Gaussian random numbers with swift 2.1

Question

I'm trying to update an old objective-c project to swift. I need to generate gaussian random numbers. In objective-c I used this:

double gaussrand()
{
    static double V1, V2, S;
    static int phase = 0;
    double X;

    if(phase == 0) {
        do {
            double U1 = (double)rand() / RAND_MAX;
            double U2 = (double)rand() / RAND_MAX;

            V1 = 2 * U1 - 1;
            V2 = 2 * U2 - 1;
            S = V1 * V1 + V2 * V2;
        } while(S >= 1 || S == 0);

        X = V1 * sqrt(-2 * log(S) / S);
    } else
        X = V2 * sqrt(-2 * log(S) / S);

    phase = 1 - phase;

    return X;
}

However this doesn't translate well into swift. Anybody know a way to generate gaussian random numbers given a mean and a standard deviation in sfift 2.1?

Answer 1

Note that as you have defined it, you want gaussRand to be a computed property. Computed properties in Swift cannot store other properties, so in the example of a Swift version of your Box-Muller transformation method implementation, I've wrapped the computed property gaussRand in a class, and kept s, v2 and cachedNumberExists as stored properties in this same class, allowing for every 2nd call gaussRand to return the cached result from previous one.

class MyRandomGenerator {
    // stored properties
    var s : Double = 0.0
    var v2 : Double = 0.0
    var cachedNumberExists = false

    // (read-only) computed properties
    var gaussRand : Double  {
        var u1, u2, v1, x : Double
        if !cachedNumberExists {
            repeat {
                u1 = Double(arc4random()) / Double(UINT32_MAX)
                u2 = Double(arc4random()) / Double(UINT32_MAX)
                v1 = 2 * u1 - 1;
                v2 = 2 * u2 - 1;
                s = v1 * v1 + v2 * v2;
            } while (s >= 1 || s == 0)
            x = v1 * sqrt(-2 * log(s) / s);
        }
        else {
            x = v2 * sqrt(-2 * log(s) / s);
        }
        cachedNumberExists = !cachedNumberExists
        return x
    }
}

We assert that we get the expected results:

// Assert expected results
var myRandomGenerator = MyRandomGenerator()

let numGaussNumbers = 1000
var myGaussArr = [Double](count: numGaussNumbers, repeatedValue: 0.0)
for (i,_) in myGaussArr.enumerate() { myGaussArr[i] = myRandomGenerator.gaussRand }

let myMean = myGaussArr.reduce(0.0, combine: +)/Double(numGaussNumbers) // 0.0.. OK
let myVar = myGaussArr.map { pow(($0 - myMean), 2) }.reduce(0.0, combine: +)/Double(numGaussNumbers) // ~1, O

print("(\(myMean),\(myVar))") // ~(0,1), OK

OK.