lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Michael Sokolov <soko...@ifactory.com>
Subject Re: Lucene sort performance roots?
Date Fri, 24 Jun 2011 22:42:40 GMT

> Because of this top-n behavior, its generally slow with Lucene to scan
> deeply into the result set. If you want to go on page 100 of your search
> results, the priority queue must at least have a size of n=docsPerPage*100.
> Because of this, most full text search engines (e.g. Google does this, too)
> prevent you from going to deep into the result set, as it would get slower
> and slower, because the PQ gets bigger and bigger. E.g. Google prevents you
> to go as far as I know beyond page 50 or like that.
>
> Uwe
>

So the trick we application developers have learned, is that, at least 
for field-sorted result sets, you can achieve the same effect as deep 
pagination by adding a range filter to your query.  That is, you can't 
page to result number 10000 quickly, but you *can* page quickly to 
results sorting after "M" in a very large alpha-sorted list (for 
example), which is usually better for users anyway.

-Mike

---------------------------------------------------------------------
To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
For additional commands, e-mail: java-user-help@lucene.apache.org


Mime
View raw message