brooklyn.internal.storage.BrooklynStorage Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of brooklyn-core Show documentation
Show all versions of brooklyn-core Show documentation
Entity implementation classes, events, and other core elements
package brooklyn.internal.storage;
import java.util.List;
import java.util.Map;
import java.util.concurrent.ConcurrentMap;
import com.google.common.annotations.Beta;
public interface BrooklynStorage {
/**
* Creates a reference to a value, backed by the storage-medium. If a reference with this
* name has already been created, then that existing reference will be returned.
*
* The returned reference is a live view: changes made to the reference will be persisted,
* and changes that others make will be reflected in the reference.
*
* The reference is thread-safe. No additional synchronization is required when getting/setting
* the reference.
*
* @param id
*/
Reference getReference(String id);
/**
* Creates a list backed by the storage-medium. If a list with this name has already been
* created, then that existing list will be returned.
*
* The returned list is not a live view. Changes are made by calling reference.set(), and
* the view is refreshed by calling reference.get(). Changes are thread-safe, but callers
* must be careful not to overwrite other's changes. For example, the code below could overwrite
* another threads changes that are made to the map between the call to get() and the subsequent
* call to set().
*
*
* {@code
* Reference> ref = storage.createNonConcurrentList("myid");
* List newval = ImmutableList.builder().addAll(ref.get()).add("another").builder();
* ref.set(newval);
* }
*
*
* TODO Aled says: Is getNonConcurrentList necessary?
* The purpose of this method, rather than just using
* {@code Reference ref = getReference(id); ref.set(ImmutableList.of())}
* is to allow control of the serialization of the things inside the list
* (e.g. switching the Location object to serialize a proxy object of some sort).
* I don't want us to have to do deep inspection of every object being added to any map/ref.
* Feels like we can use normal serialization unless the top-level object matches an
* instanceof for special things like Entity, Location, etc.
*
* Peter responds:
* What I'm a bit scared of is that we need to write some kind of meta serialization mechanism
* on top of the mechanisms provided by e.g. Hazelcast or Infinispan. Hazelcast has a very
* extensive serialization library where you can plug in all kinds of serialization mechanisms.
*
* @param id
*/
@Beta
Reference> getNonConcurrentList(String id);
/**
* Creates a map backed by the storage-medium. If a map with this name has already been
* created, then that existing map will be returned.
*
* The returned map is a live view: changes made to the map will be persisted, and changes
* that others make will be reflected in the map.
*
* The map is thread-safe: {@link Map#keySet()} etc will iterate over a snapshot view of the
* contents.
*
* @param id
*/
ConcurrentMap getMap(String id);
/**
* Removes the data stored against this id, whether it is a map, ref or whatever.
*/
void remove(String id);
/**
* Terminates the BrooklynStorage.
*/
void terminate();
}
© 2015 - 2025 Weber Informatics LLC | Privacy Policy