Jeff Watkins wrote:
Any chance of creating a table without the ID column? It just doesn't make sense in this case: I have a multicolumn primary key for this table and it seems silly to have a separate ID column which just won't ever get used.
No, not really. Anything that is a SQLObject class has to have an immutable primary key to serve as the object's identity. That might be okay, except there's no support for compound keys.
It's possible (maybe preferable) that you could implement what you want without creating classes. RelatedJoin is essentially the same thing; I think you want to add another column to the join. The join code in 0.7 is messy, though not particularly horrible. But the svn trunk join code is (IMHO) much nicer, and the joins are self-encapsulated (the SQLObject class doesn't know about the specific joins) so it should be reasonable to add new kinds of joins.
The svn trunk stuff uses events and descriptors, which you probably won't be familiar with (at least not the events). But I think they are really neat and will help make SQLObject much more extensible, and I'd definitely like to see other people try to make use of them and see how that goes.
-- Ian Bicking / [EMAIL PROTECTED] / http://blog.ianbicking.org
