[
https://issues.apache.org/jira/browse/SOLR-13289?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17100329#comment-17100329
]
Tomas Eduardo Fernandez Lobbe commented on SOLR-13289:
------------------------------------------------------
bq. Grouping
Grouping uses a different collector. It’s not included as part of this PR.
Maybe there are things that can be done to allow WAND in some cases for Grouping
bq. Custom scoring via value sources
bq. Payloads
The optimization only applies if the natural score is the first thing to be
sorted on. i.e. {{sort=score desc, id asc}}. If there is anything than “score”
there, then the optimization doesn’t apply, not OOTB at least.
bq. Is TwoPhaseIterator or custom ranking in collapse/expand mode or in
response writer the way? or have to implement a custom query overriding the
createWeight using ScoreMode.TOP_SCORES?
I don’t know exactly what you want to do. In general, block-Max/WAND requires
that the DISI can calculate the maximum contribution for a query in a block of
documents with the indexed {{impacts}}. If you want to have more signals for
calculating the score I guess those could be added as Impacts using a custom
codec and then you could have a custom scorer to use those ([~jpountz], let me
know if this is wrong). If you just need a custom similarity that uses TF/IDF,
then you should be able to use it.
> Support for BlockMax WAND
> -------------------------
>
> Key: SOLR-13289
> URL: https://issues.apache.org/jira/browse/SOLR-13289
> Project: Solr
> Issue Type: New Feature
> Reporter: Ishan Chattopadhyaya
> Assignee: Tomas Eduardo Fernandez Lobbe
> Priority: Major
> Attachments: SOLR-13289.patch, SOLR-13289.patch
>
> Time Spent: 3h
> Remaining Estimate: 0h
>
> LUCENE-8135 introduced BlockMax WAND as a major speed improvement. Need to
> expose this via Solr. When enabled, the numFound returned will not be exact.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
