Introduction
The library promises to achieve (with minor modifications - to make it very generic) a non-Windows UI look-n-feel for those who want to develop a customized UI (curves et al) by leveraging the power of images, GDI, composition, and multiple inheritance.
Inspired by...
A few years back, when I first saw the cool skins on Winamp (MP3 player), I felt excited and challenged to write a library to use in all my future development that would shock people who believe that slick UI is only feasible in web applications and Flash applications!
What's inside the library?
The library is composed of the following classes:
CSkinControl
- Parent class of all controls, containing common functionality
CSkinnedStatic
- Custom class to act as a static control or label
- Inherits from:
Cwnd
, CSkinControl
CSkinnedButton
- Custom class to act as a button control
- Inherits from:
Cwnd
, CSkinControl
CSkinnedEdit
- Custom class to act as an edit control
- Inherits from:
Cwnd
, CSkinControl
CSkinnedComboBox
- Custom class to act as a combo box control
- Inherits from:
Cwnd
, CSkinControl
- Composed of:
CSkinnedEdit
, CSkinnedButton
, and CSkinnedListBox
CSkinnedListBox
- Custom class to act as a listbox control
- Inherits from:
Cwnd
, CSkinControl
- Composed of:
CSkinnedButton
CSkinnedScrollBar
- Custom class to act as a scroll bar control
- Inherits from:
Cwnd
, CSkinControl
- Composed of:
CSkinnedButton
CSkinnedSliderCtrl
- Custom class to act as a slider control
- Inherits from:
Cwnd
, CSkinControl
- Composed of:
CSkinnedButton
Architectural details...
The idea was to store as much common functionalities as possible in one class (CSkinControl
) and then consume it, through inheritance, in concrete control classes. The base class holds references to four different images (IDs), one for normal state, one for disabled state, one for hover state, and one for pressed state. The function that stores this is SetImageResources(normal, hover, pressed, disabled)
. The base class also contains functionality for:
- Location and size:
SetCoordinates(left, top)
SetDimensions(width, height)
GetLeft()
GetTop()
GetWidth()
GetHeight()
- Colors and fonts:
GetCurrentBackgroundColor()
GetTextColor()
GetBackgroundColor(state)
SetBackgroundColor(state, color)
SetForegroundColor(color)
SetTextColor(color)
SetFontName(name)
SetFontStyle(style)
SetFontSize(size)
GetFontName()
GetFontStyle()
GetFontSize()
The most important function is UpdateMemoryDC()
which takes care of drawing and updating the visual of each control on screen, whether in default state or triggered by some user-action (mouse events).
int CSkinControl::UpdateMemoryDC()
{
HBITMAP hBitmap = NULL;
BITMAP bmpTemp;
#ifdef USE_GIF_IMAGES
hBitmap = LoadGIF(GetDllInstance((LPCTSTR)m_csDLLFileName),
MAKEINTRESOURCE(GetID()));
#else
hBitmap = LoadBitmap(GetDllInstance((LPTSTR)(LPCTSTR)m_csDLLFileName),
MAKEINTRESOURCE(GetID()));
#endif
if(hBitmap != NULL)
{
::GetObject(hBitmap, sizeof(BITMAP), &bmpTemp);
m_lImageWidth = bmpTemp.bmWidth;
m_lImageHeight = bmpTemp.bmHeight;
::SelectObject(m_dcMemory.GetSafeHdc(),hBitmap);
}
else if(m_nPressedID == -1 && m_nUnPressedID == -1 && m_nHoverID == -1)
{
m_dcMemory.SetTextColor(m_crTextColor);
m_dcMemory.DrawText(m_csText, CRect(0, 0, m_nWidth, m_nHeight), DT_CENTER);
}
return 0;
}
Concrete classes provide functionalities that are required by their standard counterparts. For example, CSkinnedEdit
supports text selection, insertion, deletion (no copy-paste implemented - sorry!!), and other customized features like "read-only", "decimal point validation", etc. Similarly, CSkinnedScrollBar
provides functionality to set minimum range, maximum range, retrieve position of scrollbar button, and so on. The code and function names are quite self-explanatory. I apologize for not providing many inline code comments, for which you can always contact me.
All the controls are created dynamically. Each one of them has a function CreateSkinControl(name, rect, parent, id, flags)
that takes parameters as mentioned. The last one (flags
) is an interesting parameter that holds any "extra" information required (as you'll see in different controls) for creation. As an example, shown below is the creation code for CSkinnedButton
control:
BOOL CSkinnedButton::CreateSkinControl(LPCTSTR lpszWindowName, LPRECT lpRect,
CWnd *pParentWnd, UINT nControlID, long lFlags)
{
m_csText = lpszWindowName;
m_nLeft = lpRect->left;
m_nTop = lpRect->top;
m_nWidth = lpRect->right - lpRect->left;
m_nHeight = lpRect->bottom - lpRect->top;
m_pParentWnd = pParentWnd;
m_nControlID = nControlID;
m_csFontName = "Arial";
m_nFontSize = 16;
m_nFontStyle = FONT_NORMAL;
m_crBackgroundColorHover = RGB(255,255,255);
m_crBackgroundColorPressed = RGB(255,255,255);
m_crBackgroundColorUnPressed = RGB(255,255,255);
m_crForegroundColor = RGB(0,0,0);
m_lButtonType = lFlags;
if(m_hWnd != NULL)
{
return FALSE;
}
if(CWnd::Create(NULL, m_csText, WS_CHILD|WS_VISIBLE|WS_CLIPSIBLINGS,
*lpRect, pParentWnd, nControlID, NULL))
{
CWnd::BringWindowToTop();
return TRUE;
}
return FALSE;
}
Steps to implement...
- In a window/dialog that you want to use a control (button, for example), define a member variable that is a pointer to the control.
CSkinnedButton* m_pOkButton;
Inside the creation logic of the dialog OnCreate()
or OnInitDialog()
, insert the creation logic of the button, after a few initializations of creating a memory DC for background painting.
int CMyDialog::OnCreate(LPCREATESTRUCT lpCreateStruct)
{
if(CDialog::OnCreate(lpCreateStruct) == -1)
{
return -1;
}
CClientDC dc(this);
m_memDC.CreateCompatibleDC(&dc);
m_memBmp.CreateCompatibleBitmap(&dc, 1024, 768);
m_memDC.SelectObject(&m_memBmp);
...
m_pOkButton = new CSkinnedButton;
m_pOkButton.SetImageResource(ID_NORMAL, ID_HOVER, ID_PRESSED, ID_DISABLED);
m_pOkButton.SetShapedFlag(TRUE);
...
}
The custom code for button creation and button rendering is implemented in the CSkinnedButton
class as shown:
int CSkinnedButton::OnCreate(LPCREATESTRUCT lpCreateStruct)
{
if (CWnd::OnCreate(lpCreateStruct) == -1)
return -1;
CClientDC dc(this);
CBitmap bmpTemp;
m_dcMemory.CreateCompatibleDC(&dc);
if(bmpTemp.CreateCompatibleBitmap(&dc, m_nWidth, m_nHeight) != 0)
{
m_dcMemory.SelectObject(&bmpTemp);
if(PrepareFont())
{
}
UpdateMemoryDC();
if(m_bShape)
{
m_hRgn = CreateRectRgn(0,0,0,0);
if(m_hRgn != NULL)
{
if(GetWindowRgn(m_hRgn) == ERROR)
{
m_hRgn = NULL;
return -1;
}
}
else
{
return -1;
}
}
}
return 0;
}
int CSkinnedButton::UpdateMemoryDC()
{
BITMAP bmpTemp;
memset(&bmpTemp, 0, sizeof(BITMAP));
if(m_dcMemory == NULL)
{
return -1;
}
#ifdef USE_GIF_IMAGES
if(m_hBitmap != NULL && m_hBitmap == GetCurrentStateBitmap())
{
return -1;
}
m_hBitmap = GetCurrentStateBitmap();
#else
hBitmap = GetCurrentStateBitmap();
#endif
if(m_hBitmap != NULL)
{
::GetObject(m_hBitmap, sizeof(BITMAP), &bmpTemp);
m_lImageWidth = bmpTemp.bmWidth;
m_lImageHeight = bmpTemp.bmHeight;
::SelectObject(m_dcMemory.GetSafeHdc(),m_hBitmap);
}
else if(m_nPressedID == -1 && m_nUnPressedID == -1 && m_nHoverID == -1)
{
CClientDC dc(this);
m_dcMemory.SetMapMode(dc.GetMapMode());
m_dcMemory.SetWindowExt(dc.GetWindowExt());
m_dcMemory.SetViewportExt(dc.GetViewportExt());
m_dcMemory.SetWindowOrg(0, 0);
CBitmap cbmpTemp;
cbmpTemp.CreateCompatibleBitmap(&dc, m_nWidth, m_nHeight);
if(m_dcMemory.SelectObject(&cbmpTemp) != NULL)
{
m_dcMemory.FillSolidRect(0, 0, m_nWidth, m_nHeight,
GetCurrentBackgroundColor());
}
}
if(m_bShape != -1 && m_bFindEdges)
{
m_bFindEdges = FALSE;
FindControlEdge(this, &m_dcMemory, COLOR_MAGENTA, m_hRgnWindow);
}
return 0;
}
FindControlEdge()
(not a very intuitive name!) implements the transparency algorithm, using a Magenta color mask, traversing through the image, and cutting out a region. You might argue that why not use the GDI function TransparentBlt()
to achieve the same. Good point! However, when I tried to implement using TransparentBlt
, it failed to run in Windows 98 SE (although MS claims to have support in that version of Windows!). Anyways, may be I didn't have the correct patch of Windows or SDK at the time. I decided to write my own. You have a choice of using TransparentBlt
which would promise an optimized performance over my technique for sure ;)
Also, my technique introduces a strict requirement of having all images bound by a four pixel magenta background!!! Example:
Those who might be facing a similar problem of TransparentBlt()
are free to use the algorithm shown here or one of your own.
BOOL FindControlEdge(CWnd* pWnd, CDC *dcControl,
COLORREF colToSkip, HRGN &hRgn)
{
int nCurrentX = 0;
int nCurrentY = 0;
int nTempX = 0;
int nTempY = 0;
BOOL bStop = FALSE;
int nDirection = 0;
int nCurDirection = 0;
int nFirstX = 0;
int nFirstY = 0;
int nXMap = 0;
int nYMap = 0;
int nIterate = 0;
POINT ptTempCoord;
CList<point,> ptCoord;
CRect rcWindow(0,0,0,0);
CRect rcClient(0,0,0,0);
pWnd->GetWindowRect(&rcWindow);
pWnd->GetClientRect(&rcClient);
pWnd->ClientToScreen(&rcClient);
nXMap = rcClient.left - rcWindow.left;
nYMap = rcClient.top - rcWindow.top;
nIterate = 0;
bStop = FALSE;
nCurrentX = 0;
nCurrentY = 0;
nDirection = SOUTHEAST;
nFirstX = 0;
nFirstY = 0;
while(!bStop)
{
if((dcControl->GetPixel(nCurrentX+1, nCurrentY+1)) != colToSkip)
{
bStop = TRUE;
if(nCurrentX == 0 && nCurrentY == 0)
{
return FALSE;
}
}
else
{
nCurrentX++;
nCurrentY++;
}
}
bStop = FALSE;
while(!bStop)
{
nIterate++;
switch(nDirection)
{
case SOUTHEAST:
if((dcControl->GetPixel(nCurrentX+1, nCurrentY+1)) != colToSkip)
{
nDirection = EAST;
continue;
}
else
{
nCurrentX++;
nCurrentY++;
}
break;
case EAST:
if((dcControl->GetPixel(nCurrentX+1, nCurrentY)) != colToSkip)
{
nDirection = NORTHEAST;
continue;
}
else
{
nCurrentX++;
}
break;
case NORTHEAST:
if((dcControl->GetPixel(nCurrentX+1, nCurrentY-1)) != colToSkip)
{
nDirection = NORTH;
continue;
}
else
{
nCurrentX++;
nCurrentY--;
}
break;
case NORTH:
if((dcControl->GetPixel(nCurrentX, nCurrentY-1)) != colToSkip)
{
nDirection = NORTHWEST;
continue;
}
else
{
nCurrentY--;
}
break;
case NORTHWEST:
if((dcControl->GetPixel(nCurrentX-1, nCurrentY-1)) != colToSkip)
{
nDirection = WEST;
continue;
}
else
{
nCurrentX--;
nCurrentY--;
}
break;
case WEST:
if((dcControl->GetPixel(nCurrentX-1, nCurrentY)) != colToSkip)
{
nDirection = SOUTHWEST;
continue;
}
else
{
nCurrentX--;
}
break;
case SOUTHWEST:
if((dcControl->GetPixel(nCurrentX-1, nCurrentY+1)) != colToSkip)
{
nDirection = SOUTH;
continue;
}
else
{
nCurrentX--;
nCurrentY++;
}
break;
case SOUTH:
if((dcControl->GetPixel(nCurrentX, nCurrentY+1)) != colToSkip)
{
nDirection = SOUTHEAST;
continue;
}
else
{
nCurrentY++;
}
break;
}
nCurDirection = nDirection;
if((dcControl->GetPixel(nCurrentX+1, nCurrentY+1)) != colToSkip)
{
nDirection = SOUTHEAST;
}
if((dcControl->GetPixel(nCurrentX+1, nCurrentY)) != colToSkip)
{
nDirection = EAST;
}
if((dcControl->GetPixel(nCurrentX+1, nCurrentY-1)) != colToSkip)
{
nDirection = NORTHEAST;
}
if((dcControl->GetPixel(nCurrentX, nCurrentY-1)) != colToSkip)
{
nDirection = NORTH;
}
if((dcControl->GetPixel(nCurrentX-1, nCurrentY-1)) != colToSkip)
{
nDirection = NORTHWEST;
}
if((dcControl->GetPixel(nCurrentX-1, nCurrentY)) != colToSkip)
{
nDirection = WEST;
}
if((dcControl->GetPixel(nCurrentX-1, nCurrentY+1)) != colToSkip)
{
nDirection = SOUTHWEST;
}
if((dcControl->GetPixel(nCurrentX, nCurrentY+1)) != colToSkip)
{
nDirection = SOUTH;
}
POINT ptTemp;
if(ptCoord.GetCount() > 0)
{
ptTemp = ptCoord.GetTail();
}
else
{
ptTemp.x = 0;
ptTemp.y = 0;
}
if(nCurrentX != ptTemp.x || nCurrentY != ptTemp.y)
{
nTempX = nCurrentX;
nTempY = nCurrentY;
switch (nCurDirection)
{
case NORTH:
case NORTHWEST:
nTempX++;
break;
case NORTHEAST:
case EAST:
nTempY++;
break;
}
ptTempCoord.x = nTempX;
ptTempCoord.y = nTempY;
ptCoord.AddTail(ptTempCoord);
}
if(nFirstX == 0 && nFirstY == 0)
{
nFirstX = nCurrentX;
nFirstY = nCurrentY;
}
else if(nCurrentX == nFirstX && nCurrentY == nFirstY)
{
break;
}
}
POINT *ptAll;
ptAll = new POINT[ptCoord.GetCount()];
int nLen = ptCoord.GetCount();
for(int idx=0; idx<nLen; idx++)
{
ptAll[idx] = ptCoord.GetHead();
ptCoord.RemoveHead();
}
hRgn = CreatePolygonRgn(ptAll, nLen, ALTERNATE);
delete []ptAll;
if(hRgn != NULL)
{
if(pWnd->SetWindowRgn(hRgn, TRUE) != 0)
{
return TRUE;
}
}
return FALSE;
}
Finally, you implement message handlers to react to control events and messages (LButtonDown/Up
for Button, OnChar
for Edit, and so on), and appropriately play with different control states (normal, hover, disabled, etc.) and updating of corresponding "look" by calling UpdateMemoryDC()
.
Some benefits...
If you design properly, you can come up with parallel "themes" for your application; basically, different sets of images to super-impose on your application and controls within, and switch easily, using configuration files.