<html><head><meta http-equiv="Content-Type" content="text/html charset=utf-8"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;" class=""><br class=""><div><blockquote type="cite" class=""><div class="">On Jun 8, 2016, at 4:47 PM, Austin Zheng <<a href="mailto:austinzheng@gmail.com" class="">austinzheng@gmail.com</a>> wrote:</div><br class="Apple-interchange-newline"><div class=""><div dir="ltr" class="">FWIW my opinion is that existentials either shouldn't be allowed to stand in for generic type parameters, or Dave's option #1 if they are.<div class=""><br class=""></div><div class="">The implied promise of a generic type parameter T right now is that T always stands for the same concrete type (modulo things like passing in a subclass where a class would do), and likewise for all of T's associated types (T.Foo is always the same type everywhere in the context where T is valid). This is what makes using anything with 'self' requirements in a generic context sound. Allowing existentials to satisfy T would violate that constraint. </div><div class=""><br class=""></div><div class="">Relaxing these semantics would make it too easy to write code that traps at runtime "without the user having to reach" (to paraphrase Jordan from the "Swift philosophy" thread). Anyone who really really wants to write code that is 'compile-time unsound' in this way should have to explicitly type erase using concrete wrappers.</div></div></div></blockquote><div><br class=""></div><div>Yes, exactly.</div><br class=""><blockquote type="cite" class=""><div class=""><div dir="ltr" class=""><div class=""><br class=""></div><div class="">Best,</div><div class="">Austin</div><div class=""><br class=""></div></div><div class="gmail_extra"><br class=""><div class="gmail_quote">On Wed, Jun 8, 2016 at 2:37 PM, Austin Zheng <span dir="ltr" class=""><<a href="mailto:austinzheng@gmail.com" target="_blank" class="">austinzheng@gmail.com</a>></span> wrote:<br class=""><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr" class=""><div class="">We might be talking past each other. I think Matthew is talking about using an existential outside the context of generic functions. For example, something like this should be trap-proof (as long as 'x' is immutable, which it is in this example):</div><div class=""><br class=""></div><div class="">func copySequenceIntoArray(x: Any<Sequence where .Iterator.Element == Int>) -> [Int] {</div><div class=""><span style="white-space:pre-wrap" class=""> </span>var buffer : [Int] = []</div><div class=""> // Stupid implementation to make a point</div><div class=""><span style="white-space:pre-wrap" class=""> </span>var iterator : x.Iterator = x.makeIterator()</div><div class=""><span style="white-space:pre-wrap" class=""> </span>while true {</div><div class=""><span style="white-space:pre-wrap" class=""> </span>let nextItem : Int? = iterator.next()</div><div class=""><span style="white-space:pre-wrap" class=""> </span>if let nextItem = nextItem {</div><div class=""><span style="white-space:pre-wrap" class=""> </span>buffer.append(nextItem)</div><div class=""><span style="white-space:pre-wrap" class=""> </span>} else {</div><div class=""><span style="white-space:pre-wrap" class=""> </span>return buffer</div><div class=""><span style="white-space:pre-wrap" class=""> </span>}</div><div class=""><span style="white-space:pre-wrap" class=""> </span>}</div><div class="">}</div><div class=""><br class=""></div><div class="">Even this would never trap as well:</div><div class=""><br class=""></div><div class="">func copySequenceIntoArray<T>(x: Any<Sequence where .Iterator.Element == T>) -> [T] {</div><div class=""><span style="white-space:pre-wrap" class=""> </span>var buffer : [T] = []</div><div class=""><span style="white-space:pre-wrap" class=""> </span>for item in x {</div><div class=""><span style="white-space:pre-wrap" class=""> </span>buffer.append(item)</div><div class=""><span style="white-space:pre-wrap" class=""> </span>}</div><div class=""><span style="white-space:pre-wrap" class=""> </span>return buffer</div><div class="">}</div><div class=""><br class=""></div><div class="">Where we run into difficulty is something like this (forgive my abuse of the collections API; I don't remember all the new indexing APIs off the top of my head):</div><div class=""><br class=""></div><div class="">func doSomething<T : Collection where T.Element == Int>(x: T, y: T) {</div><div class=""><span style="white-space:pre-wrap" class=""> </span>// Get indexes out of x and use them to index into y</div><div class=""><span style="white-space:pre-wrap" class=""> </span>var idx = x.startIndex</div><div class=""><span style="white-space:pre-wrap" class=""> </span>while (idx != x.endIndex || idx != y.endIndex) {</div><div class=""><span style="white-space:pre-wrap" class=""> </span>print(x[idx])</div><div class=""><span style="white-space:pre-wrap" class=""> </span>print(y[idx])</div><div class=""><span style="white-space:pre-wrap" class=""> </span>idx = x.nextIndex(idx)</div><div class=""><span style="white-space:pre-wrap" class=""> </span>}</div><div class="">}</div><div class="">let someSeq : Any<Collection where .Element == Int> = // ...</div><div class="">let anotherSeq : Any<Collection where .Element == Int> = // ...</div><div class="">// Trouble!</div><div class="">// someSeq and anotherSeq are the same existential type</div><div class="">// But the concrete index types within each of the existential variables may be different</div><div class="">doSomething(someSeq, anotherSeq)</div><div class=""><br class=""></div><div class="">It's this situation (using an existential type to fulfill a generic type parameter constrained to the same requirements that comprise that existential) that requires either of the two options that Dave presented, due to our lack of compile-time type information about the fulfilling type's associated types.</div><div class=""><br class=""></div><div class="">Best,</div><div class="">Austin</div></div><div class="HOEnZb"><div class="h5"><div class="gmail_extra"><br class=""><div class="gmail_quote">On Wed, Jun 8, 2016 at 2:33 PM, Matthew Johnson via swift-evolution <span dir="ltr" class=""><<a href="mailto:swift-evolution@swift.org" target="_blank" class="">swift-evolution@swift.org</a>></span> wrote:<br class=""><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><br class="">
<br class="">
Sent from my iPad<br class="">
<span class=""><br class="">
> On Jun 8, 2016, at 3:16 PM, Dave Abrahams via swift-evolution <<a href="mailto:swift-evolution@swift.org" target="_blank" class="">swift-evolution@swift.org</a>> wrote:<br class="">
><br class="">
><br class="">
>> on Wed Jun 08 2016, Thorsten Seitz <<a href="mailto:swift-evolution@swift.org" target="_blank" class="">swift-evolution@swift.org</a>> wrote:<br class="">
>><br class="">
>> Ah, thanks, I forgot! I still consider this a bug, though (will have<br class="">
>> to read up again what the reasons are for that behavior).<br class="">
><br class="">
> Yes, but in the case of the issue we're discussing, the choices are:<br class="">
><br class="">
> 1. Omit from the existential's API any protocol requirements that depend<br class="">
> on Self or associated types, in which case it *can't* conform to<br class="">
> itself because it doesn't fulfill the requirements.<br class="">
<br class="">
</span>They don't need to be omitted. They are exposed in different ways depending on how the existential is constrained. Austin's proposal was originally written to omit some members but it was modified based on feedback from Doug Gregor IIRC (Austin, is that right?). Now it contains examples showing how these members are made available in a safe way. Some members may still not be usable because you can't form an argument but IIRC the suggestion was that they be exposed anyway for consistency.<br class="">
<div class=""><div class=""><br class="">
><br class="">
> 2. Erase type relationships and trap at runtime when they don't line up.<br class="">
><br class="">
> Matthew has been arguing against #2, but you can't “fix the bug” without<br class="">
> it.<br class="">
><br class="">
>><br class="">
>> -Thorsten<br class="">
>><br class="">
>>> Am 08.06.2016 um 21:43 schrieb Austin Zheng <<a href="mailto:austinzheng@gmail.com" target="_blank" class="">austinzheng@gmail.com</a>>:<br class="">
>>><br class="">
>>> It's not possible, even with Swift's current implementation of<br class="">
>>> existentials. A protocol type P isn't considered to conform to<br class="">
>>> itself, thus the following is rejected:<br class="">
>>><br class="">
>>> let a : MyProtocol = // ...<br class="">
>>> func myFunc<T : MyProtocol>(x: T) {<br class="">
>>> // ....<br class="">
>>> }<br class="">
>>> myFunc(a) // "Cannot invoke 'myFunc' with an argument list of type MyProtocol"<br class="">
>>><br class="">
>>> Changing how this works is probably worth a proposal by itself.<br class="">
>>><br class="">
>>> Austin<br class="">
>>><br class="">
>>><br class="">
>>> On Wed, Jun 8, 2016 at 12:34 PM, Thorsten Seitz via swift-evolution<br class="">
>>> <<a href="mailto:swift-evolution@swift.org" target="_blank" class="">swift-evolution@swift.org</a><br class="">
>>> <mailto:<a href="mailto:swift-evolution@swift.org" target="_blank" class="">swift-evolution@swift.org</a>>><br class="">
>>> wrote:<br class="">
>>><br class="">
>>>> Am 08.06.2016 um 20:33 schrieb Dave Abrahams via swift-evolution<br class="">
>>>> <<a href="mailto:swift-evolution@swift.org" target="_blank" class="">swift-evolution@swift.org</a><br class="">
>>>> <mailto:<a href="mailto:swift-evolution@swift.org" target="_blank" class="">swift-evolution@swift.org</a>>>:<br class="">
>>>><br class="">
>>>><br class="">
>>>> on Tue Jun 07 2016, Matthew Johnson <<a href="http://matthew-at-anandabits.com" class="">matthew-AT-anandabits.com</a>> wrote:<br class="">
>>>><br class="">
>>>>>> On Jun 7, 2016, at 9:15 PM, Dave Abrahams<br class="">
>>>>>> <<a href="mailto:dabrahams@apple.com" target="_blank" class="">dabrahams@apple.com</a><br class="">
>>>>>> <mailto:<a href="mailto:dabrahams@apple.com" target="_blank" class="">dabrahams@apple.com</a>>><br class="">
>>>>>> wrote:<br class="">
>>>>>><br class="">
>>>>>><br class="">
>>>>>> on Tue Jun 07 2016, Matthew Johnson <<a href="http://matthew-at-anandabits.com" class="">matthew-AT-anandabits.com</a><br class="">
>>>>>> <<a href="http://matthew-at-anandabits.com/" rel="noreferrer" target="_blank" class="">http://matthew-at-anandabits.com/</a><br class="">
>>>>>> <<a href="http://matthew-at-anandabits.com/" rel="noreferrer" target="_blank" class="">http://matthew-at-anandabits.com/</a>>>> wrote:<br class="">
>>>>><br class="">
>>>>>>>> On Jun 7, 2016, at 4:13 PM, Dave Abrahams via swift-evolution<br class="">
>>>>>>>> <<a href="mailto:swift-evolution@swift.org" target="_blank" class="">swift-evolution@swift.org</a><br class="">
>>>>>>>> <mailto:<a href="mailto:swift-evolution@swift.org" target="_blank" class="">swift-evolution@swift.org</a>>><br class="">
>>>>>>>> wrote:<br class="">
>>>>>>>><br class="">
>>>>>>>><br class="">
>>>>>>>> on Tue Jun 07 2016, Matthew Johnson<br class="">
>>>>>>>> <<a href="mailto:swift-evolution@swift.org" target="_blank" class="">swift-evolution@swift.org</a><br class="">
>>>>>>>> <mailto:<a href="mailto:swift-evolution@swift.org" target="_blank" class="">swift-evolution@swift.org</a>>><br class="">
>>>>>>>> wrote:<br class="">
>>>>>>><br class="">
>>>>>>>>>> , but haven't realized<br class="">
>>>>>>>>>> that if you step around the type relationships encoded in Self<br class="">
>>>>>>>>>> requirements and associated types you end up with types that appear to<br class="">
>>>>>>>>>> interoperate but in fact trap at runtime unless used in exactly the<br class="">
>>>>>>>>>> right way.<br class="">
>>>>>>>>><br class="">
>>>>>>>>> Trap at runtime? How so? Generalized existentials should still be<br class="">
>>>>>>>>> type-safe.<br class="">
>>>>>>>><br class="">
>>>>>>>> There are two choices when you erase static type relationships:<br class="">
>>>>>>>><br class="">
>>>>>>>> 1. Acheive type-safety by trapping at runtime<br class="">
>>>>>>>><br class="">
>>>>>>>> FloatingPoint(3.0 as Float) + FloatingPoint(3.0 as Double) // trap<br class="">
>>>>>>>><br class="">
>>>>>>>> 2. Don't expose protocol requirements that involve these relationships,<br class="">
>>>>>>>> which would prevent the code above from compiling and prevent<br class="">
>>>>>>>> FloatingPoint from conforming to itself.<br class="">
>>>>>>>><br class="">
>>>>>>>>> Or are you talking about the hypothetical types / behaviors people<br class="">
>>>>>>>>> think they want when they don’t fully understand what is happening...<br class="">
>>>>>>>><br class="">
>>>>>>>> I don't know what you mean here. I think generalized existentials will<br class="">
>>>>>>>> be nice to have, but I think most people will want them to do something<br class="">
>>>>>>>> they can't possibly do.<br class="">
>>>>>>><br class="">
>>>>>>> Exactly. What I meant is that people think they want that expression<br class="">
>>>>>>> to compile because they don’t understand that the only thing it can do<br class="">
>>>>>>> is trap. I said “hypothetical” because producing a compile time error<br class="">
>>>>>>> rather than a runtime trap is the only sane thing to do. Your comment<br class="">
>>>>>>> surprised me because I can’t imagine we would move forward in Swift<br class="">
>>>>>>> with the approach of trapping.<br class="">
>>>>>><br class="">
>>>>>> I would very much like to be able to create instances of “Collection<br class="">
>>>>>> where Element == Int” so we can throw away the wrappers in the stdlib.<br class="">
>>>>>> That will require some type mismatches to be caught at runtime via<br class="">
>>>>>> trapping.<br class="">
>>>>><br class="">
>>>>> For invalid index because the existential accepts a type erased index?<br class="">
>>>><br class="">
>>>> Exactly.<br class="">
>>>><br class="">
>>>>> How do you decide where to draw the line here? It feels like a very<br class="">
>>>>> slippery slope for a language where safety is a stated priority to<br class="">
>>>>> start adopting a strategy of runtime trapping for something as<br class="">
>>>>> fundamental as how you expose members on an existential.<br class="">
>>>><br class="">
>>>> If you don't do this, the alternative is that “Collection where Element<br class="">
>>>> == Int” does not conform to Collection. That's weird and not very<br class="">
>>>> useful. You could expose all the methods that were on protocol<br class="">
>>>> extensions of Collection on this existential, unless they used<br class="">
>>>> associated types other than the element type. But you couldn't pass the<br class="">
>>>> existential to a generic function like<br class="">
>>>><br class="">
>>>> func scrambled<C: Collection>(_ c: C) -> [C.Element]<br class="">
>>><br class="">
>>> I don’t understand. Why couldn’t an existential be passed to that function?<br class="">
>>><br class="">
>>> -Thorsten<br class="">
>>><br class="">
>>><br class="">
>>><br class="">
>>>><br class="">
>>>>> IMO you should *have* to introduce unsafe behavior like that manually.<br class="">
>>>><br class="">
>>>> Collection where Element == Int & Index == *<br class="">
>>>><br class="">
>>>> ?<br class="">
>>>><br class="">
>>>>> Collection indices are already something that isn’t fully statically<br class="">
>>>>> safe so I understand why you might want to allow this.<br class="">
>>>><br class="">
>>>> By the same measure, so are Ints :-)<br class="">
>>>><br class="">
>>>> The fact that a type's methods have preconditions does *not* make it<br class="">
>>>> “statically unsafe.”<br class="">
>>>><br class="">
>>>>> But I don’t think having the language's existentials do this<br class="">
>>>>> automatically is the right approach. Maybe there is another approach<br class="">
>>>>> that could be used in targeted use cases where the less safe behavior<br class="">
>>>>> makes sense and is carefully designed.<br class="">
>>>><br class="">
>>>> Whether it makes sense or not really depends on the use-cases. There's<br class="">
>>>> little point in generalizing existentials if the result isn't very useful.<br class="">
>>>> The way to find out is to take a look at the examples we currently have<br class="">
>>>> of protocols with associated types or Self requirements and consider<br class="">
>>>> what you'd be able to do with their existentials if type relationships<br class="">
>>>> couldn't be erased.<br class="">
>>>><br class="">
>>>> We have known use-cases, currently emulated in the standard library, for<br class="">
>>>> existentials with erased type relationships. *If* these represent the<br class="">
>>>> predominant use cases for something like generalized existentials, it<br class="">
>>>> seems to me that the language feature should support that. Note: I have<br class="">
>>>> not seen anyone build an emulation of the other kind of generalized<br class="">
>>>> existential. My theory: there's a good reason for that :-).<br class="">
>>>><br class="">
>>>> --<br class="">
>>>> Dave<br class="">
>>>> _______________________________________________<br class="">
>>>> swift-evolution mailing list<br class="">
>>>> <a href="mailto:swift-evolution@swift.org" target="_blank" class="">swift-evolution@swift.org</a><br class="">
>>>> <mailto:<a href="mailto:swift-evolution@swift.org" target="_blank" class="">swift-evolution@swift.org</a>><br class="">
>>>> <a href="https://lists.swift.org/mailman/listinfo/swift-evolution" rel="noreferrer" target="_blank" class="">https://lists.swift.org/mailman/listinfo/swift-evolution</a><br class="">
>>>> <<a href="https://lists.swift.org/mailman/listinfo/swift-evolution" rel="noreferrer" target="_blank" class="">https://lists.swift.org/mailman/listinfo/swift-evolution</a>><br class="">
>>><br class="">
>>> _______________________________________________<br class="">
>>> swift-evolution mailing list<br class="">
>>> <a href="mailto:swift-evolution@swift.org" target="_blank" class="">swift-evolution@swift.org</a><br class="">
>>> <mailto:<a href="mailto:swift-evolution@swift.org" target="_blank" class="">swift-evolution@swift.org</a>><br class="">
>>> <a href="https://lists.swift.org/mailman/listinfo/swift-evolution" rel="noreferrer" target="_blank" class="">https://lists.swift.org/mailman/listinfo/swift-evolution</a><br class="">
>>> <<a href="https://lists.swift.org/mailman/listinfo/swift-evolution" rel="noreferrer" target="_blank" class="">https://lists.swift.org/mailman/listinfo/swift-evolution</a>><br class="">
>><br class="">
>> _______________________________________________<br class="">
>> swift-evolution mailing list<br class="">
>> <a href="mailto:swift-evolution@swift.org" target="_blank" class="">swift-evolution@swift.org</a><br class="">
>> <a href="https://lists.swift.org/mailman/listinfo/swift-evolution" rel="noreferrer" target="_blank" class="">https://lists.swift.org/mailman/listinfo/swift-evolution</a><br class="">
><br class="">
> --<br class="">
> Dave<br class="">
><br class="">
> _______________________________________________<br class="">
> swift-evolution mailing list<br class="">
> <a href="mailto:swift-evolution@swift.org" target="_blank" class="">swift-evolution@swift.org</a><br class="">
> <a href="https://lists.swift.org/mailman/listinfo/swift-evolution" rel="noreferrer" target="_blank" class="">https://lists.swift.org/mailman/listinfo/swift-evolution</a><br class="">
<br class="">
_______________________________________________<br class="">
swift-evolution mailing list<br class="">
<a href="mailto:swift-evolution@swift.org" target="_blank" class="">swift-evolution@swift.org</a><br class="">
<a href="https://lists.swift.org/mailman/listinfo/swift-evolution" rel="noreferrer" target="_blank" class="">https://lists.swift.org/mailman/listinfo/swift-evolution</a><br class="">
</div></div></blockquote></div><br class=""></div>
</div></div></blockquote></div><br class=""></div>
</div></blockquote></div><br class=""></body></html>