Generics

Generics overview

Generics enable writing of code abstractions which can be applied to different types at compile time. List, for example, is a basic list abstraction which allows for type-safe storage of values. Using a “List” type reference creates a specialized int32 list type.

Methods can also have generic parameters, allowing for them to be specialized either explicitly or implicitly based on callsite argument types.

public static T GetFirst<T>(List<T> list)
{
    return list[0];
}
...
let intList = new List<int32>();
intList.Add(123);
let firstVal = GetFirst(intList);

Generic constraints can be specified, which describe the ‘shape’ of the type which the generic code is intended to work with.

• Interface type - any number of interfaces can be specified for a generic parameter. The incoming type must declare implementations for all these interfaces.
• Class/struct type - a single concrete type can be specified, which the incoming type must derive from.
• Delegate type - the incoming type can either be an instance of this delegate type, or it can be a method reference whose signature conforms to the delegate (see Method References)
• operator T <op> T2 - type must result from binary operation between the specified types
• operator <op> T - type must result from unary operation on the specified
• operator implicit T - type must be implicitly convertible from the specified type
• operator explicit T - type must be explicitly convertible from the specified type
• class - type must be class
• struct - type must be a value type
• enum - type must be an enum
• interface - type must be an interface
• struct* - type must be a pointer to a value type
• new - type must define an accessible default constructor
• delete - type must define an accessible destructor
• const - type must be a constant value - see “Const Generics”
• var - type is unconstrained. This can be useful for certain kinds of “duck typing”, and can generate patterns similar to C++ templates, but in general produces less useful errors and a less pleasant development experience

public static T Abs<T>(T value) where T : IOpComparable, IOpNegatable
{
    if (value < default)
        return -value;
    else
        return value;
}

/* This method can eliminate runtime branching by specializing at compile time by incoming array size */
public static float GetSum<TCount>(float[TCount] vals) where TCount : const int
{
    if (vals.Count == 0)
    {
        return 0;
    }
    else if (vals.Count == 1)
    {
        return vals[0];
    }
    else
    {
        float total = 0;
        for (let val in vals)
            total += val;
        return total;
    }
}

static TTo Convert<TTo, TFrom>(TFrom val) where TTo : operator explicit TFrom
{
    return (TTo)val;
}

/* We use partial explicit generic args to allow inference of 'TFrom' */
var val = Convert<int...>(1.2f);