


Testing:

(f "test" 3)


Output:

"[[['test']]]"


data L=N[Char]|C L
f 0=N
f n=C. f(n-1)


Haskell's strict type system prevents returning different types (Strings vs. List of Strings vs. List of List of Strings,...), so I have to define my own type that accommodates all those cases. The main function f recursively calls n times the constructor C for nesting and N for the base case.

Usage example (with deriving (Show) added to the new data type to be able to print it): f 4 "codegolf" -> C (C (C (C (N "codegolf")))).

Edit: @Christian Sievers saved 2 bytes by rewriting the function in a point-free style for the string argument. Thanks!

• Of course Haskell's lists can be nested, but a function cannot return a string for one value and a list of lists of strings for another value of the same type. Golfing the additional deriving clause: the parens aren't needed. - Not sure if it's okay to only nest the C constructor which isn't list-like. My very similar attempt was based on a data type defined as data D x=J x|L[D x]. – Christian Sievers Oct 18 '16 at 7:10
• If you reverse the order of the arguments and don't use an infix operator, you don't need to mention the second argument: f 0=N;f n=C. f(n-1) – Christian Sievers Oct 18 '16 at 7:15
• @ChristianSievers: yes, you're right, my explanation about nested lists was not accurate - I've changed it. Regarding list-likeness: I think my data structure is list-like. Compare a native Haskell list 1:(2:(3:([]))) with C (C (C (N "codegolf"))). C is cons (:), N is nil ([]). – nimi Oct 18 '16 at 14:14
• C doesn't cons, it only embeds, your data type can't express [["a","b"],["c"]]. But maybe that is fine as this problem only needs singletons. - f n=... isn't point-free. Point-reduced? – Christian Sievers Oct 18 '16 at 14:32
• You spend 19 characters defining your data type. Wouldn't it be more sensible to use an existing type (e.g. Either) even if it meant the constructors were a little more verbose? – Periata Breatta Oct 18 '16 at 17:07

## tinylisp (repl), 34 bytes

(d F(q((S N)(i N(F(c S())(s N 1))S


Defines a function F. Technically, tinylisp doesn't have strings, but this code will work for any data type it's given.

Ungolfed (key to builtins: d = define, q = quote, i = if, c = cons, s = subtract):

(d nest
(q
((item number)
(i number
(nest (c item ()) (s number 1))
item))))


Example usage:

tl> (d F(q((S N)(i N(F(c S())(s N 1))S
F
tl> (F 2 3)
(((2)))
tl> (F () 1)
(())
tl> (F (q Hello!) 7)
(((((((Hello!)))))))
tl> (F c 3)
(((<builtin function tl_cons>)))


# Clojure, 24 bytes

#(nth(iterate list %)%2)


Clojure is somewhat competitive here. iterate creates a sequence of x, (f x), (f (f x)) ..., nth returns needed element.

See it online: https://ideone.com/2rQ166