Dynamically Loading and Using Types

Reflection provides infrastructure used by language compilers such as Microsoft Visual Basic 2005 and JScript to implement implicit late binding. Binding is the process of locating the declaration (that is, the implementation) that corresponds to a uniquely specified type. When this process occurs at run time rather than at compile time, it is called late binding. Visual Basic 2005 allows you to use implicit late binding in your code; the Visual Basic compiler calls a helper method that uses reflection to obtain the object type. The arguments passed to the helper method cause the appropriate method to be invoked at run time. These arguments are the instance (an object) on which to invoke the method, the name of the invoked method (a string), and the arguments passed to the invoked method (an array of objects).

In the following example, the Visual Basic compiler uses reflection implicitly to call a method on an object whose type is not known at compile time. A HelloWorld class has a PrintHello method that prints out "Hello World" concatenated with some text that is passed to the PrintHello method. The PrintHello method called in this example is actually a Type::InvokeMember; the Visual Basic code allows the PrintHello method to be invoked as if the type of the object (helloObj) were known at compile time (early binding) rather than at run time (late binding).

Imports System
Module Hello
    Sub Main()
        ' Sets up the variable.
        Dim helloObj As Object
        ' Creates the object.
        helloObj = new HelloWorld()
        ' Invokes the print method as if it was early bound
        ' even though it is really late bound.
        helloObj.PrintHello("Visual Basic Late Bound")
    End Sub
End Module

In addition to being used implicitly by compilers for late binding, reflection can be used explicitly in code to accomplish late binding.

The common language runtime supports multiple programming languages, and the binding rules of these languages differ. In the early-bound case, code generators can completely control this binding. However, in late binding through reflection, binding must be controlled by customized binding. The Binder class provides custom control of member selection and invocation.

Using custom binding, you can load an assembly at run time, obtain information about types in that assembly, specify the type that you want, and then invoke methods or access fields or properties on that type. This technique is useful if you do not know an object's type at compile time, such as when the object type is dependent on user input.

The following example demonstrates a simple custom binder that provides no argument type conversion. Code for Simple_Type.dll precedes the main example. Be sure to build Simple_Type.dll and then include a reference to it in the project at build time.

// Code for building SimpleType.dll. 
using namespace System;
using namespace System::Reflection;
using namespace System::Globalization;

namespace Simple_Type
{
    public ref class MySimpleClass
    {
    public:
        void MyMethod(String^ str, int i)
        {
            Console::WriteLine("MyMethod parameters: {0}, {1}", str, i);
        }

        void MyMethod(String^ str, int i, int j)
        {
            Console::WriteLine("MyMethod parameters: {0}, {1}, {2}",
                str, i, j);
        }
    };
}

using namespace Simple_Type;

namespace Custom_Binder
{
    // **************************************************** 
    //  A simple custom binder that provides no 
    //  argument type conversion. 
    // **************************************************** 
    public ref class MyCustomBinder : Binder
    {
    public:
        virtual MethodBase^ BindToMethod(
            BindingFlags bindingAttr,
            array<MethodBase^>^ match,
            array<Object^>^% args,
            array<ParameterModifier>^ modifiers,
            CultureInfo^ culture,
            array<String^>^ names,
            Object^% state) override
        {
            if (match == nullptr)
            {
                throw gcnew ArgumentNullException("match");
            }
            // Arguments are not being reordered.
            state = nullptr;
            // Find a parameter match and return the first method with 
            // parameters that match the request. 
            for each (MethodBase^ mb in match)
            {
                array<ParameterInfo^>^ parameters = mb->GetParameters();

                if (ParametersMatch(parameters, args))
                {
                    return mb;
                }
            }
            return nullptr;
        }

        virtual FieldInfo^ BindToField(BindingFlags bindingAttr,
            array<FieldInfo^>^ match, Object^ value, CultureInfo^ culture) override
        {
            if (match == nullptr)
            {
                throw gcnew ArgumentNullException("match");
            }
            for each (FieldInfo^ fi in match)
            {
                if (fi->GetType() == value->GetType())
                {
                    return fi;
                }
            }
            return nullptr;
        }

        virtual MethodBase^ SelectMethod(
            BindingFlags bindingAttr,
            array<MethodBase^>^ match,
            array<Type^>^ types,
            array<ParameterModifier>^ modifiers) override
        {
            if (match == nullptr)
            {
                throw gcnew ArgumentNullException("match");
            }

            // Find a parameter match and return the first method with 
            // parameters that match the request. 
            for each (MethodBase^ mb in match)
            {
                array<ParameterInfo^>^ parameters = mb->GetParameters();
                if (ParametersMatch(parameters, types))
                {
                    return mb;
                }
            }

            return nullptr;
        }

        virtual PropertyInfo^ SelectProperty(
            BindingFlags bindingAttr,
            array<PropertyInfo^>^ match,
            Type^ returnType,
            array<Type^>^ indexes,
            array<ParameterModifier>^ modifiers) override
        {
            if (match == nullptr)
            {
                throw gcnew ArgumentNullException("match");
            }
            for each (PropertyInfo^ pi in match)
            {
                if (pi->GetType() == returnType &&
                    ParametersMatch(pi->GetIndexParameters(), indexes))
                {
                    return pi;
                }
            }
            return nullptr;
        }

        virtual Object^ ChangeType(
            Object^ value,
            Type^ myChangeType,
            CultureInfo^ culture) override
        {
            try
            {
                Object^ newType;
                newType = Convert::ChangeType(value, myChangeType);
                return newType;
            }
            // Throw an InvalidCastException if the conversion cannot 
            // be done by the Convert.ChangeType method. 
            catch (InvalidCastException^)
            {
                return nullptr;
            }
        }

        virtual void ReorderArgumentArray(array<Object^>^% args,
            Object^ state) override
        {
            // No operation is needed here because BindToMethod does not 
            // reorder the args array. The most common implementation 
            // of this method is shown below. 

            // ((BinderState^)state).args.CopyTo(args, 0);
        }

        // Returns true only if the type of each object in a matches 
        // the type of each corresponding object in b. 
    private:
        bool ParametersMatch(array<ParameterInfo^>^ a, array<Object^>^ b)
        {
            if (a->Length != b->Length)
            {
                return false;
            }
            for (int i = 0; i < a->Length; i++)
            {
                if (a[i]->ParameterType != b[i]->GetType())
                {
                    return false;
                }
            }
            return true;
        }

        // Returns true only if the type of each object in a matches 
        // the type of each corresponding entry in b. 
        bool ParametersMatch(array<ParameterInfo^>^ a, array<Type^>^ b)
        {
            if (a->Length != b->Length)
            {
                return false;
            }
            for (int i = 0; i < a->Length; i++)
            {
                if (a[i]->ParameterType != b[i])
                {
                    return false;
                }
            }
            return true;
        }
    };

    public ref class MyMainClass
    {
    public:
        static void Main()
        {
            // Get the type of MySimpleClass.
            Type^ myType = MySimpleClass::typeid;

            // Get an instance of MySimpleClass.
            MySimpleClass^ myInstance = gcnew MySimpleClass();
            MyCustomBinder^ myCustomBinder = gcnew MyCustomBinder();

            // Get the method information for the particular overload 
            // being sought.
            MethodInfo^ myMethod = myType->GetMethod("MyMethod",
                BindingFlags::Public | BindingFlags::Instance,
                myCustomBinder, gcnew array<Type^> {String::typeid,
                int::typeid}, nullptr);
            Console::WriteLine(myMethod->ToString());

            // Invoke the overload.
            myType->InvokeMember("MyMethod", BindingFlags::InvokeMethod,
                myCustomBinder, myInstance,
                gcnew array<Object^> {"Testing...", (int)32});
        }
    };
}

int main()
{
    Custom_Binder::MyMainClass::Main();
}

Use Type::InvokeMember to invoke a member of a type. The CreateInstance methods of various classes, such as System.Activator and System.Reflection.Assembly, are specialized forms of InvokeMember that create new instances of the specified type. The Binder class is used for overload resolution and argument coercion in these methods.

The following example shows the three possible combinations of argument coercion (type conversion) and member selection. In Case 1, no argument coercion or member selection is needed. In Case 2, only member selection is needed. In Case 3, only argument coercion is needed.

public ref class CustomBinderDriver
{
public:
    static void Main()
    {
        Type^ t = CustomBinderDriver::typeid;
        CustomBinder^ binder = gcnew CustomBinder();
        BindingFlags flags = BindingFlags::InvokeMethod | BindingFlags::Instance |
            BindingFlags::Public | BindingFlags::Static;
        array<Object^>^ args;

        // Case 1. Neither argument coercion nor member selection is needed.
        args = gcnew array<Object^> {};
        t->InvokeMember("PrintBob", flags, binder, nullptr, args);

        // Case 2. Only member selection is needed.
        args = gcnew array<Object^> {42};
        t->InvokeMember("PrintValue", flags, binder, nullptr, args);

        // Case 3. Only argument coercion is needed.
        args = gcnew array<Object^> {"5.5"};
        t->InvokeMember("PrintNumber", flags, binder, nullptr, args);
    }

    static void PrintBob()
    {
        Console::WriteLine("PrintBob");
    }

    static void PrintValue(long value)
    {
        Console::WriteLine("PrintValue({0})", value);
    }

    static void PrintValue(String^ value)
    {
        Console::WriteLine("PrintValue\"{0}\")", value);
    }

    static void PrintNumber(double value)
    {
        Console::WriteLine("PrintNumber ({0})", value);
    }
};

int main()
{
    CustomBinderDriver::Main();
}

Overload resolution is needed when more than one member with the same name is available. The Binder::BindToMethod and Binder::BindToField methods are used to resolve binding to a single member. Binder.BindToMethod also provides property resolution through the get and set property accessors.

BindToMethod returns the MethodBase to invoke, or a null reference (Nothing in Visual Basic) if no such invocation is possible. The MethodBase return value need not be one of those contained in the match parameter, although that is the usual case.

When ByRef arguments are present, the caller might want to get them back. Therefore, Binder allows a client to map the array of arguments back to its original form if BindToMethod has manipulated the argument array. In order to do this, the caller must be guaranteed that the order of the arguments is unchanged. When arguments are passed by name, Binder reorders the argument array, and that is what the caller sees. For more information, see Binder::ReorderArgumentArray.

The set of available members are those members defined in the type or any base type. If BindingFlags.NonPublic is specified, members of any accessibility will be returned in the set. If BindingFlags.NonPublic is not specified, the binder must enforce accessibility rules. When specifying the Public or NonPublic binding flag, you must also specify the Instance or Static binding flag, or no members will be returned.

If there is only one member of the given name, no callback is necessary, and binding is done on that method. Case 1 of the code example illustrates this point: Only one PrintBob method is available, and therefore no callback is needed.

If there is more than one member in the available set, all these methods are passed to BindToMethod, which selects the appropriate method and returns it. In Case 2 of the code example, there are two methods named PrintValue. The appropriate method is selected by the call to BindToMethod.

ChangeType performs argument coercion (type conversion), which converts the actual arguments to the type of the formal arguments of the selected method. ChangeType is called for every argument even if the types match exactly.

In Case 3 of the code example, an actual argument of type String with a value of "5.5" is passed to a method with a formal argument of type Double. For the invocation to succeed, the string value "5.5" must be converted to a double value. ChangeType performs this conversion.

ChangeType performs only lossless or widening coercions, as shown in the following table.

Source type

Target type

Any type

Its base type

Any type

Interface it implements

Char

UInt16, UInt32, Int32, UInt64, Int64, Single, Double

Byte

Char, UInt16, Int16, UInt32, Int32, UInt64, Int64, Single, Double

SByte

Int16, Int32, Int64, Single, Double

UInt16

UInt32, Int32, UInt64, Int64, Single, Double

Int16

Int32, Int64, Single, Double

UInt32

UInt64, Int64, Single, Double

Int32

Int64, Single, Double

UInt64

Single, Double

Int64

Single, Double

Single

Double

Nonreference type

Reference type

The Type class has Get methods that use parameters of type Binder to resolve references to a particular member. Type::GetConstructor, Type::GetMethod, and Type::GetProperty search for a particular member of the current type by providing signature information for that member. Binder::SelectMethod and Binder::SelectProperty are called back on to select the given signature information of the appropriate methods.

Was this page helpful?
(1500 characters remaining)
Thank you for your feedback
Show:
© 2014 Microsoft