One major thing to keep in mind is how exactly will it behave in generic
contexts.
As a similar example, if you can do this:
func foo<A, B>(pass parameter: A, to closure: (A) -> B) -> B {
closure(parameter)
}
func bar(one: Int, two: Double, three: String) {
print(“\(one) \(two) \(three)")
}
foo(pass: (1, 2.0, “3”), to: bar)
This used to work without generics, but If I’m not mistaken, a proposal was
accepted to disallow passing tuples and the only argument of a multi-argument
function, leaving this use case kinda magical.
This is an enormously useful feature for generic programming (especially with
the lack of variadic generic types), because this allows us to define
higher-order functions that operate on any type of function, regardless of
their arity.
However we resolve the error issue, it has to behave in a similar way, so that
throwing and non-throwing functions can be dealt with in a uniform way in a
generic context without loss of information or security guarantees.
> On May 2, 2017, at 12:48 AM, Rod Brown <[email protected]> wrote:
>
>
>> On 2 May 2017, at 2:34 am, John McCall <[email protected]
>> <mailto:[email protected]>> wrote:
>>
>>>
>>> On May 1, 2017, at 9:01 AM, Rod Brown via swift-evolution
>>> <[email protected] <mailto:[email protected]>> wrote:
>>> I agree that the key problem with the current architecture that you're
>>> alluding to is it can't be easily stored and transferred. Swift errors are
>>> great for live action but holding and passing after the throwing event is
>>> problematic, and this is an elegant solution. The storage issue is when
>>> holding it as a property, and the transferring issue is when passing it to
>>> a closure as a results of an asynchronous operation etc. These are both
>>> definitely cases where storage of the type-or-error makes perfect sense.
>>>
>>> I think the key problem getting this accepted by the Swift Team will be
>>> that it doesn't currently have any specific use in the standard library. As
>>> a low level set of types, errors are generated by the lower levels but
>>> rarely stored, so the Standard library doesn't need the storage. Generally
>>> the only place we have to do that is in end user code. And currently the
>>> standard library doesn't have to support asynchronous operations natively,
>>> so there's nothing inside the kit that would require it to do completion
>>> handlers with errors.
>>
>> We've definitely considered including a Result type, but our sense was that
>> in an ideal world almost no code would be using it. It's hard to imagine an
>> ordinary API that ought to be returning a Result rather than throwing, and
>> once you've defined that away, the major remaining use case is just to shift
>> computation around, like with a completion handler. That explicit
>> computation-shifting pattern is something we're hoping to largely define
>> away with something like C#'s async/await, which would leave Result as
>> mostly just an implementation detail of such APIs. We didn't want to spend
>> a great deal of time designing a type that would end up being so marginal,
>> especially if the changing role would lead us into different directions on
>> the design itself. We also didn't want to design a type that would become
>> an obstacle to potential future language changes like, say, typed throws.
>>
>> The downside, of course, is that as long as we lack that async/await design,
>> computation-shifting isn't real great.
>>
>> John.
>
> This makes sense and is sensible. I’m curious how such an API would play with
> the existing NSURLSession completion handlers and the like, but I’m sure the
> community can design something appropriate.
>
> I think the only remaining case is simply storing a result-or-error for later
> handling, storage to disk, etc. I agree with your contention that the vast
> majority of the use case for this type is for computation shifting. I think
> it would and should be rare that we would want to “store” as a variable the
> “result-or-error” type. Errors should be handled at runtime in the vast
> majority of cases, presented to the user or otherwise handled, and then moved
> on from, with the reason no longer being relevant.
>
> As you say, in the meantime, it does leave computation-shifting a bit ad-hoc
> and convoluted, but I think the community has standardized on the Result
> temporary solution.
>
>>
>>>
>>> This would therefore be an element in the standard library purely so we
>>> don't have 50,000 different libraries with 50,000 different result types.
>>> I'd love to see this standardised so frameworks were more compatible. I'm
>>> just not sure whether the Core Team would see it as pressing to try and
>>> officiate a certain type that they themselves don't use.
>
_______________________________________________
swift-evolution mailing list
[email protected]
https://lists.swift.org/mailman/listinfo/swift-evolution
