A lightweight, git-friendly SQL database that stores data as JSON files

MicroSQL is a zero-dependency database engine for JavaScript/TypeScript projects. Run SQL-like queries on JSON files without SQLite, Postgres, or any database server. Perfect for open source projects where you want users to clone and run immediately.

• Zero setup - No database installation, no configuration files, just JSON
• Git-friendly - Human-readable diffs, easy to version control, no binary .db files
• Zero dependencies - Pure Node.js, works everywhere
• Familiar syntax - If you know SQL, you already know MicroSQL

npm install microsql

Or with pnpm:

pnpm add microsql

import { MicroSQL } from "microsql";
const db = new MicroSQL("./data");
db.query(`INSERT INTO users (id, name, age, city) VALUES (1, "Alice", 25, "Berlin")`);
db.query(`INSERT INTO users (id, name, age, city) VALUES (2, "Bob", 30, "Paris")`);
const users = db.query(`SELECT * FROM users WHERE age > 20`);
console.log(users);

Each table is stored as ./data/users.json - no database server required.

• Comparison: =, >, <, >=, <=
• Pattern matching: LIKE with % (any chars) and _ (single char)
• List membership: IN (value1, value2, ...)
• Logic: AND, OR with parentheses support

const db = new MicroSQL("./data");
db.query(`INSERT INTO products (id, name, price) VALUES (1, "Laptop", 999)`);
const products = db.query(`SELECT * FROM products`);
db.query(`UPDATE products SET price = 899 WHERE id = 1`);
db.query(`DELETE FROM products WHERE id = 1`);

db.query(`SELECT * FROM users WHERE age >= 18 AND city = "Berlin"`);
db.query(`SELECT name FROM users WHERE email LIKE "%@company.com"`);
db.query(`SELECT name FROM users WHERE id IN (1, 3, 5)`);
db.query(`SELECT name, age FROM users ORDER BY age DESC LIMIT 5`);

db.query(`
 SELECT name, email 
 FROM users 
 WHERE (city = "Berlin" AND age > 25) OR status = "premium"
 ORDER BY name ASC
 LIMIT 10
`);

MicroSQL is perfect for:

• Open source projects - Users can clone and run without database setup
• Prototypes and demos - Quick data persistence without infrastructure
• CLI tools - Store configuration and state in readable JSON
• Static site generators - Keep data in version control
• Small applications - Where SQLite is overkill

• High-traffic production applications (no concurrency control)
• Large datasets (>10,000 rows per table)
• Applications requiring transactions or ACID guarantees
• Multi-user write scenarios (no locking mechanism)

1. Each table is stored as a separate JSON file (./data/tablename.json)
2. Queries are parsed with regex-based SQL parser
3. Operations load JSON, apply logic in-memory, and save back
4. No indexes or query optimization (trades performance for simplicity)

./data/
├── users.json
├── products.json
└── orders.json

Each JSON file contains an array of objects:

[
 { "id": "1", "name": "Alice", "age": "25" },
 { "id": "2", "name": "Bob", "age": "30" }
]

• No concurrency control - Not safe for simultaneous writes
• No indexes - All queries are full table scans
• Limited SQL - Subset of SQL features only

• GROUP BY and aggregate functions (COUNT, SUM, AVG)
• Type coercion (proper number/boolean handling)

MIT