"could not initialize a collection" + @Lob + MSSQL

When using Blob-fields in n-to-m-relations Hibernate and MSSQL are failing for some reason.

SQL Error: 421, SQLState: S0001
The image data type cannot be selected as DISTINCT because it is not comparable.
...
could not initialize a collection: [Dataset.documents#someID]

My classes look as follows:

@Entity
class Dataset {
    @OneToMany(fetch = FetchType.LAZY)
    public List<Document> documents = new Ar开发者_Go百科rayList<Document>();
}

@Entity
class Document {
    @Id
    public long id;

    @Lob
    public byte[] data;
}

Any ideas on this? I already tried using a Set or Document[] to avoid the errors. It seems that Hibernate always tries a distinct SELECT on my tables. How can I workaround this?

[1] MSSQL error codes

---

It would have been interesting to post the executed queries and the tables. But one difference I can think of between the two mappings (unidirectional one-to-many vs a bidirectional) is the way they're represented at the database level.

By default, a unidirectional one-to-many will use a join table:

DATASET       DATASET_DOCUMENT       DOCUMENT
-------       ----------------       --------
ID            DATASET_ID             ID
              DOCUMENT_ID

While a bidirectional will use the following representation:

DATASET       DOCUMENT  
-------       ----------
ID            ID        
              DATASET_ID

In JPA 2.0, it is now possible to use a unidirectional association without a join table (in a standard way) by specifying a @JoinColumn on the @OneToMany side:

@Entity
class Dataset {
    @Id
    @Column(name="ID")
    public Long id;
    ...
    @OneToMany
    @JoinColumn(name="DATASET_ID", referencedColumnName="ID")
    public List<Document> documents = new ArrayList<Document>();
}

I would give the above a try.

References

• JPA 2.0 Specification
  • Section 11.1.21 "JoinColumn Annotation"
• JPA Wikibook
  • Undirectional OneToMany, No Inverse ManyToOne, No Join Table (JPA 2.0)

---

I managed to solve the problem by adding a reversed reference to my Document class. I am sure there is a more direct way, because I do not need this reference at all, so it basically only helps to workaround the problem.

@Entity
class Dataset {
    @OneToMany(fetch = FetchType.LAZY, mappedBy = "dataset")
    public List<Document> documents = new ArrayList<Document>();
}

@Entity
    class Document {
    @Id
    public long id;

    @Lob
    public byte[] data;

    @ManyToOne
    @JoinColumn
    public Dataset dataset;
}

---

I have fixed this in another way. I already had the mappedBy = "enrollment" on my OneToMay and on my ManyToOne I had a @JoinColumn(name="enrollment", nullable=true). So this was not a relation issue.

Instead I found a post on the hibernate forum suggesting overriding the supportsUnionAll() from the Dialect class. It defaults returns false, but SQL Server has since v6.5 supported UNION ALL. Remember the extend SQLServerDialect if you have not done so already. So this could be another possible solution for the "The * data type cannot be selected as DISTINCT because it is not comparable" error.