lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Michael Sokolov <msoko...@safaribooksonline.com>
Subject Re: (DocIds satisfying a query) -> (branch of the boolean query as a tree)
Date Sat, 25 Jan 2014 20:24:29 GMT
The highlighters are the only thing I know of (in trunk) that do 
something like that.

Work on this branch 
(https://issues.apache.org/jira/browse/LUCENE/fixforversion/12317158) is 
an attempt to make that more efficient.

In general the problem with doing this during scoring (the filtering 
docid phase) is you don't know which documents will be returned because 
they are being sorted by score (or something else), and only at the end 
do you have the top N.

So there always tends to be an element of re-evaluating the query; it's 
just a question of making that as optimal as possible.

-Mike

On 1/24/2014 12:31 PM, Olivier Binda wrote:
> Hello
>
> While searching a query, I guess that Lucene traverses a 
> Field->Term->DocId structure, filters the docIds that satisfy the 
> query, score them and then sort them
>
> Given a resulting docId, I would like a way to find at least a valid 
> path (or the first valid path or all valid paths) that made this docId 
> pass the filtering phase
>
> For example, if you consider that the query "A:a OR B:a OR C:a" is a 
> tree with the branch
> A:a = 1
> B:a = 2
> C:a = 3
>
> Some docId will get through 1
> others through 2
> and the others through 3
>
>
> Is there a way to do that and what what would be the most efficient 
> way (time and memory wise) ?
>
> I think that highlighting tries to solve quite the same problem (but 
> it seems quite heavy on the computationnal side).
> My idea was : isn't it possible at the "Lucene filtering docId phase", 
> to save the path (an int, if you traverse the query like a tree) that 
> made the docId pass the query ? (which would make it cheap 
> computationnaly wise)
>
> Best regards,
> Olivier
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
> For additional commands, e-mail: java-user-help@lucene.apache.org
>


---------------------------------------------------------------------
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