Question

Given this schema:
+-----------------+------------------+------+-----+-------------------+-------+
| Field           | Type             | Null | Key | Default           | Extra |
+-----------------+------------------+------+-----+-------------------+-------+
| internal_id     | int(10) unsigned | NO   | PRI |                   |       |
| external_id     | varchar(255)     | NO   | PRI |                   |       |
| mapping_type_id | int(4) unsigned  | NO   | PRI |                   |       |
| creation_date   | timestamp        | NO   |     | CURRENT_TIMESTAMP |       |
+-----------------+------------------+------+-----+-------------------+-------+

And this annotation:

@OneToMany(cascade=CascadeType.ALL, fetch=FetchType.EAGER)
@JoinTable(name="mapping",
        joinColumns = { @JoinColumn(name="internal_id") },
        inverseJoinColumns = { @JoinColumn(name="external_id") }
)
@WhereJoinTable(clause="mapping_type_id=2")
private List<Video> videos = Lists.newArrayList();

Which works fine with the legacy database, but when I run these annotated classes through hsqldb with a create-drop option (for basic in-memory unit tests), I get the following:

[elided]
Caused by: java.sql.SQLException: Column not found: VIDEOS0_.MAPPING_TYPE_ID in statement 
[elided]

So I checked with SchemaExport, and hibernate is in fact generating the DB Schema with a join table for the mapping, but it is not including the mapping_type_id field. How do I make the schema generate this with the rest of the schema without interfering with code that runs on the existing database?

Answer 1

The answer to this is to create an embedded ID and use that to map the class through the table, instead of using the join-table syntax. This is easier to fake out if the join table has a row-id, but otherwise, using @Embeddable and @EmbeddedId can make it work.