Almost.
UnitOfWork is in the singleton scope. This means there will exist exactly one
instance for the entire injector. So it does not matter if you inject it
directly or retrieve it from a provider. You will always recieve the very same
(and only) instance.
The EntityManager on the other hand has a life cycle. Therefore it is very
important to controll the time of its creation. This can be done using the
provider.
The life cycle of the entity manager is controlled by the unit of work. Because
it is a singelton and controles many entity manager it must somehow map a call
to begin() or end() to an single entity manager. This is done by the thread. So
every call to the unit of work has only an influence on the entity manager for
the same thread.
An entity manager must only be retrieved after the unit of work has been
started and may no longer be used after the unit of work has ended. One of the
open issues with guice persist is, that retrieving an entity manager outside of
a unit of work does not throw an exception but implicitly starts the unit of
work. This will almost always lead to the "unit of work already begun"
exception. But at the time the exception occures it is no longer possible to
determine where the unit of work was started.
So once again. Consider onami persist since it is based on guice persist but
improved on the lessons learned from guice persist.
And an advice to the code snipped you posted. As a rule of thumb: never store
the entity manager in a field. And never pass the entity manager to anything
(this helps avoiding to store it in a field). If you need to pass it to another
class pass the provider instead. Or even bether have the provider inject into
the other class. Calling the provider is fast and it prevents you from running
into a closed entity manager.
Am 13. Mai 2015 17:11:15 MESZ, schrieb "JB Lézoray"
<jb.lezo...@dolmen-tech.com>:
>Hi Stephan,
>
>I followed your advices, and it seems that I don't have the error when
>I inject a UnitOfWork instead of a Provider<UnitOfWork>.
>
>As far as I understand it, if a MyRunnable gets executed by
>the ExecutorService before the end of the MyServletWorker, then it
>receives a copy of the existing UnitOfWork instead of a new instance.
>Is it a bug or did I misunderstand something ? I relied on
>a Provider<UnitOfWork> because I wanted the UnitOfWork to be created
>when the Runnable is executed rather than when the Runnable is
>instantiated.
>
>Thank you for your answers, it has saved me *a lot* of debugging time.
>
>
>// A time consuming job.
>class MyRunnable implements Runnable {
>
> @Inject Provider<EntityManager> entityManagerProvider;
>
> @Inject UnitOfWork unitOfWork;
>
> @Override
> public void run()
> {
> try {
> unitOfWork.begin();
> EntityManager em = entityManagerProvider.get();
> this.doTimeConsumingStuff(em);
> } finally {
> unitOfWork.end();
> }
> }
>
> @Transactional
> public void doTimeConsumingStuff(EntityManager em) {/* ... */}
>}
>
>
>Le 13 mai 2015 à 14:51:35, Stephan Classen (st.clas...@gmx.ch) a écrit:
>
>The persist framework takes care that every thread has its own entity
>manager.
>So you don't need to do this.
>Calling unitOfWork.end() will close the entity manager for you. So also
>no need to do this manually.
>
>The unitOfWork is a singleton. So you can inject it directly and don't
>need to use the provider.
>You must use the provider for the entityManager as you do.
>
>Also important:
>@Transactional works using interception at runtime. Threrefore it can
>only be used on non private, non final and non static methods. Because
>for private, final, or static methods the code can be inlined by the
>compiler. Then it is no longer possible to intercept the method call at
>runtime.
>
>Am 13. Mai 2015 14:35:39 MESZ, schrieb "JB Lézoray"
><jb.lezo...@dolmen-tech.com>:
>Hi Laszio,
>
>Thank you for your answer.
>
>I think you pinpointed the problem, and I'm a step forward in resolving
>it.
>
>Here is the MyRunnable implementation, it already relies on two
>providers for EntityManager and UnitOfWork.
>But now my question is : how to make sure that
>the Provider<EntityManager> in the MyRunnable worker returns
>another EntityManager than the one in the MyServletWorker ?
>
>
>
>
>// A time consuming job.
>class MyRunnable implements Runnable {
>
> @Inject Provider<EntityManager> entityManagerProvider;
>
> @Inject Provider<UnitOfWork> unitOfWorkProvider;
>
> @Override
> public void run()
> {
> UnitOfWork unitOfWork = null;
> EntityManager em = null;
> try {
> unitOfWork = unitOfWorkProvider.get();
> unitOfWork.begin();
> em = entityManagerProvider.get();
> this.doTimeConsumingStuff(em);
> } finally {
> if (unitOfWork != null) unitOfWork.end();
> if (em != null && em.isOpen()) em.close(); // should it be
>explicitly closed ?
> }
> }
>
> @Transactional
> private void doTimeConsumingStuff(EntityManager em) {/* ... */}
>}
>
>
>// Executed within the http worker thread.
>// It Creates and submits the jobs.
>class MyServletWorker {
>
> @Inject Provider<EntityManager> entityManagerProvider;
>
> @Inject ExecutorService executorService;
>
> public void createAndSubmitJobs(MyData data) {
> EntityManager em = entityManagerProvider.get();
> Set<MyRunnable> jobs = createJobs(em, data);
> for (MyRunnable job : jobs)
> executorService.submit(job);
> // ....
> }
>
> @Transactional
> public Set<MyRunnable> createJobs(EntityManager em, MyData data) {/*
>... */}
>}
>
>
>Le 13 mai 2015 à 13:16:59, Laszlo Ferenczi (lferen...@gmail.com) a
>écrit:
>
>Hi,
>
>I think the issue will be in the MyRunnable class.
>You create the MyRunnable instance in the http worker thread, if you
>simply inject an EntityManager there it'll be the one bound to the
>current thread (the http worker). After submitting it to a different
>thread you essentially leak the transaction which can have many side
>effects (like when the same http worker is selected again, the
>PersistFilter will try to start a new transaction but it's already open
>for that thread - held by the long running background job).
>
>The correct way to handle this is to inject a Provider<EntityManager>
>in MyRunnable and annotate the database using methods with
>@Transactional.
>
>--
>L
>
>
>--
>L
>
>On Wed, May 13, 2015 at 11:58 AM, jbl <jb.lezo...@dolmen-tech.com>
>wrote:
>Hello,
>
>
>I'm relatively new to guice, and I've got a problem that I quite don't
>understand.
>
>I have a webapp based on guice, eclipse link, mysql that runs on a
>tomcat 7. The app uses a
>com.google.inject.persist.jpa.JpaPersistModule and
>a com.google.inject.persist.PersistFilter
>
>For some requests the webapp creates a set of java.lang.Runnable,
>passes them to a java.util.concurrent.ExecutorService, and returns.
>The runnables are numerous time-consuming jobs that must be executed in
>background to avoid blocking the request. Also, each job has to access
>the EntityManager. However, the current implementation is quite naive:
>
> @Inject private Provider<MyRunnable> myRunnableProvider;
> @Inject private ExecutorService executorService;
>
> MyRunnable runnable = myRunnableProvider.get();
> executorService.submit(runnable);
>
>This code runs fine on my development computer, but I get this
>exception when I run it in a production environment :
>
>java.lang.IllegalStateException: Work already begun on this thread.
>Looks like you have called UnitOfWork.begin() twice without a balancing
>call to end() in between.
> at
>com.google.common.base.Preconditions.checkState(Preconditions.java:150)
>~[guava-15.0.jar:na]
> at
>com.google.inject.persist.jpa.JpaPersistService.begin(JpaPersistService.java:73)
>~[guice-persist-4.0.jar:na]
> at com.jbl.MyRunnable.run(MyRunnable.java:107)
>
>I think that this has something to do with the Scope, but I don't quite
>understand how to correct this.
>
>Thank you for any insight on this !
>
>--
>You received this message because you are subscribed to the Google
>Groups "google-guice" group.
>To unsubscribe from this group and stop receiving emails from it, send
>an email to google-guice+unsubscr...@googlegroups.com.
>To post to this group, send email to google-guice@googlegroups.com.
>Visit this group at http://groups.google.com/group/google-guice.
>To view this discussion on the web visit
>https://groups.google.com/d/msgid/google-guice/141ea4f0-3f34-45c9-a8cf-c28779859a2d%40googlegroups.com.
>For more options, visit https://groups.google.com/d/optout.
>
>--
>You received this message because you are subscribed to a topic in the
>Google Groups "google-guice" group.
>To unsubscribe from this topic, visit
>https://groups.google.com/d/topic/google-guice/JKPs9sayXi4/unsubscribe.
>To unsubscribe from this group and all its topics, send an email to
>google-guice+unsubscr...@googlegroups.com.
>To post to this group, send email to google-guice@googlegroups.com.
>Visit this group at http://groups.google.com/group/google-guice.
>To view this discussion on the web visit
>https://groups.google.com/d/msgid/google-guice/CAD-udUCC6FM2w9wGHNdUcMyZD6FQykHfGpUW%3DNveLu7UTawPkw%40mail.gmail.com.
>For more options, visit https://groups.google.com/d/optout.
>
>--
>Jean-Baptiste LÉZORAY
>Ingénieur Logiciel
>Portable: 06 64 86 32 92
>jb.lezo...@dolmen-tech.com
>Bâtiment Nucléole - 85 rue de Saint-Brieuc - 35 000 RENNES
>Retrouvez-nous sur www.dolmen.bzh
>
>--
>You received this message because you are subscribed to a topic in the
>Google Groups "google-guice" group.
>To unsubscribe from this topic, visit
>https://groups.google.com/d/topic/google-guice/JKPs9sayXi4/unsubscribe.
>To unsubscribe from this group and all its topics, send an email to
>google-guice+unsubscr...@googlegroups.com.
>To post to this group, send email to google-guice@googlegroups.com.
>Visit this group at http://groups.google.com/group/google-guice.
>To view this discussion on the web visit
>https://groups.google.com/d/msgid/google-guice/845FA7E2-4B79-4FB5-9A13-DB4CC2AFA230%40gmx.ch.
>For more options, visit https://groups.google.com/d/optout.
>
>--
>
>Jean-Baptiste LÉZORAY
>Ingénieur Logiciel
>Portable: 06 64 86 32 92
>jb.lezo...@dolmen-tech.com
>Bâtiment Nucléole - 85 rue de Saint-Brieuc - 35 000 RENNES
>Retrouvez-nous sur www.dolmen.bzh
>
>--
>You received this message because you are subscribed to the Google
>Groups "google-guice" group.
>To unsubscribe from this group and stop receiving emails from it, send
>an email to google-guice+unsubscr...@googlegroups.com.
>To post to this group, send email to google-guice@googlegroups.com.
>Visit this group at http://groups.google.com/group/google-guice.
>To view this discussion on the web visit
>https://groups.google.com/d/msgid/google-guice/etPan.55536993.3d1b58ba.5aa0%40MBA-JB.local.
>For more options, visit https://groups.google.com/d/optout.
--
You received this message because you are subscribed to the Google Groups
"google-guice" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to google-guice+unsubscr...@googlegroups.com.
To post to this group, send email to google-guice@googlegroups.com.
Visit this group at http://groups.google.com/group/google-guice.
To view this discussion on the web visit
https://groups.google.com/d/msgid/google-guice/53C4ACCE-C8B8-4337-90E6-51C69FFA43B1%40gmx.ch.
For more options, visit https://groups.google.com/d/optout.
