[
https://issues.apache.org/jira/browse/SOLR-6913?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14266163#comment-14266163
]
Alexandre Rafalovitch commented on SOLR-6913:
---------------------------------------------
All fields have to be multiValued or there needs to be a way to _upgrade_ a
field definition to multiValued.
The last (single->multiValued upgrade) is what ElasticSearch does behind the
scenes. Or more like always being multivalued, but return content as value or
as array depending on how many items are there. Of course that ambiguity may
then break the clients if they expect a single value and get an array back.
The other option is to introduce dry-run URP that looks at all the values
before creating the type. As per the side-discussion in
[SOLR-6016|https://issues.apache.org/jira/browse/SOLR-6016?focusedCommentId=14060934&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-14060934].
> audit & cleanup "schema" in data_driven_schema_configs
> ------------------------------------------------------
>
> Key: SOLR-6913
> URL: https://issues.apache.org/jira/browse/SOLR-6913
> Project: Solr
> Issue Type: Task
> Reporter: Hoss Man
> Assignee: Steve Rowe
> Priority: Blocker
> Fix For: 5.0
>
> Attachments: SOLR-6913-trim-schema.patch
>
>
> the data_driven_schema_configs configset has some issues that should be
> reviewed carefully & cleaned up...
> * currentkly includes a schema.xml file:
> ** this was previously pat of the old example to show the automatic
> "bootstraping" of schema.xml -> managed-schema, but at this point it's just
> kind of confusing
> ** we should just rename this to "managed-schema" in svn - the ref guide
> explains the bootstraping
> * the effective schema as it currently stands includes a bunch of copyFields
> & dynamicFields that are taken wholesale from the techproducts example
> ** some of these might make sense to keep in a general example (ie: "\*_txt")
> but in general they should all be reviewed.
> ** a bunch of this cruft is actually commented out already, but anything we
> don't want to keep should be removed to eliminate confusion
> * SOLR-6471 added an explicit "_text" field as the default and made it a
> copyField catchall (ie: "\*")
> ** the ref guide schema API example responses need to reflect the existence
> of this field:
> https://cwiki.apache.org/confluence/display/solr/Schemaless+Mode
> ** we should draw heavy attention to this field+copyField -- both with a "/!\
> NOTE" in the refguide and call it out in solrconfig.xml & "managed-schema"
> file comments since people who start with these configs may be suprised and
> wind up with a very bloated index
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
