I use annotations and inheritance for beans and id's. This works good, but I have problem with generated SQL for creating database schema. Hibernate declares fields in primary key in "reverse" order, i.e. "nearest" fields first and then fields from superclass. This is bad, because for composite id's we need "master" fields first, because this index is better, we can for example use this index to select Shift instances by laboratory.

Is it possible change order of fields in primary key? Can you change this default behaviour to more practical?

PS. Order of fields in table itself also "reversed", and also would be good to have them in better order - from superclass to sub-class.


Hibernate version: Hibernate 3.1.1, Hibernate Annotations 3.1 beta 8

Mapping documents:

@AccessType("field")
@MappedSuperclass
public class AbstractLaboratoryId extends AbstractCompositeId {
private static final long serialVersionUID = -4959976895077908661L;
long laboratory_id;
////////////////////////////////////////////////////////////////////////////
//
// Конструктор
//
////////////////////////////////////////////////////////////////////////////
public AbstractLaboratoryId(Laboratory laboratory) {
laboratory_id = laboratory.getId();
}
}


@AccessType("field")
@MappedSuperclass
public class ShiftId extends AbstractLaboratoryId {
private static final long serialVersionUID = 605527818398859417L;
long shift_id;
////////////////////////////////////////////////////////////////////////////
//
// Constructor
//
////////////////////////////////////////////////////////////////////////////
public ShiftId(Laboratory laboratory, long shift_id) {
super(laboratory);
this.shift_id = shift_id;
}
}


@Entity
@SequenceGenerator(name = "SEQ", sequenceName = "SEQ_SHIFT")
public class Shift extends AbstractLaboratoryBean {
////////////////////////////////////////////////////////////////////////////
//
// Constructor
//
////////////////////////////////////////////////////////////////////////////
public Shift(Laboratory laboratory, long shift_id) {
super(laboratory);
m_id = new ShiftId(laboratory, shift_id);
}
////////////////////////////////////////////////////////////////////////////
//
// Object
//
////////////////////////////////////////////////////////////////////////////
@Override
public String toString() {
return BeanUtils2.toString(this, "id");
}
////////////////////////////////////////////////////////////////////////////
//
// Id
//
////////////////////////////////////////////////////////////////////////////
private ShiftId m_id;
@EmbeddedId
public ShiftId getId() {
return m_id;
}
public void setId(ShiftId id) {
m_id = id;
}
}


Name and version of the database you are using: Oracle 9i

The generated SQL (show_sql=true):

create table Shift (shift_id number(19,0) not null, laboratory_id number(19,0) not null, version number(19,0) not null, primary key (shift_id, laboratory_id));

Debug level Hibernate log excerpt:

create an explicit index.
Your choice isn't a more practical choice, it's very applciation dependant. superclass ids might not be selective enough

May be selective, may be no. But if index is here, consumes space in database we should use it as much as possible. Can I configure order of fields in primary key index? I don't want create additional index - this is just space lost.

no, ordering is undefined.