==== Question ====

As you might have noticed filter is our own implementation of the
function std::ranges::filter which was introduced in C++20. Discuss
advantages with this approach compared to the std::remove_if + erase
idiom. In particular you can focus on usability, readability or
performance.

==== Answer ====

With the std::remove_if + erase idiom and given a container (like
std::vector) we have to first populate the vector with values, and
then remove the ones that doesn't fulfill the predicate. This is
especially a problem from a performance perspective if we need to have
access to all the original values afterwards. Becuase in that case we
need to copy all the values first so that we don't remove them
completely, meaning we have to iterate the vector *at least* twice.

With our filter function we are calculating the values *lazily*
meaning we only check the predicate when we actually request a new
value. This has several benefits, one of them being that we do not
modify the original vector at all, but instead just loops over them
once and check which values fulfills the predicates and which doesn't.
This means that we don't have to spend any time actually copying or
removing values, which is better from a performance perspective.