SOLR-2412 is a fairly thin wrapper around LUCENE-2369. LUCENE-2369 was designed with the clear trade-offs
* Slow startup
* Low memory overhead
* Fast response
with the archetypal usage scenario being a large index containing one or more rich hierarchies that is batch-updated every night (see Hierarchical faceting – working code for more details). With fear of misrepresenting, SOLR-64 and SOLR-792 were created from a feature-standpoint with performance characteristics being secondary.
Feature wise, SOLR-2412 (let’s call it Exposed faceting from now on) differs markedly from pivot faceting (SOLR-792) at this time, as neither of the two solutions can do what the other one does. However, I feel confident that Exposed faceting can be tweaked to do pivot faceting later on. The main reason to use Exposed over SOLR-792 would be to change trade-offs.
Compared to SOLR-64, Exposed faceting’s features differs primarily by supporting multiple paths per document: A product belonging to multiple categories, multiple locations for a bus route and so on.
The next step is to create a test bed for doing performance measurements on Exposed vs. Solr’s different faceting implementations. Naturally the hoped-for outcome is that Exposed is markedly better under the defined trade-offs.