[news.eclipse.modeling.gmf] Re: What is the advantage of use AbstractTra

[sys2002@xxxxxxxxx (Yury)]

Hi Christian,

Thanks for your reply. I still have some questions.

> The WorkspaceCommandStackImpl implements a single, linear undo stack using
> the operation history.  It is intended to provide the same behaviour as the
> EMF CommandStack.

> However, using AbstractEMFOperations (or GMF's
> AbstractTransactionalCommands) with an operation history (not using the
> WorkspaceCommandStackImpl), you can implement more interesting undo models. 
> For example, you can have independent undo/redo stacks for different
> Resources in the same ResourceSet, if these resources are not connected. 
> Using EMF's CommandStack, you would actually need a separate ResourceSet
> and EditingDomain for each resource.
At present I need single execution stack for my application. 
IOperationHistory and OperationHistoryFactory suits fine for this. But 
working with diagram editor in general I do not know what kind of 
CommandStack it is using and if that stack is backed by global 
IOperationHistory or not. At the same time I need to show some feedback 
when the command stack changes. Currently my application listens 
CommandStack that can be retrieved from the EditingDomain of 
DiagramEditor. If all modifications are performed using the CommandStack 
then I always know when to show feedback. If I execute a sort of 
AbstractEMFOperation in the global operation history the CommandStack does 
not trigger any events. If I understand correctly you suggest to support 
custom mapping from Resource to IOperationHistory instance (like the 
mapping Resource->ResourceSet->EditingDomain->CommandStack). In my case it 
is trivial:
Resource->global IOperationHistory. Everyone who modifies my model should use 
global history and feedback provider would listen to the IOperationHistory not 
to CommandStack. Am I right?

> To answer your first question:  in a GMF application, you should probably
> never execute commands in a CommandStack, because it will not be worth the
> effort of coordinating the IUndoContexts applied to these commands and your
> GMF AbstractTransactionalCommands to ensure that the Undo/Redo menus make
> sense.
So if I want to implement a custom command which makes some changes with 
my EMF model I have to create a command inherited form 
AbstractEMFOperation and put all logic within it. Afterwards when I 
execute it in IOperationHistory I will get Undo/Redo functionality for 
free. Is it true?

Yury