fultext search and associated objects

milazzo · **Joined:** Thu Apr 12, 2007 2:34 am **Posts:** 5

I have two classes:

Code:

@Indexed
public class Book implements Serializable
{
   @DocumentId
   private Integer id;

   @ContainedIn //is this the correct annotaion??
   private Set <Keyword> keywords = new HashSet();

//whole bag of other fields
............

}

@Indexed
public class Keyword implements Serializable
{
   @DocumentId
   private Integer id;

   @ IndexedEmbedded //is this the correct annotaion??
   @Field(name="keyword", index=Index.TOKENIZED, store=Store.YES)
   private String name;

   //do I need to index this????
   private Set <Book>books = new HashSet();

..........
}

And my query code looks like this:

Code:

FullTextSession fullTextSession = Search.createFullTextSession(session); 
//why "id" here???
org.apache.lucene.queryParser.QueryParser parser = new QueryParser("id", new StopAnalyzer() );   
String queryString = "keyword:france"; //for example
org.apache.lucene.search.Query luceneQuery = parser.parse(queryString );
//do I need to add Keyword.class here????
org.hibernate.Query fullTextQuery = fullTextSession.createFullTextQuery( luceneQuery, Book.class/*, Keyword.class*/ );
result = fullTextQuery.list();

And I want to do a fulltext search returning only Book objects. This is working nicely if I just search the Book fields - but I also want it to match Keyword names.

Firstly I'm unclear about the roles of @ContainedIn and @IndexedEmbedded. Have I got it right?

At present, it builds two indexes - one for Books and one for Keywords. But I only want to get Book objects back. Is this correct? Or do I need to "embed" keyword index inside the book index?

Also, I wonder if I perhaps need to index the
private Set <Book> books = new HashSet();
in my Keywords class?

And what arguments should I pass to the method
fullTextSession.createFullTextQuery()??
I've noticed that if I pass the Keyword.class, I can get my Keyword objects back - but then I have to go through a load of laborious code to get back the corresponding Book object.

Does anyone know a good online tutorial that explains all this? I've read the docs and googled around but not found anything beyond the Hibernate Reference Guide and have spent a couple of days on this without getting much further - so any help anyone can give me would be much appreciated. Thanks

emmanuel · **Posted:** Thu Apr 12, 2007 5:48 pm

Currently Hibernate Search does not support IndexedEmbedded collections (not sure how it should look like).

IndexedEmbedded is only useful on association attributes (ie one annotated with @*ToOne) or embedded attribute (one annotated with @Embedded), so @indexedEmbedded does not make sense

What you want is all the books having a given keyword, this is not possible today but the query should look like

"keywords.name:france" filtered by Book.class only.

In the mean time you'll need to retrieve the appropriate keyword and then do a second HQL query looking for books by keyword id.

milazzo · **Joined:** Thu Apr 12, 2007 2:34 am **Posts:** 5

Thanks Emmanuel - that clears my doubts and points me in the right direction.

dgovoni · **Joined:** Wed Mar 28, 2007 2:51 pm **Posts:** 16

Hi guys, I was just trying to do this as well and it doesn't seem to work for me. I was following the example here: http://www.hibernate.org/hib_docs/searc ... pping.html

I have a class with a collection of other classes (onetomany).

The example is hard to understand because Address embeds a collection of Places which embeds Address.

So I don't understand this. Can I not index associated collections of embedded classes or not? And is there a better example of this?

For example, how would it be annotated for this.

Code:

class Customer {
 
   private String name;
   private Collection accounts;

   public Collection getAccounts() {
       return accounts;
   }

   public String getName() { return name; }

}

class Account {
    private String name;

    public String getName() { return name; }

}

What i want is to do "name:aname" and have it search Customer and all its accounts "name" field. Or I will settle for something like "name:aname AND account.name:aname".

Thanks for any help! This is an awesome feature if it works.

Darren

emmanuel · **Posted:** Fri Apr 13, 2007 3:41 pm

you want to retrieve customers with a particular name having accounts with a particular name, is that correct (correlated query)?

as I said earlier, it is not yet possible to index correlated collections

But you can do it the other way around

Look for Account.class
name:aname AND customer.name:othername

You will retrieve accounts, then you can access each customer.

I need to think about adding the colection indexing feature. It's a bit more complex than it sounds

milazzo · **Joined:** Thu Apr 12, 2007 2:34 am **Posts:** 5

dgovoni wrote:

The example is hard to understand because Address embeds a collection of Places which embeds Address.

So I don't understand this. Can I not index associated collections of embedded classes or not? And is there a better example of this?

Yes, I agree with Darren entirely - I don't completely understand the example either and spent a long time puzzling over it. A clearer example might perhaps spare a lot of the confusion.

But with Emmanuel's pointer, I got my search queries working nicely today.

Thanks again guys,

emmanuel · **Posted:** Fri Apr 13, 2007 5:06 pm

Heard
http://opensource.atlassian.com/projects/hibernate/browse/HSEARCH-38

If you have any idea on how to make it simpler, add them to the case

dgovoni · **Joined:** Wed Mar 28, 2007 2:51 pm **Posts:** 16

You are awesome. Thanks for opening the ticket.

I understood about how to query using the subfields, but based on the example I could not figure out how to "annotate" it properly because my attempts failed to return results using the proper query fields.

Thanks!!

PS. I agree that correlated collection searching is a tricky problem.

emmanuel · **Posted:** Fri Apr 20, 2007 3:08 pm

I have added support for @indexedEmbedded on collections
http://opensource.atlassian.com/projects/hibernate/browse/HSEARCH-40
It was easier (and more useful) that I thought

fabreax · **Posted:** Mon May 07, 2007 11:11 am

Where could we find the Beta 2 to have this feature ?

einherier · **Joined:** Fri Apr 20, 2007 10:31 am **Posts:** 10

You can access it via SVN.

fabreax · **Posted:** Tue May 08, 2007 4:29 am

Ok, thank you ;).

And there is the support of @indexedEmbedded ?

emmanuel · **Posted:** Tue May 08, 2007 9:26 am

yesw SVN (trunk) has support for @IndexedEmbedded in collections

fabreax · **Posted:** Tue May 08, 2007 10:33 am

Thank you, i will try ;)

hardy.ferentschik · **Posted:** Tue May 08, 2007 5:29 pm

I think I found a workaround for the collection indexing problem. I for example have a class which contains a collection of Location classes. For indexing purposes I am creating a transient dummy getter which I use for indexing and searching, eg:

Code:

    @Transient
    @Field(name = "location", index = Index.TOKENIZED)
    public String getLocationIndexString() {
        String locations = "";
        if (getLocations() != null) {
            for (Location loc : getLocations()) {
                locations += loc.getName() + " ";
            }
        }
        return locations;
    }

Making the field transient will stop Hibernate to persist a locationIndexString property, but the @Field annotation will still make sure that the locations get indexed. I just concatenate all locations for indexing. Specifying Index.TOKENIZED will then split it up into seperate tokens again.

I am not sure if this is such a good idea, but it seems to work.

Cheers,
Hardy