Hi there, I've just bumped into a new problem. There's a table which, many years ago, was created as an invisible M:N intermediate table. Later, we needed to add some information to the relationship, so now we have a table, say, Connection, which has - a number of normal attributes - a compound PK (department_id, user_id) which contains two FKs into two other tables, say, User and Market (the remaining of the original M:N intermediate) - two :1 relationships to those two tables (user and market).
Both User and Market tables model :N relationships connections (owning, PK-propagating), which long long ago replaced the original flattened M:N ones. Worked like a charm for years. Now though, I've got a new requirement: I need to be able to change the user of a given Connection. I've found that aConnection.addObjectToBothSidesOfRelationshipWithKey(newUser,'user') seems to work sort of properly — looks like all the relationships are properly updated and the key in the Connection table is changed in the database all right. The catch is, sometimes (by far not always), a short time after the change, I start getting No Connection found with globalID: <Connection: [department_id: X, user_id: Y] > with the original pre-change values of X and Y. I can't be quite sure, but I think probably there's sometimes a :N User.connections snapshot which contains the globalID of the original object. Since the user relationship change of its target actually changes the very PK of the object, the EOF synchronisation does not match the updated object (with a different PK => different globalID) with the original one and does not update the snapshot. Then, someone touches the relationship, gets the snapshot, EOF creates a fault with the original values, and when the fault fires, oops, there's nothing like that in the database. Does anybody see how to fix the problem? In principle I guess I could go programmatically through all the :N snapshots and try to find the old globalIDs and replace them by the new ones; but it would be sorta non-trivial and definitely dangerous... Thanks, OC
_______________________________________________ Do not post admin requests to the list. They will be ignored. Webobjects-dev mailing list (Webobjects-dev@lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/webobjects-dev/archive%40mail-archive.com This email sent to arch...@mail-archive.com