I have a std::variant
of possibly repeated types, and I would like to increment the active index of the variant.
Here is an example demonstrating what I would like to do:
template <typename... Ts>
bool update_variant(std::variant<Ts>& v, std::tuple<work_generator<Ts>>& gen) {
return visit_with_index([&]<size_t Index /* the active index */>(auto& e) {
if (e.done()) {
// we are done with e
constexpr size_t next_index = Index + 1;
if constexpr (next_index == sizeof...(Ts)) {
// no more other work to do
return false;
} else {
// still have other work to do so we move on to the next variant alternative
v.emplace(
std::in_place_index_t<next_index>{},
std::get<next_index>(gen).create_worker()
);
return true;
}
} else {
// do some work on e, so e gets closer to the done state
e.do_some_work();
return true;
}
}, v);
}
To implement this, I seem to need something akin to visit_with_index
above, which gives me the current index as a compile-time value.
Other than writing my own version of visit_with_index
above, which is effectively writing my own version of std::visit
, are there any simpler ways to achieve what I want without performing more than one lookup by index?
I would use visit
on the index "directly".
First
template <typename... Ts>
auto get_index_variant(const std::variant<Ts...>& v)
{
return [&]<std::size_t...Is>(std::index_sequence<Is...>){
using Ret = std::variant<std::integral_constant<std::size_t, Is>...>;
std::array<Ret, sizeof...(Is)> a = {{Ret{std::integral_constant<std::size_t, Is>{}}...}};
return a[v.index()];
}(std::make_index_sequence<sizeof...(Ts)>());
}
Then
std::visit([&](auto index){ // index() is constexpr
// ...
}, get_index_variant(v));