I’ve finally kicked my primitive obsession! I didn’t even know I had one until I got serious about making a type for everything that has a name. I’ll never write a method like this again

void MoveTo(double position)

What is position? What values of double are valid? Can it be negative, zero, not a number? Does it have a direction? Is it relative or absolute? Most importantly, what scale and system? Not very informative, huh?

If MoveTo were an absolute move on a micron scale linear motor controller, I’d rather see this

void MoveTo(Distance position)

Now we have options! Let’s see what the Distance type looks like

public class Distance : IEquatable<Distance>
    private const int MicronsPerMillimeter = 1000;

    public static readonly Distance None = new Distance(double.NaN);
    public static readonly Distance Zero = new Distance(0);

    public static Distance FromMicrons(double microns) { return new Distance(microns); }
    public static Distance FromMillimeters(double millimeters) { return new Distance(millimeters * MicronsPerMillimeter); }

    public double TotalMicrons { get; private set; }
    public double TotalMillimeters { get; private set; }

    public Distance(double microns)
        TotalMicrons = microns;
        TotalMillimeters = TotalMicrons / MicronsPerMillimeter;

    public override string ToString()
        return string.Format("{0} um", TotalMicrons);

    // equality methods... GetHashCode(), Equals(obj other)
    // arithmetic operator overloads... +, -, /, *, etc

We can make intuitive extension methods for creating instances.


We have properties for accessing the different scale primitives.


A fancy ToString implementation.

10.Millimeters().ToString();    //=> "10000 um"

And arithmetic and equality operators galore! We can even introduce other “unit” types and work them together.

// A `Velocity` times a `TimeSpan` equals a `Distance`
var distance = 100.MicronsPerSecond() * 10.Seconds();    //=> 1 mm

The toughest part of propagating these simple value types is the friction you’ll face in the framework. Like reimplementing the Math class functions (check out units of measure in F#). Although, I can make them extension methods and give them better semantics.

Math.Abs(double - double)
// or
Distance.Between(Distance, Distance)

The arithmetic and equality operator overloading is tedious. Between the same type you’ll need +, -, /, *, <, >, ==, and !=. That’s not to mention all the types that cancel or combine units (like TimeSpan, Distance, Velocity, and Acceleration`). You’ll need to implement operators between these. And, where commutative properties apply, double the number of overloads!

operator Distance *(Velocity velocity, TimeSpan timespan)
operator Distance *(TimeSpan timespan, Velocity velocity)

And where do you put these operators, on the first or second type? Maybe this is why the framework defines no operators on these kinds of types (for instance, Vector, Point, and TimeSpan) but uses static methods. I think it kills the flow of the code. Compare static instantiation and arithmetic

var distance = Velocity.PerSecond(Distance.FromMicrons(100)).DistanceAfter(TimeSpan.FromSeconds(10));

to extension methods and operator overloads

var distance = 100.MicronsPerSecond() * 10.Seconds();

Anyone could understand the units in play and the math happening here. The flow of the extension methods and operator overloads is worth more than the weak claim of static method name clarity.