Retrieve std::meta::info of an std::meta::info?

[krzikalla]

I wonder whether this is the right place to ask questions about your reflections approach. Sorry, if this is the wrong place. Anyway... I am trying to create a struct, which mirrors the non-static data members of another struct. Those reflected data member have special types. Such a type should allow me to access the original data member, when it gets an object to the original struct. You find my experiments here:

https://godbolt.org/z/bEvr1eW9x

My code always fails in substitute. IMHO the most promising approach is

template<class T, std::meta::info MemberInfo>
struct value_expression
{
};
//...
  for (std::meta::info member : old_members) {
    auto mem_type = substitute(^value_expression, {type, member});
    //...
  }

But substitute "dereflects" member expression. Writing ^member in the substitute call doesn't help: error: non-type template argument is not a constant expression. Do I have a chance to reach my goal with the current reflection implementation?

[Yaraslaut]

Hi, Here is a minimal example of copying data members https://godbolt.org/z/17rWE5MxG

[krzikalla]

Thanks for the quick reply. Thus the trick is to use expand and a templatized create_type function instead of the iterative approach shown in example §3.10 (https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2024/p2996r5.html#struct-to-struct-of-arrays). Now it works:

https://godbolt.org/z/zvbPn3b5e

[Yaraslaut]

Great that you managed to fix your code :) I am still puzzled about what the code does exactly though xD

[krzikalla]

It mimics the data member access operator (operator.). Eventually the expression data[i].x will return a std::simd value. For more details you might refer to my cppcon24 talk (github doesn't let me send private messages, but you will find it, my name is rather specific).

[Yaraslaut]

Oh, so this is a piece of puzzle i was missing to see a reason to have a layer of misdirection in data access, will definitely have a look at your talk once it will be online :) Will be happy to look at the final result, as well. Also, are you considering different memory layouts for aligned loading inside simd?

[katzdm]

@krzikalla Sounds like you figured out how to achieve what you're trying to achieve - Is there still an issue, or should we close this one?

[krzikalla]

Can be closed, thank you.