[swift-users] Swift 3 (Xcode 8 GM) issue with @escaping

Thu Sep 8 11:57:33 CDT 2016

> On Sep 7, 2016, at 6:50 PM, Jacob Bandes-Storch <jtbandes at gmail.com> wrote:
> 
> On Wed, Sep 7, 2016 at 5:54 PM, Michael Ilseman via swift-users <swift-users at swift.org <mailto:swift-users at swift.org>> wrote:
> I implemented a better (i.e. correct) diagnostic message for this at https://github.com/apple/swift/pull/4670 <https://github.com/apple/swift/pull/4670>. I want to also do a better diagnostic specifically for aggregate parameters to functions (e.g. optional closures), but that requires more work in the type checker.
> 
> Basically, @escaping is valid only on closures in function parameter position. The noescape-by-default rule only applies to these closures at function parameter position, otherwise they are escaping. Aggregates, such as enums with associated values (e.g. Optional), tuples, structs, etc., if they have closures, follow the default rules for closures that are not at function parameter position, i.e. they are escaping.
> 
> Shouldn't it be possible to allow distinguishing @escaping/@noescape for aggregates like these, at least for the simple case of Optional?  (I handled optionals in https://github.com/apple/swift/pull/4438 <https://github.com/apple/swift/pull/4438> for imported function types; see comment <https://github.com/apple/swift/pull/4438#issuecomment-243645367>.)
>  

Yes it is possible (but *not* in time for Swift 3) to address this. Optional at the very least would be the highest bang for our buck, along with other single-type aggregate structures. A general solution would require some kind of aggregate liveness solution, but we can get something very good with a bit less: we could express such things with the ‘@escaping’ still being at the immediate function parameter position, and applying it throughout aggregate types in a principled fashion. E.g.:

func foo(_ a: @escaping ((Int) -> Int)?) {}
func foo(_ a: @escaping (Int, (Int) -> Int)) {}
func foo(_ a: @escaping ((Int) -> Int, (Int) -> Int)) {} // both are escaping
func foo(_ a: @escaping [(Int) -> Int]? {} // they are all escaping

Struct aggregate members and protocol associated types would not be impacted by this, as their member types are not listed in their signature, only tuples and generic type parameters (modulo sugar).

This would re-enforce that it’s about the parameter position specifically being a special case. Also, I don’t think there’s any practical benefit to finer grained escapability, e.g. some tuple elements are escaping and some are not. Though, at this point, maybe it makes more sense being a parameter attribute, rather than a type attribute…

Any solution should be discussed further on swift-evolution.

> 
> It would be a post-Swift-3 addition to the language to be able to support more robust liveness tracking here. There may be interesting directions to take this, with optional closures being the most common beneficiary. 
> 
>> On Sep 7, 2016, at 3:33 PM, Shawn Erickson via swift-users <swift-users at swift.org <mailto:swift-users at swift.org>> wrote:
>> 
>> I see https://bugs.swift.org/browse/SR-2324 <https://bugs.swift.org/browse/SR-2324> and https://bugs.swift.org/browse/SR-2444 <https://bugs.swift.org/browse/SR-2444> which looks related to this issue and may explain the error I saw on "the other side" of this.
>> 
>> 
>> 
>> On Wed, Sep 7, 2016 at 3:28 PM Shawn Erickson <shawnce at gmail.com <mailto:shawnce at gmail.com>> wrote:
>> Yeah I actually have a few of those myself that I can no longer do.
>> 
>> On Wed, Sep 7, 2016 at 3:26 PM Jon Shier <jon at jonshier.com <mailto:jon at jonshier.com>> wrote:
>> Perhaps relatedly, it no longer seems possible to mark typealiased closures as @escaping. That was quite handy when you know that closures will always be used asynchronously.
>> 
>> 
>> Jon
>> 
>> 
>> 
>>> On Sep 7, 2016, at 6:15 PM, Shawn Erickson via swift-users <swift-users at swift.org <mailto:swift-users at swift.org>> wrote:
>>> 
>> 
>>> I should note that this issue also appeared in an earlier variant of Swift after the addition of @escaping but I was on vacation so didn't get a chance to report it then. It isn't new with the Xcode 8 GM.
>>> 
>>> On Wed, Sep 7, 2016 at 3:08 PM Shawn Erickson <shawnce at gmail.com <mailto:shawnce at gmail.com>> wrote:
>>> I like and fully supported the change to @escaping away from @noescape however in a body of code that I am porting to the latest Swift 3 variant (as found in Xcode 8 GM) I am hitting an issue for methods that take an optional completion closure. If optional is involved I can't find a way to apply @escape to the escaping closure. See the following for an basic example...
>>> 
>>> Is their a way to do what I need and/or is this an edge case in the implementation of @escaping?
>>> 
>>> typealias MyCallback = (String)->()
>>> 
>>> // Happy
>>> func foo1(bar: String, completion: ((String)->())) {
>>>     completion(bar)
>>> }
>>> 
>>> // Happy
>>> func foo2(bar: String, completion: MyCallback) {
>>>     completion(bar)
>>> }
>>> 
>>> // Happy
>>> func foo3(bar: String, completion: ((String)->())? = nil) {
>>>     completion?(bar)
>>> }
>>> 
>>> // Happy
>>> func foo4(bar: String, completion: MyCallback? = nil) {
>>>     completion?(bar)
>>> }
>>> 
>>> // Happy
>>> func foo5(bar: String, completion: Optional<MyCallback> = nil) {
>>>     completion?(bar)
>>> }
>>> 
>>> // Happy
>>> func foo6(bar: String, completion: @escaping (String)->()) {
>>>     completion(bar)
>>> }
>>> 
>>> // Happy
>>> func foo7(bar: String, completion: @escaping MyCallback) {
>>>     completion(bar)
>>> }
>>> 
>>> // Unhappy...
>>> // "@escaping attribute only applies to function types"
>>> func foo8(bar: String, completion: @escaping ((String)->())? = nil) {
>>>     completion?(bar)
>>> }
>>> 
>>> // Unhappy...
>>> // "@escaping attribute only applies to function types"
>>> func foo9(bar: String, completion: @escaping MyCallback? = nil) {
>>>     completion?(bar)
>>> }
>>> 
>>> // Unhappy...
>>> // "@escaping attribute only applies to function types"
>>> func foo10(bar: String, completion: (@escaping ((String)->()))? = nil) {
>>>     completion?(bar)
>>> }
>>> 
>>> // Unhappy...
>>> // "@escaping attribute only applies to function types"
>>> func foo11(bar: String, completion: (@escaping MyCallback)? = nil) {
>>>     completion?(bar)
>>> }
>>> 
>>> // Unhappy...
>>> // "@escaping attribute only applies to function types"
>>> func foo12(bar: String, completion: Optional<@escaping MyCallback> = nil) {
>>>     completion?(bar)
>>> }
>> 
>>> _______________________________________________
>>> swift-users mailing list
>>> swift-users at swift.org <mailto:swift-users at swift.org>
>>> https://lists.swift.org/mailman/listinfo/swift-users <https://lists.swift.org/mailman/listinfo/swift-users>
>> 
>> _______________________________________________
>> swift-users mailing list
>> swift-users at swift.org <mailto:swift-users at swift.org>
>> https://lists.swift.org/mailman/listinfo/swift-users <https://lists.swift.org/mailman/listinfo/swift-users>
> 
> 
> _______________________________________________
> swift-users mailing list
> swift-users at swift.org <mailto:swift-users at swift.org>
> https://lists.swift.org/mailman/listinfo/swift-users <https://lists.swift.org/mailman/listinfo/swift-users>
> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.swift.org/pipermail/swift-users/attachments/20160908/92a68b4f/attachment-0001.html>