Does 11g support no auto_increment / auto_assigned columns?

Question

Hi,
I am using openJPA and run into following issue when using Oracle 11g:
"Oracle" does not support auto-assigne d column values. The column may not behave as desired.
In other words: The database does not support auto_increment values for rows?
Is there a workaround for that?
The openJPA community did create some warning for Oracle in their framework:
https://issues.apache.org/jira/browse/OPENJPA-1090
However ... the warning does of course not fix anything.
Is that a workaround for that? Is there a special way of creating Oracle Database so that auto_assigned / auto_incrementor is allowed in fields? I mean currently Oracle does not seem to be compatible to openJPA so we cannot use this database or recommend it using it to our users.
You can see our sample db configuration that we have been publishing for Oracle here:
http://code.google.com/p/openmeetings/wiki/Configuration_Oracle_11g
I would be very happy if we could officially release OpenMeetings with support for Oracle.
Thank you very much,
Sebastian Wagner
Lead Developer OpenMeetings - Open-Source Web-Conferencing
http://www.openmeetings.de

Answer 1

seba.wagner wrote:
Thanks for your answers,
well of course I do know that Oracle supports sequences, my question is why openJPA does not generate those by default cause the same code works with DBs of other vendors.
However I am already posting the same question to the openJPA mailing list, maybe they have some workaround for such issue. I guess it must be some kind of configuration or setup problem, not really a coding issue.
Thanks,
SebastianActually, it would be a coding issue. For instanced, MS SQL Server does have an auto increment data type. Just define the table with that datatype for a PK column, and no further consideration needed. Oracle is not SQL Server under a different lable. To get a "auto increment" PK, you have to (1) define a sequence, and (2) be sure every insert issued against that table uses/references that sequence to populate that PK column. And the best way to make 100% sure that happens is to write an ON INSERT trigger. And one also needs to be aware that Oracle makes no guarantee that a given sequence number won't be 'wasted', so the behavior is likely different from SQL Server's AUTOINCREMENT data type.
So... any third pary app written to work with both products needs to be able to take into account all of the differences, of which this is only one. There are ways to deal with this to minimize the burden of maintaining separate code bases, but I'm afraid most vendors take the worst possible approach and simply treat the dbms as a data dump and write to the least common denominator. And lest you think that just because a vendor has a big name they know what they are talking about think again. I've mentioned several times on these forums about the experience I once had with a vendor making his pitch and claiming that his company did not recommend Oracle because "it won't support more than five concurrent connections." I made sure they didn't make the short list.