cute

Forked and approved to resolve MC, thanks!

The function failed! f(x) = 24.3+(83.3/(69.3*(x^x*x^42.3-9.7^76.5/87.9^x*92.5^2.5)^x/log(x*4.7^x/49.3)^28.4-21.2)^x/x), x = (-4.69,8.75)

Expected: equal to (24.3,7.17589e-06) (+/- (0.0078125,0.001))

Actual: (1.03767e+18,-1.20504e+19)

wolframalpha gives me result: 1.03767×10^18 - 1.20504×10^19 i

It's hard to find solution, including mine, which will have chance to pass random tests in first ten attmpts.

Such eh masterpiece

I did what you ask. No need for a reference solution

Also, I increased the number of test to 500, so the probability of getting every length from 1 to 50 is ~99.8% (before it was ~0.02%).

Everything looks great to me, but I suggest including <cstddef> since you use std::size_t. On another note, I just think it's a bit of a missed opportunity that vectors of size == 1 are not included in random tests. I think this might be a case where using a reference solution instead of pre-generating answers can simplify the tests and allow for a bit more coverage. What do you think?

It's 7 kyu kata, I don't think there's a need to make it more complicated than it already is

Is this a question? The short answer is probably "no".

The long answer is that tests are specifically written to avoid 64 bit overflow, so the answer is probably "no".

You are free to do or return whatever you want inside (and sometimes even outside) of the function, as long as the function signature stays the same

Do your magic, wizard :^)

Все тесты корректные, внимательно читай условие задачи, в частности, последний пункт.

approve his fork

approve his fork
@o2001 ty

Ah, okay.