[swift-evolution] Proposal: Namespaces

[T.J. Usiyan]

The use case described in the proposal and your response doesn't use
structs as a "bucket of values". I uses *the type* as a bucket of values.
The point of using an empty enum is to emphasize this point. The overal
point of this proposal is to introduce semantics that obviate the need to
'decide' between enum and struct since neither are the type actually
desired. I suggest using an enum as the implementation with the hope that
it will simplify implementation.

Your suggestion of `static struct` emphasizes the utility of a namespace
keyword. Is `static struct …` a declaration of a type or an instance and of
what?

TJ

On Sun, Dec 13, 2015 at 6:44 PM, Cole Kurkowski <crk at fastmail.com> wrote:

> IMO using  an enum is non-intuitive. A struct is a "bucket of values" and
> so makes fits with the usage here. If the ability to instantiate the
> NotificationNames struct is a problem, adding a
>
> private init()
> {
>
> }
>
> seems more intuitive than co-opting an enum, which has a fairly
> established use case.
>
> Your proposal would essentially be adding a keyword that is transformed
> back into "enum" during compilation. I don't think that a relatively minor
> improvement in clarity for a fairly small use case is worth adding a
> language feature.
>
> Allowing structs to be declared as static would be another way of
> simplifying this kind of declaration, i.e.
>
> static struct NotificationNames {
>      let userDataChanged = ...
>      let recievedAlert = ...
> }
>
> This would allow dropping the private init declaration and dropping the
> static from each property. I'm not sure I think this shortcut is necessary
> or even if it's a good idea, but I do think it fits with the language a
> little better than your current proposal.
>
> On Dec 11, 2015, at 04:39, T.J. Usiyan via swift-evolution <
> swift-evolution at swift.org> wrote:
>
> Namespaces
> Author(s): TJ Usiyan
>
> Introduction
>
> A `namespace` keyword for swift hold related global variables.
>
> Motivation
>
> We often want to collect related variables which do not, for whatever
> reason, fit neatly into a type provided by Swift. Importing Objective C
> 'magic strings' as Joshua Sullivan does [here] is one such example. The
> solution he arrives at is a `struct` type with many type variables and no
> fields.
> struct NotificationNames {
>  static let userDataChanged = "UserDataChangedNotificationName"
>  static let receivedAlert = "ReceivedAlertNotificationName"
>  static let peanutButterJellyTime =
> "ItsPeanutButterJellyTimeNotificationName"
> }
>
> Users of this API are not meant to create instances of this type, yet it
> is still possible. A solution to this is to use an `enum` without cases.
>
> enum NotificationNames {
>  static let userDataChanged = "UserDataChangedNotificationName"
>  static let receivedAlert = "ReceivedAlertNotificationName"
>  static let peanutButterJellyTime =
> "ItsPeanutButterJellyTimeNotificationName"
> }
>
> No instances of the `enum` can be made. That this inability to create an
> instance is intentional is only conveyed via the type system.
>
> Proposed solution
>
> Clarity would be greatly improved if we could use `namespace` as a synonym
> for an enum with no cases. This would allow us to avoid repetition of
> `static` as well.
>
> namespace NotificationNames {
>  let userDataChanged = "UserDataChangedNotificationName"
>  let receivedAlert = "ReceivedAlertNotificationName"
>  let peanutButterJellyTime = "ItsPeanutButterJellyTimeNotificationName"
> }
>
> Detailed design
>
> Namespaces can be represented in the type system as enums. All variables
> declared in a namespace's scope would be type variables.
>
> Impact on existing code
>
> This is an additive change and, as such, should not break any existing
> code. Though it doesn't matter, this addition could possibly attain ABI
> compatibility if namespaces are represented as enums without cases.
>
> Alternatives considered
>
> Don't implement namespaces. Developers could continue(begin?) using an
> empty enum to hold type variables.
>
> _______________________________________________
> 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/20151213/f1e79db0/attachment.html>