How are you auto-casting with just the = sign

[eonist]

IN your sample code you have:

cgFloatValue = doubleValue
doubleValue = intValue
doubleValue = cgFloatValue

But I can't find the auto infer code in your src. Some sort of extension or?

[seivan]

@eonist Sorry, it looks like the Readme is a bit dated. I was relying on the @conversion keyword pre Swift 2. It allowed for implicit conversion on assignment.

The only thing possible today, is what's defined in the test-suite(s). I'll fix the Readme once I migrate over to 3.0.

Sorry for the inconvenience.

[eonist]

So something like this:

let someDouble:Double = 4
let someFloat:CGFloat = someDouble

Won't work in swift 3.0?

[seivan]

Not without some Arithmetic operation, e.g addition. let someFloat:CGFloat = 0 + someInt or let someFloat:CGFloat = otherFloat + someInt or someFloat += someInt

Since we don’t got implicit conversion on assignment, the next best thing is implicit conversion through operators.

On 2016Oct10, at 12:23, Eon [email protected] wrote:

So something like this:

let someInt:Int = 4 let someFloat:CGFloat = someInt Won't work in swift 3.0?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/seivan/ScalarArithmetic/issues/17#issuecomment-252580914, or mute the thread https://github.com/notifications/unsubscribe-auth/AADYgAiUB_sUOYo-gGH5RPXZBc13-Cpnks5qyhIlgaJpZM4KR-Yj.

[eonist]

Right. Would be so awesome to not worry about casting Numerical types. I guess one could make an universal cast method by using inference and generics. like:

let someFloat:Float = 4
let someInt:Int = someFloat.cast()
let someDouble:Double = someFloat.cast()
let someCGFloat:CGFloat = someFloat.cast()

extension Float{
    func <T>cast()->T{
        if(T is Int){return Int(self)}
        else if(T is Double){return Double(self)}
        else if(T is CGFloat){return CGFloat(self)}
        else {fatalError("numerical type not supported")}
    }
}

Better yet. Make a protocol with an extension that every numerical type could extend. So that you only would have to write the extension once. extension Float:NumericalConversional... and extension CGFloat:NumericalConversional...etc

[seivan]

That’s actually been on my mind. But not sure about the implications. So far I’ve resorted to + 0 just to be explicit about it. But maybe cast() is more explicit.

On 2016Oct10, at 12:53, Eon [email protected] wrote:

Right. Would be so awesome to not worry about casting Numerical types. I guess one could make an universal cast method by using inference and generics. like:

let someInt:Int = someFloat.cast() let someDouble:Int = someFloat.cast() let someCGFloat:Int = someFloat.cast()

extension Float{ func <T>cast()->T{ if(T is Int){return Int(self)} else if(T is Double){return Double(self)} else if(T is CGFloat){return CGFloat(self)} else {fatalError("numerical type not supported")} } } — You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/seivan/ScalarArithmetic/issues/17#issuecomment-252586257, or mute the thread https://github.com/notifications/unsubscribe-auth/AADYgECRHqfA0KOvzYFj_kV11IDqOXniks5qyhlCgaJpZM4KR-Yj.

[eonist]

What i personally do Is:

let someCGFloat:CGFloat = 4
let someInt:Int = someCGFloat.int
let someDouble:Int = someCGFloat.double
let someString:String = someCGFloat.string

But this solution has code all over the place. But the suggested cast solution would be more centralised and easier to reason with. Thats why I stared your project , I tought you had found a way to infer types with the "=" sign. :)

[seivan]

I am sorry, I did in Swift 1.0. But they sorta removed it, and I understand since implicit casts could be dangerous. I actually like your way better, but a casted property could work without the huge switch case you had in mind if you rely on types. I could spike up a quick idea if you’d want that.

On 2016Oct10, at 13:00, Eon [email protected] wrote:

What i personally do Is:

let someInt:Int = cgFloat.int let someDouble:Int = cgFloat.double let someString:String = cgFloat.string But this solution has code all over the place. But the suggested cast solution would be more centralised and easier to reason with. Thats why I stared your project , I tought you had found a way to infer types with the "=" sign. :)

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/seivan/ScalarArithmetic/issues/17#issuecomment-252587665, or mute the thread https://github.com/notifications/unsubscribe-auth/AADYgJqeZJdfFNrvc6yJXuXqKTsMGW1iks5qyhrDgaJpZM4KR-Yj.

[eonist]

Yes, definitely avoid the huge if else clause with polymorphism. And just extend every Numerical type instead. If its all in one .swift class its easy enough to reason with.

[eonist]

If you have arrays with numbers you can also do something like this:

let someFloats:[Float] = [1,2,3,4]
let someCGFloats:[CGFloat] = someFloats.cast()

extension Array{
   func cast<T:NumericConversional>() -> [T]{
      return self.map { $0.cast() }
   }
}

NOTE: NumericConversional would need to be a protocol that you add when you make the extension for each Number type. etc etc

[seivan]

For Sequence extensions - You'd probably want to use either a flatMap or rethrow errors.

[eonist]

Isn't flatmap used when a value can be nil? Are you thinking about numerical values that can be NaN? I can't remember but I think Int or UInt can't be NaN

[seivan]

Yep, not to mention between unsigned and signed and such. I recall seeing optional inits. But I might be mistaken.

Sent from an iPhone on the go.

On 11 Oct 2016, at 11:18, Eon [email protected] wrote:

Isn't flatmap used when a value can be nil? Are you thinking about numerical values that can be NaN? I can't remember but I think Int or UInt can't be NaN

— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or mute the thread.

[eonist]

unsigned and signed == special cases no?

Anyways I usually just write for the simplest case first, if something needs more features then one could method overload... like for cases where the array items can be nil:

 func cast<T?:NumericConversional>() -> [T?]{
   return self.flatMap { $0.cast() }
 }

Also i'm not sure if flatMap handles NaN values. asserting for isNaN could be a substitute.

[seivan]

flatMap doesnt out of the box, but you'd have assertions in there on whatever protocol you have.