Andrei Alexandrescus Opening Keynote from Meeting C++ 2018
The next big Thing
by Andrei Alexandrescu
The next big Thing - Andrei Alexandrescu - Meeting C++ 2018 Opening Keynote
Comments (1)
0
0
Does "if constexpr" really need a fix?
In slides 27 to 32 Andrei Alexandrescu says that "if constexpr" must not introduce a scope. He gives an example that "if constexpr" cannot be used to specify types depending on template paramters (because the specified types are not visible outside of the scope). This is treu, but there are other well known techniques that can be used to achieve these goal (For example, define some kind of type traits). Using this technique, hiscode changes to
Which version is better depends on the situation and is up to some degree a matter of taste. From my understanding, both version can be considered as "Design by Introspection" and allow
"Up to 2^n possible interfaces, in compact form!".
So I am not really convinced that "if constexpr" must not introduce a scope.
In slides 27 to 32 Andrei Alexandrescu says that "if constexpr" must not introduce a scope. He gives an example that "if constexpr" cannot be used to specify types depending on template paramters (because the specified types are not visible outside of the scope). This is treu, but there are other well known techniques that can be used to achieve these goal (For example, define some kind of type traits). Using this technique, hiscode changes to
template <class K, class V, std::size_t maxLength>
struct RobinHashTable
{
using CellIdx = typename uint_t<(maxLength < 0xFFFE)>::type;
alignas(8) KV_struct<K,V, (sizeof(K) % 8 < 7)> KV;
};
Which version is better depends on the situation and is up to some degree a matter of taste. From my understanding, both version can be considered as "Design by Introspection" and allow
"Up to 2^n possible interfaces, in compact form!".
So I am not really convinced that "if constexpr" must not introduce a scope.
Add a Comment
Comments are closed.