You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@lucene.apache.org by "Erick Erickson (JIRA)" <ji...@apache.org> on 2017/03/20 16:46:41 UTC

[jira] [Commented] (SOLR-10320) Perform secondary sort using both values in and outside Solr index

    [ https://issues.apache.org/jira/browse/SOLR-10320?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15933017#comment-15933017 ] 

Erick Erickson commented on SOLR-10320:
---------------------------------------

first, before submitting a JIRA it's a good idea to discuss something on the user's list.

Do you have a patch you'd like to submit? 

You can sort by function queries, and you can write custom function query. Since based on something outside Solr is, by definition, custom for most situations, you'd have to write some custom code anyway so would that do?

And the overriding concern here is speed. Every doc that has a non-zero score will have to go through sorting. It's hard to see how going out to an external-to-solr source for, potentially, billions of documents has to be very, very efficient.

> Perform secondary sort using both values in and outside Solr index
> ------------------------------------------------------------------
>
>                 Key: SOLR-10320
>                 URL: https://issues.apache.org/jira/browse/SOLR-10320
>             Project: Solr
>          Issue Type: New Feature
>      Security Level: Public(Default Security Level. Issues are Public) 
>            Reporter: Yuchuan Zhou
>
> There are some situations that we need to sort results based on values outside of Solr (say, from a separate datastore or a data analytics service that ranks entities based on analytic results). There is also the need to return results in a deterministic order but applying a dynamic chain of sorting and/or ranking algorithms to the result set. This chain would be processed as a secondary sort implementation, where ties returned from one sorting/ranking algorithm are passed to the next sorting/ranking algorithm in the chain until all ties are resolved, resulting in a deterministic result order. This chain should have the ability to apply sorting algorithms that use data found within the solr index as well as outside of the index.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

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