The following code confuses me:
#include <iostream>
template <typename T>
struct base
{
T &self() { return *static_cast<T *>(this); }
template<typename T1> void method() {
std::cout << sizeof(T1) << "\n";
}
void hello() {
self().template method<int64_t>(); // <-- here
self().template method<int32_t>(); // well, and here too
}
};
struct derived : base<derived>
{
void hello2() {
self().method<int16_t>(); // seriously???
}
};
int main(int args, char ** argv) {
derived d;
d.hello();
}
This compiles and does what I want. But what is this "template" business on the lines highlighted by comments? What is this syntax even called? And why do i need it? And if I remove the word "template", at least with gcc I get errors:
member.cc: In member function ‘void base<T>::hello()’: member.cc:12:26: error: expected primary-expression before ‘>’ token
self().method<int64_t>();
^ member.cc:12:28: error: expected primary-expression before ‘)’ token
self().method<int64_t>();
^
And, last but not least, why does the method derived::hello2() compile just fine without the "template" word?
I should add that it clearly has something to do with base being templated. If i make it a non-templated class and get rid of the self acrobatics, i no longer need the "template" keyword.
