There is a dummy Haskell version.

Please:

• use Test.assertDeepEquals
• get your datatypes straight
• actually specify everything that will be required ( in the description )
• write random tests before publishing
• you specify. user implements. do not require the helpers you use, unless you're building towards a climax in a guided tour of a much more difficult subject than this kata's

This list is not intended to be complete. But it would be a start.

Oh, and eval is not a good idea. Never. Nor is stringifying everything. ( Maybe just pass functions instead of strings? They're first class citizens, you know? )

Please, as usual, actually finish writing a kata before publish it. There's no point publishing a half-completed kata.

Even though it's faster to write the description all lower case it's far easier to read when it is written in proper case and when code is actually formatted as code by enclosing them in grave accent characters. Apart from that this really seems to be a nice kata. =)

If "global angles" means angles relative to some globally standard vector and not relative to the previous bone, please explain that better as well.

It might be well-defined jargon in some particular domain, but I'm not familiar with that domain either, and again, it leaves me guessing. ( And even if I guessed correct, I don't know the globally standard vector. :/ )

You should make a copy of the arrays before passing in to the users in the random tests. Otherwise the results will be different since we're both operating on and mutating the same array.

As you can now see, all previous solutions has been invalidated ;-)

• Needs Random Tests
• The description is really hard to understand (other explanation is not needed if you can explain it clearly)

Have a look at how the process is described on the page on Wiki, which is much clearer