[
https://issues.apache.org/jira/browse/OAK-3591?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15006298#comment-15006298
]
Chetan Mehrotra commented on OAK-3591:
--------------------------------------
Fixed this with 1714520 in trunk
The issue was that {{fulltextEnabled}} was covering 2 cases either some
property is having analyzed=true or nodeScopeIndex= true. IndexPlanner was
relying on {{fulltextEnabled}} to decide if it should answer a fulltext query.
Here it did not considered the case that fulltext query can be nodeScoped and
there might be no nodeScoped property
As a fix now a new property {{nodeFullTextIndexed}} is introduced in
{{IndexingRule}}. This would be true only if a fulltext field for node would be
created due to either aggregation or some property having nodeScopeIndex=true.
{{indexPlanner}} would now use that to determine if it can handle query like
_contains(., 'foo')_. So if only {{analyzed}} is true then such query would not
be handled.
A similar case was coming when indexRule is for nt:base and some property had
analyzed=true. In such a case {{indexesAllNodesOfMatchingType}} was being
calculated as true as it relied on {{fulltextEnabled}} to determine if all node
matching given rule would be indexed or not. And in such a case planner would
think it can answer the query like _/jcr:root//element(*, nt:file)_ as it
think that all nodes of type nt:file would be indexed given nt:file > nt:base.
Now {{indexesAllNodesOfMatchingType}} is determined on basis of
{{nodeFullTextIndexed}} as in only that case all nodes matching given rule
would be indexed irrespective if any specific property is present or not
With this change also modified to {{PropertyDefinition}} to treat all config to
be false if {{index}} is false
> Lucene index with 'analyzed=true' sometimes used by mistake
> -----------------------------------------------------------
>
> Key: OAK-3591
> URL: https://issues.apache.org/jira/browse/OAK-3591
> Project: Jackrabbit Oak
> Issue Type: Bug
> Components: lucene, query
> Reporter: Thomas Mueller
> Assignee: Chetan Mehrotra
> Fix For: 1.3.11
>
>
> A Lucene index with a property that is configured as "analyzed = true" is
> sometimes used by mistake. Example:
> {noformat}
> oak:index/testLuceneIndex (oak:QueryIndexDefinition)
> compatVersion: 2
> type: lucene
> async: "async"
> indexRules (nt:unstructured)
> nt:base (nt:unstructured)
> properties (nt:unstructured)
> xyz (nt:unstructured)
> propertyIndex: true,
> analyzed: true,
> name: xyz
> query:
> /jcr:root/content//*[jcr:contains(., '1234')]
> {noformat}
> The index is used, but the result does not contain nodes with properties abc
> = '1234'.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
