Re: [infinispan-dev] Grouping lucene query on Infinispan

Hi Ray, yes it's the right place for such questions. This is a great example of why we expose the IndexReader so far: so that power users can do more than what we expose over simplified APIs, be it because the feature really need low level access, or simply because - like in this case - we didn't add a nicer/simpler API for the feature yet. The method getSearchFactory() was deprecated as we're trying to simplify the API, but I'm realizing now that this would also prevent accessing the IndexReader(s), and there definitely will always be good reasons to need that. Another reason for deprecating accessing the underlying search engine, is that we have plans for aggressive automatic index sharding, and it gets complex for users if we don't provide a good abstraction. But accessing an IndexReader for Grouping is a very strong case, I agree we should not take that away and created: - https://issues.jboss.org/browse/ISPN-5332 Also - while there will always be good reasons to access the IndexReader - some experiments have started in Hibernate Search to provide an higher level API to the Grouping API - https://github.com/hibernate/hibernate-search/pull/791 The API will have to change (for one because the method name is not matching the returned type anymore), and we might not want to expose the SearchIntegrator but we will expose IndexReader instances in some way. Thanks for your feedback! Glad it came before we actually removed it. Sanne On 26 March 2015 at 16:51, Raymond Rennell Douglass III <rdouglas(a)umd.edu&gt; wrote: