[jira] [Comment Edited] (OAK-4320) Use the cache tracker in the RDB Document Store

["Julian Reschke (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/OAK-4320?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15266324#comment-15266324 ] 

Julian Reschke edited comment on OAK-4320 at 5/2/16 12:14 PM:
--------------------------------------------------------------

I had another look at the change, and it removes an optimization that I think was useful; not re-parsing a document from the DB when we already have it in the case (this was in {{runThroughCache()}}. Was that change intentional (in that you think the optimization isn't worth the complexity?)?

I believe the methods in {{NodeDocumentCache}} should accept something like:
{noformat}
    public interface NodeDocumentProvider {
        public String getId();
        public NodeDocument getDocument();
        public NodeDocument getDocumentIfNewer(long modCount);
    }
{noformat}

This would allow the DS implementations to return just a thin wrapper around whatver is persisted, and to re-build the document only when it's needed.


was (Author: reschke):
I had another look at the change, and it removes an optimization that I think was useful; not re-parsing a document from the DB when we already have it in the case (this was in {{runThroughCache()}}. Was that change intentional (in that you think the optimization isn't worth the complexity?)?

> Use the cache tracker in the RDB Document Store
> -----------------------------------------------
>
>                 Key: OAK-4320
>                 URL: https://issues.apache.org/jira/browse/OAK-4320
>             Project: Jackrabbit Oak
>          Issue Type: Improvement
>          Components: documentmk, rdbmk
>            Reporter: Tomek Rękawek
>            Assignee: Tomek Rękawek
>             Fix For: 1.6
>
>         Attachments: OAK-4320.patch
>
>
> OAK-4112 introduced {{CacheChangesTracker}} mechanism, inspired by the {{QueryContext}} class, already present in the RDB DS.
> The tracker works on the {{NodeDocumentCache}} level, within the methods modifying the cache values, using the same {{NodeDocumentLocks}}, which may prevent potential concurrency problems, described in the comments to OAK-3566.
> We should synchronise both approaches, so the queries in Mongo and RDB uses the same logic to update their caches.



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