lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Simon Willnauer (JIRA)" <j...@apache.org>
Subject [jira] Commented: (LUCENE-2694) MTQ rewrite + weight/scorer init should be single pass
Date Wed, 22 Dec 2010 15:55:02 GMT

    [ https://issues.apache.org/jira/browse/LUCENE-2694?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12974251#action_12974251
] 

Simon Willnauer commented on LUCENE-2694:
-----------------------------------------

bq. I think instead of ReaderView we could change Weight.scorer API so that instead of receiving
IndexReader reader, it receives a struct that has parent reader, sub reader, ord of that sub?

so I changed the Weight#scorer API to use a class called ScoreContext that holds the parent
reader, the current sub and the subs ord. That change is absolutely massive! I don't upload
that change since I think if we do that we need to do it in a different issue anyway. I ran
into a couple of problems:

* if we pass in such a context we also need to change the explain interface since its calling
Weight#scorer here and there
* Once we pass in the Context stuff like QueryWrapperFilter doesn't work anymore since it
doesn't know which ord the incoming reader has. So Filters would need a context too. I don't
like that!
* Stuff like scoreDocsInOrder are hard to put into such a context since almost all scorers
internally are called with scoreDocsInOrder=true with a contant. meaning that nobody really
respects the incoming value for subscorers though. but if i just forward the context the member
needs to be set to true or the context needs to be cloned for subs - see BooleanQuery for
instance.

* such a context would somehow enforce that MTQ are only executed against the Reader they
where rewritten against. Which is how it should be IMO but we are also depending on that everybody
who uses a MTQ knows exactly how the query was rewritten which is kind of not obvious. I think
we need a better way to enforce stuff like that.

after all I think this must be done in a different issue though. This issue was meant to make
MTQ single pass so lets do that first.... progress over perfection ey mike :). 
Nonetheless, it seems like that we need to rethink the Weight API entirely I also don't like
that Weight operates on Searcher instead of IndexSearcher though.


> MTQ rewrite + weight/scorer init should be single pass
> ------------------------------------------------------
>
>                 Key: LUCENE-2694
>                 URL: https://issues.apache.org/jira/browse/LUCENE-2694
>             Project: Lucene - Java
>          Issue Type: Improvement
>          Components: Search
>            Reporter: Michael McCandless
>            Assignee: Simon Willnauer
>             Fix For: 4.0
>
>         Attachments: LUCENE-2694-FTE.patch, LUCENE-2694.patch, LUCENE-2694.patch, LUCENE-2694.patch,
LUCENE-2694.patch, LUCENE-2694.patch, LUCENE-2694.patch
>
>
> Spinoff of LUCENE-2690 (see the hacked patch on that issue)...
> Once we fix MTQ rewrite to be per-segment, we should take it further and make weight/scorer
init also run in the same single pass as rewrite.

-- 
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: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org


Mime
View raw message