Question

link

Given a binary tree, find the maximum path sum.

The path may start and end at any node in the tree.

For example:
Given the below binary tree,

       1
      / \
     2   3

Return 6.

Stats

Frequency 2
Difficulty 4
Adjusted Difficulty 4
Time to use --------

Ratings/Color = 1(white) 2(lime) 3(yellow) 4/5(red)

Analysis

This is a difficult DFS question.

The basic idea is:

When we are in a root node, how we decided to include current node in the maximum path sum? We need to calculate the depth in both subtree. In the meantime, we need to set a variable “max” to store the max path sum we found during the recursion.

Pay attention that when a subtree sum to negative value, we discard this subtree.

Otherwise, max path sum including current node is leftSum + cur.val + rightSum

Max depth of current node is (maximum value of leftSum and rightSum) + cur.val

Solution

There are 2 points that may incur problems during coding, they are:

  1. The recursive solution is calculating the ‘max height’ while checking ‘its own max path sum’. It’s a bit hard to explain. Basic idea is doing 2 things at same time, while traversing (are one result is returned, another is stored in a public variable, at least for code 1 below).

  2. Note that ‘max height’ of a child branch can be nagative. In this case, WE MUST SET IT TO 0.

The first point is a concept seen in previous questions, but I forgot which. The second point is important, because I missed this and got failed a few times without any clue.

Code

First, recursive solution

int max = Integer.MIN_VALUE;

public int maxPathSum(TreeNode root) {
    maxHeight(root);
    return max;
}

private int maxHeight(TreeNode node) {
    if (node == null) return 0;
    int l = Math.max(0, maxHeight(node.left));
    int r = Math.max(0, maxHeight(node.right));
    max = Math.max(max, l + r + node.val);
    return node.val + Math.max(l, r);
}

Second, previous code refactored.

I removed the global variable, and instead pass an array as reference in the recursion. In this way the ‘shared’ array achieve same functionality as a global variable.

public int maxPathSum(TreeNode root) {
    int[] shared = {Integer.MIN_VALUE};
    maxHeight(root, shared);
    return shared[0];
}

private int maxHeight(TreeNode node, int[] shared) {
    if (node == null) return 0;
    int l = Math.max(0, maxHeight(node.left, shared));
    int r = Math.max(0, maxHeight(node.right, shared));
    shared[0] = Math.max(shared[0], l + r + node.val);
    return node.val + Math.max(l, r);
}

Updated on June 10th, it’s a terrible practise to use global variable. The new solution suggested by Mr. Huang uses a new ResultType Class to solve the problem.

private class ResultType {
    int singlePath, maxPath;
    ResultType(int singlePath, int maxPath) {
        this.singlePath = singlePath;
        this.maxPath = maxPath;
    }
}

public int maxPathSum(TreeNode root) {
    ResultType result = helper(root);
    return result.maxPath;
}

private ResultType helper(TreeNode node) {
    // null case
    if (node == null) {
        return new ResultType(0, Integer.MIN_VALUE);
    }
    // divide
    ResultType ll = helper(node.left);
    ResultType rr = helper(node.right);
    // conquer
    int curSinglePath = Math.max(0, node.val +
            Math.max(ll.singlePath, rr.singlePath));
    int childMaxPath = Math.max(ll.maxPath, rr.maxPath);
    int curMaxPath = Math.max(childMaxPath, node.val +
            ll.singlePath + rr.singlePath);
    // done
    return new ResultType(curSinglePath, curMaxPath);
}