Using ()-Operator on Templates

Question

I wasn't able to find an answer by googling, since I don't know what to actually look for. So I stumbled across this:

template<typename T>
class SkipList {
    public:
        SkipList() : max_level_(16) { 
            head = new Node(T(), max_level_);
    ...
}

and I don't really know what T() means. I'm familiar with the concept of templates (at least I have basic knowledge about it), but I've never seen using the ()-Operator on them. So hat is it exactly doing?

`T` is a **type**. And a [good book](https://stackoverflow.com/questions/388242/the-definitive-c-book-guide-and-list) should cover initialization in depth. — StoryTeller - Unslander Monica, Mar 31 '18 at 14:32
`T` is a type. `T()` default-initialises a temporary object of type `T`. In the code you show, that object is passed as the first argument to a constructor of `Node`. — Peter, Mar 31 '18 at 14:45
@liliscent technically, it is doing [value-initialization](http://en.cppreference.com/w/cpp/language/value_initialization), which is related, but different. — rubenvb, Mar 31 '18 at 15:49
@liliscent: Only indirectly. There is no syntax in C++ to "call the constructor" — Lightness Races in Orbit, Mar 31 '18 at 18:06
@LightnessRacesinOrbit Agree, and it's indeed an annoying feature. It would be more convenient sometimes if constructor could be treated as a static function. — llllllllll, Mar 31 '18 at 18:41
@liliscent: I don't find it "annoying", I find it perfectly reasonable and logical. I see no use for the constructor to be treated like a static member function. Could you describe one? — Lightness Races in Orbit, Mar 31 '18 at 18:41
@LightnessRacesinOrbit Sometimes you want to take its address to pass it to a high order function, maybe do some SFINAE-like checking. — llllllllll, Mar 31 '18 at 18:43
@LightnessRacesinOrbit `std::declval` is not related to construction, it gives you a fake rvalue reference... Using a generic lambda to wrap the constructor is fine. — llllllllll, Mar 31 '18 at 18:48

Hawkeye5450 · Accepted Answer · 2018-03-31T18:05:20.323

3

What that is doing is creating a default object of that type. Depending on the object, it will mean different things, but what it comes down to is that when you want something there, but don't particularly care what it is, you can use a default object to get an object with generic values. For int and double it should be 0, but for more complex types like string, it becomes things like the empty string or all attributes are set to 0. This is part of the reason to have a default constructor in classes.

std::cout << int() << " " << std::string() << " " << double();

outputs:

0  0

edited Mar 31 '18 at 18:05

answered Mar 31 '18 at 14:51

Hawkeye5450

672
6
18

3

There is no "null string" in C++. Creating a `std::string` from `nullptr` (or `NULL`, or `0`) is undefined behaviour. You mean an *empty* string, which is something very different. – Christian Hackl Mar 31 '18 at 15:34

Christian Hackl · Answer 2 · 2018-03-31T15:42:52.617

int() creates a default int (zero).

using Ptr = void*; Ptr() creates a default void* (a nullptr).

std::string() creates a default std::string (an empty string).

Do you see the pattern?

T() creates a default T and works for every type for which objects can be created without passing any arguments. If T is int, then you get a zero int. If T is void*, then you get a nullptr. If T is std::string, then you get an empty string. And so on.

Using ()-Operator on Templates

2 Answers2