[cells-devel] proposed refinement of lazy cells

[Kenny Tilton]

I have been working on Cello and seem to have run into a problem arising 
from the Cells sub-system, specifically the "awakening" process which 
forces all non-lazy cells to be evaluated.

I have a slot which cannot be evaluated just any old time, it needs the 
right active OpenGL context (the containing window's). I /could/ have an 
"active" cell on the window itself and turn it on during GLUT callbacks, 
but that would be inefficient and a silly way to achieve something more 
elegantly (I think) conceived of as laziness: do not calculate this 
until asked. (The specific is the building of the OpenGL display list 
for a widget.)

Unfortunately, the lazy option as currently set up will not work, 
because the slot never gets sampled if no containg widget is forced to 
rebuild its own display list. (the neat thing is that simply rebuilding 
the named display list causes the Right Thing to happen when the 
containing display list next gets kicked off.)

Currenty laziness has two manifestations: cells are not awakened 
artificially, and change propagation stops at them. display-lists need 
to avoid the artificial awakening but not change propagation.

Since I can imagine another cell requiring awakening but not change 
propagation, I am considering making the lazy boolean into a code, with 
legal values:

   :until-asked           do not awaken artifically, but once asked stop 
being lazy
   :after-awakening  awaken artificially, but do not propagate change 
automatically
   :always, or t          current semantics, aka both

I guess i will just go ahead since I am not breaking anything, but 
comments/questions are welcome.

kt

-- 
Cells? Cello? Celtik?: http://www.common-lisp.net/project/cells/
Why Lisp? http://alu.cliki.net/RtL%20Highlight%20Film