[
https://issues.apache.org/jira/browse/SOLR-8767?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16461800#comment-16461800
]
Simon Rosenthal commented on SOLR-8767:
---------------------------------------
+1 on changing the behavior. Was just bitten by this real-time get behavior in
a situation where I was using a copyfield to effectively rename a -->b (a was
then defined as non-indexed, non-stored).
Not the intended use of copyfield, I know, and I could probably use a
FieldNameMutatingUpdateProcessor instead (though this results in including
field name manipulations in solrconfig.xml, where it really doesn't belong,
rather than in the schema).
> RealTimeGetComponent and stored/copyField exclusion
> ---------------------------------------------------
>
> Key: SOLR-8767
> URL: https://issues.apache.org/jira/browse/SOLR-8767
> Project: Solr
> Issue Type: Bug
> Components: SolrCloud
> Reporter: Erik Hatcher
> Priority: Critical
>
> Consider this scenario: schema has fields `a` and `b` defined, both stored.
> A copyField is defined from `a` => `b`. A document is indexed with `id=1;
> b="foo"`. A real-time /get will not return field `b` because
> RealTimeGetComponent.toSolrDoc currently excludes copyField destinations
> (despite, in this situation, the source of that copyField not being sent in).
> Granted this is a bit of a diabolical case (discovered while tinkering with
> cloud MLT tests), but isn't that far fetched to happen in the wild.
> Maybe real-time /get should return all fields set as stored, regardless of
> copyField status?
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
