Skip to content
Related Articles

Related Articles

Proto Van Emde Boas Tree | Set 2 | Construction

View Discussion
Improve Article
Save Article
Like Article
  • Difficulty Level : Medium
  • Last Updated : 05 Oct, 2021

Van Emde Boas Tree supports search, minimum, maximum, successor, predecessor, insert and delete operations in O(lglgN) time which is faster than any of related data structures like priority queue, binary search tree, etc. Proto Van Emde Boas tree is similar prototype type data structure but it fails to achieve the complexity of O(lglgN), We will learn Proto Van Emde Boas tree first to get a basic understanding of working of Van Emde Boas tree. Here N is the size of the universe over which tree is defined.

Note: Proto Van Emde Boas Data Structure’s keys must be defined over a range of 0 to n(n is positive integer of the form 22k) and it works when duplicate keys are not allowed.


  1. Proto-VEB is an abbreviation of Proto-Van Emde Boas tree.
  2. Proto-VEB(\sqrt{u}     ) is an abbreviation for Proto-VEB containing u number of keys.

Basic Understanding of the structure of Proto-VEB Tree: 
Proto Van Emde boas tree are recursively defined data structure and it shrinks to sqrt size as we go up in the tree. Please refer this article to understand the basics of it.

In Proto-VEB we use a bit array to represents whether a key is present or not, we put 1 if it is present and 0 else.
Here, the summary of a particular cluster contains whether there is any key present in a cluster if at least one key present then the summary is 1 or 0 elsewhere. Clusters are segments of a bit array. A summary is also a bit array. See the image below:

van emde boas tree basic structure

Construction of Proto VEB Tree:
Below image represents the basic Proto-VEB structure:

Van Emde Boas Tree

The recursively defined structure has two main part:

  1. summary: it is a pointer to Proto-VEB structure which has \sqrt{u}     size.
  2. cluster: it is an array of pointers to Proto-VEB structures with \sqrt{u}     size.

First of all, we have to understand some function and keywords: 

  • universe size (u) : Number of keys in Proto-VEB structure.
  • High(x): From the first image, we can see that if we want to reach at the cluster of the key then we can divide it with \sqrt{u}     .
    For example, We want to know the cluster of the key 12 then can divide it with \sqrt{16}     which is 3, so key 12 is in 3rd cluster.
High(x) = floor( x / \sqrt{u})
  • low(x): From the first image, we can see that if we want the position of the key in the cluster we can apply modulus operation x % \sqrt{u}     .
    For example, If you want to find a position of 7 in the cluster you can apply 7 % \sqrt{16}     = 3 which is a position of 7 in 2nd cluster.
low(x) = x % \sqrt{u})

Recursive Procedure of construction: 

  1. Base case : If universe size is 2 then it is a base size so there will be no more summary array, which means it is null and we will store only bit array for 2 keys.
  2. We will recursively assign summary as \sqrt{u}     sized Proto-VEB tree and \sqrt{u}     -sized Proto-VEB to all \sqrt{u}     clusters. 

See u=4 Proto-VEB structure in the image below:  


Here is the code representing the Algorithm:


#include <bits/stdc++.h>
using namespace std;
class Proto_Van_Emde_Boas {
    // Total number of keys
    int universe_size;
    // Summary
    Proto_Van_Emde_Boas* summary;
    // Clusters array of Proto-VEB pointers
    vector<Proto_Van_Emde_Boas*> clusters;
    int root(int u)
        return (int)sqrt(u);
    // Function to return cluster numbers
    // in which key is present
    int high(int x)
        return x / root(universe_size);
    // Function to return the position
    // of x in cluster
    int low(int x)
        return x % root(universe_size);
    // Function to return index form
    // cluster number and position
    int generate_index(int cluster, int position)
        return cluster * root(universe_size) + position;
    // Constructor
    Proto_Van_Emde_Boas(int size)
        universe_size = size;
        // Base case
        if (size <= 2) {
            // Set summary to nullptr as there is no
            // more summary for size 2
            summary = nullptr;
            // Vector of two pointers
            // nullptr in starting
            clusters = vector<Proto_Van_Emde_Boas*>(size, nullptr);
        else {
            // Assigning Proto-VEB(sqrt(u)) to summary
            summary = new Proto_Van_Emde_Boas(root(size));
            // Creating array of Proto-VEB Tree pointers of size sqrt(u)
            // first all nullptrs are going to assign
            clusters = vector<Proto_Van_Emde_Boas*>(root(size), nullptr);
            // Assigning Proto-VEB(sqrt(u)) to all its clusters
            for (int i = 0; i < root(size); i++) {
                clusters[i] = new Proto_Van_Emde_Boas(root(size));
// Driver code
int main()
    Proto_Van_Emde_Boas pveb(4);

Time Complexity : O(sqrt(n))

My Personal Notes arrow_drop_up
Recommended Articles
Page :

Start Your Coding Journey Now!