Processing Streaming Data - Real-Time Analytics - page 154

Finally, the chooseTasks method is implemented to rotate between
the various tasks:
public List<Integer> chooseTasks( int taskId,
List<Object> values) {
LinkedList<Integer> out = new
LinkedList<Integer>();
out.add(targetTasks.get(nextTask));
nextTask = (nextTask + 1) % numTasks;
return out;
}
Implementing Bolts
Now that data can be passed to bolts appropriately, it is time to understand
how bolts themselves work. As shown in the previous section, bolts receive
data from other bolts or spouts within the topology. They are event driven
so they cannot be used to retrieve data; that is the role of the spout.
Although it is possible to implement bolts by implementing either the
IBasicBolt or IRichBolt interfaces, the recommended method is to
extend the BaseBasicBolt or BaseRichBolt classes. These base classes
provide simple implementations for most of Storm's boilerplate methods.
Rich Bolts
Toimplement a RichBolt ,startbyextending this RichBaseBolt abstract
class and implementing the three required methods. Like the custom
grouping class from the previous section, the first is a prepare method that
passes in information about the topology. For bolts, an extra parameter is
added: the collector object. This object manages the interaction between
the bolt and the topology, especially transmitting and acknowledging tuples.
In a rich bolt, the collector is usually placed into an instance variable
for use in the execute method. The prepare method is also a good place
to create an object that cannot be serialized because they do not implement
the Serialization interface. It is a good habit to declare all variables that
are initialized by the prepare method as transient. While not all variables