Re: Correct use of pixbuf, pixmap, and image?

["muppet" <scott asofyet org>]

Ari Jolma wrote:
> muppet wrote:
>
> > Similarly, image data can be stored in "bitmaps" (1-bit images, or
> > maps of bits) or "pixmaps" (maps of more-than-one-bit pixels, or
> > pixel maps).  These are stored on the server, at device bit
> > resolution and bit depth, and this can be manipulated quite quickly.
>
>
> looking at process info, it seems that X is piling up memory, the code
> creates a new pixmap every time a new object it selected (a crude way of
> getting rid of the old selection), but pixmaps should be automatically
> unref'd since the new pixmap is put into the same place as the old, the
> code is effectively:
>
> create $self->{pixbuf}
> loop: {
>     $self->{pixmap} = $self->{pixbuf}->render_pixmap_and_mask;
>     $self->{image}->set_from_pixmap($self->{pixmap});
>
>     $self->{pixmap}->draw_line etc.
>
> }

I'm not sure why that would pile up pixmaps without seeing more of your code.

You may wish to take a different approach.  Render the pixbuf once, turn it
into a pixmap once, and then keep that as an untouched server-side resource. 
Copy the necessary portions from the source pixmap to the onscreen window on
demand, and draw on the onscreen window.  Alternatively, use a single
offscreen pixmap for the source buffer, and another for the annotate image; to
"undraw" previous annotations, copy from the clean one.  This avoids the whole
re-render stage.


> > GtkImage provides a way for you to set the source data directly from
> > a pixbuf --- $image->set_from_pixbuf($pixbuf) --- so you could avoid
> > the pixmap/mask stage.
>
> but I can't draw on an image? I need the pixmap for that, I believe.

You need the pixmap only for X's drawing primitives.  You can use gd, libart,
cairo, imagemagick, and even direct pixel manipulation to deal with the 24-bit
image data in the pixbuf.




-- 
muppet <scott at asofyet dot org>