freaking genius

I couldn't, so it seems likely :P

no one solved it in JS yet, so chances are the current JS sol is also wrong.
If you are sure about your solution, paste it here as spoiler and I will change the ref sol.

Oh sorry, didn't notice that.

But going off my results with the published version, that's still wrong?

Note I haven't yet forfeited the kata in JS, so can't comment on the actual code.

This is the old ref sol, I already updated that one in a later fork.

Reference solution is wrong. When comparing to Python and Haskell, n = n/3n should be n = (n+1n) / 3n.

When calculating expected values back to numbers, they don't match with inputs. I'm using the below code for that, which I think is correct.

function from(s) { return Array.from(s).reduceRight( (z,x) => 3 * z + "0+".indexOf(x) , 0 ); }

( Note that this abuses "0+".indexOf('-') == -1 )

Node 12 should be enabled

Missing assertion messages

It is better to use let instead of var to avoid closure issues

Preloaded should be completely empty

Didn't we have JS already?

Why did you write this in a legacy Node / test framework version?

Thank you for your translation, @Pnd72!

Done. (for some reason, sometimes messages just dissappear and reappear on tab reload in browser, so I kinda doubled the message)

Done.

That's the first solution I googled up (BigInt in random interval). It's basically an npm-package, so there should be no major problems with maintaining https://www.npmjs.com/package/random-bigint

Is the cryptographical randomness really necessary? It makes maintaining the code much harder, and it doesn't really seem to have a purpose here.

I like the redoing of the description, but code blocks should be tagged python and not if:python ( that's for normal text blocks ). Could you fix that?