[swift-evolution] Proposal: CustomConvertible protocol

[John McCall]

> On Dec 4, 2015, at 5:51 AM, Adrian Zubarev <adrian.zubarev at devandartist.com> wrote:
> We have finally a corner where people can share their ideas on how Swift can be improved, and I’m glad my ideas might be part of this fun.
> 
> Swift introduced a few Convertible protocols, but it still needs some more If you ask me. 
> 
> Something like this:
> 
> public protocol CustomConvertible {
>      
>     typealias CustomType
>      
>     public init(value: Self.CustomType) // or in a better way I can't come up with
> }
> But it should work in a way that we could create more different types like enums and ErrorType.
> 
> protocol AConvertible: CustomConvertible { /* some stuff here*/ }
> protocol BConvertible: CustomConvertible { /* some stuff here*/ }
> 
> struct A {
>     var foo: Int = 0
> }
> 
> struct B {
>     var boo: String = "Hello World"
> }
> 
> struct C: AConvertible, BConvertible {
>      
>     typealias CustomAType = A
>     typealias CustomBType = B
>      
>     var magical: SomeType
>          
>     init(value: CustomAType) {
>         // implement it
>     }
>      
>     init(value: CustomBType) {
>         // implement it
>     }
> }
> 
> //===========================
> 
> let firstC: C = A() // this is what I'm missing in Swift
> let secondC: C = B() // this would be a great syntax sugar
> What do you think?
> 
This is effectively a request for user-defined conversion operators.  We actually did support those at one point in the history of Swift, but we removed them because they both introduced a lot of confusion and errors into common user idioms and were very problematic for the type-checker.  It’s easy to look at obviously-contextually-typed examples like this and say that the type-checker should just make it work, but in the context of a general System F-sub type system, it is a massive source of added complexity.

In other words, at present, this is not an extension we feel we can deliver a satisfactory experience for, and to be taken seriously, any proposal is going need to demonstrate a lot of familiarity with type systems.

John.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.swift.org/pipermail/swift-evolution/attachments/20151204/2cd5dab6/attachment.html>