Vector - On demand vector type

Vector is a single-header library that provide math vectors to C++.

Vectors can be structural

By providing you own storage type, you can access a vector's elements easily:

template<class T, std::size_t N>
    requires (N == 3)
struct vec3_storage
{
    union {
        T x, y, z;
        T data[N];
    };

    constexpr decltype(auto) operator[](this auto&& self, std::size_t i) {
        return self.data[i];
    }
};

// VectorFeatures::extend_storage is enabled by default!
using v3f = vector::Vec<float, 3, vec3_storage>;

v3f u{7, 3.14, 0.0};
assert(u[1] == u.y);

In case you disable extend_storage, you can still access the storage using the _storage field.

Vectors are tuple like

Vectors support tuple-like operations, such as std::get, std::tuple_element, std::tuple_size and std::apply.

Vectors are array like

Vectors override the subscript vec[i] operator, so you can access their elements by index.

Vectors are iterable

Vectors provide their own contiguous iterators, accessible via Vector::iterator/Vector::const_iterator. They also have their reverse (only random access) equivalent: Vector::reverse_iterator/Vector::const_reverse_iterator. You can access them directly using std::begin/std::end, or simply vec.begin()/vec.end()

Building

Vector uses modern C++ features, it requires C++23 to compile. Vector was tested to work with GCC>=14.2 and Clang>=18.1.8. According to OpenMP documentation, OpenMP statements cannot appear in constexpr functions, however GCC allows it. Therefore the OpenMP features have to be turned off on Clang. However, the compiler is generally able to efficiently optimize the vector's iterations for -O>=1, so OpenMP doesn't make a difference.

Building the doc

Go to the docs/ directory and type doxygen default.

License

Vector is licensed under the MIT license.