[
https://issues.apache.org/jira/browse/OAK-2464?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14298368#comment-14298368
]
Vikas Saurabh commented on OAK-2464:
------------------------------------
[~chetanm], I just realized that I was mistaken -- comparator incompatibility
would actually block optimizing look up of children from list. The issue is
that if we get cached element and not find the element (the case where
incompatible comparator might go wrong), then the logic would deduce that the
queried child _does not exist_. We won't be falling back to doc store :(.
Also, I think it'd incorrect to use some implementation of sorted list (using
java comparator) to cache children because cached children are sometimes not
completely filled and the ordering is defined by doc store. So, a refill from
doc store might put a newly pulled child above already (half) cached list of
children -- I think that'd be an unacceptable behavior change.
> Optimize read of known non-existing children
> --------------------------------------------
>
> Key: OAK-2464
> URL: https://issues.apache.org/jira/browse/OAK-2464
> Project: Jackrabbit Oak
> Issue Type: Improvement
> Reporter: Vikas Saurabh
> Priority: Minor
> Attachments:
> 0001-OAK-2464-Optimize-read-of-known-non-existing-childre.patch
>
>
> Reading a non-existing node always queries down to {{DocumentStore}}.
> {{DocumentNodeStore}} should consider {{nodeChildrenCache}} to see if it
> already knows about non-existence and avoid a look up into {{DocumentStore}}.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
