We had the discussion here and we have the decision... what remain to do is :- resume - tests - implementation - reference - short line in breaking changes
2009/8/24 Roy R <royra...@gmail.com> > > I'm sorry, maybe I wasn't clear enough: Of course the filters should > change the results of the query. But there are some basic rules in the > MODEL that the programmer relies on, and when they change after > filters are enabled in the "infrastructure" (i.e., in the background > for the programmer), he or she are surprised every time. I've seen it > happen too much in the last few weeks :( > > 1. Enabling filters should not change the behavior of eager-loading > because it is strictly a performance-related feature; E.g, With the > filter on, query results should stay the same with and without eager- > loading. > 2. Enabling filters should not change the multiplicity defined in the > model: Non-null parents should always remain so, and possibly-null > parents depending on biz rules should always remain so. Filters on one- > to-many do not change the multiplicity. Filters on many-to-one do. > > Roy. > > P.S. Sorry for butting in to the dev mailing list - maybe we should > move the discussion to the users list. > > On Aug 24, 4:54 pm, Fabio Maulo <fabioma...@gmail.com> wrote: > > 2009/8/24 Roy R <royra...@gmail.com> > > > > > > > > > If filters start to change the behavior of > > > the query implicitly (e.g, when eager-loading a filtered entity), this > > > comes as a surprise to the person who writes the query. It feels like > > > magic - of the bad kind - and that we should avoid IMO. > > > > This is not the place for this discussion with you but... > > What you think the filter feature should do other than "change the > behavior > > of the query implicitly" ? > > -- > > Fabio Maulo > -- Fabio Maulo
