SKIA Rendering on Windows(SkCanvas from HDC)

Question

I'm trying to use SKIA for drawing on Windows.

Is there anyway to initiate SkCanvas from HDC ?

Folberth

Answer 1

There are several approaches "to use" skia on Windows. These approaches described in the SkOSWindow_win.cpp at the skia\src\views\win in your skia folder.

For example:

1. You can attach OpenGL viewport to your window as it has been done in function SkOSWindow::attachGL

   HDC dc = GetDC((HWND)fHWND);
   if (NULL == fHGLRC) {
       fHGLRC = SkCreateWGLContext(dc, msaaSampleCount,
           kGLPreferCompatibilityProfile_SkWGLContextRequest);
       if (NULL == fHGLRC) {
           return false;
       }
       glClearStencil(0);
       glClearColor(0, 0, 0, 0);
       glStencilMask(0xffffffff);
       glClear(GL_STENCIL_BUFFER_BIT | GL_COLOR_BUFFER_BIT);
   }
   if (wglMakeCurrent(dc, (HGLRC)fHGLRC)) {
       // use DescribePixelFormat to get the stencil bit depth.
       int pixelFormat = GetPixelFormat(dc);
       PIXELFORMATDESCRIPTOR pfd;
       DescribePixelFormat(dc, pixelFormat, sizeof(pfd), &pfd);
       info->fStencilBits = pfd.cStencilBits;
   
       // Get sample count if the MSAA WGL extension is present
       SkWGLExtensions extensions;
       if (extensions.hasExtension(dc, "WGL_ARB_multisample")) {
           static const int kSampleCountAttr = SK_WGL_SAMPLES;
           extensions.getPixelFormatAttribiv(dc,
               pixelFormat,
               0,
               1,
               &kSampleCountAttr,
               &info->fSampleCount);
       }
       else {
           info->fSampleCount = 0;
       }
   
       glViewport(0, 0,
           SkScalarRoundToInt(this->width()),
           SkScalarRoundToInt(this->height()));
       return true;
   }
   

2. You can copy contents of your SkCanvas to HDC as is has been done in SkOSWindow::doPaint, by doing BitBlt operations

   HDC hdc = (HDC)ctx;
   const SkBitmap& bitmap = this->getBitmap();
   
   BITMAPINFO bmi;
   memset(&bmi, 0, sizeof(bmi));
   bmi.bmiHeader.biSize = sizeof(BITMAPINFOHEADER);
   bmi.bmiHeader.biWidth = bitmap.width();
   bmi.bmiHeader.biHeight = -bitmap.height(); // top-down image
   bmi.bmiHeader.biPlanes = 1;
   bmi.bmiHeader.biBitCount = 32;
   bmi.bmiHeader.biCompression = BI_RGB;
   bmi.bmiHeader.biSizeImage = 0;
   
   //
   // Do the SetDIBitsToDevice.
   //
   // TODO(wjmaclean):
   //       Fix this call to handle SkBitmaps that have rowBytes != width,
   //       i.e. may have padding at the end of lines. The SkASSERT below
   //       may be ignored by builds, and the only obviously safe option
   //       seems to be to copy the bitmap to a temporary (contiguous)
   //       buffer before passing to SetDIBitsToDevice().
   SkASSERT(bitmap.width() * bitmap.bytesPerPixel() == bitmap.rowBytes());
   bitmap.lockPixels();
   int ret = SetDIBitsToDevice(hdc,
       0, 0,
       bitmap.width(), bitmap.height(),
       0, 0,
       0, bitmap.height(),
       bitmap.getPixels(),
       &bmi,
       DIB_RGB_COLORS);
   (void)ret; // we're ignoring potential failures for now.
   bitmap.unlockPixels();
   

Skirl

I think a solution 2 is best way. – Renell

This depends on what you want to get. #2 produces predictable render result and more robust, but not using OpenGL and therefore significant slower. – Skirl