Re: [Gzz] Re: ugly patch in CachedNodeFunction

[Tuomas Lukka]

On Wed, May 14, 2003 at 08:05:28AM +0300, Matti Katila wrote:
> On Wed, 14 May 2003, Tuomas Lukka wrote:
> > On Wed, May 14, 2003 at 01:05:26AM +0300, Matti Katila wrote:
> > > 
> > > How should cached nodes to be made dirty?
> > > ObsTriggering or something?
> > 
> > Yes. Could you write a test that shows where you had problems?
> 
> It was just text node view which I wanted to change color after user 
> events. So nothing happened after clicking and I though it was cached what 
> was actually the right thing.
> 
> You might need to help me with obstriggering if it's the right answer.

(We discussed this in Jkl - just summarizing for others:)

> > Are you sure you're obeying the rules in the CachedNodeFunction
> > javadoc?
> 
> Well, "Second, no function whose return value depends on anything except 
> the passed parameters (including the Graph) may be called." 
> 
> - do we need to extend cachednodefunctiono to flushablecachednodefunction 
>   where this is allowed?

No, since it's not really possible to do that in a well-defined way.
This is because the value of the function for node A might well depend
on node B which is connected to it in the structure!

> - how did you think to use this strict functionality in ui? 

By using caching node functions when there are things that
can be cached. Now, both the placeables with and without the red stuff
can be cached, but not the selection.

> > You might also have told, in this mail, which function is not working
> > and how...
> 
> Yes, the message should be something like: "Caching is fun and it gives a 
> speed but what should be done when we can't use caching?"

We use caching for that part of the operation we can.

        Tuomas