tylerh111 / vitta Goto Github PK

There is an issue where there is no recursive checking of the new types that are added to the aggregation. It occurs in the second recursive case of variant_aggregate_recursive. The NewTs... of the input variant are directly added to the final variant type. These types should be each be recursively checked in case a variant type is one of the NewTs types.

The same is true for the initial call to variant_aggregate_t. If the first argument is a variant, there is no check on its alternatives Ts.... This should be recursively checked before adding any other inputs.

The following is an examples does not compile.

using t1 = int;
using t2 = variant<variant<int>>;

// example of `variant_aggregate` not checking given variant alternatives 
static_assert(is_same_v<
    variant           < t1 >,
    variant_aggregate < t2 >
>);

// example of `variant_aggregate_recursive` not checking new variant alternatives 
static_assert(is_same_v<
    variant           < int, t1 >,
    variant_aggregate < int, t2 >
>);

The same is true for tuple_aggregate and tuple_aggregate_recursive

There is no difference between variant_aggregate.hpp and tuple_aggregate.hpp aside from variant and tuple. This can be generically represented as a type_trait that takes a variadic type as a template. Something along lines of the following could be possible. Here Agg could be variant or tuple and V is of type Agg.

template < template class<...> Agg, typename V, typename NewTV, typename... NewTVs >
struct aggregate;