Swift 5.0 and Result in Standard Library

[larryonoff]

I think it's time to start preparation for the new features of Swift 5.0, i.e. Result that has been recently accepted to the standard library, see
https://github.com/apple/swift-evolution/blob/master/proposals/0235-add-result.md.

The issue that I see that migration to the Swift.Result is a breaking change since different signatures.

https://github.com/apple/swift-evolution/blob/master/proposals/0235-add-result.md

public enum Result<Value, Error: Swift.Error> {
    case value(Value), error(Error)
}

when https://github.com/antitypical/Result

public enum Result<Value, Error: Swift.Error> {
	case success(Value)
	case failure(Error)
}

I totally respect to https://github.com/antitypical/Result, but I think that it's better using Standard Library.

Any community thought about?

[mdiep]

The accepted proposal is:

public enum Result<Success, Failure: Error> {
  case success(Success)
  case failure(Failure)
}

💯 we should use this.

antitypical/Result should also update to be a set of additions to the stdlib Result.

[larryonoff]

@mdiep the accepted signature 1-by-1 to antitypical/Result. This's really great

[andersio]

The tricky bit is NoError which could have been a typealias to Never, had it been conformed to Error in stdlib.

[mdiep]

The tricky bit is NoError which could have been a typealias to Never, had it been conformed to Error in stdlib.

Never is conformed to Error now.
swiftlang/swift#16857

[andersio]

Argh! That’s brilliant. Missed that part from the SE proposal.

[larryonoff]

Result already updated for Swift 5.0, see antitypical/Result#278

[andersio]

Assuming antitypical/Result would just be augmenting the stdlib Result since Swift 5, should we consider dropping antitypical/Result as a dependency?

RAS depends on NoError, the structural definition of Result and ResultProtocol. The former two can be sourced from the stdlib since Swift 5. The last one is used only for extension constraint for dematerialize(), which can be re-declared in RAS, and later replaced by parameterised extension when it lands.

[larryonoff]

I think it can just be extensions in ReactiveSwift without Result dependency.

[mdiep]

Assuming antitypical/Result would just be augmenting the stdlib Result since Swift 5, should we consider dropping antitypical/Result as a dependency?

Yes, I think so. 👍

[NachoSoto]

Oh I missed this. I started the process: #702. I'm excited to make ReactiveSwift dependency-free :)

[larryonoff]

@NachoSoto great to see pilots here :)

[Kalvin126]

@larryonoff I think we could close this issue now that #702 Swift.Result migration is merged?

[NachoSoto]

👍🏻