Yeah, the whole PK generation abstraction is somewhat leaky. It works, but is challenging to apply consistently across different DB engines and different generator strategies. And as a result is challenging to present to the end user in a sane fashion. Just recently had this discussion with Nikita (and we don't have a solution, just know that there's a problem).
Andrus > On Jun 15, 2021, at 12:08 PM, Tony Giaccone <t...@giaccone.org> wrote: > > When using the modeler, you can choose to pick Custom Sequence as a choice > for the primary key generation. Provide the sequence name, and then > there's this field in the modeler, "Cached PK Size". It wasn't obvious to > me, and it took a bit of sleuthing to find that this value should be the > increment size of your sequence. > > I suggest that there's text to that effect on that field. either after the > field or in a ? on the field. or just something that identifies that this > should match with what's in your sequence. The query I'm running takes a > long d*** time, and fiddling with that value to try and understand how it > works was kind of painful. A simple in modeler explanation would save a lot > of developer time. > > > > Tony
