jackrabbit-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ingomar Otter <ingomar.ot...@valtech.de>
Subject Re: Question about event ordering
Date Tue, 07 Feb 2012 19:51:52 GMT
Hi Julian! :-)
In the meantime I did for once study the JCR spec and figured that the EventIterator passed
into onEvent() in itself is just a fine event bundle demarcation.
I just reversed the events inside and now I'm good to go.

>only needed and present in Journaled observation -- are you using that?
No.
But only because I discovered it too late ;-)

Cheers,
 Ingomar

Am 07.02.2012 um 19:16 schrieb Julian Reschke:

> On 2012-02-06 21:01, Ingomar Otter wrote:
>> Hello,
>> I am currently building a simple replication between two Jackrabbit instances. I've
started off using (synchronous) Observers/Events but I am facing some challenges due to
>> the (intresting?) event order.
>> E.g. I add a file (using webdav) I see the following events (in that order):
>>> PROPERTY_ADDED,2@4376:	/Somefile.png/jcr:content/jcr:data
>>> PROPERTY_ADDED,1@4377:	/Somefile.png/jcr:content/jcr:mimeType
>>> PROPERTY_ADDED,7@4378:	/Somefile.png/jcr:content/jcr:primaryType
>>> PROPERTY_ADDED,5@4379:	/Somefile.png/jcr:content/jcr:lastModified
>>> PROPERTY_ADDED,5@4380:	/Somefile.png/jcr:created
>>> NODE_ADDED@4381:			/Somefile.png/jcr:content
>>> PROPERTY_ADDED,1@4382:	/Somefile.png/jcr:createdBy
>>> PROPERTY_ADDED,7@4383:	/Somefile.png/jcr:primaryType
>>> NODE_ADDED@4384			:/Somefile.png
>> 
>> 
>> Of course I can not replay the events in that order as some parent nodes may be missing.
I could re-order them but then this would at least require some "unit of work" demarcation
as I operate on stream of events (and thus need to cluster/chunk) the events.
>> Is there a reason why the events are emitted in 'reverse' order (I would assume per
API the node was created first)?
>> Is there a way to influence this?
>> 
>> Afaik PERSIST (event bundling) is work-in-progress in unstable, would that be a candidate
as Unit-Of-Work marker?
> 
> In the meantime I realized that PERSIST events are not in 2.4.0 (but 
> probably could be ported), and also that you may not need them; they are 
> only needed and present in Journaled observation -- are you using that?
> 
> Best regards, Julian
> 



Mime
View raw message