On Wed, Aug 25, 2010 at 2:50 PM, Yonik Seeley <yo...@lucidimagination.com> wrote: > On Wed, Aug 25, 2010 at 10:55 AM, Eric Grobler > <impalah...@googlemail.com> wrote: >> Thanks for the technical explanation. >> I will in general try to use lex and sort by count in the client if there >> are not too many rows. > > I just developed a patch that may help this scenario: > https://issues.apache.org/jira/browse/SOLR-2089 > > If you have the ability to try out trunk, I'd be interested in the results.
Oh, wait, this will only help with multiValued fields... re-reading your description, it looks like your field is single valued? I guess the time is then taken in the priority queue (actually a BoundedTreeSet) that keeps the top 5000 results sorted. Still it's surprising that it's taking that long. Was the 1 sec obtained from QTime in the Solr header, or measured externally? If measured externally, what was the QTime as Solr reported it? -Yonik http://lucenerevolution.org Lucene/Solr Conference, Boston Oct 7-8
