[swift-evolution] [Draft Proposal] Improve protocol inheritance behaviour

Dale Buckley dalebuckley86 at gmail.com
Wed Mar 15 04:14:21 CDT 2017


Thats the thing, I believe it does matter where the super implementation is
called from so the developer can reasonably reason as to what behaviour
they are invoking by calling 'super'. Because of this I'm not 100%
convinced that this is the right way to go, even if on face value it seems
like the logical way to do it. It's only when digging in that it becomes
apparent that there could be issues in moving this direction.

I'm starting to wonder if there is a third option here, something that
still explicitly defines that you are taking control over the
implementation of the protocol required method implementation but without
taking the 'override' keyword and making it do something that it's not
originally intended to do (primarily used for overriding methods in the
parent class).

I would quite like a few other people to join in on this discussion so we
can get a few minds at work (not that I don't value you opinion Nevin, I
really do, just think the more minds the better) on this problem as it is a
legitimate one that I would quite like to see solved by the time Swift 4
rolls around. Any takers?

Dale



On Tue, Mar 14, 2017 at 11:55 AM, Dale Buckley <dalebuckley86 at gmail.com>
wrote:

> So having a little more of a think about this what should happen in this
> situation:
>
> class A {
>     func method() {
>         print("A")
>     }
> }
>
> protocol P {
>     func method()
> }
>
> extension P {
>     func method() {
>         print("P")
>     }
> }
>
> class B: A, P {
>     override func method() {
>         super.method()
>         print("B")
>     }
> }
>
> B().method()
>
> When 'super' is called by the method in class B, should it call the method
> in class A or the protocol extension in extension P? As it stands by using
> override it always looks to it's superclass and ignores what happens in the
> protocol extension. This says to me that the 'override' keyword isn't right
> for solving this issue and should be reserved for inheritance, and not for
> solving this issue with choosing the subclass implementation over the
> protocol extension implementation.
>
> Dale
>
> On Tue, Mar 14, 2017 at 11:39 AM, Dale Buckley <dalebuckley86 at gmail.com>
> wrote:
>
>> So I did consider using the 'override' keyword, but it then raised the
>> question as to what should happen when 'super' is called.
>>
>> I agree that by adding 'override' it makes it more obvious that you are
>> overriding an existing implementation of the method, regardless of if it's
>> defined in the class you are inheriting from or from the protocol
>> extension. But if you add 'override' you are explicitly saying you can call
>> the super implementation.
>>
>> So the question is what should happen here? Realistically you could
>> expect calling 'super' would call the protocol extension implementation of
>> the method, but I'm trying to think of the problems that could occur if its
>> implemented in that way.
>>
>> Any thoughts on this?
>>
>> Dale
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.swift.org/pipermail/swift-evolution/attachments/20170315/d03d734e/attachment.html>


More information about the swift-evolution mailing list