lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Shai Erera (JIRA)" <>
Subject [jira] Commented: (LUCENE-1593) Optimizations to TopScoreDocCollector and TopFieldCollector
Date Mon, 27 Apr 2009 21:36:30 GMT


Shai Erera commented on LUCENE-1593:

bq.EG BooleanWeight.scorer(...) should call BS2's initCountingSumScorer (and/or somehow forward
to BS)?

Ok that "somehow forward to BS" is more problematic than I thought so initially. BS2.score(Collector)
determines whether to instantiate a new BS, add the scorers and call bs.score(Collector),
or to execute the score itself. On the other hand, it uses the same scorers in next() and
skipTo(). Therefore there's kind of a mutual exclusiveness here: either the scorers are used
by BS or by BS2. They cannot be used by both, unless we clone() them. If we want to clone
them, we need to:
* Create a BS in init().
* Clone all the Scorers and pass them to BS.
* Initialize BS2's countingSumScorer.
* In score(Collector) use the class member of BS.

bq. hmm I wonder why this wasn't done so far?

I think I understand now ... the decision on which path to take can only be determined after
score(Collector) is called, or next()/skipTo(). Before that, i.e., when BW returns BS2 it
does not know how it will be used, right? The decision is made by IndexSearcher.doSearch depending
on whether there's a filter (next()/skipTo() are used) or not (score(Collector)).

So perhaps we should revert back to having start() on DISI? Since IndexSearcher can call start
before iterating over the docs, but not if it uses scorer.score(Collector), which is delegated
to the scorer. In that case, we should check whether the countingSumScorer was initialized
and if not initialize it outselves.

Am I missing something?

> Optimizations to TopScoreDocCollector and TopFieldCollector
> -----------------------------------------------------------
>                 Key: LUCENE-1593
>                 URL:
>             Project: Lucene - Java
>          Issue Type: Improvement
>          Components: Search
>            Reporter: Shai Erera
>             Fix For: 2.9
>         Attachments: LUCENE-1593.patch,
> This is a spin-off of LUCENE-1575 and proposes to optimize TSDC and TFC code to remove
unnecessary checks. The plan is:
> # Ensure that IndexSearcher returns segements in increasing doc Id order, instead of
> # Change TSDC and TFC's code to not use the doc id as a tie breaker. New docs will always
have larger ids and therefore cannot compete.
> # Pre-populate HitQueue with sentinel values in TSDC (score = Float.NEG_INF) and remove
the check if reusableSD == null.
> # Also move to use "changing top" and then call adjustTop(), in case we update the queue.
> # some methods in Sort explicitly add SortField.FIELD_DOC as a "tie breaker" for the
last SortField. But, doing so should not be necessary (since we already break ties by docID),
and is in fact less efficient (once the above optimization is in).
> # Investigate PQ - can we deprecate insert() and have only insertWithOverflow()? Add
a addDummyObjects method which will populate the queue without "arranging" it, just store
the objects in the array (this can be used to pre-populate sentinel values)?
> I will post a patch as well as some perf measurements as soon as I have them.

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message