Segment Tree

Difficulty Level Hard
Frequently asked in Amazon CodeNation Google Microsoft Uber
Advanced Data Structure Segment-Tree TreeViews 1290

If we have performing addition on a given range of array whose element values updated any time. Then in that type of problem, we handle using a segment tree structure. Given an array a[] with n elements and you have to answer multiple queries, each of the queries is one of the two types :
1. 1 i x: Set a[i] = x
2. 2 l r: Find and print the sum of elements between l and r both inclusive

Example

Input :
a[] = {2, 5, 9, 8, 11, 3}
q = 3 (Number of queries)
2 3 5
1 2 8
2 0 5

Output :
22
37

The naive approach to solve the above problem is to run a loop from l to r to and find the sum of the range and for updating, we directly set the value of a[i] as x.

Segment Tree Approach and Representation

  1. Leaf nodes of the segment tree are the elements of the given array.
  2. Each internal node stores the sum of its children.

A segment tree is represented as an array in memory, it is a full binary tree as every node has either 2 or 0 children and all levels are filled except possibly the last level. When represented as array there are some spaces that are never been used, hence the size of a segment tree is (2*x – 1), where x is the smallest power of 2 greater than or equal to n.

Segment Tree for the above example is shown in the image.

Segment Tree

Construction

  1. We first allocate a memory equals to the size of the segment tree, that is, create an array of size equals the size of the segment tree.
  2. Then, for every node, the value of this node is equal to the sum of its left and right child.
  3. So, we write a recursive code to find the value of each node,
    value[i] = value[2 * i + 1] + value[2 * i + 2] // Left child of i is (2*i + 1) and Right child is (2*i + 2)
  4. The base case of the recurrence is when it is a leaf node, for the leaf node the value of node is equals to the value present in the array because both of its child are null(or absent).

Updating an element(Type-1 Query)

Let the ith index has to be updated to x and its original value was y, that is we have to increase its value by (x – y), and also all the sums containing this index in their range will also have to be incremented by (x – y), so we write a recursive code to do that,

  1. Start from the root.
  2. If the current node contains i within its range, then increment the value by (x- y) and recur for it’s left and right child.
  3. If the current node does not contain i within its range, then we do not make any changes to it

Sum range Query (Type-2 Query)

  1. Start from the root node, if the node range is between l and r return the value of this node.
  2. If the node’s range is completely outside the range l and r, return 0.
  3. In all the other cases return the sum of answers of the query(l, r) for it’s left child and it’s a right child.

JAVA Code for range sum using Segment Tree

public class SegmentTree {
    // Function to find the sum of given range in the segment tree
    // tree[] --> Segment Tree
    // s --> Starting index of segment tree
    // e --> Ending index of segment tree
    // i --> Current index of segment tree
    // l --> Lower index of range
    // r --> Higher index of range
    private static int rangeSum(int tree[], int s, int e, int l, int r, int i) {
        // If the current node range is within the range l and r, return its value
        if (l <= s && r >= e)
            return tree[i];

        // If current node's range is completely outside the range l and r, return 0
        if (e < l || s > r)
            return 0;

        // For all other cases return sum of answers to query for left and right child
        // Left child index = 2 * i + 1
        // Right child index = 2 * i + 2
        int mid = (s + e) / 2;
        return rangeSum(tree, s, mid, l, r, 2 * i + 1) +
                rangeSum(tree, mid + 1, e, l, r, 2 * i + 2);
    }

    // Function to update the segment tree for a given index
    // s --> Starting index of segment tree
    // e --> Ending index of segment tree
    // index --> Index to be changed in the original array
    // diff --> This is to be added in the nodes that contains index in their range
    // i --> Current index of Segment tree
    private static void updateValue(int tree[], int s, int e, int index, int diff, int i) {
        // If the current node does not contain index in its range, make no changes
        if (index < s || index > e)
            return;

        // Current node contains the index in its range, update the current nodes and its children
        // Left child index = 2 * i + 1
        // Right child index = 2 * i + 2
        tree[i] = tree[i] + diff;
        if (s != e) {
            int mid = (s + e) / 2;
            updateValue(tree, s, mid, index, diff, 2 * i + 1);
            updateValue(tree, mid + 1, e, index, diff, 2 * i + 2);
        }
    }

    // A function to create the segment tree recursively between s and e
    // i --> Index of current node in the segment tree
    private static int constructSegmentTree(int tree[], int a[], int s, int e, int i) {
        // Leaf node case
        if (s == e) {
            tree[i] = a[s];
            return a[s];
        }

        // For all other nodes its value is sum of left and right child's value
        // Left child index = 2 * i + 1
        // Right child index = 2 * i + 2
        int mid = (s + e) / 2;
        tree[i] = constructSegmentTree(tree, a, s, mid, i * 2 + 1) +
                constructSegmentTree(tree, a, mid + 1, e, i * 2 + 2);
        // Return the value of current node
        return tree[i];
    }

    // Driver function for segment tree approach
    public static void main(String args[]) {
        int a[] = {2, 5, 9, 8, 11, 3};
        int n = a.length;

        // Calculate the size of the segment tree
        int x = (int) (Math.ceil(Math.log(n) / Math.log(2)));
        int size = 2 * (int) Math.pow(2, x) - 1;

        // Allocate memory for segment tree
        int tree[] = new int[size];

        // Construct the segment tree
        constructSegmentTree(tree, a, 0, n - 1, 0);

        // Queries
        int q = 3;
        int type[] = {2, 1, 2};     // Stores the type of query to process
        int l[] = {3, 2, 0};        // Stores the index of element to be updated for type-1 query and lower range for type-2 query
        int r[] = {5, 8, 5};        // Stores the new value of element for type-1 query and higher range for type-2 query
        for (int j = 0; j < q; j++) {
            if (type[j] == 1) {
                // Type-1 query (Update the value of specified index)
                int index = l[j];
                int value = r[j];
                int diff = value - a[index];            // This diff is to be added to all the range that contains the index

                // Update the value in array a
                a[index] = value;

                // Update segment tree
                updateValue(tree, 0, n - 1, index, diff, 0);
            } else {
                // Type-2 query (Find the Sum of given range)
                int sum = rangeSum(tree, 0, n - 1, l[j], r[j], 0);
                System.out.println(sum);
            }
        }
    }
}

C++ Code for range sum using Segment Tree

#include <bits/stdc++.h> 
using namespace std; 

// Function to find the sum of given range in the segment tree
// tree[] --> Segment Tree
// s --> Starting index of segment tree
// e --> Ending index of segment tree
// i --> Current index of segment tree
// l --> Lower index of range
// r --> Higher index of range
int rangeSum(int *tree, int s, int e, int l, int r, int i) {
  // If the current node range is within the range l and r, return its value
    if (l <= s && r >= e)
        return tree[i];

    // If current node's range is completely outside the range l and r, return 0
    if (e < l || s > r)
        return 0;
    
  // For all other cases return sum of answers to query for left and right child
    // Left child index = 2 * i + 1
    // Right child index = 2 * i + 2
    int mid = (s + e) / 2;
    return rangeSum(tree, s, mid, l, r, 2 * i + 1) +
                rangeSum(tree, mid + 1, e, l, r, 2 * i + 2);
}

// Function to update the segment tree for a given index
// s --> Starting index of segment tree
// e --> Ending index of segment tree
// index --> Index to be changed in the original array
// diff --> This is to be added in the nodes that contains index in their range
// i --> Current index of Segment tree
void updateValue(int *tree, int s, int e, int index, int diff, int i) {
    // If the current node does not contain index in its range, make no changes
    if (index < s || index > e)
        return;

    // Current node contains the index in its range, update the current nodes and its children
    // Left child index = 2 * i + 1
    // Right child index = 2 * i + 2
    tree[i] = tree[i] + diff;
    if (s != e) {
        int mid = (s + e) / 2;
        updateValue(tree, s, mid, index, diff, 2 * i + 1);
        updateValue(tree, mid + 1, e, index, diff, 2 * i + 2);
    }
}

// A function to create the segment tree recursively between s and e
// i --> Index of current node in the segment tree
int constructSegmentTree(int *tree, int *a, int s, int e, int i) {
    // Leaf node case
    if (s == e) {
        tree[i] = a[s];
        return a[s];
    }

    // For all other nodes its value is sum of left and right child's value
    // Left child index = 2 * i + 1
    // Right child index = 2 * i + 2
    int mid = (s + e) / 2;
    tree[i] = constructSegmentTree(tree, a, s, mid, i * 2 + 1) +
                constructSegmentTree(tree, a, mid + 1, e, i * 2 + 2);
    // Return the value of current node
    return tree[i];
}

// Driver function for segment tree approach  
int main() {
  int a[] = {2, 5, 9, 8, 11, 3};
  int n = sizeof(a)/sizeof(a[0]);
  
  // Calculate the size of the segment tree
  int x = (int)(ceil(log2(n)));  
    int size = 2*(int)pow(2, x) - 1;
  
  // Allocate memory for segment tree
  int tree[size];
  
  // Construct the segment tree
  constructSegmentTree(tree, a, 0, n - 1, 0);
  
  // Queries
    int q = 3;
    int type[] = {2, 1, 2};     // Stores the type of query to process
    int l[] = {3, 2, 0};        // Stores the index of element to be updated for type-1 query and lower range for type-2 query
    int r[] = {5, 8, 5};        // Stores the new value of element for type-1 query and higher range for type-2 query
    for (int j = 0; j < q; j++) {
        if (type[j] == 1) {
            // Type-1 query (Update the value of specified index)
            int index = l[j];
            int value = r[j];
            int diff = value - a[index];            // This diff is to be added to all the range that contains the index

            // Update the value in array a
            a[index] = value;

            // Update segment tree
            updateValue(tree, 0, n - 1, index, diff, 0);
        } else {
            // Type-2 query (Find the Sum of given range)
            int sum = rangeSum(tree, 0, n - 1, l[j], r[j], 0);
            cout<<sum<<endl;
        }
    }
  return 0;
}
22
37

Complexity Analysis

The time complexity for the type-1 query is O(1) and for the type-2 query, it is O(n), this can be optimized by the use of a segment tree.

References

Translate ยป