Arithmetic/Rational
The objective of this task is to create a reasonably complete implementation of rational arithmetic in the particular language using the idioms of the language.
You are encouraged to solve this task according to the task description, using any language you may know.
For example: Define an new type called frac with binary operator "//" of two integers that returns a structure made up of the numerator and the denominator (as per a rational number).
Further define the appropriate rational unary operators abs and '-', with the binary operators for addition '+', subtraction '-', multiplication '×', division '/', integer division '÷', modulo division, the comparison operators (e.g. '<', '≤', '>', & '≥') and equality operators (e.g. '=' & '≠').
Define standard coercion operators for casting int to frac etc.
If space allows, define standard increment and decrement operators (e.g. '+:=' & '-:=' etc.).
Finally test the operators: Use the new type frac to find all perfect numbers less then 219 by summing the reciprocal of the factors.
See also
Ada
ALGOL 68
<lang algol68> MODE FRAC = STRUCT( INT num #erator#, den #ominator#);
FORMAT frac repr = $g(-0)"//"g(-0)$;
PROC gcd = (INT a, b) INT: # greatest common divisor #
(a = 0 | b |: b = 0 | a |: ABS a > ABS b | gcd(b, a MOD b) | gcd(a, b MOD a));
PROC lcm = (INT a, b)INT: # least common multiple #
a OVER gcd(a, b) * b;
PROC raise not implemented error = ([]STRING args)VOID: (
put(stand error, ("Not implemented error: ",args, newline));
stop
);
PRIO // = 9; # higher then the ** operator #
OP // = (INT num, den)FRAC: ( # initialise and normalise #
INT common = gcd(num, den);
IF den < 0 THEN
( -num OVER common, -den OVER common)
ELSE
( num OVER common, den OVER common)
FI
);
OP + = (FRAC a, b)FRAC: (
INT common = lcm(den OF a, den OF b);
FRAC result := ( common OVER den OF a * num OF a + common OVER den OF b * num OF b, common );
num OF result//den OF result
);
OP - = (FRAC a, b)FRAC: a + -b,
* = (FRAC a, b)FRAC: (
INT num = num OF a * num OF b,
den = den OF a * den OF b;
INT common = gcd(num, den);
(num OVER common) // (den OVER common)
);
OP / = (FRAC a, b)FRAC: a * FRAC(den OF b, num OF b),# real division #
% = (FRAC a, b)INT: ENTIER (a / b), # integer divison #
%* = (FRAC a, b)FRAC: a/b - FRACINIT ENTIER (a/b), # modulo division #
** = (FRAC a, INT exponent)FRAC:
IF exponent >= 0 THEN
(num OF a ** exponent, den OF a ** exponent )
ELSE
(den OF a ** exponent, num OF a ** exponent )
FI;
OP REALINIT = (FRAC frac)REAL: num OF frac / den OF frac,
FRACINIT = (INT num)FRAC: num // 1,
FRACINIT = (REAL num)FRAC: (
# express real number as a fraction # # a future execise! #
raise not implemented error(("Convert a REAL to a FRAC","!"));
SKIP
);
OP < = (FRAC a, b)BOOL: num OF (a - b) < 0,
> = (FRAC a, b)BOOL: num OF (a - b) > 0,
<= = (FRAC a, b)BOOL: NOT ( a > b ),
>= = (FRAC a, b)BOOL: NOT ( a < b ),
= = (FRAC a, b)BOOL: (num OF a, den OF a) = (num OF b, den OF b),
/= = (FRAC a, b)BOOL: (num OF a, den OF a) /= (num OF b, den OF b);
# Unary operators #
OP - = (FRAC frac)FRAC: (-num OF frac, den OF frac),
ABS = (FRAC frac)FRAC: (ABS num OF frac, ABS den OF frac),
ENTIER = (FRAC frac)INT: (num OF frac OVER den OF frac) * den OF frac;
COMMENT Operators for extended characters set, and increment/decrement:
OP +:= = (REF FRAC a, FRAC b)REF FRAC: ( a := a + b ),
+=: = (FRAC a, REF FRAC b)REF FRAC: ( b := a + b ),
-:= = (REF FRAC a, FRAC b)REF FRAC: ( a := a - b ),
*:= = (REF FRAC a, FRAC b)REF FRAC: ( a := a * b ),
/:= = (REF FRAC a, FRAC b)REF FRAC: ( a := a / b ),
%:= = (REF FRAC a, FRAC b)REF FRAC: ( a := FRACINIT (a % b) ),
%*:= = (REF FRAC a, FRAC b)REF FRAC: ( a := a %* b );
# OP aliases for extended character sets (eg: Unicode, APL, ALCOR and GOST 10859) #
OP × = (FRAC a, b)FRAC: a * b,
÷ = (FRAC a, b)INT: a OVER b,
÷× = (FRAC a, b)FRAC: a MOD b,
÷* = (FRAC a, b)FRAC: a MOD b,
%× = (FRAC a, b)FRAC: a MOD b,
≤ = (FRAC a, b)FRAC: a <= b,
≥ = (FRAC a, b)FRAC: a >= b,
≠ = (FRAC a, b)BOOL: a /= b,
↑ = (FRAC frac, INT exponent)FRAC: frac ** exponent,
÷×:= = (REF FRAC a, FRAC b)REF FRAC: ( a := a MOD b ),
%×:= = (REF FRAC a, FRAC b)REF FRAC: ( a := a MOD b ),
÷*:= = (REF FRAC a, FRAC b)REF FRAC: ( a := a MOD b );
# BOLD aliases for CPU that only support uppercase for 6-bit bytes - wrist watches #
OP OVER = (FRAC a, b)INT: a % b,
MOD = (FRAC a, b)FRAC: a %*b,
LT = (FRAC a, b)BOOL: a < b,
GT = (FRAC a, b)BOOL: a > b,
LE = (FRAC a, b)BOOL: a <= b,
GE = (FRAC a, b)BOOL: a >= b,
EQ = (FRAC a, b)BOOL: a = b,
NE = (FRAC a, b)BOOL: a /= b,
UP = (FRAC frac, INT exponent)FRAC: frac**exponent;
# the required standard assignment operators #
OP PLUSAB = (REF FRAC a, FRAC b)REF FRAC: ( a +:= b ), # PLUS #
PLUSTO = (FRAC a, REF FRAC b)REF FRAC: ( a +=: b ), # PRUS #
MINUSAB = (REF FRAC a, FRAC b)REF FRAC: ( a *:= b ),
DIVAB = (REF FRAC a, FRAC b)REF FRAC: ( a /:= b ),
OVERAB = (REF FRAC a, FRAC b)REF FRAC: ( a %:= b ),
MODAB = (REF FRAC a, FRAC b)REF FRAC: ( a %*:= b );
END COMMENT Example: searching for Perfect Numbers.
FRAC sum:= FRACINIT 0;
FORMAT perfect = $b(" perfect!","")$;
FOR i FROM 2 TO 2**19 DO
INT candidate := i;
FRAC sum := 1 // candidate;
REAL real sum := 1 / candidate;
FOR factor FROM 2 TO ENTIER sqrt(candidate) DO
IF candidate MOD factor = 0 THEN
sum := sum + 1 // factor + 1 // ( candidate OVER factor);
real sum +:= 1 / factor + 1 / ( candidate OVER factor)
FI
OD;
IF den OF sum = 1 THEN
printf(($"Sum of reciprocal factors of "g(-0)" = "g(-0)" exactly, about "g(0,real width) f(perfect)l$,
candidate, ENTIER sum, real sum, ENTIER sum = 1))
FI
OD</lang>
Output:
Sum of reciprocal factors of 6 = 1 exactly, about 1.0000000000000000000000000001 perfect! Sum of reciprocal factors of 28 = 1 exactly, about 1.0000000000000000000000000001 perfect! Sum of reciprocal factors of 120 = 2 exactly, about 2.0000000000000000000000000002 Sum of reciprocal factors of 496 = 1 exactly, about 1.0000000000000000000000000001 perfect! Sum of reciprocal factors of 672 = 2 exactly, about 2.0000000000000000000000000001 Sum of reciprocal factors of 8128 = 1 exactly, about 1.0000000000000000000000000001 perfect! Sum of reciprocal factors of 30240 = 3 exactly, about 3.0000000000000000000000000002 Sum of reciprocal factors of 32760 = 3 exactly, about 3.0000000000000000000000000003 Sum of reciprocal factors of 523776 = 2 exactly, about 2.0000000000000000000000000005
C
C++
Boost provides a rational number template.
<lang c++>
- include <iostream>
- include "math.h"
- include "boost/rational.hpp"
typedef boost::rational<int> frac;
bool is_perfect(int c) {
frac sum(1, c);
for (int f = 2;f < sqrt(static_cast<float>(c)); ++f){
if (c % f == 0) sum += frac(1,f) + frac(1, c/f);
}
if (sum.denominator() == 1){
return (sum == 1);
}
return false;
}
int main() {
for (int candidate = 2; candidate < 0x80000; ++candidate){
if (is_perfect(candidate))
std::cout << candidate << " is perfect" << std::endl;
} return 0;
} </lang>
Common Lisp
Common Lisp has rational numbers built-in and integrated with all other number types. Common Lisp's number system is not extensible so reimplementing rational arithmetic would require all-new operator names.
<lang lisp>(loop for candidate from 2 below (expt 2 19)
for sum = (+ (/ candidate)
(loop for factor from 2 to (isqrt candidate)
when (zerop (mod candidate factor))
sum (+ (/ factor) (/ (floor candidate factor)))))
when (= sum 1)
collect candidate)</lang>
Forth
<lang forth>\ Rationals can use any double cell operations: 2!, 2@, 2dup, 2swap, etc. \ Uses the stack convention of the built-in "*/" for int * frac -> int
- numerator drop ;
- denominator nip ;
- s>rat 1 ; \ integer to rational (n/1)
- rat>s / ; \ integer
- rat>frac mod ; \ fractional part
- rat>float swap s>f s>f f/ ;
- rat. swap 1 .r [char] / emit . ;
\ normalize: factors out gcd and puts sign into numerator
- gcd ( a b -- gcd ) begin ?dup while tuck mod repeat ;
- rat-normalize ( rat -- rat ) 2dup gcd tuck / >r / r> ;
- rat-abs swap abs swap ;
- rat-negate swap negate swap ;
- 1/rat over 0< if negate swap negate else swap then ;
- rat+ ( a b c d -- ad+bc bd )
rot 2dup * >r rot * >r * r> + r> rat-normalize ;
- rat- rat-negate rat+ ;
- rat* ( a b c d -- ac bd )
rot * >r * r> rat-normalize ;
- rat/ swap rat* ;
- rat-equal d= ;
- rat-less ( a b c d -- ad<bc )
-rot * >r * r> < ;
- rat-more 2swap rat-less ;
- rat-inc tuck + swap ;
- rat-dec tuck - swap ;</lang>
Fortran
<lang fortran>module module_rational
implicit none private public :: rational public :: rational_simplify public :: assignment (=) public :: operator (//) public :: operator (+) public :: operator (-) public :: operator (*) public :: operator (/) public :: operator (<) public :: operator (<=) public :: operator (>) public :: operator (>=) public :: operator (==) public :: operator (/=) public :: abs public :: int public :: modulo type rational integer :: numerator integer :: denominator end type rational interface assignment (=) module procedure assign_rational_int, assign_rational_real end interface interface operator (//) module procedure make_rational end interface interface operator (+) module procedure rational_add end interface interface operator (-) module procedure rational_minus, rational_subtract end interface interface operator (*) module procedure rational_multiply end interface interface operator (/) module procedure rational_divide end interface interface operator (<) module procedure rational_lt end interface interface operator (<=) module procedure rational_le end interface interface operator (>) module procedure rational_gt end interface interface operator (>=) module procedure rational_ge end interface interface operator (==) module procedure rational_eq end interface interface operator (/=) module procedure rational_ne end interface interface abs module procedure rational_abs end interface interface int module procedure rational_int end interface interface modulo module procedure rational_modulo end interface
contains
recursive function gcd (i, j) result (res)
integer, intent (in) :: i
integer, intent (in) :: j
integer :: res
if (j == 0) then
res = i
else
res = gcd (j, modulo (i, j))
end if
end function gcd
function rational_simplify (r) result (res) type (rational), intent (in) :: r type (rational) :: res integer :: g g = gcd (r % numerator, r % denominator) res = r % numerator / g // r % denominator / g end function rational_simplify
function make_rational (numerator, denominator) result (res) integer, intent (in) :: numerator integer, intent (in) :: denominator type (rational) :: res res = rational (numerator, denominator) end function make_rational
subroutine assign_rational_int (res, i) type (rational), intent (out), volatile :: res integer, intent (in) :: i res = i // 1 end subroutine assign_rational_int
subroutine assign_rational_real (res, x)
type (rational), intent(out), volatile :: res
real, intent (in) :: x
integer :: x_floor
real :: x_frac
x_floor = floor (x)
x_frac = x - x_floor
if (x_frac == 0) then
res = x_floor // 1
else
res = (x_floor // 1) + (1 // floor (1 / x_frac))
end if
end subroutine assign_rational_real
function rational_add (r, s) result (res)
type (rational), intent (in) :: r
type (rational), intent (in) :: s
type (rational) :: res
res = r % numerator * s % denominator + r % denominator * s % numerator // &
& r % denominator * s % denominator
end function rational_add
function rational_minus (r) result (res) type (rational), intent (in) :: r type (rational) :: res res = - r % numerator // r % denominator end function rational_minus
function rational_subtract (r, s) result (res)
type (rational), intent (in) :: r
type (rational), intent (in) :: s
type (rational) :: res
res = r % numerator * s % denominator - r % denominator * s % numerator // &
& r % denominator * s % denominator
end function rational_subtract
function rational_multiply (r, s) result (res) type (rational), intent (in) :: r type (rational), intent (in) :: s type (rational) :: res res = r % numerator * s % numerator // r % denominator * s % denominator end function rational_multiply
function rational_divide (r, s) result (res) type (rational), intent (in) :: r type (rational), intent (in) :: s type (rational) :: res res = r % numerator * s % denominator // r % denominator * s % numerator end function rational_divide
function rational_lt (r, s) result (res)
type (rational), intent (in) :: r
type (rational), intent (in) :: s
type (rational) :: r_simple
type (rational) :: s_simple
logical :: res
r_simple = rational_simplify (r)
s_simple = rational_simplify (s)
res = r_simple % numerator * s_simple % denominator < &
& s_simple % numerator * r_simple % denominator
end function rational_lt
function rational_le (r, s) result (res)
type (rational), intent (in) :: r
type (rational), intent (in) :: s
type (rational) :: r_simple
type (rational) :: s_simple
logical :: res
r_simple = rational_simplify (r)
s_simple = rational_simplify (s)
res = r_simple % numerator * s_simple % denominator <= &
& s_simple % numerator * r_simple % denominator
end function rational_le
function rational_gt (r, s) result (res)
type (rational), intent (in) :: r
type (rational), intent (in) :: s
type (rational) :: r_simple
type (rational) :: s_simple
logical :: res
r_simple = rational_simplify (r)
s_simple = rational_simplify (s)
res = r_simple % numerator * s_simple % denominator > &
& s_simple % numerator * r_simple % denominator
end function rational_gt
function rational_ge (r, s) result (res)
type (rational), intent (in) :: r
type (rational), intent (in) :: s
type (rational) :: r_simple
type (rational) :: s_simple
logical :: res
r_simple = rational_simplify (r)
s_simple = rational_simplify (s)
res = r_simple % numerator * s_simple % denominator >= &
& s_simple % numerator * r_simple % denominator
end function rational_ge
function rational_eq (r, s) result (res) type (rational), intent (in) :: r type (rational), intent (in) :: s logical :: res res = r % numerator * s % denominator == s % numerator * r % denominator end function rational_eq
function rational_ne (r, s) result (res) type (rational), intent (in) :: r type (rational), intent (in) :: s logical :: res res = r % numerator * s % denominator /= s % numerator * r % denominator end function rational_ne
function rational_abs (r) result (res) type (rational), intent (in) :: r type (rational) :: res res = sign (r % numerator, r % denominator) // r % denominator end function rational_abs
function rational_int (r) result (res) type (rational), intent (in) :: r integer :: res res = r % numerator / r % denominator end function rational_int
function rational_modulo (r) result (res) type (rational), intent (in) :: r integer :: res res = modulo (r % numerator, r % denominator) end function rational_modulo
end module module_rational</lang> Example: <lang fortran>program perfect_numbers
use module_rational implicit none integer, parameter :: n_min = 2 integer, parameter :: n_max = 2 ** 19 - 1 integer :: n integer :: factor type (rational) :: sum
do n = n_min, n_max
sum = 1 // n
factor = 2
do
if (factor * factor >= n) then
exit
end if
if (modulo (n, factor) == 0) then
sum = rational_simplify (sum + (1 // factor) + (factor // n))
end if
factor = factor + 1
end do
if (sum % numerator == 1 .and. sum % denominator == 1) then
write (*, '(i0)') n
end if
end do
end program perfect_numbers</lang> Output: <lang>6 28 496 8128</lang>
GAP
Rational numbers are built-in. <lang gap>2/3 in Rationals;
- true
2/3 + 3/4;
- 17/12</lang>
Go
Go's big package implements arbitrary-precision integers and rational numbers.
<lang go>package main
import (
"fmt" "big" "math"
)
func main() {
max := int64(1<<19)
for candidate := int64(2); candidate < max; candidate++ {
sum := big.NewRat(1, candidate)
max2 := int64(math.Sqrt(float64(candidate)))
for factor := int64(2); factor <= max2; factor++ {
if candidate % factor == 0 {
sum.Add(sum, new(big.Rat).Add(big.NewRat(1, factor), big.NewRat(1, candidate / factor)))
}
}
if sum.Denom().Int64() == 1 {
perfectstring := ""
if sum.Num().Int64() == 1 { perfectstring = "perfect!" }
fmt.Printf("Sum of recipr. factors of %d = %d exactly %s\n",
candidate, sum.Num().Int64(), perfectstring)
}
}
}</lang>
It might be implemented like this:
[insert implementation here]
Haskell
Haskell provides a Rational type, which is really an alias for Ratio Integer (Ratio being a polymorphic type implementing rational numbers for any Integral type of numerators and denominators). The fraction is constructed using the % operator.
<lang haskell>import Data.Ratio
-- simply prints all the perfect numbers main = mapM_ print [candidate
| candidate <- [2 .. 2^19],
getSum candidate == 1]
where getSum candidate = 1 % candidate +
sum [1 % factor + 1 % (candidate `div` factor)
| factor <- [2 .. floor(sqrt(fromIntegral(candidate)))],
candidate `mod` factor == 0]</lang>
For a sample implementation of Ratio, see the Haskell 98 Report.
Icon and Unicon
Icon
The IPL provides support for rational arithmetic
- The data type is called 'rational' not 'frac'.
- Use the record constructor 'rational' to create a rational. Sign must be 1 or -1.
- Neither Icon nor Unicon supports operator overloading. Augmented assignments make little sense w/o this.
- Procedures include 'negrat' (unary -), 'addrat' (+), 'subrat' (-), 'mpyrat' (*), 'divrat' (modulo /).
Additional procedures are implemented here to complete the task:
- 'makerat' (make), 'absrat' (abs), 'eqrat' (=), 'nerat' (~=), 'ltrat' (<), 'lerat' (<=), 'gerat' (>=), 'gtrat' (>)
<lang Icon>procedure main()
limit := 2^19
write("Perfect numbers up to ",limit," (using rational arithmetic):")
every write(is_perfect(c := 2 to limit))
write("End of perfect numbers")
# verify the rest of the implementation
zero := makerat(0) # from integer
half := makerat(0.5) # from real
qtr := makerat("1/4") # from strings ...
one := makerat("1")
mone := makerat("-1")
verifyrat("eqrat",zero,zero)
verifyrat("ltrat",zero,half)
verifyrat("ltrat",half,zero)
verifyrat("gtrat",zero,half)
verifyrat("gtrat",half,zero)
verifyrat("nerat",zero,half)
verifyrat("nerat",zero,zero)
verifyrat("absrat",mone,)
end
procedure is_perfect(c) #: test for perfect numbers using rational arithmetic
rsum := rational(1, c, 1)
every f := 2 to sqrt(c) do
if 0 = c % f then
rsum := addrat(rsum,addrat(rational(1,f,1),rational(1,integer(c/f),1)))
if rsum.numer = rsum.denom = 1 then
return c
end</lang>
Sample output:
Perfect numbers up to 524288 (using rational arithmetic): 6 28 496 8128 End of perfect numbers Testing eqrat( (0/1), (0/1) ) ==> returned (0/1) Testing ltrat( (0/1), (1/2) ) ==> returned (1/2) Testing ltrat( (1/2), (0/1) ) ==> failed Testing gtrat( (0/1), (1/2) ) ==> failed Testing gtrat( (1/2), (0/1) ) ==> returned (0/1) Testing nerat( (0/1), (1/2) ) ==> returned (1/2) Testing nerat( (0/1), (0/1) ) ==> failed Testing absrat( (-1/1), ) ==> returned (1/1)
The following task functions are missing from the IPL: <lang Icon>procedure verifyrat(p,r1,r2) #: verification tests for rational procedures return write("Testing ",p,"( ",rat2str(r1),", ",rat2str(\r2) | &null," ) ==> ","returned " || rat2str(p(r1,r2)) | "failed") end
procedure makerat(x) #: make rational (from integer, real, or strings) local n,d static c initial c := &digits++'+-'
return case type(x) of {
"real" : real2rat(x)
"integer" : ratred(rational(x,1,1))
"string" : if x ? ( n := integer(tab(many(c))), ="/", d := integer(tab(many(c))), pos(0)) then
ratred(rational(n,d,1))
else
makerat(numeric(x))
}
end
procedure absrat(r1) #: abs(rational)
r1 := ratred(r1) r1.sign := 1 return r1
end
invocable all # for string invocation
procedure xoprat(op,r1,r2) #: support procedure for binary operations that cross denominators
local numer, denom, div
r1 := ratred(r1) r2 := ratred(r2)
return if op(r1.numer * r2.denom,r2.numer * r1.denom) then r2 # return right argument on success
end
procedure eqrat(r1,r2) #: rational r1 = r2 return xoprat("=",r1,r2) end
procedure nerat(r1,r2) #: rational r1 ~= r2 return xoprat("~=",r1,r2) end
procedure ltrat(r1,r2) #: rational r1 < r2 return xoprat("<",r1,r2) end
procedure lerat(r1,r2) #: rational r1 <= r2 return xoprat("<=",r1,r2) end
procedure gerat(r1,r2) #: rational r1 >= r2 return xoprat(">=",r1,r2) end
procedure gtrat(r1,r2) #: rational r1 > r2 return xoprat(">",r1,r2) end
link rational</lang>
The
provides rational and gcd in numbers. Record definition and usage is shown below:
<lang Icon> record rational(numer, denom, sign) # rational type
addrat(r1,r2) # Add rational numbers r1 and r2. divrat(r1,r2) # Divide rational numbers r1 and r2. medrat(r1,r2) # Form mediant of r1 and r2. mpyrat(r1,r2) # Multiply rational numbers r1 and r2. negrat(r) # Produce negative of rational number r. rat2real(r) # Produce floating-point approximation of r rat2str(r) # Convert the rational number r to its string representation. real2rat(v,p) # Convert real to rational with precision p (default 1e-10). Warning: excessive p gives ugly fractions reciprat(r) # Produce the reciprocal of rational number r. str2rat(s) # Convert the string representation (such as "3/2") to a rational number subrat(r1,r2) # Subtract rational numbers r1 and r2.
gcd(i, j) # returns greatest common divisor of i and j</lang>
Unicon
The Icon solution works in Unicon.
J
J implements rational numbers:
<lang j> 3r4*2r5 3r10</lang>
That said, note that J's floating point numbers work just fine for the stated problem: <lang j> is_perfect_rational=: 2 = (1 + i.) +/@:%@([ #~ 0 = |) ]</lang>
faster version (but the problem, as stated, is still tremendously inefficient): <lang j>factors=: */&>@{@((^ i.@>:)&.>/)@q:~&__ is_perfect_rational=: 2= +/@:%@,@factors</lang>
Exhaustive testing would take forever: <lang j> I.is_perfect_rational@"0 i.2^19 6 28 496 8128
I.is_perfect_rational@x:@"0 i.2^19x
6 28 496 8128</lang>
More limited testing takes reasonable amounts of time: <lang j> (#~ is_perfect_rational"0) (* <:@+:) 2^i.10x 6 28 496 8128</lang>
JavaScript
Lua
<lang lua>function gcd(a,b) return a == 0 and b or gcd(b % a, a) end
do
local function coerce(a, b)
if type(a) == "number" then return rational(a, 1), b end
if type(b) == "number" then return a, rational(b, 1) end
return a, b
end
rational = setmetatable({
__add = function(a, b)
local a, b = coerce(a, b)
return rational(a.num * b.den + a.den * b.num, a.den * b.den)
end,
__sub = function(a, b)
local a, b = coerce(a, b)
return rational(a.num * b.den - a.den * b.num, a.den * b.den)
end,
__mul = function(a, b)
local a, b = coerce(a, b)
return rational(a.num * b.num, a.den * b.den)
end,
__div = function(a, b)
local a, b = coerce(a, b)
return rational(a.num * b.den, a.den * b.num)
end,
__pow = function(a, b)
if type(a) == "number" then return a ^ (b.num / b.den) end
return rational(a.num ^ b, a.den ^ b) --runs into a problem if these aren't integers
end,
__concat = function(a, b)
if getmetatable(a) == rational then return a.num .. "/" .. a.den .. b end
return a .. b.num .. "/" .. b.den
end,
__unm = function(a) return rational(-a.num, -a.den) end}, {
__call = function(z, a, b) return setmetatable({num = a / gcd(a, b),den = b / gcd(a, b)}, z) end} )
end
print(rational(2, 3) + rational(3, 5) - rational(1, 10) .. "") --> 7/6 print((rational(4, 5) * rational(5, 9)) ^ rational(1, 2) .. "") --> 2/3 print(rational(45, 60) / rational(5, 2) .. "") --> 3/10 print(5 + rational(1, 3) .. "") --> 16/3
function findperfs(n)
local ret = {}
for i = 1, n do
sum = rational(1, i)
for fac = 2, i^.5 do
if i % fac == 0 then
sum = sum + rational(1, fac) + rational(fac, i)
end
end
if sum.den == sum.num then
ret[#ret + 1] = i
end
end
return table.concat(ret, '\n')
end print(findperfs(2^19))</lang>
Mathematica
Mathematica has full support for fractions built-in. If one divides two exact numbers it will be left as a fraction if it can't be simplified. Comparison, addition, division, product et cetera are built-in: <lang Mathematica>4/16 3/8 8/4 4Pi/2 16!/10! Sqrt[9/16] Sqrt[3/4] (23/12)^5 2 + 1/(1 + 1/(3 + 1/4))
1/2+1/3+1/5 8/Pi+Pi/8 //Together 13/17 + 7/31 Sum[1/n,{n,1,100}] (*summation of 1/1 + 1/2 + 1/3 + 1/4+ .........+ 1/99 + 1/100*)
1/2-1/3 a=1/3;a+=1/7
1/4==2/8 1/4>3/8 Pi/E >23/20 1/3!=123/370 Sin[3]/Sin[2]>3/20
Numerator[6/9] Denominator[6/9]</lang> gives back: <lang Mathematica>1/4 3/8 2 2 Pi 5765760 3/4 Sqrt[3]/2 6436343 / 248832 47/17
31/30 (64+Pi^2) / (8 Pi) 522 / 527 14466636279520351160221518043104131447711 / 2788815009188499086581352357412492142272
1/6 10/21
True False True True True
2 3</lang> As you can see, Mathematica automatically handles fraction as exact things, it doesn't evaluate the fractions to a float. It only does this when either the numerator or the denominator is not exact. I only showed integers above, but Mathematica can handle symbolic fraction in the same and complete way: <lang Mathematica>c/(2 c) (b^2 - c^2)/(b - c) // Cancel 1/2 + b/c // Together</lang> gives back: <lang Mathematica>1/2 b+c (2 b+c) / (2 c)</lang> Moreover it does simplification like Sin[x]/Cos[x] => Tan[x]. Division, addition, subtraction, powering and multiplication of a list (of any dimension) is automatically threaded over the elements: <lang Mathematica>1+2*{1,2,3}^3</lang> gives back: <lang Mathematica>{3, 17, 55}</lang> To check for perfect numbers in the range 1 to 2^25 we can use: <lang Mathematica>found={}; CheckPerfect[num_Integer]:=If[Total[1/Divisors[num]]==2,AppendTo[found,num]]; Do[CheckPerfect[i],{i,1,2^25}]; found</lang> gives back: <lang Mathematica>{6, 28, 496, 8128, 33550336}</lang> Final note; approximations of fractions to any precision can be found using the function N.
Objective-C
OCaml
OCaml's Num library implements arbitrary-precision rational numbers:
<lang ocaml>#load "nums.cma";; open Num;;
for candidate = 2 to 1 lsl 19 do
let sum = ref (num_of_int 1 // num_of_int candidate) in
for factor = 2 to truncate (sqrt (float candidate)) do
if candidate mod factor = 0 then
sum := !sum +/ num_of_int 1 // num_of_int factor
+/ num_of_int 1 // num_of_int (candidate / factor)
done;
if is_integer_num !sum then
Printf.printf "Sum of recipr. factors of %d = %d exactly %s\n%!"
candidate (int_of_num !sum) (if int_of_num !sum = 1 then "perfect!" else "")
done;;</lang>
It might be implemented like this:
[insert implementation here]
PARI/GP
Pari handles rational arithmetic natively. <lang>for(n=2,1<<19,
s=0; fordiv(n,d,s+=1/d); if(s==2,print(n))
)</lang>
Perl
Perl's Math::BigRat core module implements arbitrary-precision rational numbers. The bigrat pragma can be used to turn on transparent BigRat support:
<lang perl>use bigrat;
foreach my $candidate (2 .. 2**19) {
my $sum = 1 / $candidate;
foreach my $factor (2 .. sqrt($candidate)+1) {
if ($candidate % $factor == 0) {
$sum += 1 / $factor + 1 / ($candidate / $factor);
}
}
if ($sum->denominator() == 1) {
print "Sum of recipr. factors of $candidate = $sum exactly ", ($sum == 1 ? "perfect!" : ""), "\n";
}
}</lang>
It might be implemented like this:
[insert implementation here]
Perl 6
Perl 6 supports rational arithmetic natively. <lang perl6>for 2..2**19 -> $candidate {
my $sum = 1 / $candidate;
for 2 .. ceiling(sqrt($candidate)) -> $factor {
if $candidate %% $factor {
$sum += 1 / $factor + 1 / ($candidate / $factor);
}
}
if $sum.denominator == 1 {
say "Sum of reciprocal factors of $candidate = $sum exactly", ($sum == 1 ?? ", perfect!" !! ".");
}
} </lang> Note also that ordinary decimal literals are stored as Rats, so the following loop always stops exactly on 10 despite 0.1 not being exactly representable in floating point:
<lang perl6>for 1.0, 1.1, 1.2 ... 10 { .say }</lang> The arithmetic is all done in rationals, which are converted to floating-point just before display so that people don't have to puzzle out what 53/10 means.
PicoLisp
<lang PicoLisp>(load "@lib/frac.l")
(for (N 2 (> (** 2 19) N) (inc N))
(let (Sum (frac 1 N) Lim (sqrt N))
(for (F 2 (>= Lim F) (inc F))
(when (=0 (% N F))
(setq Sum
(f+ Sum
(f+ (frac 1 F) (frac 1 (/ N F))) ) ) ) )
(when (= 1 (cdr Sum))
(prinl
"Perfect " N
", sum is " (car Sum)
(and (= 1 (car Sum)) ": perfect") ) ) ) )</lang>
Output:
Perfect 6, sum is 1: perfect Perfect 28, sum is 1: perfect Perfect 120, sum is 2 Perfect 496, sum is 1: perfect Perfect 672, sum is 2 Perfect 8128, sum is 1: perfect Perfect 30240, sum is 3 Perfect 32760, sum is 3 Perfect 523776, sum is 2
Python
Python 3's standard library already implements a Fraction class:
<lang python>from fractions import Fraction
for candidate in range(2, 2**19):
sum = Fraction(1, candidate)
for factor in range(2, int(candidate**0.5)+1):
if candidate % factor == 0:
sum += Fraction(1, factor) + Fraction(1, candidate // factor)
if sum.denominator == 1:
print("Sum of recipr. factors of %d = %d exactly %s" %
(candidate, int(sum), "perfect!" if sum == 1 else ""))</lang>
It might be implemented like this:
<lang python>def lcm(a, b):
return a // gcd(a,b) * b
def gcd(u, v):
return gcd(v, u%v) if v else abs(u)
class Fraction:
def __init__(self, numerator, denominator):
common = gcd(numerator, denominator)
self.numerator = numerator//common
self.denominator = denominator//common
def __add__(self, frac):
common = lcm(self.denominator, frac.denominator)
n = common // self.denominator * self.numerator + common // frac.denominator * frac.numerator
return Fraction(n, common)
def __sub__(self, frac):
return self.__add__(-frac)
def __neg__(self):
return Fraction(-self.numerator, self.denominator)
def __abs__(self):
return Fraction(abs(self.numerator), abs(self.denominator))
def __mul__(self, frac):
return Fraction(self.numerator * frac.numerator, self.denominator * frac.denominator)
def __div__(self, frac):
return self.__mul__(frac.reciprocal())
def reciprocal(self):
return Fraction(self.denominator, self.numerator)
def __cmp__(self, n):
return int(float(self) - float(n))
def __float__(self):
return float(self.numerator / self.denominator)
def __int__(self):
return (self.numerator // self.denominator)</lang>
Ruby
Ruby's standard library already implements a Rational class:
<lang ruby>require 'rational'
for candidate in 2 .. 2**19:
sum = Rational(1, candidate)
for factor in 2 ... candidate**0.5
if candidate % factor == 0
sum += Rational(1, factor) + Rational(1, candidate / factor)
end
end
if sum.denominator == 1
puts "Sum of recipr. factors of %d = %d exactly %s" %
[candidate, sum.to_i, sum == 1 ? "perfect!" : ""]
end
end</lang>
It might be implemented like this:
[insert implementation here]
Scala
<lang scala> class Rational(n: Long, d:Long) extends Ordered[Rational] {
require(d!=0) private val g:Long = gcd(n, d) val numerator:Long = n/g val denominator:Long = d/g
def this(n:Long)=this(n,1)
def +(that:Rational):Rational=new Rational(
numerator*that.denominator + that.numerator*denominator,
denominator*that.denominator)
def -(that:Rational):Rational=new Rational(
numerator*that.denominator - that.numerator*denominator,
denominator*that.denominator)
def *(that:Rational):Rational=
new Rational(numerator*that.numerator, denominator*that.denominator)
def /(that:Rational):Rational=
new Rational(numerator*that.denominator, that.numerator*denominator)
def unary_~ :Rational=new Rational(denominator, numerator)
def unary_- :Rational=new Rational(-numerator, denominator)
def abs :Rational=new Rational(Math.abs(numerator), Math.abs(denominator))
override def compare(that:Rational):Int=
(this.numerator*that.denominator-that.numerator*this.denominator).toInt
override def toString()=numerator+"/"+denominator
private def gcd(x:Long, y:Long):Long=
if(y==0) x else gcd(y, x%y)
}
object Rational {
def apply(n: Long, d:Long)=new Rational(n,d) def apply(n:Long)=new Rational(n) implicit def longToRational(i:Long)=new Rational(i)
} </lang>
<lang scala> def find_perfects():Unit= {
for (candidate <- 2 until 1<<19)
{
var sum= ~Rational(candidate)
for (factor <- 2 until (Math.sqrt(candidate)+1).toInt)
{
if (candidate%factor==0)
sum+= ~Rational(factor)+ ~Rational(candidate/factor)
}
if (sum.denominator==1 && sum.numerator==1)
printf("Perfect number %d sum is %s\n", candidate, sum)
}
} </lang>
Slate
Slate uses infinite-precision fractions transparently.
<lang slate>54 / 7. 20 reciprocal. (5 / 6) reciprocal. (5 / 6) as: Float.</lang>
Smalltalk
Smalltalk uses naturally and transparently fractions (through the class Fraction):
st> 54/7 54/7 st> 54/7 + 1 61/7 st> 54/7 < 50 true st> 20 reciprocal 1/20 st> (5/6) reciprocal 6/5 st> (5/6) asFloat 0.8333333333333334
<lang smalltalk>| sum | 2 to: (2 raisedTo: 19) do: [ :candidate |
sum := candidate reciprocal.
2 to: (candidate sqrt) do: [ :factor |
( (candidate \\ factor) = 0 )
ifTrue: [
sum := sum + (factor reciprocal) + ((candidate / factor) reciprocal)
]
].
( (sum denominator) = 1 )
ifTrue: [
('Sum of recipr. factors of %1 = %2 exactly %3' %
{ candidate printString .
(sum asInteger) printString .
( sum = 1 ) ifTrue: [ 'perfect!' ]
ifFalse: [ ' ' ] }) displayNl
]
].</lang>
Tcl
TI-89 BASIC
While TI-89 BASIC has built-in rational and symbolic arithmetic, it does not have user-defined data types.