>
> While I like the general idea of what you're proposing, I'm a bit lost as to
> where the default pk generation strategy will come from. As it stands now,
> data objects are shielded from rdbms-specific things. The various pk
> generators are actually quite rdbms-bound. This may not be a problem if
> there was a default generator for all sequence-based pks or what have you,
> as Andrus suggested. Or, are you simply suggesting the db adapter checks
> the return value of getAlternatePkGenerator() and if it's null, use its own
> default pk generator?
Yes.. CayenneDataObject would provide a default "null" implementation.
If you're providing your own PkGenerator, its probably not rdbms
specific, after all.. at least as far as Cayenne is concerned.
As far as the modeler / xml debate, the reason I disagree is that it
is a class that is not visible to the modeler. This is similar to
extended types, which are not defined in the modeler. In fact, if you
look at the modeler, there isn't a single class listed, except for the
ones generated by velocity. I think that paradigm is worth preserving,
mostly for the reasons I listed previously - but I see the good points
about having it in the modeler, so I won't argue any further than that
:)
Cris
This archive was generated by hypermail 2.0.0 : Mon Dec 12 2005 - 21:53:06 EST