Another option is to use faceting (via the facet.prefix param) for your
auto-suggest. It's not as fast and scalable as using one of the
Suggester implementations, but it does allow arbitrary fq parameters to
be included in the request to limit the results.
http://wiki.apache.org/solr/SimpleFacetParameters#Facet_prefix_.28term_suggest.29
Doug
On 04/24/2012 04:30 PM, Erick Erickson wrote:
I don't know if there is a really good solution here. The problem is that
suggester (and the trunk FST version) simply traverse the terms in
the index. there's not even a real concept of those terms belonging to
any document. Since your security level is on a document basis, that
makes things hard.
How many users do you have? And do you ever expect to search
across more than one user's files? If not, you could consider having
one core per user. Then the suggestions would be correct and since
the searches would be against the user's core, they'd never see
any documents they didn't own.
But that solution has some complexity involved, and if you have a zillion
users it can be difficult to get right.
You could consider having separate (dynamically-defined) fields that
had the suggestion list for each individual user. that would be
administratively easier. Then you suggestions would simply go against
that user's suggestion field (suggestion_user1 e.g.).
None of this is elegant, but this is not an elegant problem given how
Solr is structured.
Best
Erick
On Tue, Apr 24, 2012 at 2:31 PM, prakash_ajp<prakash_...@yahoo.com> wrote:
I read on a couple of other web pages that fq is not supported for suggester.
I even tried the query and it doesn't help. My understanding was, when the
suggest (spellcheck) index is built, only the field chosen is considered for
queries and the other fields from the main index are not available for
filtering purposes once the index is created.
--
View this message in context:
http://lucene.472066.n3.nabble.com/Auto-suggest-on-indexed-file-content-filtered-based-on-user-tp3934565p3936144.html
Sent from the Solr - User mailing list archive at Nabble.com.
