The absolute KOTY = Kata Of The Year !!

Tkx...
Note if I may - in c++ - (I counted and outputed the values for testing)
fixed_tests : 20
random_tests

• small_primes: 100
• small_composites: 100
• big_composites: 100
• primes_32_bits: 100
• semiprimes_64_bits: 75??

That could be in the description... (especially the fact that some 64 bits semiprimes will be tested IS actually important for the factorization algorithm...)

75 such tests

(C++)
Is it possible to get the number of tests of each category ?? Timing out @semiprimes_64_bits tests.
Note: I already think that it's an absolutely excellent kata, BTW.
Still not solved but ... keep trying :-)

Approved

LGTM! Approved. One minor quirk is that operations like (disb(gen)%(extra/2))+1 return a lower bound that is one-less than the JS version (ex: 1 here instead of 2). It infinitesimally increases the likelihood of integer divisions, but that seems harmless to me (since the finite division section in JS also generates integer divisions every now and then). If another reviewer finds this a deal breaker, it can be forked

OK thanks, saw that you authored a lot of katas...
Moves the ref solution to the private section (learnt a thing there), also implemented the other ideas.
Note: I'm not confortable (at all) with range v3 and fmt/... with C++17 clang (therefore pretty comfortable with c++20/23 stuff ) but with clang there is a hell of a lot of shi.. to include...
You #4 is indeed nice !

Very faithful translation, but I would like to point out one issue and offer 3 suggestions:

1. Issue: referenceSolution being a free public function means it is visible to the solver, who can invoke your function to pass the tests (after a simple forward declaration). Consider moving it into the private section of the Describe struct
2. I think it'd be beneficial change the body of the test function to this so the solvers receive error feedback:

   Assert::That(fractionToPeriodic(n,d), Equals(s), ExtraMessage(fmt::format("Incorrect result for fractionToPeriodic({}, {}):", n, d)));
   

3. Some header inclusions are missing (<cmath>, <utility>)
4. The entire code with stringstream and substr-ing out the zeroes can be replaced with a clearer one-liner

   const auto stringified = fmt::format("{}", static_cast<double>(n) / d);
   

   The library handles the zeroes for you, and returns a result identical to what you've hand-rolled

Everything else LGTM
(#include <fmt/core.h> assumed for #2 and #4)

:-)

Haha! You know it is possible, but your cached bitset works better. Bravo!

C++ translation waiting for comments or ... approval :-)

It did when I posted it :)

Can't figure out why the description says and you must verify whether any pairs of primes satisfy the condition. ?
IMHO this should be removed.

Somewhere in the description we may find the concept of unordered pair or pair set , as it is not clear wether you expect ordered or unordered pairs.
See https://en.wikipedia.org/wiki/Ordered_pair or https://en.wikipedia.org/wiki/Unordered_pair