I have some comments:

1. No need to include dist/* to PR
2. Tests needed (with duplicated nested keys too)
3. storeMappingPair is called not with different sets of params. Need to check benchmarks before and after changes.

@puzrin I added tests! 😎

Can you elaborate on what you mean in your third point?

Can you elaborate on what you mean in your third point?

JIT can deoptimize functions if they are not monomorphic. Here is good reading http://mrale.ph/.

So, it worth to check, if you change did not slowed down anything. Run ./benchmark/benchmark.js.

If you see that things become bad, try to add , 0, 0 to end of all storeMappingPair calls.

Hey @puzrin, thanks for clarifying.

I ran the benchmark tests before and after my commits, here's the results: https://www.diffchecker.com/Y3Idvbal

Looks similar. Difference can be caused by noise.

Can i ask you do one more thing? Could you change other calls to storeMappingPair(..., 0, 0) and compare benchmarks again? I don't like such change, because it looks dirty, but speed make sense. Make sure to close other programs and run twice to check if difference is real or just a noise.

PS. Please, squash your commits into one, i think this PR is valuable enougth and my fantasies about speed should not delay merge :)

Yes! I'll change the calls to storeMappingPair(..., null, null) instead, because the logic for falling back to what's in state needs a falsy value to kick in.

Edit: here's the logic I'm taking about

Here's the benchmarks where the number of arguments is constant across the library:

https://www.diffchecker.com/8UMWNP3m

Also, GitHub now has a way to squash pull requests when you merge on GitHub. Maybe that will help?

Otherwise, let me know and I can rebase my branch 😎

Here's the benchmarks where the number of arguments is constant

Equal number of arguments is not enougth. Those should be of the same type in all calls, and typeof null !== typeof 0.

As alternative, i'd suggest to pass (..., -1, -1) and change condition to (if >= 0)

You can just squash commits and i'll merge result as is. It's ok for me. I can test benchmarks myself to save your time.

Otherwise, let me know and I can rebase my branch

I tried to use github's squash somewhere and don't like garbaged comments it does. So, prefer manual rebase. Please, do it, if not difficult :)

Done 🎉

Do you mind doing an npm publish once this gets merged? We'd love to start using this fix ASAP

LOL. Rebased commit message contains lines i wished to avoid :). I mean second one and below (github adds the same on squash). Could you remove those and leave only first one?

Do you mind doing an npm publish once this gets merged? We'd love to start using this fix ASAP

If that's not deadly critical, i'd like to take a pause for 2 days (not more).

^ There's a one-line commit message 👍

Published. Thanks for PR!