C / C++ / MFC

Re: OpenGL question : wglUseFontBitmaps and glGenLists

5-Jan-05 4:55

I'm trying to use wglUseFontBitmaps to display some text in OpenGL.

now, it's working fine, I'm doing something like :

 m_iDisplayListBase = 1000;

 // Delete existing display-list numbers.
 glDeleteLists(m_iDisplayListBase, 256);
 // Generate new display lists.
 m_iDisplayListStart = glGenLists(256);

 BOOL b = wglUseFontBitmaps (m_hDC, 0, 256, m_iDisplayListBase);

...

 glListBase(m_iDisplayListBase);
// now draw the characters in a string
 glColor4f( 0.8, 0.1, 0.8, 1);
 glRasterPos3i( 20, iHeight-10 - m_iFontHeight, 1);
 glCallLists (24, GL_UNSIGNED_BYTE, "Hello Win32 OpenGL World");

Now, the questions.

1 - When using glGenLists to generate a number of empty display lists ( to be filled by wglUseFontBitmaps ), it returns the number (ID?) of the first display list ? so, should the argument to glListBase be ID + someArbitraryValue ? or will this be done automatically ?

2- When using wglUseFontBitmaps, do I need to do something else if I want to recreate the display lists for a different font size ? can I reuse the same "font bitmaps" ?

Thanks.

Max.

Maximilien Lincourt
Your Head A Splode - Strong Bad

El Corazon5-Jan-05 7:03

El Corazon

5-Jan-05 7:03

Maximilien wrote:
1 - When using glGenLists to generate a number of empty display lists ( to be filled by wglUseFontBitmaps ), it returns the number (ID?) of the first display list ? so, should the argument to glListBase be ID + someArbitraryValue ? or will this be done automatically ?

What is actually happening is that the glListBase(n) is setting the base of the next set of calls to n. In your case your Display list base index for your font. the function glCallLists() will generate one call per ascii value, thus you generate a full 256 count for your font. But the actual call is the combination of the last glListBase(n) adding the ascii value from the glCallLists() for each character.

Maximilien wrote:
2- When using wglUseFontBitmaps, do I need to do something else if I want to recreate the display lists for a different font size ? can I reuse the same "font bitmaps" ?

I have 3 different font selections in my program. Remember that the callLists is the sum of the base plus the ascii value. So you allocate a new set of 256 lists, I would suggest switching to an index variable name that reflects the font so you don't get confused.

However, remember that your driver may handle display lists in system memory or compiled and stored on the graphics card. You are generating 256 bitmap image display lists for every font. If you want to have multiple fonts on the screen at the same time, you will need three different sets of display lists. If you want the user to choose his font, do so ahead of building the font, and generate only one from the user definition. Don't use more memory than you need to. You can also cheat, make sure the text is in 7bit form (0-127 ASCII) and only generate 128 display lists. If you do, don't forget to check the string or you will get unknown results for any >127 values.

I rewrote this code as I pasted to remove some "stuff" hope it helps. You can also look into texture based fonts which may be much faster on some hardware. Google for OpenGL tutorials at "Nehe" and you will find text tutorials.

<br />
int             BuildFont(char *fontname, int point, int bold)<br />
{<br />
  GLint           base;<br />
  HFONT__        *font;<br />
  struct HDC__   *h_DC;<br />
<br />
  base = glGenLists(256);<br />
  font = CreateFont(point,            // Height Of Font<br />
                    0,                 // Width Of Font<br />
                    0,                 // Angle Of Escapement<br />
                    0,                 // Orientation Angle<br />
                    bold ? FW_BOLD : FW_NORMAL, // Font Weight<br />
                    0,                 // Italic<br />
                    0,                 // Underline<br />
                    0,                 // Strikeout<br />
                    ANSI_CHARSET,      // Character Set Identifier<br />
                    OUT_TT_PRECIS,     // Output Precision<br />
                    CLIP_DEFAULT_PRECIS,  // Clipping Precision<br />
                    ANTIALIASED_QUALITY,  // Output Quality<br />
                    FIXED_PITCH,     // Family And Pitch<br />
                    fontname);         // Font Name<br />
<br />
  h_DC = GetDC(NULL);<br />
  SelectObject(h_DC, font);            // Selects The Font We Want<br />
<br />
  wglUseFontBitmaps(h_DC, 0, 256, base);  // Builds 256 Characters for font<br />
  return base;<br />
}<br />
  m_LargeTextBase = BuildFont("Arial Bold", LargeTextPitch, LargeTextBold);<br />
  m_SmallTextBase = BuildFont("Arial Bold", SmallTextPitch, SmallTextBold);<br />
  m_MediumTextBase = BuildFont("Arial Bold", MediumTextPitch, MediumTextBold);<br />
...<br />
  glListBase(m_SmallTextBase );<br />
  glCallLists(strlen(string), GL_UNSIGNED_BYTE, string);<br />
...<br />
  glListBase(m_MediumTextBase);<br />
  glCallLists(strlen(string), GL_UNSIGNED_BYTE, string);<br />
...<br />
  glListBase(m_LargeTextBase);<br />
  glCallLists(strlen(string), GL_UNSIGNED_BYTE, string);<br />
<br />

_________________________
Asu no koto o ieba, tenjo de nezumi ga warau.
Talk about things of tomorrow and the mice in the ceiling laugh. (Japanese Proverb)

Re: OpenGL question : wglUseFontBitmaps and glGenLists

Maximilien5-Jan-05 7:21

5-Jan-05 7:21

Thanks for the Nehe reference, I will look at those tutorials tonight ...

I thought about using a couple of different font size display lists. I will make my life a bit easier; maybe that will be enough.

Thanks again.
Big Grin | :-D

Maximilien Lincourt
Your Head A Splode - Strong Bad

From C to C++ Pointer

jw815-Jan-05 4:55

5-Jan-05 4:55

I have C background but I need something like pointer so store two double variable data. Could anyone show me how or post some tutorial website here? Thanks.

David Crow5-Jan-05 5:24

5-Jan-05 5:24

One way:

double d1 = 5.1;
double *p1 = &d1;

Another:

double *p1 = new double;
*p1 = 1.8;

"Opinions are neither right nor wrong. I cannot change your opinion. I can, however, change what influences your opinion." - David Crow

jw815-Jan-05 5:26

5-Jan-05 5:26

Thx

I am using VC++, any easier class in MFC? The way you used seemed like C.

David Crow5-Jan-05 5:54

5-Jan-05 5:54

jw81 wrote:
any easier class in MFC?

What I provided was not easy?

Are you sure that you understand what you are asking? A pointer in C is the same pointer in C++. Encapsulating a simple pointer into a C++ class is unnecessary.

"Opinions are neither right nor wrong. I cannot change your opinion. I can, however, change what influences your opinion." - David Crow

jw815-Jan-05 15:38

5-Jan-05 15:38

My mistake, I was thinking abt linked list. Sorry abt the confusion. Does VC++ linked list same as C?

Thanks.

Maximilien5-Jan-05 16:14

5-Jan-05 16:14

yes; all C constructs are C++ compatible.

but I would suggest using either MFC collections or STL collections

Maximilien Lincourt
Your Head A Splode - Strong Bad

jw815-Jan-05 17:16

5-Jan-05 17:16

Thanks Maximilien,

This is the actual question I am asking, could you suggest any MFC function that I can use for such case (linked list)? There are quite a lot of solution from MFC but I dont know which one is the best. If it is too difficult to show here could you post any url related to such case? Thanks.

ThatsAlok5-Jan-05 22:05

ThatsAlok

5-Jan-05 22:05

jw81 wrote:
I can use for such case (linked list)?

Look for Collection Classes in MSDN. Some of them are

Clist
CPtrArray
CArray

or In STL

std::list
std::vector

"I Think this Will Help"
Alok Gupta
visit me at http://www.thisisalok.tk

Maximilien5-Jan-05 5:43

Re: How to draw static controls and check box buttons transparently ?

5-Jan-05 5:43

store where and how ?

double* d1;
double* d2;

d1 = new double;
*d1 = 0.4;

d2 = new double;
*d2 = 0.3;

Maximilien Lincourt
Your Head A Splode - Strong Bad

rrrado5-Jan-05 3:41

rrrado

5-Jan-05 3:41

Here is the
Screen Shot

The problem is that background is not solid, it has gradient so I can't erase background of my controls with solid rectrangle as defauld controls do.
(The code you've posted before also seems to be filling only solid rectangle)

rrrado

Re: How to draw static controls and check box buttons transparently ?

YoSilver7-Jan-05 0:35

YoSilver

7-Jan-05 0:35

Then the solution is to forward the WM_ERASEBKGND message of a control to the wnd proc of the parent dialog that is responsible for drawing a gradient background. Before passing a DC handle, set the viewport origin to match the control position in the dialog.

Another problem with group boxes is correct painting of contained sibling controls. If this problem will arise, try this :

void myGroupBox::PreSubclassWindow() 
{
    ASSERT(::GetWindowLong(::GetParent(m_hWnd), GWL_STYLE) & WS_CLIPCHILDREN);
    ASSERT((::GetWindowLong(m_hWnd, GWL_STYLE) & BS_TYPEMASK) == BS_GROUPBOX);

    ModifyStyle(0, WS_CLIPSIBLINGS);
    SetWindowPos(&CWnd::wndBottom, 0,0,0,0, SWP_NOSIZE|SWP_NOMOVE);

    CButton::PreSubclassWindow();
}

Group boxes are a common headache for all programmers. The simplest solution is to avoid gradient background and make it just lighter than common COLOR_BTNFACE, 75 percent - that will do and look neat.

One always gets the deserved.

http://www.silveragesoftware.com/hffr.html
Update your source code with my tool HandyFile Find And Replace!

Re: How to draw static controls and check box buttons transparently ?

rrrado11-Jan-05 4:04

rrrado

11-Jan-05 4:04

Thank you very much for your help, but finally I've found the easiest solution, without owner drawing nor any message handling, it seems that
all I need to use this function on my dialog in OnInitDialog():

EnableThemeDialogTexture()

Laugh | :laugh:

rrrado

Re: How to draw static controls and check box buttons transparently ?

YoSilver11-Jan-05 6:35

YoSilver

11-Jan-05 6:35

WOW !!!! in fact I have to thank You!!! I always used to invent a wheel for my dialogs in respect to they look, and never thought there's a function like this!!! Smile | :)

One always gets the deserved.

http://www.silveragesoftware.com/hffr.html
Update your source code with my tool HandyFile Find And Replace!

Accessing Large CMapStringToString's inside Critical Sections

MJWhiteman25-Jan-05 2:56

MJWhiteman2

5-Jan-05 2:56

I have a critical section that enables thread-safe reading from / writing to a CMapStringToString object.

The CMapStringToString could possibly grow in size to have around 200000 entries, where the 'key' is a string of <=6 characters, and the value will, on average, be around 300 characters.

Firstly, does anyone know how much memory a CMapStringToString object with 200000 entries with keys 6 characters long and values 300 characters long will actually use?

Secondly, when I call the Map's Lookup() function, will this start to take a long time to return if my map grows to this kind of size? Would this slowing down make it unsuitable to use inside a critical section which is probably being accessed by a different thread eg. 4 times per second?

Thanks for any thoughts anyone might have,

Matt

Re: Accessing Large CMapStringToString's inside Critical Sections

basementman5-Jan-05 4:38

basementman

5-Jan-05 4:38

Firstly, you should set your initial block size to a large number to create more hash buckets, thereby separating your strings into smaller sets and decreasing the search time.

If you need to access this many elements so frequently via multiple threads, I would suggest that you create a higher level hash table (of smaller size) that would contain locks (critical sections) and pointers to individual sections (hash tables) of the entire data set. By doing this, you will get higher concurrency at the cost of more memory usage. This way, you would lock only a section of the set that you were interested in, rather than the entire set.

onwards and upwards...

Critical Sections

MJWhiteman25-Jan-05 2:47

MJWhiteman2

5-Jan-05 2:47

I have a couple of questions about using the CCriticalSection class. Firstly, is there any difference between the way the critical section works by using it in the following two manners:

1)
CCriticalSection blocker;
blocker.Lock();
// Do something
blocker.Unlock();

2)
CCriticalSection blocker;
CSingleLock singleLock(&blocker);
singleLock.Lock();
// Do something
singleLock.Unlock();

Also, am I correct in saying that using either of these methods, no thread will time-out when waiting for the critical section to become free - they will just wait until it's they're turn to use the code?

Any comments would be much appreciated,

Thanks,

Matt

Re: Critical Sections

Martin Koorts5-Jan-05 3:25

Martin Koorts

5-Jan-05 3:25

Hi Matt

The CCriticalSection class is the lock that is accessed from more than 1 thread and ensures that if 1 thread has called CCriticalSection::Lock (not having called CCriticalSection::Unlock yet), another thread's call to CCriticalSection::Lock will block - effectively suspending the second thread until the first calls CCriticalSection::Unlock.

The CSingleLock class is simply a smart-pointer style helper class that acts on the CSyncObject that is passed in with it's constructor - in essence, it just calls the CSyncObject::Lock in it's constructor and CSyncObject::Unlock in the destructor. That way you can have the SyncObject's Unlock method automatically called when the CSingleLock local variable goes out-of-scope....
but an example says a thousand words;

CCriticalSection csGlobal; // global variable, or such..

void Thread1()
{
CSingleLock lock(&csGlobal);
// do stuff
}

void Thread2()
{
CSingleLock lock(&csGlobal);
// do stuff
}

So you can see that the smart-pointer style local CSingleLock instances really just capitalise on the C++ destructor to help you remember to call csGlobal.Unlock. Without the CSingleLock, you could end up writing (faulty) code like this;

void Thread3()
{
csGlobal.Lock();
// do some stuff
if (bFailed)
return; // !!!! didn't call csGlobal.Unlock();
// do more stuff
csGlobal.Unlock();
}

Lastly, the explicit Lock/Unlock methods on CSingleLock is really if you can't wait for the destructor, and need to Unlock and possibly Lock again.

I hope this makes sense.
Martin

C++ code size

Defenestration5-Jan-05 2:19

Defenestration

5-Jan-05 2:19

If I use a class which has 10 methods and 10 attributes, but I only access/use 5 methods and 5 attributes from that class in my program, will the resultant executable also include code/data for the other 5 methods and attributes ?

I suppose I'm asking if the linker is clever enough to discard unused functions and data.

PS. I'm using Visual C++ 6.

Re: C++ code size

David Crow5-Jan-05 5:22

5-Jan-05 5:22

At one point in time very long ago, I seem to remember that the linker would remove unused code from EXEs but not from DLLs. How it works today is unknown to me. Is this causing an issue with you?

"Opinions are neither right nor wrong. I cannot change your opinion. I can, however, change what influences your opinion." - David Crow

Re: C++ code size

Defenestration5-Jan-05 7:11

Defenestration

5-Jan-05 7:11

It's not a big problem. I'm new to C++ having come from a C background. With the availablility of all the classes on CodeProject (and other sites), it's easy to just drop in classes to facilitate easier programming. However, if I only use a small fraction of the functions in the different classes, then I'm thinking that my application will end up being unnecessarily bloated with code that never gets executed and data members that never get accessed.

I suppose can test it easily enough by creating a test app and then seeing if the code size is larger or the same size when I access more class member functions/variables. I was just curious if anyone knew and had a bit more info on how it works.

Re: C++ code size

David Crow5-Jan-05 7:52