Consider this code with some custom macros:
@CouchbaseModel
final class Foo
{
@Relationship var child: Bar? = nil
}
@Relationship expands to nothing; it's a marker that helps me decide some things about Bar, since type information isn't available in the AST at the time macros expand.
@CouchbaseModel has several roles, but ONE of the things it's supposed to do is add a second macro attribute, @CouchbasePersisted to stored properties. Here's that implementation:
extension CouchbaseModelMacro: MemberAttributeMacro
{
public static func expansion(...) throws -> [SwiftSyntax.AttributeSyntax]
{
// Some logic to check attributes of the member is omitted for brevity. It's not relevant to the problem.
return [
AttributeSyntax(stringLiteral:
"""
@CouchbasePersisted
"""
)
]
}
}
The Problem
This works great, unless the stored property already has some attributes. In that case, the macro expands to this:
final class Foo
{
@Relationship var child: Bar? = nil
@CouchbasePersisted
}
This, of course, is useless. @CouchbasePersisted is in the wrong spot and attached to nothing. @CouchbasePersisted (an Accessor and Peer macro) therefore does nothing. But as far as I can see, I don't have any control over how the memberAttribute macro positions/orders the attributes on members.
Am I misusing the memberAttribute macro? Is there a way to work around this issue?
Another Approach
I also tried constructing the AttributeSyntax this way, which produced the same mis-positioned result when a member has existing attributes:
AttributeSyntax(
attributeName: IdentifierTypeSyntax(name: .identifier("CouchbasePersisted"))
)