Given an array of numbers, return array of products of all other numbers (no division)

An explanation of polygenelubricants method is: The trick is to construct the arrays (in the case for 4 elements)

{              1,         a[0],    a[0]*a[1],    a[0]*a[1]*a[2],  }
{ a[1]*a[2]*a[3],    a[2]*a[3],         a[3],                 1,  }

Both of which can be done in O(n) by starting at the left and right edges respectively.

Then multiplying the two arrays element by element gives the required result

My code would look something like this:

int a[N] // This is the input
int products_below[N];
p=1;
for(int i=0;i<N;++i) {
  products_below[i]=p;
  p*=a[i];
}

int products_above[N];
p=1;
for(int i=N-1;i>=0;--i) {
  products_above[i]=p;
  p*=a[i];
}

int products[N]; // This is the result
for(int i=0;i<N;++i) {
  products[i]=products_below[i]*products_above[i];
}

If you need to be O(1) in space too you can do this (which is less clear IMHO)

int a[N] // This is the input
int products[N];

// Get the products below the current index
p=1;
for(int i=0;i<N;++i) {
  products[i]=p;
  p*=a[i];
}

// Get the products above the curent index
p=1;
for(int i=N-1;i>=0;--i) {
  products[i]*=p;
  p*=a[i];
}

Here is a small recursive function (in C++) to do the modofication in place. It requires O(n) extra space (on stack) though. Assuming the array is in a and N holds the array length, we have

int multiply(int *a, int fwdProduct, int indx) {
    int revProduct = 1;
    if (indx < N) {
       revProduct = multiply(a, fwdProduct*a[indx], indx+1);
       int cur = a[indx];
       a[indx] = fwdProduct * revProduct;
       revProduct *= cur;
    }
    return revProduct;
}

Here's my attempt to solve it in Java. Apologies for the non-standard formatting, but the code has a lot of duplication, and this is the best I can do to make it readable.

import java.util.Arrays;

public class Products {
    static int[] products(int... nums) {
        final int N = nums.length;
        int[] prods = new int[N];
        Arrays.fill(prods, 1);
        for (int
           i = 0, pi = 1    ,  j = N-1, pj = 1  ;
           (i < N)         && (j >= 0)          ;
           pi *= nums[i++]  ,  pj *= nums[j--]  )
        {
           prods[i] *= pi   ;  prods[j] *= pj   ;
        }
        return prods;
    }
    public static void main(String[] args) {
        System.out.println(
            Arrays.toString(products(1, 2, 3, 4, 5))
        ); // prints "[120, 60, 40, 30, 24]"
    }
}

The loop invariants are pi = nums[0] * nums[1] *.. nums[i-1] and pj = nums[N-1] * nums[N-2] *.. nums[j+1]. The i part on the left is the "prefix" logic, and the j part on the right is the "suffix" logic.

---

Recursive one-liner

Jasmeet gave a (beautiful!) recursive solution; I've turned it into this (hideous!) Java one-liner. It does in-place modification, with O(N) temporary space in the stack.

static int multiply(int[] nums, int p, int n) {
    return (n == nums.length) ? 1
      : nums[n] * (p = multiply(nums, nums[n] * (nums[n] = p), n + 1))
          + 0*(nums[n] *= p);
}

int[] arr = {1,2,3,4,5};
multiply(arr, 1, 0);
System.out.println(Arrays.toString(arr));
// prints "[120, 60, 40, 30, 24]"