[
https://issues.apache.org/jira/browse/JCR-2744?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jukka Zitting updated JCR-2744:
-------------------------------
Attachment: TransientManyChildNodesTest-2.png
I've now gotten rid of the separate Element objects. The AbstractPath class now
implements both the Path and Element interfaces (the latter properly only when
dealing with a single-element path).
I also improved the getElements() method to avoid the O(n^2) behaviour.
The attached plot shows the current performance improvements (about 20% on
large transient operations) that are mostly due to these changes. There's still
room for additional improvements, but I believe the biggest wins have already
been achieved here.
> Avoid element arrays in PathImpl
> --------------------------------
>
> Key: JCR-2744
> URL: https://issues.apache.org/jira/browse/JCR-2744
> Project: Jackrabbit Content Repository
> Issue Type: Improvement
> Components: jackrabbit-spi, jackrabbit-spi-commons
> Reporter: Jukka Zitting
> Attachments: TransientManyChildNodesTest-2.png,
> TransientManyChildNodesTest.png
>
>
> The path handling code in spi-commons shows quite often in thread dumps and
> profiling results, as the current implementation does quite a bit of
> repetitive allocating and copying of path element arrays. We should be able
> to streamline and simplify the path handling code by only tracking the latest
> path element and a reference to the parent path. To do this efficiently we
> may need to adjust some of the Path and PathFactory method declarations (that
> currently assume element array -based paths) also in the SPI.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
