minimizing contact with the database

I'm trying to minimize the contact with my RDBMS from my application, only updating, inserting and removing entries that needs to be touched.

My problem is when I have an unlimited sized list of attributes associated with a table. Example:

MyClass:
|| PK || ...

MyClass_attribute:
|| MyClassFK || attribute ||

At commit-time, I know the value of attributes have been added and removed, but the attribute isn't unique and there might be duplicates. I can't just pass a "DELETE FROM ... WHERE MyClassFK=n AND attribute=x", since that might remove more than the one attribute I had in mind.

I want to solve this in the buisness layer, not in the persistent layer, thus I don't want to extend MyClass_attribute with unique PK:s.

One solution would be to check if there are any duplicates when I gather the attributes from the persistent layer. If none, it safe to pass above query. If there are duplicates, and any duplicate has been touched, I have to remove ALL attributes and then insert them again.

Are there any better solutions?

karl Received on Thu May 08 2003 - 07:57:30 CEST