Postgres Lite 🚀

An Ultra-Lightweight Embedded PostgreSQL Engine that Runs Natively in Node.js and the Browser — No WASM, No Emulation, Just Pure Performance.

PostgresLite is a high-performance, in-process, zero-dependency embedded PostgreSQL database engine for Bun, Node.js, and the Browser. It provides a PostgreSQL-compatible SQL interface with the simplicity of a local storage engine—effectively bringing the power of Postgres to the developer experience, serving as a robust SQLite alternative.

Unlike traditional PostgreSQL, PostgresLite requires no server, no network overhead, and zero configuration. It utilizes a custom-built storage engine designed for low-latency I/O, high concurrency, and full ACID compliance.

🚀 Key Features

• Cross-Runtime: Native support for Bun, Node.js, and Modern Browsers (via IndexedDB).
• Serverless: In-process execution; no connection strings, background processes, or docker containers needed.
• PostgreSQL Dialect: Supports a vast subset of the Postgres syntax including Joins, CTEs, and Window Functions.
• Performance: Capable of handling 1M+ records per table via B-Tree indexing and advanced Buffer Pool management.
• ACID Compliant: Supports full transactions with BEGIN, COMMIT, and ROLLBACK via Write-Ahead Logging (WAL).
• Schema Isolation: Multi-schema support (public, pg_catalog, information_schema).

🛠 Supported Syntax

🏗 Optimization Technologies

PostgresLite is built with several advanced database engineering techniques to ensure high performance:

1. Slotted Page Layout: Data is stored in fixed 4KB pages using a slotted-page architecture. This allows for efficient management of variable-length records (like JSONB or TEXT) and prevents page fragmentation.
2. Write-Ahead Logging (WAL): Every mutation is logged to a persistent WAL file before being applied to the main database. This ensures durability and allows for automatic crash recovery.
3. B-Tree Indexing: Primary keys are automatically indexed using a B-Tree, enabling $O(\log n)$ point lookups even as datasets scale into the millions.
4. Volcano Execution Model: The engine uses an iterator-based processing model. Rows are "pulled" through the execution plan one by one, ensuring that complex queries (like SELECT *) use a constant and minimal memory footprint.
5. External Merge Sort: For large ORDER BY operations that exceed available RAM, the engine automatically spills to disk and performs a multi-way merge sort to maintain memory safety.
6. LRU Buffer Pool: A sophisticated Least-Recently-Used (LRU) cache minimizes physical disk I/O by keeping frequently accessed pages in memory.
7. Predicate Pushdown: The execution engine optimizes filters by pushing them down to the storage layer, utilizing indices for $O(1)$ lookups whenever possible.

📦 Installation

Using NPM

npm install @pglite/core

Using Bun

bun add @pglite/core

🛠 Usage

import { PGLite } from "@pglite/core";

// For Node.js/Bun:
import { NodeFSAdapter } from "@pglite/core/node-fs";
const db = new PGLite("app.db", { adapter: new NodeFSAdapter() });

// For Browser:
import { BrowserFSAdapter } from "@pglite/core/browser";
const db = new PGLite("app.db", { adapter: new BrowserFSAdapter() });

// 1. DDL & Data Mutation
await db.exec(`
  CREATE TABLE users (
    id SERIAL PRIMARY KEY,
    name TEXT NOT NULL,
    metadata JSONB
  )
`);

await db.exec(`
  CREATE TABLE posts (
    id SERIAL PRIMARY KEY,
    user_id INTEGER REFERENCES users(id),
    title TEXT NOT NULL,
    content TEXT
  )
`);

// 2. Parameterized Queries (SQL Injection Protected)
await db.exec(
  "INSERT INTO users (name, metadata) VALUES ($1, $2)", 
  ["Alice", { role: "admin", active: true }]
);

// 3. Complex Querying (Joins, Aggregates, Grouping)
const results = await db.query(`
  SELECT u.name, COUNT(p.id) as post_count
  FROM users u
  LEFT JOIN posts p ON u.id = p.user_id
  WHERE u.name LIKE $1
  GROUP BY u.name
  ORDER BY post_count DESC
`, ["Al%"]);

console.table(results);

📉 Scalability & Performance Benchmarks

PostgresLite is designed for heavy lifting in local environments:

• Point Lookups: $O(1)$ to $O(\log n)$ via Primary Key B-Tree Index.
• Sequential Scans: High-throughput streaming via the Pager's Buffer Pool.
• Memory Efficiency: The Volcano-style iterator ensures that running a SELECT * on a 1,000,000 row table does not result in an OutOfMemory error.
• Large Sorts: Automatically triggers disk-backed sorting when result sets exceed the configurable buffer threshold.

🔧 Engineering Deep-Dive

The Slotted Page Layout

PostgresLite does not store rows as raw strings. Each 4KB page contains a header, a slot array pointing to record offsets, and the data area. This prevents fragmentation and allows for variable-length records (like JSONB or TEXT) to be updated in place efficiently.

Write-Ahead Logging (WAL)

Every mutation is first appended to a .wal file. In the event of a process crash, the engine automatically replays the WAL on the next initialization, ensuring your database state remains consistent and corruption-free.

🤝 Contributing

We welcome contributions to the core engine, specifically in the following areas:

• Expansion of the SQL Parser for more complex PostgreSQL dialects.
• Implementation of Secondary Indexes.
• Full-text search (TSVECTOR) integration.

📄 License

MIT © Senior Systems Programming Team.