Hi, thanks again for your clarifications. [...] > > > I looked at the order-by attribute of the reference > > descriptor, but I can't > > use that. There is no attribute in my element-classes that I > > can sort on; > > it's the order that I put them in the list that is relevant. > > Mhh, In an RDBMS there is no notion of "position in a table". If you > want to > have your elements to show up in a given sequence you have to use an > additional column that can be used for sorting. > This is not a restriction with OJB but with relational calculus in > general!
Yes, I understand this quite well... :-) [...] > You are right, the DListImpl provides such an indirection mechanism > (and > also maintains order of the entries in the way that you'd like to see. > > The only problem is that DListImpl & friends are coded to cooperate > with > ODMG transaction manager. > They lookup the current ODMG transaction and perform locks etc. > > This won't be of any help in a JDO managed situation (could also cause > errors). > But it would be an option to reuse most of the DlistImpl code to write a > collection class that does all your housekeeping, but does not try to > work > with ODMG transactions. As I said, I have to move away from the JDO-managed situation (until there is a sufficiently high-quality open-source JDO implementation). I'll have to go to the ODMG-managed situation. Only I don't like that I have to change my code in several places now to explicitly use DList instead of java.util.List and it's standard implementations. I want to keep my code independant of the actual persistence mechanism used, so that in future I can switch back to a JDO managed solution. It would be nice if there was a List implementation that is not ODMG-specific but provides the same functionality, so that it can be used transparently (to the largest degree possible) without modifications to the domain objects. I guess that for a JDO implementation in OJB 2.0 such proxy-objects for Set, List and Map will have to be created. But since I don't have time now to do this myself, I'll have to change what needs to be changed! Thanks for the advice and explanations - Thomas, you're a great help to everyone on the list. A big thanks. --Tim --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
