[ 
https://issues.apache.org/jira/browse/SLING-4611?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Joel Richard updated SLING-4611:
--------------------------------
    Attachment: SLING-4611_experimental.patch

I have attached an experimental patch which improves the rendering performance 
in combination with the patch from SLING-4596 for 28%-39% (tested on two 
different pages). This patch depends on the other one and must not be applied 
alone.

In order to make the resourcemerger tests passing again, the patch from 
SLING-4626 has to be applied as well.

> Performance: Consider optimizing MergedResource#getParent and getChild
> ----------------------------------------------------------------------
>
>                 Key: SLING-4611
>                 URL: https://issues.apache.org/jira/browse/SLING-4611
>             Project: Sling
>          Issue Type: Improvement
>          Components: ResourceResolver
>    Affects Versions: Resource Merger 1.2.8
>            Reporter: Joel Richard
>            Priority: Critical
>              Labels: perfomance
>         Attachments: SLING-4611_experimental.patch
>
>
> For some pages up to 29% of the rendering time is spent in 
> AbstractResource.getChild. In my case, 75% of the resources are 
> MergedResources and the rest are JcrNodeResources (for which I have already 
> opened SLING-4596).
> Therefore, I would suggest implementing the following optimization:
> The merged resources should be stored in MergedResource and when getChild is 
> called, it would use getChild on them directly and merge the child resources. 
> This would have the advantage that the ParentHidingHandler would not have to 
> be called again for the parent resources and could even make SLING-4568 
> obsolete. Moreover, it would leverage optimizations of the merged resource 
> implementations (e.g. SLING-4596 for JcrNodeResource#getChild).
> A problem with this approach is that possible in JCR (and maybe also some 
> other resource providers) to set ACLs which allow to read children but not 
> their parents. Therefore, getChild will in addition have to check for each 
> search path which isn't covered with a merged resource if this path does 
> really not exist.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to