With the following entities:

@Entity
@table(name="a")
public class A {
private int id;
private List<B> bs;

@Id
public int getId() {return id;}
public void setId(int pId){id = pId;}

@OneToMany(targetEntity=B.class, mappedBy="a")
public List<B> getBs() {return bs;}
public void setBs(List<B> pBs){bs = pBs;}
}

and


@Entity
@table(name="b")
public class B {
private int id;
private A a;

@Id
public int getId() {return id;}
public void setId(int pId){id = pId;}

@ManyToOne(targetEntity=A.class)
@JoinColumn(name="a")
public A getA() {return a;}
public void setA(A pA){a = pA;}
}

when the tables are:

a

id
------
1

b

id | a
---|---
1 | 1
2 | 1


one might expect that the query created from

SELECT DISTINCT a
FROM A a
LEFT JOIN FETCH a.bs

will return a list containing one entity.
Instead, the query returns a list containing twice the same instance of the entity, since it creates one entity for every row (although it doesn't recreates entities, so it is the same instance twice).
It seems like a bug.
It is possible to fix it in Loader (while populating the list), what should be done?

Hibernate version:3.1.rc2

Name and version of the database you are using:oracle9i

The generated SQL (show_sql=true):
select distinct a0_.id as id2_0_,
bs1_.id as id8_1_,
bs1_.a as a8_1_,
bs1_.a as a0__,
bs1_.id as id0__,
from a a0_ left outer join b bs1_ on a0_.id = bs1_.a

This is not a bug regarding the way SQL and hence EJBQL works.
But Steve has changed this behavior in hibernate core SVN trunk, you can try and use it.

But EJBQL is a typed language, so how can It return the same instance twice, If I use "distinct"?
In the ejb-persistence spec (final draft) it is written: "The DISTINCT keyword is used to specify that duplicate values must be eliminated from the query result", so this behaviour does not match the (temporary) spec!!!

The spec has benn changed in this area :-)

Do you mean the ditinct does not mean what I said it means? The quote is from the "public final draft" the last version of the spec. Do you say it will be changed in the next version, or that it was changed and you haven't managed to match the change yet?

There are on going discussions on whether allowing distinct when a join is used.