[swift-evolution] Changing postfix "self" to something clearer like "type"

Taras Zakharko taras.zakharko at uzh.ch
Thu Dec 17 06:03:43 CST 2015


(Sorry, I pressed the send button accidentally)

What I wanted to say is that maybe one can have a more transparent system, e.g. typeOf(x) that always returns type of x? For example, typeOf(3) can be Int, type of typeOf(Int) will be the metatype of Int, etc. Similarly, to refer to a type, one should just be able to use the type name. Does anyone know the reason why the system was set up like it is in the first place? 

Best, 

 Taras

> On 17 Dec 2015, at 13:00, Taras Zakharko <taras.zakharko at uzh.ch> wrote:
> 
> I also find this aspect of Swift very confusing. I have worked a lot with metatypes (and metaprogramming in general) in Python and R, but I can’t really wrap my head around how this works in Swift. 
> 
>> On 17 Dec 2015, at 11:12, Gwendal Roué via swift-evolution <swift-evolution at swift.org <mailto:swift-evolution at swift.org>> wrote:
>> 
>> I have two questions:
>> 
>> 1. why do we need the postfix self?
>> 2. why is the postfix self sometimes required, and sometimes not? (see sample code below):
>> 
>> 	func f1<T>(type: T.Type) { print(type) }
>> 	f1(Int)                     // Int
>> 
>> 	func f2<T>(type: T.Type, extra: Bool) { print(type) }
>> 	f2(Int, extra: true)        // Compiler error
>> 	f2(Int.self, extra: true)   // Int
>> 
>> Gwendal Roué
>> 
>> 
>>> Le 15 déc. 2015 à 22:29, Marc Knaup via swift-evolution <swift-evolution at swift.org <mailto:swift-evolution at swift.org>> a écrit :
>>> 
>>> The word "Type" is also blocked in most cases. So still being able to use the word "type" is less valuable anyway because you cannot create a nested type called "Type" and will most likely rename it to something like "Kind".
>>> 
>>> This is actually very annoying:
>>> 
>>> struct Attachment {
>>> 
>>> 	let type: Type
>>> 
>>> 
>>> 	enum Type {
>>> 		case Image
>>> 		case Video
>>> 	}
>>> }
>>> 
>>> let attachmentType = Attachment.Type // error: expected member name or constructor call after type name
>>> 
>>> 
>>> 
>>> 
>>> On Tue, Dec 15, 2015 at 7:24 PM, Jordan Rose via swift-evolution <swift-evolution at swift.org <mailto:swift-evolution at swift.org>> wrote:
>>> ".self" was chosen for a few reasons:
>>> 
>>> - The obvious choice was ".class", given precedent in Objective-C and Java, but not all types are classes.
>>> - 'type' is a very common property name, so we have tried very hard to avoid taking it as a general keyword.
>>> - 'type' also always implies going up a level. "obj.dynamicType" gives you back the type of 'obj', so wouldn't "SomeClass.type" give you back the metaclass <http://sealiesoftware.com/blog/archive/2009/04/14/objc_explain_Classes_and_metaclasses.html>? (Alternately, "SomeType.staticType' not being the same as 'SomeType.dynamicType" seems weird.)
>>> - 'self' is already a keyword.
>>> - ".self" actually works in Objective-C as well.
>>> - ".self" currently also applies to instances, doing exactly what you think it does. This is nearly useless. In theory you could use it to unwrap one level of optionality ("doubleOpt?.self") but that doesn't actually work today.
>>> 
>>> I read "SomeType.self" as "SomeType itself, rather than an instance of it (or associated type)".
>>> 
>>> (And before someone brings it up, we chose not to just allow "SomeType" on its own because "let x = SomeType" is a likely typo for "let x: SomeType".)
>>> 
>>> I think coming up with a clearer name is possible here, but there's plenty to consider. Still, certainly a reasonable thing to bring up.
>>> 
>>> Best,
>>> Jordan
>>> 
>>>> On Dec 15, 2015, at 8:42 , Brandon Knope via swift-evolution <swift-evolution at swift.org <mailto:swift-evolution at swift.org>> wrote:
>>>> 
>>>> Doh! staticType is the obvious choice!
>>>> 
>>>> I agree that adding more keywords can be bad, but in this case I think the clarity outweighs any downside:
>>>> 
>>>> SomeType.staticType
>>>> SomeType.self 
>>>> 
>>>> To me (and I'm sure many others) one is vastly more obvious and easier to understand. 
>>>> 
>>>> I still don't really understand what SomeType.self is trying to convey upon first glance
>>>> 
>>>> Brandon 
>>>> 
>>>> Sent from my iPad
>>>> 
>>>> On Dec 15, 2015, at 11:34 AM, Dennis Lysenko <dennis.s.lysenko at gmail.com <mailto:dennis.s.lysenko at gmail.com>> wrote:
>>>> 
>>>>> +1. Side effects can be eliminated through code migration if a suitable property name is chosen. Perhaps `staticType` to continue in the vein of `dynamicType`? 
>>>>> 
>>>>> Main detractor is that creating more keywords isn't necessarily a good thing.
>>>>> 
>>>>> On Tue, Dec 15, 2015 at 11:19 AM Brandon Knope via swift-evolution <swift-evolution at swift.org <mailto:swift-evolution at swift.org>> wrote:
>>>>> One area of swift that is really not clear to me is when you want to use the type of a class, struct, enum, etc as a value. 
>>>>> 
>>>>> Metatyping is explained here: https://developer.apple.com/library/ios/documentation/Swift/Conceptual/Swift_Programming_Language/Types.html#//apple_ref/doc/uid/TP40014097-CH31-XID_1022 <https://developer.apple.com/library/ios/documentation/Swift/Conceptual/Swift_Programming_Language/Types.html#//apple_ref/doc/uid/TP40014097-CH31-XID_1022>
>>>>> 
>>>>> Example: 
>>>>> let metatype: SomeClass.Type = SomeClass.self
>>>>> 
>>>>> Is there a reason why this isn't SomeClass.type? Everywhere in the document this is explained as returning the type yet it's using a postfix self to access the type. 
>>>>> 
>>>>> I propose changing the postfix self to something more obvious like "type"
>>>>> 
>>>>> Going back to the example:
>>>>> let metatype: SomeClass.Type = SomeClass.type
>>>>> 
>>>>> Several reasons why I think this is better:
>>>>> 1. Postfix self is not obvious as an option as you never see a postfix self anywhere else 
>>>>> 2. "self" does not clearly explain that the type is being returned 
>>>>> 3. ObjC programmers are familiar with accessing the class type by sending the "class" method to the class type. In this case it needs to work on structs and enums, so a "type" method would make more sense. 
>>>>> 4. Instances have a dynamicType method. For consistency, classes, structs, etc., should have a type method
>>>>> 
>>>>> Any other suggestions would be welcome. 
>>>>> 
>>>>> Brandon 
>>>>> 
>>>>> Sent from my iPad
>>>>> 
>>>>> _______________________________________________
>>>>> 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>
>>> 
>>> _______________________________________________
>>> 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>
>>  _______________________________________________
>> 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/20151217/7a1c096d/attachment.html>


More information about the swift-evolution mailing list