Home » Modeling » EMF » ChangeRecorder issue ?
|
| Re: ChangeRecorder issue ? [message #416396 is a reply to message #416394] |
Wed, 30 January 2008 11:12   |
 Ed Merks
Messages: 33142
Registered: July 2009 |
Senior Member |
|
|
This is a multi-part message in MIME format.
--------------020205050809080102080802
Content-Type: text/plain; charset=ISO-8859-15; format=flowed
Content-Transfer-Encoding: 7bit
Stephane,
Comments below.
Stephane wrote:
> Hi,
>
> I'm using the ChangeRecorder API and I have got a weird problem.
As opposed to those non-wierd problems. :-P
>
> I monitor an EMF object "Canvas" that has a containment reference
> named "figures" of type Figure .
>
> Let's imagine, one instance of this "Canvas" named canvas1 has 2
> objects contained in the "figures" feature.
>
> a) I start a change recorder for canvas1.
> b) I unset the "figures" containment feature.
> c) I end the change recorder and get the change description.
>
> Result : no change happens for canvas1 :
> ChangeDescription#getObjectChanges(),
> ChangeDescription#getObjectsToAttach or
> ChangeDescription#getObjectsToDetach are empty collections !! Weird
> isn'it ?
Hmm. Is it an unsettable multi-valued containment? Do you need that?
>
> When ending the change recorder, the
> BasicChangeRecorder#eliminateEmptyChanges() is called.
> I debug the code, and the featureChange is retrieved according the
> unset on the "figures" feature. Nevertheless, It compares the value
> registered into the featureChange with the current value of the
> related feature on canvas1. Both lists are empty ones and hence
> equals, the featureChange is removed... What's a pitty !!
>
> Is it a bug or normal behavior ? How can i figure out that issue ?
It doesn't sound good. I would expect the unset to fire a REMOVE_MANY
and for this code in the ChangeRecorder to kick in to record the old
value of the list.
case Notification.REMOVE_MANY:
{
if (change == null && changes != null)
{
@SuppressWarnings("unchecked") List<Object> removedValues =
(List<Object>)notification.getOldValue();
List<Object> oldValue = new
BasicEList<Object>((Collection<?>)eObject.eGet(feature));
int[] positions = (int[])notification.getNewValue();
if (positions == null)
{
oldValue.addAll(removedValues);
}
else
{
for (int i = 0; i < positions.length; ++i)
{
oldValue.add(positions[i], removedValues.get(i));
}
}
change = createFeatureChange(eObject, feature, oldValue,
notification.wasSet());
((InternalEList<FeatureChange>)changes).addUnique(change);
}
break;
}
Does a clear cause the same problems as an unset? Does the unset case
kick in?
case Notification.UNSET:
{
if (change == null && changes != null)
{
if (feature.isMany())
{
List<Object> oldValue = new
BasicEList<Object>((Collection<?>)eObject.eGet(feature));
int index = notification.getPosition();
if (index != Notification.NO_INDEX)
{
oldValue.set(index, notification.getOldValue());
}
change = createFeatureChange(eObject, feature, oldValue,
notification.wasSet());
}
else
{
Object oldValue = notification.getOldValue();
change = createFeatureChange(eObject, feature, oldValue,
notification.wasSet());
}
((InternalEList<FeatureChange>)changes).addUnique(change);
}
if (containment != null)
{
Object newValue = notification.getNewValue();
if (newValue != null && newValue != Boolean.TRUE &&
newValue != Boolean.FALSE)
{
addAdapter((Notifier)newValue);
}
}
break;
}
>
> Stephane.
>
--------------020205050809080102080802
Content-Type: text/html; charset=ISO-8859-15
Content-Transfer-Encoding: 8bit
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-15"
http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
Stephane,<br>
<br>
Comments below.<br>
<br>
Stephane wrote:
<blockquote
cite="mid:ea209def52072b9f557fa0aa79cf8d41$1@www.eclipse.org"
type="cite">Hi,
<br>
<br>
I'm using the ChangeRecorder API and I have got a weird problem.
<br>
</blockquote>
As opposed to those non-wierd problems.
Ed Merks
Professional Support: https://www.macromodeling.com/
|
|
|
| Re: ChangeRecorder issue ? [message #416404 is a reply to message #416396] |
Wed, 30 January 2008 13:59 |
Stephane fournier
Messages: 340
Registered: July 2009 |
Senior Member |
|
|
Ed,
Answers below.
I found the issue : my mistake. The change recorder is started at a
moment, where the EReference is still not loaded (i.e the list is null).
When my wizard is displayed, depending on the pages, data are lazy loaded
(i'm not using the EMF serialization). Hence, the change recorder detects
the EReference is loaded and then depending on end-user actions, the
EReference can be cleared. At the end, in the case where the EReference
were cleared, the state for the EReference does not change : stay empty.
It is too complex and risky to change the way data are loaded in my
application. Thus I create my own ChangeRecorder that extends the one
provided by EMF and I override the eliminateEmptyChanges method.
Thank you for your answers, it helps me to come up with ideas to figure
out my issue.
Ed Merks wrote:
> Stephane,
> Comments below.
> Stephane wrote:
>> Hi,
>>
>> I'm using the ChangeRecorder API and I have got a weird problem.
> As opposed to those non-wierd problems. :-P
>>
>> I monitor an EMF object "Canvas" that has a containment reference
>> named "figures" of type Figure .
>>
>> Let's imagine, one instance of this "Canvas" named canvas1 has 2
>> objects contained in the "figures" feature.
>>
>> a) I start a change recorder for canvas1.
>> b) I unset the "figures" containment feature.
>> c) I end the change recorder and get the change description.
>>
>> Result : no change happens for canvas1 :
>> ChangeDescription#getObjectChanges(),
>> ChangeDescription#getObjectsToAttach or
>> ChangeDescription#getObjectsToDetach are empty collections !! Weird
>> isn'it ?
> Hmm. Is it an unsettable multi-valued containment? Do you need that?
>>
The "figures" EReference is not unsettable. When I called
eUnset(EReference), it does the same thing as the method clear on the
list.
>> When ending the change recorder, the
>> BasicChangeRecorder#eliminateEmptyChanges() is called.
>> I debug the code, and the featureChange is retrieved according the
>> unset on the "figures" feature. Nevertheless, It compares the value
>> registered into the featureChange with the current value of the
>> related feature on canvas1. Both lists are empty ones and hence
>> equals, the featureChange is removed... What's a pitty !!
>>
>> Is it a bug or normal behavior ? How can i figure out that issue ?
> It doesn't sound good. I would expect the unset to fire a REMOVE_MANY
> and for this code in the ChangeRecorder to kick in to record the old
> value of the list.
The notification with the eUnset or clear() does not trigger a REMOVE_MANY
event but a REMOVE one. (I'm using EMF 2.3.1) The
ChangeRecorder#notifyChanged delegates to handleFeature in case of EObject
&& EReference.
> case Notification.REMOVE_MANY:
> {
> if (change == null && changes != null)
> {
> @SuppressWarnings("unchecked") List<Object> removedValues =
> (List<Object>)notification.getOldValue();
> List<Object> oldValue = new
> BasicEList<Object>((Collection<?>)eObject.eGet(feature));
> int[] positions = (int[])notification.getNewValue();
> if (positions == null)
> {
> oldValue.addAll(removedValues);
> }
> else
> {
> for (int i = 0; i < positions.length; ++i)
> {
> oldValue.add(positions[i], removedValues.get(i));
> }
> }
> change = createFeatureChange(eObject, feature, oldValue,
> notification.wasSet());
> ((InternalEList<FeatureChange>)changes).addUnique(change);
> }
> break;
> }
> Does a clear cause the same problems as an unset? Does the unset case
> kick in?
A clear does the same problem. The clear or unset does not kick in the
following case.
> case Notification.UNSET:
> {
> if (change == null && changes != null)
> {
> if (feature.isMany())
> {
> List<Object> oldValue = new
> BasicEList<Object>((Collection<?>)eObject.eGet(feature));
> int index = notification.getPosition();
> if (index != Notification.NO_INDEX)
> {
> oldValue.set(index, notification.getOldValue());
> }
> change = createFeatureChange(eObject, feature, oldValue,
> notification.wasSet());
> }
> else
> {
> Object oldValue = notification.getOldValue();
> change = createFeatureChange(eObject, feature, oldValue,
> notification.wasSet());
> }
> ((InternalEList<FeatureChange>)changes).addUnique(change);
> }
> if (containment != null)
> {
> Object newValue = notification.getNewValue();
> if (newValue != null && newValue != Boolean.TRUE &&
> newValue != Boolean.FALSE)
> {
> addAdapter((Notifier)newValue);
> }
> }
> break;
> }
>>
>> Stephane.
>>
|
|
|
Goto Forum:
Current Time: Fri Apr 26 15:52:40 GMT 2024
Powered by FUDForum. Page generated in 0.03478 seconds |
] 
Search
Help
Register
Login
Home
