I agree this is a very important use case. I wish attributes were not ignorable, the Clang implementer veto here is actually quite frustrating. I think Java actually got this one right with their annotations (ref: https://docs.oracle.com/javase/1.5.0/docs/guide/language/annotations.html).

struct serializable
{
    bool value;
};

struct Foo
{
    @serializable(true)
    int data;
};

My major concern is that this will just end up subsuming attributes anyways with folks writing things like:

@deprecated() [[deprecated]] struct foo {};

Perhaps to avoid that, there are (at least for attributes that folks might want to reflect) corresponding "standard annotations" that are implied by using the attribute (if the attribute is not "being ignored").

e.g.,

[[deprecated]] struct foo {};
/* implies */
@std::deprecated() struct foo {};

Uses of the annotation directly would be not be ignorable.

- Wyatt

Jeremy Ong wrote:

Thinking outloud, as an example, we might declare:

template <bool>
class Serializable {};

And later use this class as a decorator like so:

struct Foo
{
    @Serializable<true>
    int data;
}:

We can do this, but it isn't necessary. @type is not
substantially easier to implement than @expr, so we
can just do

struct serializable
{
    bool value;
};

struct Foo
{
    @serializable(true)
    int data;
};

instead.