Solution for overloaded operator constraint in .NE

2019-01-01 08:58发布

问题:

What would I do if I want to have a generic method that only accepts types that have overloaded an operator, for instance the subtraction operator. I tried using an interface as a constraint but interfaces can\'t have operator overloading.

What is the best way to achieve this?

回答1:

There is no immediate answer; operators are static, and cannot be expressed in constraints - and the existing primatives don\'t implement any specific interface (contrast to IComparable[<T>] which can be used to emulate greater-than / less-than).

However; if you just want it to work, then in .NET 3.5 there are some options...

I have put together a library here that allows efficient and simple access to operators with generics - such as:

T result = Operator.Add(first, second); // implicit <T>; here

It can be downloaded as part of MiscUtil

Additionally, in C# 4.0, this becomes possible via dynamic:

static T Add<T>(T x, T y) {
    dynamic dx = x, dy = y;
    return dx + dy;
}

I also had (at one point) a .NET 2.0 version, but that is less tested. The other option is to create an interface such as

interface ICalc<T>
{
    T Add(T,T)() 
    T Subtract(T,T)()
} 

etc, but then you need to pass an ICalc<T>; through all the methods, which gets messy.



回答2:

I found that IL can actually handle this quite well. Ex.

ldarg.0
ldarg.1
add
ret

Compiled in a generic method, the code will run fine as long as a primitive type is specified. It may be possible to extend this to call operator functions on non-primitive types.

See here.



回答3:

There is a piece of code stolen from the internats that I use a lot for this. It looks for or builds using IL basic arithmetic operators. It is all done within an Operation<T> generic class, and all you have to do is assign the required operation into a delegate. Like add = Operation<double>.Add.

It is used like this:

public struct MyPoint
{
    public readonly double x, y;
    public MyPoint(double x, double y) { this.x=x; this.y=y; }
    // User types must have defined operators
    public static MyPoint operator+(MyPoint a, MyPoint b)
    {
        return new MyPoint(a.x+b.x, a.y+b.y);
    }
}
class Program
{
    // Sample generic method using Operation<T>
    public static T DoubleIt<T>(T a)
    {
        Func<T, T, T> add=Operation<T>.Add;
        return add(a, a);
    }

    // Example of using generic math
    static void Main(string[] args)
    {
        var x=DoubleIt(1);              //add integers, x=2
        var y=DoubleIt(Math.PI);        //add doubles, y=6.2831853071795862
        MyPoint P=new MyPoint(x, y);
        var Q=DoubleIt(P);              //add user types, Q=(4.0,12.566370614359172)

        var s=DoubleIt(\"ABC\");          //concatenate strings, s=\"ABCABC\"
    }
}

Operation<T> Source code courtesy of paste bin: http://pastebin.com/nuqdeY8z

with attribution below:

/* Copyright (C) 2007  The Trustees of Indiana University
 *
 * Use, modification and distribution is subject to the Boost Software
 * License, Version 1.0. (See accompanying file LICENSE_1_0.txt or copy at
 * http://www.boost.org/LICENSE_1_0.txt)
 *  
 * Authors: Douglas Gregor
 *          Andrew Lumsdaine
 *          
 * Url:     http://www.osl.iu.edu/research/mpi.net/svn/
 *
 * This file provides the \"Operations\" class, which contains common
 * reduction operations such as addition and multiplication for any
 * type.
 *
 * This code was heavily influenced by Keith Farmer\'s
 *   Operator Overloading with Generics
 * at http://www.codeproject.com/csharp/genericoperators.asp
 *
 * All MPI related code removed by ja72. 
 */