Fetching contributors…
Cannot retrieve contributors at this time
132 lines (86 sloc) 12.6 KB

# Amicable Numbers

This is derived from my blog post made in answer to the Week 20 of the Perl Weekly Challenge organized by Mohammad S. Anwar as well as answers made by others to the same challenge.

Amicable numbers are two different numbers so related that the sum of the proper divisors of each is equal to the other number. (A proper divisor of a number is a positive factor of that number other than the number itself. For example, the proper divisors of 6 are 1, 2, and 3.)

## My Solution

We'll use the sum-divisors subroutine to find the divisors of a given number and return their sum. This subroutine uses trial division, i.e. tries division by all integers below a certain limit (here, half of the input number), and then sums up all those that evenly divide the input number. Then, we just loop over a lazy infinite list of integers from 2 onward and call sum_divisors subroutine. If the sum of divisors is larger than the integer being examined (if it is smaller, then it is a pair of numbers that we have already checked), then we check the sum of divisors of the sum of divisors. If it is equal to the current integer, then we've found two amicable numbers and can print them and exit the loop.

use v6;

sub sum-divisors (Int $num) { my @divisors = grep {$num %% $_ }, 2..($num / 2).Int;
return [+] 1, | @divisors;
}

for 2..Inf -> $i { my$sum_div = sum-divisors $i; if$sum_div > $i and$i == sum-divisors $sum_div { say "$i and $sum_div are amicable numbers"; last; } } The sum-divisors subroutine is not optimal and its performance could be greatly enhanced, but we don't need to do that since we're looking only for the first pair of amicable numbers, which are rather small. Also, I did not have much time on that week because I was quite busy with the Perl Conference in Riga. This program prints almost instantly the first pair of amicable numbers: $ perl6 amicable_nrs.p6
220 and 284 are amicable numbers


Note that the Wikipedia article shows some rules to quickly find some amicable numbers (Thābit ibn Qurra's theorem and Euler's rule), but I did not want to use any of these because neither Thābit ibn Qurra's theorem, nor Euler's method will produce all amicable numbers, so that it is not guaranteed that we would find the first pair with such methods (although it so happens that both methods do produce the first pair).

## Alternative Solutions

Arne Sommer used essentially the same method, except that his proper-divisors subroutine returns a list of proper divisors which are then added in the calling code.

Ruben Westerberg also uses essentially the same method, but his proper subroutine is particularly concise:

sub proper(\n) {
sum (1..n-1).grep({ n%%$_}); } Francis Whittle used multi MAIN subroutines to use either Euler's rule or the naive trial division algorithm described in my solution above. One interesting thing in Francis's naive implementation is that is proper-divisor-sum is using the is cached trait to avoid recomputing the proper divisors of a number if it has already been calculated. Kevin Colyer and Randy Lauen also cached the sums of proper divisors, but did it manually by storing them into a hash. Noud also cached the results returned by prop_div subroutine, but did that in an original way by adding a wrapper to the subroutine. Adam Russell used the same general algorithm, but interestingly used promises to parallelize part of the work over several concurrent threads and presumably speed up the search. Athanasius used the divisor_sum subroutine of the CPAN Math::Prime::Util Perl 5 module and further created an alias on it presumably to make its use easier: use Math::Prime::Util:from<Perl5> <divisor_sum>; my Sub$divisor-sum := &Math::Prime::Util::divisor_sum;     # Alias

This is really cool, as it is a good example showing how a fairly large part of the Perl 5 ecosystem can be used in Perl 6. Note, however that the divisor_sum subroutine returns the sum of all divisors of a number (including itself), so that he had to subtract the input number from the sums of its divisors in order to obtain the sum of proper divisors.

Feng Chang made a very nice and concise script using infinite lists:

my @a = (0..∞).map:  { sod($_) }; my @b = (0..∞).grep: { @a[$_] > $_ and @a[@a[$_]] == $_ }; say(@b[0], ' ', @a[@b[0]]); # sum of proper divisors sub sod(UInt$n) {
[+] (1 .. $n/2).grep: {$n %% \$_ }
}

Jaldhar H. Vyas used Thābit ibn Qurra's theorem to generate the first pair of amicable numbers. As mentioned above, I believe this is a bit dangerous since this method cannot generate an exhaustive list of such pairs, so we are not really guaranteed to find the first pair. As it turns out, though, this method does yield the first pair, so that Jaldhar's result is nonetheless correct.

Joelle Maslak's factorsum subroutine is probably more efficient than my sum-divisors subroutine (and is certainly faster for large input integers) because it iterates up to the square root of the input number (instead of going up to half the input number in my subroutine) and then computes the other larger divisors (for example, when looking for the divisors of 12, my subroutine iterates until 6, whereas Joelle's subroutine only needs to iterate until 3, obtains 2 and 3, and then finds 6 by dividing 12 by 2 and 4 by dividing 12 by 3; this probably doesn't improve very much performance for a small number such as 12, but it undoubtedly does for larger numbers). In addition, she caches the results with the is cached trait. In addition, her code is also a pretty beautiful clockwork of functional programming with two nice data pipelines.

Roger Bell West's divisors_unself subroutine also iterates only up to the square root of the input number, using essentially the same method as Joelle to generate the missing divisors above the square root. As noted just above, this is likely to be significantly faster for large input integers.

## Enter Damian Conway

Also known as TheDamian, Damian Conway wrote (as usual) the most definitive analysis on the subject in his With friends as these... blog post. On top of that, it is really well-written.

I will not try to summarize Damian's master piece, you should really follow the link and read it, but I will only highlight a few points.

Given a function that returns the divisors of a number, Damian says it is easy enough to iterate through each integer from 1 to infinity, find its divisors, then check to see if the sum-of-divisors of that number is identical to the original number (pretty much what I did in the main code on my script at the beginning of this post). Then, the article goes on to study how to do it if we want to find more than one amicable pair. Of course, in such a case, the performance issues that I originally brushed aside because we only needed the first pair now need to be re-examined.

The problem is that, since Perl 6 does not have a built-in function that returns the proper divisors of a number, we obviously need to write one. Damian comes up with a handful of such subroutines using the naive trial division algorithm, for example:

multi divisors (\N) { (1..N).grep(N %% *) }

The problem with it is that it is very inefficient when the input numbers become large. The first thing that Damian does is more or less the same improvement that Joelle and Roger used: iterate up to the square root of the input number and compute the larger missing divisors as the division of the input integer by each of the smaller divisors. For example:

multi divisors (\N) {
my \small-divisors = (1..sqrt N).grep(N %% *);
my \big-divisors   = N «div« small-divisors;
return unique flat small-divisors, big-divisors;
}

This is vastly more efficient. The performance of the divisors subroutine up to divisors(10**9) is entirely acceptable at under 0.1 seconds, but starts to fall off rapidly after that point. What if we want to go further?

Cryptographers have designed a number of more efficient algorithms to find the factors of an integer. One of them is Pollard’s 𝜌 algorithm, which is implemented in the prime-factors function of a Perl 6 module named Prime::Factor by Stephen Schulze. This function finds all the prime factors of a large input number very quickly. But the prime factors of a number isn't the same thing as its divisors. But we can use prime factors to compute the divisors by combining the factors in all possible ways. Except that if we use the naive approach of finding all combinations of prime factors, the program becomes catastrophically slow because of combinatorial explosion. Damian then shows how storing the prime factors in a Bag data structure makes it possible to test every possible combination of the prime factors exponents (the values of the bag). And the new version based on this idea scales incredibly better than any previous implementation: the program can find the divisors of a number with 100 digits in less than half a second.

Damian makes two additional points. One is that the new version of the divisors is a lot faster for very large numbers, but that, for smaller numbers (below about 10,000), iterating trial division to the square root of the input integer is faster. He then shows how easy it is to make two multi versions of the divisors subroutine, one for small numbers and one for big ones. Finally, Damian demonstrates how adding a single word, the hyper prefix, to his main for loop enables concurrency and makes the overall program about twice as fast.

I hope that I have opened your appetite. Once again, you should really follow the link to Damian's post and read it in detail. It's really worth it.