This repository contains a never-valueless C++14 implementation of std::variant in variant.hpp and an implementation of expected<T, E...> in expected.hpp that is an extended version of expected<T, E> as proposed in P0323R1 and the subsequent D0323R2.
The code requires mp11 and Boost.Config. The repository is intended to be placed into the libs/variant2 directory of a Boost clone or release, with mp11 in libs/mp11, but the headers will also work standalone if mp11.hpp or mp11_single.hpp is included beforehand.
Supported compilers:
- g++ 5 or later with -std=c++14 or -std=c++1z
- clang++ 3.5 or later with -std=c++14 or -std=c++1z
- Visual Studio 2017
Tested on Travis and Appveyor.
The class boost::variant2::variant<T...> is an almost conforming implementation of std::variant with the following differences:
- A converting constructor from, e.g.
variant<int, float>tovariant<float, double, int>is provided as an extension; - The reverse operation, going from
variant<float, double, int>tovariant<int, float>is provided as the member functionsubset<U...>. (This operation can throw if the current state of the variant cannot be represented.)
To avoid going into a valueless-by-exception state, this implementation falls back to using double storage unless
- all the contained types are nothrow move constructible, or
- the first alternative is the type
valueless.
If the second bullet doesn't hold, but the first does, the variant uses single storage, but emplace constructs a temporary and moves it into place if the construction of the object can throw. In case this is undesirable, one can force emplace into always constructing in-place by adding valueless as a first alternative.
The class boost::variant2::expected<T, E...> represents the return type of an operation that may potentially fail. It contains either the expected result of type T, or a reason for the failure, of one of the error types in E.... Internally, this is stored as variant<T, E...>.
See its documentation for more information.