Hi there.

I'm looking for a way to add arbitrary documents into my Hibernate search application.
By "arbitrary", I mean documents that are not entities.
I understood that I'll have to handle manually the document updates and removal, and it's ok.

I looked on such facilities in the SearchFactory, but everything is related to an entity.

Any idea to performs such a thing ? I really don't want to set up another native lucene manually.

Thanks a lot.

Hi,

I am just wondering why you wanted to do that. Wouldn't it be better from a resource management point of view and application design to have a separate Lucene index? Why adding unrelated documents to the HSearch managed index?

Not only would you have to manage the document updates manually, but the problem of concurrent access to the index. Maybe you could describe your use case a little more.

That all said, you can always gain access to the underlying Lucene resources via the SearchFactory. There you can get hold of the underlying Lucene directory, but I would not recommend going down this path.

--Hardy

The Hibernate Search in Action book has some examples about indexing Word and PDF documents, but as Hardy said it makes mostly sense when they are structured as "binary attachments" of entities, so that they are not totally unrelated to the entities.
Ideally I'd map them as content of a lazily loaded Blob field of an entity, as that doesn't necessarily mean they are stored on the database: using a custom persister you can have the entity represent the metadata (going in index and database), and have the original document stored elsewhere (FS?), have Search handle a consistent way to index metadata+content and have easy to use pointers to both in the resultsets.
There's no secret in the book, just some examples: you still have to use a custom fieldbridge to extract the text from the binary/proprietary formats, it shows how to plug in converters such as TIKA or POI.

You're right, I need to give some context.

I'm writing a game engine in Java, which can be manage in a dynamic way : you define all the gaming stuff while the application is running.
I massively used Groovy for storing game rules, and JPA for storing types and objects.

So my application stores to kind of POJOs:
- classical entities, persisted in database through JPA: "types and objects"
- other ones are "actions", a POJO linked to a groovy script and representing a gaming rule

Thus I have some classical DAOs for types and objects, and other DAOs that persists actions on the file system, in groovy script.

But all that stuff is inner mechanics, and from the user's point of view, he manipulates types, objects and rules.
That is why I wanted to have in a same request results classical entities and "arbitraty documents" : my action POJO.

Unfortunatly, as I said, the SearchFactory does not allow access or creation of a DirectoryProvider unrelated to an persisted Entity class. Unless I missed a method ?

thanks for the context, tough it's not clear to me what you need, I'm assuming you want to index annotated POJOS which are not entities:

Consider that Indexed entities don't necessarily need to be JPA entities, as long as they are POJOs with an @Indexed annotation then Search will be able to provide you with an indexed service for it (transformations into and from Lucene's documents and DirectoryProvider/Reader management). Any object though needs an identifier, when it's a JPA/Hibernate we figure it out easily via the @Id mapped property, but you can use a plain @DocumentId and avoid the @Id, or you can play with @ProvidedId.
The downside of not managing entities, is that Hibernate won't automatically create events to update your object, you'll have to fire events yourself.
Use programmatic configuration to add more POJOs to be indexed (autodiscovery is also not going to work as it won't do annotation scanning on @Indexed objects).

A great source of examples is the Infinispan Query module:
https://github.com/infinispan/infinispa ... ster/query
This project sets up it's own event listeners and indexes objects put in the Infinispan cache (a key/value store), have a look at the examples - it's still a new, small and simple project, just a bunch of classes, so I hope it gets you an idea without getting lost in it.

Thank Sane.

Indeed, InfiniSpan/query seems pretty hard to understand from the first sight, but I'll have a deeper look as soon as possible.

I tried to add my annotated POJO indexed by Hibernate Search (I run Hibernate with Spring):
applicationContext.xml


ConfigurationPostProcessor.java


Action.java


But when I run th following code :


I got an error indicating that the Action.class is not recognized :

Hi ! As explained, it really seems that Hibernate Search is unable to index POJO that are just annotated with @Indexed, and that are not @Entity or @MappedSuperclass.

Am I wrong ?

Damien.

No it should work pretty well, all tests on Infinispan are green :)

I don't know much about Spring, but it seems you're registering your pojos as entities, don't mess with the persistenceContext and just start the SearchFactory:

But then it's up to you to send updates to the engine in form of "update", "delete", "add" to the index operations, and to run searches using the createHSQuery() method.