[eclipselink-users] Removing from @ManyToMany collection doesn't work wh

[Andreas Joseph Krogh <andreas@xxxxxxxxxx>]

Hi all, I have the following ManyToMany which fails to remove an entry in the join-table when an element in the list is removed:

 

I have a user, with natural key userName (unique) and each user can be member of several groups. Group has the natural-key groupName, and the join-table joins on their natural-keys.

 

The join-table is like this:

 

 groupname | character varying | not null
 username  | character varying | not null
Indexes:
    "onp_user_group_pkey" PRIMARY KEY, btree (groupname, username)
    "onp_user_group_gropuname_idx" btree (groupname)
    "onp_user_group_username_idx" btree (username)
Foreign-key constraints:
    "$1" FOREIGN KEY (groupname) REFERENCES onp_group(groupname) ON UPDATE CASCADE ON DELETE CASCADE
    "$2" FOREIGN KEY (username) REFERENCES onp_user(username) ON UPDATE CASCADE ON DELETE CASCADE

 

 

The entities, User and Group, are defined as:

 

@Entity
@Table(name = "onp_user")
@PrimaryKeyJoinColumn(name = "id")
@DiscriminatorValue("USER")
class User

 

    @Id

    @Column(name = "id", nullable = false)

    Long id;

 

    @Column(name = "username", unique = true, nullable = false)

    String userName

...

    @ManyToMany(fetch = FetchType.LAZY)
    @JoinTable(name = "onp_user_group",
        joinColumns = {@JoinColumn(name = "username", referencedColumnName = "username")},
        inverseJoinColumns = {@JoinColumn(name = "groupname", referencedColumnName = "groupname")}
        )

        List<Group> groups;

 

 

 

@Entity
@Table(name = "onp_group")
@PrimaryKeyJoinColumn(name = "id")
@DiscriminatorValue("GROUP")
class Group

 

    @Id

    @Column(name = "id", nullable = false)

    Long id;

 

    @Column(name = "groupname", unique = true, nullable = false)
    String name;

 

 

 

The problem is that when removing an entry in user.groups, then the entry is not removed from the join-table "onp_user_group", because the query EL ends up executing is:

LOG:  execute cached-531203064: DELETE FROM onp_user_group WHERE ((groupname = $1) AND (username = $2))
DETAIL:  parameters: $1 = 'Somegroup', $2 = NULL

 

So NULL is passed as the value for "userName".

 

I see in the method (EL-source) CollectionMapping.writeChanges that the WriteObjectQuery passed as an argument (with argument-name "query") has a property "translationRow" which is used to map keys, and this Map contains the following when trying to persist changes:

 

onp_user.id=134

 

But RelationTableMechanism.addRelationTableSourceRow has the value "onp_user.username" in its sourceKeyFields, which doesn't map to an entry in the passed sourceRow. This results in the addRelationTableSourceRow method adding NULL as a value to key="onp_user.username", because sourceValue is assigned the value NULL:

 

    public AbstractRecord addRelationTableSourceRow(AbstractRecord sourceRow, AbstractRecord databaseRow) {
        int size = sourceKeyFields.size();
        for(int i=0; i < size; i++) {
            Object sourceValue = sourceRow.get(sourceKeyFields.get(i));
            databaseRow.put(sourceRelationKeyFields.get(i), sourceValue);
        }
        return databaseRow;
    }

 

So, is there a way to make this work, or do I have to use primary keys in the join-table? Everything else works nicely, just not removing.

 

Thanks.

 

--

Andreas Joseph Krogh

CTO / Partner - Visena AS

Mobile: +47 909 56 963

andreas@xxxxxxxxxx

www.visena.com