C++ version. Contrary to the description, negative arguments are fed to the function.
The expected results for negative fractions look strange:
Incorrect result for fractionToPeriodic(-1689246217, 36291): Expected: equal to "-46547.(-2-4-90-9-7-5-7-2-400-...)"

Suddenly it looks so simple.

Should it be mentioned that the function is pure or am I missing something?

I get a SIGABRT with my code. Surely, my code is wrong but I do not allocate any memory so how is it possible?

Besides, I'm seeing

Expected: "*&ikh)ihg&j'kj%&hij'-&'*ih)h'"

which should not be possible ("Every String consists of only lowercase letters and spaces.")

Using the very big hammer, nice.

Ruby version: For "" (empty string) the expected return is false although all gs are happy. Is that correct?

It's always good to have a π along ;-)

indeed

Thanks, that was fast.

Haskell version

The code template says "balanced :: [Bool] -> Bool" but the required type is
"balanced :: [Word] -> Bool".

Also it is unclear why tube sizes are passed if they are all the same (or 0).

This also groups repeated digits in the integral part.

repeating_fractions(221, 1) => "(2)1.0"

The regexp should probably be /(\d)(\1+)(?!..)/.

Finally a solution in O(n).

Ruby version: Had to require 'set' in order to make the test framework work.

That's disappointing indeed. The fitness Proc expects an integer rather than a binary string. Call it with fitness.call(c.to_i(2)).

This can generate a small subset of the valid strings only.