[Gegl-developer] DAGs make users' eyes cross
Øyvind Kolås
pippin at gimp.org
Wed Oct 18 03:31:04 PDT 2006
On 10/17/06, Ken Bateman <kbateman at cymbym.com> wrote:
> I've independently been thinking of a similar imaging core for the last month or
> so. Of course, I think it's a good idea. Logically, behind the scenes,
> everything should be in the form of a DAG. But I think that a raw DAG does not
> make up a part of a good user interface.
The current user interface used in the prototype GUI of GEGL is an
abstraction above the DAG that is a tree with clones. This allows
expressing all DAGs that have sources (one output) filters (one input
and one output), composers (two inputs and one output). This is the
interface model I used in bauxite[1] , with the added ability to make
some of the nodes in this tree be embedded graphs I think you have the
best of both worlds. It shold also be noted that one of the operations
in GEGL at the moment is the "layer", which is a composition of over
(or another compositing operator), translate and opacity.
> Spreadsheets are in essence a DAG for performing numerical calculations, and I
> would suggest that the user interface should take the form of a spreadsheet
> since this is a model familiar to many users. Some cells would contain source
> rasters, other cells would contain paths, vector graphics, or masks (or even
> references to other functions), some cells would contain a function that
> referred to other cells for inputs, some cells could contain labels or comments,
> and most cells would be empty.
Such a user inteface does not lend itself well to a user that wants to
do image manipulation in the manner traditionally done by users of
GIMP, an application that does something along these lines is nip2[2]
the interface of the vips projects.
1: http://pippin.gimp.org/bauxite/
2: http://www.vips.ecs.soton.ac.uk/index.php?title=Nip2
/OEyvind K.
--
«The future is already here. It's just not very evenly distributed»
-- William Gibson
http://pippin.gimp.org/ http://ffii.org/
More information about the Gegl-developer
mailing list