SQL Learning Path: Start Here

In short: you learn SQL in 7 milestones. Each one has a small dataset and a set of tasks with the expected answers, so you always know whether you're right. Then you apply SQL to your role.

How to use this page

Read this page once to see the plan. Then, for each milestone, follow the same four steps: learn → solve → check → commit. Come back here whenever you're unsure what to do next.

The pages in this learning path

In short: each page has one job, so nothing is explained twice.

Page	What it's for	When to open it
This roadmap	The plan: what each milestone covers and how to check yourself	At the start of each milestone
SQL cheat sheet	Learn each idea: short explanation, example, exercise	The "learn" step of every milestone
Milestones & Mini-Projects	A dataset and tasks with expected results for each milestone	The "solve" and "check" steps
Quick Reference	Look up syntax, functions, and other databases' spellings	Any time you're writing SQL
SQL Best Practices	Habits that keep SQL readable, fast, and safe	After Milestone 4, then before every merge
SQL: The Complete Guide	Internals and interview questions	When you want the deeper "why"

Part 1 — Core SQL: Milestones 1–7

In short: everyone does these seven, in order.

Milestone	You learn	You work on
1	Choosing, sorting, and limiting rows	Library catalogue
2	Counting, grouping, and summarising	Weather report
3	Joining tables	Library loans
4	Creating tables, constraints, changing data	To-do app schema
5	Subqueries, CTEs, CASE	Customer segments
6	Window functions	Game leaderboard
7	Transactions, indexes, query plans	Bank ledger

For each milestone:

1. Learn — read the cheat-sheet sections listed below, and run every example on the practice database.
2. Solve — run the milestone's setup in Milestones & Mini-Projects, then write a query for each task.
3. Check — compare your result with the expected one. Only then open the solution, and compare the two queries too.
4. Commit — save your queries in milestone-N.sql and commit:

git add milestone-1.sql
git commit -m "Milestone 1: library catalogue, 8/8 tasks correct"

Milestone 1: Choosing rows

Learn: What SQL is · Choosing rows · Sorting & limiting · Text, numbers & NULL

Work on: Library catalogue

Check yourself:

• Why does WHERE email = NULL never match anything?
• What's the difference between LIKE 'The%' and LIKE 'The %'?
• Why can't you rely on row order without ORDER BY?

Milestone 2: Counting & grouping

Learn: Counting & grouping · Quick Reference: Aggregates, Clause order

Work on: Weather report

Check yourself:

• What's the difference between COUNT(*) and COUNT(column)?
• When do you use WHERE, and when HAVING?
• Why must every selected column be grouped or aggregated?

Milestone 3: Joining tables

Learn: Joining tables · Quick Reference: Joins

Work on: Library loans

Check yourself:

• When do you need a LEFT JOIN instead of a JOIN?
• How do you find rows in one table with no match in another?
• Why can a join make a SUM or COUNT too big?

Milestone 4: Building tables & changing data

Learn: Changing data · Creating tables · Insert or update (upsert)

Work on: To-do app schema

Then read: SQL Best Practices, sections 1–5 and 8–9.

Check yourself:

• What does each constraint type protect against?
• What does ON DELETE CASCADE do, and when is it dangerous?
• What's your routine before running an UPDATE on real data?

Milestone 5: Subqueries, CTEs & CASE

Learn: Subqueries · CTEs · CASE · Combining results · Views

Work on: Customer segments

Check yourself:

• When is a CTE clearer than a subquery?
• Why is NOT EXISTS safer than NOT IN?
• How do you count several conditions in one pass?

Milestone 6: Window functions

Learn: Window functions · Dates & times · Quick Reference: Window functions

Work on: Game leaderboard

Check yourself:

• What's the difference between GROUP BY and PARTITION BY?
• When do RANK, DENSE_RANK, and ROW_NUMBER give different answers?
• How do you get the top row per group?

Milestone 7: Transactions & performance

Learn: Indexes · Transactions · Reading query plans · Isolation & locking · Recursive queries

Work on: Bank ledger

Check yourself:

• What do the four letters of ACID promise?
• Which columns would you index first, and what does each index cost?
• How do you spot a full-table scan in a plan?

Part 2 — Your role track

In short: SQL is used differently by each role. Pick yours.

Role	Read next	Final project
SDET (test engineer)	SQL for SDET — checking what the app saved, test data, data-quality checks, comparing tables, testing migrations	A database test suite that runs in CI
SDE (software developer)	SQL Best Practices sections 6–11, then the complete guide on indexes and transactions	A small app with migrations and parameterised queries
SRE (reliability engineer)	Isolation & locking, Reading query plans, and the Quick Reference's monitoring one-liners	A runbook for "the database is slow": find the query, read its plan, fix it

Also finish the cheat sheet's Part 3: JSON columns, Designing tables, and SQL from code.

How you know you're ready to move on

After Milestone 7:

• Every task in all seven milestones gives the expected result.
• You can write a three-table join with grouping without looking anything up.
• You can explain a query plan's main line to someone else.
• Your git history shows steady progress.

If you get stuck

Problem	What to do
"column … does not exist"	Check the clause order: WHERE can't use a SELECT alias. Also check spelling and table aliases.
"must appear in the GROUP BY clause"	Every selected column must be grouped or inside COUNT/SUM/….
Too many rows after a join	A join condition is missing or matches more than you expected — count rows after each join.
Right numbers, wrong order	Add or fix ORDER BY.
Your answer differs from the expected one	Run the setup again on a fresh database — earlier experiments may have changed the data.

Start now

Install PostgreSQL (or use SQLite, already on macOS and most Linux), create a database, and load the practice data from the cheat sheet.

createdb shop                       # PostgreSQL: create a database
psql shop                           # open it, then paste the practice setup
# or, with SQLite:
sqlite3 shop.db                     # creates the file on first use

Helpful tools:

Tool	What it does
DBeaver or pgAdmin	Free desktop apps to browse tables and run queries
psql / sqlite3	Command-line clients — see the command table
Docker	Run PostgreSQL without installing it: docker run -e POSTGRES_PASSWORD=pw -p 5432:5432 postgres

Good books to read alongside: Learning SQL (Alan Beaulieu) to build foundations, and SQL Performance Explained (Markus Winand, also free online as "Use The Index, Luke") for indexes and speed.