Fuzion

Runtime implementation of shortcut fusion for JavaScript and TypeScript

Fuzion is an optimization technique that merges multiple array operations into a single loop, eliminating intermediate data structures and providing significant performance improvements over chained native methods.

📖 What is Shortcut Fusion?

Definition (haskell docs): Shortcut fusion is an optimizer method that merges some function calls into one. E.g., map f * map g can be substituted by map (f * g), and filter p * filter q can be substituted by filter (\x -> q x && p x). It can also help to remove intermediate data structures. E.g., computing sum [1..n] does not require an explicit list structure, and the expression is actually translated into a simple loop.

🛠️ API

• fuzion(input, ...operators) - Main function for shortcut fusion
• map(fn) - Transform each element
• filter(predicate) - Include elements that pass the test
• forEach(fn) - Execute function for each element
• take(count) - Limit result to first N elements

💡 Usage Examples

Basic Operations

import { fuzion, map, filter, take } from 'fuzion';

// Simple transformation
const doubled = fuzion([1, 2, 3, 4, 5], map(x => x * 2));
// Result: [2, 4, 6, 8, 10]

// Filter and transform
const evensDoubled = fuzion([1, 2, 3, 4, 5],
  filter(x => x % 2 === 0),
  map(x => x * 2)
);
// Result: [4, 8]

// Complex chain with early exit
const result = fuzion([1, 2, 3, 4, 5, 6, 7, 8, 9, 10],
  map(x => x * 2),           // Double each number
  filter(x => x > 5),         // Keep only > 5
  take(3)                     // Take first 3 results
);
// Result: [6, 8, 10] - stops processing after finding 3 results

TypeScript Support

Fuzion provides full TypeScript inference with up to 13 operator chains:

interface User {
  id: number;
  name: string;
  age: number;
}

const users: User[] = [
  { id: 1, name: 'Alice', age: 25 },
  { id: 2, name: 'Bob', age: 30 },
  { id: 3, name: 'Charlie', age: 35 }
];

// Full type inference
const adultNames = fuzion(users,
  filter(user => user.age >= 30),    // TypeScript knows this is User[]
  map(user => user.name)             // TypeScript infers string[]
);
// Type: string[] = ['Bob', 'Charlie']

⚡ Performance

• MAP-only chains: Competitive with chained native (often faster)
• TAKE operator: Up to 58x faster than chained native with early exit
• Memory efficient: No intermediate arrays created
• Optimized constants: Numeric operator types for faster comparisons
• Lightweight: Only 599 B gzipped (tree-shakeable for smaller bundles)

Operation Type	Performance vs Chained Native	Notes
MAP-only chains	1.03x faster	Pre-allocated arrays
TAKE operations	Up to 58x faster	Early exit optimization
Mixed operations	1.18x slower	Function call overhead
Single operations	1.18-1.81x slower	Baseline overhead

Performance Comparison

Traditional chaining (creates intermediate arrays):

const result = array
  .map(x => x * 2)      // Creates new array
  .map(x => x + 1)      // Creates another array
  .filter(x => x > 5);  // Creates final array

Fuzion (single loop, no intermediate arrays):

const result = fuzion(array,
  map(x => x * 2),      // Fused into single loop
  map(x => x + 1),      // No intermediate arrays
  filter(x => x > 5)    // Direct processing
);

Real-world example:

const data = Array.from({ length: 10000 }, (_, i) => i + 1);

// Chained native (slower - creates intermediate arrays)
const chained = data
  .map(x => x * 2)
  .map(x => x + 1)
  .filter(x => x > 5)
  .slice(0, 100);

// Fuzion (faster - single loop with early exit)
const fused = fuzion(data,
  map(x => x * 2),
  map(x => x + 1),
  filter(x => x > 5),
  take(100)
);

🎯 When to Use Fuzion

✅ Great for:

• Complex data transformation pipelines
• Operations with take() for early exit
• MAP-only chains (often faster than native)
• Memory-constrained environments
• Functional programming patterns

⚠️ Consider alternatives for:

• Simple single operations (use native methods)
• Performance-critical single loops (manual implementation)
• Very small arrays (< 100 elements)

🔧 Installation

npm install fuzion
# or
yarn add fuzion

📄 License

MIT License - see LICENSE file for details.