[Kst] dependencies continued....

[George Staikos]

On Wednesday 28 January 2004 00:17, C. Barth Netterfield wrote:
> Yup, and nope...
>
> But I shall ruminate further upon this problem (which has been pestering me
> for... a long time.)

   It seems to me that the logical thing is to create a directed acyclic graph 
and update by walking the graph.  The problems there involve overcomplicating 
anything that manipulates objects, since they have to break the tree each 
time they change object dependencies.  The nice thing is that we can very 
easily check for cycles when new objects are created and deny the changes to 
the user if they will create such a cycle.

  Actually I guess it will be a forest of DAGs since there could be many 
disjoint graphs in practice...  Anyhow, I guess this is fairly obvious stuff.  
My other approach, which I fooled around with here today, was to use signals 
and slots, but I think that's extremely expensive, unfortunately, and we 
could very easily enter infinite recursion.

-- 
George Staikos
KDE Developer				http://www.kde.org/
Staikos Computing Services Inc.		http://www.staikos.net/