1

This question (more or less) has been asked before (here and here), but AFAICS it has never been answered. So I ask it again.

We can put @inlinable at all of these places:

public protocol Foo {
  @inlinable var v: Int {
    @inlinable get
    @inlinable set
  }
}

public struct S: Foo {
  @inlinable public var v: Int {
    @inlinable get { ... }
    @inlinable set { ... }
  }
}

But, for example, in what way is the above different than:

public protocol Foo {
  var v: Int { get set }
}

public struct S: Foo {
  public var v: Int {
    @inlinable get { ... }
    @inlinable set { ... }
  }
}

?

Are the rules for this documented somewhere?

1 Like
2

AFAICT, it has absolutely no implication. You still need to reannotate the implementation declaration, or else it’ll not get inlined.

3

If that is indeed the case, then perhaps the compiler should simply disallow any @inlinable (or other attribute) that has no effect, to avoid confusion, ie in my above example:

public protocol Foo {
  @inlinable var v: Int {        // ERROR: '@inlinable' attribute cannot be applied to this declaration
    @inlinable get               // ERROR: '@inlinable' attribute cannot be applied to this declaration
    @inlinable set               // ERROR: '@inlinable' attribute cannot be applied to this declaration
  }
}

public struct S: Foo {
  @inlinable public var v: Int { // ERROR: '@inlinable' attribute cannot be applied to this declaration
    @inlinable get { ... }
    @inlinable set { ... }
  }
}

?

1 Like