maybe.cpp

// how I'd "fix" https://akrzemi1.wordpress.com/2017/01/02/not-detecting-bugs/
// 
// * Explicitly state "code assumes input vectors are same length"
// * Explicitly request inlining of loop's inner funcion
// * real world code will probably not use "v1, v2" naming
// * instead of relying "compiler will optimize away helper constructs", 
//   just don't use any.
//
// in general, it's a similar case, when instead of ...
//    struct x = {a[++i], b[i]}
//
// you write
//    i++;
//    struct x = {a[i], b[i]}
//
// .. and skip the guesswork of what the compiler will choose to do.
#include <vector>

inline void process(const int sourceId, const int destinationId) {
}

int main() {
  std::vector<int> sourceIndexes = {1, 2, 3, 4, 0};
  std::vector<int> destinationIndexes = {2, 3, 5, 7, 0};

  // verify assumptions in debug
  assert(sourceIndexes.size()==destinationIndexes.size());
  
  {
    auto currentSource = sourceIndexes.begin();
    const auto end = sourceIndexes.end();
    auto currentDestination = destinationIndexes.end();

    for (;currentSource!=end;currentSource++, currentDestination++) {
      process(currentSource,currentDestination);
    }
  }
}