Overloads

[ArturG via Digitalmars-d-learn]

On Sunday, 26 June 2016 at 11:23:14 UTC, Márcio Martins wrote:
> Consider this snippet:
>
> struct X {
>   int foo(Args...)(Args args) if (Args.length > 1) { return 
> Args.length; }
>
>   int foo() { return 0; }
>
>   int foo(int y) { return 1; }
>
>   alias Name = string;
>
>   int field_;
> }
>
>
> void listMembers(T)(ref T x) {
>   foreach (Member; __traits(derivedMembers, T)) {
>     pragma(msg, Member, " ", __traits(getOverloads, x, 
> Member).length);
>     //pragma(msg, __traits(getProtection, __traits(getMember, 
> x, Member))); // Error: argument string has no protection
>   }
> }
>
> void main() {
>   X x;
>   listMembers(x);
>   //auto fptr = &x.foo; // Error: x.foo(Args...)(Args args) if 
> (Args.length > 0) is not an lvalue
> }
>
> Output:
> foo 0LU
> Name 0LU
> field_ 0LU
> foo 0LU
> Name 0LU
> field_ 0LU
>
>
> There seems to be a few problems here:
> 1. It seems like getOverloads is returning 0 for 'foo' - is 
> this a bug? Was expecting a 3 or at least a 2 if the template 
> would be ignored.
> 2. That alias breaks getProtection - is this bug? Seems like it 
> should be public.
>
> These two make it quite hard to iterate over and collect info 
> about arbitrary aggregates.
>
> I want to get a list of all *public* members, including 
> pointers to all public member functions and their overloads, 
> excluding template member functions. This is turning out to be 
> hard due to these "unexpected behaviors".
>
> Is there anything else I can do?

__traits(getOverloads, x, Member).length works if you place the 
template after a function of the overloads and then it returns 2.

it fails as soon as the first member of the overload set is any 
template, so i guess it must be a bug.
e.g.

struct Fails
{
     void foo()(){}
     void foo(int){}
}

struct Works
{
     void foo(int){}
     void foo()(){}
}

__traits(getOverloads, Fails, "foo").length.writeln; // 0
__traits(getOverloads, Works, "foo").length.writeln; // 1