[SR-8401] Generic class, type aliases and type inference at compile for a static function on a generic class

[allenhumphreys]

Previous ID	SR-8401
Radar	None
Original Reporter	@allenhumphreys
Type	Bug
Status	Resolved
Resolution	Duplicate

Environment

Swift 4.1 Xcode 9.4.1

Additional Detail from JIRA

Votes	0
Component/s	Compiler
Labels	Bug
Assignee	None
Priority	Medium

md5: cd7c569a6c30091ff8deac1ccd6486bd

duplicates:

• [SR-1789] Swift can't infer types in generic methods that reuse part of the struct/class' parameters #44398

Issue Description:

I'm not sure if this is a bug, but it kind of threw me for a loop so I thought I'd pose the question.

class CallWrapper<T, U: Decodable> {

    static func function() -> CallWrapper<Float, Float> {
        return CallWrapper() // Compiler Error: Cannot convert return expression of type 'SubCall<T>' to return type 'SubCall<Float>'
        // Fixit: Insert ' as! SubCall<Float>'
    }

    typealias TypeAlias<V, W: Decodable> = CallWrapper<V, W>

    static func function2() -> TypeAlias<Float, Float> {
        return TypeAlias()
    }
}

With function, I kind of expect type inference to pick up the return type of the function and use it for the initializer. But I understand kind of why it maybe shouldn't or can't because the type of the initializer in the static context is determined at the call site, i.e. CallWrapper<Int, Int>.function().

Things got strange for me, though, when I made a type alias, as in function2, where I was able to use type inference to only specify the complete type as the return of the function. My goal is convenience because in my case the types are pretty long and using the inference makes it easy to change the code.

I don't understand why type inference could work with the alias but not with the bare type, and was curious if one or the other is a bug.

[belkadan]

Yep, you can read more about this in #44398.