TONL (Token-Optimized Notation Language)

TONL is a production-ready data platform that combines compact serialization with powerful query, modification, indexing, and streaming capabilities. Designed for LLM token efficiency while providing a rich API for data access and manipulation.

🎉 Latest Release: v2.0.6 - Nested Array Length Fix

✨ Key Features in v2.0.6:

🐛 Fixed nested array length preservation - Perfect round-trip for [[]], [[[]]], etc.
🔄 Revolutionary dual-mode system (quoting + preprocessing)
✅ Perfect round-trip safety - 100% data preservation in default mode
🛠️ Advanced quoting for special characters (#, @, "", etc.)
🌐 Browser playground now handles all JSON inputs flawlessly
📋 Optional --preprocess flag for clean, readable output
🔄 Zero data loss guaranteed for both modes

🏠 Homepage: tonl.dev 📦 GitHub: github.com/tonl-dev/tonl 📖 Documentation: Complete Guides

Why TONL?

🗜️ Up to 60% Smaller - Reduce JSON size and LLM token costs 👁️ Human-Readable - Clear text format, not binary 🚀 Blazingly Fast - 10-1600x faster than targets 🔒 Production Secure - 100% security hardened (v2.0.3) 🛠️ TypeScript-First - Full type safety & IntelliSense 📦 Zero Dependencies - Pure TypeScript, no bloat 🌐 Browser Ready - 10.5 KB gzipped bundle (IIFE/UMD) ✅ 100% Tested - 496/496 tests passing (core functionality)

🚀 Quick Start

Installation

npm install tonl

Basic Usage

import { TONLDocument, encodeTONL, decodeTONL } from 'tonl';

// Create from JSON
const doc = TONLDocument.fromJSON({
  users: [
    { id: 1, name: "Alice", role: "admin", age: 30 },
    { id: 2, name: "Bob", role: "user", age: 25 }
  ]
});

// Query with JSONPath-like syntax
doc.get('users[0].name');                          // 'Alice'
doc.query('users[*].name');                        // ['Alice', 'Bob']
doc.query('users[?(@.role == "admin")]');          // [{ id: 1, ... }]
doc.query('$..age');                               // All ages recursively

// Modify data
doc.set('users[0].age', 31);
doc.push('users', { id: 3, name: "Carol", role: "editor", age: 28 });

// Navigate and iterate
for (const [key, value] of doc.entries()) {
  console.log(key, value);
}

doc.walk((path, value, depth) => {
  console.log(`${path}: ${value}`);
});

// Export
const tonl = doc.toTONL();
const json = doc.toJSON();
await doc.save('output.tonl');

// Classic API (encode/decode)
const data = { users: [{ id: 1, name: "Alice" }] };
const tonlText = encodeTONL(data);
const restored = decodeTONL(tonlText);

// Advanced Optimization (v2.0.1+)
import { AdaptiveOptimizer, BitPacker, DeltaEncoder } from 'tonl/optimization';

// Automatic optimization
const optimizer = new AdaptiveOptimizer();
const result = optimizer.optimize(data);  // Auto-selects best strategies

// Specific optimizers
const packer = new BitPacker();
const packed = packer.packBooleans([true, false, true]);

const delta = new DeltaEncoder();
const timestamps = [1704067200000, 1704067201000, 1704067202000];
const compressed = delta.encode(timestamps, 'timestamp');

CLI Usage

# Get started (shows help)
tonl

# Version info
tonl --version

# Encode JSON to TONL (perfect round-trip, quotes special keys)
tonl encode data.json --out data.tonl --smart --stats

# Encode with preprocessing (clean, readable keys)
tonl encode data.json --preprocess --out data.tonl

# Decode TONL to JSON
tonl decode data.tonl --out data.json

# Query data
tonl query users.tonl "users[?(@.role == 'admin')]"
tonl get data.json "user.profile.email"

# Validate against schema
tonl validate users.tonl --schema users.schema.tonl

# Format and prettify
tonl format data.tonl --pretty --out formatted.tonl

# Compare token costs
tonl stats data.json --tokenizer gpt-5

📊 Format Overview

Arrays of Objects (Tabular Format)

JSON (245 bytes, 89 tokens):

{
  "users": [
    { "id": 1, "name": "Alice", "role": "admin" },
    { "id": 2, "name": "Bob, Jr.", "role": "user" },
    { "id": 3, "name": "Carol", "role": "editor" }
  ]
}

TONL (158 bytes, 49 tokens - 45% reduction):

#version 1.0
users[3]{id:u32,name:str,role:str}:
  1, Alice, admin
  2, "Bob, Jr.", user
  3, Carol, editor

Nested Objects

JSON:

{
  "user": {
    "id": 1,
    "name": "Alice",
    "contact": {
      "email": "alice@example.com",
      "phone": "+123456789"
    },
    "roles": ["admin", "editor"]
  }
}

TONL:

#version 1.0
user{id:u32,name:str,contact:obj,roles:list}:
  id: 1
  name: Alice
  contact{email:str,phone:str}:
    email: alice@example.com
    phone: +123456789
  roles[2]: admin, editor

✨ Complete Feature Set

🔄 Core Serialization

Compact Format - 32-45% smaller than JSON (bytes + tokens)
Human-Readable - Clear text format with minimal syntax
Round-Trip Safe - Perfect bidirectional JSON conversion
Smart Encoding - Auto-selects optimal delimiters and formatting
Type Hints - Optional schema information for validation

🔍 Query & Navigation API

JSONPath Queries - users[?(@.age > 25)], $..email
Filter Expressions - ==, !=, >, <, &&, ||, contains, matches
Wildcard Support - users[*].name, **.email
Tree Traversal - entries(), keys(), values(), walk()
LRU Cache - >90% cache hit rate on repeated queries

✏️ Modification API

CRUD Operations - set(), get(), delete(), push(), pop()
Bulk Operations - merge(), update(), removeAll()
Change Tracking - diff() with detailed change reports
Snapshots - Document versioning and comparison
Atomic File Edits - Safe saves with automatic backups

⚡ Performance & Indexing

Hash Index - O(1) exact match lookups
BTree Index - O(log n) range queries
Compound Index - Multi-field indexing
Stream Processing - Handle multi-GB files with <100MB memory
Pipeline Operations - Chainable filter/map/reduce transformations

🗜️ Advanced Optimization

Dictionary Encoding - Value compression via lookup tables (30-50% savings)
Delta Encoding - Sequential data compression (40-60% savings)
Run-Length Encoding - Repetitive value compression (50-80% savings)
Bit Packing - Boolean and small integer bit-level compression (87.5% savings)
Numeric Quantization - Precision reduction for floating-point numbers (20-40% savings)
Schema Inheritance - Reusable column schemas across data blocks (20-40% savings)
Hierarchical Grouping - Common field extraction for nested structures (15-30% savings)
Tokenizer-Aware - LLM tokenizer optimization for minimal token usage (5-15% savings)
Column Reordering - Entropy-based ordering for better compression
Adaptive Optimizer - Automatic strategy selection based on data patterns

✅ Schema & Validation

Schema Definition - .schema.tonl files with TSL (TONL Schema Language)
13 Constraints - required, min, max, pattern, unique, email, etc.
TypeScript Generation - Auto-generate types from schemas
Runtime Validation - Validate data programmatically or via CLI
Strict Mode - Enforce schema compliance

🛠️ Developer Tools

Interactive REPL - Explore data interactively in terminal
CLI Suite - encode, decode, query, validate, format, stats
Browser Support - ESM, UMD, IIFE builds (8.84 KB gzipped)
VS Code Extension - Syntax highlighting for .tonl files
TypeScript-First - Full IntelliSense and type safety

📊 Performance Comparison

Metric	JSON	TONL	TONL Smart	Improvement
Size (bytes)	245	167	158	36% smaller
Tokens (GPT-5)	89	54	49	45% fewer
Encoding Speed	1.0x	15x	12x	12-15x faster
Decoding Speed	1.0x	10x	10x	10x faster
Query Speed	-	-	1600x	Target: <1ms

Benchmarks based on typical e-commerce product catalog data

🔒 Security & Quality

✅ Tests:          496/496 passing (100% coverage)
✅ Security:       All vulnerabilities fixed (100%)
✅ Security Tests: 96 security tests passing
✅ Code Quality:   TypeScript strict mode
✅ Dependencies:   0 runtime dependencies
✅ Bundle Size:    10.5 KB gzipped (browser)
✅ Performance:    10-1600x faster than targets
✅ Production:     Ready & Fully Secure

Security:

✅ ReDoS, Path Traversal, Buffer Overflow protection
✅ Prototype Pollution, Command Injection prevention
✅ Integer Overflow, Type Coercion fixes
✅ Comprehensive input validation and resource limits

See SECURITY.md and CHANGELOG.md for details.

🎯 Use Cases

LLM Prompts

Reduce token costs by 32-45% when including structured data in prompts:

const prompt = `Analyze this user data:\n${doc.toTONL()}`;
// 45% fewer tokens = lower API costs

Configuration Files

Human-readable configs that are compact yet clear:

config{env:str,database:obj,features:list}:
  env: production
  database{host:str,port:u32,ssl:bool}:
    host: db.example.com
    port: 5432
    ssl: true
  features[3]: auth, analytics, caching

API Responses

Efficient data transmission with schema validation:

app.get('/api/users', async (req, res) => {
  const doc = await TONLDocument.load('users.tonl');
  const filtered = doc.query('users[?(@.active == true)]');
  res.type('text/tonl').send(encodeTONL(filtered));
});

Data Pipelines

Stream processing for large datasets:

import { createEncodeStream, createDecodeStream } from 'tonl/stream';

createReadStream('huge.json')
  .pipe(createDecodeStream())
  .pipe(transformStream)
  .pipe(createEncodeStream({ smart: true }))
  .pipe(createWriteStream('output.tonl'));

Log Aggregation

Compact structured logs:

logs[1000]{timestamp:i64,level:str,message:str,metadata:obj}:
  1699564800, INFO, "User login", {user_id:123,ip:"192.168.1.1"}
  1699564801, ERROR, "DB timeout", {query:"SELECT...",duration:5000}
  ...

🌐 Browser Usage

ESM (Modern Browsers)

<script type="module">
  import { encodeTONL, decodeTONL } from 'https://cdn.jsdelivr.net/npm/tonl@2.0.6/+esm';

  const data = { users: [{ id: 1, name: "Alice" }] };
  const tonl = encodeTONL(data);
  console.log(tonl);
</script>

UMD (Universal)

<script src="https://unpkg.com/tonl@2.0.6/dist/browser/tonl.umd.js"></script>
<script>
  const tonl = TONL.encodeTONL({ hello: "world" });
  console.log(tonl);
</script>

Bundle Sizes:

ESM: 15.5 KB gzipped
UMD: 10.7 KB gzipped
IIFE: 10.6 KB gzipped

📚 Complete API Reference

TONLDocument Class

// Creation
TONLDocument.fromJSON(data)
TONLDocument.fromTONL(text)
TONLDocument.load(filepath)

// Query
doc.get(path: string)                              // Single value
doc.query(query: string)                           // Multiple values
doc.has(path: string)                              // Check existence

// Modification
doc.set(path: string, value: any)                  // Set value
doc.delete(path: string)                           // Delete value
doc.push(path: string, value: any)                 // Append to array
doc.pop(path: string)                              // Remove last from array
doc.merge(path: string, value: object)             // Deep merge objects

// Navigation
doc.entries()                                      // Iterator<[key, value]>
doc.keys()                                         // Iterator<string>
doc.values()                                       // Iterator<any>
doc.walk(callback: WalkCallback)                   // Tree traversal
doc.find(predicate: Predicate)                     // Find single value
doc.findAll(predicate: Predicate)                  // Find all matching
doc.some(predicate: Predicate)                     // Any match
doc.every(predicate: Predicate)                    // All match

// Indexing
doc.createIndex(field: string, type?: IndexType)   // Create index
doc.removeIndex(field: string)                     // Remove index
doc.getIndex(field: string)                        // Get index

// Export
doc.toTONL(options?: EncodeOptions)                // Export as TONL
doc.toJSON()                                       // Export as JSON
doc.save(filepath: string, options?)               // Save to file
doc.getSize()                                      // Size in bytes
doc.getStats()                                     // Statistics object

Encode/Decode API

// Encoding
encodeTONL(data: any, options?: {
  delimiter?: "," | "|" | "\t" | ";";
  includeTypes?: boolean;
  version?: string;
  indent?: number;
  singleLinePrimitiveLists?: boolean;
}): string

// Smart encoding (auto-optimized)
encodeSmart(data: any, options?: EncodeOptions): string

// Decoding
decodeTONL(text: string, options?: {
  delimiter?: "," | "|" | "\t" | ";";
  strict?: boolean;
}): any

Schema API

import { parseSchema, validateTONL } from 'tonl/schema';

// Parse schema
const schema = parseSchema(schemaText: string);

// Validate data
const result = validateTONL(data: any, schema: Schema);

if (!result.valid) {
  result.errors.forEach(err => {
    console.error(`${err.field}: ${err.message}`);
  });
}

Streaming API

import { createEncodeStream, createDecodeStream, encodeIterator, decodeIterator } from 'tonl/stream';

// Node.js streams
createReadStream('input.json')
  .pipe(createEncodeStream({ smart: true }))
  .pipe(createWriteStream('output.tonl'));

// Async iterators
for await (const line of encodeIterator(dataStream)) {
  console.log(line);
}

✅ Schema Validation

Define schemas with the TONL Schema Language (TSL):

@schema v1
@strict true
@description "User management schema"

# Define custom types
User: obj
  id: u32 required
  username: str required min:3 max:20 pattern:^[a-zA-Z0-9_]+$
  email: str required pattern:email lowercase:true
  age: u32? min:13 max:150
  roles: list<str> required min:1 unique:true

# Root schema
users: list<User> required min:1
totalCount: u32 required

13 Built-in Constraints:

required - Field must exist
min / max - Numeric range or string/array length
length - Exact length
pattern - Regex validation (or shortcuts: email, url, uuid)
unique - Array elements must be unique
nonempty - String/array cannot be empty
positive / negative - Number sign
integer - Must be integer
multipleOf - Divisibility check
lowercase / uppercase - String case enforcement

See docs/SCHEMA_SPECIFICATION.md for complete reference.

🛠️ Development

Build & Test

# Install dependencies
npm install

# Build TypeScript
npm run build

# Run all tests (496 tests)
npm test

# Watch mode
npm run dev

# Clean build artifacts
npm run clean

Benchmarking

# Byte size comparison
npm run bench

# Token estimation (GPT-5, Claude 3.5, Gemini 2.0, Llama 4)
npm run bench-tokens

# Comprehensive performance analysis
npm run bench-comprehensive

CLI Development

# Install CLI locally
npm run link

# Test commands
tonl encode test.json
tonl query data.tonl "users[*].name"
tonl format data.tonl --pretty

🗺️ Roadmap

✅ v2.0+ - Complete

✅ Advanced optimization module (60% additional compression)
✅ Complete query, modification, indexing, streaming APIs
✅ Schema validation & TypeScript generation
✅ Browser support (10.5 KB bundles)
✅ 100% test coverage & security hardening

🚀 Future

Enhanced VS Code extension (IntelliSense, debugging)
Web playground with live conversion
Python, Go, Rust implementations
Binary TONL format for extreme compression

See ROADMAP.md for our comprehensive development vision.

📖 Documentation

For Users

Getting Started Guide - Beginner-friendly tutorial with examples
API Reference - Complete API documentation with examples
CLI Documentation - Command-line tool guide
Query API - JSONPath-like query syntax reference
Modification API - CRUD operations guide
Navigation API - Tree traversal methods
Use Cases - Real-world scenarios and solutions

For Implementers (Other Languages)

Implementation Reference - Language-agnostic implementation guide
Transformation Examples - 20+ JSON↔TONL conversion examples
Format Specification - Technical format specification
Schema Specification - TSL (TONL Schema Language) spec

Implementing TONL in Python, Go, Rust, or another language? Check out the Implementation Reference for complete algorithms, pseudo-code, and test requirements!

🤝 Contributing

Contributions are welcome! Please read CONTRIBUTING.md for:

Development setup
Code style guidelines
Testing requirements
Pull request process
Architecture overview

📄 License

MIT License - see LICENSE file for details.

🌟 Links

Website: tonl.dev
npm Package: npmjs.com/package/tonl
GitHub: github.com/tonl-dev/tonl
Issues: github.com/tonl-dev/tonl/issues
Discussions: github.com/tonl-dev/tonl/discussions
VS Code Extension: [Coming Soon]

TONL: Making structured data LLM-friendly without sacrificing readability. 🚀

Built with ❤️ by Ersin Koc

Name		Name	Last commit message	Last commit date
Latest commit History 236 Commits
.github		.github
archive/debug-files		archive/debug-files
bench		bench
docs		docs
examples		examples
scripts		scripts
src		src
test		test
vscode-extension		vscode-extension
website		website
.gitignore		.gitignore
.npmignore		.npmignore
CHANGELOG.md		CHANGELOG.md
CLAUDE.md		CLAUDE.md
CONTRIBUTING.md		CONTRIBUTING.md
HOW_IT_WORKS.md		HOW_IT_WORKS.md
LICENSE		LICENSE
README.md		README.md
SECURITY.md		SECURITY.md
fix-imports.js		fix-imports.js
package-lock.json		package-lock.json
package.json		package.json
readme.png		readme.png
tsconfig.json		tsconfig.json
vite.config.ts		vite.config.ts

License

sidarkincal/tonl

Folders and files

Latest commit

History

Repository files navigation