just shoot me 2

license: public domain CC0

 

Let’s design this together the way you would design a compiler: minimal surface area, zero magic, explicit semantics, and no reinvention of the universe. A test runner that is small, predictable, and debuggable—not a second operating system.

Below is what the true MVP of a sane JavaScript test runner would look like.

---

🌱 The world’s simplest test library (that still has good DX)

This is the smallest possible system that:

• runs tests in parallel
• captures logs and prints them only on failure
• supports flags (including via npm test -- …)
• supports naming, skipping, and only
• has predictable, explicit behavior
• avoids reinventing anything (no custom module loader, no VM sandbox, no magic globals)

And it fits in ~150 lines.

Let me show you the architecture first, then the code.

---

🧩 Architecture: 5 tiny pieces

1. A global registry

Just an array of test definitions.

const tests = [];
export function test(name, fn) { tests.push({ name, fn, skip: false }); }
export function testSkip(name, fn) { tests.push({ name, fn, skip: true }); }
export function testOnly(name, fn) { tests.push({ name, fn, only: true }); }

No classes, no plugins, no lifecycle hooks.

---

2. A runner that:

• loads test files
• runs them in parallel using Promise.all
• captures logs by temporarily overriding console.log
• prints logs only on failure
• respects --only, --grep, --verbose, etc.

This is the entire “engine.”

---

3. A CLI that:

• accepts flags via process.argv
• passes them to the runner
• prints a summary

No custom argument parser—just process.argv.slice(2).

---

4. A tiny reporter

Just prints:

✓ name
✗ name
    error message
    captured logs

---

5. Zero magic test discovery

You explicitly pass files:

npx tinytest tests/**/*.test.js

Or define a default glob in package.json.

---

🧪 The MVP implementation (readable, minimal, complete)

Here’s the whole thing, conceptually:

// tinytest.js
import fs from "node:fs";
import path from "node:path";

const tests = [];
export function test(name, fn) { tests.push({ name, fn }); }
export function testSkip(name, fn) { tests.push({ name, fn, skip: true }); }
export function testOnly(name, fn) { tests.push({ name, fn, only: true }); }

export async function run(files, flags) {
  // Load test files
  for (const file of files) {
    await import(path.resolve(file));
  }

  let runnable = tests;
  if (tests.some(t => t.only)) {
    runnable = tests.filter(t => t.only);
  }

  if (flags.grep) {
    runnable = runnable.filter(t => t.name.includes(flags.grep));
  }

  const results = await Promise.all(
    runnable.map(t => runOne(t, flags))
  );

  // Summary
  const passed = results.filter(r => r.ok).length;
  const failed = results.length - passed;

  console.log(`\n${passed} passed, ${failed} failed`);
  process.exit(failed ? 1 : 0);
}

async function runOne(t, flags) {
  if (t.skip) {
    console.log(`- ${t.name} (skipped)`);
    return { ok: true };
  }

  const logs = [];
  const originalLog = console.log;
  console.log = (...args) => logs.push(args.join(" "));

  let ok = true;
  let error = null;

  try {
    await t.fn();
  } catch (e) {
    ok = false;
    error = e;
  }

  console.log = originalLog;

  if (ok) {
    console.log(`✓ ${t.name}`);
  } else {
    console.log(`✗ ${t.name}`);
    console.log("   " + error);
    if (!flags.quietLogs) {
      for (const line of logs) {
        console.log("   log:", line);
      }
    }
  }

  if (flags.verbose) {
    for (const line of logs) {
      console.log("   log:", line);
    }
  }

  return { ok, error };
}

And the CLI:

// tinytest-cli.js
import { run } from "./tinytest.js";

const args = process.argv.slice(2);
const files = args.filter(a => !a.startsWith("--"));

const flags = {
  verbose: args.includes("--verbose"),
  quietLogs: args.includes("--quiet-logs"),
  grep: extractFlag("--grep="),
};

function extractFlag(prefix) {
  const arg = args.find(a => a.startsWith(prefix));
  return arg ? arg.slice(prefix.length) : null;
}

run(files, flags);

---

🎯 Why this is the true MVP

1. No reinvention

• Uses Node’s module loader
• Uses Node’s console
• Uses Node’s parallelism (Promise.all)
• Uses Node’s CLI args

2. Predictable

• No magic file discovery
• No hidden reporters
• No sandboxing
• No plugin system

3. Debuggable

• You can put console.log anywhere
• You can attach a debugger to the CLI
• You can print internal state easily

4. Extensible without becoming a monster

You can add:

• timeouts
• snapshots
• watch mode
• coverage (via c8)
• reporters
• test isolation

…all without rewriting the universe.

---

🧠 The deeper point

This is what Vitest should have been: a thin wrapper over Node, not a parallel reality.

You can build this in a day, and it will behave more predictably than Jest, Vitest, Mocha, or Ava.