jackrabbit-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Stefan Guggisberg (JIRA)" <j...@apache.org>
Subject [jira] Commented: (JCR-1242) Improve serialization of NodeReferences for BundleDB PMs
Date Thu, 29 Nov 2007 11:03:43 GMT

    [ https://issues.apache.org/jira/browse/JCR-1242?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12546655

Stefan Guggisberg commented on JCR-1242:

> >- the serialization format used by ItemStateBinding is changed, but the
> >  changes are *not* backwards compatible (see read/writePropertyId(...)).
> read/writePropertyIds methods have not been ever used, there is no need to ensure backward

those methods are public so we can't really tell whether they're used or not.

> > - version of new serialization format is not incremented/persisted
> yes it is, only method affected is writeState(DataOutputStream out, NodeReferences state),
first byte contains new version (if new format wouldn't be persistent I couldn't use this

sorry, i can't follow you here. you changed ItemStateBinding's serialization format but you
didn't increment the version. that's IMO incorrect (and dangerous). 

> >- the proposed changes are IMO rather 'hacky'; for example
> >  ItemStateBinding#readState(..., NodeReferencesId, ...) assumes
> >  implementation details of Serializer#serialize(NodeReferences refs, ...)
> that's true, but to ensure 'backward compatibility' for bundle PMs readState method must
assume something :-), it assumes previous implementation, what do you suggest ?

ItemStateBinding now assumes  that Serializer had been used previously; 
even worse, it assumes implementation details of Serializer. that's IMO a 
rather nasty hack. 

a cleaner solution would e.g. be to extend Serializer, optimize the serialization format and
a version id in order to guarantee backwards compatibility.

> > double wrapping of DataOutputStream's (i guess that's just an oversight)
> yes, double wrapping is a oversight, I originally prepared the patch for 1.3 branch (it's
the version i'm currently using)

> Improve serialization of NodeReferences for BundleDB PMs
> --------------------------------------------------------
>                 Key: JCR-1242
>                 URL: https://issues.apache.org/jira/browse/JCR-1242
>             Project: Jackrabbit
>          Issue Type: Improvement
>          Components: jackrabbit-core
>    Affects Versions: 1.3.3, 1.4
>            Reporter: Przemo Pakulski
>            Priority: Minor
>             Fix For: 1.4
>         Attachments: JCR-1242.patch
> BudleDB PMs use currently Serializer class to serialize, deserialize node references.
Those methods are unefficient, it use string representation of UUID, namespaceURI and localName.

> For UUID rawBytes should be used and for namespaceURI, localName namespaceIndex/nameIndex
should be used to improve efficiency of serialization/deserialization.

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

View raw message