At 04:32 PM 8/21/00 +0400, Jephte CLAIN wrote:

"Phillip J. Eby" wrote:

...

I don't see a need for a mass invalidation operation, just more documentation on these inner workings. :)

or the lack of an attribute depencies mechanism :-) if attribute x depends on attribute y from another generic attribute provider, invalidation of y doesn't make y to be recomputed.

You can do that with a custom attribute provider also. At this point, however, I have not run into a situation where I or my developers have needed that. In general, we have taken the approach that "dependencies (such as triggers, indexing, etc.) are only validated/executed at (sub)transaction boundaries." In other words, an approach in keeping with Zope's optimistic transaction approach. If there is enough demand, I might implement a SkinScript construct to specify code to be executed immediately after an attribute changes, or to (semi-)automatically invalidate dependent attributes. IMHO, however, I think this sort of situation is one to be avoided if at all possible, because dependencies have a way of becoming spaghetti (think complex makefiles).