Improvement: cached for loops

[mster]

Hello there!

I was looking into speeding up native functions and came across your package. Awesome work so far. I wanted to inquire about some for loops you're using, for example: https://github.com/kb-dev/sanic.js/blob/master/lib/array/map.js#L30

Would caching the length values help you improve performance even further? I got the idea from your article: https://medium.com/kbdev/voyage-to-the-most-efficient-loop-in-nodejs-and-a-bit-js-5961d4524c2e

If you think so, I would love to open up a PR with the improvement. Cheers!

[AamuLumi]

Hey @mster!

Firstly, thanks a lot for your message! :)

About the length caching, as written in the edit of this article, it's now kind of a useless hint. Evolution of v8 engine (especially 5.9 if my memory is good) leads to an entire change of the optimization engine by introducing Turbofan compiler which fixed the trick.

So in the case of WebKit-based browsers and Node.js, we don't need to do this. But I don't try on Gecko-based browsers with SpiderMonkey engine like Firefox. Maybe you can try the hint on Firefox with some JS browser benchmark. I can't tell you which tool you should use because the good old jsperf is now offline.

I'm thinking about something bigger like memoization, but we will need to transform all structures as immutables structures. And we will lose the fact that you just have to import the library to be used.

Anyway, I'll be happy to see a positive result for the test on Firefox. And I stay open to any new ideas to push this library forward! :D

[mster]

@AamuLumi oof! I must've missed your edit. Good thing v8 is getting smarter every day.

Maybe you can try the hint on Firefox with some JS browser benchmark.

I'll check out the length caching on FireFox 👍

I'm thinking about something bigger like memoization, but we will need to transform all structures as immutables structures. And we will lose the fact that you just have to import the library to be used.

The fact that you only need to import sanic.js to use the library is really slick. It would be to experiment with memoization in a similar context.

Thanks for getting back to me so quick and taking the time to provide a solid explanation. I really appreciate it. :) If I come up with anything worth showing, I'll report back.