I guess the relationship is considered for the key part present in select.
e.g. if you look at the current CompoundKeyRelationshipTests in java rdb
das,
If the relationship is considered only for the keypart present in select,
the populated graph will possible not correctly represent the data on the
database.
For example:
Table A:
Metadata = id, b_id1, b_id2 (composed foreign key)
Row1 = 1, 1, 1
Row2 = 2, 2, 1
Row3 = 3, 1, 2
Row4 = 4, 2, 2
Table B:
Metadata = id1, id2 (composed primary key)
Row1 = 1, 1
Row2 = 2, 1
Row3 = 1, 2
Row4 = 2, 2
To clarify the references between tables mapped to ERD(Entity Relationship
Diagram):
B(Row1) -> A(Row1)
B(Row2) -> A(Row2)
B (Row3) -> A (Row3)
B(Row4) -> A(Row4)
Relationship definition:
<Relationship name="*A->B*" primaryKeyTable="B" foreignKeyTable="A"many
="*true*">
<KeyPair primaryKeyColumn="id1" foreignKeyColumn ="*b_id1*" />
<KeyPair primaryKeyColumn ="id2" foreignKeyColumn="*b_id2*" />
</Relationship>
The query "SELECT A.id, A.b_id1 from A, B;"
Supposing that the relationship is created even if not all the keypairs
are present on the result, the sdo graph references created, if I'm not
missing anything, as described bellow:
B(1, 1) -> A(1, 1)
B(3, 1) -> A(1, 1)
B(2, 1) -> A(2, 2)
B(2, 2) -> A(2, 2)
B(1, 1) -> A(3, 1)
B(3, 1) -> A(3, 1)
B(2, 1) -> A(4, 2)
B(2, 2) -> A(4, 2)
Concluding, the data related on sdo graph are not correctly described,
cause it is different from how they are related on the database. Though, if
the sdo user does not know how the data are on the database, the user would
get a wrong idea of how data are linked. On the example above the sdo user
may think that a row of table A is referenced by two different rows of table
B, but on the database it does not exist.
Adriano Crestani
On 4/27/07, Amita Vadhavkar <[EMAIL PROTECTED]> wrote:
>
> I am not sure if I am interpreting the question correct, but please see
> below findings.
>
> I guess the relationship is considered for the key part present in
> select.
> e.g. if you look at the current CompoundKeyRelationshipTests in java rdb
> das,
> Data:
> -----
> protected static Object[][] orderDetailsData = {{ Integer.valueOf
> (1),
> Integer.valueOf(1), new Float(1.1)},
> {Integer.valueOf(1), Integer.valueOf(2), new Float(1.2)},
> {Integer.valueOf(2), Integer.valueOf(1), new Float(2.1)},
> {Integer.valueOf (2), Integer.valueOf(2), new Float(2.2)}};
>
>
> protected static Object[][] orderDetailsDescriptionData =
> { {Integer.valueOf(1), Integer.valueOf(1), Integer.valueOf(1),
> "Descr
> 1,1,1"},
> { Integer.valueOf(2), Integer.valueOf(1), Integer.valueOf(1),
> "Descr
> 2,1,1"},
> {Integer.valueOf(3), Integer.valueOf(1), Integer.valueOf(2),
> "Descr
> 3,1,2"},
> {Integer.valueOf(4), Integer.valueOf(1), Integer.valueOf(2),
> "Descr
> 4,1,2"},
> {Integer.valueOf(5), Integer.valueOf(2), Integer.valueOf(1),
> "Descr
> 6,2,1"},
> {Integer.valueOf(6), Integer.valueOf(2), Integer.valueOf(2),
> "Descr
> 6,2,2"}};
>
> SQL:
> -----
> the below statement's graph does consider relationship
> String statement = "SELECT ORDERDETAILS.ORDERID,
> ORDERDETAILSDESC.ORDERID
> FROM ORDERDETAILS LEFT JOIN ORDERDETAILSDESC "
> + "ON ORDERDETAILS.ORDERID = ORDERDETAILSDESC.ORDERIDAND "
> +
> " ORDERDETAILS.PRODUCTID = ORDERDETAILSDESC.PRODUCTID" +
> " WHERE ORDERDETAILS.ORDERID = 1 AND
> ORDERDETAILS.PRODUCTID = 1";
>
> where the compound key is {ORDERID, PRODUCTID}.
>
> Similar is case when we have
> SELECT ORDERDETAILS.ORDERID, ORDERDETAILSDESC.PRODUCTID....
>
> OR
> basically at lease any one column from ORDERDETAILS.ORDERID/PRODUCTID
> and at
> least any one
> column from ORDERDETAILSDESC.ORDERID/PRODUCTID - makes DAS to look at
> relationship.
>
> But I see there one more confusion:-
> even if the relationship is getting considered, in above cases or even
> case
> like below which covers
> the "complete" compound key:-
> "SELECT ORDERDETAILS.ORDERID, ORDERDETAILS.PRODUCTID,
> ORDERDETAILSDESC.ORDERID,ORDERDETAILSDESC.PRODUCTID"+
> " FROM ORDERDETAILS LEFT JOIN ORDERDETAILSDESC "
> + "ON ORDERDETAILS.ORDERID = ORDERDETAILSDESC.ORDERID AND " +
> " ORDERDETAILS.PRODUCTID = ORDERDETAILSDESC.PRODUCTID " +
> " WHERE ORDERDETAILS.ORDERID = 1 AND ORDERDETAILS.PRODUCTID = 1";
>
> As, TableData.columnData is a java Map, the returned count of
> orderdetailsdesc for the orderdetail(1,1) is
> shown "1" in the data graph, where as in the database table rows it is
> "2".
>
> This is merely happening as Map does not allow duplicates. Is this a
> code
> bug or there is
> some more consideration behind using Map?
> Regards,
> Amita
>
>
> On 4/27/07, Adriano Crestani < [EMAIL PROTECTED]> wrote:
> >
> > I wonder what happens if a relationship has composite keys and the
> query
> > does not returns all the relationship keys. Is the relationship
> ignored
> > and
> > the keys returned are treated only as containment attributes on the
> sdo
> > graph?
> >
> > Adriano Crestani
> >
>
