For security reasons, our oracle db objects normally belong to a different schema than the logged in user. Eg. tables are in xx_core and the user we log in is xx_app_yy .
In my persistence.xml I define a orm file so that I can specify the schema name at deploy time eg.:
<mapping-file>xx_schema_orm.xml</mapping-file>
Then in the xx_schema_orm.xml I can define the object-owning-schema eg.:
<persistence-unit-metadata>
<persistence-unit-defaults>
<schema>xx_core</schema>
</persistence-unit-defaults>
</persistence-unit-metadata>
This works great for tables, but I can’t find the equivalent for sequences. It tries to use the sequence without the schema name and then I get an exception:
2010-10-14 03:04:05,423:DEBUG could not get next sequence value [select xx_SEQ.nextval from dual] - org.hibernate.util.JDBCExceptionReporter
java.sql.SQLException: ORA-02289: sequence does not exist
at oracle.jdbc.driver.DatabaseError.throwSqlException(DatabaseError.java:145)
I’ve tried to set the schema name as part of the sequence name for a generator in the xx_schema_orm.xml, but could not get it working eg.:
<sequence-generator name="xx_SEQ_GEN"
sequence-name="xx_core.xx_SEQ"/>
Workarounds I may try:
- make a database SYNONYM for the sequences in the user’s schema.
- stop using sequences and use some other way to generate ids.
In JPA 2.0:
@SequenceGeneratorannotation and the equivalentsequence-generatorelement do allow to specify aschema(andcatalog) name.schemasubelement should be honored by sequence generators as well.But this doesn’t apply to JPA 1.0.
I’ll just quote the sections about the
schemasubelement to illustrate the differences (other relevant sections are mentioned in the references below). From the JPA 2.0 specification:From the JPA 1.0 specification:
So, unless your provider offers some specific extensions, my suggestions are:
schemasubelement will do the trick ~or~TableGeneratorif you have to stick with JPA 1.0 ~or~References