-- dedicated to those who pursue after the academic development

Mephisto, April 4th 2005

It is inevitable for us college students to have some trouble in the process of gaining good knowledge, I think. To commence be doubtful of the capability to deal with the complexity, the intellect to comprehend the abstract or advanced theory just for a trivial frustration in study is unwanted. We should know this phenomenon is normal and we should accept it willingly. From the day when we entered the ivory tower, the classic theories founded in the recent 3 centuries confronted with us, which are abstract and systematic to master in a short period, which own their subjects' developing history we need time to learn about, and which is different from the knowledge in the elementary schools or secondary schools we should distinguish from it. For example, Newton and Leibnitz discovered the Calculous, which showed us the infinitesimal conception. Another example is Turing's computable model theory, which and the Church's recursion theory were viewed as the theoretic rationales for the birth of the computer. These were the big breakthroughs in that day. If they can mastered easily by an average man, they will not be what they are. And it decided that we cannot grasp them in one day, i.e., we should learn them step by step in order to make a very solid foundation.

As a matter of fact, learning the knowledge for the textbook is relatively easier compared with the research. Why? When we are in the phase of learning, the only thing we’d better do is imitation. But in the researching, we face the open problems, which may have no standard answer, and there is no one who can tell you what to do and how to do. Many people told me that there is no oracle in the frontier of science. So as a student not a researcher is fortunate, because we can find so many classic textbook, good reference books, and famous papers, because our teacher can tell us the solution how to do an exercise even how to study well. When I was an undergraduate, studying the Calculous, my teacher recommended me a series of exercise books complied by Gemmidovic.

However, judging from the score you got I can tell you be a top student in your class. And little by little the courses you touched or will touch are further than ever. Now we can see that the problem of computing on the rational domain is so simple that we can say that this knowledge we will understand in a few hours because you have own enough ability to deal with this kind of problems. When you climb the summit you will discover the problem once viewed as hard is easy and another new challenge will burden on your shoulder. That is the case.

Having known your configuration on the courses, it is unreasonable indeed. I only suggest you schedule all the things soundly and give up some non-significant doings. Since you select to proceed and preserve your leading position, I think, giving up the weekend not to relax but to study is a must, giving up the money not to purchase luxury but to use it in copying or printing the papers or other helpful books and giving up some course temporarily just for concentrating on the most important courses, which you can complement later. Communicating with the successful man is another means to guide you go onto a shortcut.

When it comes to my experience, I just want to tell you to read a good book as many times as possible. More than one year before, I noted the model checking technique for the first time and at the same time the fixed-point theory occurred to me. Some papers on this abstract and complex theory were searched and gained, since this is a brand-new field for me, a layman. In the first a few weeks, I even did not know what the papers said, including the strange formulae printed on which. I analyzed the main causes must be that the authors are the experts in the model checking areas and they write the paper to communicate with others assumed their readers must be the peers. I know I want to do model checking but I still have not power enough to fulfill it. Till now I still cannot understand the fixed-point theory completely and I was little worried about it since I have read this theory for many times, but every time I went into it I made some new cognition on it. And when we faced a problem, there are different sceneries from different viewpoint or different hierarchies. As it is well known that the mathematical induction method is a general method to prove some proofs. Today I will tell you that mathematical induction proof is just a concrete form or special instance of the well-founded induction method, don’t be astonished, my pal, because maybe we stand at respective spot.

English learning is also important. Once it have wasted me much energy, and I really found its huge effect when I began to do the research and the large amount of English literature often make me depressed. I can recollected that at that time I just read one or two pages in a whole afternoon, however, after a long-term training for reading and writing English now I can cover 30 pages or more in one night almost no barriers. Through reading and writing theses in English, I leaned about the authors’ mode in thinking and their precise and rigorous work in research.

At last, I wish you to bring up happily, and I wish you can steer the canoe in the ocean freely and eventually reach the other shore as you desire.

Using C++ Multiple Inheritance with MFC

This note describes how to use Multiple Inheritance (MI) with the Microsoft Fo
undation Classes.

Why Multiple Inheritance?

There is an ongoing debate in the C++ and object-oriented communities over the
value of multiple inheritance (MI). The Visual C++ compiler and development e
nvironment fully support MI.

The MFC class library has been designed so that you do not need to understand
MI to use MFC. MI is not used in any of the MFC classes. We have found that MI
is not required to write a class library, nor is it required for writing seri
ous applications. To use MI or not can be a personal decision, so we leave tha
t decision to you.

So You Want to Use MI?

If you already understand how to use MI, understand the performance trade-offs
, and want to use MFC, this tech note will tell you what you must do. Some of
the restrictions are general C++ restrictions, others are imposed by the MFC a
rchitecture.

The following describes some of the technical issues of how MI affects the use
of common MFC idioms. At the end of this technical note a complete MFC applic
ation using MI is included which you can extract and compile.

CRuntimeClass

The persistence and dynamic object creation mechanisms of MFC use the CRuntime
Class data structure to uniquely identify classes. MFC associates one structur
e of this type with each dynamic and/or serializable class in the application.
These structures are initialized at application startup time using a special
static object of type AFX_CLASSINIT. You need not concern yourself with the im
plementation of this information, as it is likely to change between revisions
of MFC.

The current implementation of CRuntimeClass does not support Multiple Inherita
nce runtime type information. This does not mean you cannot use MI in your MFC
application, but if you do, you will have certain responsibilities when worki
ng with objects that have more than one base class.

The CObject::IsKindOf member function will not correctly determine the type of
an object if it has multiple base classes. Therefore, you cannot use CObject
as a virtual base class, and all calls to CObject member functions such as Ser
ialize and operator new will need to have scope qualifiers so that C++ can dis
ambiguate the appropriate function call. If you do find the need to use MI wit
hin MFC, then you should be sure to make the class containing the CObject base
class the left-most class in the list of base classes.

For advice on the uses and abuses of MI, see Advanced C++ Programming Styles a
nd Idioms by James O. Coplien (Addison Wesley, 1992).

CObject - The Root of all Classes

As you know, all significant classes derive directly or indirectly from class
CObject. CObject does not have any member data, but does have some default fun
ctionality. When using MI, it will be common to inherit from two or more CObje
ct-derived classes, for example, a CFrameWnd and a CObList:

class CListWnd : public CFrameWnd, public CObList
{
...
};
CListWnd myListWnd;
In this case CObject is included twice, which leads to two problems:

Any reference to CObject member functions must be disambiguated.
myListWnd.Dump(afxDump);
    // compile time error, CFrameWnd::Dump or CObList::Dump ?
Static member functions, including 'operator new' and 'operator delete' must a
lso be disambiguated.
Recommended Steps

When creating a new class with two or more CObject derived base classes, reimp
lement those CObject members that you expect people to use. Operators new and
delete are mandatory, and Dump is recommended. For example:

class CListWnd : public CFrameWnd, public CObList
{
public:
    void* operator new(size_t nSize)
        { return CFrameWnd::operator new(nSize); }
    void operator delete(void* p)
        { CFrameWnd::operator delete(p); }

    void Dump(CDumpContent& dc)
        { CFrameWnd::Dump(dc);
          CObList::Dump(dc); }
     ...
};
Virtual Inheritance of CObject ?

You may ask, "If you inherit CObject virtually, won't all of the ambiguity pro
blems go away?"

Even in the efficient Microsoft Object Model, virtual inheritance is not as ef
ficient as nonvirtual inheritance (just as Multiple Inheritance is not as effi
cient as single inheritance in certain cases). Since there is no member data i
n CObject, virtual inheritance is not needed to prevent multiple copies of a b
ase class's member data.

The real answer is no, virtual inheritance will not solve the ambiguity proble
ms illustrated above. For example, the Dump virtual member function is still a
mbiguous (since CFrameWnd and CObList implement it differently).

Therefore we recommend following the steps above to provide disambiguation:

CObject::IsKindOf and Run-Time Typing

The run-time typing mechanism supported by MFC in CObject uses the macros DECL
ARE_DYNAMIC, IMPLEMENT_DYNAMIC, DECLARE_DYNCREATE, IMPLEMENT_DYNCREATE, DECLAR
E_SERIAL and IMPLEMENT_SERIAL. These give the ability to do a run-time type ch
eck to allow for safe cast-downs.

These macros only support a single base class and will work in a limited way f
or multiply inherited classes. The base class you specify in IMPLEMENT_DYNAMIC
or IMPLEMENT_SERIAL should be the first (or left-most) base class. For exampl
e,

class CListWnd : public CFrameWnd, public CObList
{
    DECLARE_DYNAMIC(CListWnd)
    ...
};
IMPLEMENT_DYNAMIC(CListWnd, CFrameWnd)
This will allow you to do type checking for the left-most base class only. The
run-time type system will know nothing about additional bases (CObList in thi
s case).

CWnd and Message Maps

For the MFC message map system to work correctly, there are two additional req
uirements:

There must be only one CWnd-derived base class.
The CWnd-derived base class must be the first (or left-most) base class.
In the example above, CFrameWnd is the first base class.

Some examples that will not work:

class CTwoWindows : public CFrameWnd, public CEdit
    { ... };
        // error : two copies of CWnd

class CListEdit : public CObList, public CEdit
    { ... };
        // error : CEdit (derived from CWnd) must be first
A Sample Program using MI

The following sample is a stand-alone application that consists of one class d
erived from CFrameWnd and CWinApp. This way of structuring an application is n
ot a recommended, but this is an example of the smallest MFC application with
one class.

You can cut the following program and copy it on top of HELLOAPP.CPP in the si
ngle-inheritance MFC General sample HELLOAPP. Then build the program as you wo
uld normally.

#include <afxwin.h>

class CHelloAppAndFrame : public CFrameWnd, public CWinApp
{
public:
    CHelloAppAndFrame()
        { }

    // Necessary evil for MI disambiguity
    void* operator new(size_t nSize)
        { return CFrameWnd::operator new(nSize); }
    void operator delete(void* p)
        { CFrameWnd::operator delete(p); }

// Implementation
    // CWinApp overrides
    virtual BOOL InitInstance();
    // CFrameWnd overrides
    virtual void PostNcDestroy();
    afx_msg void OnPaint();

    DECLARE_MESSAGE_MAP()

};

BEGIN_MESSAGE_MAP(CHelloAppAndFrame, CFrameWnd)
    ON_WM_PAINT()
END_MESSAGE_MAP()

// since the frame window is not allocated on the heap, we must
// override PostNCDestroy not to delete the frame object
void CHelloAppAndFrame::PostNcDestroy()
{
    // do nothing (do not call base class)
}

void CHelloAppAndFrame::OnPaint()
{
    CPaintDC dc(this);
    CRect rect;
    GetClientRect(rect);

    CString s = "Hello, Windows!";
    dc.SetTextAlign(TA_BASELINE | TA_CENTER);
    dc.SetTextColor(::GetSysColor(COLOR_WINDOWTEXT));
    dc.SetBkMode(TRANSPARENT);
    dc.TextOut(rect.right / 2, rect.bottom / 2, s);
}

// Application initialization
BOOL CHelloAppAndFrame::InitInstance()
{
    // first create the main frame
    if (!CFrameWnd::Create(NULL, "Multiple Inheritance Sample",
        WS_OVERLAPPEDWINDOW, rectDefault))
        return FALSE;

    // the application object is also a frame window
    m_pMainWnd = this;         
    ShowWindow(m_nCmdShow);
    return TRUE;
}

CHelloAppAndFrame theHelloAppAndFrame;

一时间有点接受不了，我不希望这是真的。百合上的朋友，百合上的博客，当然还有百合上的那些文章，可能都要离我们而去了。

在百合上了解了南大，受到了一些启发，当然学到了不少东西。

的确应该有个纪念。