[swift-users] Very unexpected automatic behaviour between StringLiteralConvertible and pattern matching!

Joe Groff jgroff at apple.com
Wed Jan 6 13:05:31 CST 2016


> On Jan 5, 2016, at 9:28 AM, David Hart via swift-users <swift-users at swift.org> wrote:
> 
> How is it that Swift allows code like this:
> 
> struct Sneaky: StringLiteralConvertible {
> 	init(stringLiteral value: String) {}
> 	init(extendedGraphemeClusterLiteral value: String) {}
> 	init(unicodeScalarLiteral value: String) {}
> }
> 
> func ~=(sneaky: Sneaky, string: String) -> Bool {
> 	return false
> }
> 
> enum NormalEnum: String {
> 	case Super = "super"
> 	case Mario = "mario"
> }
> 
> let value = NormalEnum(rawValue: "super”) // return nil!!!!

I see two bugs here. When an enum has a raw value type, the compiler generates this initializer:

init(rawValue: String) {
  switch rawValue {
  case "super":
    self = .Super
  ...
  }
}

so uses ~= pattern matching to match the raw value. It would be more sensible to always use `==` comparison in the synthesized initializer. However, I'm surprised too that the type checker favors ~=(Sneaky, String) over ~=(String, String); it should at best be ambiguous. Do you have time to file these two bugs?

-Joe

> 
> It hit completely by surprise today because of of a Regex library:
> 
> struct Regex: StringLiteralConvertible {
> 	init(stringLiteral value: String) {}
> 	init(extendedGraphemeClusterLiteral value: String) {}
> 	init(unicodeScalarLiteral value: String) {}
> 
> 	//...
> }
> 
> func ~=(regex: Regex, string: String) -> Bool {
> 	return regex.matches(string)
> }
> 
> If I was not already a Swift enthusiast, this behaviour would have left me completely dumbfounded.
> What can we do about it?
> 
> David.
> 
> _______________________________________________
> swift-users mailing list
> swift-users at swift.org
> https://lists.swift.org/mailman/listinfo/swift-users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.swift.org/pipermail/swift-users/attachments/20160106/d830f7dc/attachment.html>


More information about the swift-users mailing list