What Is Hashing and Its Techniques in Data Structure?


Heather Bennett

What Is Hashing and Its Techniques in Data Structure?

Hashing is a fundamental concept in data structures that allows for efficient storage and retrieval of data. It involves using a function, called a hash function, to map data to a fixed-size array, known as a hash table or hash map. In this article, we will explore the basics of hashing and its various techniques.

Hash Function

A hash function is a mathematical function that takes an input (or key) and produces a fixed-size output value, known as the hash code or hash value. The primary objective of a hash function is to minimize collisions by distributing the keys uniformly across the array.

Let’s consider an example where we have an array of size 10. A simple hash function could be taking the remainder when dividing the key by 10:

hash(key) = key % 10

Collision Resolution Techniques

In hashing, collisions occur when two different keys generate the same index in the array. To handle collisions efficiently, various collision resolution techniques are used:

1. Separate Chaining

In separate chaining, each element in the hash table is linked to a linked list or some other data structure. When collisions occur, new elements are simply appended to the existing list at that index. This technique ensures constant-time complexity for insertion and deletion.

2. Open Addressing

  • Linear Probing: If there is a collision at index ‘i’, linear probing checks the next index ‘i+1’.

    If it is also occupied, it continues checking subsequent indices until it finds an empty slot.

  • Quadratic Probing: Similar to linear probing, but instead of checking consecutive indices, quadratic probing checks indices based on a quadratic equation.
  • Double Hashing: Double hashing uses two hash functions. If there is a collision at index ‘i’, it calculates a new index by applying the second hash function to the key and increments ‘i’ accordingly.

Advantages of Hashing

  • Fast access and retrieval time: Hashing provides constant-time complexity for insertions, deletions, and lookups in the average case.
  • Efficient memory usage: Hash tables can be dynamically resized to accommodate more elements while maintaining optimal performance.
  • Hash functions for data integrity: Hash functions are commonly used to ensure data integrity by generating unique hash codes for data.


Hashing is a powerful technique used in data structures to efficiently store and retrieve data. It employs a hash function to map keys to array indices, ensuring fast operations and optimal memory usage.

Understanding different collision resolution techniques is crucial for implementing efficient hash tables. By incorporating hashing into your applications, you can significantly improve performance and ensure data integrity.


[1] Cormen, T. H., Leiserson, C. E., Rivest, R. L., & Stein, C. (2009). Introduction to Algorithms (3rd ed.).

The MIT Press.

[2] Sahni, S., & Anderson-Freed, S. (1996). Fundamentals of Data Structures in C++. Silicon Press.

Discord Server - Web Server - Private Server - DNS Server - Object-Oriented Programming - Scripting - Data Types - Data Structures

Privacy Policy