Re: [OJB-developers] Question about Identity

[Georg S. <ge...@me...>]

On Mon, 13 May 2002, Thomas Mahler wrote:

> Say we have a baseclass A and a class B extending A.
> say we have an instance of B, object b with primary key pk_b.
>
> Now, if we don't have the restriction that identities are on the extend
> level we could build non-equivalent identies pointing to the same object b:
> 1. id_A = new Identity(A,pk_b)
> 2. id_B = new Identity(B,pk_b)
>
> as id_A is not equivalent with id_B OJB won't be able to detect that
> they are referring to the same instance.

OK, I get the point

> I have not thought about all deeper implications, but this could be a
> performance improvement! You can give it a try and check if all JUnit
> tests are still OK!

I added a field objectsRealClass to Identity and changed the
getDBObject(Identity id) method in PersistenceBrokerImpl, so that it will
first look if that field is set and use that concrete class.

I used the cvs-version from this morning and had the problem that the
following tests failed without having changed a line of code yet:

testTimestampLock()

testPathExpressionsMtoNDecomposed()
testPathExpressionsMtoN()

When applying my changes the same 3 failures showed up, so I suppose the
changes at least don't do anything bad for the rest. Of course I can't
talk for the 3 tests mentioned above.

By the way, Identity.java is an example of a file which seems to be in the
cvs-repository with the wrong line ending format. It contains an extra
Carriage Return for each line ending. PersistenceBrokerImpl on the other
hand doesn't show that behaviour.

Cheers

Georg

P.D.: Attached are the 2 diff files