You are viewing a plain text version of this content. The canonical link for it is here.
Posted to oak-issues@jackrabbit.apache.org by "Chetan Mehrotra (JIRA)" <ji...@apache.org> on 2017/05/02 08:50:04 UTC
[jira] [Comment Edited] (OAK-4637) Property index: include/exclude
key pattern list
[ https://issues.apache.org/jira/browse/OAK-4637?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15992553#comment-15992553 ]
Chetan Mehrotra edited comment on OAK-4637 at 5/2/17 8:49 AM:
--------------------------------------------------------------
The usecase I have in mind is the one related to slingResourceType. In Sling the "sling:resourceType" property is added to each node created via Sling api. Due to this even the nodes created for Sling Job Management also have sling:resourceType=slingevent:Job or sling:Folder causing such nodes to be indexed.
However no code does a query for these values. So we need a way to exclude such resource type values from getting indexed i.e. way to provide a list of values which should not be indexed
Also note that there may be query like below which should still be able to use the index on sling:resourceType.
{noformat}
/jcr:root/etc/replication//*[jcr:like(@sling:resourceType, "cq/replication/components/%agent")]
{noformat}
From my current understanding of code in property index such queries gets converted to property != null (untill OAK-5897 is implemented). The current implementation (rev 1793097) looks like would opt out of handling such a query. But we need a way such that index still gets used for these queries.
{noformat}
if (valuePattern.matchesAll()) {
// matches all values: not a problem
} else if (values == null) {
// "is not null" condition, but we have a value pattern
// that doesn't match everything
// so we can't use that index
continue;
}
{noformat}
was (Author: chetanm):
The usecase I have in mind is the one related to slingResourceType. In Sling the "sling:resourceType" property is added to each node created via Sling api. Due to this even the nodes created for Sling Job Management also have sling:resourceType=slingevent:Job or sling:Folder causing such nodes to be indexed.
However no code does a query for these values. So we need a way to exclude such resource type values from getting indexed.
Do note that there may be query like below which should still be able to use the index on sling:resourceType.
{noformat}
/jcr:root/etc/replication//*[jcr:like(@sling:resourceType, "cq/replication/components/%agent")]
{noformat}
From my current understanding of code in property index such queries gets converted to property != null (untill OAK-5897 is implemented). The current implementation (rev 1793097) looks like would opt out of handling such a query. But we need a way such that index still gets used.
{noformat}
if (valuePattern.matchesAll()) {
// matches all values: not a problem
} else if (values == null) {
// "is not null" condition, but we have a value pattern
// that doesn't match everything
// so we can't use that index
continue;
}
{noformat}
> Property index: include/exclude key pattern list
> ------------------------------------------------
>
> Key: OAK-4637
> URL: https://issues.apache.org/jira/browse/OAK-4637
> Project: Jackrabbit Oak
> Issue Type: Improvement
> Components: property-index
> Reporter: Thomas Mueller
> Assignee: Thomas Mueller
> Priority: Critical
> Fix For: 1.8
>
>
> In some cases, property indexes contain many nodes, and updating them can be slow. Right now we have filters for node and mixin types, path (include and exclude).
> An include and exclude list of values (patterns) would be useful. For example the property "status", if we only ever run queries with the condition "status = 'ACTIVE'", then nodes with status INACTIVE and DONE don't need to be indexed.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)