[jira] [Commented] (SOLR-15049) Optimize TopLevelJoinQuery "self joins"

["Jason Gerlowski (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SOLR-15049?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17253081#comment-17253081 ] 

Jason Gerlowski commented on SOLR-15049:
----------------------------------------

I did a few performance tests this morning in advance of getting this committed. Under appropriate circumstances to take advantage of the optimization I'm seeing throughput nearly double. e.g. each self-joins takes half the time on average that they had before. (Specifically my test uses a 500K doc index, with 100K unique values in the string field being joined on, 5K of which are matched by each 'fromQuery')

Anyway, a nice tidy boost for at least some join usecases.

Hoping to commit today or tomorrow after tests and precommit

> Optimize TopLevelJoinQuery "self joins"
> ---------------------------------------
>
>                 Key: SOLR-15049
>                 URL: https://issues.apache.org/jira/browse/SOLR-15049
>             Project: Solr
>          Issue Type: Improvement
>      Security Level: Public(Default Security Level. Issues are Public) 
>          Components: query parsers
>    Affects Versions: master (9.0)
>            Reporter: Jason Gerlowski
>            Assignee: Jason Gerlowski
>            Priority: Minor
>          Time Spent: 20m
>  Remaining Estimate: 0h
>
> The "TopLevelJoinQuery" join implementation added recently performs well in a variety of circumstances.  However we can improve the performance further by adding logic to handle the special case where fields being joined are the same (e.g. {{fromIndex}} is null and {{fromField}} == {{toField}}).
> Currently this self-join usecase still does a lot of work to convert "from" ordinals into "to" ordinals.  If we make the query (or QParserPlugin?) smart enough to recognize this special case we can skip this step and improve performance greatly in this case.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

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