STL/ranges algorithm to calculate weighted average

Question

Assume I have a vector of grades, where the grade is

struct Grade{
   const int grade;
   const int ECTS; // weight
};

Is there a STL/range-v3 algorithm/algorithms that enable me to do this?

I know I could do it with std:: accumulate with some fancy type as an accumulator(that remembers the sum of weights), but I am looking for a simpler alternative if one exists.

Constrained numeric algorithms are currently being developed: [P1813](http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2019/p1813r0.pdf). They'll enable something like: `ranges::inner_product(grades, grades, 0, {}, {}, &Grade::grade, &Grade::ECTS);` — metalfox, Aug 30 '19 at 07:29

score 17 · Answer 1 · edited Sep 03 '19 at 11:38

17

The Grade type itself is fancy enough to act as the accumulator type.

auto [grade_sum, ects] = std::accumulate(
    grages.begin(), grades.end(), Grade {0,0}, 
    [] (Grade acc, Grade g) -> Grade {
        return { g.grade*g.ECTS + acc.grade,
                 g.ECTS         + acc.ECTS  };
});
// auto average_grade = grade_sum/ects;

C++17 structured binding can be replaced by std::tie if necessary.

edited Sep 03 '19 at 11:38

Jarod42

203,559
14
181
302

answered Sep 03 '19 at 09:45

Tom

3,281
26
33

1

Probably `auto average_grade = double(grade_sum) / ects`. – Jarod42 Sep 03 '19 at 11:36
That is is a possibility indeed. Thanks for fixing the accumulator/value order by the way ;) – Tom Sep 03 '19 at 11:57

score 7 · Answer 2 · answered Sep 03 '19 at 12:01

7

With range-v3, it might be:

auto average = ranges::inner_product(grades, grades, 0, {}, {}, &Grade::grade, &Grade::ECTS)
        / double(ranges::accumulate(grades, 0, {}, &Grade::ECTS));

Demo

answered Sep 03 '19 at 12:01

Jarod42

203,559
14
181
302

1

Currently, the compilation time is a show-stopper. https://godbolt.org/z/I1_kGh vs https://godbolt.org/z/eVvMJz or 14448ms (ranges::inner_product) vs 3890ms (std::accumulate) – Porsche9II Sep 16 '19 at 15:48

STL/ranges algorithm to calculate weighted average

2 Answers2