> That said, the warning was pointing to the Entry entity, but I think the bad > mapping was in the User entity. Maybe it's a matter of perspective? Or is > that a bug?
From the top of my head, Cayenne detects a cycle in relationship joins, but it can't tell which relationship in the pair a user might prefer to change (e.g. a relationship may only be relevant in subclass). So it takes a random guess. We should probably improve the error message to mention both relationships are candidates for refactoring. Andrus On Mar 27, 2012, at 12:48 AM, Robert Zeigler wrote: > After examining my project some more, I found the offending relationship. > > I have (generically): > Entry (to one)-> User > EntryType1 -> user (inherited from Entry) > EntryType2 -> user (inherited from Entry) > > And: > User (to many) => EntryType1 > > The offending relationship was the User => EntryType1. Once I changed that to > User => Entry, the project validated fine. > This makes sense: having User -> Entry and Entry -> User makes more sense > than User -> EntryType1 and Entry -> User. The "bad" mapping is an artifact > of having refactored the class hierarchy awhile ago. > > That said, the warning was pointing to the Entry entity, but I think the bad > mapping was in the User entity. Maybe it's a matter of perspective? Or is > that a bug? > > Robert > > On Mar 26, 2012, at 3/264:41 PM , Robert Zeigler wrote: > >> I'm gradually migrating my projects to 3.1M3 from the 3.0 series. Today, I >> converted a project in the modeler and it now complains: >> >> "Usage of super entity's relationships 'Entry.user' as reversed >> relationships for sub entity is discouraged" >> >> I'm using single-table inheritance in this case, and "Entry" is the >> superclass, with two sub-classes. Every subclass needs a reference to the >> "user" table, so I put the relationship in the superclass rather than the >> subclasses. So my questions: >> >> 1) Why is this discouraged? >> 1b) What are the side-effects of maintaining the current mapping? >> 2) I suppose the "correct" thing is to remap this as one relationship to the >> user table for each subclass? >> 3) What change in behavior from 3.0.2 -> 3.1M3 warrants this new >> warning/preference? Or has it always been bad practice but now we're letting >> people know? >> >> Robert > >
