Problem:

Merge k sorted linked lists and return it as one sorted list. Analyze and describe its complexity.

Method 1:

MergeSort. This is a very typical implementation for MergeSort by using array.

Code for method 1:

/**
 * Definition for singly-linked list.
 * public class ListNode {
 *     int val;
 *     ListNode next;
 *     ListNode(int x) { val = x; }
 * }
 */
public class Solution {
    public ListNode mergeKLists(ListNode[] lists) {
        if (lists == null || lists.length == 0) return null;
        return helper(lists, 0, lists.length-1);
    }
    
    public ListNode helper (ListNode[] lists, int l, int h){
        if (l < h){
             int mid = (l + h)/2;
             return merge(helper(lists, l, mid), helper(lists, mid+1, h));
        }
        return lists[l];
    }
    
    public ListNode merge(ListNode l1, ListNode l2){
        ListNode dummy = new ListNode(0);
        ListNode temp = dummy;
        while (l1 != null && l2 != null){
            if (l1.val < l2.val){
                temp.next = l1;
                temp = temp.next;
                l1 = l1.next;
            }else{
                temp.next = l2;
                temp = temp.next;
                l2 = l2.next;
            }
        }
        if (l1 != null){
            temp.next = l1;
        }
        if (l2 != null){
            temp.next = l2;
        }
        return dummy.next;
    }
}

Analysis for time complexity:

Suppose there are k lists, and there are n elements inside each list. The time complexity is O(nklogk).

Method 2:

Use heap. Suppose we have k lists, and for each list, there are at most n elements.

Maintain a list of k elements. And always remove the first element from the list. After removing an element, add the element after it to the heap.

Code for method 2:


/**
 * Definition for singly-linked list.
 * public class ListNode {
 *     int val;
 *     ListNode next;
 *     ListNode(int x) { val = x; }
 * }
 */
public class Solution {
    public ListNode mergeKLists(ListNode[] lists) {
        PriorityQueue<ListNode> heap = new PriorityQueue<ListNode>(10, new Comparator<ListNode>(){
            public int compare (ListNode l1, ListNode l2){
                return l1.val - l2.val;
            }
        });
        
        for (int i = 0; i < lists.length; i++){
            ListNode temp = lists[i];
            if (temp != null){
                heap.offer(temp);
            }
        }
        ListNode dummy = new ListNode(0);
        ListNode cur = dummy;
        
        while (heap.size() != 0){
            ListNode n = heap.poll();
            cur.next = n;
            cur = cur.next;
            if (n.next != null){
                heap.offer(n.next);
            }
        }
        return dummy.next;
    }
}

Analysis for time complexity:

We have to insert at most k*n nodes, and for each node to be in order in the heap, we need log(k) time (“Bottom Up Insertion”). So total time complexity is also knlog(k).