🚀 go-pugleaf

When I paint my client area in response to WM_PAINT, I scale my 'world'
co-ordinates (~1e6 metres) to logical units of 0.1 mm and then paint with
MM_LOMETRIC.  This is fine.  Up until now, I have ignored the clipping
rectangle in PAINTSTRUCT, just to get it working (slow).  Now I want to use the
rect to redraw only what is required.

Now I'm confused about how to use this clipping rectangle because it is
provided for the default hdc of MM_TEXT and I work in MM_LOMETRIC.
How do I convert those MM_TEXT co-ordinates to MM_LOMETRIC co-ordinates?
I thought of using screen co-ordinates as an intermediate, i.e.:

case WM_PAINT:
	hdc=BeginPaint(hwnd,&ps);
	LPtoDP(hdc,(LPPOINT)&ps.rcPaint,2);	//to pixels

	//now set up my co-ordinate system
	SetMapMode(hdc,MM_LOMETRIC);
	SetWindowExt...
	SetViewport...
	SetViewportOrg...

	//convert pixels back to 0.1 mm logical units
	DPtoLP(hdc,(LPPOINT)&ps.rcPaint,2);

	//convert again to get world co-ordinates
	...
	
	//paint my picture using world co-ordinates to clip
	...

	//finish up
	EndPaint(hwnd,&ps);

Is this the correct way to do this?  I was hoping there would be a more
transparent way.

Preferences...
  It appears that there are several cases where a redraw is required:
	(a) WM_SIZE/WM_PAINT
		Since the window has changed size, the only way to
		do this is to read the data all over again and redraw the
		whole client area from scratch.  True?
	
	(b) WM_MOVE/WM_PAINT (possibly not as common)
	(c) WM_PAINT after an occluding window is removed
		If I BitBlt a copy of the picture to memory after each
		WM_SIZE/WM_PAINT or initial WM_PAINT, then I could repaint
		a part of it by just BitBlting from memory to the client area?

Is it better to redraw the rectangle by reading the data and drawing it again
or attempt to BitBlt the picture to memory the first time, so that subsequent
times it can be redrawn with more BitBlts.  Is the BitBlt seamless?

Sorry for the rambling - I can't seem to keep it short for this one.