[swift-evolution] [Discussion] A Problem With SE-0025?

Matthew Johnson matthew at anandabits.com
Wed Jun 15 16:30:38 CDT 2016


> On Jun 15, 2016, at 4:27 PM, Xiaodi Wu <xiaodi.wu at gmail.com> wrote:
> 
> On Wed, Jun 15, 2016 at 4:14 PM, Matthew Johnson <matthew at anandabits.com <mailto:matthew at anandabits.com>> wrote:
> 
>> On Jun 15, 2016, at 4:08 PM, Xiaodi Wu <xiaodi.wu at gmail.com <mailto:xiaodi.wu at gmail.com>> wrote:
>> 
>> On Wed, Jun 15, 2016 at 3:09 PM, Matthew Johnson <matthew at anandabits.com <mailto:matthew at anandabits.com>> wrote:
>> 
>>> On Jun 15, 2016, at 2:55 PM, Xiaodi Wu <xiaodi.wu at gmail.com <mailto:xiaodi.wu at gmail.com>> wrote:
>>> 
>>> On Wed, Jun 15, 2016 at 2:48 PM, Matthew Johnson via swift-evolution <swift-evolution at swift.org <mailto:swift-evolution at swift.org>>wrote:
>>> 
>>>> On Jun 15, 2016, at 2:46 PM, Adrian Zubarev via swift-evolution <swift-evolution at swift.org <mailto:swift-evolution at swift.org>> wrote:
>>>> 
>>>> I was referencing to the issue Robert discovered in his implementation. 
>>>> 
>>>> I do understand what the proposal is all about, but thank you for re-clarifying that to me. :)
>>>> 
>>>> 
>>> 
>>> I don’t think it’s a bug, but it is definitely something that isn’t as clear as it should have been.
>>> 
>>> Was it intentional on the part of the proposal, then, that there should be two modifiers meaning the same thing for a top level declaration in a file? Or was it intended that only one or the other be used in that scenario?
>> 
>> I don’t think it was carefully considered, although I think it did come up at some point during discussion in the context of compatibility with existing code (i.e. nothing changes for current top-level `private` declarations).  
>> 
>> It is in some sense a “coincidence” that they mean the same thing at file scope.  The proposal would have had to introduce a specific prohibition to prevent this situation and it did not do so.  That said, I think this kind of issue falls well within the discretion of the core team to make a call without violating the spirit of the proposal.
>> 
>> There are two reasonable options here: 
>> 
>> 1. Allow both `private` and `fileprivate` at file scope despite the fact that they have the same meaning.  This is more consistent in the sense that we are not introducing a special case that arbitrarily prohibits an otherwise valid access modifier.  It also means that nothing needs to change for top level `private` declarations in existing code.
>> 
>> 2. Prohibit `private` at file scope.  Given that it appears as if the behavior of `private` at file scope may not be intuitive and is equivalent to `fileprivate` it might be reasonable to just disallow it.  This would result in more consistent *code* (even if there needs to be a special case in the language).
>> 
>> I don’t have a strong opinion on which option we choose.  But I do feel strongly that the semantics of `private` need to properly respect the scope in which the keyword is written and into which the associated declaration is introduced (rather than the scope *inside* the declaration it is attached to).
>> 
>> Right, I think both would be OK. More radically, we might want to re-evaluate the continued utility of a `fileprivate` scope. It seems the use cases for such a scope not adequately served by either `internal` or the new `private` would be exceedingly rare.
> 
> I disagree with that.  `fileprivate` is indispensable when you need it.  There are times when you want to keep visibility limited to the current file but the new `private` is too restrictive (for example, you need to access a member of one type in a closely related extension of a different type that lives in the same file).
> 
> Sure. This was more of a thought for the future. As we move towards fully embracing a scope-based model for organizing code, modules will no longer need to be strictly "single units of code distribution," and a move towards supporting submodules could serve your use case without `fileprivate`. That, IMO, would be a logical endpoint of moving from file-based access to scope-based access.

I would really like to see submodules, but I think there would still be valid uses for `fileprivate` even with them.  But of course we would need to know the details of submodules to have a good discussion about that so it’s a topic for the future. :)

> 
> 
>>  
>> -Matthew
>>>> 
>>>> 
>>>> 
>>>> -- 
>>>> Adrian Zubarev
>>>> Sent with Airmail
>>>> 
>>>> Am 15. Juni 2016 um 21:40:37, Matthew Johnson (matthew at anandabits.com <mailto:matthew at anandabits.com>) schrieb:
>>>> 
>>>>> What seems like a nasty bug missed during review?  I don’t follow you there.
>>>>> 
>>>>> This proposal was specifically designed to follow Swift’s design of a scope-based access control mechanism rather than a type-based access control mechanism that is common in other languages.
>>>> 
>>>> 
>>>> _______________________________________________
>>>> swift-evolution mailing list
>>>> swift-evolution at swift.org <mailto:swift-evolution at swift.org>
>>>> https://lists.swift.org/mailman/listinfo/swift-evolution <https://lists.swift.org/mailman/listinfo/swift-evolution>
>>> 
>>> 
>>> _______________________________________________
>>> swift-evolution mailing list
>>> swift-evolution at swift.org <mailto:swift-evolution at swift.org>
>>> https://lists.swift.org/mailman/listinfo/swift-evolution <https://lists.swift.org/mailman/listinfo/swift-evolution>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.swift.org/pipermail/swift-evolution/attachments/20160615/d984d12b/attachment.html>


More information about the swift-evolution mailing list