General Considerations for Interoperation
All calls made between managed and unmanaged code must negotiate the requirements imposed by each respective programming model. Managed and unmanaged programming models are dissimilar in many respects. The following table shows the defining characteristics of each model.
Characteristic |
Unmanaged model |
Managed model |
|---|---|---|
Coding model |
Interface-based |
Object-based |
Identity |
GUIDs |
Strong names |
Error handling mechanism |
HRESULTs |
Exceptions |
Type compatibility |
Binary standard |
Type standard |
Type definition |
Type library |
Metadata |
Type safety |
Not type safe |
Optionally safe |
Versioning |
Immutable |
Resilient |
Some programming model characteristics have associated entities that you can view, such as type libraries and metadata. Some you can pass as arguments, such as GUIDs and strong names. Other characteristics are more conceptual; you will undoubtedly consider their impact on your application design, but you will not encounter direct mapping between the managed and unmanaged models for these characteristics.
The following sections explain each characteristic in more detail.
Coding models
Unmanaged objects always communicate through interfaces; managed objects and classes can pass data directly without implementing interfaces.By default, COM interop generates a class interface to expose managed functionality through an interface to COM when the object or class does not implement one.
Error handling mechanisms
COM methods usually return an HRESULT, indicating that the call succeeded or failed. Managed code incorporates exceptions. By default, COM interop maps managed exceptions to failure HRESULTs.Identities
GUIDs identify a specific unmanaged type and provide no location information for that type. Strong names consist of a unique assembly name in addition to a type name. Because the assembly name uniquely identifies the type, you can reuse a type name across multiple assemblies. An assembly also introduces publisher key, version, and location information to a managed type. Interoperation services generate GUIDs and are strong-named as required.Plain old data structures (PODS)
Platform invoke cannot be used to return structures or classes by value if they contain a constructor. In general, user-defined types that contain non-PODS elements should be returned by reference. PODS are data structures that contain only passive collections of field values, as defined by the ISO/IEC standard 14882, "Programming Languages — C++." They do not contain constructors, copy assignment operators, destructors, or non-static data members that are not themselves PODS.Type compatibility
Types vary between managed and unmanaged code, and also among languages.Type definitions
If you are accustomed to working with type libraries, you know that they contain only public types. Moreover, a type library is optional. In the managed programming model, type information is mandatory for all types. Interoperation services provide tools that convert type libraries to metadata in assemblies and metadata to type libraries.Type safety
Unmanaged compilers provide no type checking on pointer types, making the code susceptible to potentially harmful activity. In general, managed code requires a higher level of trust. Programmers can continue to use pointers in managed code, although the code has restrictions because of its unsafe behavior. Interoperation services prevent untrusted, managed code from accessing unmanaged code.Versioning
COM interfaces are immutable. If you change an interface, you must rename it with a new GUID. Managed types can evolve while keeping the same name.