3.5.2 Basics of Relational Databases

Elective Chapter

This chapter is optional. If you only want to do data analysis and modeling, you can skip it for now. But if you plan to build AI applications in the future (such as RAG systems or AI Agents), database knowledge is essential.

Learning Objectives

• Understand what a database is and why we need it
• Master the core concepts of relational databases
• Understand the meaning of tables, rows, columns, primary keys, and foreign keys
• Learn about common database management systems

---

Why do AI engineers need to learn databases?

You may be thinking: "I already know how to use Pandas to read CSV files, so why do I still need to learn databases?"

Scenario	CSV files	Database
Data size	Okay for tens of thousands of rows, slows to a crawl at millions	Easily handles hundreds of millions of rows
Collaboration	Hard to know who changed what, easy to conflict	Supports concurrent read/write with access control
Data security	If the file is deleted, it is gone	Has backups, transactions, and crash recovery
Query speed	Must scan everything each time	Has indexes, millisecond-level queries
Data relationships	Manually merge multiple files	Use one SQL JOIN to get it done

Real-world examples:

• Building a RAG system → user Q&A records need to be stored in a database
• Building an AI Agent → the memory system needs persistent storage
• Building a recommendation system → user behavior data lives in the database
• Doing data analysis → 99% of enterprise data is stored in databases

---

What is a relational database?

A comparison with Excel

If you have used Excel before, you already understand 80% of the concepts of relational databases:

Excel concept	Database concept	Explanation
One Excel file	One database	A container that stores all data
One Sheet	One table	Stores one type of data
One row	One record (Row / Record)	One specific data entity
One column	One field (Column / Field)	One attribute of the data
Column header	Column name	The attribute name
Cell data type	Data type	Integer, text, date, and so on

Understand it with an example

Suppose you run an online store and need to manage users and orders:

Users table (users):

id	name	email	age	city
1	Zhang San	[email protected]	28	Beijing
2	Li Si	[email protected]	35	Shanghai
3	Wang Wu	[email protected]	22	Guangzhou

Orders table (orders):

order_id	user_id	product	amount	order_date
101	1	iPhone 16	7999	2024-11-01
102	1	AirPods	999	2024-11-05
103	2	MacBook	14999	2024-11-10

These two tables are connected through user_id—this is where the name relational database comes from: tables have relationships with each other.

---

Core concepts

Primary Key

A primary key is the unique identifier for each record, like an ID card number: it cannot be repeated and cannot be empty.

In the users table: id is the primary key → each user has a unique id
In the orders table: order_id is the primary key → each order has a unique order_id

Why do we need a primary key?

Imagine there is no primary key: if two users are both named "Zhang San", how would you tell them apart? The primary key solves this problem—even if the names are the same, the ids must be different.

Foreign Key

A foreign key is a field that references the primary key of another table and is used to build relationships between tables.

The user_id in the orders table is the foreign key—it points to the id in the users table and indicates "which user this order belongs to."

Common data types

Type	Explanation	Example
INTEGER	Integer	1, 42, -100
REAL / FLOAT	Floating-point number	3.14, 99.9
TEXT / VARCHAR	Text string	"Zhang San", "hello"
DATE	Date	2024-11-01
DATETIME	Date and time	2024-11-01 14:30:00
BOOLEAN	Boolean value	TRUE / FALSE
BLOB	Binary data	Images, files (less commonly used)

Constraints

Constraints are rules for data that help ensure data quality:

Constraint	Purpose	Example
PRIMARY KEY	Primary key, unique and not null	id
NOT NULL	Cannot be empty	name NOT NULL
UNIQUE	Value cannot be duplicated	email UNIQUE
DEFAULT	Default value	city DEFAULT 'Unknown'
FOREIGN KEY	Foreign key, references another table	user_id REFERENCES users(id)

---

Common database management systems

Database	Features	Use cases
SQLite	Zero configuration, stored in a single file	Learning, small apps, mobile apps
MySQL	The most popular open-source database	Web applications, small to medium projects
PostgreSQL	The most powerful open-source database	Large projects, AI applications (supports vector search)
SQL Server	Made by Microsoft	Enterprise Windows environments

This chapter uses SQLite

SQLite does not require installing any server, and Python comes with the sqlite3 module, making it ideal for learning. All SQL syntax is also applicable in other databases.

The safest default order when learning databases for the first time

A more stable order is usually:

1. First get familiar with “tables, primary keys, and foreign keys”
2. Then learn SQL queries
3. Then connect Python to the database
4. Finally, learn database design

This is less confusing than trying to memorize a lot of SQL details right from the start.

---

Hands-on: Create your first database

import sqlite3

# 1. Connect to the database (it will be created automatically if it does not exist)
conn = sqlite3.connect("my_shop.db")
cursor = conn.cursor()

# 2. Create the users table
cursor.execute("""
    CREATE TABLE IF NOT EXISTS users (
        id INTEGER PRIMARY KEY AUTOINCREMENT,
        name TEXT NOT NULL,
        email TEXT UNIQUE,
        age INTEGER,
        city TEXT DEFAULT 'Unknown'
    )
""")

# 3. Insert data
cursor.execute("INSERT INTO users (name, email, age, city) VALUES ('Zhang San', '[email protected]', 28, 'Beijing')")
cursor.execute("INSERT INTO users (name, email, age, city) VALUES ('Li Si', '[email protected]', 35, 'Shanghai')")
cursor.execute("INSERT INTO users (name, email, age, city) VALUES ('Wang Wu', '[email protected]', 22, 'Guangzhou')")

# 4. Commit changes
conn.commit()

# 5. Query data
cursor.execute("SELECT * FROM users")
rows = cursor.fetchall()
for row in rows:
    print(row)
# (1, 'Zhang San', '[email protected]', 28, 'Beijing')
# (2, 'Li Si', '[email protected]', 35, 'Shanghai')
# (3, 'Wang Wu', '[email protected]', 22, 'Guangzhou')

# 6. Close the connection
conn.close()

Congratulations! You just created a database, created a table, and inserted 3 rows of data.

What is the most important thing to learn from this small example first?

The most important thing is not every SQL keyword, but that the smallest database workflow is actually very straightforward:

1. Connect to the database
2. Create a table
3. Insert data
4. Query the results

Once this flow makes sense, learning SQL and Python database connections later will feel much less abstract.

---

Summary

Concept	One-sentence understanding
Database	A "folder" that stores all tables
Table	An "Excel worksheet" for one kind of data
Primary key	The "ID card number" of each record
Foreign key	The "link" connecting two tables
Constraint	A "rule" that ensures data quality

What should you take away from this lesson?

• The most important thing about relational databases is not "storing many tables," but that tables can establish relationships through keys
• Primary keys uniquely identify records, and foreign keys connect tables
• Once this foundation is clear, SQL and multi-table analysis will make much more sense

---

Hands-on practice

Exercise 1: Design table structures

Design two tables for a library management system:
- books table: title, author, publication year, price, category
- borrows table: borrowing records (who borrowed which book, borrow date, return date)

Think about:
1. What is the primary key of each table?
2. Which foreign keys does the borrows table need?
3. Which fields should have NOT NULL constraints?

Exercise 2: Practice with SQLite

# Use sqlite3 to create the books table and borrows table designed above
# Insert 5 books and 3 borrowing records
# Query all data and print it