[
https://issues.apache.org/jira/browse/SOLR-9822?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15726360#comment-15726360
]
Yonik Seeley commented on SOLR-9822:
------------------------------------
I tried yet another approach of bulk gathering the id+ord in an array and then
looping over that in the calling code, but it was much slower than the lambda
(although still faster than current master w/o patching by 10%). Still slow
enough I won't bother attaching the patch.
In the spirit of progress over perfection, we should probably just commit the
first approach (since it gives a 50% speedup in those cases.), but limited to
the two call sites in that patch (in FacetFieldProcessorByArrayDV). We
shouldn't over-generalize the results found here. It may be that a lambda-type
approach will work better in other contexts, and those will need to be tested.
It's also the case that encapsulating this logic will make it easier to
introduce/maintain additional optimizations such as actually using the skipping
of the docvalues iterator when it's sparse vs our domain set).
> Improve faceting performance with FieldCache
> --------------------------------------------
>
> Key: SOLR-9822
> URL: https://issues.apache.org/jira/browse/SOLR-9822
> Project: Solr
> Issue Type: Sub-task
> Security Level: Public(Default Security Level. Issues are Public)
> Components: Facet Module
> Reporter: Yonik Seeley
> Assignee: Yonik Seeley
> Fix For: master (7.0)
>
> Attachments: SOLR-9822.patch, SOLR-9822_OrdValues.patch,
> SOLR-9822_lambda.patch
>
>
> This issue will try to specifically address the performance regressions of
> faceting on FieldCache fields observed in SOLR-9599.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
