Just thinking that "loop the same long string 26 times" already is O(n).

Not sure if I agree. The case satement referencing each index in each when is much more readable and more immediately understandable.

Beautiful! It's calling 'placeholder'?

Yeah absolutely sexy, this also prevents nasty user input because it enforces the use of a certain datatype. Thanks for the upload!

This is just beautiful to look at. My personal favourite!

Cool, thanks!

the language versions are inconsistent with each other

Yes. I let the kata as it is...

Well, either way the language versions are inconsistent with each other, so you'll have to decide whether to enforce it or not.

It's kind of a work but if you read at the very bottom of the page you will find a post by @ZozoFouchtra who made the JS translation:

I think initial kata assumed a1 contains unique search strings. 
That's why author of the kata didn't initialy test the ["arp", "arp"] case.

One of the translators added this constraint (checking a1).
Does this mean all previous translations should be now modified and several passed solutions invalidated ?

This is why this is not tested (the constraint was added from the HS translation).

Beware: r must be without duplicates.

At least in JS this is not tested and even the reference solution does not handle them correctly.

I will look at it but it is a suggestion, not an issue.

What is the result of solution("foo#bar", ["#"]) supposed to be?
I would expect that it should be "foo".

But there are solutions of the kata, where the result is "foo#bar",
so a test might be missing to enforce this.

The final tests dont catch it if you forget to implement zero.