How to use Apache Storm Trident API

Sometime back i blogged about HelloWorld - Apache Storm Word Counter program , which demonstrates how to build WordCount program using Apache Storm. Now problem with that project was that it was not Maven project instead i had screen shot of all the jars that you will have to include in the program. So i changed it to use Apache Maven as build framework. You can download the source code. In addition to normal API, storm also provides trident API, which allows us to build much compact code, i wanted to try that out so i built this simple Word Count program using Trident API. While using Trident API you will have to start by creating object of TridentTopology, you still need LineReaderSpout that takes file path as input, reads and emits one line of file at a time. But the part that is different is you dont need WordSpitterBolt and WordCounterBolt, instead you can use compact code like this

 topology.newStream("spout",lineReaderSpout)
.each(new Fields("line"), new Split(), new Fields("word"))
.groupBy(new Fields("word"))
.aggregate(new Fields("word"), new Count(), new Fields("count"))
.each(new Fields("word","count"), new Debug());

The each(new Fields("line"), new Split(), new Fields("word")) line takes the line emitted by the LineReaderSpout and uses built in storm.trident.operation.builtin.Split function to split the lines into words and emits each word as Tuple. The groupBy(new Fields("word")) line takes the tuples and groups them by word's. The aggregate(new Fields("word"), new Count(), new Fields("count")) line takes care of aggregating the words and counts them(At this point you have a tuple like {word,count}), for that it uses storm.trident.operation.builtin.Count class. The last part is .each(new Fields("word","count"), new Debug());, which takes care of printing each tuple which in WORD count format. Trident API provides set of sample classes that makes developing WordCount type of program very easy. But you could have created your own version of Split and Count program and the code would still look significantly compact