search upper_bound by function of element

Question

Lets say a have a class A that can be evaluated to an integer.

We get as input a vector of A that is sorted by it's evaluate() value.

How can I find the upper bound by the evaluated value?

I tried this but it doesn't compile.

class A;
int evaluate(const A& a);
// ...
vector<A>::iterator foo(vector<A>& v, int k)
{
    return upper_bound(v.begin(), v.end(), k,
        [](const A& a, int k)
        {
            return evaluate(a) < k;
        }
    );
}

Answer 1

While other users have correctly pointed out the problem (eg "You got the order of the arguments in the lambda wrong" by @StoryTeller), please allow me to put down a piece of code that compiles on my machine, for other viewers' reference.

#include <algorithm>
#include <iostream>
#include <vector>

class A
{
public:
    int data;
    A(int data) : data(data) {}
};

int evaluate(const A & a)
{
    return a.data;
}

std::vector<A>::iterator foo(std::vector<A> & v, int k)
{
    return std::upper_bound(v.begin(), v.end(), k,
           [](int k, A & a) { return evaluate(a) > k; });   // first argument always val
}

int main()
{
    std::vector<A> vec {36, 42, 57};
    std::cout << foo(vec, 42) - vec.begin() << std::endl;   // 2
}

Answer 2

According to [upper.bound] in the C++17 standard:

... the following corresponding conditions hold: ... or comp(value, *j) == false .

That means you will get calls to comp (your predicate) with the first argument being the value you passed; and the second argument the dereferenced iterator.

In other words, you likely want your predicate to have type:

bool (int, const A&)