I love writing unit tests against an in-memory database, so I usually pull out HSQLDB for that. But I get the impression that the support for HSQLDB in Eclipselink is very minimal. The GeneratedValue is converted to a identify column, but the generated DDL code is incorrect (syntax error) when letting EclipseLink generate the database.
Time for me to get the 1.1 source distribution and work on the HSQLDB implementation?
Hmmm, HSQLDB only supports IDENTITY on integers (Eclipselink mirror that from a Java Integer class) or bigints (no idea what Java class Eclipselink will mirror to that). So sequences would be very welcome.
HSQLDB only allows IDENTITY on columns of type Integer or BigInt. EclipseLink generates a column of type Decimal-something. I can get you the exact error tomorrow.
> For database independent sequencing you can use TABLE sequencing. It is
> also more efficient than IDENTITY sequencing.
Is that so? Hm. It sure it more portable, but I intended to enable sequences in the HSQLDB database platform. Table or sequence? Hm, table is more portable and that is the goal here, so I think I'll be using table.
> You might also consider Derby/JavaDB as a test database.
Those do not work in memory; this is unittesting, so I need a new clean database everytime. In memory is very practical.