Category Theory via C# (9) Bifunctor

[LINQ via C# series]

[Category Theory via C# series]

Latest version: https://weblogs.asp.net/dixin/category-theory-via-csharp-5-bifunctor

Bifunctor

As discussed in all the previous functor parts, a functor is a wrapper of a object with a “Select” ability to preserve a morphism to another‘

image_thumb[2]

A bifunctor, as the name implies, is a wrapper of 2 objects, with a “Select” ability to preserve 2 morphisms to another morphism:

image

As above diagram represented, F:

  • maps objects X ∈ ob(C), Y ∈ ob(D) to objects F(X, Y) ∈ ob(E)
  • also maps morphism mC: X → X’ ∈ hom(C), mD: Y → Y’ ∈ hom(D) to a new morphism mE: F(X, Y) → F(X’, Y’) ∈ hom(E)

and satisfies the functor laws:

  1. Select(idX, idY) ≌ idF(X, Y)
    image
  2. Select(m2 ∘ m1, n2 ∘ n1) ≌ Select(m2, n2) ∘ F(m1, n1)
    image

Remember the pseudo C# definition of functor:

// Cannot be compiled.
public interface IFunctor<in TSourceCategory, out TTargetCategory, TFunctor<>>
    where TSourceCategory : ICategory<TSourceCategory>
    where TTargetCategory : ICategory<TTargetCategory>
    where TFunctor<> : IFunctor<TSourceCategory, TTargetCategory, TFunctor<>>
{
    IMorphism<TFunctor<TSource>, TFunctor<TResult>, TTargetCategory> Select<TSource, TResult>(
        IMorphism<TSource, TResult, TSourceCategory> selector);
}

Similarly, bifunctor can be defined as:

// Cannot be compiled
public interface IBinaryFunctor<in TSourceCategory1, in TSourceCategory2, out TTargetCategory, TBinaryFunctor< , >>
    where TSourceCategory1 : ICategory<TSourceCategory1>
    where TSourceCategory2 : ICategory<TSourceCategory2>
    where TTargetCategory : ICategory<TTargetCategory>
    where TBinaryFunctor< , > : IBinaryFunctor<TSourceCategory1, TSourceCategory2, TTargetCategory, TBinaryFunctor< , >>
{
    IMorphism<TBinaryFunctor<TSource1, TSource2>, TBinaryFunctor<TResult1, TResult2>, TTargetCategory> Select<TSource1, TSource2, TResult1, TResult2>(
        IMorphism<TSource1, TResult1, TSourceCategory1> selector1, IMorphism<TSource2, TResult2, TSourceCategory2> selector2);
}

As above definition mentioned, bifunctor wraps 2 objects. So here TBinaryFunctor< , > takes 2 parameters so that it can wrap 2 types. Later the Select function will be implemented as extension method for each bifunctor, the same as how functors are handled.

Tri-functor, and multi-functor can be defined and implemented similarly.

C#/.NET bifunctor

Theoretically, the intuitive bifunctor is Tuple< , >. However, as a previous part mentioned, Tuple< , > can have unexpected behavior in C#/LINQ context, so it will only be considered functor-like. So, to be consistent, Tuple<> or Tuple< , >, … will only be used as utilities in the category theory via C# posts, instead of as functor or bifunctor. Here is a scenario for Tuple< , >, so its lazy version Lazy< , > can be created:

public class Lazy<T1, T2>
{
    private readonly Lazy<Tuple<T1, T2>> lazy;

    public Lazy(Func<T1> factory1, Func<T2> factory2)
        : this(() => Tuple.Create(factory1(), factory2()))
    {
    }

    public Lazy(T1 value1, T2 value2)
        : this(() => Tuple.Create(value1, value2))
    {
    }

    public Lazy(Func<Tuple<T1, T2>> factory)
    {
        this.lazy = new Lazy<Tuple<T1, T2>>(factory);
    }

    public T1 Value1
    {
        [Pure]get { return this.lazy.Value.Item1; }
    }

    public T2 Value2
    {
        [Pure]get { return this.lazy.Value.Item2; }
    }
}

The difference from Lazy<> functor is, as the definition said, Lazy< , > wraps 2 types of values.

To make Lazy< , > a bifunctor, just create these bi-Select extension methods (in Haskell this is called bimap):

// [Pure]
public static partial class LazyExtensions
{
    public static Lazy<TResult1, TResult2> Select<TSource1, TSource2, TResult1, TResult2>
        (this Lazy<TSource1, TSource2> source, 
            Func<TSource1, TResult1> selector1, 
            Func<TSource2, TResult2> selector2) =>
                new Lazy<TResult1, TResult2>(() => selector1(source.Value1), () => selector2(source.Value2));

    public static IMorphism<Lazy<TSource1, TSource2>, Lazy<TResult1, TResult2>, DotNet> Select<TSource1, TSource2, TResult1, TResult2>
        (IMorphism<TSource1, TResult1, DotNet> selector1, IMorphism<TSource2, TResult2, DotNet> selector2) => 
            new DotNetMorphism<Lazy<TSource1, TSource2>, Lazy<TResult1, TResult2>>(
                source => source.Select(selector1.Invoke, selector2.Invoke));
}

The difference from Lazy<> functor is - there are 2 selectors, one selector for each wrapped type.

Unit tests

The following unit test demonstrates the usage and laziness of Lazy< , >:

[TestClass()]
public class BinaryFunctorTests
{
    [TestMethod()]
    public void LazyTest()
    {
        bool isExecuted1 = false;
        bool isExecuted2 = false;
        Lazy<int, string> lazyBinaryFunctor = new Lazy<int, string>(1, "abc");
        Func<int, bool> selector1 = x => { isExecuted1= true; return x > 0; };
        Func<string, int> selector2 = x => { isExecuted2 = true; return x.Length; };

        Lazy<bool, int> query = lazyBinaryFunctor.Select(selector1, selector2);
        Assert.IsFalse(isExecuted1); // Laziness.
        Assert.IsFalse(isExecuted2); // Laziness.

        Assert.AreEqual(true, query.Value1); // Execution.
        Assert.AreEqual("abc".Length, query.Value2); // Execution.
        Assert.IsTrue(isExecuted1);
        Assert.IsTrue(isExecuted2); 
    }
}

Please notice Tuple< , > does not have such laziness.

2 Comments

Add a Comment

As it will appear on the website

Not displayed

Your website