What makes SQL special

Written by Chris Zetter

Published 2025-07-20

I wrote a book that guides you through building a database server that can run SQL.

But why write a book about SQL? And why use SQL as the interface to a database? Over the last 50 years, SQL has become the dominant way to communicate with databases. Here are some of the characteristics that make SQL special and have contributed to its popularity:

Relational databases

In the 1970s, the relational model was introduced. The purpose of the model was to provide an abstraction so that you could query data without knowing how it was physically stored. This was done by representing data as tables made up of rows and having a defined set of operations that can transform this data.

SQL embraced the ideas of the relational model and became the language of relational databases. There are other query languages, but none have had the popularity and longevity of SQL.

The relational model uses ideas from set theory to precisely describe relational databases and operations you can perform on them. The image above shows an ‘orders’ table made up of rows. Using the terminology from the relational model, you would call the ‘orders’ a relation made up of tuples. A tuple is a list of values, just like a row. A relation is a set of tuples with the same structure, just like a table is a set of rows.

These terms aren’t always equivalent. For example, In the relational model, tuples should be unique within a relation. In practice, rows in databases don’t usually enforce this constraint. You’re likely to see these terms in database research and in the source code of database implementations.

WHILE TRUE  
  department = NEXT_RECORD('department')  
  BREAK IF EMPTY(department)  
  WHILE TRUE  
    employee = NEXT_RECORD_WITHIN(department, 'employee')  
    BREAK IF EMPTY(employee)  
    PRINT(employee.name)  
  END  
END  

Designed for everyone

The authors of SQL (originally called SEQUEL) aimed to make a language that was easier to use by people who weren’t trained programmers. They did this by creating a language around the set of operations in the relational model and avoiding having to manage variables when writing queries.

There were other languages based on the relational model. QUEL is a query language used for the INGRES database that was made at a similar time to SQL. Here’s a QUEL query that computes the pay for an employee with a certain name:

RANGE OF E IS EMPLOYEE  
RETRIEVE INTO W  
(PAY = E.RATE*E.HOURS)  
WHERE E.NAME = "Jones"  

And an equivalent query in SQL:

SELECT RATE*HOURS AS PAY FROM EMPLOYEE  
WHERE NAME = "Jones"  

Both SQL and QUEL allow you to perform similar operations. In QUEL you need to explicitly define variables to represent rows of the employee table, for example the E variable for the EMPLOYEE table. While programmers are used to dealing with variables like this in code, the creators of SQL thought that it would be a barrier to wider use.

A declarative language

When writing SQL, you describe the data you want using high-level operations that are based on the relational model.

This gives databases the power to choose the details of their implementations. For example, databases get to decide what sorting algorithm to use when ordering data. They also choose how to organize data they save to physical storage.

Databases might make decisions so they can specialize for certain uses. They might also have multiple implementations of the same operation and will choose the one that might be fastest for an individual query.

This is great if you want to build your own database server, since you’ll have a large amount of freedom when choosing how to execute SQL statements.

Widespread implementations

Whatever you want to do with data, there’s probably an SQL database suited to your needs.

There are transactional databases (sometimes called OLTP databases for OnLine Transaction Processing) which are designed to quickly read and write data so are well suited to responding to user actions such as creating an account or paying for a purchase. MySQL, PostgreSQL, Oracle Database, Microsoft SQL Server are examples of transactional databases.

There are also analytical databases (sometimes called OLAP databases for OnLine Analytical Processing). These are optimized to run analytical queries across large amounts of data. Google’s BigQuery, Amazon’s RedShift, Clickhouse are all analytical databases.

Transactional and analytical databases are examples of two specializations. There are also databases that are designed to automatically scale in a cloud environment, or databases that are designed to run on resource-constrained devices.

Despite being built for different purposes, all these databases use SQL. This makes being able to write SQL and understand the concepts behind it a valuable skill.

CREATE TABLE orders(id INTEGER);  
SELECT id FROM orders;  
SELECT ID FROM orders;  

SELECT "id" FROM orders; -- lowercase databases only  
SELECT "ID" FROM orders; -- uppercase databases only  

More

This post is adapted from the first chapter of Build a Database Server, a hands-on guide to implementing a working SQL database from scratch. Preview or buy the book to learn more about how databases work.