[jdemmon]

The Hibernate mapping tool generated this equals() method for AbstractFoo.java:

 public boolean equals(Object rhs)
    {
        if (rhs == null)
            return false;
        if (! (rhs instanceof Foo))
            return false;
        Foo that = (Foo) rhs;
        if (this.getRowId() != null && that.getRowId() != null)
        {
            if (! this.getRowId().equals(that.getRowId()))
            {
                return false;
            }
        }
        return true;
    }

The primary key in this case is a String, which is null until Hibernate assigns a value, which may not occur until the session is flushed.

This method returns true between any two new unsaved instances of Foo. It also returns true if (this.getRowId() == null && that.getRowId() != null). The way equals() is written, I’ll never be able to add more than one unsaved instance of Foo to a Set. These observations also hold true for when the right hand side is a subclass of Foo.

Also, take a look at the hash code method:

public int hashCode()
    {
        if (this.hashValue == 0)
        {
            int result = 17;
            int rowIdValue = this.getRowId() == null ? 0 : this.getRowId().hashCode();
            result = result * 37 + rowIdValue;
            this.hashValue = result;
        }
        return this.hashValue;
   }

So for all new (unsaved) instances of Foo the hashcode is the same? Better hope you don’t want to put more than a few of these in a Set or a Map!

[Riyad Kalla]

Thank you for bringing this to our attention we will have a look right away.

[support-jeff]

Excellent points. A lot of this has been hash-out on the hibernate wiki and in the forums for hibernate:

http://hibernate.org/109.html

Point being that there is not much we can do but rely on the identity.

As for the equals() code, definitely a bug. This is the pattern generated for objects with composite keys. I am surprised that no one had noticed it before. Fortunately for you (and other users) you can modify the Velocity template in the com.genuitec.eclipse.hibernate plugin (templates/mappedobject.vm) to do the right thing when generating the code. We will get a fix in as soon as possilble.

[support-jeff]

Doing some more follow up:

(1) Can you identify a case where a subclass POJO and its base class POJO would have identical PK id values (other than the case of unsaved)? I cannot think of a case where this would occur.

(2) The equals() bug is present for any mapped POJO with a non-primitive identifier (and the case you note above is for String, not for composite id classes; the latter are also broken in a similar way). We will be fixing this in the upcoming 4.0 release (probably won’t make the first milestone release; definitely the second tho).

[airframes]

This message has not been recovered.

[Brian Fernandes]

This message has not been recovered.

[Author]

Posts