Dan Espen <dane_at_bae.bellcore.com> writes:
> However, if the window manager attempted to draw the icon, I think
> it would fail miserably. When it went to reduce a spot in the window
> containing a character to a pixel, the color in the background would
> overwhelm the color of the character. So when it tried to draw a
> picture of the XEmacs window, with all its glorious color, all you
> would see is the background, menubar, toolbar and status line.
I do not think, it would look _that_ bad. Of course, an application
knows a lot more about its data and is thus much better suited to
calculate the icon, but even without this knowledge it is possible to
come up with a generic algorithm that looks quite passable. Do a
couple of experiments with "xv" and you will see what I mean.
For instance, open a window of your prefered application and grab its
contents. Then calculate the size that this window will have in the
pager and shrink the grabbed image to this size. If you tell "xv" to
use dithering and to internally compute with true-color values, then
the resulting icon can be told apart from other icons. This will not
help in telling exactly what is displayed in the original window, but
it should give you a basic impression.
Also, an algorithm that wants to do this, could probably enhance the
image quality by using first and second derivatives of the color
gradients in order to detect edges. It would then have to use this
information in order to emphasize structures.
I see a couple of other problems, though:
- this algorithm takes a lot of time. In order to not eat up all of
the cpu resources, the process would have to artificially slow down
and return cpu performance to the system. This means that updates
do not happen very often.
- in order to come up with good results it is essential to use lots
of colors. While this is no problem in a true color environment, it
could eat up valuable color entries in an 8bit environment. I
imagine, that there would have to be some rather tricky code to use
existing color entries. Maybe, the icons could be limited to
grayscale entries. It might be able to get away with just a few
discreet values.
- I am uncertain, whether it would be possible to reflect changes in
partially or totally obscured windows. Somebody else, who is more
experienced with the X protocol, will have to provide this
information.
Markus
P.S.: Sorry, I do not have the time to try implementing this
algorithm. Even money wont convince me :-)
--
Visit the official FVWM web page at <URL:http://www.hpc.uh.edu/fvwm/>.
To unsubscribe from the list, send "unsubscribe fvwm" in the body of a
message to majordomo_at_hpc.uh.edu.
To report problems, send mail to fvwm-owner_at_hpc.uh.edu.
Received on Fri Oct 11 1996 - 21:05:13 BST