AFAIK, you can legally access the compatible
<http://en.cppreference.com/w/c/language/type#Compatible_types> prefix of
structure fields in C. (Regardless of whether you have a pointer to a Sample1
or to a Sample2, casting it to either and accessing a or b is guaranteed to
work; c or d is UB.)
You can port your second example to Swift as well, by pulling out the common
fields in a struct and using an enum as a union for the variable part:
struct Sample {
enum CD {
case float(Float)
case double(Double)
}
var a: Int
var b: Int
var cd: CD
}
Félix
> Le 22 juil. 2017 à 17:07, Daryle Walker via swift-evolution
> <[email protected]> a écrit :
>
> Way back in the macOS pre-X days, I remember some APIs had types like:
>
> struct Sample1 {
> int a;
> int b;
> float c;
> };
>
> struct Sample2 {
> int a;
> int b;
> double d;
> };
>
> The types are supposed to be related, and it was supposed to be guaranteed
> that if you type-punned Sample1 to or from Sample2, the common members (“a”
> and “b” here) would have the same semantics. I guess it could be done like
> this:
>
> struct Sample {
> int a;
> int b;
> union {
> float c;
> double d;
> }
> };
>
> too.
>
> I wonder if we need something similar in Swift:
>
> enum Sample {
> @common case withFloat(a: Int, b: Int, c: Float)
> @common case withDouble(a: Int, b: Int, d: Double)
> }
>
> The attribute would ensure that the shared initial members would map to the
> same internal offset. And, if there were no cases that didn’t have the common
> members, you can use them as instance level properties:
>
> var x: Sample = .withFloat(1, 2, 3.0)
> //…
> assert(x.a == 1) // Instead of “x.withFloat.a"
>
> Good idea? Does this already exist?
>
> —
> Daryle Walker
> Mac, Internet, and Video Game Junkie
> darylew AT mac DOT com
>
> _______________________________________________
> swift-evolution mailing list
> [email protected]
> https://lists.swift.org/mailman/listinfo/swift-evolution
_______________________________________________
swift-evolution mailing list
[email protected]
https://lists.swift.org/mailman/listinfo/swift-evolution
