I am struggling with making a design choice in the following setup: I am writing (C++) functions which take a pair of iterators (to template containers) and compute a return value of the same type that the iterators are pointing to. Let's say I want to implement a sum function (this is just an example to illustrate the point). The way I see it I have two options and both have a downside: Option 1 Make the function template parameter T defining the type the container stores. I only need one parameter but I can only use the function on iterators of MyContainer .
template T sum(typename MyContainer::IteratorRange range) < T sum; for (auto it = range.first; it < range.second; ++it) < sum += *it; >return sum; > Option 2 Have the function take an arbitrary IteratorRange template class and a second class U determining the return type. While the iterators can be sourced from any container, I need two template classes even though the return type is always the same type as pointed to by the iterators.
template U sum(IteratorRange range) < U sum; for (auto it = range.first; it < range.second; ++it) < sum += *it; >return sum; > Which option is cleaner or is there an alternative offering the benefits of both options? Bonus Question How should one initialize the sum variable?