Get Ready to Unleash the Power of UCanCode .NET

Data types definition

A TTF file consists of several tables, each table represent some data, regarding of its type. Some tables are required, some are not. We actually need only one of them, called "name", e.g. names table. This is the place where the font information is stored, like font name, copyright, trademark and more.

//This is TTF file header
typedef struct _tagTT_OFFSET_TABLE{
    USHORT uMajorVersion;
    USHORT uMinorVersion;
    USHORT uNumOfTables;
    USHORT uSearchRange;
    USHORT uEntrySelector;
    USHORT uRangeShift;
}TT_OFFSET_TABLE;

//Tables in TTF file and there placement and name (tag)
typedef struct _tagTT_TABLE_DIRECTORY{
    char szTag[4]; //table name
    ULONG uCheckSum; //Check sum
    ULONG uOffset; //Offset from beginning of file
    ULONG uLength; //length of the table in bytes
}TT_TABLE_DIRECTORY;

//Header of names table
typedef struct _tagTT_NAME_TABLE_HEADER{
    USHORT uFSelector; //format selector. Always 0
    USHORT uNRCount; //Name Records count
    USHORT uStorageOffset; //Offset for strings storage, 
                           //from start of the table
}TT_NAME_TABLE_HEADER;

//Record in names table
typedef struct _tagTT_NAME_RECORD{
    USHORT uPlatformID;
    USHORT uEncodingID;
    USHORT uLanguageID;
    USHORT uNameID;
    USHORT uStringLength;
    USHORT uStringOffset; //from start of storage area
}TT_NAME_RECORD;

Macros

Now only thing left is macros I was talking before. The macros definition looks like:

#define SWAPWORD(x) MAKEWORD(HIBYTE(x), LOBYTE(x))
#define SWAPLONG(x) MAKELONG(SWAPWORD(HIWORD(x)), SWAPWORD(LOWORD(x)))

Now what is that? The reason we need those macros is that TTF files are stored in Big-Endian format, unlike in Windows systems, where all files are in Little Endian. Yeah I know it sounds silly with all those "endians" :). Big Endian is used by Motorolla processors for example, where the higher byte is stored first, while in Little Endian (for Intel processors) the higher byte is the last. For example you have an integer variable 1 (which is 4 bytes long). Try to save it to file and open in any hexadecimal editor, you will see:

01 00 00 00    //Little Endian - Intel

This is Little Endian system (Intel). But for Big-Endian (Motorolla), the number will be stored vise versa:

00 00 00 01    //Big Endian - Motorolla

So these formats are incompatible. And TTF file as I said, is stored in Motorolla style (Big Endian). That's why we need those 2 macros to rearrange bytes in variables retrieved from TrueType font file.

Reading the file

Now we are prepared to read the TTF file. So let's get started.

First of all we need to read the file header (TT_OFFSET_TABLE structure):

CFile f;
CString csRetVal;

//lpszFilePath is the path to our font file
if(f.Open(lpszFilePath, CFile::modeRead|CFile::shareDenyWrite)){

    //define and read file header
    TT_OFFSET_TABLE ttOffsetTable;
    f.Read(&ttOffsetTable, sizeof(TT_OFFSET_TABLE));

    //remember to rearrange bytes in the field you gonna use
    ttOffsetTable.uNumOfTables = SWAPWORD(ttOffsetTable.uNumOfTables);
    ttOffsetTable.uMajorVersion = SWAPWORD(ttOffsetTable.uMajorVersion);
    ttOffsetTable.uMinorVersion = SWAPWORD(ttOffsetTable.uMinorVersion);

    //check is this is a true type font and the version is 1.0
    if(ttOffsetTable.uMajorVersion != 1 || ttOffsetTable.uMinorVersion != 0)
        return csRetVal;

Right after the file header goes Offsets Table. You can find here an offset to interesting you table, "name" in our case.

    TT_TABLE_DIRECTORY tblDir;
    BOOL bFound = FALSE;
    CString csTemp;

    for(int i=0; i< ttOffsetTable.uNumOfTables; i++){
        f.Read(&tblDir, sizeof(TT_TABLE_DIRECTORY));
        csTemp.Empty();

        //table's tag cannot exceed 4 characters
        strncpy(csTemp.GetBuffer(4), tblDir.szTag, 4);
        csTemp.ReleaseBuffer();
        if(csTemp.CompareNoCase(_T("name")) == 0){
            //we found our table. Rearrange order and quit the loop
            bFound = TRUE;
            tblDir.uLength = SWAPLONG(tblDir.uLength);
            tblDir.uOffset = SWAPLONG(tblDir.uOffset);
            break;
        }
    }

We finally found the names table, so let's read its header:

    if(bFound){
        //move to offset we got from Offsets Table
        f.Seek(tblDir.uOffset, CFile::begin);
        TT_NAME_TABLE_HEADER ttNTHeader;
        f.Read(&ttNTHeader, sizeof(TT_NAME_TABLE_HEADER));

        //again, don't forget to swap bytes!
        ttNTHeader.uNRCount = SWAPWORD(ttNTHeader.uNRCount);
        ttNTHeader.uStorageOffset = SWAPWORD(ttNTHeader.uStorageOffset);
        TT_NAME_RECORD ttRecord;
        bFound = FALSE;

Right after the Names Table header, go records in it. So we need to run through all records to find information interesting to us - font name.

    for(int i=0; i<ttNTHeader.uNRCount; i++){
    f.Read(&ttRecord, sizeof(TT_NAME_RECORD));
    ttRecord.uNameID = SWAPWORD(ttRecord.uNameID);

    //1 says that this is font name. 0 for example determines copyright info
    if(ttRecord.uNameID == 1){
        ttRecord.uStringLength = SWAPWORD(ttRecord.uStringLength);
        ttRecord.uStringOffset = SWAPWORD(ttRecord.uStringOffset);

        //save file position, so we can return to continue with search
        int nPos = f.GetPosition();
        f.Seek(tblDir.uOffset + ttRecord.uStringOffset + 
                 ttNTHeader.uStorageOffset, CFile::begin);

        //bug fix: see the post by SimonSays to read more about it
        TCHAR lpszNameBuf = csTemp.GetBuffer(ttRecord.uStringLength + 1);
        ZeroMemory(lpszNameBuf, ttRecord.uStringLength + 1);
        f.Read(lpszNameBuf, ttRecord.uStringLength);
        csTemp.ReleaseBuffer();

        //yes, still need to check if the font name is not empty
        //if it is, continue the search
        if(csTemp.GetLength() > 0){
            csRetVal = csTemp;
            break;
        }
        f.Seek(nPos, CFile::begin);
    }
}