I think they are wicket-specific, because where outside of wicket would
you need a wicket interface?
And if you implement it yourself you will have to think of versioning,
otherwise you will have to think of versioning everytime you just use it.
For now I'd say have for POJO IFilterLocator a
rememberVersion(IStateRecorder) method (but name it different) Than it is
explicit what is going on and it is easier to use. (The same could be done
for ISortState). (Again IFilterLocator should take a Serializable and
maybe Ojbects.clone() could check if the cloneable implements Cloneable)
Generally I think IModels which need versioning and the underlying object
does not do it itself (which POJOs) should implement an interface and this
should be as:
IPOJOVersionable{
rememberVersion(IStateRecorder recorder);
}
The user than has to call this method on the Model.
I think this is better because:
a) it is more explicit. which ordinary user - like me knows - what is
going on in modelChanging()
b) currently only the model gets cloned which for pojos only has an effect
if you use a model with a non transient reference
c) you can not share the same underlying POJO instance between different
components because after the roleback you will end with two instances
(transparently)
d) I often use more than one POJO ie in a Form so if I need versioning I
want to version them both without the need of an extra component for
registering the Pojo.
Christian
On Tue, 6 Dec 2005 15:36:55 -0800, Igor Vaynberg <[EMAIL PROTECTED]>
wrote:
I guess the question is do we want to infact make it something
wicket-specific. right now its just an ordinary pojo you can use anywhere
you like. but if we roll the versioning logic into it, it really does
become
locked to wicket. that is why i struggled to find a way to keep
versioning
out of it - it allows users to work with simple pojos.
by keeping it an pojo there is another advantage - it is easier to expand
on. For example IFilterLocator works with ordinary objects that are used
as
a model for the filter form. everytime you write a filter model
(normally a
simple bean) do you want to deal with versioning everytime or would you
rather let the filter toolbar do it for you?
by keeping the versioning out of models we shift the versioning into
toolbars. most users will use toolbars and not create them so i think its
better to keep stuff like that out and let the users who create toolbars
know that they should version the model they are provided if they want
the
backbutton to work - but that is not any different then creating any
other
custom component that modifies data - so this isnt even anything special
about toolbars.
-Igor
On 12/6/05, Christian Essl <[EMAIL PROTECTED]> wrote:
On Tue, 6 Dec 2005 12:25:23 -0800, Igor Vaynberg
<[EMAIL PROTECTED]>
wrote:
> So to summarize your idea it would be like this
>
> class SingleSortState
> IChangeRecorder recorder;
> SortState(IChangeRecorder) {...}
>
> setPropertyState(...) {
> recorder.addChange( new Change() {...} )
> }
>
> OrderByLink { onclick () { getState().setProperty(...); } }
>
> this has the same net affect as
>
> OrderByLink { onclick() { addstatechange(new Change() { ...
> clone(getState()); }); getState().setProperty(...); } }
>
Right.
>
> Your way shifts a lot more responsibility into the model because it
> needs to
> know how to version itself, and most models in wicket need to do that
to
> support the back button. That means you can no longer user your pojo
> models
> as wicket models. for example i can no longer reuse my Person pojo
from
> the
> domain layer because it has no concept of versioning. am i totally off
> here?
> Thats the advantage of doing it on component side - your models can
stay
> dumb pojos and the component takes care of versioning them by using
the
> only
> fool-proof way it can - cloning.
>
> -Igor
You are right I definately went too far. ModelChanging triggers a
versioning clone - did not know that. Propably I did not see it because
I
mainly use my Pojos from the DB and I do not version them. Thanks for
the
advice I now understand versioning a bit better.
But back to the SortState: I think for such Wicket only models it is
still
more convinient to do it directly inside the model. In the end you will
have to 'version' it somehow and while you are right that the only entry
is through a component there can be many entry-points. And than it is a
good candidate to refactor it in one place - why not the model?
Christian
--
Christian Essl
___________________________________________________________
Gesendet von Yahoo! Mail - Jetzt mit 1GB Speicher kostenlos - Hier
anmelden: http://mail.yahoo.de
-------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc. Do you grep through log
files
for problems? Stop! Download the new AJAX search engine that makes
searching your log files as easy as surfing the web. DOWNLOAD SPLUNK!
http://ads.osdn.com/?ad_id=7637&alloc_id=16865&op=click
_______________________________________________
Wicket-user mailing list
Wicket-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wicket-user
--
Christian Essl
___________________________________________________________
Gesendet von Yahoo! Mail - Jetzt mit 1GB Speicher kostenlos - Hier anmelden: http://mail.yahoo.de
-------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc. Do you grep through log files
for problems? Stop! Download the new AJAX search engine that makes
searching your log files as easy as surfing the web. DOWNLOAD SPLUNK!
http://ads.osdn.com/?ad_id=7637&alloc_id=16865&op=click
_______________________________________________
Wicket-user mailing list
Wicket-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wicket-user
