[swift-evolution] ternary operator ?: suggestion

Alex Lew alexl.mail+swift at gmail.com
Sun Dec 6 14:24:29 CST 2015


Thanks for the feedback, Matthew. It's sensible to me to consider dropping
the ternary operator. I like it because the analogy "C's if is to Swift's
if as C's ternary operator is to Swift's ternary operator" is (sort of)
satisfied. But it is also confusing, both for the reasons you mention, and
because ? has other meanings in Swift:

 // compiler error without space betw thatColor and ?
let thisColor = thatColor?
     case .Red: .Green
     default: .Blue

On the other hand, is it really worth it to have control flow expressions
if they don't let your code look nicer?

let thisColor = switch thatColor {
     case .Red:
          return .Green;
     default:
          return .Yellow;
}

really isn't much nicer than

let thisColor: Color
switch thatColor {
     case .Red:
          thisColor = .Green
     default:
          thisColor = .Yellow
}

Maybe we could do a compromise, something like

let thisColor = switch thatColor
     case .Red: .Green // must be an expression
     default: .Yellow      // must be an expression

Or we could introduce a new keyword? Like *match*:

let thisColor = match thatColor
      case .Red: .Green    // must be an expression
      default: .Yellow         // must be an expression


I sort of like the new-keyword approach, because even though this is
similar to a switch, it's not a switch: there's no fallthrough, you can't
put statements inside, etc.

The problem with all these proposals:

let thisColor = match thatColor
     case .Red: match thatOtherColor
                           case .Blue: .Green
                           case .Pink: .Yellow
                           default: .Orange
     default: .Orange

is ambiguous. (Does case .Pink match thatColor or thatOtherColor? We can
know because of exhaustiveness checking, but this won't always work.) You
could solve this problem either by using parentheses around the whole
expression when necessary

let thisColor = match thatColor
     case .Red: (match thatOtherColor
                           case .Blue: .Green
                           case .Pink: .Yellow
                           default: .Orange)
     default: .Orange

or by adding curly braces in again

let thisColor = match thatColor {
     case .Red: match thatOtherColor {
                           case .Blue: .Green
                           case .Pink: .Yellow
                           default: .Orange
                        }
     default: .Orange
}

But that starts to look like switch again. (Of course, the best way to
handle this is as a programmer is to just switch on the tuple (thatColor,
thatOtherColor), but the language should allow for nested control
expressions.)


On Sun, Dec 6, 2015 at 2:48 PM, Matthew Johnson via swift-evolution <
swift-evolution at swift.org> wrote:

> I am not a fan of this approach based on the ternary operator.  The
> ternary operator is already a bit of an anomaly in that all other operators
> are unary or binary and do not perform any control flow (beyond possibly
> short circuiting an autoclosure argument).
>
> I would much rather features that perform control flow continue to use
> keywords, but allow them to be expressions.
>
> Once we have control flow expressions I would like to see the ternary
> operator removed from the language as it would no longer server a purpose.
> Removing the ternary operator seems to fit nicely with the direction to
> remove some features that are carried over from C-based languages but don’t
> necessarily fit with the direction Swift is heading.
>
>
> On Dec 6, 2015, at 1:19 PM, Kevin Lundberg via swift-evolution <
> swift-evolution at swift.org> wrote:
>
> Ostensibly, case may not be necessary if you could delimit each case on
> one line with something (perhaps a comma, or something else if that would
> not fit well within the grammar):
>
> let thisColor = thatColor ? .Blue: .Red, .Green: .Blue, .Red: .Green,
> default: .Yellow
>
> On Sun, Dec 6, 2015, at 01:57 PM, Paul Ossenbruggen via swift-evolution
> wrote:
>
> I like this too, seems more powerful.  Also, would single line expressions
> be allowed?  If not would case be required for example:
>
> let myFavoriteColor = yourFavoriteColor ?
>     case .Blue: .Red
>     case .Green: .Blue
>     case .Red: .Green
>     default: .Yellow
>
>
>
> On Dec 6, 2015, at 9:11 AM, Sean Heber via swift-evolution <
> swift-evolution at swift.org> wrote:
>
> I really like this train of thought. +1
>
> l8r
> Sean
>
> On Dec 6, 2015, at 11:02 AM, Alex Lew via swift-evolution <
> swift-evolution at swift.org> wrote:
>
> What if we left the if { ...} else { ... } syntax alone (as a statement),
> and updated the ternary expression to be a more general pattern matching
> expression (closer to "switch")? Something like
>
> let x = condition ?
>    true: "Hello"
>    false: "Goodbye"
>
> let x = optionalValue ?
>    .Some(let unwrapped): "Hello, \(unwrapped)"
>    .None: "To Whom It May Concern"
>
> let myFavoriteColor = yourFavoriteColor ?
>     .Blue: .Red
>     .Green: .Blue
>     .Red: .Green
>
> let quadrant = (x, y) ?
>     let (x, y) where x < 50 && y < 50: "top left"
>     let (x, y) where x < 50 && y > 50: "bottom left"
>     let (x, y) where x > 50 && y < 50: "top right"
>     default: "bottom right"
>
> The colon comes from the fact that this is sort of a light-weight
> expression-based "switch" statement, where each branch can only contain an
> expression, not a series of statements.
>
> This is very similar to pattern matching expressions in languages like
> Haskell, ML, and Coq.
>
> On Sun, Dec 6, 2015 at 11:25 AM, Thorsten Seitz <thorsten.seitz at web.de> wrote:
>
>
>
>
> Am 06.12.2015 um 01:28 schrieb Alex Lew via swift-evolution <
> swift-evolution at swift.org>:
>
> I don't think you can just get rid of the if statement in favor of an
> expression. You still want to be able to do this:
>
> if (condition) {
>     funcWithSideEffectsThatReturnsInt()
> } else {
>     funcWithSideEffectsThatReturnsString()
> }
>
> but that's not a valid expression (what is its type?).
>
>
>
> That would actually be no problem if Swift’s type system would have union
> types (Ceylon has union and intersection types which are quite awesome and
> enable lots of nice things quite naturally, see
> http://ceylon-lang.org/documentation/1.2/tour/types/).
>
> In that case the type of such an expression would just be the union of
> both types, which is written Int | String in Ceylon.
>
>
> -Thorsten
>
>
>
>
>
> _______________________________________________
> swift-evolution mailing list
> swift-evolution at swift.org
> https://lists.swift.org/mailman/listinfo/swift-evolution
>
> _______________________________________________
> swift-evolution mailing list
> swift-evolution at swift.org
> https://lists.swift.org/mailman/listinfo/swift-evolution
>
>
> *_______________________________________________*
> swift-evolution mailing list
> swift-evolution at swift.org
> https://lists.swift.org/mailman/listinfo/swift-evolution
>
>
> _______________________________________________
> swift-evolution mailing list
> swift-evolution at swift.org
> https://lists.swift.org/mailman/listinfo/swift-evolution
>
>
> Untracked with Trackbuster <https://trackbuster.com/?sig>
>
> _______________________________________________
> 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/20151206/aa4516e4/attachment.html>


More information about the swift-evolution mailing list