[swift-evolution] Proposal: Allow Type Annotations on Throws

Alejandro Martinez alexito4 at gmail.com
Thu Dec 24 19:51:07 CST 2015 

Absolutely agree Matthew.
My point was that if this is the temporary tradeoff that we have to support for me is worth it. Having typed throws looks like a great next step to improving the current error system.
Curious to see your ideas about that ;)

Cheers,

Sent from my iPad

> On 24 Dec 2015, at 14:26, Matthew Johnson <matthew at anandabits.com> wrote:
> 
> 
> 
> Sent from my iPad
> 
>> On Dec 24, 2015, at 3:24 AM, Alejandro Martinez <alexito4 at gmail.com> wrote:
>> 
>> Hi!
>> I've been following the conversation closely so here are my 2c
>> 
>> TL;TR: +1
>> 
>> After proposing solutions for the concerns about conversions this is a clear +1. Overloading the conversion functions is how it works in Rust and I think is a good tradeoff, even without the magical "try!" That Rust has. Maybe we could explore something like that when Swift give us a macro like system. But for now I will be happy with this.
> 
> Rust actually uses a trait (the Rust equivalent of a protocol) to do this.  It isn't possible in Swift because protocols cannot have type parameters like traits can in Rust and there is no other mechanism in Swift with equivalent capability.  
> 
> This means that Rust's From mechanism is generic and is part of the standard library.  It is part of the common vocabulary of Rust programmers.  This is a significant advantage over a solution that uses ad-hoc overloading relying on a convention that is likely to vary between teams.  Some teams probably won't even be aware of the possibility to adopt such a convention if it isn't part of our common vocabulary.
> 
> Here is what the Rust documentation has to say about this:
> 
>> From is very useful because it gives us a generic way to talk about conversion from a particular type T to some other type (in this case, “some other type” is the subject of the impl, or Self). The crux of From is the set of implementations provided by the standard library.
> 
> 
> Swift is not Rust and I personally believe a different approach to handling error translation during propagation would be more Swifty for a number of reasons.  I believe that would be the case even if we had macros and protocols with type parameters.  I am continuing to think about this and will probably submit a proposal if / when David's proposal is accepted.
> 
> IMO it is very important that we have a standard mechanism for translating errors during propagation.
> 
> 
>> 
>> The most important thing to remember is that this changes will not affect people that prefers the untyped throws, as they can continue doing it like now. I think this is really important as will allow us to ignore typed errors when writing prototyping code, but it will give the option to improve the safety of our code when is time for that.
>> 
>> Thanks for the proposal and the interesting discussion.
>> 
>> Sent from my iPad
>> 
>> PS: i've been playing on porting the example from Rust errors page to Swift. 
>> http://alejandromp.com/blog/2015/12/23/rust-error-handling-swift-script/
>> Will be interesting to try it again with the changes from this proposal ;)
>> 
>>> On 23 Dec 2015, at 05:58, Thorsten Seitz via swift-evolution <swift-evolution at swift.org> wrote:
>>> 
>>> David's proposal looks good enough for me.
>>> 
>>> With regards to Matthew's worry of cluttering the code with conversion I'd like to remark that this conversion code should only live on the border of your API (the facade), so it should probably not be too invasive to your business logic.
>>> 
>>> -Thorsten 
>>> 
>>> 
>>> 
>>>> Am 22.12.2015 um 20:08 schrieb Matthew Johnson via swift-evolution <swift-evolution at swift.org>:
>>>> 
>>>> 
>>>>>> On Dec 22, 2015, at 12:09 PM, David Owens II <david at owensd.io> wrote:
>>>>>> 
>>>>>> 
>>>>>> On Dec 22, 2015, at 9:50 AM, Matthew Johnson <matthew at anandabits.com> wrote:
>>>>>> 
>>>>>> Unfortunately, I don’t see a way to make it safe.  You had to use fatalError in a default case to make it work.  An alternative would have been to include an ‘UnknownError’ case in ‘PublishedError’.  Neither is not an acceptable solution IMO.
>>>>> 
>>>>> I need the fatalError() because the sample is a working example in Swift today. If we had typed errors, this would simply work:
>>>>> 
>>>>>    static func from<T>(@autoclosure fn: () throws InternalError -> T) throws PublishedError -> T {
>>>>>         do {
>>>>>             return try fn()
>>>>>         }
>>>>>         catch InternalError.Internal(let value) {
>>>>>             throw PublishedError.Converted(value: value)
>>>>>         }
>>>>>     }
>>>>> 
>>>>> This states that the only closure accepted is one that throws an InternalError. 
>>>> 
>>>> Ok, so you suggest writing a specific overload for each combination of error types that are convertible.  Got it.  Not sure why I didn’t think of overloads.  I was too focused on a general try function dispatching to an initializer.
>>>> 
>>>> That is indeed safe and I can live with it.  Thanks for taking the time to work through these examples with me and help to identify patterns that address my concerns!
>>>> 
>>>> I still find it unfortunate that this is in the realm of a pattern though.  IMO it would be much better if it was part of the common Swift vocabulary, either as a language feature or a library function but that isn’t possible as a generic implementation isn’t possible.
>>>> 
>>>>> 
>>>>>> This top level `from` example also brings up a couple of points that I don’t recall being addressed in your proposal.  Specifically, the interaction of typed errors with generics and type inferrence.
>>>>> 
>>>>> I call out in the proposal that errors work with generics no differently than other types.
>>>> 
>>>> Great, I must have missed that.
>>>> 
>>>>> 
>>>>>> I still consider this to be an unresolved concern.  I would like to have a safe way to perform error conversion during propagation without cluttering up my control flow and seriously degrading readability.  This is a problem that can and has been solved in other languages.  IMO it is should be considered an essential element of a proposal introducing typed errors.
>>>>> 
>>>>> When Swift has a macro as powerful as Rust, then this is a solved problem as well. However, Swift isn’t there yet. 
>>>> 
>>>> I would prefer a solution to this that didn’t require macros which would fit better in Swift.  This feature is buried in the `try!` macro in Rust as Rust doesn’t have built-in language level error handling support.  
>>>> 
>>>> Swift already has `try` built into the language.  IMO it would be better to have it handled by the built-in language level error handling support in Swift.  That seems like the more “Swifty” approach.  We could have a Swift macro `tryAndConvert` or something, but that seems inelegant.
>>>> 
>>>> We’ve gone back and forth on this quite a bit but nobody else has chimed in.  I’m curious to hear what others thing.  I would love it if any lurkers would jump in and comment!
>>>> 
>>>> Matthew
>>>> 
>>>> _______________________________________________
>>>> 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/20151225/d7c73f92/attachment.html>

More information about the swift-evolution mailing list