PrefCheck

Benchmarks for runtime type validation libraries.

Website

Test cases

Type	Valid	Invalid
parseSafe	Parses and returns cleaned data	Throws on invalid
parseStrict	Like safe, but rejects unknown keys	Throws on unknown keys
assertLoose	Returns true without parsing	Throws on invalid
assertStrict	Returns true, rejects extra keys	Throws on invalid or extra keys

Score

The score is the normalized geometric mean of valid tests — the industry standard for benchmarking (used by SPEC CPU, Phoronix, AnandTech). Each test contributes equal weight.

• Library with the highest geomean → 100
• Others scaled proportionally
• Bars show operations per second (higher = better)
• Margin (±%) is the 95% confidence interval

Quick start

# Install
npm install

# Run benchmarks
bun run schema:run:bun   # Bun
npm run schema:run       # Node (via tsx)

Results are saved to src/modules/schema/results/ as per-platform gzipped JSON files.

# Dev server
npm run dev

Open http://localhost:5173 — select a runtime, toggle valid/invalid, filter by test type.

Adding a library

1. Create a directory: src/modules/schema/cases/<slug>/
2. Add meta.json:

{
  "package": "my-lib",
  "npm": "https://www.npmjs.com/package/my-lib",
  "github": "https://github.com/user/my-lib"
}

3. (Optional) Add build.ts to compile the library.
4. Add index.ts — call createCase() for each scenario:

import { createCase } from '../../benchmarks/case.ts';
import { parse, assert } from 'my-lib';

createCase('parseSafe', () => {
  return (data) => parse(data)
});
createCase('parseStrict', () => {
  return (data) => parse(data, { strict: true })
});
createCase('assertLoose', () => {
  return (data) => { 
    assert(data);
    return true;
  }
});
createCase('assertStrict', () => {
  return (data) => { 
    assert(data, { strict: true });
    return true;
  }
});

5. Run test for check output is correct: npm run test