[
https://issues.apache.org/jira/browse/JCR-1885?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12651551#action_12651551
]
Marcel Reutegger commented on JCR-1885:
---------------------------------------
A Lucene IndexReader allows one to set a 'termInfosIndexDivisor', which
controls how many terms are loaded into memory. The search index should have a
configuration parameter that reflects this property.
> Make termInfosIndexDivisor configurable
> ---------------------------------------
>
> Key: JCR-1885
> URL: https://issues.apache.org/jira/browse/JCR-1885
> Project: Jackrabbit
> Issue Type: New Feature
> Components: jackrabbit-core
> Reporter: Marcel Reutegger
> Priority: Minor
>
> Workspaces with large indexes may consume considerable heap memory. Lucene
> implements multi level skip lists for terms in the index. The first level of
> the skip list is kept in memory. This is usually not an issue, but when terms
> consist of long Strings the memory consumption increases drastically.
> Jackrabbit not just tokenizes string properties, but it also creates a single
> term, based on the complete string property value (needed for jcr:like
> function). These long terms are the reason for the increased memory
> consumption.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
