The transaction framework provides the capability of managing access to an editing domain by
multiple reading and writing threads. It also provides a facility to register and share an
editing domain amongst different clients and listeners. Resource set listeners are defined
in the transaction layer and are provided with notifications in batches. The resource set listener
has the option to append changes before the transaction is committed (pre-commit) or receive only
the notifications of transactions that were validated and not rolled-back (post-commit).
The following are the main extension points and classes to be used with the transaction framework:
The extension point
is used to declare an transactional editing domain unique identifier. This identifier can be used by clients to
either construct the transactional editing domain if it doesn't already exist or find the editing domain from memory
if it already exists.
The extension point
is used to declare listeners that should be automatically attached to an editing domain when it is first
constructed. A listener can be registered against all, one or many editing domains.
TransactionalEditingDomain has two interfaces called
can be used to add new transactional editing domains to the registry using a unique identifier or
to retrieve (or construct) an editing domain from its unique identifier. The
Factory has a default
implementation that is used
to construct new transactional editing domains. A
TransactionalEditingDomain allows clients to
add and remove their
ResourceSetListeners as well as allowing clients to perform exclusive reads blocking
any other reading / writing threads. Write access is obtained by executing an EMF-style command
on the editing domain's command stack.
Clients can use the
RecordingCommand to automatically record the changes made by their subclass
so that undo and redo can be done automatically. The client does not need to implement undo and
redo in this case.
RunnableWithResult class is provided so that clients can pass instances of this special
runExclusive() method on a
TransactionalEditingDomain and have their result propagated back
to the caller of
runExclusive(). This can help clients who are querying models and returning
results back to the caller.
Please refer to the tutorial
Copyright (c) 2006, 2007 IBM Corporation and others. All Rights Reserved.