Question

What is the difference between the code snippets labeled “version 1” and “version 2” in the following code section:

int main() {
  using namespace std;
  typedef istream_iterator<int> input;

  // version 1)
  //vector<int> v(input(cin), input());

  // version 2)
  input endofinput;
  vector<int> v(input(cin), endofinput);
}

As far as I understand “version 1” is treated as function declaration. But I don’t understand why nor what the arguments of the resulting function v with return type vector<int> are.

Answer 1

why

Because the Standard says, more or less, that anything that can possibly be interpreted as a function declaration will be, in any context, no matter what.

what the arguments… are

You might not believe this, but it’s true. input(cin) is treated as input cin; in this spot, parentheses are allowed and simply meaningless. However, input() is not treated as declaring a parameter of type input with no name; instead, it is a parameter of type input(*)(), i.e. a pointer to a function taking no arguments and returning an input. The (*) part is unnecessary in declaring the type, apparently. I guess for the same reason that the & is optional when you use a function name to initialize the function pointer…

Another way to get around this, taking advantage of the fact that we’re declaring the values separately anyway to justify skipping the typedef:

istream_iterator<int> start(cin), end;
vector<int> v(start, end);

Another way is to add parentheses in a way that isn’t allowed for function declarations:

vector<int> v((input(cin)), input());

For more information, Google “c++ most vexing parse”.