AKS test for primes You are encouraged to solve this task according to the task description, using any language you may know. AKS test for primes From Rosetta Code The AKS algorithm (http://www.cse.iitk.ac.in/users/manindra/algebra /primality_v6.pdf) for testing whether a number is prime is a polynomial-time algorithm based on an elementary theorem about Pascal triangles. The theorem on which the test is based can be stated as follows: a number p is prime if and only if all the coefficients of the polynomial expansion of (x − 1) p − (x p − 1) are divisible by p. For example, trying p = 3: (x − 1) 3 − (x 3 − 1) = (x 3 − 3x 2 + 3x − 1) − (x 3 − 1) = − 3x 2 + 3x And all the coefficients are divisible by 3 so 3 is prime. Note: This task is not the AKS primality test. It is an inefficient exponential time algorithm discovered in the late 1600s and used as an introductory lemma in the AKS derivation. The task Create a function/subroutine/method that given p generates the coefficients of the expanded polynomial representation of (x − 1) p . 1. Use the function to show here the polynomial expansions of (x − 1) p for p in the range 0 to at least 7, inclusive. 2. Use the previous function in creating another function that when given p returns whether p is prime using the theorem. 3. Use your test to generate a list of all primes under 35. 4. As a stretch goal, generate all primes under 50 (Needs greater than 31 bit integers). 5. References Agrawal-Kayal-Saxena (AKS) primality test (https://en.wikipedia.org/wiki/AKS_primality_test) (Wikipedia) Fool-Proof Test for Primes (http://www.youtube.com/watch?v=HvMSRWTE2mI) - Numberphile (Video). The accuracy of this video is disputed -- at best it is an oversimplification. Contents 1 ALGOL 68 2 AutoHotkey 1 of 53
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
AKS testfor primesYou areencouragedto solve this
task according to thetask description, usingany language you mayknow.
AKS test for primesFrom Rosetta Code
The AKS algorithm (http://www.cse.iitk.ac.in/users/manindra/algebra/primality_v6.pdf) for testing whether a number is prime is apolynomial-time algorithm based on an elementary theorem aboutPascal triangles.
The theorem on which the test is based can be stated as follows:
a number p is prime if and only if all the coefficients of thepolynomial expansion of
(x − 1)p − (xp − 1)
are divisible by p.
For example, trying p = 3:
(x − 1)3 − (x3 − 1) = (x3 − 3x2 + 3x − 1) − (x3 − 1) = − 3x2 + 3xAnd all the coefficients are divisible by 3 so 3 is prime.
Note:This task is not the AKS primality test. It is an inefficient exponential timealgorithm discovered in the late 1600s and used as an introductory lemma
in the AKS derivation.
The task
Create a function/subroutine/method that given p generates the coefficients of the expandedpolynomial representation of (x − 1)p.
1.
Use the function to show here the polynomial expansions of (x − 1)p for p in the range 0 to atleast 7, inclusive.
2.
Use the previous function in creating another function that when given p returns whether pis prime using the theorem.
3.
Use your test to generate a list of all primes under 35.4. As a stretch goal, generate all primes under 50 (Needs greater than 31 bit integers).5.
References
Agrawal-Kayal-Saxena (AKS) primality test (https://en.wikipedia.org/wiki/AKS_primality_test)(Wikipedia)Fool-Proof Test for Primes (http://www.youtube.com/watch?v=HvMSRWTE2mI) -Numberphile (Video). The accuracy of this video is disputed -- at best it is anoversimplification.
The code below uses Algol 68 Genie which provides arbitrary precision arithmetic for LONGLONG modes.
BEGINCOMMENT Mathematical preliminaries.
First note that the homogeneous polynomial (a+b)^n is symmetrical(to see this just swap the variables a and b). Therefore its
coefficients need be calculated only to that of (ab)^{n/2} for even n or (ab)^{(n-1)/2} for odd n.
Second, the coefficients are the binomial coefficients C(n,k) where the coefficient of a^k b^(n-k) is C(n,k) = n! / k! (k-1)!. This leads to an immediate and relatively efficient implementation for which we do not need to compute n! before dividing by k! and (k-1)! but, rather cancel common factors as we go along. Further, the well-known symmetry identity C(n,k) = C(n, n-k) allows a significant reduction in computational effort.
Third, (x-1)^n is the value of (a + b)^n when a=x and b = -1. The powers of -1 alternate between +1 and -1 so we may as well compute
(x+1)^n and negate every other coefficient when printing.COMMENT
PR precision=300 PRMODE LLI = LONG LONG INT; CO For brevity COPROC choose = (INT n, k) LLI :BEGIN
LLI result := 1;INT sym k := (k >= n%2 | n-k | k); CO Use symmetry COIF sym k > 0 THEN
FOR i FROM 0 TO sym k-1DO result TIMESAB (n-i); result OVERAB (i+1)OD
FI; result
END;PROC coefficients = (INT n) [] LLI :BEGIN
[0:n] LLI a;FOR i FROM 0 TO n%2DO a[i] := a[n-i] := choose (n, i) CO Use symmetry CO
OD; a
END;COMMENT First print the polynomials (x-1)^n, remembering to alternate signs and to tidy up the constant term, the x^1 term and the x^n term. This means we must treat (x-1)^0 and (x-1)^1 speciallyCOMMENT
FOR n FROM 0 TO 7DO
[0:n] LLI a := coefficients (n);printf (($"(x-1)^", g(0), " = "$, n));CASE n+1 IN
OUTprintf (($"x^", g(0)$, n));FOR i TO n-2DO printf (($xax, g(0), "x^", g(0)$, (ODD i | "-" | "+"), a[i], n-i))OD;printf (($xax, g(0), "x"$, (ODD (n-1) | "-" | "+"), a[n-1]));printf (($xaxg(0)l$, (ODD n | "-" | "+"), a[n]))
ESACOD;COMMENT Finally, for the "AKS" portion of the task, the sign of the coefficient has no effect on its divisibility by p so, once again, we may as well use the positive coefficients. Symmetry clearly
3 of 53
reduces the necessary number of tests by a factor of two.COMMENT
PROC is prime = (INT n) BOOL :BEGIN
BOOL prime := TRUE;FOR i FROM 1 TO n%2 WHILE prime DO prime := choose (n, i) MOD n = 0 OD;
primeEND;print ("Primes < 50 are ");FOR n FROM 2 TO 50 DO (is prime (n) | printf (($g(0)x$, n)) ) OD;print (newline);print ("And just to show off, the primes between 900 and 1000 are ");FOR n FROM 900 TO 1000 DO IF is prime (n) THEN printf (($g(0)x$, n)) FI OD;print (newline)
; 1. Create a function/subroutine/method that given p generates the coefficients of the expanded polynomial representation of (x-1)^p. ; Function modified from http://rosettacode.org/wiki/Pascal%27s_triangle#AutoHotkeypascalstriangle(n=8) ; n rows of Pascal's triangle{
p := Object(), z:=Object()Loop, % n
Loop, % row := A_Indexcol := A_Index, p[row, col] := row = 1 and col = 1
? 1: (p[row-1, col-1] = "" ; math operations on blanks return blanks; I want to assume zero
? 0: p[row-1, col-1])
- (p[row-1, col] = ""? 0: p[row-1, col])
Return p}
; 2. Use the function to show here the polynomial expansions of p for p in the range 0 to at least 7, inclusive.For k, v in pascalstriangle(){
s .= "`n(x-1)^" k-1 . "="For k, w in v
s .= "+" w "x^" k-1}s := RegExReplace(s, "\+-", "-")s := RegExReplace(s, "x\^0", "")s := RegExReplace(s, "x\^1", "x")Msgbox % clipboard := s
; 3. Use the previous function in creating another function that when given p returns whether p is prime using the AKS test.aks(n){
isnotprime := FalseFor k, v in pascalstriangle(n+1)[n+1]
(k != 1 and k != n+1) ? isnotprime |= !(v // n = v / n) ; if any is not divisible, returns true
4 of 53
Return !isnotprime}
; 4. Use your AKS test to generate a list of all primes under 35. i := 49p := pascalstriangle(i+1)Loop, % i{
n := A_Indexisnotprime := FalseFor k, v in p[n+1]
(k != 1 and k != n+1) ? isnotprime |= !(v // n = v / n) ; if any is not divisible, returns truet .= isnotprime ? "" : A_Index " "
Function maxes out at i = 61 as AutoHotkey supports up to 64-bit signed integers.
Bracmat
Bracmat automatically normalizes symbolic expressions with the algebraic binary operators +, *, ^and \L (logartithm). It can differentiate such expressions using the \D binary operator. (Theseoperators were implemented in Bracmat before all other operators!). Some algebraic values canexist in two evaluated forms. The equivalent x*(a+b) and x*a+x*b are both considered "normal", butx*(a+b)+-1 is not, and therefore expanded to -1+a*x+b*x. This is used in the forceExpansion function toconvert e.g. x*(a+b) to x*a+x*b.
The primality test uses a pattern that looks for a fractional factor. If such a factor is found, thetest fails. Otherwise it succeeds.
Small primes using the AKS test:[2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59, 61, 67, 71, 73, 79, 83, 89, 97]
EchoLisp
We use the math.lib library and the poly functions to compute and display the requiredpolynomials. A polynomial P(x) = a0 +a1*x + .. an*x^n is a list of coefficients (a0 a1 .... an).
(let ((P (poly-add (mono p) (aks-poly p))) (test (lambda(a) (zero? (modulo a p))))) ;; p divides a[i] ? (apply and (map test P)))) ;; returns #t if true for all a[i]
The Erlang io module can print out lists of characters with any level of nesting as a flat string.(e.g. ["Er", ["la", ["n"]], "g"] prints as "Erlang") which is useful when constructing the strings toprint out for the binomial expansions. The program also shows how lazy lists can be implementedin Erlang.
! Point #4do n = 2, 35if (is_prime(n)) write(*, '(I4)', advance='no') n
end do write(*, *)
! Point #5do n = 2, 124if (is_prime(n)) write(*, '(I4)', advance='no') n
end do write(*, *)
if (allocated(coeffs)) deallocate(coeffs)contains! Calculate coefficients of (x - 1)^n using binomial theoremsubroutine polynomial_expansion(n, coeffs)integer(kind=16), intent(in) :: ninteger(kind=16), dimension(:), allocatable, intent(out) :: coeffsinteger(kind=16) :: i, j
if (allocated(coeffs)) deallocate(coeffs)
allocate(coeffs(n + 1))
do i = 1, n + 1 coeffs(i) = binomial(n, i - 1)*(-1)**(n - i - 1)
end doend subroutine
! Calculate binomial coefficient using recurrent relation, as calculation! using factorial overflows too quickly.function binomial(n, k) result (res)integer(kind=16), intent(in) :: n, kinteger(kind=16) :: resinteger(kind=16) :: i
if (k == 0) then res = 1
returnend if
13 of 53
res = 1do i = 0, k - 1
res = res*(n - i)/(i + 1)end do
end function
! Outputs polynomial with given coefficientssubroutine print_polynom(coeffs)integer(kind=16), dimension(:), allocatable, intent(in) :: coeffsinteger(kind=4) :: i, pcharacter(len=40) :: cbuf, pbuflogical(kind=1) :: non_zero
if (.not. allocated(coeffs)) return
non_zero = .false.
do i = 1, size(coeffs)if (coeffs(i) .eq. 0) cycle
p = i - 1 write(cbuf, '(I40)') abs(coeffs(i)) write(pbuf, '(I40)') p
if (non_zero) thenif (coeffs(i) .gt. 0) then
write(*, fmt='(A)', advance='no') ' + 'else
write(*, fmt='(A)', advance='no') ' - 'endif
elseif (coeffs(i) .gt. 0) then
write(*, fmt='(A)', advance='no') ' 'else
write(*, fmt='(A)', advance='no') ' - 'endif
endif
if (p .eq. 0) then write(*, fmt='(A)', advance='no') trim(adjustl(cbuf))
elseif (p .eq. 1) thenif (coeffs(i) .eq. 1) then
write(*, fmt='(A)', advance='no') 'x'else
write(*, fmt='(A)', advance='no') trim(adjustl(cbuf))//'x'end if
! Test if n is prime using AKS test. Point #3.function is_prime(n) result (res)integer(kind=16), intent (in) :: nlogical(kind=1) :: resinteger(kind=16), dimension(:), allocatable :: coeffsinteger(kind=16) :: i
'METHOD -- Use the Pascal triangle to retrieve the coefficients'UPPER LIMIT OF FREEBASIC ULONGINT GETS PRIMES UP TO 70Sub string_split(s_in As String,char As String,result() As String)
Dim As String s=s_in,var1,var2Dim As Integer n,pst#macro split(stri,char,var1,var2)
pst=Instr(stri,char) var1="":var2=""
If pst<>0 Then var1=Mid(stri,1,pst-1) var2=Mid(stri,pst+1)
Else var1=stri
End IfRedim Preserve result(1 To 1+n-((Len(var1)>0)+(Len(var2)>0)))
result(n+1)=var1#endmacroDo
split(s,char,var1,var2):n=n+1:s=var2Loop Until var2=""Redim Preserve result(1 To Ubound(result)-1)
End Sub
'Get Pascal triangle componentsFunction pasc(n As Integer,flag As Integer=0) As String n+=1
Dim As Ulongint V(n):V(1)=1ulDim As String s,signFor r As Integer= 2 To n
s=""For i As Integer = r To 1 Step -1
V(i) += V(i-1)If i Mod 2=1 Then sign="" Else sign="-"
s+=sign+Str(V(i))+","Next i
Next rIf flag Then 'formatted output
Dim As String i,i2,i3,gRedim As String a(0)
16 of 53
string_split(s,",",a())For n1 As Integer=1 To Ubound(a)
If Left(a(n1),1)="-" Then sign="" Else sign="+"If n1=Ubound(a) Then i2="" Else i2=a(n1)If n1=2 Then i3="x" Else i3="x^"+Str(n1-1)If n1=1 Then i="":sign=" " Else i=i3
g+=sign+i2+i+" "Next n1
g="(x-1)^"+Str(n-1)+" = "+gReturn g
End IfReturn s
End Function
Function isprime(num As Integer) As IntegerRedim As String a(0)
string_split(pasc(num),",",a())For n As Integer=Lbound(a)+1 To Ubound(a)-1
If (Valulng(Ltrim(a(n),"-"))) Mod num<>0 Then Return 0Next nReturn -1
End Function'==================================== 'Formatted outputFor n As Integer=1 To 9
Print pasc(n,1)Next n
Print'Limit of Freebasic Ulongint sets about 70 maxPrint "Primes up to 70:"For n As Integer=2 To 70
-- Primes up to 100:[2,3,5,7,11,13,17,19,23,29,31,37,41,43,47,53,59,61,67,71,73,79,83,89,97]
J
Solution:
binomialExpansion =: (!~ * _1 ^ 2 | ]) i.&.:<: NB. 1) Create a function that gives the coefficients of (x-1)^p. testAKS =: 0 *./ .= ] | binomialExpansion NB. 3) Use that function to create another which determines whether p is prime usin
Examples:
binomialExpansion&.> i. 8 NB. 2) show the polynomial expansions p in the range 0 to at 7 inclusive.+-++--+----+-------+-----------+---------------+------------------+|0||_2|_3 3|_4 6 _4|_5 10 _10 5|_6 15 _20 15 _6|_7 21 _35 35 _21 7|+-++--+----+-------+-----------+---------------+------------------+
(#~ testAKS&> ) 2+i. 35 NB. 4) Generate a list of all primes under 35.2 3 5 7 11 13 17 19 23 29 31
In the #Prolog section of this page, it is shown how the symmetry of rows in a Pascal triangle canbe used to yield a more efficient test of primality than is apparently envisioned by the problemstatement. The key concept is the "OptPascal row", which is just the longest non-decreasingsequence of the corresponding Pascal row. In this article, the focus will therefore be on OptPascalrows.
NOTE: jq uses IEEE 754 64-bit numbers and thus if builtin arithmetic is used, is_prime will onlybe accurate up to 96 by this method because of loss of precision. The program below, however,can easily be adapted to use a BigInt library such as the one at https://github.com/joelpurra/jq-bigint
# add_pairs is a helper function for optpascal/0# Input: an OptPascal array# Output: the next OptPascal array (obtained by adding adjacent items, # but if the last two items are unequal, then their sum is repeated)def add_pairs: if length <= 1 then . elif length == 2 then (.[0] + .[1]) as $S | if (.[0] == .[1]) then [$S] else [$S,$S] end else [.[0] + .[1]] + (.[1:]|add_pairs) end;
# Input: an OptPascal row# Output: the next OptPascalRowdef next_optpascal: [1] + add_pairs;
# generate a stream of OptPascal arrays, beginning with []def optpascals: [] | recurse(next_optpascal);
# generate a stream of Pascal arraysdef pascals: # pascalize takes as input an OptPascal array and produces # the corresponding Pascal array; # if the input ends in a pair, then peel it off before reversing it. def pascalize: . + ((if .[-2] == .[-1] then .[0:-2] else .[0:-1] end) | reverse);
Task 1: "A method to generate the coefficients of (x-1)^p"
def coefficients: def alternate_signs: . as $in | reduce range(0; length) as $i ([]; . + [$in[$i] * (if $i % 2 == 0 then 1 else -1 end )]); (.+1) | pascal | alternate_signs;
Task 2: "Show here the polynomial expansions of (x − 1)^p for p in the range 0 to at least 7,inclusive."
range(0;8) | "Coefficient for (x - 1)^\(.): \(coefficients)"
Output:
Coefficients for (x - 1)^0: [1]Coefficients for (x - 1)^1: [1,-1]Coefficients for (x - 1)^2: [1,-2,1]Coefficients for (x - 1)^3: [1,-3,3,-1]Coefficients for (x - 1)^4: [1,-4,6,-4,1]Coefficients for (x - 1)^5: [1,-5,10,-10,5,-1]Coefficients for (x - 1)^6: [1,-6,15,-20,15,-6,1]Coefficients for (x - 1)^7: [1,-7,21,-35,35,-21,7,-1]
Task 3: Prime Number Test
For brevity, we show here only the relatively efficient solution based on optpascal/0:
def is_prime: . as $N | if . < 2 then false else (1+.) | optpascal | all( .[2:][]; . % $N == 0 ) end;
Task 4: "Use your AKS test to generate a list of all primes under 35."
range(0;36) | select(is_prime)
Output:
235711131719232931
Task 5: "As a stretch goal, generate all primes under 50."
[range(0;50) | select(is_prime)]
Output:
23 of 53
[2,3,5,7,11,13,17,19,23,29,31,37,41,43,47]
Julia
Task 1
function polycoefs(n::Int64) pc = typeof(n)[] if n < 0 return pc end sgn = one(n) for k in n:-1:0 push!(pc, sgn*binomial(n, k)) sgn = -sgn end return pcend
Perhaps this should be done with a comprehension, but properly accounting for the sign is trickyin that case.
Task 2
function stringpoly(n::Int64) if n < 0 return "" end st = @sprintf "(x - 1)^{%d} & = & " n for (i, c) in enumerate(polycoefs(n)) if i == 1 op = "" ac = c elseif c < 0 op = "-" ac = abs(c) else op = "+" ac = abs(c) end p = n + 1 - i if p == 0 st *= @sprintf " %s %d\\\\" op ac elseif ac == 1 st *= @sprintf " %s x^{%d}" op p else st *= @sprintf " %s %dx^{%d}" op ac p end end return stend
Of course this could be simpler, but this produces a nice payoff in typeset equations that do oninclude extraneous characters (leading pluses and coefficients of 1).
Task 3
function isaksprime(n::Int64) if n < 2
24 of 53
return false end for c in polycoefs(n)[2:(end-1)] if c%n != 0 return false end end return trueend
Task 4
println("<math>")println("\\begin{array}{lcl}")for i in 0:10 println(stringpoly(i))endprintln("\\end{array}")println("</math>\n")
L = 50print("AKS primes less than ", L, ": ")sep = ""for i in 1:L if isaksprime(i) print(sep, i) sep = ", " endendprintln()
Output:
AKS primes less than 50: 2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47
Maple
Maple handles algebraic manipulation of polynomials natively.
> for xpr in seq( expand( (x-1)^p ), p = 0 .. 7 ) do print( xpr ) end: 1
x - 1
2
25 of 53
x - 2 x + 1
3 2 x - 3 x + 3 x - 1
4 3 2 x - 4 x + 6 x - 4 x + 1
5 4 3 2 x - 5 x + 10 x - 10 x + 5 x - 1
6 5 4 3 2 x - 6 x + 15 x - 20 x + 15 x - 6 x + 1
7 6 5 4 3 2 x - 7 x + 21 x - 35 x + 35 x - 21 x + 7 x - 1
To implement the primality test, we write the following procedure that uses the (built-in)polynomial expansion to generate a list of coefficients of the expanded polynomial.
Use polc to implement prime? which does the primality test.
prime? := n -> n > 1 and {op}( map( modp, polc( n ), n ) ) = {0}
Of course, rather than calling polc, we can inline it, just for the sake of making the whole thing aone-liner (while adding argument type-checking for good measure):
prime? := (n::posint) -> n > 1 and {op}( map( modp, [coeffs]( expand( (x-1)^n - (x^n-1) ) ), n ) ) = {0}
This agrees with the built-in primality test isprime:
> evalb( seq( prime?(i), i = 1 .. 1000 ) = seq( isprime( i ), i = 1 .. 1000 ) ); true
Use prime? with the built-in Maple select procedure to pick off the primes up to 50:
use warnings;# Select one of these lines. Math::BigInt is in core, but quite slow.use Math::BigInt; sub binomial { Math::BigInt->new(shift)->bnok(shift) }# use Math::Pari "binomial";# use ntheory "binomial";
sub binprime {my $p = shift;return 0 unless $p >= 2;# binomial is symmetric, so only test half the termsfor (1 .. ($p>>1)) { return 0 if binomial($p,$_) % $p }1;
}sub coef { # For prettier printingmy($n,$e) = @_;return $n unless $e;$n = "" if $n==1;$e==1 ? "${n}x" : "${n}x^$e";
The ntheory module has implementations of the full AKS algorithm in Perl, C, and C+GMP. This isvastly faster than the method used in this task and is polynomial time, but like all current AKSimplementations is still much slower than other methods such as BPSW, APR-CL, and ECPP.
Library: ntheory
use ntheory ":all";# Uncomment next line to see the r and s values used. Set to 2 for more detail.# prime_set_config(verbose => 1);say join(" ", grep { is_aks_prime($_) } 1_000_000_000 .. 1_000_000_100);
sub polyprime($p where 2..*) { so expansions[$p].[1 ..^ */2].all %% $p }
The expansions are generated similarly to how most FP languages generate sequences thatresemble Pascal's triangle, using a zipwith meta-operator (Z) with subtraction, applied betweentwo lists that add a 0 on either end to the prior list. Here we define a constant infinite sequenceusing the ... sequence operator with a "whatever" endpoint. In fact, the second term [1,-1] couldhave been generated from the first term, but we put it in there for documentation so the readercan see what direction things are going.
The polyprime function pretty much reads like the original description. Is it "so" that the p'thexpansion's coefficients are all divisible by p? The .[1 ..^ */2] slice is done simply to weed outdivisions by 1 or by factors we've already tested (since the coefficients are symmetrical in termsof divisibility). If we wanted to write polyprime even more idiomatically, we could have made itanother infinite constant list that is just a mapping of the first list, but we decided that would justbe showing off. :-)
-- Does not work for primes above 53, which is actually beyond the original task anyway.-- Translated from the C version, just about everything is (working) out-by-1, what fun.
sequence c = repeat(0,100)
procedure coef(integer n)-- out-by-1, ie coef(1)==^0, coef(2)==^1, coef(3)==^2 etc. c[n] = 1 for i=n-1 to 2 by -1 do c[i] = c[i]+c[i-1] end forend procedure
function is_prime(integer n) coef(n+1); -- (I said it was out-by-1) for i=2 to n-1 do -- (technically "to n" is more correct) if remainder(c[i],n)!=0 then return 0 end if end for return 1end function
procedure show(integer n)-- (As per coef, this is (working) out-by-1)object ci for i=n to 1 by -1 do ci = c[i] if ci=1 then if remainder(n-i,2)=0 then if i=1 then if n=1 then ci = "1" else ci = "+1" end if else ci = "" end if else ci = "-1" end if else if remainder(n-i,2)=0 then ci = sprintf("+%d",ci) else ci = sprintf("-%d",ci) end if end if if i=1 then -- ie ^0 printf(1,"%s",{ci}) elsif i=2 then -- ie ^1 printf(1,"%sx",{ci}) else printf(1,"%sx^%d",{ci,i-1}) end if end forend procedure
procedure AKS_test_for_primes() for n=1 to 10 do -- (0 to 9 really) coef(n); printf(1,"(x-1)^%d = ", n-1); show(n);
32 of 53
puts(1,'\n'); end for
puts(1,"\nprimes (<=53):");-- coef(2); -- (needed to reset c, if we want to avoid saying 1 is prime...) c[2] = 1 -- (this manages "", which is all that call did anyway...) for n = 2 to 53 do if is_prime(n) then printf(1," %d", n); end if end for puts(1,'\n'); if getc(0) then end ifend procedure
The theorem as stated ties together two elementary concepts in mathematics: prime numbers andthe Pascal triangle. The simplicity of the connection can be expressed directly in Prolog by thefollowing prime number generator:
prime(P) :- pascal([1,P|Xs]), append(Xs, [1], Rest), forall( member(X,Xs), 0 is X mod P).
where pascal/1 is a generator of rows of the Pascal triangle, for example as defined below; theother predicates used above are standard.
This solution to the Rosetta Code problems will accordingly focus on the Pascal triangle, but toillustrate a number of points, we shall exploit its symmetry by representing each of its rows bythe longest initial non-decreasing segment of that row, as illustrated in the third column of thefollowing table:
We shall refer to this condensed representation of a row as an "optpascal list". Using it, we cansimplify and improve the above prime number generator by defining it as follows:
prime(N) :- optpascal([1,N|Xs]), forall( member(X,Xs), 0 is X mod N).
Using SWI-Prolog without modifying any of the memory management parameters, this primenumber generator was used to generate all primes up to and including 75,659.
Since Pascal triangles are the foundation of our approach to addressing the specific Rosetta Codeproblems, we begin by defining the generator pascal/2 that is required by the first problem, butwe do so by defining it in terms of an efficient generator, optpascal/1.
Pascal Triangle Generator
% To generate the n-th row of a Pascal triangle% pascal(+N, Row)pascal(0, [1]).pascal(N, Row) :- N > 0, optpascal( [1, N|Xs] ),!,
pascalize( [1, N|Xs], Row ).
pascalize( Opt, Row ) :-% if Opt ends in a pair, then peel off the pair:
% optpascal_successor(+P, -Q) is true if Q is an optpascal list beneath the optpascal list P:optpascal_successor(P, Q) :- optpascal(P, NextP),(Q = NextP ; optpascal_successor(NextP, Q)).
% optpascal(+Row, NextRow) is true if Row and NextRow are adjacent rows in the Pascal triangle.% optpascal(+Row, NextRow) where the optpascal representation is usedoptpascal(X, [1|Y]) :- add_pairs(X, Y).
% add_pairs(+OptPascal, NextOptPascal) is a helper function for optpascal/2.% Given one OptPascal list, it generates the next by adding adjacent% items, but if the last two items are unequal, then their sum is% repeated. This is intended to be a deterministic predicate, and to% avoid a probable compiler limitation, we therefore use one cut.add_pairs([], []).add_pairs([X], [X]).add_pairs([X,Y], Ans) :- S is X + Y,(X = Y -> Ans=[S] ; Ans=[S,S]),!. % To overcome potential limitation of compiler
add_pairs( [X1, X2, X3|Xs], [S|Ys]) :- S is X1 + X2, add_pairs( [X2, X3|Xs], Ys).
Solutions
Solutions with output from SWI-Prolog:
%%% Task 1: "A method to generate the coefficients of (1-X)^p"
% As required by the problem statement, but necessarily very inefficient::- between(0, 7, N), coefficients(N, Coefficients), writeln(Coefficients), fail ; true.
The following would be more efficient because backtracking saves recomputation:
:- coefficients(Coefficients), writeln(Coefficients), Coefficients = [_,N|_], N = -7.
%%% Task 3. Use the previous function in creating [sic]%%% another function that when given p returns whether p is prime%%% using the AKS test.
% Even for testing whether a given number, N, is prime, % this approach is inefficient, but here is a Prolog implementation:
prime_test_per_requirements(N) :- coefficients(N, [1|Coefficients]), append(Cs, [_], Coefficients), forall( member(C, Cs), 0 is C mod N).
The following is more efficient (because it relies on optpascal lists rather than the full array ofcoefficients), and more flexible (because it can be used to generate primes without requiringrecomputation):
prime(N) :- optpascal([1,N|Xs]), forall( member(X,Xs), 0 is X mod N).
%%% Task 4. Use your AKS test to generate a list of all primes under 35.
# small primes using the aks test[2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59, 61, 67, 71, 73, 79, 83, 89, 97]
Python: Output formatted for wiki
Using a wikitable and math features with the following additional code produces better formattedpolynomial output:
print('''{| class="wikitable" style="text-align:left;"|+ Polynomial Expansions and AKS prime test|-! <math>p</math>! <math>(x-1)^p</math>|-''')for p in range(12):
Borrowing heavily from Python listing. Optimized for the fact that the vector of the coefficients isa palindrome.
Is.Prime<-function(x){ expand<-function(p){ ex = 1 for (i in 0:(p/2-1)){ ex<-c(ex[1]*(p-i)/(i+1),ex) } return(rev(ex)[-1]) } return(as.logical(min(expand(x)%%x==0))) }
Racket
With copious use of the math/number-theory library...
#lang racket(require math/number-theory)
;; 1. coefficients of expanded polynomial (x-1)^p;; produces a vector because in-vector can provide a start;; and stop (of 1 and p) which allow us to drop the (-1)^p;; and the x^p terms, respectively.;;;; (vector-ref (coefficients p) e) is the coefficient for p^e(define (coefficients p) (for/vector ((e (in-range 0 (add1 p)))) (define sign (expt -1 (- p e))) (* sign (binomial p e))))
;; 2. Show the polynomial expansions from p=0 .. 7 (inclusive)
39 of 53
;; (it's possible some of these can be merged...)(define (format-coefficient c e leftmost?) (define (format-c.x^e c e) (define +c (abs c)) (match* (+c e) [(_ 0) (format "~a" +c)] [(1 _) (format "x^~a" e)] [(_ _) (format "~ax^~a" +c e)])) (define +/- (if (negative? c) "-" "+")) (define +c.x^e (format-c.x^e c e)) (match* (c e leftmost?) [(0 _ _) ""] [((? negative?) _ #t) (format "-~a" +c.x^e)] [(_ _ #t) +c.x^e] [(_ _ _) (format " ~a ~a" +/- +c.x^e)]))
(define (format-polynomial cs) (define cs-length (sequence-length cs)) (apply string-append (reverse ; convention is to display highest exponent first (for/list ((c cs) (e (in-naturals))) (format-coefficient c e (= e (sub1 cs-length)))))))
;; there is some discussion (see Discussion) about what to do with the perennial "1";; case. This is my way of saying that I'm ignoring it(define lowest-tested-number 2)
;; 4. list of numbers < 35 that are prime (note that 1 is prime;; by the definition of the AKS test for primes):(displayln (for/list ((i (in-range lowest-tested-number 35)) #:when (prime?/AKS i)) i))
;; 5. stretch goal: all prime numbers under 50(displayln (for/list ((i (in-range lowest-tested-number 50)) #:when (prime?/AKS i)) i))(displayln (for/list ((i (in-range lowest-tested-number 100)) #:when (prime?/AKS i)) i))
This REXX version is an optimized version (of version 1) and modified to address each of therequirements.The program determines programmatically the required number of digits (precision) for the large
41 of 53
coefficients.
/*REXX pgm calculates primes via the Agrawal-Kayal-Saxena (AKS) primality test*/parse arg Z .; if Z=='' then Z=200 /*Z not specified? Then use default.*/OZ=Z; tell=Z<0; Z=abs(Z) /*Is Z negative? Then show expression.*/numeric digits max(9,Z%3) /*define a dynamic # of decimal digits.*/$.0='-'; $.1="+"; @.=1 /*$.x: sign char; default coefficients.*/#= /*define list of prime numbers (so far)*/do p=3 for Z; pm=p-1; pp=p+1 /*PM & PP: used as a coding convenience*/
do m=2 for pp%2-1; mm=m-1 /*calculate coefficients for a power. */ @[email protected] + @.pm.m; h=pp-m /*calculate left side of coefficients*/ @[email protected] /* " right " " " */
end /*m*/ /* [↑] The M DO loop creates both */end /*p*/ /* sides in the same loop, saving */
/* a bunch of execution time. */if tell then say '(x-1)^0: 1' /*possibly display the first expression*/
/* [↓] test for primality by division.*/do n=2 for Z; nh=n%2; d=n-1 /*create expressions; find the primes.*/
do k=3 to nh while @.n.k//d==0 /*are coefficients divisible by N-1 ? */end /*k*/ /* [↑] skip the 1st & 2nd coefficients*/
/* [↓] multiple THEN─IF faster than &s*/if d\==1 then if d\==4 then if k>nh then #=# d /*add number to prime list.*/if \tell then iterate /*Don't tell? Don't show expressions.*/
y='(x-1)^'d": " /*define first part of the expression. */ s=1 /*S: is the sign indicator (-1│+1).*/
do j=n to 2 by -1 /*create the higher powers first. */if j==2 then xp='x' /*if power=1, then don't show the power*/
else xp='x^' || (j-1) /* ··· else show power with ^ */if j==n then y=y xp /*no sign (+│-) for the 1st expression.*/
else y=y $.s @.n.j'∙'xp /*build the expression with sign (+|-).*/ s=\s /*flip the sign for the next expression*/
end /*j*/ /* [↑] the sign (now) is either 0 │ 1,*//* and is displayed either - │ + */
say y $.s 1 /*just show the first N expressions, */end /*n*/ /* [↑] ··· but only for negative Z. */
say /* [↓] Has Z a leading + ? Then show.*/is="isn't"; if Z==word(. #,words(#)+1) then is='is' /*is or isn't a prime.*/if left(OZ,1)=='+' then say Z is 'prime.' /*tell if OZ has a +. */
else say 'primes:' # /*display prime # list. */say /* [↓] size of big 'un.*/say 'Found ' words(#) ' primes and the largest coefficient has' ,
length(@.pm.h) "decimal digits." /*stick a fork in it, we're all done. */
output for requirement #2, showing twenty expressions using as input: -20
Found 8 primes and the largest coefficient has 6 decimal digits.
output for requirement #3, showing if 2221 is prime (or not) using for input: +2221
42 of 53
(Output note: this number is really pushing at the limits of REXX's use of virtual memory; theversion ofRegina REXX used herein has a limit of around 2 Gbytes.)
2221 is prime.
Found 331 primes and the largest coefficient has 668 decimal digits.
output for requirement #4, showing all primes under 35 using the input: 35
primes: 2 3 5 7 11 13 17 19 23 29 31
Found 11 primes and the largest coefficient has 10 decimal digits.
output for requirement #5 (stretch goal), showing all primes under 50 using the input: 50
primes: 2 3 5 7 11 13 17 19 23 29 31 37 41 43 47
Found 15 primes and the largest coefficient has 15 decimal digits.
def prime?(p)return false if p < 2(x_minus_1_to_the(p) - Polynomial.from_string("x**#{p}-1")).coefs.all?{|n| n%p==0}
end
8.times do |n|# the default Polynomial#to_s would be OK here; the substitutions just make the# output match the other version below.puts "(x-1)^#{n} = #{x_minus_1_to_the(n).to_s.gsub(/\*\*/,'^').gsub(/\*/,'')}"
n=7; //For the expansion up to power of ng=50; //For test of primes up to g
function X = pascal(g) //Pascal´s triangle X(1,1)=1; //Zeroth power X(2,1)=1; //First power X(2,2)=1; for q=3:1:g+1 //From second power use this loop X(q,1)=1; X(q,q)=1; for p=2:1:q-1 X(q,p)=X(q-1,p-1)+X(q-1,p); end endendfunction
function cf=coef(Z,q,p) //Return coeffiecents for nicer display of expansion without "ones" if Z(q,p)==1 then cf=""; else cf=string(Z(q,p)); endendfunction
for q=3:n+1 //Generate and display the expansions Q(q)=strcat(["(x-1)^",string(q-1)," = "]); sing=""; //Sign of coeff. for p=1:q-1 //Number of coefficients equals power minus 1 Q(q)=strcat([Q(q),sing,coef(Z,q,p),"x^",string(q-p)]); if sing=="-" then sing="+"; else sing="-"; end end Q(q)=strcat([Q(q),sing,string(1)]); disp(Q(q)) clear Qend
function prime=prime(Z,g) prime="true"; for p=2:g if abs(floor(Z(g+1,p)/g)-Z(g+1,p)/g)>0 then prime="false"; break; end endendfunction
R="2"; //For nicer displayfor r=3:g
46 of 53
if prime(Z,r)=="true" then R=strcat([R, ", ",string(r)]); endenddisp(R)
const func array integer: expand_x_1 (in integer: p) is func result var array integer: ex is [] (1); local var integer: i is 0; begin for i range 0 to p - 1 do ex := [] (ex[1] * -(p - i) div (i + 1)) & ex; end for; end func;
const func boolean: aks_test (in integer: p) is func result var boolean: aks_test is FALSE; local var array integer: ex is 0 times 0; var integer: idx is 0; begin if p >= 2 then ex := expand_x_1(p); ex[1] +:= 1; for idx range 1 to pred(length(ex)) until ex[idx] rem p <> 0 do noop; end for; aks_test := idx = length(ex); end if; end func;
const proc: main is func local var integer: p is 0; var integer: n is 0; var integer: e is 0; begin writeln("# p: (x-1)^p for small p"); for p range 0 to 11 do write(p lpad 3 <& ": "); for n key e range expand_x_1(p) do write(" "); if n >= 0 then
47 of 53
write("+"); end if; write(n); if e > 1 then write("x^" <& pred(e)); end if; end for; writeln; end for; writeln; writeln("# small primes using the aks test"); for p range 0 to 61 do if aks_test(p) then write(p <& " "); end if; end for; writeln; end func;
func polynomialCoeffs(n: Int) -> [Int] { var result = [Int](count : n+1, repeatedValue : 0)
result[0]=1 for i in 1 ..< n/2+1 { //Progress up, until reaching the middle value result[i] = result[i-1] * (n-i+1)/i; } for i in n/2+1 ..< n+1 { //Copy the inverse of the first part result[i] = result[n-i]; } // Take into account the sign for i in stride(from: 1, through: n, by: 2) { result[i] = -result[i] }
return result}
func isPrime(n: Int) -> Bool {
var coeffs = polynomialCoeffs(n)
coeffs[0]-- coeffs[n]++
for i in 1 ... n { if coeffs[i]%n != 0 { return false } }
return true}
for i in 0...10 {
let coeffs = polynomialCoeffs(i)
print("(x-1)^\(i) = ") if i == 0 { print("1") } else { if i == 1 { print("x") } else { print("x^\(i)") if i == 2 { print("\(coeffs[i-1])x") } else { for j in 1...(i - 2) { if j%2 == 0 { print("+\(coeffs[j])x^\(i-j)") } else { print("\(coeffs[j])x^\(i-j)") } } if (i-1)%2 == 0 { print("+\(coeffs[i-1])x") } else {
primes (never mind the 1): 1 2 3 5 7 11 13 17 19 23 29 31
zkl
Translation of: Python
var BN=Import("zklBigNum");fcn expand_x_1(p){ ex := L(BN(1)); foreach i in (p){ ex.append(ex[-1] * -(p-i) / (i+1)) } return(ex.reverse()) }fcn aks_test(p){ if (p < 2) return(False); ex := expand_x_1(p);
52 of 53
ex[0] = ex[0] + 1; return(not ex[0,-1].filter('%.fp1(p)));}println("# p: (x-1)^p for small p");foreach p in (12){ println("%3d: ".fmt(p),expand_x_1(p).enumerate() .pump(String,fcn([(n,e)]){"%+d%s ".fmt(e,n and "x^%d".fmt(n) or "")}));}
println("\n# small primes using the aks test");println([0..110].filter(aks_test).toString(*));
# small primes using the aks testL(2,3,5,7,11,13,17,19,23,29,31,37,41,43,47,53,59,61,67,71,73,79,83,89,97,101,103,107,109)
Retrieved from "http://rosettacode.org/mw/index.php?title=AKS_test_for_primes&oldid=210996"Categories: Programming Tasks Prime Numbers ALGOL 68 AutoHotkey Bracmat CC sharp Clojure CoffeeScript Common Lisp D EchoLisp Erlang Fortran Go FreeBASICHaskell J Java JavaScript Jq Julia Maple Mathematica Wolfram Language OforthPARI/GP Pascal Perl Ntheory Perl 6 Phix PicoLisp Prolog PureBasic Python R RacketREXX Ruby Rust Scala Scilab Seed7 Sidef Swift Tcl UBasic/4tH Zkl
This page was last modified on 5 September 2015, at 17:36.Content is available under GNU Free Documentation License 1.2.