I was playing around with toy tuple implementations and eventually stuck with how get
function works.
Consider this simple example
#include <iostream>
#include <utility>
template <size_t Tag, typename ValueType>
struct TagedValue { ValueType value; };
struct Test : TagedValue<0, int>, TagedValue<1, std::string>, TagedValue<2, double> {};
template <size_t Idx, typename T>
auto& get(Test& test) {
((TagedValue<Idx, T>&)(test)).value;
}
template <size_t Idx, typename T>
auto& get_impl(TagedValue<Idx, T>& tagged_value) {
return tagged_value.value;
}
template <size_t Idx>
auto& get_2(Test& test) {
return get_impl<Idx>(test);
}
int main()
{
Test test;
get_2<0>(test);
get<0>(test);
}
<source>: In function 'int main()':
<source>:29:16: error: no matching function for call to 'get<0>(Test&)'
29 | get<0>(test);
| ^
<source>:10:7: note: candidate: 'template<long unsigned int Idx, class T> auto& get(Test&)'
10 | auto& get(Test& test) {
| ^~~
<source>:10:7: note: template argument deduction/substitution failed:
<source>:29:16: note: couldn't deduce template parameter 'T'
29 | get<0>(test);
| ^
I do have couple of questions:
get_2
works and get
doesn't compile. To me it looks like get_2
does exactly what I'm trying to do inside get
get_2
take O(1) time, if yes how is it possible? Does compiler store some kind of map internally?When you call a function template then all template arguments must either be specified explicitly or be deduced from the function arguments. When you call
get<0>(test);
Then Idx
is 0
, but there is no way for the compiler to know what T
is supposed to be. The parameter is just Test
, and T
cannot be deduced from that.