Swift Primitive Value Extension

Swift, unlike many other languages, does not separate primitive values from classes and structs. As we’ll see in this post this choice allows for some interesting and powerful constructs. In other languages, such as Java, the primitive types int, double, float, and boolean are distinctly different from classes and, because of that, you can’t use them in the same way.

In Swift “primitive types” are actually structs which unlocks a lot of interesting features that are otherwise not available for primitives. The code in this post utilizes extensions to extend 32 bit integer values.

Colors

A color is usually represented as a tuple of 3 or 4 values in the range 0, 255. In OSX and iOS development we have UIColor and NSColor.

A straightforward way to implement a color value is shown below. It’s most likely similar to the UIColor and NSColor implementation.

// Example color implementation
struct Color {
  let r: UInt8
  let g: UInt8
  let b: UInt8
}

In lower level graphics programming it’s common to pack color values into integer values by doing bit manipulation. This is done primarily for memory efficiency and compatibility with hardware. A framebuffer is commonly a 1 dimensional array of ints with some kind of packing. There are different ways to pack a color value into an int.

For example using a 32 bit unsigned integer with 4 8 bit values. This packing is usually referred to as ABRG for Alpha, Blue, Green, and Red. The alternative RGBA is also used.

ABGR packing is show in the image above.

In Swift rather than using a struct we can pack color values into integer values. With extensions and typealiasing we can create something that behaves like a struct while retaining the compact memory layout and speed of a packed integer.

#if os(Linux)
    import Glibc
#else
    import Darwin.C
#endif

typealias Color = UInt32

extension Color {
    var r: UInt8 {
        get {
            return UInt8((self & 0x000000FF) >> 0)
        }
    }

    var g: UInt8 {
        get {
            return UInt8((self & 0x0000FF00) >> 8)
        }
    }

    var b: UInt8 {
        get {
            return UInt8((self & 0x00FF0000) >> 16)
        }
    }

    var a: UInt8 {
        get {
            return UInt8((self & 0xFF000000) >> 24)
        }
    }

    init(r: UInt8, g: UInt8, b: UInt8) {
        self = 0xFF000000
        self = self | UInt32(r) << 0
        self = self | UInt32(g) << 8
        self = self | UInt32(b) << 16
    }
}

In other languages this pattern can still be used, but it requires that everyone is a aware when a 32 bit int is actually a color and not a normal int. The fact that Swift supports extensions of primitive types like this, is extremely powerful and gives us type safety, information hiding, and auto completion. Memory efficient layout is retained as well as comatibility with C APIS that expect arrays of ints as color values.

This code was written as a part of my Swift Ray tracer. The full source of this color packing is available here