Exponentiation operator
![Task](http://static.miraheze.org/rosettacodewiki/thumb/b/ba/Rcode-button-task-crushed.png/64px-Rcode-button-task-crushed.png)
You are encouraged to solve this task according to the task description, using any language you may know.
Most all programming languages have a built-in implementation of exponentiation. Re-implement integer exponentiation for both intint and floatint as both a procedure, and an operator (if your language supports operator definition).
If the language supports operator (or procedure) overloading, then an overloaded form should be provided for both intint and floatint variants.
Ada
First we declare the specifications of the two procedures and the two corresponding operators (written as functions with quoted operators as their names): <lang ada>package Integer_Exponentiation is
-- int^int procedure Exponentiate (Argument : in Integer; Exponent : in Natural; Result : out Integer); function "**" (Left : Integer; Right : Natural) return Integer;
-- real^int procedure Exponentiate (Argument : in Float; Exponent : in Integer; Result : out Float); function "**" (Left : Float; Right : Integer) return Float;
end Integer_Exponentiation;</lang>
Now we can create a test program:
<lang ada>with Ada.Float_Text_IO, Ada.Integer_Text_IO, Ada.Text_IO; with Integer_Exponentiation;
procedure Test_Integer_Exponentiation is
use Ada.Float_Text_IO, Ada.Integer_Text_IO, Ada.Text_IO; use Integer_Exponentiation; R : Float; I : Integer;
begin
Exponentiate (Argument => 2.5, Exponent => 3, Result => R); Put ("2.5 ^ 3 = "); Put (R, Fore => 2, Aft => 4, Exp => 0); New_Line;
Exponentiate (Argument => -12, Exponent => 3, Result => I); Put ("-12 ^ 3 = "); Put (I, Width => 7); New_Line;
end Test_Integer_Exponentiation;</lang>
Finally we can implement the procedures and operations:
<lang ada>package body Integer_Exponentiation is
-- int^int procedure Exponentiate (Argument : in Integer; Exponent : in Natural; Result : out Integer) is begin Result := 1; for Counter in 1 .. Exponent loop Result := Result * Argument; end loop; end Exponentiate;
function "**" (Left : Integer; Right : Natural) return Integer is Result : Integer; begin Exponentiate (Argument => Left, Exponent => Right, Result => Result); return Result; end "**";
-- real^int procedure Exponentiate (Argument : in Float; Exponent : in Integer; Result : out Float) is begin Result := 1.0; if Exponent < 0 then for Counter in Exponent .. -1 loop Result := Result / Argument; end loop; else for Counter in 1 .. Exponent loop Result := Result * Argument; end loop; end if; end Exponentiate;
function "**" (Left : Float; Right : Integer) return Float is Result : Float; begin Exponentiate (Argument => Left, Exponent => Right, Result => Result); return Result; end "**";
end Integer_Exponentiation;</lang>
ALGOL 68
<lang algol68>main:(
INT two=2, thirty=30; # test constants # PROC VOID undefined;
- First implement exponentiation using a rather slow but sure FOR loop #
PROC int pow = (INT base, exponent)INT: ( # PROC cannot be over loaded # IF exponent<0 THEN undefined FI; INT out:=( exponent=0 | 1 | base ); FROM 2 TO exponent DO out*:=base OD; out );
printf(($" One Gibi-unit is: int pow("g(0)","g(0)")="g(0)" - (cost: "g(0) " INT multiplications)"l$,two, thirty, int pow(two,thirty),thirty-1));
- implement exponentiation using a faster binary technique and WHILE LOOP #
OP ** = (INT base, exponent)INT: ( BITS binary exponent:=BIN exponent ; # do exponent arithmetic in binary # INT out := IF bits width ELEM binary exponent THEN base ELSE 1 FI; INT sq := IF exponent < 0 THEN undefined; ~ ELSE base FI;
WHILE binary exponent := binary exponent SHR 1; binary exponent /= BIN 0 DO sq *:= sq; IF bits width ELEM binary exponent THEN out *:= sq FI OD; out );
printf(($" One Gibi-unit is: "g(0)"**"g(0)"="g(0)" - (cost: "g(0) " INT multiplications)"l$,two, thirty, two ** thirty,8));
OP ** = (REAL in base, INT in exponent)REAL: ( # ** INT Operator can be overloaded # REAL base := ( in exponent<0 | 1/in base | in base); INT exponent := ABS in exponent; BITS binary exponent:=BIN exponent ; # do exponent arithmetic in binary # REAL out := IF bits width ELEM binary exponent THEN base ELSE 1 FI; REAL sq := base;
WHILE binary exponent := binary exponent SHR 1; binary exponent /= BIN 0 DO sq *:= sq; IF bits width ELEM binary exponent THEN out *:= sq FI OD; out );
printf(($" One Gibi-unit is: "g(0,1)"**"g(0)"="g(0,1)" - (cost: "g(0) " REAL multiplications)"l$, 2.0, thirty, 2.0 ** thirty,8));
OP ** = (REAL base, REAL exponent)REAL: ( # ** REAL Operator can be overloaded # exp(ln(base)*exponent) );
printf(($" One Gibi-unit is: "g(0,1)"**"g(0,1)"="g(0,1)" - (cost: " "depends on precision)"l$, 2.0, 30.0, 2.0 ** 30.0))
)</lang> Output
One Gibi-unit is: int pow(2,30)=1073741824 - (cost: 29 INT multiplications) One Gibi-unit is: 2**30=1073741824 - (cost: 8 INT multiplications) One Gibi-unit is: 2.0**30=1073741824.0 - (cost: 8 REAL multiplications) One Gibi-unit is: 2.0**30.0=1073741824.0 - (cost: depends on precision)
Recursive operator calls
<lang algol68>main:(
INT two=2, thirty=30; # test constants # PROC VOID undefined;
- First implement exponentiation using a rather slow but sure FOR loop #
PROC int pow = (INT base, exponent)INT: ( # PROC cannot be over loaded # IF exponent<0 THEN undefined FI; INT out:=( exponent=0 | 1 | base ); FROM 2 TO exponent DO out*:=base OD; out ); printf(($" One Gibi-unit is: int pow("g(0)","g(0)")="g(0)" - (cost: "g(0) " INT multiplications)"l$,two, thirty, int pow(two,thirty),thirty-1));
- implement exponentiation using a faster binary technique and WHILE LOOP #
OP ** = (INT base, exponent)INT: IF base = 0 THEN 0 ELIF base = 1 THEN 1 ELIF exponent = 0 THEN 1 ELIF exponent = 1 THEN base ELIF ODD exponent THEN (base*base) ** (exponent OVER 2) * base ELSE (base*base) ** (exponent OVER 2) FI;
printf(($" One Gibi-unit is: "g(0)"**"g(0)"="g(0)" - (cost: "g(0) " INT multiplications)"l$,two, thirty, two ** thirty,8)); OP ** = (REAL in base, INT in exponent)REAL: ( # ** INT Operator can be overloaded # REAL base := ( in exponent<0 | 1/in base | in base); INT exponent := ABS in exponent; IF base = 0 THEN 0 ELIF base = 1 THEN 1 ELIF exponent = 0 THEN 1 ELIF exponent = 1 THEN base ELIF ODD exponent THEN (base*base) ** (exponent OVER 2) * base ELSE (base*base) ** (exponent OVER 2) FI ); printf(($" One Gibi-unit is: "g(0,1)"**"g(0)"="g(0,1)" - (cost: "g(0) " REAL multiplications)"l$, 2.0, thirty, 2.0 ** thirty,8)); OP ** = (REAL base, REAL exponent)REAL: ( # ** REAL Operator can be overloaded # exp(ln(base)*exponent) ); printf(($" One Gibi-unit is: "g(0,1)"**"g(0,1)"="g(0,1)" - (cost: " "depends on precision)"l$, 2.0, 30.0, 2.0 ** 30.0))
)</lang> Output:
One Gibi-unit is: int pow(2,30)=1073741824 - (cost: 29 INT multiplications) One Gibi-unit is: 2**30=1073741824 - (cost: 8 INT multiplications) One Gibi-unit is: 2.0**30=1073741824.0 - (cost: 8 REAL multiplications) One Gibi-unit is: 2.0**30.0=1073741824.0 - (cost: depends on precision)
AutoHotkey
<lang AutoHotkey>MsgBox % Pow(5,3) MsgBox % Pow(2.5,4)
Pow(x, n){ r:=1 loop %n% r *= x return r }</lang>
AWK
This one-liner reads base and exponent from stdin, one pair per line, and writes the result to stdout: <lang awk>$ awk 'func pow(x,n){r=1;for(i=0;i<n;i++)r=r*x;return r}{print pow($1,$2)}' 2.5 2 6.25 10 6 1000000 3 0 1</lang>
BASIC
The vast majority of BASIC implementations don't support defining custom operators, or overloading of any kind.
<lang qbasic>DECLARE FUNCTION powL& (x AS INTEGER, y AS INTEGER) DECLARE FUNCTION powS# (x AS SINGLE, y AS INTEGER)
DIM x AS INTEGER, y AS INTEGER DIM a AS SINGLE
RANDOMIZE TIMER a = RND * 10 x = INT(RND * 10) y = INT(RND * 10) PRINT x, y, powL&(x, y) PRINT a, y, powS#(a, y)
FUNCTION powL& (x AS INTEGER, y AS INTEGER)
DIM n AS INTEGER, m AS LONG IF x <> 0 THEN m = 1 IF SGN(y) > 0 THEN FOR n = 1 TO y m = m * x NEXT END IF END IF powL& = m
END FUNCTION
FUNCTION powS# (x AS SINGLE, y AS INTEGER)
DIM n AS INTEGER, m AS DOUBLE IF x <> 0 THEN m = 1 IF y <> 0 THEN FOR n = 1 TO y m = m * x NEXT IF y < 0 THEN m = 1# / m END IF END IF powS# = m
END FUNCTION</lang>
Sample outputs:
0 8 0 7.768213 8 13260781.61887441 1 9 1 2.707636 9 7821.90151734948 8 2 64 9.712946 2 94.34131879665438
C
Two versions are given - one for integer bases, the other for floating point. The integer version returns 0 when the abs(base) is != 1 and the exponent is negative. <lang c>#include <stdio.h>
- include <assert.h>
int ipow(int base, int exp) {
int pow = base; int v = 1; if (exp < 0) { assert (base != 0); /* divide by zero */ return (base*base != 1)? 0: (exp&1)? base : 1; } while(exp > 0 ) { if (exp & 1) v *= pow; pow *= pow; exp >>= 1; } return v;
}
double dpow(double base, int exp) {
double v=1.0; double pow = (exp <0)? 1.0/base : base; if (exp < 0) exp = - exp;
while(exp > 0 ) { if (exp & 1) v *= pow; pow *= pow; exp >>= 1; } return v;
}
int main() {
printf("2^6 = %d\n", (int)ipow(2,6)); printf("2^-6 = %lf\n", ipow(2,-6)); printf("2.71^6 = %lf\n", dpow(2.71,6)); printf("2.71^-6 = %lf\n", dpow(2.71,-6));
}</lang>
C#
In C# it is possible to overload operators (+, -, *, etc..), but to do so requires the overload to implement at least one argument as the calling type.
What this means, is that if we have the class, A, to do an overload of + - we must set one of the arguments as the type "A". This is because in C#, overloads are defined on a class basis - so when doing an operator, .Net looks at the class to find the operators. In this manner, one of the arguments must be of the class, else .Net would be looking there in vain.
This again means, that a direct overloading of the ^-character between two integers / double and integer is not possible.
However - coming to think of it, one could overload the "int" class, and enter the operator there. --LordMike 17:45, 5 May 2010 (UTC)
<lang csharp> static void Main(string[] args) { Console.WriteLine("5^5 = " + Expon(5, 5)); Console.WriteLine("5.5^5 = " + Expon(5.5, 5)); Console.ReadLine(); }
static double Expon(int Val, int Pow) { return Math.Pow(Val, Pow); } static double Expon(double Val, int Pow) { return Math.Pow(Val, Pow); } </lang>
Example output:
5^5 = 3125 5.5^5 = 5032,84375
C++
While C++ does allow operator overloading, it does not have an exponentiation operator, therefore only a function definition is given. For non-negative exponents the integer and floating point versions are exactly the same, for obvious reasons. For negative exponents, the integer exponentiation would not give integer results; therefore there are several possibilities:
- Use floating point results even for integer exponents.
- Use integer results for integer exponents and give an error for negative exponents.
- Use integer results for integer exponents and return just the integer part (i.e. return 0 if the base is larger than one and the exponent is negative).
The third option somewhat resembles the integer division rules, and has the nice property that it can use the exact same algorithm as the floating point version. Therefore this option is chosen here. Actually the template can be used with any type which supports multiplication, division and explicit initialization from int. Note that there are several aspects about int which are not portably defined; most notably it is not guaranteed
- that the negative of a valid int is again a valid int; indeed for most implementations, the minimal value doesn't have a positive counterpart,
- whether the result of a%b is positive or negative if a is negative, and in which direction the corresponding division is rounded (however, it is guaranteed that (a/b)*b + a%b == a)
The code below tries to avoid those platform dependencies. Note that bitwise operations wouldn't help here either, because the representation of negative numbers can vary as well. <lang cpp>template<typename Number>
Number power(Number base, int exponent)
{
int zerodir; Number factor; if (exponent < 0) { zerodir = 1; factor = Number(1)/base; } else { zerodir = -1; factor = base; }
Number result(1); while (exponent != 0) { if (exponent % 2 != 0) { result *= factor; exponent += zerodir; } else { factor *= factor; exponent /= 2; } } return result;
}</lang>
Chef
See Basic integer arithmetic#Chef.
Clojure
Operators in Clojure are functions, so this satisfies both requirements. Also, this is polymorphic- it will work with integers, floats, etc, even ratios. (Since operators are implemented as functions they are used in prefix notation) <lang lisp>(defn ** [x n] (reduce * (repeat n x)))</lang> Usage:
(** 2 3) ; 8 (** 7.2 2.1) ; 373.24800000000005 (** 7/2 3) ; 343/8
Common Lisp
Common Lisp has a few forms of iteration. One of the more general is the do loop. Using the do loop, one definition is given below: <lang lisp>(defun my-expt-do (a b)
(do ((x 1 (* x a)) (y 0 (+ y 1))) ((= y b) x)))</lang>
do takes three forms. The first is a list of variable iniitializers and incrementers. In this case, x, the eventual return value, is initialized to 1, and every iteration of the do loop replaces the value of x with x * a. Similarly, y is initialized to 0 and is replaced with y + 1. The second is a list of conditions and return values. In this case, when y = b, the loop stops, and the current value of x is returned. Common Lisp has no explicit return keyword, so x ends up being the return value for the function. The last form is the body of the loop, and usually consists of some action to perform (that has some side-effect). In this case, all the work is being done by the first and second forms, so there are no extra actions.
Of course, Lisp programmers often prefer recursive solutions. <lang lisp>(defun my-expt-rec (a b)
(cond ((= b 0) 1) (t (* a (my-expt-rec a (- b 1))))))</lang>
This solution uses the fact that a^0 = 1 and that a^b = a * a^{b-1}. cond is essentially a generalized if-statement. It takes a list of forms of the form (cond result). For instance, in this case, if b = 0, then function returns 1. t is the truth constant in Common Lisp and is often used as a default condition (similar to the default keyword in C/C++/Java or the else block in many languages).
Common lisp has much more lenient rules for identifiers. In particular, ^ is a valid CL identifier. Since it is not already defined in the standard library, we can simply use it as a function name, just like any other function. <lang lisp>(defun ^ (a b)
(do ((x 1 (* x a)) (y 0 (+ y 1))) ((= y b) x)))</lang>
D
From the Python and C++ versions.
D V.2 has a built-in exponentiation operator: ^^ <lang d>import std.stdio: writeln; import std.conv: to;
pure T power(T)(T base, int exponent)
in { // precondition if (exponent < 0) assert (base != 0, "Division by zero"); } body { int e = exponent; int zerodir; T factor;
if (e < 0) { zerodir = 1; factor = (cast(T)1) / base; } else { zerodir = -1; factor = base; }
T result = 1;
while (e != 0) { if (e % 2 != 0) { result *= factor; e += zerodir; } else { factor *= factor; e /= 2; } }
return result; }
struct Int {
int x; alias x this;
Int opBinary(string op : "^^")(int e) { writeln("opBinary"); return Int(power(x, e)); }
string toString() { return to!string(x); }
}
struct Double {
double x; alias x this;
Double opBinary(string op : "^^")(int e) { writeln("opBinary"); return Double(power(x, e)); }
string toString() { return to!string(x); }
}
void main() {
writeln(Int(3) ^^ 3); writeln(Double(2.5) ^^ 5); writeln(Int(0) ^^ -2);
}</lang> Output:
core.exception.AssertError@exp_op.d(7): Division by zero opBinary 27 opBinary 97.6563 opBinary
E
Simple, unoptimized implementation which will accept any kind of number for the base. If the base is an int
, then the result will be of type float64
if the exponent is negative, and int
otherwise.
<lang e>def power(base, exponent :int) {
var r := base if (exponent < 0) { for _ in exponent..0 { r /= base } } else if (exponent <=> 0) { return 1 } else { for _ in 2..exponent { r *= base } } return r
}</lang>
Factor
Simple, unoptimized implementation which accepts a positive or negative exponent: <lang factor>: pow ( f n -- f' )
dup 0 < [ abs pow recip ] [ [ 1 ] 2dip swap [ * ] curry times ] if ;</lang>
Here is a recursive implementation which splits the exponent in two: <lang factor>: pow ( f n -- f' )
{ { [ dup 0 < ] [ abs pow recip ] } { [ dup 0 = ] [ 2drop 1 ] } [ [ 2 mod 1 = swap 1 ? ] [ [ sq ] [ 2 /i ] bi* pow ] 2bi * ] } cond ;</lang>
This implementation recurses only when an odd factor is found: <lang factor>USING: combinators kernel math ; IN: test
- (pow) ( f n -- f' )
[ dup even? ] [ [ sq ] [ 2 /i ] bi* ] while dup 1 = [ drop ] [ dupd 1 - (pow) * ] if ;
- pow ( f n -- f' )
{ { [ dup 0 < ] [ abs (pow) recip ] } { [ dup 0 = ] [ 2drop 1 ] } [ (pow) ] } cond ;</lang>
A non-recursive version of (pow) can be written as: <lang factor>: (pow) ( f n -- f' )
[ 1 ] 2dip [ dup 1 = ] [ dup even? [ [ sq ] [ 2 /i ] bi* ] [ [ [ * ] keep ] dip 1 - ] if ] until drop * ;</lang>
Forth
<lang forth>: ** ( n m -- n^m )
1 swap 0 ?do over * loop nip ;</lang>
<lang forth>: f**n ( f n -- f^n )
dup 0= if drop fdrop 1e else dup 1 and if 1- fdup recurse f* else 2/ fdup f* recurse then then ;</lang>
Fortran
<lang fortran>MODULE Exp_Mod IMPLICIT NONE
INTERFACE OPERATOR (.pow.) ! Using ** instead would overload the standard exponentiation operator
MODULE PROCEDURE Intexp, Realexp
END INTERFACE
CONTAINS
FUNCTION Intexp (base, exponent) INTEGER :: Intexp INTEGER, INTENT(IN) :: base, exponent INTEGER :: i
IF (exponent < 0) THEN IF (base == 1) THEN Intexp = 1 ELSE Intexp = 0 END IF RETURN END IF Intexp = 1 DO i = 1, exponent Intexp = Intexp * base END DO END FUNCTION IntExp
FUNCTION Realexp (base, exponent) REAL :: Realexp REAL, INTENT(IN) :: base INTEGER, INTENT(IN) :: exponent INTEGER :: i Realexp = 1.0 IF (exponent < 0) THEN DO i = exponent, -1 Realexp = Realexp / base END DO ELSE DO i = 1, exponent Realexp = Realexp * base END DO END IF END FUNCTION RealExp
END MODULE Exp_Mod
PROGRAM EXAMPLE USE Exp_Mod
WRITE(*,*) 2.pow.30, 2.0.pow.30
END PROGRAM EXAMPLE</lang> Output
1073741824 1.073742E+09
Go
Go doesn't support operator defintion. Other notes: While I left the integer algorithm simple, I used the shift and square trick for the float algorithm, just to show an alternative. <lang go> package main
import (
"fmt" "os"
)
func expI(b, p int) (int, os.Error) {
if p < 0 { return 0, os.NewError("negative power not allowed") } r := 1 for i := 1; i <= p; i++ { r *= b }
return r, nil
}
func expF(b float, p int) float {
var neg bool if p < 0 { neg = true p = -p } r := 1. for pow := b; p > 0; pow *= pow { if p & 1 == 1 { r *= pow } p >>= 1 } if neg { r = 1/r } return r
}
func main() {
ti := func(b, p int) { fmt.Printf("%d^%d: ", b, p) e, err := expI(b, p) if err != nil { fmt.Println(err) } else { fmt.Println(e) } }
fmt.Println("expI tests") ti(2, 10) ti(2, -10) ti(-2, 10) ti(-2, 11) ti(11, 0)
fmt.Println("overflow undetected") ti(10, 10)
tf := func(b float, p int) {
fmt.Printf("%g^%d: %g\n", b, p, expF(b, p)) }
fmt.Println("\nexpF tests:") tf(2, 10) tf(2, -10) tf(-2, 10) tf(-2, 11) tf(11, 0)
fmt.Println("disallowed in expI, allowed here") tf(0, -1)
fmt.Println("other interesting cases for 32 bit float type") tf(10, 39) tf(10, -39) tf(-10, 39)
} </lang> Output:
expI tests 2^10: 1024 2^-10: negative power not allowed -2^10: 1024 -2^11: -2048 11^0: 1 overflow undetected 10^10: 1410065408 expF tests: 2^10: 1024 2^-10: 0.0009765625 -2^10: 1024 -2^11: -2048 11^0: 1 disallowed in expI, allowed here 0^-1: +Inf other interesting cases for 32 bit float type 10^39: +Inf 10^-39: 0 -10^39: -Inf
Haskell
Here's the exponentiation operator from the Prelude: <lang haskell>(^) :: (Num a, Integral b) => a -> b -> a _ ^ 0 = 1 x ^ n | n > 0 = f x (n-1) x where
f _ 0 y = y f a d y = g a d where g b i | even i = g (b*b) (i `quot` 2) | otherwise = f b (i-1) (b*y)
_ ^ _ = error "Prelude.^: negative exponent"</lang>
There's no difference in Haskell between a procedure (or function) and an operator, other than the infix notation. This routine is overloaded for any integral exponent (which includes the arbitrarily large Integer type) and any numeric type for the bases (including, for example, Complex). It uses the fast "binary" exponentiation algorithm. For a negative exponent, the type of the base must support division (and hence reciprocals):
<lang haskell>(^^) :: (Fractional a, Integral b) => a -> b -> a x ^^ n = if n >= 0 then x^n else recip (x^(negate n))</lang>
This rules out e.g. the integer types as base values in this case. Haskell also has a third exponentiation operator,
<lang haskell>(**) :: Floating a => a -> a -> a x ** y = exp (log x * y)</lang> which is used for floating point arithmetic.
HicEst
<lang hicest>WRITE(Clipboard) pow(5, 3) ! 125 WRITE(ClipBoard) pow(5.5, 7) ! 152243.5234
FUNCTION pow(x, n)
pow = 1 DO i = 1, n pow = pow * x ENDDO
END </lang>
Icon and Unicon
The procedure below will take an integer or real base and integer exponent and return base ^ exponent. If exponent is negative, base is coerced to real so as not to return 0. Operator overloading is not supported and this is not an efficient implementation. <lang Icon>procedure main() bases := [5,5.] numbers := [0,2,2.,-1,3] every write("expon(",b := !bases,", ",x := !numbers,")=",(expon(b,x) | "failed") \ 1) end
procedure expon(base,power) local op,res
base := numeric(base) | runerror(102,base) power := power = integer(power) | runerr(101,power)
if power = 0 then return 1 else op := if power < 1 then
(base := real(base)) & "/" # force real base else "*"
res := 1 every 1 to abs(power) do
res := op(res,base)
return res end</lang>
J
J is concretely specified, which makes it easy to define primitives in terms of other primitives (this is especially true of mathematical primitives, given the language's mathematical emphasis).
So we have any number of options. Here's the simplest, equivalent to the for each number, product = product * number
of other languages. The base may be any number, and the exponent may be any non-negative integer (including zero):
<lang j> exp =: */@:#~
10 exp 3
1000
10 exp 0
1</lang>
We can make this more general by allowing the exponent to be any integer (including negatives), at the cost of a slight increase in complexity:
<lang j> exp =: *@:] %: */@:(#~|)
10 exp _3
0.001</lang>
Or, we can define exponentiation as repeated multiplication (as opposed to multiplying a given number of copies of the base)
<lang J> exp =: dyad def 'x *^:y 1'
10 exp 3
1000
10 exp _3
0.001</lang>
Here, when we specify a negative number of repetitions, multiplication's inverse is used that many times.
J's calculus of functions permits us to define exponentiation in its full generality, as the inverse of log (i.e. exp = log-1):
<lang j> exp =: ^.^:_1
81 exp 0.5
9</lang>
Note that the definition above does not use the primitive exponentiation function ^
. The carets in it represent different (but related) things . The function is composed of three parts: ^. ^: _1
. The first part, ^.
, is the primitive logarithm operator (e.g. 3 = 10^.1000
) .
The second part, ^:
, is interesting: it is a "meta operator". It takes two arguments: a function f
on its left, and a number N
on its right. It produces a new function, which, when given an argument, applies f
to that argument N
times. For example, if we had a function increment
, then increment^:3 X
would increment X
three times, so the result would be X+3
.
In the case of ^. ^: _1
, f
is ^.
(i.e. logarithm) and N
is -1. Therefore we apply log negative one times or the inverse of log once (precisely as in log-1).
Similarly, we can define exponentiation as the reverse of the inverse of root. That is, x pow y = y root-1 x:
<lang j> exp =: %:^:_1~
81 exp 0.5
9</lang>
Compare this with the previous definition: it is the same, except that %:
, root, has been substituted for ^.
, logarithm, and the arguments have been reversed (or reflected) with ~
.
That is, J is telling us that power is the same as the reflex of the inverse of root, exactly as we'd expect.
One last note: we said these definitions are the same as ^
in its full generality. What is meant by that? Well, in the context of this puzzle, it means both the base and exponent may be any real number. But J goes further than that: it also permits complex numbers.
Let's use Euler's famous formula, epi*i = -1 as an example:
<lang j> pi =: 3.14159265358979323846
e =: 2.71828182845904523536 i =: 2 %: _1 NB. Square root of -1 e^(pi*i)
_1</lang>
And, as stated, our redefinition is equivalent:
<lang j> exp =: %:^:_1~
e exp (pi*i)
_1</lang>
Java
Java does not support operator definition. This example is unoptimized, but will handle negative exponents as well. It is unnecessary to show intint since an int in Java will be cast as a double. <lang java>public class Exp{
public static void main(String[] args){ System.out.println(pow(2,30)); System.out.println(pow(2.0,30)); //tests System.out.println(pow(2.0,-2)); }
public static double pow(double base, int exp){ if(exp < 0) return 1 / pow(base, -exp); double ans = 1.0; for(;exp > 0;--exp) ans *= base; return ans; }
}</lang> Output:
1.073741824E9 1.073741824E9 0.25
JavaScript
<lang javascript>function pow(base, exp) {
if (exp != Math.floor(exp)) throw "exponent must be an integer"; if (exp < 0) return 1 / pow(base, -exp); var ans = 1; while (exp > 0) { ans *= base; exp--; } return ans;
}</lang>
Logo
<lang logo>to int_power :n :m
if equal? 0 :m [output 1] if equal? 0 modulo :m 2 [output int_power :n*:n :m/2] output :n * int_power :n :m-1
end</lang>
Lucid
Some misconceptions about Lucid
<lang lucid>pow(n,x)
k = n fby k div 2; p = x fby p*p; y =1 fby if even(k) then y else y*p; result y asa k eq 0;
end</lang>
M4
M4 lacks floating point computation and operator definition. <lang M4>define(`power',`ifelse($2,0,1,`eval($1*$0($1,decr($2)))')') power(2,10)</lang> Output:
1024
Mathematica
Define a function and an infix operator \[CirclePlus] with the same definition: <lang Mathematica>exponentiation[x_,y_Integer]:=Which[y>0,Times@@ConstantArray[x,y],y==0,1,y<0,1/exponentiation[x,-y]] CirclePlus[x_,y_Integer]:=exponentiation[x,y]</lang> Examples: <lang Mathematica>exponentiation[1.23,3] exponentiation[4,0] exponentiation[2.5,-2] 1.23\[CirclePlus]3 4\[CirclePlus]0 2.5\[CirclePlus]-2</lang> gives back: <lang Mathematica>1.86087 1 0.16 1.86087 1 0.16</lang> Note that \[CirclePlus] shows up as a special character in Mathematica namely a circle divided in 4 pieces. Note also that this function supports negative and positive exponents.
Modula-3
<lang modula3>MODULE Expt EXPORTS Main;
IMPORT IO, Fmt;
PROCEDURE IntExpt(arg, exp: INTEGER): INTEGER =
VAR result := 1; BEGIN FOR i := 1 TO exp DO result := result * arg; END; RETURN result; END IntExpt;
PROCEDURE RealExpt(arg: REAL; exp: INTEGER): REAL =
VAR result := 1.0; BEGIN IF exp < 0 THEN FOR i := exp TO -1 DO result := result / arg; END; ELSE FOR i := 1 TO exp DO result := result * arg; END; END; RETURN result; END RealExpt;
BEGIN
IO.Put("2 ^ 4 = " & Fmt.Int(IntExpt(2, 4)) & "\n"); IO.Put("2.5 ^ 4 = " & Fmt.Real(RealExpt(2.5, 4)) & "\n");
END Expt.</lang> Output:
2 ^ 4 = 16 2.5 ^ 4 = 39.0625
OCaml
The following operators only take nonnegative integer exponents. <lang ocaml>let expt_int a b =
let rec aux x i = if i = b then x else aux (x * a) (i + 1) in aux 1 0
let expt_float a b =
let rec aux x i = if i = b then x else aux (x *. a) (i + 1) in aux 1. 0
let ( ** ) = expt_int ;; (* hides the default infix operator *) let ( **. ) = expt_float ;;</lang>
<lang ocaml># 2 ** 3 ;; - : int = 8
- 2.4 **. 3 ;;
- : float = 13.824</lang>
It is possible to create a generic exponential. For this, one must know the multiplication function, and the unit value. Here, the usual fast algorithm is used:
<lang ocaml>let pow one mul a n =
let rec g p x = function | 0 -> x | i -> g (mul p p) (if i mod 2 = 1 then mul p x else x) (i/2) in g a one n
pow 1 ( * ) 2 16;; (* 65536 *) pow 1.0 ( *. ) 2.0 16;; (* 65536. *)
(* pow is not limited to exponentiation *) pow 0 ( + ) 2 16;; (* 32 *) pow "" ( ^ ) "abc " 10;; (* "abc abc abc abc abc abc abc abc abc abc " *) pow [ ] ( @ ) [ 1; 2 ] 10;; (* [1; 2; 1; 2; 1; 2; 1; 2; 1; 2; 1; 2; 1; 2; 1; 2; 1; 2; 1; 2] *)
(* Thue-Morse sequence *) Array.init 32 (fun n -> (1 - pow 1 ( - ) 0 n) lsr 1);;
(* [|0; 1; 1; 0; 1; 0; 0; 1; 1; 0; 0; 1; 0; 1; 1; 0;
1; 0; 0; 1; 0; 1; 1; 0; 0; 1; 1; 0; 1; 0; 0; 1|]
See http://en.wikipedia.org/wiki/Thue-Morse_sequence
- )</lang>
See also Matrix-exponentiation operator#OCaml for a matrix usage.
PARI/GP
This version works for integer and floating-point bases (as well as intmod bases, ...). <lang>ex(a,b)={
my(c=1); while(b>1, if(b%2, c *= a); a = a^2; b >>= 1 ); a * c
};</lang>
Perl
<lang Perl>#!/usr/bin/perl -w use strict ;
sub expon {
my ( $base , $expo ) = @_ ; if ( $expo == 0 ) { return 1 ; } elsif ( $expo == 1 ) { return $base ; } elsif ( $expo > 1 ) { my $prod = 1 ; foreach my $n ( 0..($expo - 1) ) {
$prod *= $base ;
} return $prod ; } elsif ( $expo < 0 ) { return 1 / ( expon ( $base , -$expo ) ) ; }
} print "3 to the power of 10 as a function is " . expon( 3 , 10 ) . " !\n" ; print "3 to the power of 10 as a builtin is " . 3**10 . " !\n" ; print "5.5 to the power of -3 as a function is " . expon( 5.5 , -3 ) . " !\n" ; print "5.5 to the power of -3 as a builtin is " . 5.5**-3 . " !\n" ; </lang> Output:
3 to the power of 10 as a function is 59049 ! 3 to the power of 10 as a builtin is 59049 ! 5.5 to the power of -3 as a function is 0.00601051840721262 ! 5.5 to the power of -3 as a builtin is 0.00601051840721262 !
Perl 6
<lang perl6>subset Natural of Int where { $^n >= 0 }
multi pow (0, 0) { fail '0**0 is undefined' } multi pow ($base, Natural $exp) { [*] $base xx $exp } multi pow ($base, Int $exp) { 1 / pow $base, -$exp }
sub infix:<***> ($a, $b) { pow $a, $b }</lang>
Examples of use:
<lang perl6>say pow .75, -5; say .75 *** -5;</lang>
PL/I
<lang PL/I> declare exp generic
(iexp when (fixed, fixed), fexp when (float, fixed) );
iexp: procedure (m, n) returns (fixed binary (31));
declare (m, n) fixed binary (31) nonassignable; declare exp fixed binary (31) initial (m), i fixed binary; if m = 0 & n = 0 then signal error; if n = 0 then return (1); do i = 2 to n; exp = exp * m; end; return (exp);
end iexp; fexp: procedure (a, n) returns (float (15));
declare (a float, n fixed binary (31)) nonassignable; declare exp float initial (a), i fixed binary; if a = 0 & n = 0 then signal error; if n = 0 then return (1); do i = 2 to n; exp = exp * a; end; return (exp);
end fexp;</lang>
PicoLisp
This uses Knuth's algorithm (The Art of Computer Programming, Vol. 2, page 442) <lang PicoLisp>(de ** (X N) # N th power of X
(let Y 1 (loop (when (bit? 1 N) (setq Y (* Y X)) ) (T (=0 (setq N (>> 1 N))) Y ) (setq X (* X X)) ) ) )</lang>
PowerShell
<lang powershell>function pow($a, [int]$b) {
if ($b -eq -1) { return 1/$a } if ($b -eq 0) { return 1 } if ($b -eq 1) { return $a } if ($b -lt 0) { $rec = $true # reciprocal needed $b = -$b }
$result = $a 2..$b | ForEach-Object { $result *= $a }
if ($rec) { return 1/$result } else { return $result }
}</lang> The function works for both integers and floating-point values as first argument.
PowerShell does not support operator overloading directly (and there wouldn't be an exponentiation operator to overload).
Output:
PS> pow 2 15 32768 PS> pow 2.71 -4 0,018540559532257 PS> pow (-1.35) 3 −2,460375
The negative first argument needs to be put in parentheses because it would otherwise be passed as string. This can be circumvented by declaring the first argument to the function as double
, but then the return type would be always double while currently pow 2 3
returns an int
.
PureBasic
PureBasic does not allow an operator to be redefined or operator overloading. <lang PureBasic>Procedure powI(base, exponent)
Protected i, result.d If exponent < 0 If base = 1 result = 1 EndIf ProcedureReturn result EndIf result = 1 For i = 1 To exponent result * base Next ProcedureReturn result
EndProcedure
Procedure.f powF(base.f, exponent)
Protected i, magExponent = Abs(exponent), result.d If base <> 0 result = 1.0 If exponent <> 0 For i = 1 To magExponent result * base Next If exponent < 0 result = 1.0 / result EndIf EndIf EndIf ProcedureReturn result
EndProcedure
If OpenConsole()
Define x, a.f, exp x = Random(10) - 5 a = Random(10000) / 10000 * 10 For exp = -3 To 3 PrintN(Str(x) + " ^ " + Str(exp) + " = " + Str(powI(x, exp))) PrintN(StrF(a) + " ^ " + Str(exp) + " = " + StrF(powF(a, exp))) PrintN("--------------") Next Print(#CRLF$ + #CRLF$ + "Press ENTER to exit") Input() CloseConsole()
EndIf</lang> Sample output:
-3 ^ -3 = 0 6.997000 ^ -3 = 0.002919 -------------- -3 ^ -2 = 0 6.997000 ^ -2 = 0.020426 -------------- -3 ^ -1 = 0 6.997000 ^ -1 = 0.142918 -------------- -3 ^ 0 = 1 6.997000 ^ 0 = 1.000000 -------------- -3 ^ 1 = -3 6.997000 ^ 1 = 6.997000 -------------- -3 ^ 2 = 9 6.997000 ^ 2 = 48.958012 -------------- -3 ^ 3 = -27 6.997000 ^ 3 = 342.559235 -------------
Python
<lang python>>>> import operator >>> class num(int):
def __pow__(self, b): print "Empowered" return operator.__pow__(self+0, b)
>>> x = num(3)
>>> x**2
Empowered
9
>>> class num(float):
def __pow__(self, b): print "Empowered" return operator.__pow__(self+0, b)
>>> x = num(2.5)
>>> x**2
Empowered
6.25
>>></lang>
R
<lang r># Method pow <- function(x, y) {
x <- as.numeric(x) y <- as.integer(y) prod(rep(x, y))
}
- Operator
"%pow%" <- function(x,y) pow(x,y)
pow(3, 4) # 81 2.5 %pow% 2 # 6.25</lang>
Retro
Retro has no floating point support in the standard VM.
<lang Retro>: pow ( bp-n ) 1 swap [ over * ] times nip ;</lang>
REXX
<lang rexx> say 'digits=' digits() say '17**65' say 17**65 say numeric digits 100 say say 'digits=' digits() say '17**65' say 17**65 say numeric digits 10 say say 'digits=' digits() say '2 ** -10' say 2 ** -10 say numeric digits 30 say say 'digits=' digits() say '-3.1415926535897932384626433 ** 3' say -3.1415926535897932384626433 ** 3 say numeric digits 1000 say say 'digits=' digits() say '2 ** 1000' say 2 ** 1000 say </lang> Output:
digits= 9 17**65 9.53190909E+79 digits= 100 17**65 95319090450218007303742536355848761234066170796000792973413605849481890760893457 digits= 10 2 ** -10 0.0009765625 digits= 30 -3.1415926535897932384626433 ** 3 -31.0062766802998201754763126013 digits= 1000 2 ** 1000 10715086071862673209484250490600018105614048117055336074437503883703510511249361224931983788156958581275946729175531468251871452856923140435984577574698574803934567774824230985421074605062371141877954182153046474983581941267398767559165543946077062914571196477686542167660429831652624386837205668069376
Ruby
As a method: <lang ruby>class Numeric
def pow(m) raise ArgumentError, "exponent must be an integer: #{m}" unless m.is_a?(Integer) puts "pow!!"
# below requires ruby 1.8.7 or greater Array.new(m, self).inject(1){|res, n| res*n}
# for earlier versions of ruby without the "enum.inject(initial, sym)" method: # res = 1 # m.times {res *= self} # res end
end
p 5.pow(3) p 5.5.pow(3) p 5.pow(3.1)</lang> outputs
pow!! 125 pow!! 166.375 ArgumentError: exponent must be an integer: 3.1
To overload the ** exponentiation operator, this might work, but doesn't: <lang ruby>class Numeric
def **(m) pow(m) end
end</lang> It doesn't work because the ** method is defined independently for Numeric subclasses Fixnum, Bignum and Float. One must: <lang ruby>class Fixnum
def **(m) print "Fixnum " pow(m) end
end class Bignum
def **(m) print "Bignum " pow(m) end
end class Float
def **(m) print "Float " pow(m) end
end
p i=2**64 p i ** 2 p 2.2 ** 3</lang> which outputs
Fixnum pow!! 18446744073709551616 Bignum pow!! 340282366920938463463374607431768211456 Float pow!! 10.648
Scala
There's no distinction between an operator and a method in Scala. Alas, there is no way of adding methods to a class, but one can make it look like a method has been added, through a method commonly known as Pimp My Library.
Therefore, we show below how that can be accomplished. We define the operator ↑ (unicode's uparrow), which is written as \u2191 below, to make cut&paste easier.
To use it, one has to import the implicit from the appropriate object. ExponentI will work for any integral type (Int, BigInt, etc), ExponentF will work for any fractional type (Double, BigDecimal, etc). Importing both at the same time won't work. In this case, it might be better to define implicits for the actual types being used, such as was done in Exponents.
<lang scala>object Exponentiation {
import scala.annotation.tailrec @tailrec def powI[N](n: N, exponent: Int)(implicit num: Integral[N]): N = { import num._ exponent match { case 0 => one case _ if exponent % 2 == 0 => powI((n * n), (exponent / 2)) case _ => powI(n, (exponent - 1)) * n } } @tailrec def powF[N](n: N, exponent: Int)(implicit num: Fractional[N]): N = { import num._ exponent match { case 0 => one case _ if exponent < 0 => one / powF(n, exponent.abs) case _ if exponent % 2 == 0 => powF((n * n), (exponent / 2)) case _ => powF(n, (exponent - 1)) * n } } class ExponentI[N : Integral](n: N) { def \u2191(exponent: Int): N = powI(n, exponent) }
class ExponentF[N : Fractional](n: N) { def \u2191(exponent: Int): N = powF(n, exponent) }
object ExponentI { implicit def toExponentI[N : Integral](n: N): ExponentI[N] = new ExponentI(n) } object ExponentF { implicit def toExponentF[N : Fractional](n: N): ExponentF[N] = new ExponentF(n) } object Exponents { implicit def toExponent(n: Int): ExponentI[Int] = new ExponentI(n) implicit def toExponent(n: Double): ExponentF[Double] = new ExponentF(n) }
}</lang>
Scheme
This definition of the exponentiation procedure ^
operates on bases of all numerical types that the multiplication procedure *
operates on, i. e. integer, rational, real, and complex. The notion of an operator does not exist in Scheme. Application of a procedure to its arguments is always expressed with a prefix notation.
<lang scheme>(define (^ base exponent)
(define (*^ exponent acc) (if (= exponent 0) acc (*^ (- exponent 1) (* acc base)))) (*^ exponent 1))
(display (^ 2 3)) (newline) (display (^ (/ 1 2) 3)) (newline) (display (^ 0.5 3)) (newline) (display (^ 2+i 3)) (newline)</lang> Output:
8 1/8 0.125 2+11i
Seed7
In Seed7 the ** operator is overloaded for both intint and floatint. The following re-implementation of both functions does not use another exponentiation function to do the computation. Instead the the exponentiation by squaring algorithm is used.
<lang seed7>const func integer: intPow (in var integer: base, in var integer: exponent) is func
result var integer: result is 0; begin if exponent < 0 then raise(NUMERIC_ERROR); else if odd(exponent) then result := base; else result := 1; end if; exponent := exponent div 2; while exponent <> 0 do base *:= base; if odd(exponent) then result *:= base; end if; exponent := exponent div 2; end while; end if; end func;</lang>
Original source: [1]
<lang seed7>const func float: fltIPow (in var float: base, in var integer: exponent) is func
result var float: result is 0.0; local var boolean: neg_exp is FALSE; begin if base = 0.0 then if exponent < 0 then result := Infinity; elsif exponent = 0 then result := 1.0; else result := 0.0; end if; else if exponent < 0 then exponent := -exponent; neg_exp := TRUE; end if; # In the twos complement representation the most # negative number is the only one where both the # number and its negation are negative. When the # exponent is proven to be to the most negative # number fltIPow returns 0.0 . if exponent >= 0 then if odd(exponent) then result := base; else result := 1.0; end if; exponent >>:= 1; while exponent <> 0 do base *:= base; if odd(exponent) then result *:= base; end if; exponent >>:= 1; end while; if neg_exp then result := 1.0 / result; end if; end if; end if; end func;</lang>
Original source: [2]
Since Seed7 supports operator and function overloading a new exponentiation operator like ^* can be defined for both intint and floatint:
<lang seed7>$ syntax expr: .(). ^* .() is <- 4;
const func integer: (in var integer: base) ^* (in var integer: exponent) is
return intPow(base, exponent);
const func float: (in var float: base) ^* (in var integer: exponent) is
return fltIPow(base, exponent);</lang>
Slate
This code is from the current slate implementation: <lang slate>x@(Number traits) raisedTo: y@(Integer traits) [
y isZero ifTrue: [^ x unit]. x isZero \/ [y = 1] ifTrue: [^ x]. y isPositive ifTrue: "(x * x raisedTo: y // 2) * (x raisedTo: y \\ 2)" [| count result | count: 1. [(count: (count bitShift: 1)) < y] whileTrue. result: x unit. [count isPositive]
whileTrue: [result: result squared. (y bitAnd: count) isZero ifFalse: [result: result * x]. count: (count bitShift: -1)].
result] ifFalse: [(x raisedTo: y negated) reciprocal]
].</lang>
For floating numbers:
<lang slate>x@(Float traits) raisedTo: y@(Float traits) "Implements floating-point exponentiation in terms of the natural logarithm and exponential primitives - this is generally faster than the naive method." [
y isZero ifTrue: [^ x unit]. x isZero \/ [y isUnit] ifTrue: [^ x]. (x ln * y) exp
].</lang>
Smalltalk
Extending the class Number, we provide the operator for integers, floating points, rationals numbers (and any other derived class) <lang smalltalk>Number extend [
** anInt [ | r | ( anInt isInteger ) ifFalse: [ '** works fine only for integer powers'
displayOn: stderr . Character nl displayOn: stderr ].
r := 1. 1 to: anInt do: [ :i | r := ( r * self ) ]. ^r ]
].
( 2.5 ** 3 ) displayNl. ( 2 ** 10 ) displayNl. ( 3/7 ** 3 ) displayNl.</lang>
Outputs
15.625 1024 27/343
Standard ML
The following operators only take nonnegative integer exponents. <lang sml>fun expt_int (a, b) = let
fun aux (x, i) = if i = b then x else aux (x * a, i + 1)
in
aux (1, 0)
end
fun expt_real (a, b) = let
fun aux (x, i) = if i = b then x else aux (x * a, i + 1)
in
aux (1.0, 0)
end
val op ** = expt_int infix 6 ** val op *** = expt_real infix 6 ***</lang>
<lang sml>- 2 ** 3; val it = 8 : int - 2.4 *** 3; val it = 13.824 : real</lang>
Tcl
Tcl already has both an exponentiation function (set x [expr {pow(2.4, 3.5)}]
) and operator (set x [expr {2.4 ** 3.5}]
). The operator cannot be overridden. The function may be overridden by a procedure in the tcl::mathfunc
namespace, relative to
the calling namespace.
This solution does not consider negative exponents. <lang tcl>package require Tcl 8.5 proc tcl::mathfunc::mypow {a b} {
if { ! [string is int -strict $b]} {error "exponent must be an integer"} set res 1 for {set i 1} {$i <= $b} {incr i} {set res [expr {$res * $a}]} return $res
} expr {mypow(3, 3)} ;# ==> 27 expr {mypow(3.5, 3)} ;# ==> 42.875 expr {mypow(3.5, 3.2)} ;# ==> exponent must be an integer</lang>
- Programming Tasks
- Arithmetic operations
- Ada
- ALGOL 68
- AutoHotkey
- AWK
- BASIC
- C
- C sharp
- C++
- Chef
- Clojure
- Common Lisp
- D
- E
- Factor
- Forth
- Fortran
- Go
- Haskell
- HicEst
- Icon
- Unicon
- J
- Java
- JavaScript
- Logo
- Lucid
- M4
- Mathematica
- Modula-3
- OCaml
- PARI/GP
- Perl
- Perl 6
- PL/I
- PicoLisp
- PowerShell
- PureBasic
- Python
- R
- Retro
- REXX
- Ruby
- Scala
- Scheme
- Seed7
- Slate
- Smalltalk
- Standard ML
- Tcl