Version 3.0.0 (planned breaking changes and architecture improvements)

[wellwelwel]

In favor of better maintenance, performance and a more cohesive architecture, I'm planning to make some major updates and migrate the opening issues to the next Poku version.

📦 Installing

# Node.js
npm i -D poku

# Node.js + TypeScript
npm i -D poku tsx

# Bun
bun add poku

# Deno (optional)
deno add npm:poku # use "deno run npm:poku" to run tests

---

⚙️ Migrating

• For Node.js, make sure you have version 14.x.x or higher.
• Instead of poku --parallel or poku -p, just use poku.
• Instead of poku (sequential in v2), use poku --sequential.
• If you have some flag using kebab case (e.g., --fail-fast), use camel case instead (e.g., --failFast).
• Instead of setting the platform manually (poku --platform=..., poku --deno, poku --bun or poku --node), just use the platform/runtime to call Poku, for example: npx poku, bun poku or deno run poku.

---

🗺️ Roadmap

✅ v3.0.0-rc.0

Focused on performance and breaking changes:

• Drop the support for Node.js 8, 10 and 12 (for performance reasons).
  • #885
  • #896
• Remove all methods and options marked as deprecated.
  • #877
  • #883
• Improve performance.
  • #886
  • Investigate the use of fork instead of spawn:
    • Benchmarks were all negative against fork.
• Use parallel run as default, such as Node.js, Jest, etc.
  • #888
• Remove the need to manually define the platform for environments with multiple runtimes (--platform, --node, --bun, --deno).
  • #894

✅ v3.0.0-rc.1

Note

From this point on, there is no intention of breaking changes, except for some inevitable case.

Focused on improvements and major features:

• Revisit early access and experimental features.
  • #901
  • #905
• Improve the global cache to increase initial performance.
  • #907

⏳ v3.0.0 (fresh out of the oven)

• Improve architecture to allow external plugins.
  • Investigate the use of events approach:
    • Benchmarks were all negative against events (more than 2x slower).
  • #921

Note

I'm extremely satisfied with the current benchmarks (#926). I'll wait to write the extensive report after v3.