Creation and memory mapping existing DBF files as alternative of data serialization during work with modified CListCtrl classes in virtual mode on dialogs of MDI application

• Download ModifiedListCtrl04.zip - 121.34 KB

Introduction

At last we can set ourselves the task of conservation and read-out of necessary information for our application from files of data. In typical MFC program standard serialization of data is usually supported with CDocument classes. In principle nothing interferes to us to use this method for our aims. However, at once does a question get up in what format to keep information? Certainly we can invent an own data format, appropriate it our extension, for example, *.eee and to work with it as this will be comfortably us. Perhaps, for educational aims or there where principle of serialization of information has a substantial value, this method is very quite good. But, as we design working application in which it is necessary to have an random access to file memory, the use of serialization does not seem an optimum enough decision for us already. In addition, for development of own format of data are needed very weighty grounds. I think that satisfaction of own ego is not such cause.

So, at first, we must be determined with the existent format of data, and secondly, to define a method of random access to file information. At such wording, decision of these tasks are fully obviously. The simplest and well known format of data is structure of dbf-file which gives a good account of itself. And random access to content of files of data easily to carry out taking advantages of technology of memory mapped files – MMF. Main desert of this technology is in that we work with files as with ordinary memory. For databases files it is an excellent decision. There is not a necessity preliminary to read information in ordinary memory for their further use, although we can do it for small but often using information, with the purpose of optimization, for example for meta data, describing structure of a database. In addition, an obvious necessity to be engaged in spooling of data falls off, that is very comfortably for filling of CListCtrl classes and its descendants in the virtual mode. Except for it, we can directly use static structure of database for random access to its elements of data. Plus lightness of the use of technology of MMF. Advantages are so much, that has no sense to talk about serialization of information, while practically at those efforts we get all of advantages of random access to memory mapped files.

On this stage we define organization of work thus. At opening a certain table form, the program searches proper by it a dbf-file, described in our meta data. If finds, tries to open its and use its information for filling of a proper list, otherwise creates this dbf-file, reading information from our static variables. If to change such file by some external editor program of dbf-files, at repeated his read-out we will see new information already. Later we will learn to edit dbf-files directly in our program. This project is already supplied with the prepared files, located in the Dbf folder. Its names are «First.dbf», «Second.dbf» and «Third.dbf». They have an alike structure (in the third file the fields «Name» and «Title» moved between itself) but its data are sorted variously. In all from them there are 360 records. If for example we delete the file «Third .dbf», instead of its our program will be to create the same file and will fill its data with our static variables. As the result we will get an example indicated on a Fig. 1.

1. Structure of DBF file

In the Internet is not very much hardness to find necessary information about dbf-files, as this format is opened and a long ago known. However succeeded finding an official specification was not. Therefore we will expound short results about their structure, see Fig. 2.

If we know beforehand an amount of fields FLDCOUNT = N in a database, their lengths FLDLEN1 = M1, FLDLEN2 = M2, . . ., FLDLENn = Mn and number of records RECCOUNT = K, the proper structure of dbf-file can be defined as:

typedef struct {
    DBF_HEADER DbfHdr;
    DBF_FIELD aDbfField[FLDCOUNT];  // FLDCOUNT = (DbfHdr.wFirstRec-296)/32 for VFP
    BYTE cHdrEnd;  // Header record terminator = 0x0D (13)
    BYTE acDbcFile[263];  // A data range for name of associated *.dbc file or contains 0x00
    DBF_RECORD aDbfRec[RECCOUNT];  // RECCOUNT = DbfHdr.nRecCount
    BYTE cFileEnd;  // File record terminator = 0x1A (26)
} DBF_FILE;

where the proper structures are determined as:

typedef struct {
    /*00-00*/ BYTE cType;  // File type = 0x30 (48) for Visual FoxPro (VFP)
    /*01-01*/ BYTE cYear;  // Year of last update (for VFP nFullYear = 2000 + cYear)
    /*02-02*/ BYTE cMonth;  // Month of last update
    /*03-03*/ BYTE cDay;  // Day of last update
    /*04-07*/ ULONG nRecCount;    // Number of records in file
    /*08-09*/ WORD wFirstRec;  // Position of first data record
    /*10-11*/ WORD wRecSize;  // Length of one data record, including delete flag
    /*12-27*/ BYTE cReserv1227[16];  // Reserved, contains 0x00
    /*28-28*/ BYTE cFlag;  // Table Flags (Only Visual FoxPro)
    /*29-29*/ BYTE cCodePage;  // Code page mark = 0x03 (3) for code page 1252 (Windows ANSI)
    /*30-31*/ WORD wReserv3031;  // Reserved, contains 0x00
} DBF_HEADER;

typedef struct {
    /*00-10*/ BYTE cName[11];  // Field name with null terminator (0x00)
    /*11-11*/ BYTE cType;  // Field type (C – Character; N – Numeric; D – Date and so on)
    /*12-15*/ ULONG nOffset;  // Displacement of field in record
    /*16-16*/ BYTE cLen;  // Length of field (in bytes)
    /*17-17*/ BYTE cDec;  // Number of decimal places
    /*18-18*/ BYTE cFlag;  // Field flag
    /*19-22*/ LONG nNext;  // Value of auto increment Next value 
    /*23-23*/ BYTE cStep;  // Value of auto increment Step value 
    /*24-31*/ BYTE cReserv2431[8];  // Reserved, contains 0x00
} DBF_FIELD;

typedef struct {
    BYTE cDelete;  // Delete flag
    BYTE cData1[FLDLEN1];  // Where FLDLEN1 = aDbfField[0].cLen
    BYTE cData2[FLDLEN2],  // Where FLDLEN2 = aDbfField[1].cLen
    . . .
    BYTE cDataN[FLDLENN];  // Where FLDLENN = aDbfField[N-1].cLen
} DBF_RECORD;

Here is showed the static structure of dbf-file actually (we are oriented on the VFP format with the file type 0x30 (48)). For dynamic structures in ordinary memory, organization of data must be other – through pointers (size of which is known) to complete structures of data, sizes of which are unknown on the stage of compiling of program (or not desirable for the use, not to limit to ourselves with a concrete structure of one file). And as the compiler of VS6 C++ can not create dynamic types of data, size of which turns out only on the stage of executing of program (runtime), we can not use the structure of data obviously described higher for direct manipulation file information. But we also do not wish preliminary to read information of records in ordinary memory, to analyze and convert its directly to use in future. The same we lose advantages of the MMF. Thus, gets up the task of maximal description of dynamic structures of data with static ones, with the purpose of effective selection of elements of data from databases (in this case dbf-files) on the stage of executing of the program.

2. Determination of dynamic structures (unknown sizes) via static ones

This theme is important enough in itself, to spare it a little bit of attention. Ordinary compilers can not create dynamic types of data, when sizes of structures are ascertained in runtime. Just the «data type» bears in mind but not a dynamical storage area is reserved under it. We will illustrate said the known example. We will assume want to create an array of bytes for length which we determine dynamically, for example:

UINT nX = 100;
BYTE acData[nX];

But on this very simple construction a compiler will «swear». instead of nX it wants to see constant expression, as though

BYTE acData[100];

Thus, constant expression must be indicated obviously. For example, expression:

UINT nCount  = sizeof(aKnownStructure)/ sizeof(aKnownStructure[0]);
acData[nCount];

will be erroneous, while the direct use:

acData[sizeof(aKnownStructure)/ sizeof(aKnownStructure[0])];

will be faithful.

Nevertheless, building the indicated array of the dynamically calculated size is possible if only to give up obvious static determination of array and substitute it by non-obvious static determination. I.e. instead of

BYTE acData[nX];  // Correctly, BYTE acData[100];

we write

BYTE *acData = new BYTE[nX];  // UINT nX = 100;

A difference is here in that in first case of sizeof(acData) = 100, and in second sizeof(acData) = 4. I.e. in second case, actually we has not dynamic type, measuring 100 bytes, and static one of length 4 bytes, because it is a size of pointer to a dynamically selected storage area measuring 100 bytes. From viewpoint of the use of these elements of arrays are differences no, and from viewpoint of sizes of in-use structures difference is substantial. For an example will show yet how dynamically to build a bi-dimensional array.

//*** Dynamic array of field names of dbf file
BYTE **aaсFldName = new BYTE *[m_nFldCount];

//*** Initializes arrays of names of dbf fields
for(int i = 0; i < m_nFldCount; i++) {
    //*** Field name with a maximum of 10 characters with null terminator
    m_aacFldName[i] = new BYTE[11];

    //*** Copies field name
    for(int j = 0; j < 11; j++)
            m_aacFldName[i][j] = aDbfField[i].acName[j];
}

It is the real code from our application. Apparently, this technique is good for copying of dynamic structures of unknown beforehand (on the stage of compiling) size. But that well for often in-use and not largeness of meta data, is not very much well for enormous arrays of elements of data. It would be desirable directly to handle to them at their use (without a preliminary copying and analysis of structure of data). However for this purpose it is necessary beforehand to know key parameters of file of database, what on their basis to build the static types of complete structures of data. As it applies to the structure of dbf-file described higher, obvious knowledge of the parameters indicated there is needed. But as mentioned already, even if these parameters for this file know us, it is not interestingly us to specify them statically in our program, because it limits to us consideration only exactly this file. Certainly, we can beforehand describe the whole great number of dbf-files, in-use in our program, as it is ordinary and done, but however, if we wish to have an access to random dbf-file, this technique does not arrange us.

So, we walked up to the contradiction. From one side, static description of structure of database (as it is done at the beginning of the first section) allows us to take advantage memory mapped files, but limits to us the fixed set of these structures (i.e. by the certain amount of concrete types of files). From other side dynamic description of structures of data through pointers to structures, unknown beforehand sizes, results in the necessity of the use of flow of entrance data of unknown structure, that deprives us obvious advantages of MMF technology, namely random access to elements of a database.

For an example will specify that second a way was went author of «The alxBase classes for work with DBF files» by Alexey, therefore, MMF technician was not used him. He applies the classic method of retrieval of data from a dbf-file by the obvious positioning of file pointer and then reading of data elements. However one thing when we directly write a sort of (in right part it is file memory elements):

BYTE *acDataElement = acRecord[j].acField[i];  // ~ BYTE acDataElement[aDbfField[i].cLen];

and another, that at first we must calculate file pointer of current element of data, positioned on it and only after have read information. But even it is not main (for the dbf-file it is not difficult to do it by ourselves), but circumstance that it is needed to work at organization of share (multi-user , multi-thread) access to common data yet. As far as I understood Alexey limited to monopolistic access to database, that limits the use of his library substantially. It is visible already because he withdrew support his project from 2005 (at least, in public). In MMF technology considerably anymore possibilities for organization of share access to data, so that, I think, it is not needed to ignore this possibility.

But will go back to our «contradiction». Clear, that we are not arranged by neither first nor second way. If someone knows a third way, it would be interestingly to know about it. We will endeavour to unite these methods, so, as far as it is possible. As we do not can beforehand fully statically to describe a dbf-file structure, limited to then partial static description, thus by not one similar structure, but two. If we once again attentively will look at the structure of dbf-file resulted higher (which in such kind will not be «understood» by a compiler), easily to see that in it it’s possible to select two static parts, which already will be «clear» a compiler. Namely:

typedef struct {
    DBF_HEADER DbfHdr;  // Dbf header structure
    DBF_FIELD aDbfField[1];  // Really FLDCOUNT = (DbfHdr.wFirstRec-296)/32 fields
} DBF_FILE1;

typedef struct {
    BYTE cHdrEnd;  // Header record terminator = 0x0D (13)
    BYTE acDbcFile[263];  // A data range for associated *.dbc file or contains 0x00
    BYTE aDbfRec[1];  // Really DbfHdr.wRecSize * DbfHdr.nRecCount records
    //BYTE cFileEnd;  // File record terminator = 0x1A (26)
} DBF_FILE2;

Further, we will take advantage of the known «hacker» technique – conscious output outside the statically declared array of data. «Paying» for such combined approach will be a transition from «flat», «bi-dimensional» indexation (j, i) to «one-dimensional», linear indexation of ji = j*nColCount + i (for a count of linear number of cell of information) and necessity of calculation of general linear displacement nLineInd = j*nRowSize + m_anOff[i] (in bytes). But, I think, that it is not too large «price», for such approach.

Exactly these structures we will really attach to a dbf-file, applying after a conscious output outside indexes of the static arrays of structures aDbfField[1] and aDbfRec[1], by the mentioned linear indexes. As practice showed, it is fully good approach and it can be used for other data files, for example, for memo-fields (fpt-files), index cdx-files, database containers (dbc-files) etc.

3. MMF and reading of existent DBF files

Reading of having dbf-file takes a place by MMF technology, the use of which is presented in the function CMainDoc::OnOpenDocument:

/////////////////////////////////////////////////////////////////////////////
// OpenDocumentFile
/////////////////////////////////////////////////////////////////////////////
BOOL CMainDoc::OnOpenDocument(LPCTSTR szFileName) {  // szFileName isn't using
    TCHAR *szDbfName = m_MetaTable.szDbfName;

    CFileStatus FileStatus;

    //*** If file szDbfName does not exist creates its
    if(!CFile::GetStatus(szDbfName, FileStatus)) {
        //*** Creates current document (dbf) file on physical disk
        if(!CreateDocumentFile(szDbfName)) {
            //_M("CMainDoc: Failed to create a document file!");
            return FALSE;
        }
    }

    //*** Gets handle of dbf file
    m_hDbfFile = ::CreateFile(
            szDbfName,  // Name of dbf file
            GENERIC_READ | GENERIC_WRITE,  // Access (read-write) mode
            FILE_SHARE_READ | FILE_SHARE_WRITE,  // Share mode
            NULL,  // Pointer to security attributes
            OPEN_EXISTING,  // How to create
            FILE_ATTRIBUTE_NORMAL,  // File attributes
            NULL  // HANDLE hTemplateFile - Handle to file with attributes to copy
    );

    if(m_hDbfFile == INVALID_HANDLE_VALUE) {
        _M("CMainDoc: Failed to call ::CreateFile function!");
        return FALSE;
    }

    //*** The message buffer
    TCHAR szStr[MAXITEMTEXT];

    //*** Gets file size in bytes
    ULONG nDbfSize = ::GetFileSize(
        m_hDbfFile,  // Handle of file to get size of
        NULL  // Pointer to high-order word for file size
    );

    //*** Checks file size
    if(nDbfSize == 0) {
        swprintf(
                szStr,
                _T("CMainDoc: File '%s' is empty!"),
                szDbfName
        );

        _M(szStr);
        ::CloseHandle(m_hDbfFile);
        return FALSE;
    }

    //*** Creates file mapping
    m_hDbfMap = ::CreateFileMapping(
            m_hDbfFile,  // Handle to file to map
            NULL,  // Optional security attributes
            PAGE_READWRITE,  // Protection for mapping object
            0,  // High-order 32 bits of object size
            0,  // Low-order 32 bits of object size
            NULL  // Name of file-mapping object
    );

    if(!m_hDbfMap) {
        _M("CMainDoc: Failed to call ::CreateFileMapping function!");
        return FALSE;
    }

    //*** Maps view of dbf file for its first part (where are header)
    m_pDbfView1 = reinterpret_cast<DBF_FILE1 *>(::MapViewOfFile(
            m_hDbfMap,  // File-mapping object to map into address space
            FILE_MAP_WRITE,  // Access mode
            0,  // High-order 32 bits of file offset
            0,  // Low-order 32 bits of file offset
            0  // Number of bytes to map (if it is zero, the entire file is mapped)
    ));

    if(!m_pDbfView1) {
        _M("CMainDoc: Failed to call ::MapViewOfFile function!");
        return FALSE;
    }

    //*** Dbf header structure
    m_pDbfHdr = &m_pDbfView1->DbfHdr;

    //*** Checks dbf file type
    if(m_pDbfHdr->cType != VFPTYPE) {  // = 0x30 (48)
        swprintf(
                szStr,
                _T("CMainDoc: Dbf type: %d doesn't equal to VFP type: %d!"),
                m_pDbfHdr->cType,
                VFPTYPE
        );

        _M(szStr);
        return FALSE;
    }
    /*
    //*** Shows date of last update
    
    swprintf(
            szStr, 
            _T("Date of last update is %0.2d.%0.2d.%d"), 
            m_pDbfHdr->cDay, 
            m_pDbfHdr->cMonth, 
            BASEYEAR + m_pDbfHdr->cYear
    );

    _M(szStr);
    */
    //*** Number of records in file
    m_nRecCount = m_pDbfHdr->nRecCount;

    //*** Length of one data record (including delete flag)
    m_nRecSize = m_pDbfHdr->wRecSize;

    //*** Checks record size
    if((m_nRecSize == 0 && m_nRecCount != 0) ||
         (m_nRecSize != 0 && m_nRecCount == 0)) {
        swprintf(
                szStr,
                _T("CMainDoc: Not matches record size (%d) and record count (%d)!"),
                m_nRecSize,
                m_nRecCount
        );

        _M(szStr);
        return FALSE;
    }

    //*** Calculates size of all records
    ULONG nDataSize = nDbfSize - 1 - m_pDbfHdr->wFirstRec;

    //*** Checks record parameters
    if(nDataSize != m_nRecCount*m_nRecSize) {
        swprintf(
                szStr,
                _T("CMainDoc: Data size (%d) doesn't equal record count (%d) * record size (%d)!"),
                nDataSize,
                m_nRecCount,
                m_nRecSize
        );

        _M(szStr);
        return FALSE;
    }

    //*** Number of fields in file (for Visual FoxPro only)
    m_nFldCount = (m_pDbfHdr->wFirstRec - 296)/32;

    //*** Checks field count
    if(m_nFldCount > m_nRecSize - 1) {
        _M("CMainDoc: Field count is very large!");
        return FALSE;
    }

    //*** Dbf field structure
    DBF_FIELD *aDbfField = m_pDbfView1->aDbfField;

    //*** Maps view of dbf file for its first part (where are data)
    m_pDbfView2 = reinterpret_cast<DBF_FILE2 *>(
            &m_pDbfView1->aDbfField[m_nFldCount].acName[0]
    );

    BYTE cHdrEnd = 0;

    //*** Checks dbf reading
    try {
        cHdrEnd = m_pDbfView2->cHdrEnd;
  } catch(...) {
        _M("CMainDoc: Dbf file has wrong structure!");
        return FALSE;
    }

    //*** Checks dbf header record terminator
    if(cHdrEnd != HEADEREND) {  // = 0x0D (13)
        swprintf(
                szStr, 
                _T("CMainDoc: Header record terminator: %d doesn't equal to: %d!"),
                m_pDbfView2->cHdrEnd,
                HEADEREND
        );

        _M(szStr);
        return FALSE;
    }

    //*** Delete flag  // = " " or "*"
    //_M(pDbfFile2->aDbfRec[0]);

    //*** Dynamic array of field names
    m_aacFldName = new BYTE *[m_nFldCount];

    //*** Dynamic array of field types
    m_acFldType = new BYTE[m_nFldCount];

    //*** Dynamic array of offsets
    m_anOff = new UINT[m_nFldCount];
    
    //*** Dynamic array of lengths
    m_acLen = new BYTE[m_nFldCount];
    
    //*** Dynamic array of decimal places
    m_acDec = new BYTE[m_nFldCount];

    //*** Initializes arrays of length, offsets and etc. of dbf fields
    for(int i = 0; i < m_nFldCount; i++) {
        //*** Field name with a maximum of 10 characters with null terminator
        m_aacFldName[i] = new BYTE[11];

        //*** Copies field name
        for(int j = 0; j < 11; j++)
                m_aacFldName[i][j] = aDbfField[i].acName[j];

        //*** Field type
        m_acFldType[i] = aDbfField[i].cType;

        //*** Field lenght (in bytes)
        m_acLen[i] = aDbfField[i].cLen;

        //*** Number of decimal places (in bytes)
        m_anOff[i] = aDbfField[i].nOffset;

        //*** Number of decimal places (in bytes)
        m_acDec[i] = aDbfField[i].cDec;
    }

    //*** Testing for all field of j-th record
    /*
    //*** j-th record
    ULONG j = 11;
    
    //*** Line displacement of i-th field of j-th record
    ULONG ji = 0;
    
    for(i = 0; i < m_nFldCount; i++) {
        ji = j*m_nRecSize + m_anOff[i];

        //*** The copy of (j, i) field value of m_anLen[i]-th length
        // As it hasn't null terminator
        CString sFldVal((LPCSTR) &m_pDbfMap2->aDbfRec[ji], m_anLen[i]);

        sFldVal.TrimLeft();
        sFldVal.TrimRight();

        swprintf(
                szStr, 
                _T("%s : %c : %d : %d.%d :: '%s'"),
                (CString) aszFldName[i],  // As it has null terminator
                acFldType[i],
                m_anOff[i],
                m_anLen[i],
                m_anDec[i],
                sFldVal
        );

        _M(szStr);
    }
    */
    BYTE cFileEnd = 0;

    //*** Checks dbf reading
    try {
        cFileEnd = m_pDbfView2->aDbfRec[m_nRecCount * m_nRecSize];
  } catch(...) {
        _M("CMainDoc: Dbf file has wrong structure!");
        return FALSE;
    }

    //*** Checks dbf file record terminator
    if(cFileEnd != DBFEND) {  // = 0x1A (26)
        swprintf(
                szStr, 
                 _T("CMainDoc: Header record terminator: %d doesn't equal to: %d!"),
                m_pDbfView2->aDbfRec[m_nRecCount * m_nRecSize], 
                DBFEND
        );

        _M(szStr);
        return FALSE;
    }

    //*** Current table
    CListCtrlEx *pTable = m_pMainApp->m_apTable[m_eTable];
    
    if(!pTable) {
        _M("CMainDoc: Empty a CListCtrlEx object!");
        return FALSE;
    }

    //*** Sets the table rows count in the virtual mode (LVS_OWNERDATA)
    //*** Send messages LVN_GETDISPINFOW & HDM_LAYOUT
    //*** Calls the CListCtrlEx::DrawItem
    pTable->SetItemCount(m_nRecCount);

    //*** Shows the vertical scroll bar always
    //pTable->ShowScrollBar(SB_VERT);

    //*** Saves the current document
    m_pMainApp->m_apDoc[m_eTable] = this;

    return TRUE;
}  // OnOpenDocument

4. Creation of DBF files and filling their with static data

To create a file of database, it is needed preliminary to know its structure. In our demonstration program there are three such static structures which allow to create three different dbf-files. Changing their amount and content it is possible to create random enough files of the Visual FoxPro format of databases. Here are basic structures of our meta data:

typedef struct {
    TCHAR *szFldName;  // Field name
    TCHAR *szFldType;  // Field type
    UINT nFldLen;  // Field length (in bytes)
    UINT nDecLen;  // Number of decimal places (in bytes)
} META_DATA;

typedef struct {
    TCHAR *szHdrName;  // Column name
    DWORD nAdjust;  // Text formatting
    UINT nWidth;  // Column width
} META_HEADER;

typedef struct {
    TCHAR *szDbfName;  // Dbf name
    META_DATA *aMetaData;  // Dbf-file fields data structure
    TCHAR *szTblName;  // Table name
    META_HEADER *apMetaHeader;  // Meta table header structure
    DWORD dwStyle;  // Table style
    DWORD dwExStyle;  // Extended table style
    RECT *pFrmRect;  // Frame rectangle pointer
    RECT *pViewRect;  // View rectangle pointer
    CFont *pHdrFont;  // Table header font pointer
    CFont *pListFont;  // Table list font pointer
    UINT nHdrHeight;  // Table header height
    UINT nListHeight;  // Table list height
    UINT nColCount;  // Table header columns count
    UINT nRowCount;  // Table list row count
    TCHAR **apRowText;  // Table rows text array
} META_TABLE;

Using these structures the function CMainApp::CreateDocumentFile creates a necessary dbf-file:

/////////////////////////////////////////////////////////////////////////////
// CreateDocumentFile
/////////////////////////////////////////////////////////////////////////////
BOOL CMainDoc::CreateDocumentFile(TCHAR *szDbfName) {
    //*** The dbf file structure
    /*
    typedef struct {
        DBF_HEADER DbfHdr;
        DBF_FIELD aDbfField[FLDCOUNT];  // FLDCOUNT = (DbfHdr.wFirstRec-296)/32 for VFP
        BYTE cHdrEnd;  // Header record terminator = 0x0D (13)
        BYTE acDbcFile[263];  // A data range for associated *.dbc file, contains 0x00
        DBF_RECORD aDbfRec[RECCOUNT];  // RECCOUNT = DbfHdr.nRecCount
        BYTE cFileEnd;  // File record terminator = 0x1A (26)
    } DBF_FILE;
    */
    CFileStatus FileStatus;

    //*** If file szDbfName does exist simply return
    if(CFile::GetStatus(szDbfName, FileStatus))
            return TRUE;

    //*** Number of records in file
    ULONG nRecCount = m_MetaTable.nRowCount;

    //*** Data table fields count
    UINT nFldCount = m_MetaTable.nColCount;

    SYSTEMTIME SysTime = {0};

    //*** Gets system date and time
    GetSystemTime(&SysTime);

    //*** Dbf header structure

    DBF_HEADER DbfHdr = {0};

    DbfHdr.cType = VFPTYPE;  // DBF type
    DbfHdr.cYear = SysTime.wYear%BASEYEAR;  // Year of last update
    DbfHdr.cMonth = SysTime.wMonth;  // Month of last update
    DbfHdr.cDay = SysTime.wDay;  // Day of last update
    DbfHdr.nRecCount = nRecCount;  // Number of records in file
    //DbfHdr.wFirstRec = 0;    // Position of first data record
    //DbfHdr.wRecSize = 0;  // Length of one data record, including delete flag
    //DbfHdr.cFlag = 0;     // Table Flags (Only Visual FoxPro)
    DbfHdr.cCodePage = CP1252;    // Windows ANSI

    //*** Dbf field structure

    DBF_FIELD DbfField = {0};

    //*** Gets handle of dbf file
    HANDLE hDbfFile = ::CreateFile(
            szDbfName,  // Pointer to name of the dbf file
            GENERIC_WRITE,  // Access (read-write) mode
            0,  // Share mode
            NULL,  // Pointer to security attributes
            CREATE_ALWAYS,  // How to create
            FILE_ATTRIBUTE_NORMAL,  // File attributes
            NULL  // HANDLE hTemplateFile - Handle to file with attributes to copy
    );

    //*** Message buffer
    TCHAR szStr[MAXITEMTEXT];

    //*** Checks file creation
    if(hDbfFile == INVALID_HANDLE_VALUE) {
        swprintf(
                szStr,
                _T("CMainApp: Failed to create new file: '%s'!"),
                szDbfName
        );

        _M(szStr);
        ::CloseHandle(hDbfFile);
        return FALSE;
    }

    //*** Number of written bytes
    DWORD dwBytes = 0;

    //*** Writes bytes into file
    WriteFile(hDbfFile, &DbfHdr, sizeof(DbfHdr), &dwBytes, NULL);

    //*** Dinamic array of field length
    BYTE *acFldLen = new BYTE[nFldCount];

    //*** Dinamic array of field types
    BYTE *acFldType = new BYTE[nFldCount];

    UINT nOffset = 1;  // Skips delete byte

    //*** Writes array of DBF_FIELD aDbfField[FLDCOUNT] structures
    for(int i = 0; i < nFldCount; i++) {
        META_DATA MetaData = m_MetaTable.aMetaData[i];
        BYTE *acName = DbfField.acName;
        TCHAR *szFldName = MetaData.szFldName;

        //*** Field name with a maximum of 10 characters, a rest is padded 
        // with 0x00
        //for(int j = 0; j < 11; j++)
                //acName[j] = szFldName[j];
        
        //*** Simply copies
        while(*acName++ = *szFldName++);

        acFldType[i] = MetaData.szFldType[0];  // Field type
        DbfField.cType = acFldType[i];  // Field type
        DbfField.nOffset = nOffset;  // Displacement of field in record
        acFldLen[i] = MetaData.nFldLen;  // Length of field (in bytes)
        DbfField.cLen = acFldLen[i];  // Length of field (in bytes)
        DbfField.cDec = MetaData.nDecLen;  // Number of decimal places

        nOffset += acFldLen[i];

        //*** Writes bytes into file
        WriteFile(hDbfFile, &DbfField, sizeof(DbfField), &dwBytes, NULL);
    }

    //*** Length of one data record, including delete flag
    DbfHdr.wRecSize = nOffset;  // ARE NOT WRITTEN YET!

    //*** Header record terminator
    BYTE cHdrEnd = HEADEREND;  // = 0x0D (13)
    
    //*** Writes bytes into file
    WriteFile(hDbfFile, &cHdrEnd, sizeof(cHdrEnd), &dwBytes, NULL);

    //*** A data range for associated *.dbc file, contains 0x00
    BYTE acDbcFile[263] = {0};

    //*** Writes bytes into file
    WriteFile(hDbfFile, &acDbcFile, sizeof(acDbcFile), &dwBytes, NULL);

    //*** Gets current file pointer
    DWORD nCurOffset = SetFilePointer(hDbfFile, 0, NULL, FILE_CURRENT);

    //*** Position of first data record
    DbfHdr.wFirstRec = nCurOffset;  // ARE NOT WRITTEN YET!

    //*** The delete flag
    BYTE cDelete = 32;  // = 0x20 (" ")

    //*** Line table cell index
    UINT ji = 0;

    //*** Writes array of DBF_RECORD aDbfRec[RECCOUNT] structures
    for(ULONG j = 0; j < nRecCount; j++) {
        //*** Writes bytes into file
        WriteFile(hDbfFile, &cDelete, sizeof(cDelete), &dwBytes, NULL);

        //*** Writes array of BYTE's strings
        for(i = 0; i < nFldCount; i++) {
            ji = j*nFldCount + i;  // Line table cell index
            TCHAR *acRowText = m_MetaTable.apRowText[ji];

            BYTE cFldLen = acFldLen[i];
            BYTE cFldType = acFldType[i];

            //*** Dinamic array of field data
            BYTE *acFldData = new BYTE[cFldLen + 2];  // +2 for sake date format

            //*** Copies field data (into BYTEs from TCHARs)
            for(int k = 0; k < cFldLen; k++)
                    acFldData[k] = acRowText[k];

            //*** Formates our date string (DD.MM.YYYY) into dbf style (YYYYMMDD)
            if(cFldType == 68) {  // = 0x44 ("D") - Date
                if(cFldLen != 8) {
                    swprintf(
                            szStr, 
                            _T("CMainApp: Length of date format is %d. Must be 8!"), 
                            cFldLen
                    );

                    _M(szStr);
                    return FALSE;
                }

                if(cFldLen == 8) {  // Date length for dbf date format
                    //*** Our static date has 10 characters
                    acFldData[8] = acRowText[8];
                    acFldData[9] = acRowText[9];
                    
                    //*** Date format is d1d2.m1m2.y1y2y3y4 . Must be y1y2y3y4m1m2d1d2
                    acFldData[2] = acFldData[8];  // Writes y3
                    acFldData[5] = acFldData[4];  // Writes m2
                    acFldData[4] = acFldData[3];  // Writes m1
                    acFldData[3] = acFldData[9];  // Writes y4
                    acFldData[8] = acFldData[0];  // Saves d1
                    acFldData[9] = acFldData[1];  // Saves d2
                    acFldData[0] = acFldData[6];  // Writes y1
                    acFldData[1] = acFldData[7];  // Writes y2
                    acFldData[6] = acFldData[8];  // Writes d1
                    acFldData[7] = acFldData[9];  // Writes d2
                } 
                //*** Else do nothing
            }

            //*** Writes bytes into file
            WriteFile(hDbfFile, acFldData, cFldLen, &dwBytes, NULL);
        }
    }

    //*** File record terminator
    BYTE cFileEnd = DBFEND;  // = 0x1A (26)

    //*** Writes bytes into file
    WriteFile(hDbfFile, &cFileEnd, sizeof(cFileEnd), &dwBytes, NULL);

    //*** Calculates file pointer to DbfHdr.wFirstRec
    ULONG nPos = (ULONG) &DbfHdr.wFirstRec - (ULONG) &DbfHdr.cType;  // = 8
    WORD wFirstRec = DbfHdr.wFirstRec;
    WORD wRecSize = DbfHdr.wRecSize;

    //*** Sets file pointer in DbfHdr.wRecSize position
    SetFilePointer(hDbfFile, nPos, NULL, FILE_BEGIN);

    //*** Writes NOT WRITTEN YET bytes into file
    WriteFile(hDbfFile, &wFirstRec, sizeof(wFirstRec), &dwBytes, NULL);

    //*** Writes next NOT WRITTEN YET bytes into file
    WriteFile(hDbfFile, &wRecSize, sizeof(wRecSize), &dwBytes, NULL);

    ::CloseHandle(hDbfFile);
    
    return TRUE;
}  // CreateDocumentFile

5. Processing of data in the Virtual Mode

In conclusion we will show the code of LVN_GETDISPINFO handler, which is «responsible» for the virtual mode.

/////////////////////////////////////////////////////////////////////////////
// OnChildNotify
/////////////////////////////////////////////////////////////////////////////
BOOL CListCtrlEx::OnChildNotify(UINT message, WPARAM wParam, LPARAM lParam, LRESULT *pResult) {
    NMHDR *pNMHdr = reinterpret_cast<NMHDR *>(lParam);

    LV_DISPINFO *pLVDI = reinterpret_cast<LV_DISPINFO *>(lParam);
    LV_ITEM *pItem = &pLVDI->item;

    if(message == WM_NOTIFY) {
        switch(pNMHdr->code) {
            case LVN_GETDISPINFO: {
                if(pItem->mask & LVIF_TEXT) {
                    //*** Item row
                    UINT nRow = pItem->iItem;

                    //*** Item column
                    UINT nCol = pItem->iSubItem;

                    //*** The message buffer
                    TCHAR szStr[MAXITEMTEXT];

                    //*** Current document
                    m_pDoc = m_pMainApp->m_apDoc[m_eTable];

                    if(!m_pDoc) {
                        _M("CListCtrlEx::CListCtrlEx : Empty document!");
                        //*** Forces to exit from the application as else will be a lot messages
                        exit(-1);
                    }

                    //*** Number of fields in dbf file
                    ULONG nColCount = m_pDoc->m_nFldCount;  // = m_MetaTable.nColCount;

                    if(nColCount == 0) {
                        _M("CListCtrlEx::OnChildNotify : Column count = 0!");
                        //*** Forces to exit from the application as else will be a lot messages
                        exit(-1);
                    }

                    if(m_MetaTable.nColCount != nColCount) {
                        swprintf(
                                szStr, 
                                _T("CListCtrlEx::OnChildNotify : Table (%d) and Dbf (%d) columns are different!"), 
                                m_MetaTable.nColCount, 
                                nColCount
                        );

                        _M(szStr);
                        //*** Forces to exit from the application as else will be a lot messages
                        exit(-1);
                    }

                    //*** Number of records in dbf file
                    ULONG nRowCount = m_pDoc->m_nRecCount;

                    if(nRowCount == 0) {
                        _M("CListCtrlEx::OnChildNotify : Row count = 0!");
                        //*** Forces to exit from the application as else will be a lot messages
                        exit(-1);
                    }

                    //*** Sets row count into meta table
                    //if(m_MetaTable.nRowCount != nRowCount)
                    m_MetaTable.nRowCount = nRowCount;
                    
                    //*** Length of one data record (including delete flag)
                    ULONG nRowSize = m_pDoc->m_nRecSize;

                    if(nRowSize == 0) {
                        _M("CListCtrlEx::OnChildNotify : Row size = 0!");
                        //*** Forces to exit from the application as else will be a lot messages
                        exit(-1);
                    }

                    //*** Line displacement of nCol-th field of nRow-th record
                    ULONG nLineInd = nRow*nRowSize + m_pDoc->m_anOff[nCol];

                    if(nLineInd < 0) {
                        _M("CListCtrlEx::OnChildNotify : Line index < 0!");
                        //*** Forces to exit from the application as else will be a lot messages
                        exit(-1);
                    }

                    //*** Maps view of dbf file for its second part (where are data)
                    DBF_FILE2 *m_pDbfView2 = m_pDoc->m_pDbfView2;

                    if(!m_pDbfView2) {
                        _M("CListCtrlEx::OnChildNotify : Empty view of dbf file!");
                        //*** Forces to exit from the application as else will be a lot messages
                        exit(-1);
                    }

                    //*** The Visual FoxPro records
                    BYTE *aDbfRec = m_pDbfView2->aDbfRec;

                    if(!aDbfRec) {
                        _M("CListCtrlEx::OnChildNotify : Empty array of dbf file!");
                        //*** Forces to exit from the application as else will be a lot messages
                        exit(-1);
                    }

                    if(!aDbfRec) {
                        _M("CListCtrlEx::OnChildNotify : Empty records into dbf file!");
                        //*** Forces to exit from the application as else will be a lot messages
                        exit(-1);
                    }

                    //*** Line table cell index
                    UINT ji = nRow*nColCount + nCol;

                    if(ji < 0) {
                        _M("CListCtrlEx::OnChildNotify : Line cell index < 0!");
                        //*** Forces to exit from the application as else will be a lot messages
                        exit(-1);
                    }

                    if(ji < nRowCount*nColCount) {
                        try {
                            //*** Copies (j, i) field value of m_anLen[i]-th length
                            // We do this as it hasn't null terminator
                            CString sFldVal(
                                    (LPCSTR) &aDbfRec[nLineInd], 
                                    m_pDoc->m_acLen[nCol]
                            );

                            //*** Formates date string into German style
                            if(m_pDoc->m_acFldType[nCol] == 68) {  // = 0x44 ("D") - Date
                                CString sDate = 
                                        sFldVal.Right(2) + _T(".") + 
                                        sFldVal.Mid(4, 2) + _T(".") + 
                                        sFldVal.Left(4);
                                
                                sFldVal = sDate;
                            }

                            wcscpy((wchar_t *)(pItem->pszText), sFldVal);
                        } catch (...) {
                            //*** Forces to exit from the application as else will be a lot messages
                            exit(-1);
                        }
                    } else {
                        wcscpy((wchar_t *)(pItem->pszText), _T("***"));
                    }
                }

                break;
            }
        }
    }

    return CListCtrl::OnChildNotify(message, wParam, lParam, pResult);
}  // OnChildNotify