[swift-evolution] PITCH: Export _JSONEncoder / _JSONDecoder

Florent Vilmart florent at flovilmart.com
Thu Nov 9 18:34:31 CST 2017


I really like this idea of scoped import, but I’m not sure it’s part of anything supported widely. I was about to suggest ‘namespaces’ but that’s not a thing in swift.

On Nov 9, 2017, 17:06 -0500, Rod Brown <rodney.brown6 at icloud.com>, wrote:
>
> > On 7 Nov 2017, at 6:24 am, Tony Parker via swift-evolution <swift-evolution at swift.org> wrote:
> >
> > Hi Florent,
> >
> > We definitely thought about this while designing the set of types with the Codable proposals.
> >
> > One serious concern was just how much API surface area there already is with Codable. If we open up the internal classes as well, we risk confusing the majority of people who are just adopting Codable with APIs that are intended only for the minority of people who are trying to create their own custom encoders and decoders.
> >
> > Any thoughts on how to mitigate this?
> >
> > - Tony
>
> I’ve been curious for some time about if we can do something about an opt-in import in Swift?
>
> For example, currently UIGestureRecognizer in UIKit has “subclass only” methods that are protected and opt in. Importing UIKit itself doesn’t bring it in, and instead you need to specifically import UIKit.UIGestureRecognizerSubclass.
>
> I realise this is a standalone case, but I’m wondering whether we can generalise this into something we can propose, to actively support nested scopes in the same way?
>
> This would lend well to disclosing the internals like this. It would avoid users jumping straight for the internal types because they wouldn’t be there with “import Foundation” - it would require something eg “import Foundation.xyz”.
>
> Was this part of an earlier discussion of modules etc?
>
> - Rod
>
> >
> > > On Nov 3, 2017, at 9:58 AM, Florent Vilmart via swift-evolution <swift-evolution at swift.org> wrote:
> > >
> > > At Swift Summit, we discussed with Joe and Jordan about opening up the Encoder/Decoder classes in order to make the work of an encoder designer easier.
> > >
> > > As I was working on an API project, I found myself into the situation of needing to tweak so slightly the encoding strategy this required a full copy/paste of the JSONEncoder.swift file and playing with the internals. I also wanted to implement a simple QueryStringEncoder/Decoder that would properly encode / decode a query string.
> > >
> > > The internally defined classes are proven a very powerful tool of reflection as well, being able to collect / untransform a series of containers safely into a strongly typed swift object.
> > >
> > > The pitch:
> > >
> > > - Keep JSONEncoder / JSONDecoder as 'proxies' to encoding to Data
> > > - Make _JSONEncoder / _JSONDecoder open classes
> > > - Mark public all container implementations of UnkeyedEncodingContainers etc...
> > > - Find a good naming for the _JSONEncoder and _JSONDecoder, that doesn't conflict with JSONEncoder / JSONDecoder but also denotes they conform to Encoder.
> > >
> > > Opening those API's isn't for the general Codable implementation, the JSONEncoder/JSONDecoder should stay as-is but it's intended to reduce the amount of boiler plate one would need to implement in order to provide different serialization mechanism / strategy.
> > >
> > >
> > > _______________________________________________
> > > swift-evolution mailing list
> > > swift-evolution at swift.org
> > > https://lists.swift.org/mailman/listinfo/swift-evolution
> >
> > _______________________________________________
> > swift-evolution mailing list
> > swift-evolution at swift.org
> > https://lists.swift.org/mailman/listinfo/swift-evolution
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.swift.org/pipermail/swift-evolution/attachments/20171109/12f89e3e/attachment.html>


More information about the swift-evolution mailing list