Performance

Benchmark Results

All benchmarks run on Bun, measuring operations per second:

Operation	Vex	Zod	Valibot	vs Zod	vs Valibot
string	367M	50M	50M	7.4x	7.3x
email	128M	12M	17M	11x	7.5x
url	110M	4M	5M	28x	23x
uuid	117M	16M	6M	7.3x	19x
object (3 fields)	18M	5M	7M	3.9x	2.7x
object (nested)	11M	5M	5M	2x	2.1x
array[50]	8.5M	780K	1.4M	11x	6x
safeParse (valid)	15M	7M	6M	2.1x	2.5x
safeParse (invalid)	44M	376K	2.6M	118x	17x

Average: 12x faster than Zod, 6x faster than Valibot

Why is Vex Fast?

1. Pure Functional Design

Validators are constants, not factory functions:

// Vex - zero allocation
str                  // constant
str(email)           // compose constants

// Others - allocates on every call
z.string().email()   // creates new objects
v.string()           // creates new objects

2. No Wrapper Objects

Direct function composition without intermediate objects:

// Vex internals (simplified)
const str = (value) => {
  if (typeof value !== 'string') throw error
  return value
}

// Others create class instances
class StringSchema {
  email() { return new EmailSchema(this) }
  // ...
}

3. Error Path Optimization

Error paths are only computed when validation fails:

// Only allocate on error, not success path
if (!valid) {
  throw new ValidationError(message, path)
}

Running Benchmarks

# Clone the repo
git clone https://github.com/SylphxAI/vex.git
cd vex

# Install dependencies
bun install

# Run benchmarks
bun run bench

Tree-shaking

Vex is fully tree-shakeable:

// Only imports what you use
import { str, num, object } from '@sylphx/vex'

// Bundle size: ~2KB (minified + gzipped)

Compare to Zod which imports the entire library regardless of usage.