After doing some digging into our search service I see that the group we
outsourced this project to wrote a custom hit collector and a separate
sorting method. This is probably due to the requirements of being able to
sort by up to 2 fields with either field being ascending or descending. I
see it is the sorting method they wrote is the problem.  Thanks for every
one's input. Sorry I didn't research it further before I jumped to the
conclusion it was a lucene issue.

Mark

On Jan 12, 2008 6:26 AM, DIGY <[EMAIL PROTECTED]> wrote:

> Lucene uses cached values for sorting and warm-up time of the cache takes
> nearly 99% of the search time. So, If you do not close the IndexSearcher
> other searches will be very fast.
>
> DIGY
>
> -----Original Message-----
> From: Mark Burks [mailto:[EMAIL PROTECTED]
> Sent: Thursday, January 10, 2008 2:36 PM
> To: [email protected]
> Subject: Re: Search while sorting is slow
>
> Thanks Michael. I have considered doing as you have suggested but  I was
> hoping that i was doing something wrong with the index build and it would
> be
> a quick fix. Guess I will be going down that road now.
>
> I don't understand why lucene is slow at sorting when that is one of
> things
> that it is supposed to do efficiently.
> Mark
>
> On Jan 10, 2008 12:04 AM, Michael Mitiaguin <[EMAIL PROTECTED]> wrote:
>
> > I also found built-in sorting very slow and ended up with populating
> > certain C# collection from hits in default order and then sorting that
> > collection.  Apart from greater speed , more possibilities to
> > implement any kind of sorting.
> >
> > On Jan 10, 2008 3:19 PM, Mark Burks <[EMAIL PROTECTED]> wrote:
> > > I have an index that when i search for term with out sorting the
> results
> > are
> > > returned in .7 seconds or less. When I choose to search with a column
> > chosen
> > > for sorting the results will return in 7 - 10 seconds.  The larger the
> #
> > of
> > > hits the longer it takes when sorting.  The fields that can be sorted
> on
> > are
> > > unique to the document,stored in the index as un_tokenized. I am using
> > > lucene.net 2.0. Index size is normally around 2.5 gig but i have the
> > > same problem on a 500 meg index too.  Has anyone else experienced
> this?
> > >
> > > Any help will be greatly appreciated.
> > >
> > > Thanks,
> > > Mark Burks
> > >
> >
>
>

Reply via email to