Hacker News Reader

Comment by ploggingdev | Hacker News Reader

In https://github.com/sangupta/ps/blob/master/solutions/2016/fa... the solution for fastest integer sorting is wrong.

    / build the boolean result array
    boolean[] result = new boolean[array.length];
    for(int index = 0; index < array.length; index++) {
      int num = array[index];
      result[num] = true;
    }

The result array is the same size as the original array.

But when you do:

    int num = array[index];
    result[num] = true;

The index being referenced in the result array is dependent on the value of num. So if num is greater than or equal to the length of the array, it throws an ArrayOutOfBoundsException. And what happens when you try sorting an array with elements having the same value?

Eg- try sorting {10,5,9}

Unrelated: HN needs to allow highlighting code inline.

View on HN

In the prompt, he says we are looking at a "bounded" integer set, so presumably we know the largest possible number before we start

My bad - that was a typo when writing. Updating it.

Yep, should be

boolean[] result = new boolean[Math.max(array)];

That won't be efficient memorywise since the size of the result array is dependant on the largest value in the array of values.

Consider sorting an array of 2 large integers {65000,30000}. To sort this array, the result array will be initialized with 65000 elements. Very inefficient.

Yeah, well, now you get an exception. Not much better...

You should look into bucketsort. That allows you to sort in finite universes in Theta(n+k) without allocating huge arrays for large integers.

I will read more about bucketsort and update the article as necessary.

I have a simple Python implementation of Countingsort, Bucketsort and Radixsort, if it helps: https://github.com/MauriceGit/Advanced_Algorithms

Thanks a bunch for the link.

Wouldn't using a sparse bit-array be equivalent to using Bucketsort in space-complexity, but more efficient in time-complexity, or am I missing something here.

Using a sparse bit-array will reduce the memory space of the problem.

Yes, but do you have an implementation of a sparse array with constant time lookups?

I can work up one. Choosing the right bucket will be O(1) as the index is integer. We can store the buckets for lookup in a hash-table or an array-of-arrays, which again will be O(1). Reaching to the actual index should again be O(1) - unless am messing something here.

Reply on news.ycombinator.com