Hello and Happy New Year :-)

I have the following query:
the documentation recommends NOT to split a given entity index in multiple ones... however I want to do that for
a specific reason:

I have an entity Document which needs to be indexed: a given document can only belong to a project, so
my application will deal with set of documents grouped by project.

The users will only search within a SINGLE project at the time, so I am wondering if in this case it makes
sense to split the Document index per project, so each search will open and search across the project index only
(why searching for documents in projects Alpha & Beta if the user is currently working in project Gamma??)

Any thought?
If this is a viable option how do I set (at run-time) which index must be indexed/searched?

Tnx in advance

Beppe

Hi,

We recommend index sharding only for large indexes. Your case sounds like a perfect case for a Lucene filter. Depending on the user you initialize you filter with the right project to filter on. You could probably cache in this case the filter as well as its result bitset (FilterCacheModeType.INSTANCE_AND_DOCIDSETRESULTS).
Using a filter will also make it easier two for example change across two projects in case your requirements will change.

--Hardy

while I agree with Hardy in recommending to avoid sharding until you're sure it's needed, you might want to play with a new feature released in latest beta:
http://opensource.atlassian.com/projects/hibernate/browse/HSEARCH-251

So you still use the filter approach at API level but get the benefits of having separated indexes.