Hi,
I understand my post wasn't very clear. I was going to explain it further but I have found the cause of my problems.
DataRowStore has a snapshots hashmap, keyed on object id. It had 'duplicate' objectIds in it with Short and String values for pk.
One coming from SelectQuery (with correct Short skingleKeyValue), and other from DataObjectUtils.objectForPK, where i was mistakenly passing a string (in an Object variable so i didn't notice a few times).
Is it possible to prevent creating ObjectIDs with singleKeyValue of wrong data type? Does anyone depend on this behavior?
Damir
----- Original Message ----
From: Andrus Adamchik <andru..bjectstyle.org>
To: cayenne-user@incubator.apache.org
Sent: Thursday, 9 November, 2006 11:24:38 PM
Subject: Re: Nested data context problems
I don't think I fully follow the sequence of events here, so my
comments are a bit random...
> 3. I commit it on the first page, so it goes to the database -
> here i can see sql updates but also indication of event i think is
> used for cache synchronisation
> DataRowStore [DEBUG] postSnapshotsChangeEvent: [SnapshotEvent]
> source: org.objectstyle.cayenne.access.DataRowStor..a8f43,
> modified 1 id(s)
This message is only related to synching between peer DataContexts at
the top level, so it is benign.
> 4. I go to the second page and it gets rendered with current
> (modified) version of dataobjects
> - when i debug converting of dataobjects to strings i notice
> that my dataobject is in the same, new, state as i would expect, i
> get the expected data with
> DataObjectUtils.objectForPK(object.getDataContext(),
> "Object", DataObjectUtils.pkForObject(object))
This looks suspicious. No objects should be in the "new" state after
commit.
Andrus
On Nov 8, 2006, at 8:53 AM, Damir Bijuklic wrote:
> Hi,
>
> I'm playing around with nested data context's and tapestry.
> Using tapestry 4.1.1 and cayenne 1.2.1.
>
> I have encountered an issue with cayenne that i don't quite
> understand. I will describe my setup and what I'm trying to do.
>
> I'm trying to use nested data context's on subforms/subpages.
> This way i can discard changes inside subforms although my
> dataobject that is rendered on the form is updated (i either call
> commitChangesToParent on nested data context, or i don't).
>
> I'm also using tapestry data squeezer, which is mechanism in
> tapestry for converting from data object to string and vice versa.
>
> DataSqueezer has two methods that look similar to this: String
> squeeze(Object) and Object unsqueeze(String) .
>
> In testing i have encountered an issue which i have tried to
> reproduce with test case, but have not been able to do so, so far.
>
> When i navigate between three pages, each having it's own
> datacontext, where top page uses session bound datacontext and
> subpages use ones derived with
> createChildDataContext (so third page has datacontext that derives
> from derived datacontext from second page).
>
> 1. I go to the third page,change some data and commit it
> 2. I commit it on the second page also
> 3. I commit it on the first page, so it goes to the database -
> here i can see sql updates but also indication of event i think is
> used for cache synchronisation
> DataRowStore [DEBUG] postSnapshotsChangeEvent: [SnapshotEvent]
> source: org.objectstyle.cayenne.access.DataRowStor..a8f43,
> modified 1 id(s)
> 4. I go to the second page and it gets rendered with current
> (modified) version of dataobjects
> - when i debug converting of dataobjects to strings i notice
> that my dataobject is in the same, new, state as i would expect, i
> get the expected data with
> DataObjectUtils.objectForPK(object.getDataContext(),
> "Object", DataObjectUtils.pkForObject(object))
> - it gets converted to something like "Object:200" where Object
> is entity name and 200 is PK
> 5. during the next request i try to parse string back into the
> object, here the funny stuff happens:
> - i parse "Object:200" and use DataObjectUtils.objectForPk
> (dataContext, "Object", 200) to fetch the object
> - i have checked few times that the datacontext i use here is
> the same as one above
> - the object i receive is NOT the one i have just rendered to
> the string, it has it's properties changed to what they were before
> step 1
> - during this step i notice message in tho log
> DataRowStore [DEBUG] postSnapshotsChangeEvent: [SnapshotEvent]
> source: org.objectstyle.cayenne.access.DataRowStor..a8f43,
> modified 1 id(s)
> - tracing objectForPK shows that it reads the old values from
> the cache though they should not be there and it seems they weret
> there a few moments ago??
>
> I'm using shared cache, single jvm. What else could be wrong. If i
> restart the app to reload the caches it's all ok. Also i would like
> to load this object from the cache because it should have correct
> values in it.
>
> Any ideas what I'm missing? Any further info i could post?
>
> Damir
Send instant messages to your online friends http://uk.messenger.yahoo.com
This archive was generated by hypermail 2.0.0 : Sun Nov 12 2006 - 07:58:06 EST