I figured that would fix it, but I also figured that was in there for a reason
and I didn't want to change something I didn't completely understand. That's
mainly why I asked why it had to be that way.
I'm the one that has the patch for deep-prefetching via joins, so I don't
really care, because that usually works better for me. I just wanted to bring
the issue up in case anyone else had the same problem. Thanks for the
comments.
John
>===== Original Message From Jakob Braeuchi <[EMAIL PROTECTED]> =====
>hi john,
>
>have you tried disabling the collection-desriptor tweaking in
>RelationshipPrefetcherImpl#prepareRelationshipSettings() ?
>
>public void prepareRelationshipSettings()
>{
>setCascadeRetrieve(getObjectReferenceDescriptor().getCascadeRetrieve());
>getObjectReferenceDescriptor().setCascadeRetrieve(false); // comment it
>}
>
>
>jakob
>
>John wrote:
>> I sensed that, but I'm not sure the penalty for loading an object twice is
is
>> bad as the penalty I was getting. A lot of places in our code assume the
>> object reference is loaded on a required relationship. One thread was
>> prefetching and turned off auto-retrieve, then another came in and thought
>> that was the right value and set it back to that. So from then on none of
the
>> objects had the relationship loaded and I got NullPointerException
everywhere.
>> I suppose that multiple loads could be more of an issue with the global
>> cache, but with cache-per-broker I'm not sure if that's an issue. At least
to
>> me double-load isn't as big a deal as the NPE's. I had to discontinue
using
>> prefetch in the couple of places I was using it. (Not sure that's the
worst
>> thing anyway, since I don't know that it helped a whole lot.)
>>
>> John
>>
>>
>>>===== Original Message From Jakob Braeuchi <[EMAIL PROTECTED]> =====
>>>hi john,
>>>
>>>
>>>John wrote:
>>>
>>>
>>>Has there been any solution to this issue (OJB188)? This bit me in the
>>
>> butt,
>>
>>>but of course it took quite a bit of digging and debugging to realize this
>>
>> is
>>
>>>what was happening. Why exactly is the value changed while prefetching?
>>>
>>>
>>>auto-retrieve is disabled during prefetching of a relationship to avoid
>>>loading the same obj multiple times.
>>>
>>>jakob
>>>
>>>
>>>I have a patched version of OJB that was based on HEAD from the middle of
>>>August, so I haven't been able to update for a while (don't want to
>>
>> repatch).
>>
>>>It seems that there have been some changes to OJB since then with regards
>>
>> to
>>
>>>(proxy) prefetching, autoretrieve, etc, especially in how those are
>>
>> configured
>>
>>>in the repository. Are these summarized anywhere?
>>>
>>>John Marshall
>>>Connectria
>>>
>>>
>>>=============================================
>>>Date: Fri, 11 Jul 2003 18:19:33 +0200
>>>From: Jakob Braeuchi <[EMAIL PROTECTED]>
>>>Subject: Potential problem with prefetch-relationships ?
>>>Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>>>
>>>
>>>hi theo,
>>>
>>>during prefetch auto-retrieve is disabled. when an other thread uses the
>>>relationship-desriptor it will find auto-retrive off.
>>>this is a known problem but i do not have a solution for it :(
>>>
>>>
>>>jakob
>>>
>>>Theo Niemeijer wrote:
>>>
>>>
>>>
>>>>Hi all,
>>>>I seemed to have stumbled on a potential problem with
>>>>the "prefetch relationship" option in PB query.
>>>>
>>>>After using prefetchRelationship for retrieval of big list of results,
>>>>and at the same time performing other queries,
>>>>I seemed to have lost the "auto-retrieve" attribute on
>>>>the collection descriptor.
>>>>
>>>>The result was that subsequent queries did not
>>>>retrieve that collection anymore, probably because the
>>>>"auto-retrieve" was disabled.
>>>>
>>>>The problem is quite hard to reproduce, but my guess is
>>>>that different threads modified the repository descriptor
>>>>in the wrong sequence, by means of the setCascadeRetrieve
>>>>method used in prepareRelationshipSettings.
>>>>
>>>>I might be wrong, because I simply do not oversee all aspects
>>>>of OJB. But am I right to view this as a potential problem
>>>>in "concurrency situations" like websites ?
>>>>I do not feel too comfortable that OJB makes these "temporary"
>>>>changes to the repository model.
>>>>
>>>>Cheers,
>>>> Theo Niemeijer
>>>
>>>
>>>
>>>---------------------------------------------------------------------
>>>To unsubscribe, e-mail: [EMAIL PROTECTED]
>>>For additional commands, e-mail: [EMAIL PROTECTED]
>>>
>>>
>>>
>>>
>>>---------------------------------------------------------------------
>>>To unsubscribe, e-mail: [EMAIL PROTECTED]
>>>For additional commands, e-mail: [EMAIL PROTECTED]
>>
>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: [EMAIL PROTECTED]
>> For additional commands, e-mail: [EMAIL PROTECTED]
>>
>>
>
>
>---------------------------------------------------------------------
>To unsubscribe, e-mail: [EMAIL PROTECTED]
>For additional commands, e-mail: [EMAIL PROTECTED]
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
