[4.1][stdlib] Eliminate IndexDistance #13344
Conversation
* Eradicate IndexDistance associated type, replacing with Int everywhere * Consistently use Int for ExistentialCollection’s IndexDistance type. * Fix test for IndexDistance removal * Remove a handful of no-longer-needed explicit types * Add compatibility shims for non-Int index distances * Test compatibility shim * Move IndexDistance typealias into the Collection protocol
|
@swift-ci please test |
|
@swift-ci please benchmark |
|
@swift-ci please test source compatibility |
|
@swift-ci please smoke compiler performance |
| @@ -782,6 +776,9 @@ public protocol Collection: Sequence where SubSequence: Collection { | |||
| /// - Parameter i: A valid index of the collection. `i` must be less than | |||
| /// `endIndex`. | |||
| func formIndex(after i: inout Index) | |||
|
|
|||
| @available(swift, deprecated, message: "all index distances are now of type Int") | |||
| typealias IndexDistance = Int | |||
There was a problem hiding this comment.
public
UPD: Oh wait, it's a protocol. Nevermind.
There was a problem hiding this comment.
Yup. I moved it back and forth between the protocol and the extension multiple times and forgot this every. single. time.
| } | ||
| */ | ||
| @available(*, deprecated, message: "all index distances are now of type Int") | ||
| public func formIndex<T: BinaryInteger>(_ i: inout Index, offsetBy n: T, limitedBy limit: Index) -> Bool { |
There was a problem hiding this comment.
I think formIndex should have a @discardableResult attribute, otherwise it will issue 2 warnings.
There was a problem hiding this comment.
eh, they're both legit warnings though
There was a problem hiding this comment.
Take that back. Turns out we do not have @discardableResult on this method anywhere...
| return formIndex(&i, offsetBy: Int(n), limitedBy: limit) | ||
| } | ||
| @available(*, deprecated, message: "all index distances are now of type Int") | ||
| public func distance<T: BinaryInteger>(from start: Index, to end: Index) -> T { |
There was a problem hiding this comment.
Surprised it does not result in ambiguity in expressions like let d = xs.distance(from: someIndex, to: someOtherIndex).
There was a problem hiding this comment.
The other distance method isn't generic (T is Int) so it wins.
There was a problem hiding this comment.
Isn't generic other than on the protocol, that is...
| @@ -2898,6 +2898,9 @@ public struct ${Self} | |||
| : FixedWidthInteger, ${Unsigned}Integer, | |||
| _ExpressibleByBuiltinIntegerLiteral { | |||
|
|
|||
| public typealias IntegerLiteralType = ${Self} | |||
There was a problem hiding this comment.
Hum. It isn't. I probably meant to do this on my clean-up branch. But since this is a cherry-pick from master where it's merged, probably should just leave it in.
|
@swift-ci please nominate |
Build comment file:Build failed before running benchmark. |
IndexDistanceassociated type, implementing SE-191.IndexDistancewithInt. It also introduces some compatibility shims (anIndexDistancetypealias equal toInt, and generic conversions for calls using non-Intindex distances).Intis not already used, and the majority of changes are simplifications of the code to no longer need to perform casts between different distance types.