Gegl-ers: I just checked in some changes to make things go more smoothly in the direction of things discussed at gimpcon. I removed many of the ops, except for roughly one of each kind we have so far. I left a couple of extra Unary ops, since they have different parameters/properties and it is good to experiment with these. I removed the ChannelSpace class and its subclasses, and introduced the Color class (a single color object). In the process I renamed the Color op (constant color image) to Fill. It takes a color object, and a data type for the underlying image data desired. Right now the color object can only be passed floats to initialize it. This will change to doubles eventually. Anyway, during an op's process routine, it can extract a color in any color model it wants, though it is stored internally as floats (eventually doubles). I haven't updated the gtk-docs completely so for the latest information look at the unit tests. The major change from the descriptions in the docs is that Color op had its name changed to Fill, and takes the new Color object during its creation. You can see the filltest unit tests for details. The other big change, as mentioned above, that ChannelSpace went away. And there are less ops overall right now while we make changes to images, data buffers etc. Anyway, work on the Images, DataBuffers, Tiles, ImageIterators, Storage and so one can now probably move ahead with a little more ease. I will update the gtk-docs on the gegl.org shortly to reflect the new changes. Calvin
