3-way Merge Sort

Prerequisite – Merge Sort

Merge sort involves recursively splitting the array into 2 parts, sorting and finally merging them. A variant of merge sort is called 3-way merge sort where instead of splitting the array into 2 parts we split it into 3 parts.
Merge sort recursively breaks down the arrays to subarrays of size half. Similarly, 3-way Merge sort breaks down the arrays to subarrays of size one third.

Examples:

Input  : 45, -2, -45, 78, 30, -42, 10, 19 , 73, 93 
Output : -45 -42 -2 10 19 30 45 73 78 93 

Input  : 23, -19
Output : -19  23

// Java program to perform 3 way Merge Sort
public class MergeSort3Way
{
    // Function  for 3-way merge sort process
    public static void mergeSort3Way(Integer[] gArray)
    {
        // if array of size is zero returns null
        if (gArray == null)
            return;

        // creating duplicate of given array
        Integer[] fArray = new Integer[gArray.length];

        // copying alements of given array into
        // duplicate array
        for (int i = 0; i < fArray.length; i++)
            fArray[i] = gArray[i];

        // sort function
        mergeSort3WayRec(fArray, 0, gArray.length, gArray);

        // copy back elements of duplicate array
        // to given array
        for (int i = 0; i < fArray.length; i++)
            gArray[i] = fArray[i];
    }

    /* Performing the merge sort algorithm on the
       given array of values in the rangeof indices
       [low, high).  low is minimum index, high is
       maximum index (exclusive) */
    public static void mergeSort3WayRec(Integer[] gArray,
                  int low, int high, Integer[] destArray)
    {
        // If array size is 1 then do nothing
        if (high - low < 2)
            return;

        // Splitting array into 3 parts
        int mid1 = low + ((high - low) / 3);
        int mid2 = low + 2 * ((high - low) / 3) + 1;

        // Sorting 3 arrays recursively
        mergeSort3WayRec(destArray, low, mid1, gArray);
        mergeSort3WayRec(destArray, mid1, mid2, gArray);
        mergeSort3WayRec(destArray, mid2, high, gArray);

        // Merging the sorted arrays
        merge(destArray, low, mid1, mid2, high, gArray);
    }

    /* Merge the sorted ranges [low, mid1), [mid1,
       mid2) and [mid2, high) mid1 is first midpoint
       index in overall range to merge mid2 is second
       midpoint index in overall range to merge*/
    public static void merge(Integer[] gArray, int low,
                           int mid1, int mid2, int high,
                                   Integer[] destArray)
    {
        int i = low, j = mid1, k = mid2, l = low;

        // choose smaller of the smallest in the three ranges
        while ((i < mid1) && (j < mid2) && (k < high))
        {
            if (gArray[i].compareTo(gArray[j]) < 0)
            {
                if (gArray[i].compareTo(gArray[k]) < 0)
                    destArray[l++] = gArray[i++];

                else
                    destArray[l++] = gArray[k++];
            }
            else
            {
                if (gArray[j].compareTo(gArray[k]) < 0)
                    destArray[l++] = gArray[j++];
                else
                    destArray[l++] = gArray[k++];
            }
        }

        // case where first and second ranges have
        // remaining values
        while ((i < mid1) && (j < mid2))
        {
            if (gArray[i].compareTo(gArray[j]) < 0)
                destArray[l++] = gArray[i++];
            else
                destArray[l++] = gArray[j++];
        }

        // case where second and third ranges have
        // remaining values
        while ((j < mid2) && (k < high))
        {
            if (gArray[j].compareTo(gArray[k]) < 0)
                destArray[l++] = gArray[j++];

            else
                destArray[l++] = gArray[k++];
        }

        // case where first and third ranges have
        // remaining values
        while ((i < mid1) && (k < high))
        {
            if (gArray[i].compareTo(gArray[k]) < 0)
                destArray[l++] = gArray[i++];
            else
                destArray[l++] = gArray[k++];
        }

        // copy remaining values from the first range
        while (i < mid1)
            destArray[l++] = gArray[i++];

        // copy remaining values from the second range
        while (j < mid2)
            destArray[l++] = gArray[j++];

        // copy remaining values from the third range
        while (k < high)
            destArray[l++] = gArray[k++];
    }

    // Driver function
    public static void main(String args[])
    {
        // test case of values
        Integer[] data = new Integer[] {45, -2, -45, 78,
                               30, -42, 10, 19, 73, 93};
        mergeSort3Way(data);

        System.out.println("After 3 way merge sort: ");
        for (int i = 0; i < data.length; i++)
            System.out.print(data[i] + " ");
    }
}

Output:

After 3 way merge sort: 
-45 -42 -2 10 19 30 45 73 78 93 

Here, we first copy the contents of data array to another array called fArray. Then, sort the array by finding midpoints that divide the array into 3 parts and called sort function on each array respectively. The base case of recursion is when size of array is 1 and it returns from the function. Then merging of arrays starts and finally the sorted array will be in fArray which is copied back to gArray.

Time Complexity: In case of 2-way Merge sort we get the equation: T(n) = 2T(n/2) + O(n)
Similarly, in case of 3-way Merge sort we get the equation: T(n) = 3T(n/3) + O(n)
By solving it using Master method, we get its complexity as O(n log 3n).. Although time complexity looks less compared to 2 way merge sort, the time taken actually may become higher because number of comparisons in merge function go higher. Please refer Why is Binary Search preferred over Ternary Search? for details.

Similar article :
3 way Quick Sort

This article is contributed by Pavan Gopal Rayapati. If you like GeeksforGeeks and would like to contribute, you can also write an article using contribute.geeksforgeeks.org or mail your article to contribute@geeksforgeeks.org. See your article appearing on the GeeksforGeeks main page and help other Geeks.

Please write comments if you find anything incorrect, or you want to share more information about the topic discussed above.

GATE CS Corner    Company Wise Coding Practice

Recommended Posts:







Writing code in comment? Please use ide.geeksforgeeks.org, generate link and share the link here.