Don't require members of structs to also impl TypeName

[Boscop]

It shouldn't be necessary for members to also impl TypeName.

[antiguru]

IIRC, we don't require for generic types implementing TypeName that their generic parameters also implement TypeName. Could you provide a MWE so we can have a look?

What might happen is that somewhere a type name of a GetSessionData type is requested when K doesn't implement TypeName.

[Boscop]

What might happen is that somewhere a type name of a GetSessionData type is requested when K doesn't implement TypeName.

@antiguru But that's what I mean, K shouldn't have to impl TypeName in order for GetSessionData to impl TypeName! Why is it required?

[antiguru]

Currently, we include generic parameters in the type name, as they are effectively part of the type. Instead of using derive(TypeName), you could implement the TypeName trait yourself for GetSessionData. This implementation then doesn't need to have the requriement for K: TypeName.

[Boscop]

@antiguru But why are you including generic parameters in the type name?

(I want to avoid having to manually implement TypeName.)

[gandro]

We have found it useful to differentiate between Vec<i32> and Vec<TcpStream>, which are two entirely different things. It is also the main reason we need a formatter, as we cannot return a static string.

In addition, I also wanted to be roughly compatible with std's type_name intrinsic (which will be stabilized soon, making this crate obsolete).

If you need type names without generic parameters, I recommend you use the named_type crate instead.