Here is my code:

struct L
{
    string file;
    size_t line;
}

auto W(string f = __FILE__,size_t l= __LINE__)()
{
    return L(f,l);
}

struct A
{
  @W       // (1): line# 16
  {
    int a;
    int b;
  }
  @W()     // (2): line# 21
  {
    int c;
    int d;
  }
}

void main()
{
    pragma(msg, __traits(getAttributes, A.a)); // tuple(W(string f = __FILE__, ulong l = __LINE__)())
    pragma(msg, __traits(getAttributes, A.b)); // tuple(W(string f = __FILE__, ulong l = __LINE__)())
    pragma(msg, __traits(getAttributes, A.c)); // tuple(L("app.d", 21LU))
    pragma(msg, __traits(getAttributes, A.d)); // tuple(L("app.d", 21LU))
}

W() (2) works as expected but is it possible to achieve the same without parenthesis so getAttributes trait returns tuple(L("app.d", 16LU)) for (1)?

On Sunday, 14 November 2021 at 05:12:58 UTC, Andrey Zherikov wrote:

No, without parens this is really the function template, as a symbol, that becomes the UDA.

On Sunday, 14 November 2021 at 07:06:25 UTC, user1234 wrote:

Can this be achieved with any other type of W? For example I can get the same result with struct (also changing @W() to @W!() on line (2)):

struct W(string f = __FILE__, size_t l = __LINE__)
{
    auto loc = L(f,l);
}