[swift-evolution] [Proposal Draft] Literal Syntax Protocols

[Nate Cook]

I like the namespace-based approach to group these protocols together and I'm very happy to see some clarification happening in this group of protocols. However, I don't think the proposed new names communicate what they need to. The names listed in the "Alternatives Considered" section do a better job of describing the use and behavior of these protocols.

Primarily, the new names read like we're saying that a conforming type is a literal, compounding a common existing confusion between literals and types that can be initialized with a literal. Swift's type inference can sometimes make it seem like dark magic is afoot with literal conversions—for example, you need to understand an awful lot about the standard library to figure out why line 1 works here but not line 2:

    var x = [1, 2, 3, 4, 5]
    let y = [10, 20]

    x[1..<2] = [10, 20]     // 1
    x[1..<2] = y            // 2

These new names are a (small) step in the wrong direction. While it's true that the type system doesn't have an IntegerLiteral type, the language does have integer literals. If someone reads:

    extension MyInt : Syntax.IntegerLiteral { ... }

the implication is that MyInt is an integer literal, and therefore instances of MyInt should be usable wherever an integer literal is usable. The existing "Convertible" wording may be a red herring, but it at least suggests that there's a difference between a literal and a concrete type.

In sum, I support a change like this and strongly recommend keeping some sort of adjective in the protocol name.

Nate


> On Jun 23, 2016, at 10:31 AM, Matthew Johnson via swift-evolution <swift-evolution at swift.org> wrote:
> 
> I have completed a draft of a proposal to move the `*LiteralConvertible` protocols into a `Syntax` namespace as suggested by the standard library team. 
> 
> The draft can be found here: https://gist.github.com/anandabits/99dad2305d310874bd613b72b14eee56 <https://gist.github.com/anandabits/99dad2305d310874bd613b72b14eee56>.  It is also included below.
> 
> I will submit a PR in the next day or two after incorporating any feedback.
> 
> -Matthew
> 
> Literal Syntax Protocols
> 
> Proposal: SE-NNNN <file:///Users/Matthew/Library/Containers/com.brettterpstra.marked2/Data/Library/Caches/Marked%202/Watchers/NNNN-literal-syntax-protocols.md>
> Author: Matthew Johnson <https://github.com/anandabits>
> Status: Awaiting review
> Review manager: TBD
> Introduction
> 
> This proposal renames the *LiteralConvertible protocols to Syntax.*Literal. 
> 
> Swift-evolution thread: Revisiting SE–0041 Names <http://thread.gmane.org/gmane.comp.lang.swift.evolution/21290>
> Motivation
> 
> The standard library currently has protocols that use the term Convertible in two different ways. The *LiteralConvertible protocols use the meaning of converting from a literal. The Custom(Debug)StringConvertible protocols use the meaning of converting to a String. This causes confusion for developers attempting to name their own protocols following the precedence established by the standard library.
> 
> Further, the standard library team has observed:
> 
> The "literal" protocols are not about conversion, they are about adopting a certain syntax provided by the language.  "Convertible" in the name is a red herring: a type can't be convertible from an integer literal because there is no "IntegerLiteral" entity in the type system.  The literal *becomes* typed as the corresponding literal type (e.g., Int or String), and as far as the user at the call site is concerned, there is no visible conversion (even if one is happening behind the scenes)
> An earlier proposal <https://github.com/apple/swift-evolution/blob/master/proposals/0041-conversion-protocol-conventions.md> was intended to address the first problem by introducing strong naming conventions for three kinds of conversion protocols (from, to, and bidirectional). The review highlighted the difficulity in establishing conventions that everyone is happy with. This proposal takes a different approach to solving the problem that originally inspired that proposal while also solving the awkwardness of the current names described by the standard library team.
> 
> Proposed solution
> 
> This proposal addresses both problems by introducing a Syntax “namespace” and moving the *LiteralConvertible protocols into that “namespace” while also renaming them. The proposal does not make any changes to the requirements of the protocols.
> 
> Detailed design
> 
> All of the *LiteralConvertible protocols will receive new *Literal names inside a Syntaxnamespace. 
> 
> This namespace will initially be implemented using a case-less enum, but this detail may change in the future if submodules or namespaces are added to Swift. Swift does not currently allow protocols to be declared inside the scope of a type. In order to work around this limitation the protocols themselves will be declared using underscore-prefixed names internal to the standard library. Typealiases inside the Syntax enum will declare the names intended to be visible to user code.
> 
> This proposal does not change any requirements of these protocols. All requirements of all *LiteralConvertible protocols will remain exactly the same.
> 
> The following protocol declarations and names:
> 
> public protocol NilLiteralConvertible { ... }
> public protocol BooleanLiteralConvertible { ... }
> public protocol FloatLiteralConvertible { ... }
> public protocol IntegerLiteralConvertible { ... }
> public protocol UnicodeScalarLiteralConvertible { ... }
> public protocol ExtendedGraphemeClusterConvertible { ... }
> public protocol StringLiteralLiteralConvertible { ... }
> public protocol StringInterpolationLiteralConvertible { ... }
> public protocol ArrayLiteralConvertible { ... }
> public protocol DictionaryLiteralConvertible { ... }
> Are changed as follows:
> 
> 
> public protocol _NilLiteralSyntax { ... }
> public protocol _BooleanLiteralSyntax { ... }
> public protocol _IntegerLiteralSyntax { ... }
> public protocol _FloatLiteralSyntax { ... }
> public protocol _UnicodeScalarLiteralSyntax { ... }
> public protocol _ExtendedGraphemeClusterLiteralSyntax { ... }
> public protocol _StringLiteralLiteralSyntax { ... }
> public protocol _StringInterpolationLiteralSyntax { ... }
> public protocol _ArrayLiteralSyntax { ... }
> public protocol _DictionaryLiteralSyntax { ... }
> 
> public /* closed */ enum Syntax {
>   public typealias NilLiteral = _NilLiteralSyntax
>   public typealias BooleanLiteral = _BooleanLiteralSyntax
>   public typealias IntegerLiteral = _IntegerLiteralSyntax
>   public typealias FloatLiteral = _FloatLiteralSyntax
>   public typealias UnicodeScalarLiteral = _UnicodeScalarLiteralSyntax
>   public typealias ExtendedGraphemeClusterLiteral = _ExtendedGraphemeClusterLiteralSyntax
>   public typealias StringLiteralLiteral = _StringLiteralLiteralSyntax
>   public typealias StringInterplolationLiteral = _StringInterpolationLiteralSyntax
>   public typealias ArrayrLiteral = _ArrayLiteralSyntax
>   public typealias DictionaryLiteral = _DictionaryLiteralSyntax
> }
> Impact on existing code
> 
> All code that references any of the *LiteralConvertible protocols will need to be modified to reference the protocol via the new Syntax.*Literal name.
> 
> Alternatives considered
> 
> Protocol names
> 
> Several commenters have suggested that the names in this proposal are confusing at the site of use:
> 
> struct Foo: Syntax.IntegerLiteral { ... }
> One alternative naming scheme would emphasize the semantic of initializing the type with a literal:
> 
> struct Foo: Syntax.IntegerLiteralInitializable { ... }
> Discussion of the pros and cons of the proposed and alternative naming schemes is encouraged. The core team should feel free to make a final decision on the exact naming scheme used if they choose to accept this proposal.
> 
> Previous proposal
> 
> This proposal is a follow up to Updating Protocol Naming Conventions for Conversions <https://github.com/apple/swift-evolution/blob/master/proposals/0041-conversion-protocol-conventions.md>. Many related alternatives were explored during the discussion and review of that proposal.
> 
> Acknowledgements
> 
> The design described in this proposal was suggested by Dave Abrahams, Dmitri Gribenko, and Maxim Moiseev.
> 
> _______________________________________________
> swift-evolution mailing list
> swift-evolution at swift.org
> https://lists.swift.org/mailman/listinfo/swift-evolution

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.swift.org/pipermail/swift-evolution/attachments/20160623/68af7044/attachment.html>