Zig vs D generics

On Sunday, 9 October 2022 at 21:48:38 UTC, Walter Bright wrote: > https://news.ycombinator.com/item?id=33142751 Can we create a real derived structure with templates at compile time? Still, I remembered that there was such a thing as a static structure. What can the Zig programming language add to D? SDB@79

>In contrast, C++ code attempting to call non-constexpr functions (i.e. functions only accessible at run time) in a constexpr context (i.e. functions which are available both at compile time and run time) will give an error at the call site. Eh? So can C++ support an allocator that works both at compile time and at run time? I use such allocator in my D code so it's ctfeable.

On Monday, 10 October 2022 at 07:30:38 UTC, Kagamin wrote: >>In contrast, C++ code attempting to call non-constexpr functions (i.e. functions only accessible at run time) in a constexpr context (i.e. functions which are available both at compile time and run time) will give an error at the call site. > > Eh? So can C++ support an allocator that works both at compile time and at run time? I use such allocator in my D code so it's ctfeable. Depends on how much you expect from it, featurewise, https://www.cppstories.com/2021/constexpr-new-cpp20

On Monday, 10 October 2022 at 07:30:38 UTC, Kagamin wrote: > Eh? So can C++ support an allocator that works both at compile time and at run time? I use such allocator in my D code so it's ctfeable. No, only encapsulated types can hold memory. This is the right approach when you don't have garbage collection. If you allow manual compile time allocation you end up with something that doesn't scale in terms of debugging. Or you need a very advanced debugger.

October 10, 2022

Re: Zig vs D generics

Posted by Paulo Pinto
in reply to Ola Fosheim Grøstad

Permalink

Paulo Pinto

Posted in reply to Ola Fosheim Grøstad

Permalink

On Monday, 10 October 2022 at 10:23:44 UTC, Ola Fosheim Grøstad wrote:

On Monday, 10 October 2022 at 09:55:32 UTC, Ola Fosheim Grøstad wrote:

On Monday, 10 October 2022 at 07:30:38 UTC, Kagamin wrote:

Eh? So can C++ support an allocator that works both at compile time and at run time? I use such allocator in my D code so it's ctfeable.

No, only encapsulated types can hold memory. This is the right approach when you don't have garbage collection. If you allow manual compile time allocation you end up with something that doesn't scale in terms of debugging.

Or you need a very advanced debugger.

Argh, I couldn't make this work either in C++20. I thought std::string("hello world) should be constexpr returnable. Apparently not, unless I did something wrong. (That is a big weakness, so I really hope I did something wrong. ;-)

I guess you did something wrong, :)

https://godbolt.org/z/qKPGvbd1W

#include <string>
#include <iostream>

constexpr std::string hello() {
    return std::string("hello world");
}

int main() {
    static constexpr auto greeting = hello();

    std::cout << greeting << std::endl;
}

#include <string> #include <iostream> constexpr std::string hello() { return std::string("hello world"); } int main() { static constexpr auto greeting = hello(); std::cout << greeting << std::endl; }

Forums