Serialization: Making a Serializable Class

HomeOverviewHow Do ITutorial

Five main steps are required to make a class serializable. They are listed below and explained in the following sections:

  1. Deriving your class from CObject (or from some class derived from CObject).

  2. Overriding the Serialize member function.

  3. Using the DECLARE_SERIAL macro in the class declaration. 

  4. Defining a constructor that takes no arguments.

  5. Using the IMPLEMENT_SERIAL macro in the implementation file for your class.

If you call Serialize directly rather than through the >> and << operators of CArchive, the last three steps are not required for serialization.

Deriving Your Class from CObject

The basic serialization protocol and functionality are defined in the CObject class. By deriving your class from CObject (or from a class derived from CObject), as shown in the following declaration of class CPerson, you gain access to the serialization protocol and functionality of CObject.

Overriding the Serialize Member Function

The Serialize member function, which is defined in the CObject class, is responsible for actually serializing the data necessary to capture an object’s current state. The Serialize function has a CArchive argument that it uses to read and write the object data. The CArchive object has a member function, IsStoring, which indicates whether Serialize is storing (writing data) or loading (reading data). Using the results of IsStoring as a guide, you either insert your object’s data in the CArchive object with the insertion operator (<<) or extract data with the extraction operator (>>).

Consider a class that is derived from CObject and has two new member variables, of types CString and WORD. The following class declaration fragment shows the new member variables and the declaration for the overridden Serialize member function:

class CPerson : public CObject
{
public:
    DECLARE_SERIAL( CPerson )
    // empty constructor is necessary
    CPerson(){};

    CString m_name;
    WORD   m_number;

    void Serialize( CArchive& archive );
    
    // rest of class declaration
};

To override the Serialize member function

  1. Call your base class version of Serialize to make sure that the inherited portion of the object is serialized.

  2. Insert or extract the member variables specific to your class.

    The insertion and extraction operators interact with the archive class to read and write the data. The following example shows how to implement Serialize for the CPerson class declared above:

    void CPerson::Serialize( CArchive& archive )
    {
        // call base class function first
        // base class is CObject in this case
        CObject::Serialize( archive );
    
        // now do the stuff for our specific class
        if( archive.IsStoring() )
            archive << m_name << m_number;
        else
            archive >> m_name >> m_number;
    }
    

You can also use the CArchive::Read and CArchive::Write member functions to read and write large amounts of untyped data.

Using the DECLARE_SERIAL Macro

The DECLARE_SERIAL macro is required in the declaration of classes that will support serialization, as shown here:

class CPerson : public CObject
{
    DECLARE_SERIAL( CPerson )
    // rest of declaration follows...
};

Defining a Constructor with No Arguments

MFC requires a default constructor when it re-creates your objects as they are deserialized (loaded from disk). The deserialization process will fill in all member variables with the values required to re-create the object.

This constructor can be declared public, protected, or private. If you make it protected or private, you ensure that it will only be used by the serialization functions. The constructor must put the object in a state that allows it to be safely deleted if necessary.

Note   If you forget to define a constructor with no arguments in a class that uses the DECLARE_SERIAL and IMPLEMENT_SERIAL macros, you will get a “no default constructor available” compiler warning on the line where the IMPLEMENT_SERIAL macro is used.

Using the IMPLEMENT_SERIAL Macro in the Implementation File

The IMPLEMENT_SERIAL macro is used to define the various functions needed when you derive a serializable class from CObject. You use this macro in the implementation file (.CPP) for your class. The first two arguments to the macro are the name of the class and the name of its immediate base class.

The third argument to this macro is a schema number. The schema number is essentially a version number for objects of the class. Use an integer greater than or equal to 0 for the schema number. (Don’t confuse this schema number with database terminology.)

The MFC serialization code checks the schema number when reading objects into memory. If the schema number of the object on disk does not match the schema number of the class in memory, the library will throw a CArchiveException, preventing your program from reading an incorrect version of the object.

If you want your Serialize member function to be able to read multiple versions — that is, files written with different versions of the application — you can use the value VERSIONABLE_SCHEMA as an argument to the IMPLEMENT_SERIAL macro. For usage information and an example, see the GetObjectSchema member function of class CArchive.

The following example shows how to use IMPLEMENT_SERIAL for a class, CPerson, that is derived from CObject:

IMPLEMENT_SERIAL( CPerson, CObject, 1 )

Once you have a serializable class, you can serialize objects of the class, as discussed in the article Serialization: Serializing an Object.