Simple Use Case

This section demonstrates how to define a simple job using the Job Specification Language (JSL) and how to implement the corresponding batch artifacts. Refer to the rest of the sections in this chapter for detailed descriptions of the elements in the batch framework.

The following job definition specifies a chunk step and a task step as follows:

<?xml version="1.0" encoding="UTF-8"?>
<job id="simplejob" xmlns="http://xmlns.jcp.org/xml/ns/javaee"
                    version="1.0">
  <properties>
    <property name="input_file" value="input.txt"/>
    <property name="output_file" value="output.txt"/>
  </properties>
  <step id="mychunk" next="mytask">
    <chunk>
      <reader ref="MyReader"></reader>
      <processor ref="MyProcessor"></processor>
      <writer ref="MyWriter"></writer>
    </chunk>
  </step>
  <step id="mytask">
    <batchlet ref="MyBatchlet"></batchlet>
    <end on="COMPLETED"/>
  </step>
</job>

Chunk Step

In most cases, you have to implement a checkpoint class for chunk-oriented steps. The following class just keeps track of the line number in a text file:

public class MyCheckpoint implements Serializable {
    private long lineNum = 0;
    public void increase() { lineNum++; }
    public long getLineNum() { return lineNum; }
}

The following item reader implementation continues reading the input file from the provided checkpoint if the job was restarted. The items consist of each line in the text file (in more complex scenarios, the items are custom Java types and the input source can be a database):

@Dependent
@Named("MyReader")
public class MyReader implements javax.batch.api.chunk.ItemReader {
    private MyCheckpoint checkpoint;
    private BufferedReader breader;
    @Inject
    JobContext jobCtx;

    public MyReader() {}

    @Override
    public void open(Serializable ckpt) throws Exception {
        if (ckpt == null)
            checkpoint = new MyCheckpoint();
        else
            checkpoint = (MyCheckpoint) ckpt;
        String fileName = jobCtx.getProperties()
                                .getProperty("input_file");
        breader = new BufferedReader(new FileReader(fileName));
        for (long i = 0; i < checkpoint.getLineNum(); i++)
            breader.readLine();
    }

    @Override
    public void close() throws Exception {
        breader.close();
    }

    @Override
    public Object readItem() throws Exception {
        String line = breader.readLine();
        return line;
    }
}

In the following case, the item processor only converts the line to uppercase. More complex examples can process items in different ways or transform them into custom output Java types:

@Dependent
@Named("MyProcessor")
public class MyProcessor implements javax.batch.api.chunk.ItemProcessor {
    public MyProcessor() {}

    @Override
    public Object processItem(Object obj) throws Exception {
        String line = (String) obj;
        return line.toUpperCase();
    }
}

Note:

The batch processing API does not support generics. In most cases, you need to cast items to their specific type before processing them.

The item writer writes the processed items to the output file. It overwrites the output file if no checkpoint is provided; otherwise, it resumes writing at the end of the file. Items are written in chunks:

@Dependent
@Named("MyWriter")
public class MyWriter implements javax.batch.api.chunk.ItemWriter {
    private BufferedWriter bwriter;
    @Inject
    private JobContext jobCtx;

    @Override
    public void open(Serializable ckpt) throws Exception {
        String fileName = jobCtx.getProperties()
                                .getProperty("output_file");
        bwriter = new BufferedWriter(new FileWriter(fileName,
                                                    (ckpt != null)));
    }

    @Override
    public void writeItems(List<Object> items) throws Exception {
        for (int i = 0; i < items.size(); i++) {
            String line = (String) items.get(i);
            bwriter.write(line);
            bwriter.newLine();
        }
    }

    @Override
    public Serializable checkpointInfo() throws Exception {
        return new MyCheckpoint();
}

Task Step

The task step displays the length of the output file. In more complex scenarios, task steps perform any task that does not fit the chunk processing programming model:

@Dependent
@Named("MyBatchlet")
public class MyBatchlet implements javax.batch.api.chunk.Batchlet {
    @Inject
    private JobContext jobCtx;

    @Override
    public String process() throws Exception {
        String fileName = jobCtx.getProperties()
                                .getProperty("output_file");
        System.out.println(""+(new File(fileName)).length());
        return "COMPLETED";
    }
}