boost-ext / mp Goto Github PK

C++20 was specified and tried with both GCC11.3 and 12.3, but the same error occurs.
Maybe I'm doing something wrong, please let me know.

Actual Behavior

root@monica003:/home/monica/test/build# cmake --build .
[ 50%] Building CXX object CMakeFiles/main.dir/test.cpp.o
In file included from /home/monica/test/test.cpp:3:
/home/monica/mp/installed/x64-linux/include/mp-0.0.1/boost/mp.hpp:208:7: error: expected identifier before ‘typeof’
  208 | using typeof = decltype(expr(std::declval<Ts>()...));
      |       ^~~~~~
/home/monica/mp/installed/x64-linux/include/mp-0.0.1/boost/mp.hpp:208:14: error: expected primary-expression before ‘=’ token
  208 | using typeof = decltype(expr(std::declval<Ts>()...));
      |              ^
/home/monica/mp/installed/x64-linux/include/mp-0.0.1/boost/mp.hpp:208:14: error: expected unqualified-id before ‘=’ token
gmake[2]: *** [CMakeFiles/main.dir/build.make:76: CMakeFiles/main.dir/test.cpp.o] Error 1
gmake[1]: *** [CMakeFiles/Makefile2:83: CMakeFiles/main.dir/all] Error 2
gmake: *** [Makefile:91: all] Error 2

Steps to Reproduce the Problem

microsoft/vcpkg#29975 (comment)

Specifications

• Version:

  root@monica003:/home/monica/test/build# lsb_release -a
  No LSB modules are available.
  Distributor ID: Ubuntu
  Description:    Ubuntu 22.04.1 LTS
  Release:        22.04
  Codename:       jammy
  

• Platform: Linux

Unfortunately, this only works currently with clang because of a bug. It's not supposed to (according to http://eel.is/c++draft/temp#deduct.general-9.sentence-1, particularly the Note below that)

great library, I found what I think is a little bug in the readme:

Expected Behavior

In the first godbolt link in the readme, this example compiles:

static_assert(mp::list<val const>() ==
  hello_world(mp::list<int, foo, val, bar>(), add_const, has_value)
);

Actual Behavior

However, in the readme itself, the example is written without the mp::list::operator(), which doesn't compile:

static_assert(mp::list<val const> ==
  hello_world(mp::list<int, foo, val, bar>, add_const, has_value)
);

Steps to Reproduce the Problem

1. Follow the godbolt link in the readme: https://godbolt.org/z/3T86zvcEn
2. Remove the two () operators after mp::list<...> to match the example as it is written in the readme
3. observe that it doesn't compile anymore

Specifications

same as in the original godbolt link

I think the solution is to add the () operators in the readme, too.

Is it possible to filter and reorder a parameter pack?

void foo(auto&&... args)
{
    qoo(std::forward_as_tuple(std::forward<decltype(args)>(args)...)
        | std::views::filter(???)
        | std::views::reverse);
}

Currently boost::mp::list is not usable in range based for.

So this is not compiles:

    for (char c : mp::list<__PRETTY_FUNCTION__>()) {
        std::cout << c;
    }

This can be enabled on types where the template arguments are the same.

Possible implementation

#include <https://raw.githubusercontent.com/boost-ext/mp/main/include/boost/mp.hpp>
namespace boost::mp {
template<auto...As>
constexpr static inline auto list_array = std::array{As...};

template<class ...>
constexpr bool is_same_all_v = true;
template<class Ref, class...Oth>
constexpr bool is_same_all_v<Ref, Oth...> = (std::is_same_v<Ref, Oth> && ...);

template<template <auto...> class A, auto...As>
    requires is_same_all_v<decltype(As)...>
auto begin(const A<As...>& v) {
    return std::begin(list_array<As...>);
}

template<template <auto...> class A, auto...As>
    requires is_same_all_v<decltype(As)...>
auto end(const A<As...>& v) {
    return std::end(list_array<As...>);
}
} // boost::mp

>------ Build All started: Project: CMakeFindUsage, Configuration: x64-Release ------
  MSBuild version 17.4.1+9a89d02ff for .NET Framework
    Checking Build System
    Building Custom Rule C:/Users/monica/source/repos/CMakeFindUsage/CMakeFindUsage/CMakeLists.txt
    CMakeFindUsage.cpp
E:\boost-ext-mp\installed\x64-windows\include\mp-0.0.1\boost\mp.hpp(171,13): error C3527: 'Cs' is not a valid operand for 'sizeof...'. Did you mean to use 'sizeof'? 
E:\boost-ext-mp\installed\x64-windows\include\mp-0.0.1\boost\mp.hpp(569,24): error C3861: '__PRETTY_FUNCTION__': identifier not found 
E:\boost-ext-mp\installed\x64-windows\include\mp-0.0.1\boost\mp.hpp(569,24): error C2065: '__PRETTY_FUNCTION__': undeclared identifier 
E:\boost-ext-mp\installed\x64-windows\include\mp-0.0.1\boost\mp.hpp(590,8): error C2065: '__PRETTY_FUNCTION__': undeclared identifier 
E:\boost-ext-mp\installed\x64-windows\include\mp-0.0.1\boost\mp.hpp(590,8): error C3861: '__PRETTY_FUNCTION__': identifier not found 
E:\boost-ext-mp\installed\x64-windows\include\mp-0.0.1\boost\mp.hpp(591,14): error C2065: '__PRETTY_FUNCTION__': undeclared identifier 
E:\boost-ext-mp\installed\x64-windows\include\mp-0.0.1\boost\mp.hpp(591,14): error C3861: '__PRETTY_FUNCTION__': identifier not found 
E:\boost-ext-mp\installed\x64-windows\include\mp-0.0.1\boost\mp.hpp(597,8): error C2065: '__PRETTY_FUNCTION__': undeclared identifier 
E:\boost-ext-mp\installed\x64-windows\include\mp-0.0.1\boost\mp.hpp(597,8): error C3861: '__PRETTY_FUNCTION__': identifier not found 
E:\boost-ext-mp\installed\x64-windows\include\mp-0.0.1\boost\mp.hpp(598,14): error C2065: '__PRETTY_FUNCTION__': undeclared identifier 
E:\boost-ext-mp\installed\x64-windows\include\mp-0.0.1\boost\mp.hpp(598,14): error C3861: '__PRETTY_FUNCTION__': identifier not found 

Build All failed.

For details, please se: microsoft/vcpkg#29975 (comment).

See https://github.com/ldionne/metabench.

This is so cool! A few random comments (or questions on what I'm missing if I'm wrong about these things):

I think this should be and here?

This (and several places above) can just use CTAD, right? [Answering my own question: you can't do CTAD because of the sizeof...(Ts) == 0 case 🤦🏼‍♀️]

Probably not relevant, but this implies that types like int and float* are callable

Unsolicited suggestion: while you're at it, I like using operator+ for concatenation (so that things like join<TLs...> can just be implemented as decltype(TLs{} + ... + type_list<>{})). Maybe not relevant here, but fun thought for later. Here's where I do this in my toy library cutempl: https://github.com/dhollman/cutempl/blob/5bb6cae4423f0456c21e50454bcba210a94226b3/include/cutempl/containers/type_list.hpp#L14-L21

I believe expr_fn doesn't need to be captured here since it's constexpr? (In fact, I think it's technically incorrect to do so, but the compiler is just okay with it?)

Also, shouldn't declval be a non-ADL call here since it could be ambiguous with std::declval if T is in std?

Seems like this one could just delegate to the previous overload? Something like value_list<Vs...> | fn | []<auto... Vvs> { return declval<T<R, Vvs...>>() }; (Obviously omitting some corner cases).

Seems like requires requires { []<class... Ts>(std::tuple<Ts...>){}(t()); } would be a more descriptive/self-documenting constraint here.

This deduction guide should be unnecessary in C++20 and later

Is there a reason to do this instead of using the comma operator? (Here and elsewhere; Also, if you do use the comma operator, you should do ((void)expr(Ts), ...) to work around overloaded comma operators)

Really nit-picky, but you should ignore digit separators in this loop.