Add traits to type_list

[rolandreichweinbmw]

This change adds several traits to type_list.h:

• type_list_contains
• type_list_index_of
• type_list_max_sizeof_type (long name to prevent confusion with type_list_size)
• type_list_max_alignment

Further: test_type_list.cpp including tests for already existing traits.

[jwellbelove]

Removed the review comment.
I thought I'd had it compiling OK with etl::is_one_of, but I hadn't.

[jwellbelove]

A corresponding 'type from index' class should be implemented.

template<typename TypeList, size_t Index>
struct type_list_type_at

i.e.

using TypeList = etl::type_list<char, short, int>;
using second_type = etl::type_list_type_at<TypeList, 1>::type; // short

[rolandreichweinbmw]

Isn't that available via the already implemented etl::nth_type<>?

[jwellbelove]

Well, yes, but that's a much older class, originally designed for a narrower use case.
I think a class named consistently with other type_list_x classes could be less confusing.
There is another etl::type_list style class called etl::parameter_pack which will eventually be deprecated, as etl::type_list covers everything it does, and more.

[jwellbelove]

I would prefer TypeList over TL

[rolandreichweinbmw]

Adjusted accordingly.

[jwellbelove]

For C++17, add _v definitions for types that return a value
i.e.

#if ETL_USING_CPP17
  template<typename TL, typename T>
  inline constexpr bool type_list_contains_v = etl::type_list_contains<TL, T>::value;
#endif

[rolandreichweinbmw]

Adjusted accordingly.

[jwellbelove]

Doesn't this give the same index for the first type in a non-empty type_list, and an empty type_list?

using NonEmpty = etl::type_list<char, int>;
using Empty    = etl::type_list<>;

size_t nonEmptyIndex = etl::type_list_index_of<NonEmpty, char>::value;

size_t emptyIndex = etl::type_list_index_of<Empty, char>::value;

nonEmptyIndex == emptyIndex

Maybe add static constexpr size_t npos = etl::integral_limits<size_t>::max; to etl::type_list<> and return etl::type_list<>::npos for emptyIndex.

For the furture, I'm thinking that an ETL generic etl::npos may be useful

[rolandreichweinbmw]

Adjusted accordingly.

The "npos" is implemented as etl::type_list_npos. Reasoning:

• etl::type_list<>::npos causes linker errors on application level / unit tests since the symbol needs to be defined in a single compile unit and in ETL we only work header-only. This is different from plain global symbols, or in an abstract template.
• etl::npos causes compile warnings and therefore errors because etl::basic_string_view::npos shadows etl::npos