Mailing-List: contact user-help@cayenne.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@cayenne.apache.org
Received-SPF: pass (athena.apache.org: domain of marcin@ish.com.au designates
 59.167.240.32 as permitted sender)
Message-ID: <4EA8FE75.2000606@ish.com.au>
Date: Thu, 27 Oct 2011 16:47:17 +1000
From: Marcin Skladaniec <marcin@ish.com.au>
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.7;
 rv:6.0) Gecko/20110812 Thunderbird/6.0
MIME-Version: 1.0
To: user@cayenne.apache.org
Subject: Re: problem with ROP query performance
References: <4EA10F89.3040000@ish.com.au>
 <3A8423E9-C298-477B-A2E4-335E9E60CDE8@objectstyle.org>
In-Reply-To: <3A8423E9-C298-477B-A2E4-335E9E60CDE8@objectstyle.org>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit

I will Jira it, just will try to be more specific and include some code 
or maybe unit test.
Thanks
Marcin
PS: using cayenne 3.1m2

On 27/10/11 5:22 AM, Andrus Adamchik wrote:
> Yeah sounds odd. It fetches paintings of that artist. Maybe we can jira that and take a closer look?
>
> On a side note, I noticed that due to the weak references used in the DataContext to store committed objects, you may see lots of faulting of objects that were fetched in the same context just a few seconds ago. This is probably completely unrelated, and for that we are experimenting with using soft references (instead of weak) via a new ObjectMapRetainStrategy (available in 3.1M3).
>
> Andrus
>
>
> On Oct 21, 2011, at 9:22 AM, Marcin Skladaniec wrote:
>
>> Hi
>>
>> There is a discrepancy in behaviour of performQuery between ROP and non-ROP setup.
>>
>> lets consider this simple code:
>>
>>         DataContext newContext = ...
>>
>>         Artist artist = newContext.newObject(Artist.class);
>>         newContext.commitChanges();
>>
>>         Painting painting = newContext.newObject(Painting.class);
>>         painting.setArtist(artist);
>>
>>         assertEquals(artist.getPersistenceState(), PersistenceState.MODIFIED); // this is true
>>
>>         SelectQuery sq = new SelectQuery(Painting.class, ExpressionFactory.matchExp(Painting.ARTIST_PROPERTY, artist));
>>
>>         newContext.performQuery(sq);
>>
>>
>> I have added some logging to the performQuery() to check what happens. This is what I see in ROP setup:
>>
>> DEBUG - select query for Painting with qualifier: artist =<ObjectId:Artist, id=200>
>> DEBUG - RelationshipQuery:paintings for:<ObjectId:Artist, id=200>
>>
>> Seems straightforward, select query is executed... but then cayenne faults the Artist object and its relationships. This would be ok, but the Artist object is already in the context, no trip to the databse required!
>>
>> Same code in non-ROP setup does not behave teh same - the RelatioshipQuery is not executed (unless artist.getPaintings() is called explicitly).
>>
>> For us this Relationship Query is a big big performance problem.
>> Is that a bug in ROP? Is there a way to prevent faulting of the relatioship?
>>
>> With regards
>> Marcin
>>
>>