Inconsistent field mappings used in joined columns

[talios]

'lo,

I noticed some inconsistency with field column names used by envers for joined tables, I've entered this currently in JIRA as http://jira.jboss.com/jira/browse/ENVERS-15 but thought I'd comment here as well.

Maybe the problem stems from me not using Hibernate (or envers extension) to generate my schema.

My classes have joined annotated as:

 @ManyToOne(cascade = {CascadeType.ALL})
 @JoinColumn(name = "contact_mech_id", referencedColumnName = "contact_mech_id", nullable = false, insertable = true, updatable = true)
 public ContactMech getContactMech() {
 return contactMech;
 }

so that the column name is contact_mech_id, however envers generates the mapping to be contactMech_contact_mech_id instead.

After walking through the code and trying to work out whats going on, changing org.jboss.envers.metadata.VersionsMetadataGenerator#addToOne to use:

Column column = (Column) property.getValue().getColumnIterator().next();
IdMapper relMapper = new SingleIdMapper(property.getName(), column.getName());

instead of the prefixed version things started to work again (which revealed some other issues, which I'll post separately).

Obviously this is a rather hackish alteration. Ideally I'd like to keep my column names in the versions table consistent with my normal table, so I wondering if there was any particular reason for the current behaviour?

[talios]

Interesting. I just reworked our build environment so I can use the Envers schema generator and it dished out the schema as I expected:

 create table smx3.contact_mech_versions (
 _versions_id int4 not null,
 contact_mech_id int4,
 _revision int4 not null,
 info_string varchar(255),
 _revision_type int2,
 contact_mech_type_id int4,
 primary key (_versions_id),
 unique (contact_mech_id, _revision)
 );

Is the hibernate schema generator code using the same VersionsMetadataGenerator that envers uses at run-time?

[adamw]

Hello,

yes, it's exactly the same. Weird :)

I'll take a look.

--
Adam