[cairo] Scrolling/Copying

[Bill Spitzak]

Carl Worth wrote:

> So, let's say you've got a cairo_surface_t *surface and you want to
> copy a rectangle, (x1,y1,width,height) to a position (x2,y2) on a
> cairo_t *cr.
> 
> I would write something like the following code to do that:
> 
> 	cairo_set_source_surface (cr, x2 - x1, y2 - y1);
> 	cairo_rectangle (cr, x2, y2, width, height);
> 	cairo_fill (cr);

I just realized that if this is going to work while the CTM is not the 
identity, it implies that somehow the CTM is cancelled out when the 
source and destination surface are the same, or that all surfaces 
contain a CTM. I don't think either of these are wanted for any uses 
other than this copy-area.

So it may be that the best solution is to add another operator. It 
should take a dx,dy and move all pixels in the clip region by dx,dy (a 
vector transformed by the CTM). Pixels that are not in the clip region, 
or where the source is outside the clip, would be unaffected. (I prefer 
this over having the destination be the clip as this way the clip is 
already set up for drawing the missing piece). I suspect dx,dy should be 
rounded to the nearest integer after being transformed, since 
misalignment is better than filtering the pixels, and obviously easier 
to implement.

This would allow you to continue the current behavior where the results 
are undefined if the source and destination are the same area of the 
same surface, which I assumme makes the code much easier to write and 
faster.

Another solution is to just say that copying areas is not supported by 
Cairo. On modern systems with double-buffering and modern GUI's where 
displays are hardly ever entirely static, it may not be worth it to try 
to figure out what can be copied. Existing API's that support copying 
usually have an interface to indicate what part was damaged (by 
overlapping windows) so it would be easy to have these use Cairo by 
doing nothing and returning the fact that the entire area was damaged.