Fair enough, I'll try to replicate the python tests in js when I can.

ahhhumm.. js? I don't have the js-fu to sniff out incorrect solutions at a glance especially with all the odd things js does and the existing tests are nearly non-existant considering that the kata sets out to mix types. its reference solution is using isNaN to do type checking which makes me want to send them a strongly worded dead tree letter. no I don't want to touch it.

it's still easy to find slightly incorrect python solutions (especially if willing to go for pathological test cases) but mixing types is smelly anyway. I scratched my itch.

Approved, could you take a look at the two issues below too? I think you already addressed the second one.

Just the same, the "should be" wording is poor presentation and potentially misleading. The beginner solver doesn't know what to expect and has other things on their mind. A 8 kyu kata should not behave this way. (or any kyu really but whatever)

So I'm fixing the root cause of this and a couple other blemishes and some tests that I think were missing although I'm sure there are some more stuff to test if one sifts through solutions to find some wrong ones. Please approve my python fork

Not a kata issue, print the input values to debug your code.