I'm working on modernizing Rosetta Code's infrastructure. Starting with communications. Please accept this time-limited open invite to RC's Slack.. --Michael Mol (talk) 20:59, 30 May 2020 (UTC)

Erdős–Woods numbers

From Rosetta Code
Erdős–Woods numbers is a draft programming task. It is not yet considered ready to be promoted as a complete task, for reasons that should be found in its talk page.
Description

A positive integer k is said to be an Erdős–Woods number if it has the following property: there exists a positive integer a such that in the sequence (a, a + 1, …, a + k) of consecutive integers, each of the elements has a non-trivial common factor with one of the endpoints. In other words, k is an Erdős–Woods number if there exists a positive integer a such that for each integer i between 0 and k, at least one of the greatest common divisors gcd(a, a + i) or gcd(a + i, a + k) is greater than 1.

It can be shown that there are infinitely many such numbers. Moreover, if k is an E-W number for some integer a, then one can find an infinite number of other a's using the formula a(jq + 1) where:

  • q is the product of all odd prime factors of a + k; and
  • j is any positive integer.


Example

16 is an E-W number (and also the first) for which the smallest value of a is 2184. This is because in the sequence 2184 to 2200 inclusive, the prime factors of the endpoints are:

  • 2³ x 3 x 7 x 13 = 2184
  • 2³ x 5² x 11 = 2200

and, if you check all the numbers between them, you will find that they all have a prime factor in common with at least one of the endpoints (2189 is divisible by 11, 2191 by 7, 2197 by 13 and the rest by either 2, 3, or 5).

Task

Compute and show here the first 20 E-W numbers together with their corresponding smallest value of a. If your language doesn't support arbitrary precision arithmetic, just compute and show as many as you can.

Extra credit

Do the same for the next 20 E-W numbers.

Note

Although all the E-W numbers relevant to this task are even, odd examples do exist the first one being k = 903.

References




Julia[edit]

Translation of: Python
""" modified from https://codegolf.stackexchange.com/questions/230509/find-the-erd%C5%91s-woods-origin/ """
 
using BitIntegers
 
"""
Returns the smallest value for `a` of Erdős-Woods number n, -1 if n is not in sequence
"""
function erdős_woods(n)
primes = Int[]
P = BigInt(1)
k = 1
while k < n
P % k != 0 && push!(primes, k)
P *= k * k
k += 1
end
divs = [evalpoly(2, [Int(a % p == 0) for p in primes]) for a in 0:n-1]
np = length(primes)
partitions = [(Int256(0), Int256(0), Int256(2)^np - 1)]
ort(x) = trailing_zeros(divs[x + 1] | divs[n - x + 1])
for i in sort(collect(1:n-1), lt = (b, c) -> ort(b) > ort(c))
new_partitions = Tuple{Int256, Int256, Int256}[]
factors = divs[i + 1]
other_factors = divs[n - i + 1]
for p in partitions
set_a, set_b, r_primes = p
if (factors & set_a != 0) || (other_factors & set_b != 0)
push!(new_partitions, p)
continue
end
for (ix, v) in enumerate(reverse(string(factors & r_primes, base=2)))
if v == '1'
w = Int256(1) << (ix - 1)
push!(new_partitions, (set_a ⊻ w, set_b, r_primes ⊻ w))
end
end
for (ix, v) in enumerate(reverse(string(other_factors & r_primes, base=2)))
if v == '1'
w = Int256(1) << (ix - 1)
push!(new_partitions, (set_a, set_b ⊻ w, r_primes ⊻ w))
end
end
end
partitions = new_partitions
end
result = Int256(-1)
for (px, py, _) in partitions
x, y = Int256(1), Int256(1)
for p in primes
isodd(px) && (x *= p)
isodd(py) && (y *= p)
px ÷= 2
py ÷= 2
end
newresult = ((n * invmod(x, y)) % y) * x - n
result = result == -1 ? newresult : min(result, newresult)
end
return result
end
 
function test_erdős_woods(startval=3, endval=116)
arr = fill((0, Int256(-1)), endval - startval + 1)
@Threads.threads for k in startval:endval
arr[k - startval + 1] = (k, erdős_woods(k))
end
ewvalues = filter(x -> last(x) > 0, arr)
println("The first $(length(ewvalues)) Erdős–Woods numbers and their minimum interval start values are:")
for (k, a) in ewvalues
println(lpad(k, 3), " -> $a")
end
end
 
test_erdős_woods()
 
Output:
Same as Wren example.

Phix[edit]

Translation of: Wren
Translation of: Julia

Using flag arrays instead of bigint bitfields

with javascript_semantics -- takes about 47s (of blank screen) though, also note the line 
                          -- below which triggered an unexpected violation on desktop/Phix
                          -- (the fix/hoist for which meant a need to swap result and tmp)
requires("1.0.1") -- mpz_invert()
include mpfr.e

function coppy(sequence p, integer ab, j)
    p = deep_copy(p)
    p[ab][j] = 1
    p[3][j] = 0
    return p
end function

function erdos_woods(integer n)
    integer n1 = n-1
    sequence primes = get_primes_le(n1),
             divs = repeat(0,n),
             trailing_zeros = repeat(0,n1)
    integer np = length(primes)
    for a=1 to n do
        sequence d = {}
        for i=np to 1 by -1 do
            d &= (remainder(a-1,primes[i])=0)
        end for
        divs[a] = d
    end for
    for i=1 to n1 do
        trailing_zeros[i] = rfind(1,sq_or(divs[i+1],divs[n-i+1]))
    end for
    sequence smc = custom_sort(trailing_zeros,tagset(n1)),
             partitions = {{repeat(0,np),repeat(0,np),repeat(1,np)}}
    for s=1 to length(smc) do
        integer i = smc[s]
        sequence new_partitions = {},
                 facts = divs[i+1],
                 other = divs[n-i+1]
        for p=1 to length(partitions) do
            sequence partp = partitions[p],
                     {setA,setB,rPrimes} = partp,
                     fsa = sq_and(facts,setA),
                     fob = sq_and(other,setB)
            if find(1,fsa) or find(1,fob) then
                new_partitions = append(new_partitions,partp)
            else
                for j=1 to np do
                    if facts[j] and rPrimes[j] then
                        new_partitions = append(new_partitions,coppy(partp,1,j))
                    end if
                    if other[j] and rPrimes[j] then
                        new_partitions = append(new_partitions,coppy(partp,2,j))
                    end if
                end for
            end if
        end for
        partitions = new_partitions
    end for
    mpz result = null
    mpz {x,y,tmp} = mpz_inits(3,1)
    for p=1 to length(partitions) do
        sequence {px,py} = partitions[p]
        -- triggers "p2js violation: JavaScript does not support string subscript destructuring"...
--      mpz {x,y,tmp} = mpz_inits(3,1)
        mpz_set_si(x,1)
        mpz_set_si(y,1)
        for i=1 to np do
            integer pi = primes[np-i+1]
            if px[i] then mpz_mul_si(x,x,pi) end if
            if py[i] then mpz_mul_si(y,y,pi) end if
        end for
        mpz_invert(tmp,x,y)
        mpz_mul_si(tmp,tmp,n)
        mpz_mod(tmp,tmp,y)
        mpz_mul(tmp,tmp,x)
        mpz_sub_si(tmp,tmp,n)
        if result=null then
            result = tmp
            tmp = mpz_init()
        elsif mpz_cmp(tmp,result)=-1 then
            {result,tmp} = {tmp,result}
        end if
    end for
    return result
end function
 
integer k = 3, count = 0
printf(1,"The first 20 Erdos–Woods numbers and their minimum interval start values are:\n")
while count<20 do
    mpz a = erdos_woods(k)
    if a then
        printf(1,"%3d -> %s\n", {k, mpz_get_str(a)})
        count += 1
    end if
    k += 1
end while
Output:
Same as Wren example.

Python[edit]

Original author credit to the Stackexchange website user ovs, who in turn credits user xash.

""" modified from https://codegolf.stackexchange.com/questions/230509/find-the-erd%C5%91s-woods-origin/ """
 
def erdős_woods(n):
""" Returns the smallest value for `a` of the Erdős-Woods number n, or Inf if n is not in the sequence """
primes = []
P = k = 1
while k < n:
if P % k:
primes.append(k)
P *= k * k
k += 1
divs = [
int(''.join(str((a%p==0) + 0) for p in primes)[::-1], 2)
for a in range(n)
]
np = len(primes)
partitions = [(0, 0, 2**np-1)]
for i in sorted(
range(1,n),
key = lambda x: bin(divs[x] | divs[n-x])[::-1].find('1'),
reverse=True
):
new_partitions = []
factors = divs[i]
other_factors = divs[n-i]
for p in partitions:
set_a, set_b, r_primes = p
if factors & set_a or other_factors & set_b:
new_partitions += (p,)
continue
for ix, v in enumerate(bin(factors & r_primes)[2:][::-1]):
if v=='1':
w = 1 << ix
new_partitions += ((set_a^w, set_b, r_primes^w),)
for ix, v in enumerate(bin(other_factors & r_primes)[2:][::-1]):
if v=='1':
w = 1 << ix
new_partitions += ((set_a, set_b^w, r_primes^w),)
partitions = new_partitions
result = float('inf')
for px, py, _ in partitions:
x = y = 1
for p in primes:
if px % 2:
x *= p
if py % 2:
y *= p
px //= 2
py //= 2
result = min(result, n*pow(x,-1,y)%y*x-n)
return result
 
 
K = 3
COUNT = 0
print('The first 20 Erdős–Woods numbers and their minimum interval start values are:')
while COUNT < 20:
a = erdős_woods(K)
if a != float('inf'):
print(f"{K: 3d} -> {a}")
COUNT += 1
K += 1
 
Output:
Same as Wren example.

Wren[edit]

Wren-cli[edit]

Library: Wren-big
Library: Wren-fmt
Library: Wren-sort
Library: Wren-trait

It's not easy to find a way of doing this in a reasonable time.

I ended up translating the Python 3.8 code (the more readable version) here, 6th post down, and found the first 20 E-W numbers in around 93 seconds. Much slower than Python which has arbitrary precision numerics built-in nowadays but acceptable for Wren.

import "./big" for BigInt
import "./fmt" for Conv, Fmt
import "./sort" for Sort
import "./trait" for Indexed
 
var zero = BigInt.zero
var one = BigInt.one
var two = BigInt.two
 
var ew = Fn.new { |n|
var primes = []
var k = 1
var P = one
while (k < n) {
if ((P % k) != 0) primes.add(k)
P = P * k * k
k = k + 1
}
var divs = []
var np = primes.count
if (np > 0) {
for (a in 0...n) {
var A = BigInt.new(a)
var s = primes.map { |p| Conv.btoi(A % p == 0).toString }.join()[-1..0]
divs.add(BigInt.new(Conv.atoi(s, 2)))
}
}
var partitions = [ [zero, zero, two.pow(np) - one] ]
var key = Fn.new { |x| (divs[x] | divs[n-x]).toBaseString(2)[-1..0].indexOf("1") }
var cmp = Fn.new { |i, j| (key.call(j) - key.call(i)).sign }
for (i in Sort.merge((1...n).toList, cmp)) {
var newPartitions = []
var factors = divs[i]
var otherFactors = divs[n-i]
for (p in partitions) {
var setA = p[0]
var setB = p[1]
var rPrimes = p[2]
if ((factors & setA) != zero || (otherFactors & setB) != zero) {
newPartitions.add(p)
continue
}
for (se in Indexed.new((factors & rPrimes).toBaseString(2)[-1..0])) {
var ix = se.index
var v = se.value
if (v == "1") {
var w = one << ix
newPartitions.add([setA ^ w, setB, rPrimes ^ w])
}
}
for (se in Indexed.new((otherFactors & rPrimes).toBaseString(2)[-1..0])) {
var ix = se.index
var v = se.value
if (v == "1") {
var w = one << ix
newPartitions.add([setA, setB ^ w, rPrimes ^ w])
}
}
}
partitions = newPartitions
}
var result = null
for (p in partitions) {
var px = p[0]
var py = p[1]
var x = one
var y = one
for (p in primes) {
if ((px % two) == one) x = x * p
if ((py % two) == one) y = y * p
px = px / two
py = py / two
}
var N = BigInt.new(n)
var temp = x.modInv(y) * N % y * x - N
result = result ? BigInt.min(result, temp) : temp
}
return result
}
 
var k = 3
var count = 0
System.print("The first 20 Erdős–Woods numbers and their minimum interval start values are:")
while (count < 20) {
var a = ew.call(k)
if (a) {
Fmt.print("$3d -> $i", k, a)
count = count + 1
}
k = k + 1
}
Output:
The first 20 Erdős–Woods numbers and their minimum interval start values are:
 16 -> 2184
 22 -> 3521210
 34 -> 47563752566
 36 -> 12913165320
 46 -> 21653939146794
 56 -> 172481165966593120
 64 -> 808852298577787631376
 66 -> 91307018384081053554
 70 -> 1172783000213391981960
 76 -> 26214699169906862478864
 78 -> 27070317575988954996883440
 86 -> 92274830076590427944007586984
 88 -> 3061406404565905778785058155412
 92 -> 549490357654372954691289040
 94 -> 38646299993451631575358983576
 96 -> 50130345826827726114787486830
100 -> 35631233179526020414978681410
106 -> 200414275126007376521127533663324
112 -> 1022681262163316216977769066573892020
116 -> 199354011780827861571272685278371171794


Embedded[edit]

Library: Wren-gmp

Takes about 15.4 seconds which is significantly faster than Wren-cli, but still nowhere near as fast as the Python version - a majestic 1.2 seconds!

import "./gmp" for Mpz
import "./fmt" for Conv, Fmt
import "./sort" for Sort
import "./trait" for Indexed
 
var zero = Mpz.zero
var one = Mpz.one
var two = Mpz.two
 
var ew = Fn.new { |n|
var primes = []
var k = 1
var P = Mpz.one
while (k < n) {
if (!P.isDivisibleUi(k)) primes.add(k)
P.mul(k * k)
k = k + 1
}
var divs = []
var np = primes.count
if (np > 0) {
for (a in 0...n) {
var A = Mpz.from(a)
var s = primes.map { |p| Conv.btoi(A % p == 0).toString }.join()[-1..0]
divs.add(Mpz.from(Conv.atoi(s, 2)))
}
}
var partitions = [ [Mpz.zero, Mpz.zero, Mpz.two.pow(np) - one] ]
var key = Fn.new { |x| (divs[x] | divs[n-x]).toString(2)[-1..0].indexOf("1") }
var cmp = Fn.new { |i, j| (key.call(j) - key.call(i)).sign }
for (i in Sort.merge((1...n).toList, cmp)) {
var newPartitions = []
var factors = divs[i]
var otherFactors = divs[n-i]
for (p in partitions) {
var setA = p[0]
var setB = p[1]
var rPrimes = p[2]
if ((factors & setA) != zero || (otherFactors & setB) != zero) {
newPartitions.add(p)
continue
}
for (se in Indexed.new((factors & rPrimes).toString(2)[-1..0])) {
var ix = se.index
var v = se.value
if (v == "1") {
var w = one << ix
newPartitions.add([setA ^ w, setB, rPrimes ^ w])
}
}
for (se in Indexed.new((otherFactors & rPrimes).toString(2)[-1..0])) {
var ix = se.index
var v = se.value
if (v == "1") {
var w = one << ix
newPartitions.add([setA, setB ^ w, rPrimes ^ w])
}
}
}
partitions = newPartitions
}
var result = null
for (p in partitions) {
var px = p[0].copy()
var py = p[1].copy()
var x = Mpz.one
var y = Mpz.one
for (p in primes) {
if (px.isOdd) x.mul(p)
if (py.isOdd) y.mul(p)
px.div(2)
py.div(2)
}
var N = Mpz.from(n)
var x2 = x.copy()
var temp = x.modInv(y).mul(N).rem(y).mul(x2).sub(N)
result = result ? Mpz.min(result, temp) : temp
}
return result
}
 
var k = 3
var count = 0
System.print("The first 20 Erdős–Woods numbers and their minimum interval start values are:")
while (count < 20) {
var a = ew.call(k)
if (a) {
Fmt.print("$3d -> $i", k, a)
count = count + 1
}
k = k + 1
}
Output:
Same as Wren-cli version.