Iterating through List Containers

.

When using containers in a program, I declare a class which represents an item in a container:

// Can also be a struct
class CMyClass : public CObject {
    // Program specific
    ...
};

typedef CTypedPtrList<CPtrList,CMyClass*> TMyList;

Separate class (container) has the actual container object:

class CMyClassList : public CObject {
    private:
        TMyList List;
        // Maybe some other data members too
    public:
        // Usual stuff (construction, destruction etc.)
        void Flush(void);
        BOOL Add(CMyClass *ptr);
        CMyClass *Find(...);
        BOOL Del(...);
};

If the container is associated with a user interface object (tree control, list control etc.), one must iterate through the container and perform an action for each item (or for those that satisfy some criteria). Easy way to do it is to move a List variable to a public part of the class declaration and then access it directly. This is bad since it violates a data encapsulation principle (in all but trivial examples, class declaration has a lot more member variables and methods and performs some useful job too).

One approach is to use a public member function and a supplied callback function which is executed for each item. This is messy since supplied callback function is usually a member function of another class so for each different callback function, class CMyClassList must have an overloaded member function. If a callback function is a static non-member function, one must use a DWORD function argument to pass a "this" pointer to a static callback function, then from within a callback function cast a DWORD to pointer to a class object and then invoke a method which actually performs some action with a container item.

Much easier and more elegant approach is to use special iterator class. New class declaration is slightly modified:

class CMyClassListIterator;

class CMyClassList : public CObject {
    private:
        TMyClassList List;
        // Maybe something else
    public:
        // Usual stuff (construction, destruction etc.)
        void Flush(void);
        BOOL Add(CMyClass *ptr);
        CMyClass *Find(...);
        BOOL Del(...);
    friend CMyClassListIterator;        // New stuff
};

New iterator class is declared as follows:

class CMyClassListIterator : public CObject {
  private:
    CMyClassList& Owner;
    POSITION Pos;
  public:
    CMyClassListIterator(TProcess& obj)
              :CObject(),Owner(obj) { Reset(); }
    void Reset(void)                { Pos = Owner.List.GetHeadPosition(); }
    void Next(void)                 { Owner.List.GetNext(Pos); }
    CMyClass *Current(void)         { return (CMyClass*)Owner.List.GetAt(Pos); }
    BOOL IsDone(void)               { return (Pos == NULL) ? TRUE : FALSE; }
};

Now, all you need to do to iterate thru the list is to implement the following peace of code:

CMyClassList MyList;
........
CMyClassListIterator iterator(MyList);
while (!iterator.IsDone()) {
    CMyClass *ptr = iterator.Current();
    // Use ptr but do not delete it.
    // You can modify its contents however.
    iterator.Next();
};

You can also easily implement a nested iteration. Good side effect of this implementation is that your code for list iteration is focused on the job that must be done since all the code is implemented in one function (no callbacks). Also, data encapsulation is preserved since List object is not directly visible.