https://issues.dlang.org/show_bug.cgi?id=21137
--- Comment #3 from Simen Kjaeraas <[email protected]> --- You're right, my mind was somewhere else when I did that. The problem is, D's templates are complicated beasts (and untyped lambdas are templates), and the information is undecidable in the general case: template foo(bool b, T) { static if (b) { void foo(@(T) T i) {} } else { enum T foo = T.init; } } We don't know if the above even is a function, and even if we remove the bool parameter, there's no possible type for the UDA on the parameter 'i' without an explicit instantiation (and figuring out what template arguments are valid is also a hard problem). We could possibly expose a __traits that gives access to this information in the limited cases where the template body has exactly one member which is a function and whose parameters' UDAs can be determined. IMO, this is material for a DIP on template introspection, possibly AST macros. --
