The issue with the SQL-OS is that each data type needs its own custom "valuie 
mapper" that handles read/writes from/to the database tables.

The native Joda classes would need to be implemented before they can be 
persisted...

Regards,
Kevin

Jeroen van der Wal <[email protected]> wrote:

>First of all there's no intention to break other components by using
>Joda
>natively. We've added support for Joda in the JDO object store already
>and
>I recommend that other object stores and viewers will start supporting
>Joda
>types too, it shouldn't be hard to accomplish that. Marking the applib
>date/time classes being deprecated will hint developers to use Joda
>over
>applib but will not break anything. It's up to the community to do a
>vote
>on the final removal of the classes but there's no rush to do that.
>
>Your quick fix to expose the underlying Joda classes would IMHO make
>things
>more complicated than it should be. One of the drivers to move to Joda
>is
>that in the current application we're building we have to do a lot of
>time
>and date based calculations and Joda excels in that.
>
>A little refactoring should not stop the improvement of the framework
>but
>perhaps I'm missing some more relevant.
>
>Kind Regards,
>
>Jeroen
>
>On Tue, Sep 18, 2012 at 7:05 PM, Kevin Meyer - KMZ <[email protected]>
>wrote:
>
>> An immediate quick fix would be to provide a getter on the applib
>date
>> classes that exposes the underlying Joda class... then you could get
>> the full access you want, without affecting any other component (e.g.
>> the Isis sql object store ;-) )
>>
>> Regards,
>> Kevin
>>
>> On 18 Sep 2012 at 16:50, Jeroen van der Wal wrote:
>>
>> > The reason that I use Joda is because I want to access it full
>> capabilities.
>> >
>> > +1 to mark the applib date/times as deprecated and leave it around
>for a
>> > while. I'm volunteering to do some refactoring of Isis.
>> >
>> > Regards,
>> >
>> > Jeroen
>> >
>> >
>> >
>> >
>> > On Tue, Sep 18, 2012 at 4:20 PM, Dan Haywood
>> > <[email protected]>wrote:
>> >
>> > > I wonder if we should just get rid of the Isis applib date/times?
>>  It's not
>> > > like our implementation is in any way superior to Joda (rather:
>it is
>> > > substantially inferior).
>> > >
>> > >
>> > >
>> > > On 18 September 2012 14:54, Kevin Meyer - KMZ <[email protected]>
>wrote:
>> > >
>> > > > Hi all,
>> > > >
>> > > > This probably does not have any direct bearing on this
>discussion,
>> but
>> > > > I have converted all Isis date / time variables to (at their
>core)
>> use
>> > > > Joda time instead of Java time classes... the core Joda time
>property
>> > > > might not be exposed, though.
>> > > >
>> > > > Regards,
>> > > > Kevin
>> > > >
>> > > > On 18 Sep 2012 at 15:38, Jeroen van der Wal wrote:
>> > > >
>> > > > > Hi Minto,
>> > > > >
>> > > > > 1. UUID is not persisted by default but will be if you mark
>it with
>> > > > > @Persistent. Don't know from the top of my head if that's
>also the
>> case
>> > > > for
>> > > > > the applib.DateTime. As an alternative you can use
>> > > org.joda.time.DateTime
>> > > > > (which also must be annotated with @Persistent). See the
>> Datanucleus
>> > > > > documentation on Java types[1] for more information.
>> > > > >
>> > > > > 2. Collections should be marked with
>> > > > > @Persistent(mappedBy="<parent class>") for a 1-N
>relationship. Add
>> an
>> > > > > additional @Join to create a separate join table. Again,
>> Datanucleus
>> > > > > provides loads of documentation on this topic [2].
>> > > > >
>> > > > > Hope this helps.
>> > > > >
>> > > > > Jeroen
>> > > > >
>> > > > > [1]
>http://www.datanucleus.org/products/datanucleus/jdo/types.html
>> > > > > [2]
>> > > > >
>> > > >
>> > >
>>
>http://www.datanucleus.org/products/datanucleus/jdo/orm/one_to_many_set.html
>> > > > >
>> > > > >
>> > > > >
>> > > > >
>> > > > >
>> > > > > --
>> > > > > Jeroen van der Wal
>> > > > > Stromboli b.v.
>> > > > > +31 655 874050
>>
>>

--
Sent from my Android phone with K-9 Mail. Please excuse my brevity.

Reply via email to