Implement hyperexponentiation/tetration without the use of '^'

Question

The Challenge

Implement tetration (aka Power Tower or Hyperexponentiation) with the least amount of characters.

The Conditions

• Don't use the 'power' operator or its equivalents (such as pow(x,y), x^y, x**y, etc.)
• Input given as: x y (separated by a space)
• x is exponentiated by itself y times.
• Your method must be able to compute at least 4 3 (4 exponentiated by itself 3 times)

The Scoring

• Lowest score wins: (# of characters) + (10 * (# of 'power' operators))

• Bonus deduction if you do not use the multiplication operator (-5 points).

• No Speed/Memory requirements. Take as long as you want.

Examples

x 0 = 1

2 2 = 2^2 = 4

2 4 = 2^(2^(2^2)) = 65536

4 3 = 4^(4^4) = 4^256 = 13407807929942597099574024998205846127479365820592393377723561443721764030073546976801874298166903427690031858186486050853753882811946569946433649006084096

Open to suggestions/alterations/questions

Answer 1

J, 16 19 12 characters

*/@$~/1,~$~/

or as a verb (17 characters):

h=:[:*/@$~/1,~$~/

usage:

   h 2 4
65536

or taking input from keyboard (24 27 20 characters):

*/@$~/1,~$~/".1!:1]1

with thanks to FUZxxl for pointing out my stupidity. :-)

Explanation:

J is read from right to left, so using 2 4:

/ is used to insert the verb $~ between each pair of items in the list. $~ takes the left item and shapes it $ using the right item (the ~ reverses the arguments) - so this would be equivalent to 4 $ 2 which gives you a list of 2s which is four items long 2 2 2 2.

Now we append 1 to the list 1,~ and then do the same thing again; / insert a verb */@$~ between each pair of items in the list. This verb starts in the same way $~ but this time it / inserts a * between each item of the newly generated list. The @ just makes sure that the */@$~ works as one verb instead of two. This gives 2 multiplied by itself enough times to be equivalent to 2^4.

J's vocabulary page - I find solving problems with J fun just because of the different way it sometimes does things.

Adding one further iteration to remove the * operator has 2 problems

• It comes out at 17 characters (+/@$~/,@$~/1,~$~/) which, even with the -5 bonus, is too long
• It runs out of memory if the number gets too large so doesn't meet the requirement of being able to calculate 4 3

Answer 2

GolfScript (24 chars - 5 = 19 points)

~\1{1{0{+}?}?}{@\+@*}:?~

is insanely slow.

(or 20 chars)

~\1{1{*}?}{@\+@*}:?~

is much faster.

Answer 3

Ruby, 66 59 characters

def e(x,y)
r=1
(1..y).each{t=x
(2..r).each{t*=x}
r=t}
r
end

Answer 4

Python, 161 - 5 (no * operator) = 156

r=xrange
def m(x,y):
 i=0
 for n in r(y):i+=x
 return i
def e(x,y):
 i=1
 for n in r(1,y+1):i=m(i,x)
 return i
def t(x,y):
 i=1
 for n in r(y):i=e(x,i)
 return i

invoke:

t(2, 4)

Answer 5

Python, 112 chars

The numbers should be the 1st and 2nd argument: python this.py 4 3
** operator not used.
* used. It's quite trivial to implement, exactly like **, but costs more than 5 chars.

import sys
p=lambda y:y and x*p(y-1)or 1
t=lambda y:y>1 and p(t(y-1))or x
x,y=map(long,sys.argv[1:])
print t(y)

Answer 6

C, 117 105 99 chars

EDIT: Merged the two functions p and r into one, saving some chars.
Of 99 chars, 52 do the actual calculation (including variable definitions). The other 47 are for handling input and output.
BUG: Badly handles powers of 0 (e.g. 0 2). Should find a minimum cost fix. This isn't a bug, I forgot that 0 2 is undefined.

Successfully handles 4 3, and even gives an exact result. However, can be inaccurate for some smaller numbers.
Prints the number with a trailing .000000.

x,y,z;
double R(){return--y?!z?y=R(),R(z=1):x*R():x;}
main(){
    scanf("%d%d",&x,&y);
    printf("%f\n",R());
}

Answer 7

Perl, 61 chars

here's a bizarre one

sub t
{
  ($x,$y,$z)=@_;
  $y>1&&t($x,$y-1,eval$x."*$x"x($z-1||1))||$z
}

usage:

print t(2,4,1)

Answer 8

Haskell, 87 85 - 5 == 80 82

import Data.List
t x=genericLength.(iterate(sequence.map(const$replicate x[]))[[]]!!)

Uses neither of exponentiation, multiplication or addition (!), just list operations. Demonstration:

Prelude> :m +Data.List
Prelude Data.List> let t x=genericLength.(iterate(sequence.map(const$replicate x[]))[[]]!!)
Prelude Data.List> t 2 2
4
Prelude Data.List> t 2 4
65536
Prelude Data.List> t 4 3

...
ahm... you didn't say anything about performance or memory, did you? But given enough billions of years and some petabytes of RAM, this would still yield the correct result (genericLength can use a bigInt to count the length of the list).

Answer 9

GolfScript, 15 18 chars

~])*1\+{[]+*{*}*}*

Yes, one of the *s is a multiplication operator (exercise: which one?) so I don't qualify for the 5 char bonus. Still, it's just barely shorter than Peter's solution.

This earlier 15-char version is otherwise the same, but produces no output when the second argument is 0. Thanks to r.e.s. for spotting the bug.

~])*{[]+*{*}*}*

Answer 10

Scala:110

type B=BigInt
def r(a:B,b:B,f:(B,B)=>B):B=if(b>1)f(a,r(a,b-1,f))else a
def h(a:B,b:B)=r(a,b,r(_,_,r(_,_,(_+_))))

ungolfed:

type B=BigInt
def recursive (a:B, b:B, f:(B,B)=>B): B = 
  if (b>1) f (a, recursive (a, b-1, f)) 
  else a
recursive (2, 3, recursive (_, _, recursive (_, _, (_ + _))))

explanation:

type B=BigInt
def p (a:B, b:B):B = a+b
def m (a:B, b:B):B = if (b>1) p (a, m (a, b-1)) else a
def h (a:B, b:B):B = if (b>1) m (a, h (a, b-1)) else a
def t (a:B, b:B):B = if (b>1) h (a, t (a, b-1)) else a

plus, mul, high(:=pow), tetration all work in the same manner. The common pattern can be extracted as recursive method, which takes two BigInts and a basic function:

def r (a:B, b:B, f:(B,B)=>B):B = 
  if (b>1) f(a, r(a, b-1, f)) else a
r (4, 3, r (_,_, r(_,_, (_+_))))

The underlines are placeholder for something which gets called in this sequence, for example the addition plus(a,b)=(a+b); therefore (+) is a function which takes two arguments and adds them (a+b).

unfortunately, I get issues with the stack size. It works for small values for 4 (for example: 2) or if I reduce the depth for one step:

def h(a:B,b:B)=r(a,b,r(_,_,(_*_))) // size -7, penalty + 5
def h(a:B,b:B)=r(a,b,r(_,_,r(_,_,(_+_)))) 

The original code is 112 characters and would score, if valid, 107. Maybe I find out how to increase the stack.

The expanded algorithm can be transformed to tailrecursive calls:

type B=BigInt
def p(a:B,b:B):B=a+b
import annotation._
@tailrec
def m(a:B,b:B,c:B=0):B=if(b>0)m(a,b-1,p(a,c))else c
@tailrec
def h(a:B,b:B,c:B=1):B=if(b>0)h(a,b-1,m(a,c))else c
@tailrec
def t(a:B,b:B,c:B=1):B=if(b>0)t(a,b-1,h(a,c))else c

The tailrecursive call is longer than the original method, but didn't raise a stackoverflow in the long version - however it doesn't yield a result in reasonable time. t(2,4) is fine, but t(3,3) already was stopped by me after 5 min. However, it is very elegant, isn't it?

// 124 = 119-5 bonus
type B=BigInt
def r(a:B,b:B,c:B,f:(B,B)=>B):B=if(b>0)r(a,b-1,f(a,c),f)else c
def t(a:B,b:B)=r(a,b,1,r(_,_,1,r(_,_,0,(_+_))))

And now the same as above: use stinky multiplication (we even profit while rejecting the bonus of 5, because we save 7 characters: win=4 chars:)

// 115 without bonus
type B=BigInt
def r(a:B,b:B,c:B,f:(B,B)=>B):B=if(b>0)r(a,b-1,f(a,c),f)else c
def t(a:B,b:B)=r(a,b,1,r(_,_,1,(_*_)))

invocation:

timed ("t(4,3)")(t(4,3)) 
t(4,3): 1
scala> t(4,3)
res89: B = 13407807929942597099574024998205846127479365820592393377723561443721764030073546976801874298166903427690031858186486050853753882811946569946433649006084096

runtime: 1ms.

Answer 11

Python, 70

This uses nested eval calls, eventually producing a string "a*a*a*a...*a" which gets evaluated. Almost half of the score is wasted on getting the arguments... though I've noticed that a few other solutions don't bother with that.

a,b=map(int,raw_input().split())
exec"eval('*'.join('a'*"*b+'1'+'))'*b

Answer 12

*Perhaps the no multiplication bonus is not good enough. I think the no multiplication answer is much more interesting, but this is code golf, so it's not the best. Here's an answer without *, and a better (shorter, and better scoring) answer with it:

VBA:

90 chars, no power operators, uses multiplication = 90

Sub c(x,y)
f=IIf(y,x,1):For l=2 To y:b=x:For j=2 To f:b=b*x:Next:f=b:Next:MsgBox f
End Sub

116 chars, no power operators, no multiplication bonus (-5) = 111

Sub c(x,y)
f=IIf(y,x,1):For l=2 To y:b=x:For j=2 To f:For i=1 To x:a=a+b:Next:b=a:a=0:Next:f=b:Next:MsgBox f
End Sub

NOTE: VBA has issues printing the number when the result is very large (i.e. 4, 3), but it does calculate correctly, so if, for example, you wanted to USE that number, you would be good to go. Also, even BIGGER numbers overflow (i.e. 3, 4).

Answer 13

Javascript: 116 chars

function t(i){y=i.split(' ');a=y[0];b=y[1];return+b&&p(a,t(a+' '+(b-1)))||1}function p(a,b){return+b&&a*p(a,b-1)||1}

t('4 3') Outputs:

1.3407807929942597e+154

Answer 14

Factor, 187 characters

USING: eval io kernel locals math prettyprint sequences ;
IN: g
:: c ( y x o! -- v )
o 0 = [ x y * ] [ o 1 - o!
y x <repetition> 1 [ o c ] reduce ] if ;
contents eval( -- x y ) swap 2 c .

Before golf:

USING: eval io kernel locals math prettyprint sequences ;
IN: script

! Calculate by opcode:
!   0 => x * y, multiplication
!   1 => x ^ y, exponentiation
!   2 => x ^^ y, tetration
:: calculate ( y x opcode! -- value )
    opcode 0 = [
        x y *
    ] [
        ! Decrement the opcode. Tetration is repeated exponentiation,
        ! and exponentiation is repeated multiplication.
        opcode 1 - opcode!

        ! Do right-associative reduction. The pattern is
        !   seq reverse 1 [ swap ^ ] reduce
        ! but a repetition equals its own reverse, and 'calculate'
        ! already swaps its inputs.
        y x <repetition> 1 [ opcode calculate ] reduce
    ] if ;

contents eval( -- x y )         ! Read input.
swap 2 calculate .              ! Calculate tetration. Print result.

I did not remove the multiplication operator *. If I did so, then I would need to add some logic expressing that the sum of an empty sequence is 0, not 1. This extra logic would cost more than the -5 bonus.

---

Rule breaker, 124 + 10 = 134 characters

USING: eval kernel math.functions prettyprint sequences ;
contents eval( -- x y ) swap <repetition> 1 [ swap ^ ] reduce .

This program has a lower score, but the exponentiation operator ^ breaks the rules. The rules say "(# of characters) + (10 * (# of 'power' operators))", so I applied the +10 penalty. However, the rules also say "Don't use the 'power' operator", so any program taking this penalty does break the rules. Therefore, this program of 134 characters is not a correct answer, and I must present my longer program of 187 characters as my answer.

Answer 15

Clojure - 122 + 10 = 132 Points

(def v(read))(prn((fn h[x,y,z](if(= y 0)1(if(= y 1)z(h x(- y 1)((fn p[n m](if(= m 1)n(*'(p n(- m 1))n)))x z)))))v(read)v))

Definitely not the shortest as I rewrote the power function using multiplication. But, still, a fun, entirely recursive, example.

Run it as a script, not through REPL.

Answer 16

Python (111) (113) no *

r=lambda x,y:(x for _ in range(y));t=lambda x,y:reduce(lambda y,x:reduce(lambda x,y:sum(r(x,y)),r(x,y)),r(x,y),1)

6***3 - 36k digits))

Upd: Have to add initial value, to fit t(X,0)=1

Answer 17

Ruby, 47 46 45

t=->x,n{r=x;2.upto(n){r=([x]*r).inject :*};r}

Answer 18

Haskell 110 - 5 = 105

Tetration Peano Style. This is the most insanely slow solution possible, just a warning, but also avoids even addition.

data N=Z|S N
a&+Z=a
a&+S b=S$a&+b
_&*Z=Z
a&*S b=a&+(a&*b)
_&^Z=S Z
a&^S b=a&*(a&^b)
_&>Z=S Z
a&>S b=a&^(a&>b)

This relies on you having the patience to type out Peano numbers (and won't show the answer, If you actually want to run it, add these few lines (90 chars):

f 0=Z
f a=S$f$a-1
t Z=0
t(S a)=1+t a
main=interact$show.f.(\[x,y]->x&>y).map(f.read).words

Answer 19

Mathematica, 40 33

This doesn't quite conform to the rules but it is not in contention for the shortest code anyway, and I hope that it will be of interest to someone.

m@f_:=Fold[f,1,#2~Table~{#}]&;

m[m@Sum]

This builds a "tetration" function when it is run, but the arguments must be given in reverse order. Example:

m[m@Sum][3, 4]

1340780792994259709957402499820584612747936582059239337772356144372176 4030073546976801874298166903427690031858186486050853753882811946569946 433649006084096

Answer 20

Haskell:  58  51 chars, with or without multiplication.

i f x 1=x;i f x n=f$i f x$n-1
t=i(\o n->i(o n)n)(+)4

Ungolfed:

bump op n a = iterate (op n) n !! (fromIntegral $ a-1)
tetrate = iterate bump (+) !! 3

Shorter definition comes from inlining “bump”, and defining a custom version of “iterate”. Unfortunately the result is impossibly inefficient, but starting with (*) instead of (+) gives decent speed. In ghci:

Prelude> let i f x 1=x;i f x n=f$i f x$n-1
(0.00 secs, 1564024 bytes)
Prelude> let t=i(\o n->i(o n)n)(*)3
(0.00 secs, 1076200 bytes)
Prelude> t 4 3
13407807929942597099574024998205846127479365820592393377723561443721764030073546
976801874298166903427690031858186486050853753882811946569946433649006084096
(0.01 secs, 1081720 bytes)

Answer 21

J, score is 7 (12 chars - 5 points for avoiding multiplication)

+/@$/@$~/@$~

usage:

   4 +/@$/@$~/@$~ 3
1.34078e154
t=.+/@$/@$~/@$~  NB. define a function
   4 t 3
1.34078e154
   2 t 2
4

Just few nested folds:

• Using multiplication it would be */@$~/@$~
• Using power it would be ^/@$~ where $~ creates array, / is a fold function.

Answer 22

Common Lisp 85 chars

(lambda(b c)(let((r b)u)(dotimes(c c r)(setf u 1 r(dotimes(c b u)(setf u(* u r)))))))

I tried doing the multiplications through repeated addition, but it was way more than 5 characters. Same thing with macrolets, the declarations were not worth the gains.

Another solution, inspired by boothby's python solution. It's 1 character less than the above solution.

(lambda(a b)(eval`(*,@(loop for x below b nconc(loop for x below a nconc`(,a,a))))))

Answer 23

Python, 64

Using some of boothby's ideas.

a,b=map(int,raw_input().split())
r=1;exec"r=eval('a*'*r+'1');"*b

The result is stored in r.