: doesn't compass use multiple indexes?
I don't relaly know anything about Compass, but if it's managing seperate
index partitions itself then yes it would certainly be hard to try and
point a Solr instance at it's index (since you'd just have ot pick one)
Solr was designed with teh assumption that it would *own* the index and be
bale to manage everything itself -- butthe master/slave model built in to
it makes it easy to configure a 'slave" whose "master" isn't actaully
another Solr index -- with the constraint that it still expects the index
to be a single FSDirectory.
The topic of using a MultiReader (whose primary usage is opening
multi-segment indexes) to search multiple standalone indexes in seperate
directories has come up on the lucene-java lists before ... i've never
tried it myself but if:
a) it really is that easy
b) there is a desire among the Solr user base for Solr to support
searching of indexes not created by Solr (ie: Compass)
...then perhaps we could add two new options:
1) a way to configure solr so that it knows it should treat the index as
read only (this might be as simple as not registering any UpdateHandlers)
2) a hook to register a factory Solr would use when it wants/needs to
open a new IndexReader ... the default implimentation being the current
code that opens an FSDIrectory in ${dataDir}/index ... but another more
complex implimentation could be configured to take in a list of
directories and open a MultiReader across all of them. (or even an
implimentation that used a pure RAMDirectory
-Hoss
