there was a bug in the random tests, due to careless division that could yield a NaN when the length of the array is 0. can you check if it is good for you know ?

Thanks, looking back it's most probably too clever. Negative detection could simply be xneg = (x < 0); more readable and code generation would be same or better because it wouldn't confuse optimizers about the intent.

Despite its name, the C standard does not mandate that implementations of qsort() use the quicksort algorithm. In fact, it does not even require a O(n log n) complexity for qsort(), though of course most clibs will make it as fast as possible because it is an important function. So you cannoy rely on its stability.

That being said, it seems like the glibc standard library (the one used on Codewars, which runs on Linux) uses mergesort in most cases; and "most merge sort implementations produce a stable sort". Do not rely on it though, it is an implementation detail. If you need a stable sort, make it stable yourself, e.g. by comparing the pointers themselves in case of structural equality.

This may give you some insights

Merged recently on C standard

Thank you! These days I wouldn't even abbreviate the loop counters, I'd write day and week instead of d and w. This solution must be almost 10 years old. I have absolutely no memory of doing this kata.

.