Wrapping

This is the manual for older MicroStream versions (Version < 5.0).
The new documentation (Version >= 5.0) is located at:
​https://docs.microstream.one/​
MicroStream uses a strictly interface-based architecture. All types in the public API are, whenever possible, interfaces. This offers the best possibilities to extend or exchange parts of the engine. A good ways to enrich a type with features, is the wrapper (decorator) pattern.
For example, let's say we want to add logging to the PersistenceStoring's store(object) method.
PersistenceStoring.java
public interface PersistenceStoring
{
    public long store(Object instance);
    
    public long[] storeAll(Object... instances);
    
    public void storeAll(Iterable<?> instances);
    
    public void storeSelfStoring(SelfStoring storing);
}
Conventionally it would be done that way: A new type, implementing the original interface, would be handed over the wrapped instance, all interface methods have to be implemented and delegated. And in the single method, we wanted to add functionality; the actual implementation of the logging is done.
public class PersistenceStoringWithLogging implements PersistenceStoring
{
	private final PersistenceStoring wrapped;
	
	public PersistenceStoringWithLogging(final PersistenceStoring wrapped)
	{
		super();
		
		this.wrapped = wrapped;
	}
	
	@Override
	public long store(final Object instance)
	{
		Logger.getLogger(PersistenceStoring.class.getName())
			.info("Object stored: " + instance);
		
		return this.wrapped.store(instance);
	}
	
	@Override
	public long[] storeAll(final Object... instances)
	{
		return this.wrapped.storeAll(instances);
	}
	
	@Override
	public void storeAll(final Iterable<?> instances)
	{
		this.wrapped.storeAll(instances);
	}
	
	@Override
	public void storeSelfStoring(final SelfStoring storing)
	{
		this.wrapped.storeSelfStoring(storing);
	}
}
This produces a lot of overhead. In this case, three methods are just boilerplate code to delegate the calls to the wrapped instance. A common solution for that is to create an abstract base wrapper type for the designated interface, and to reuse it whenever needed.
public abstract class BaseWrapperPersistenceStoring implements PersistenceStoring
{
	private final PersistenceStoring wrapped;
	
	public BaseWrapperPersistenceStoring(final PersistenceStoring wrapped)
	{
		super();
		
		this.wrapped = wrapped;
	}
	
	@Override
	public long store(final Object instance)
	{
		return this.wrapped.store(instance);
	}
	
	@Override
	public long[] storeAll(final Object... instances)
	{
		return this.wrapped.storeAll(instances);
	}
	
	@Override
	public void storeAll(final Iterable<?> instances)
	{
		this.wrapped.storeAll(instances);
	}
	
	@Override
	public void storeSelfStoring(final SelfStoring storing)
	{
		this.wrapped.storeSelfStoring(storing);
	}
}
And then, based on that, the implementation of the logger type would look like this:
public class PersistenceStoringWithLogging extends BaseWrapperPersistenceStoring
{
	public PersistenceStoringWithLogging(PersistenceStoring wrapped)
	{
		super(wrapped);
	}
	
	@Override
	public long store(Object instance)
	{
		Logger.getLogger(PersistenceStoring.class.getName())
			.info("Object stored: " + instance);
			
		return super.store(instance);
	}
}
That's better. No more boilerplate code. Just overwrite the methods you want to extend. 
The only work left is, to generate the base wrapper types. One way is to let your IDE generate the wrapper or delegation code. Disadvantage of that is, it has to be redone every time your interfaces change. A code generator, which does it automatically would be nice. And that's what the base module brings along. Like the layered entity code generator, it is an annotation processor.
Multiple Layers
Configuration
Last modified 1yr ago