Copying objects from pm to pm, relations

Question

Hi
We want to copy some data from a relational database to a Versant database
using Kodo JDO. The basic approach is to read some objects, detach them
from the relational PM, and attach them to the Versant PM.
Because the database is BIG, letting JDO take care of the relations is not
feasible, this would take too much memory. Doing the copy process in small
bytes won__t work either, since this will result in double entries in the
target database for tables like __country__ (same value, different oid).
So we are now thinking of another approach. We want to read and write the
tables one by one, without making relations between objects. After copying
to Versant, we will run queries there for all relations using the key
fields (that we will have mapped as simple values).
When trying to do this, we ran into this problem: if we include in our
Java class fields for the relationship objects, Kodo will complain there
is no mapping for them when reading from the relational db.
One solution I have come up with is to declare these fields transient,
copy the data to Versant, then change the Java code and Versant schema to
make them non transient. Very cumbersome and not straightforward to
automate.
Another solution is to describe all the mappings in the JDO files, and
make sure the fields are not in the default fetch group. But this is a lot
of work, especially since we then don__t use the relations at all !
Is there another solution to this, a sort of mapping that we can use both
in the relational case (there it would be a sort of null mapping) and the
Versant case ?
Regards, Luc

Answer 1

luc wrote:
Another solution is to describe all the mappings in the JDO files, and
make sure the fields are not in the default fetch group. But this is a lot
of work, especially since we then don__t use the relations at all !By default, relations are not in the dfg.
Is there another solution to this, a sort of mapping that we can use both
in the relational case (there it would be a sort of null mapping) and the
Versant case ?In the Kodo mapping information, you can set a persistent field to use a
'none' mapping:
<jdbc-field-map type="none"/>
That might do what you're looking for.
-Patrick