Mailing-List: contact user-help@cayenne.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@cayenne.apache.org
Received-SPF: pass (nike.apache.org: local policy)
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0 (Apple Message framework v1084)
Subject: Re: Getting null instead of the actual value
From: Andrus Adamchik <andrus@objectstyle.org>
In-Reply-To: 
 <CAPFNckhsUM1a+1zS12z1JOxg9tz2f0xSnq4SLtHO2TSJkO0L_w@mail.gmail.com>
Date: Fri, 16 Sep 2011 22:16:40 +0300
Content-Transfer-Encoding: quoted-printable
Message-Id: <4654784B-B27F-46A0-853D-2931FFA306AD@objectstyle.org>
References: 
 <CAPFNckicWsvKb8JME4XhL8YemvZ+h7jzNj4MdbFEs7N+9LLAgw@mail.gmail.com>
 <CAA-upzqzQDKEj9K2iv8V-3Kea+q7m-Rk3V2RgjirQwA2vfH8rw@mail.gmail.com>
 <CAPFNckgzc0vX-VaAzAZV9s4o4M6hzR8DzX1F8tdazJb-y1X=kQ@mail.gmail.com>
 <480D6A08-A61C-4F8A-B72D-BF8175A1A1E4@objectstyle.org>
 <CAPFNckg=Eezu8Va6PSmkgnmzmDMM+h6cnpw-pDw1GZnHFK-1dA@mail.gmail.com>
 <4F560F6C-51CD-4B38-83E1-AE4178E0816B@objectstyle.org>
 <CAPFNckgujQQ_sLmZ5GPRQwXzDQtDnYOZtrSCXnbTaNTpAPh_WQ@mail.gmail.com>
 <CAM1yOjYo65uuL-_kXr9kV+AQrW_TE4-j8_i6bNDFZQRSzjenhA@mail.gmail.com>
 <5A179745-EE5A-486C-994E-6EC4D6CA25C3@objectstyle.org>
 <CAA-upzoBWc1hfrJ=N6DN0rJvC-xYrj3dNVKf59uU6J+snNJJ9A@mail.gmail.com>
 <CAPFNckhsUM1a+1zS12z1JOxg9tz2f0xSnq4SLtHO2TSJkO0L_w@mail.gmail.com>
To: user@cayenne.apache.org

One side of prefetching is saving us uneeded queries, when we know =
upfront which related objects we need in addition to the main query =
result. This is the angle shown in the docs where prefetching is =
presented as a performance optimization tool.=20

But there is another side of it, which is just as important. It is about =
fine-grained control of refreshing parts of the object graph already in =
memory. When a query is run, Cayenne can't automatically refresh all =
objects directly or indirectly connected to the query result objects, as =
it would cause a massive fetch. With prefetching Cayenne knows how much =
of the object graph we want to refresh.

So I don't find the result all that surprising. (Although if things =
happen in the same JVM I am still surprised the relationship isn't =
refreshed via Cayenne cross-context event mechanism. This could be a =
race condition actually, as events are delivered asynchronously)

Andrus=20


On Sep 16, 2011, at 10:01 PM, Christian Grobmeier wrote:

> I have not had duplication issues so far. But I will double check now.
> I added this code b/c of the problems with the null values, so it isnt
> the root cause for my problem.
> At the moment it still looks good. It seems that the prefetch helped.
> I am very puzzled about it :-)
>=20
> On Fri, Sep 16, 2011 at 8:30 PM, Michael Gentry =
<mgentry@masslight.net> wrote:
>> I seem to recall doing a setX() followed by an addToY() causes a
>> duplicate Y to be added.  I'd have to double check this, but that is
>> my recollection.
>>=20
>> On Fri, Sep 16, 2011 at 2:19 PM, Andrus Adamchik =
<andrus@objectstyle.org> wrote:
>>>> but may cause duplication issues for addToY().
>>>=20
>>> I hope it doesn't. But since I am trained to just set the rel in one =
direction, I may not be aware of the issue.
>>>=20
>>>=20
>>> On Sep 16, 2011, at 9:12 PM, Mike Kienenberger wrote:
>>>=20
>>>> On Fri, Sep 16, 2011 at 4:22 AM, Christian Grobmeier
>>>> <grobmeier@gmail.com> wrote:
>>>>> userRoles.setUser(user);
>>>>> userRoles.setRole(role);
>>>>> user.addToUserRoles(userRoles);
>>>>> role.addToRolesToJoin(userRoles);
>>>>=20
>>>> Again, unless, you've specifically made both relationships
>>>> unidirectional (the default is bidirectional), you are actually
>>>> setting both sides of the relationship twice.   Not an issue for a
>>>> SetX(), but may cause duplication issues for addToY().
>>>>=20
>>>=20
>>>=20
>>=20
>=20
>=20
>=20
> --=20
> http://www.grobmeier.de
>=20