z3 number of solutions - python

How do I use z3 to count the number of solutions? For example, I want to prove that for any n, there are 2 solutions to the set of equations {x^2 == 1, y_1 == 1, ..., y_n == 1}. The following code shows satisfiability for a given n, which isn't quite what I want (I want number of solutions for an arbitrary n).
#!/usr/bin/env python
from z3 import *
# Add the equations { x_1^2 == 1, x_2 == 1, ... x_n == 1 } to s and return it.
def add_constraints(s, n):
assert n > 1
X = IntVector('x', n)
s.add(X[0]*X[0] == 1)
for i in xrange(1, n):
s.add(X[i] == 1)
return s
s = Solver()
add_constraints(s, 3)
s.check()
s.model()

If there are a finite number of solutions, you can use the disjunct of the constants (your x_i's) not equal to their assigned model values to enumerate all of them. If there are infinite solutions (which is the case if you want to prove this for all natural numbers n), you can use the same technique, but of course couldn't enumerate them all, but could use this to generate many solutions up to some bound you pick. If you want to prove this for all n > 1, you will need to use quantifiers. I've added a discussion of this below.
While you didn't quite ask this question, you should see this question/answer as well: Z3: finding all satisfying models
Here's your example doing this (z3py link here: http://rise4fun.com/Z3Py/643M ):
# Add the equations { x_1^2 == 1, x_2 == 1, ... x_n == 1 } to s and return it.
def add_constraints(s, n, model):
assert n > 1
X = IntVector('x', n)
s.add(X[0]*X[0] == 1)
for i in xrange(1, n):
s.add(X[i] == 1)
notAgain = []
i = 0
for val in model:
notAgain.append(X[i] != model[val])
i = i + 1
if len(notAgain) > 0:
s.add(Or(notAgain))
print Or(notAgain)
return s
for n in range(2,5):
s = Solver()
i = 0
add_constraints(s, n, [])
while s.check() == sat:
print s.model()
i = i + 1
add_constraints(s, n, s.model())
print i # solutions
If you want to prove there are no other solutions for any choice of n, you need to use quantifiers, since the previous approach will only work for finite n (and it gets very expensive quickly). Here is an encoding showing this proof. You could generalize this to incorporate the model generation capability in the previous part to come up with the +/- 1 solution for a more general formula. If the equation has a number of solutions independent of n (like in your example), this would allow you to prove equations have some finite number of solutions. If the number of solutions is a function of n, you'd have to figure that function out. z3py link: http://rise4fun.com/Z3Py/W9En
x = Function('x', IntSort(), IntSort())
s = Solver()
n = Int('n')
# theorem says that x(1)^2 == 1 and that x(1) != +/- 1, and forall n >= 2, x(n) == 1
# try removing the x(1) != +/- constraints
theorem = ForAll([n], And(Implies(n == 1, And(x(n) * x(n) == 1, x(n) != 1, x(n) != -1) ), Implies(n > 1, x(n) == 1)))
#s.add(Not(theorem))
s.add(theorem)
print s.check()
#print s.model() # unsat, no model available, no other solutions

Related

Let n be a square number. Using Python, how we can efficiently calculate natural numbers y up to a limit l such that n+y^2 is again a square number?

Using Python, I would like to implement a function that takes a natural number n as input and outputs a list of natural numbers [y1, y2, y3, ...] such that n + y1*y1 and n + y2*y2 and n + y3*y3 and so forth is again a square.
What I tried so far is to obtain one y-value using the following function:
def find_square(n:int) -> tuple[int, int]:
if n%2 == 1:
y = (n-1)//2
x = n+y*y
return (y,x)
return None
It works fine, eg. find_square(13689) gives me a correct solution y=6844. It would be great to have an algorithm that yields all possible y-values such as y=44 or y=156.
Simplest slow approach is of course for given N just to iterate all possible Y and check if N + Y^2 is square.
But there is a much faster approach using integer Factorization technique:
Lets notice that to solve equation N + Y^2 = X^2, that is to find all integer pairs (X, Y) for given fixed integer N, we can rewrite this equation to N = X^2 - Y^2 = (X + Y) * (X - Y) which follows from famous school formula of difference of squares.
Now lets rename two factors as A, B i.e. N = (X + Y) * (X - Y) = A * B, which means that X = (A + B) / 2 and Y = (A - B) / 2.
Notice that A and B should be of same odditiy, either both odd or both even, otherwise in last formulas above we can't have whole division by 2.
We will factorize N into all possible pairs of two factors (A, B) of same oddity. For fast factorization in code below I used simple to implement but yet quite fast algorithm Pollard Rho, also two extra algorithms were needed as a helper to Pollard Rho, one is Fermat Primality Test (which allows fast checking if number is probably prime) and second is Trial Division Factorization (which helps Pollard Rho to factor out small factors, which could cause Pollard Rho to fail).
Pollard Rho for composite number has time complexity O(N^(1/4)) which is very fast even for 64-bit numbers. Any faster factorization algorithm can be chosen if needed a bigger space to be searched. My fast algorithm time is dominated by speed of factorization, remaining part of algorithm is blazingly fast, just few iterations of loop with simple formulas.
If your N is a square itself (hence we know its root easily), then Pollard Rho can factor N even much faster, within O(N^(1/8)) time. Even for 128-bit numbers it means very small time, 2^16 operations, and I hope you're solving your task for less than 128 bit numbers.
If you want to process a range of possible N values then fastest way to factorize them is to use techniques similar to Sieve of Erathosthenes, using set of prime numbers, it allows to compute all factors for all N numbers within some range. Using Sieve of Erathosthenes for the case of range of Ns is much faster than factorizing each N with Pollard Rho.
After factoring N into pairs (A, B) we compute (X, Y) based on (A, B) by formulas above. And output resulting Y as a solution of fast algorithm.
Following code as an example is implemented in pure Python. Of course one can use Numba to speed it up, Numba usually gives 30-200 times speedup, for Python it achieves same speed as optimized C++. But I thought that main thing here is to implement fast algorithm, Numba optimizations can be done easily afterwards.
I added time measurement into following code. Although it is pure Python still my fast algorithm achieves 8500x times speedup compared to regular brute force approach for limit of 1 000 000.
You can change limit variable to tweak amount of searched space, or num_tests variable to tweak amount of different tests.
Following code implements both solutions - fast solution find_fast() described above plus very tiny brute force solution find_slow() which is very slow as it scans all possible candidates. This slow solution is only used to compare correctness in tests and compare speedup.
Code below uses nothing except few standard Python library modules, no external modules were used.
Try it online!
def find_slow(N):
import math
def is_square(x):
root = int(math.sqrt(float(x)) + 0.5)
return root * root == x, root
l = []
for y in range(N):
if is_square(N + y ** 2)[0]:
l.append(y)
return l
def find_fast(N):
import itertools, functools
Prod = lambda it: functools.reduce(lambda a, b: a * b, it, 1)
fs = factor(N)
mfs = {}
for e in fs:
mfs[e] = mfs.get(e, 0) + 1
fs = sorted(mfs.items())
del mfs
Ys = set()
for take_a in itertools.product(*[
(range(v + 1) if k != 2 else range(1, v)) for k, v in fs]):
A = Prod([p ** t for (p, _), t in zip(fs, take_a)])
B = N // A
assert A * B == N, (N, A, B, take_a)
if A < B:
continue
X = (A + B) // 2
Y = (A - B) // 2
assert N + Y ** 2 == X ** 2, (N, A, B, X, Y)
Ys.add(Y)
return sorted(Ys)
def trial_div_factor(n, limit = None):
# https://en.wikipedia.org/wiki/Trial_division
fs = []
while n & 1 == 0:
fs.append(2)
n >>= 1
all_checked = False
for d in range(3, (limit or n) + 1, 2):
if d * d > n:
all_checked = True
break
while True:
q, r = divmod(n, d)
if r != 0:
break
fs.append(d)
n = q
if n > 1 and all_checked:
fs.append(n)
n = 1
return fs, n
def fermat_prp(n, trials = 32):
# https://en.wikipedia.org/wiki/Fermat_primality_test
import random
if n <= 16:
return n in (2, 3, 5, 7, 11, 13)
for i in range(trials):
if pow(random.randint(2, n - 2), n - 1, n) != 1:
return False
return True
def pollard_rho_factor(n):
# https://en.wikipedia.org/wiki/Pollard%27s_rho_algorithm
import math, random
fs, n = trial_div_factor(n, 1 << 7)
if n <= 1:
return fs
if fermat_prp(n):
return sorted(fs + [n])
for itry in range(8):
failed = False
x = random.randint(2, n - 2)
for cycle in range(1, 1 << 60):
y = x
for i in range(1 << cycle):
x = (x * x + 1) % n
d = math.gcd(x - y, n)
if d == 1:
continue
if d == n:
failed = True
break
return sorted(fs + pollard_rho_factor(d) + pollard_rho_factor(n // d))
if failed:
break
assert False, f'Pollard Rho failed! n = {n}'
def factor(N):
import functools
Prod = lambda it: functools.reduce(lambda a, b: a * b, it, 1)
fs = pollard_rho_factor(N)
assert N == Prod(fs), (N, fs)
return sorted(fs)
def test():
import random, time
limit = 1 << 20
num_tests = 20
t0, t1 = 0, 0
for i in range(num_tests):
if (round(i / num_tests * 1000)) % 100 == 0 or i + 1 >= num_tests:
print(f'test {i}, ', end = '', flush = True)
N = random.randrange(limit)
tb = time.time()
r0 = find_slow(N)
t0 += time.time() - tb
tb = time.time()
r1 = find_fast(N)
t1 += time.time() - tb
assert r0 == r1, (N, r0, r1, t0, t1)
print(f'\nTime slow {t0:.05f} sec, fast {t1:.05f} sec, speedup {round(t0 / max(1e-6, t1))} times')
if __name__ == '__main__':
test()
Output:
test 0, test 2, test 4, test 6, test 8, test 10, test 12, test 14, test 16, test 18, test 19,
Time slow 26.28198 sec, fast 0.00301 sec, speedup 8732 times
For the easiest solution, you can try this:
import math
n=13689 #or we can ask user to input a square number.
for i in range(1,9999):
if math.sqrt(n+i**2).is_integer():
print(i)

How to find sum of cubes of the divisors for every number from 1 to input number x in python where x can be very large

Examples,
1.Input=4
Output=111
Explanation,
1 = 1³(divisors of 1)
2 = 1³ + 2³(divisors of 2)
3 = 1³ + 3³(divisors of 3)
4 = 1³ + 2³ + 4³(divisors of 4)
------------------------
sum = 111(output)
1.Input=5
Output=237
Explanation,
1 = 1³(divisors of 1)
2 = 1³ + 2³(divisors of 2)
3 = 1³ + 3³(divisors of 3)
4 = 1³ + 2³ + 4³(divisors of 4)
5 = 1³ + 5³(divisors of 5)
-----------------------------
sum = 237 (output)
x=int(raw_input().strip())
tot=0
for i in range(1,x+1):
for j in range(1,i+1):
if(i%j==0):
tot+=j**3
print tot
Using this code I can find the answer for small number less than one million.
But I want to find the answer for very large numbers. Is there any algorithm
for how to solve it easily for large numbers?
Offhand I don't see a slick way to make this truly efficient, but it's easy to make it a whole lot faster. If you view your examples as matrices, you're summing them a row at a time. This requires, for each i, finding all the divisors of i and summing their cubes. In all, this requires a number of operations proportional to x**2.
You can easily cut that to a number of operations proportional to x, by summing the matrix by columns instead. Given an integer j, how many integers in 1..x are divisible by j? That's easy: there are x//j multiples of j in the range, so divisor j contributes j**3 * (x // j) to the grand total.
def better(x):
return sum(j**3 * (x // j) for j in range(1, x+1))
That runs much faster, but still takes time proportional to x.
There are lower-level tricks you can play to speed that in turn by constant factors, but they still take O(x) time overall. For example, note that x // j == 1 for all j such that x // 2 < j <= x. So about half the terms in the sum can be skipped, replaced by closed-form expressions for a sum of consecutive cubes:
def sum3(x):
"""Return sum(i**3 for i in range(1, x+1))"""
return (x * (x+1) // 2)**2
def better2(x):
result = sum(j**3 * (x // j) for j in range(1, x//2 + 1))
result += sum3(x) - sum3(x//2)
return result
better2() is about twice as fast as better(), but to get faster than O(x) would require deeper insight.
Quicker
Thinking about this in spare moments, I still don't have a truly clever idea. But the last idea I gave can be carried to a logical conclusion: don't just group together divisors with only one multiple in range, but also those with two multiples in range, and three, and four, and ... That leads to better3() below, which does a number of operations roughly proportional to the square root of x:
def better3(x):
result = 0
for i in range(1, x+1):
q1 = x // i
# value i has q1 multiples in range
result += i**3 * q1
# which values have i multiples?
q2 = x // (i+1) + 1
assert x // q1 == i == x // q2
if i < q2:
result += i * (sum3(q1) - sum3(q2 - 1))
if i+1 >= q2: # this becomes true when i reaches roughly sqrt(x)
break
return result
Of course O(sqrt(x)) is an enormous improvement over the original O(x**2), but for very large arguments it's still impractical. For example better3(10**6) appears to complete instantly, but better3(10**12) takes a few seconds, and better3(10**16) is time for a coffee break ;-)
Note: I'm using Python 3. If you're using Python 2, use xrange() instead of range().
One more
better4() has the same O(sqrt(x)) time behavior as better3(), but does the summations in a different order that allows for simpler code and fewer calls to sum3(). For "large" arguments, it's about 50% faster than better3() on my box.
def better4(x):
result = 0
for i in range(1, x+1):
d = x // i
if d >= i:
# d is the largest divisor that appears `i` times, and
# all divisors less than `d` also appear at least that
# often. Account for one occurence of each.
result += sum3(d)
else:
i -= 1
lastd = x // i
# We already accounted for i occurrences of all divisors
# < lastd, and all occurrences of divisors >= lastd.
# Account for the rest.
result += sum(j**3 * (x // j - i)
for j in range(1, lastd))
break
return result
It may be possible to do better by extending the algorithm in "A Successive Approximation Algorithm for Computing the Divisor Summatory Function". That takes O(cube_root(x)) time for the possibly simpler problem of summing the number of divisors. But it's much more involved, and I don't care enough about this problem to pursue it myself ;-)
Subtlety
There's a subtlety in the math that's easy to miss, so I'll spell it out, but only as it pertains to better4().
After d = x // i, the comment claims that d is the largest divisor that appears i times. But is that true? The actual number of times d appears is x // d, which we did not compute. How do we know that x // d in fact equals i?
That's the purpose of the if d >= i: guarding that comment. After d = x // i we know that
x == d*i + r
for some integer r satisfying 0 <= r < i. That's essentially what floor division means. But since d >= i is also known (that's what the if test ensures), it must also be the case that 0 <= r < d. And that's how we know x // d is i.
This can break down when d >= i is not true, which is why a different method needs to be used then. For example, if x == 500 and i == 51, d (x // i) is 9, but it's certainly not the case that 9 is the largest divisor that appears 51 times. In fact, 9 appears 500 // 9 == 55 times. While for positive real numbers
d == x/i
if and only if
i == x/d
that's not always so for floor division. But, as above, the first does imply the second if we also know that d >= i.
Just for Fun
better5() rewrites better4() for about another 10% speed gain. The real pedagogical point is to show that it's easy to compute all the loop limits in advance. Part of the point of the odd code structure above is that it magically returns 0 for a 0 input without needing to test for that. better5() gives up on that:
def isqrt(n):
"Return floor(sqrt(n)) for int n > 0."
g = 1 << ((n.bit_length() + 1) >> 1)
d = n // g
while d < g:
g = (d + g) >> 1
d = n // g
return g
def better5(x):
assert x > 0
u = isqrt(x)
v = x // u
return (sum(map(sum3, (x // d for d in range(1, u+1)))) +
sum(x // i * i**3 for i in range(1, v)) -
u * sum3(v-1))
def sum_divisors(n):
sum = 0
i = 0
for i in range (1, n) :
if n % i == 0 and n != 0 :
sum = sum + i
# Return the sum of all divisors of n, not including n
return sum
print(sum_divisors(0))
# 0
print(sum_divisors(3)) # Should sum of 1
# 1
print(sum_divisors(36)) # Should sum of 1+2+3+4+6+9+12+18
# 55
print(sum_divisors(102)) # Should be sum of 2+3+6+17+34+51
# 114

An implementation for the double factorial

I have written this piece of code that implements the double factorial in Python both iteratively and recursively; the code works without problems, but I'm interested in improving my overall programming style. Here's the code:
def semif_r(n): #recursive implementation
if n == 0 or n == 1:
z = 1
else:
z= n * semif_r(n-2)
return z
def semif_i(n): #iterative implementation
N = 1
if n == 0 or n == 1:
return 1
elif n%2 == 1:
for i in range(0,n/2):
N = (2*i + 1)*N
VAL = N
return n*VAL
elif n%2 == 0:
for i in range(0,n/2):
N = (2*i+2)*N
VAL = N
return VAL
I hope that some experienced programmers can give me some feedback about improving my code!
from operator import mul
semif_pythonic = lambda x: reduce(mul, xrange(x, 1, -2))
I don't really understand why you need the VAL variable since it is equal to N; just use N.
You may write: N *= (2*i + 1) rather than N = (2*i + 1)*N but if you don't want to use this way, maybe it would still be better to write N = N * (2*i + 1) because it is easier to read.
For arithmetic functions, write n//2 rather than n/2 because both are different in Python 3; writing n//2 is more portable accross the different versions of Python.
As a challenge, you may want to try writing a third version as a tail-recursive function by using the tco module: http://baruchel.github.io/python/2015/11/07/explaining-functional-aspects-in-python/

Can someone explain to me this part of Dixon's factorization algorithm?

I've been trying to implement Dixon's factorization method in python, and I'm a bit confused. I know that you need to give some bound B and some number N and search for numbers between sqrtN and N whose squares are B-smooth, meaning all their factors are in the set of primes less than or equal to B. My question is, given N of a certain size, what determines B so that the algorithm will produce non-trivial factors of N? Here is a wikipedia article about the algorithm, and if it helps, here is my code for my implementation:
def factor(N, B):
def isBsmooth(n, b):
factors = []
for i in b:
while n % i == 0:
n = int(n / i)
if not i in factors:
factors.append(i)
if n == 1 and factors == b:
return True
return False
factor1 = 1
while factor1 == 1 or factor1 == N:
Bsmooth = []
BsmoothMod = []
for i in range(int(N ** 0.5), N):
if len(Bsmooth) < 2 and isBsmooth(i ** 2 % N, B):
Bsmooth.append(i)
BsmoothMod.append(i ** 2 % N)
gcd1 = (Bsmooth[0] * Bsmooth[1]) % N
gcd2 = int((BsmoothMod[0] * BsmoothMod[1]) ** 0.5)
factor1 = gcd(gcd1 - gcd2, N)
factor2 = int(N / factor1)
return (factor1, factor2)
Maybe someone could help clean my code up a bit, too? It seems very inefficient.
This article discusses the optimal size for B: https://web.archive.org/web/20160205002504/https://vmonaco.com/dixons-algorithm-and-the-quadratic-sieve/. Briefly, the optimal value is thought to be exp((logN loglogN)^(1/2)).
[ I wrote this for a different purpose, but you might find it interesting. ]
Given x2 ≡ y2 (mod n) with x ≠ ± y, about half the time gcd(x−y, n) is a factor of n. This congruence of squares, observed by Maurice Kraitchik in the 1920s, is the basis for several factoring methods. One of those methods, due to John Dixon, is important in theory because its sub-exponential run time can be proven, though it is too slow to be useful in practice.
Dixon's method begins by choosing a bound b &approx; e√(log n log log n) and identifying the factor base of all primes less than b that are quadratic residues of n (their jacobi symbol is 1).
function factorBase(n, b)
fb := [2]
for p in tail(primes(b))
if jacobi(n, p) == 1
append p to fb
return fb
Then repeatedly choose an integer r on the range 1 < r < n, calculate its square modulo n, and if the square is smooth over the factor base add it to a list of relations, stopping when there are more relations than factors in the factor base, plus a small reserve for those cases that fail. The idea is to identify a set of relations, using linear algebra, where the factor base primes combine to form a square. Then take the square root of the product of all the factor base primes in the relations, take the product of the related r, and calculate the gcd to identify the factor.
struct rel(x, ys)
function dixon(n, fb, count)
r, rels := floor(sqrt(n)), []
while count > 0
fs := smooth((r * r) % n, fb)
if fs is not null
append rel(r, fs) to rels
count := count - 1
r := r + 1
return rels
A number n is smooth if all its factors are in the factor base, which is determined by trial division; the smooth function returns a list of factors, which is null if n doesn't completely factor over the factor base.
function smooth(n, fb)
fs := []
for f in fb
while n % f == 0
append f to fs
n := n / f
if n == 1 return fs
return []
A factor is determined by submitting the accumulated relations to the linear algebra of the congruence of square solver.
For example, consider the factorization of 143. Choose r = 17, so r2 ≡ 3 (mod 143). Then choose r = 19, so r2 ≡ 75 ≡ 3 · 52. Those two relations can be combined as (17 · 19)2 ≡ 32 · 52 ≡ 152 (mod 143), and the two factors are gcd(17·19 − 15, 143) = 11 and gcd(17·19 + 15, 143) = 13. This sometimes fails; for instance, the relation 212 ≡ 22 (mod 143) can be combined with the relation on 19, but the two factors produced, 1 and 143, are trivial.
Thanks for very interesting question!
In pure Python I implemented from scratch Dixon Factorization Algorithm in 3 different flavors:
Using simplest sieve. I'm creating u64 array with all numbers in range [N; N * 2), which signify z^2 value. This array hold result of multiplication of prime numbers. Then through sieving process I iterate all factor base prime numbers and do array[k] *= p in those k positions that are divisible by p. Finally when sieved array is ready I check both that a) array index k is a perfect square, b) and array[k] == k - N. Second b) condition means that all multiplied p primes give final number, this is only true if number is divisible only by factor-base primes, i.e. it is B-smooth. This is simplest and most slowest out of my 3 solutions.
Second solution uses SymPy library to factorize every z^2. I iterate all possible z and do sympy.factorint(z * z), this gives factorization of z^2. If this factorization contains only small primes, i.e. from factor base, then I collect such z and z^2 for later processing. This version of algorithm is also slow, but much faster than first one.
Third solution uses a kind of sieving used in Quadratic Sieve. This sieving process is fastest of all three algorithms. Basically what it does, it finds all roots of equation x^2 = N (mod p) for all primes in factor base, as I have just few primes root finding is done through simple loop through all variants, for bigger primes one can use Shanks Tonelli algorithm of finding root, which is really fast. Only around 50% of primes give a root solution at all, hence only half of primes are actually used in Quadratic Sieve. Roots of such equation can be used to generate lots of solutions at once, because root + k * p is also a valid solution for all k. Sieving is done through array[offset(root) :: p] += Log2(p). Here instead of multiplication of first algorithm I used adding a logarithm of prime. First it is a bit faster to add a number than to multiply. Secondly, what is more important is that it supports any size of number, e.g. even 256-bit. While multiplying is possible only till 64-bit number, because Numpy has no 128 or 256 bit integers support. After logartithms are added, I check which logarithms are equal to logarithm of original z^2 number, this numbers are final sieved numbers.
After all three algorithms above have sieved all z^2 then I do Linear Algebra stage through Gaussian Elemination algorithm. This stage is meant to find such combination of B-smooth z^2 numbers which after multiplication of their prime factors give final number with all EVEN prime powers.
Lets call a Relation a triple z, z^2, prime factors of z^2. Basically all relations are given to Gaussian Elemination stage, where even combinations are found.
Even powers of prime numbers give us equality a^2 = b^2 (mod N), from where we can get a factor by doing factor = GCD(a + b, N), here GCD is Greatest Common Divisor found through Euclidean Algorithm. This GCD sometimes gives trivial factors 1 and N, in this case other even combinations should be checked.
To be 100% sure to get even combinations I do Sieving stage till I find a bit more than amount of prime numbers amount of relations, actually around 105% of amount of prime numbers. This extra 5% of relations ensure us that we certainly will get dependent linear equations in Gaussian stage. All these dependent equation form even combinations.
Actually we need a bit more dependent equations, not just 1 more than amount of primes, but around 5%-10% more, only because some (50-60% of them as I can see experimentally) dependencies give only trivial factor 1 or N. Hence extra equations are needed.
Put a look at console output at the end of my post. This console output shows all the impressions from my program. There I run in parallel (multi-threaded) both 2nd (Sieve_B) and 3rd (Sieve_C) algorithms. 1st one (Sieve_A) is not run by my program because it is so slow that you'll wait forever for it to finish.
At the very end of source file you can tweak variable bits = 64 to some other size, like bits = 96. This is amount of bits in composite number N. This N is created as a product of just two random prime numbers of equal size. Such a composite consisting of two equal in size primes is usually called RSA Number.
Also find B = 1 << 10, this tells degree of B-smoothness, basically factor base consists of all possible primes < B. You may increase this B limit, this will give more frequent answers of sieved z^2 hence whole factoring becomes much faster. The only limitation of huge size of B is Linear Algebra stage (Gaussian Elemination), because with bigger factor base you have to solve more linear equations of bigger size. And my Gauss is done not in very optimal way, for example instead of keeping bits as np.uint8 you may keep bits as dense np.uint64, this will increase Linear Algebra speed by 8x times more.
You may also find variable M = 1 << 23, which tells how large is sieving array size, in other words it is block size that is processed at once. Bigger block is a bit faster, but not much. Bigger values of M will not give much difference because it only tells what size of tasks sieving process is split into, it doesn't influence any computation power. More than that bigger M will occupy more memory, so you can't increases it infinitely, only till you have enough memory.
Besides all mentioned above algorithms I also used Fermat Primality Test, also Sieve of Eratosthenes (for generating prime factor base).
Plus also implemented my own algorithm of filtering square numbers. For this I take some composite modulus that looks close to Primorial, like mod = 2 * 2 * 2 * 3 * 3 * 5 * 7 * 11 * 13. And inside boolean array I mark all numbers modulus mod that are squares. Later when any number K should be checked if it is square or not I get flag_array[K % mod] and if it is True then number is "Possibly" squares, while if it is False then number is "Definitely" not square. Thus this filter gives false positives sometimes but never false negatives. This filter checking stage filters out 95% of non-squares, remaining 5% of possibly squares can be double-checked through math.isqrt().
Please, click below on Try it online! link, to test run my program on online server of ReplIt. This will give you best impression, especially if you have no Python or no personal laptop. My code below can be just run straight away after only PIP-installing python -m pip numpy sympy.
Try it online!
import threading
def GenPrimes_SieveOfEratosthenes(end):
import numpy as np
composites = np.zeros((end,), dtype = np.uint8)
for p in range(2, len(composites)):
if composites[p]:
continue
if p * p >= end:
break
composites[p * p :: p] = 1
primes = []
for p in range(2, len(composites)):
if not composites[p]:
primes.append(p)
return np.array(primes, dtype = np.uint32)
def Print(*pargs, __state = (threading.RLock(),), **nargs):
with __state[0]:
print(*pargs, flush = True, **nargs)
def IsSquare(n, *, state = []):
if len(state) == 0:
import numpy as np
Print('Pre-computing squares filter...')
squares_filter = 2 * 2 * 2 * 3 * 3 * 5 * 7 * 11 * 13
squares = np.zeros((squares_filter,), dtype = np.uint8)
squares[(np.arange(0, squares_filter, dtype = np.uint64) ** 2) % squares_filter] = 1
state.extend([squares_filter, squares])
if not state[1][n % state[0]]:
return False, None
import math
root = math.isqrt(n)
return root ** 2 == n, root
def FactorRef(x):
import sympy
return dict(sorted(sympy.factorint(x).items()))
def CheckZ(z, N, primes):
z2 = pow(z, 2, N)
factors = FactorRef(z2)
assert all(p <= primes[-1] for p in factors), (primes[-1], factors, N, z, z2)
return z
def SieveSimple(N, primes):
import time, math, numpy as np
Print('Simple Sieve of B-smooth z^2...')
sieve_block = 1 << 21
rep0_time = 0
for iiblock, iblock in enumerate(range(N, N * 2, sieve_block)):
if time.time() - rep0_time >= 30:
Print(f'Block {iiblock:>3} (2^{math.log2(max(iblock - N, 1)):>5.2f})')
rep0_time = time.time()
iblock_end = iblock + sieve_block
sieve_arr = np.ones((sieve_block,), dtype = np.uint64)
iblock_modN = iblock % N
for p in primes:
mp = 1
while True:
if mp * p >= sieve_block:
break
mp *= p
off = (mp - iblock_modN % mp) % mp
sieve_arr[off :: mp] *= p
for i in range(1 if iblock == N else 0, sieve_block):
num = iblock + i
z2 = num - N
if sieve_arr[i] < z2:
continue
assert sieve_arr[i] == z2, (sieve_arr[i], round(math.log2(sieve_arr[i]), 3), z2)
is_square, z = IsSquare(num)
if not is_square:
continue
#Print('z', z, 'z^2', z2)
yield CheckZ(z, N, primes)
def SieveFactor(N, primes):
import math
Print('Factor Sieve of B-smooth z^2...')
for iz, z in enumerate(range(math.isqrt(N - 1) + 1, math.isqrt(N * 2 - 1) + 1)):
z2 = z ** 2 - N
assert 0 <= z2 and z2 < N, (z, z2)
factors = FactorRef(z2)
if any(p > primes[-1] for p in factors):
continue
#Print('iz', iz, 'z', z, 'z^2', z2, 'z^2 factors', factors)
yield CheckZ(z, N, primes)
def BinarySearch(begin, end, Test):
while begin + 1 < end:
mid = (begin + end - 1) >> 1
if Test(mid):
end = mid + 1
else:
begin = mid + 1
assert begin + 1 == end and Test(begin), (begin, end, Test(begin))
return begin
def ModSqrt(n, p):
n %= p
def Ret(x):
if pow(x, 2, p) != n:
return []
nx = (p - x) % p
if x == nx:
return [x]
elif x <= nx:
return [x, nx]
else:
return [nx, x]
#if p % 4 == 3 and sympy.isprime(p):
# return Ret(pow(n, (p + 1) // 4, p))
for i in range(p):
if pow(i, 2, p) == n:
return Ret(i)
return []
def SieveQuadratic(N, primes):
import math, numpy as np
# https://en.wikipedia.org/wiki/Quadratic_sieve
# https://www.rieselprime.de/ziki/Multiple_polynomial_quadratic_sieve
M = 1 << 23
def Log2I(x):
return int(round(math.log2(max(1, x)) * (1 << 24)))
def Log2IF(li):
return li / (1 << 24)
Print('Quadratic Sieve of B-smooth z^2...')
plogs = {}
for p in primes:
plogs[int(p)] = Log2I(int(p))
qprimes = []
B = int(primes[-1]) + 1
for p in primes:
p = int(p)
res = []
mp = 1
while True:
if mp * p >= B:
break
mp *= p
roots = ModSqrt(N, mp)
if len(roots) == 0:
if mp == p:
break
continue
res.append((mp, tuple(roots)))
if len(res) > 0:
qprimes.append(res)
qprimes_lin = np.array([pinfo[0][0] for pinfo in qprimes], dtype = np.uint32)
yield qprimes_lin
Print('QSieve num primes', len(qprimes), f'({len(qprimes) * 100 / len(primes):.1f}%)')
x_begin0 = math.isqrt(N - 1) + 1
assert N <= x_begin0 ** 2
for iblock in range(1 << 30):
if (x_begin0 + (iblock + 1) * M) ** 2 - N >= N:
break
x_begin = x_begin0 + iblock * M
if iblock != 0:
Print('\n', end = '')
Print(f'Block {iblock} (2^{math.log2(max(1, x_begin ** 2 - N)):>6.2f})...')
a = np.zeros((M,), np.uint32)
for pinfo in qprimes:
p = pinfo[0][0]
plog = np.uint32(plogs[p])
for imp, (mp, roots) in enumerate(pinfo):
off_done = set()
for root in roots:
for off in range(mp):
if ((x_begin + off) ** 2 - N) % mp == 0 and off not in off_done:
break
else:
continue
a[off :: mp] += plog
off_done.add(off)
logs = np.log2(np.array((np.arange(M).astype(np.float64) + x_begin) ** 2 - N, dtype = np.float64))
logs2if = Log2IF(a.astype(np.float64))
logs_diff = np.abs(logs - logs2if)
for ix in range(M):
if logs_diff[ix] > 0.3:
continue
z = x_begin + ix
z2 = z * z - N
factors = FactorRef(z2)
assert all(p <= primes[-1] for p, c in factors.items())
#Print('iz', ix, 'z', z, 'z^2', z2, f'(2^{math.log2(max(1, z2)):>6.2f})', ', z^2 factors', factors)
yield CheckZ(z, N, primes)
def LinAlg(N, zs, primes):
import numpy as np
Print('Linear algebra...')
Print('Factoring...')
m = np.zeros((len(zs), len(primes) + len(zs)), dtype = np.uint8)
def SwapRows(i, j):
t = np.copy(m[i])
m[i][...] = m[j][...]
m[j][...] = t[...]
def MatToStr(m):
s = '\n'
for i in range(len(m)):
for j in range(len(m[i])):
s += str(m[i, j])
s += '\n'
return s[1:-1]
for iz, z in enumerate(zs):
z2 = z * z - N
fs = FactorRef(z2)
for p, c in fs.items():
i = np.searchsorted(primes, p, 'right') - 1
assert i >= 0 and i < len(primes) and primes[i] == p, (i, primes[i])
m[iz, i] = (int(m[iz, i]) + c) % 2
m[iz, len(primes) + iz] = 1
Print('Gaussian elemination...')
#Print(MatToStr(m)); Print()
one_col, one_rows = 0, 0
while True:
while True:
for i in range(one_rows, len(m)):
if m[i, one_col]:
break
else:
one_col += 1
if one_col >= len(primes):
break
continue
break
if one_col >= len(primes):
break
assert m[i, one_col]
assert np.all(m[i, :one_col] == 0)
for j in range(len(m)):
if i == j:
continue
if not m[j, one_col]:
continue
m[j][...] ^= m[i][...]
SwapRows(one_rows, i)
one_rows += 1
one_col += 1
assert np.all(m[one_rows:, :len(primes)] == 0)
zeros = m[one_rows:, len(primes):]
Print(f'Even combinations ({len(m) - one_rows}):')
Print(MatToStr(zeros))
return zeros
def ProcessResults(N, zs, la_zeros):
import math
Print('Computing final results...')
factors = []
for i in range(len(la_zeros)):
zero = la_zeros[i]
assert len(zero) == len(zs)
cz = []
for j in range(len(zero)):
if not zero[j]:
continue
z = zs[j]
z2 = z * z - N
cz.append((z, z2, FactorRef(z2)))
a = 1
for z, z2, fs in cz:
a = (a * z) % N
cnts = {}
for z, z2, fs in cz:
for p, c in fs.items():
cnts[p] = cnts.get(p, 0) + c
cnts = dict(sorted(cnts.items()))
b = 1
for p, c in cnts.items():
assert c % 2 == 0, (p, c, cnts)
b = (b * pow(p, c // 2, N)) % N
factor = math.gcd(a + b, N)
Print('a', str(a).rjust(len(str(N))), ' b', str(b).rjust(len(str(N))), ' factor', factor if factor != N else 'N')
if factor != 1 and factor != N:
factors.append(factor)
return factors
def SieveCollectResults(N, its):
import time, threading, queue, traceback, math
K = len(its)
qs = [queue.Queue() for i in range(K)]
last_dot, finish = False, False
def Get(it, ty, need, compul):
nonlocal last_dot, finish
try:
cnt = 0
for iz, z in enumerate(it):
if finish:
break
if iz < 4:
z2 = z * z - N
Print(('\n' if last_dot else '') + 'Sieve_' + ('C', 'B', 'A')[K - 1 - ty], ' iz', iz,
'z', z, 'z^2', z2, f'(2^{math.log2(max(1, z2)):>6.2f})', ', z^2 factors', FactorRef(z2))
last_dot = False
else:
Print(('.', 'b', 'a')[K - 1 - ty], end = '')
last_dot = True
qs[ty].put(z)
cnt += 1
if cnt >= need:
break
except:
Print(traceback.format_exc())
thr = []
for ty, (it, need, compul) in enumerate(its):
thr.append(threading.Thread(target = Get, args = (it, ty, need, compul), daemon = True))
thr[-1].start()
for ithr, t in enumerate(thr):
if its[ithr][2]:
t.join()
finish = True
if last_dot:
Print()
zs = [[] for i in range(K)]
for iq, q in enumerate(qs):
while not qs[iq].empty():
zs[iq].append(qs[iq].get())
return zs
def DixonFactor(N):
import time, math, numpy as np, sys
B = 1 << 10
primes = GenPrimes_SieveOfEratosthenes(B)
Print('Num primes', len(primes), 'last prime', primes[-1])
IsSquare(0)
it = SieveQuadratic(N, primes)
qprimes = next(it)
zs = SieveCollectResults(N, [
#(SieveSimple(N, primes), 3, False),
(SieveFactor(N, primes), 3, False),
(it, round(len(qprimes) * 1.06 + 0.5), True),
])[-1]
la_zeros = LinAlg(N, zs, qprimes)
fs = ProcessResults(N, zs, la_zeros)
if len(fs) > 0:
Print('Factored, factors', sorted(set(fs)))
else:
Print('Failed to factor! Try running program again...')
def IsPrime_Fermat(n, *, ntrials = 32):
import random
if n <= 16:
return n in (2, 3, 5, 7, 11, 13)
for i in range(ntrials):
if pow(random.randint(2, n - 2), n - 1, n) != 1:
return False
return True
def GenRandom(bits):
import random
return random.randrange(1 << (bits - 1), 1 << bits)
def RandPrime(bits):
while True:
n = GenRandom(bits) | 1
if IsPrime_Fermat(n):
return n
def Main():
import math
bits = 64
N = RandPrime(bits // 2) * RandPrime((bits + 1) // 2)
Print('N to factor', N, f'(2^{math.log2(N):>5.1f})')
DixonFactor(N)
if __name__ == '__main__':
Main()
Console output:
N to factor 10086068308526249063 (2^ 63.1)
Num primes 172 last prime 1021
Pre-computing squares filter...
Quadratic Sieve of B-smooth z^2...
Factor Sieve of B-smooth z^2...
QSieve num primes 78 (45.3%)
Block 0 (2^ 32.14)...
Sieve_C iz 0 z 3175858067 z^2 6153202727426 (2^ 42.48) , z^2 factors {2: 1, 29: 2, 67: 1, 191: 1, 487: 1, 587: 1}
Sieve_C iz 1 z 3175859246 z^2 13641877439453 (2^ 43.63) , z^2 factors {31: 1, 61: 1, 167: 1, 179: 1, 373: 1, 647: 1}
Sieve_C iz 2 z 3175863276 z^2 39239319203113 (2^ 45.16) , z^2 factors {31: 1, 109: 1, 163: 1, 277: 1, 311: 1, 827: 1}
Sieve_C iz 3 z 3175867115 z^2 63623612174162 (2^ 45.85) , z^2 factors {2: 1, 29: 1, 41: 1, 47: 1, 61: 1, 127: 1, 197: 1, 373: 1}
.........................................................................
Sieve_B iz 0 z 3175858067 z^2 6153202727426 (2^ 42.48) , z^2 factors {2: 1, 29: 2, 67: 1, 191: 1, 487: 1, 587: 1}
......
Linear algebra...
Factoring...
Gaussian elemination...
Even combinations (7):
01000000000000000000000000000000000000000000000000001100000000000000000000000000000
11010100000010000100100000010011100000000001001001001001011001000000110001010000000
11001011000101111100011111001011010011000111101000001001011000001111100101001110000
11010010010000110110101100110101000100001100010011100011101000100010011011001001000
00010110111010000010000010000111010001010010111001000011011011101110110001001100100
00000010111000110010100110001111010101001000011010110011101000110001101101100100010
10010001111111101100011110111110110100000110111011010001010001100000010100000100001
Computing final results...
a 9990591196683978238 b 9990591196683978238 factor 1
a 936902490212600845 b 3051457985176300292 factor 3960321451
a 1072293684177681642 b 8576178744296269655 factor 2546780213
a 1578121372922149955 b 1578121372922149955 factor 1
a 2036768191033218175 b 8049300117493030888 factor N
a 1489997751586754228 b 2231890938565281666 factor 3960321451
a 9673227070299809069 b 3412883990935144956 factor 3960321451
Factored, factors [2546780213, 3960321451]

Rabin-Miller Strong Pseudoprime Test Implementation won't work

Been trying to implement Rabin-Miller Strong Pseudoprime Test today.
Have used Wolfram Mathworld as reference, lines 3-5 sums up my code pretty much.
However, when I run the program, it says (sometimes) that primes (even low such as 5, 7, 11) are not primes. I've looked over the code for a very long while and cannot figure out what is wrong.
For help I've looked at this site aswell as many other sites but most use another definition (probably the same, but since I'm new to this kind of math, I can't see the same obvious connection).
My Code:
import random
def RabinMiller(n, k):
# obviously not prime
if n < 2 or n % 2 == 0:
return False
# special case
if n == 2:
return True
s = 0
r = n - 1
# factor n - 1 as 2^(r)*s
while r % 2 == 0:
s = s + 1
r = r // 2 # floor
# k = accuracy
for i in range(k):
a = random.randrange(1, n)
# a^(s) mod n = 1?
if pow(a, s, n) == 1:
return True
# a^(2^(j) * s) mod n = -1 mod n?
for j in range(r):
if pow(a, 2**j*s, n) == -1 % n:
return True
return False
print(RabinMiller(7, 5))
How does this differ from the definition given at Mathworld?
1. Comments on your code
A number of the points I'll make below were noted in other answers, but it seems useful to have them all together.
In the section
s = 0
r = n - 1
# factor n - 1 as 2^(r)*s
while r % 2 == 0:
s = s + 1
r = r // 2 # floor
you've got the roles of r and s swapped: you've actually factored n − 1 as 2sr. If you want to stick to the MathWorld notation, then you'll have to swap r and s in this section of the code:
# factor n - 1 as 2^(r)*s, where s is odd.
r, s = 0, n - 1
while s % 2 == 0:
r += 1
s //= 2
In the line
for i in range(k):
the variable i is unused: it's conventional to name such variables _.
You pick a random base between 1 and n − 1 inclusive:
a = random.randrange(1, n)
This is what it says in the MathWorld article, but that article is written from the mathematician's point of view. In fact it is useless to pick the base 1, since 1s = 1 (mod n) and you'll waste a trial. Similarly, it's useless to pick the base n − 1, since s is odd and so (n − 1)s = −1 (mod n). Mathematicians don't have to worry about wasted trials, but programmers do, so write instead:
a = random.randrange(2, n - 1)
(n needs to be at least 4 for this optimization to work, but we can easily arrange that by returning True at the top of the function when n = 3, just as you do for n = 2.)
As noted in other replies, you've misunderstood the MathWorld article. When it says that "n passes the test" it means that "n passes the test for the base a". The distinguishing fact about primes is that they pass the test for all bases. So when you find that as = 1 (mod n), what you should do is to go round the loop and pick the next base to test against.
# a^(s) = 1 (mod n)?
x = pow(a, s, n)
if x == 1:
continue
There's an opportunity for optimization here. The value x that we've just computed is a20 s (mod n). So we could test it immediately and save ourselves one loop iteration:
# a^(s) = ±1 (mod n)?
x = pow(a, s, n)
if x == 1 or x == n - 1:
continue
In the section where you calculate a2j s (mod n) each of these numbers is the square of the previous number (modulo n). It's wasteful to calculate each from scratch when you could just square the previous value. So you should write this loop as:
# a^(2^(j) * s) = -1 (mod n)?
for _ in range(r - 1):
x = pow(x, 2, n)
if x == n - 1:
break
else:
return False
It's a good idea to test for divisibility by small primes before trying Miller–Rabin. For example, in Rabin's 1977 paper he says:
In implementing the algorithm we incorporate some laborsaving steps. First we test for divisibility by any prime p < N, where, say N = 1000.
2. Revised code
Putting all this together:
from random import randrange
small_primes = [2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31] # etc.
def probably_prime(n, k):
"""Return True if n passes k rounds of the Miller-Rabin primality
test (and is probably prime). Return False if n is proved to be
composite.
"""
if n < 2: return False
for p in small_primes:
if n < p * p: return True
if n % p == 0: return False
r, s = 0, n - 1
while s % 2 == 0:
r += 1
s //= 2
for _ in range(k):
a = randrange(2, n - 1)
x = pow(a, s, n)
if x == 1 or x == n - 1:
continue
for _ in range(r - 1):
x = pow(x, 2, n)
if x == n - 1:
break
else:
return False
return True
In addition to what Omri Barel has said, there is also a problem with your for loop. You will return true if you find one a that passes the test. However, all a have to pass the test for n to be a probable prime.
I'm wondering about this piece of code:
# factor n - 1 as 2^(r)*s
while r % 2 == 0:
s = s + 1
r = r // 2 # floor
Let's take n = 7. So n - 1 = 6. We can express n - 1 as 2^1 * 3. In this case r = 1 and s = 3.
But the code above finds something else. It starts with r = 6, so r % 2 == 0. Initially, s = 0 so after one iteration we have s = 1 and r = 3. But now r % 2 != 0 and the loop terminates.
We end up with s = 1 and r = 3 which is clearly incorrect: 2^r * s = 8.
You should not update s in the loop. Instead, you should count how many times you can divide by 2 (this will be r) and the result after the divisions will be s. In the example of n = 7, n - 1 = 6, we can divide it once (so r = 1) and after the division we end up with 3 (so s = 3).
Here's my version:
# miller-rabin pseudoprimality checker
from random import randrange
def isStrongPseudoprime(n, a):
d, s = n-1, 0
while d % 2 == 0:
d, s = d/2, s+1
t = pow(a, d, n)
if t == 1:
return True
while s > 0:
if t == n - 1:
return True
t, s = pow(t, 2, n), s - 1
return False
def isPrime(n, k):
if n % 2 == 0:
return n == 2
for i in range(1, k):
a = randrange(2, n)
if not isStrongPseudoprime(n, a):
return False
return True
If you want to know more about programming with prime numbers, I modestly recommend this essay on my blog.
You should also have a look at Wikipedia, where known "random" sequences gives guaranteed answers up to a given prime.
if n < 1,373,653, it is enough to test a = 2 and 3;
if n < 9,080,191, it is enough to test a = 31 and 73;
if n < 4,759,123,141, it is enough to test a = 2, 7, and 61;
if n < 2,152,302,898,747, it is enough to test a = 2, 3, 5, 7, and 11;
if n < 3,474,749,660,383, it is enough to test a = 2, 3, 5, 7, 11, and 13;
if n < 341,550,071,728,321, it is enough to test a = 2, 3, 5, 7, 11, 13, and 17;

Categories