Bug 21138 – Difficult to treat UDAs given to parameters.

Function arguments can be given UDAs, but they are very difficult to treat. ---------------------------------------------------------- import std; template Temp1(alias params) { alias Temp1 = params; } template Temp2(params...) { alias Temp2 = params; } template Foo(alias f) { // OK1 alias params = Parameters!(f); pragma(msg, params); // (@(10) int, @(15) long) // OK2 pragma(msg, params[0..1]); // (@(10) int) pragma(msg, __traits(getAttributes, params[0..1])); // tuple(10) // NG1: Lack UDAs alias param2 = params[0]; pragma(msg, params[0]); // int //pragma(msg, __traits(getAttributes, params[0])); // Error // NG2: Error //pragma(msg, Temp1!(params)); // NG3: This is able to pass, but lack UDAs pragma(msg, Temp2!(params)); } void main() { alias X = Foo!( (@(10) int x, @(15) long b) => x ); } ---------------------------------------------------------- In NG1 and NG3, UDAs should not lack. NG2 should allow them to be passed to an alias parameter, as is possible in OK1. But maybe this should be considered separately from this issue. (Allow tuple to be passed to an alias template parameter)

Changing the behavior of template alias parameters to allow them to bind to tuples is a large breaking change, and is simply not going to happen, though I see the benefit in making it work the same there as alias works elsewhere. NG1 and NG3 are the samme issue, and is described in issue 19787 *** This issue has been marked as a duplicate of issue 19787 ***