Spring Batch(0)——控制Step执行流程

Conditional Flow in Spring Batch

I just announced the new Learn Spring course, focused on the fundamentals of Spring 5 and Spring Boot 2:

CHECK OUT THE COURSE

1. Introduction

We use Spring Batch to compose jobs from multiple steps that read, transform, and write data. If the steps in a job have multiple paths, similar to using an if statement in our code, we say that the job flow is conditional.

In this tutorial, we'll look at two ways to create Spring Batch jobs with a conditional flow.

2. Exit Status and Batch Status

When we specify a conditional step with Spring's Batch framework, we're using the exit status of a step or job. Therefore, we need to understand the difference between Batch Status and Exit Status in our steps and jobs:

• BatchStatus is an Enum representing the status of a step/job and is used by the Batch framework internally
• Possible values are: ABANDONED, COMPLETED, FAILED, STARTED, STARTING, STOPPED, STOPPING, UNKNOWN
• ExitStatus is the status of a step when execution is completed and is used to conditionally determine the flow

By default, the ExitStatus of a step or job is the same as its BatchStatus. We can also set a custom ExitStatus to drive flow.

3. Conditional Flow

Let's say we have an IOT device that sends us measurements. Our device measurements are arrays of integers, and we need to send notifications if any of our measurements contain positive integers.

In other words, we need to send a notification when we detect a positive measurement.

3.1. ExitStatus

Importantly, we use the exit status of a step to drive conditional flow.

To set the exit status of a step, we need to use the StepExecution object's setExitStatus method. In order to do that, we need to create an ItemProcessor that extends ItemListenerSupport and gets the step's StepExecution.

We use this to set our step's exit status to NOTIFY when we find a positive number. When we determine our exit status based on data within the batch job, we can use an ItemProcessor.

Let's look at our NumberInfoClassifier to see the three methods we need:

public class NumberInfoClassifier extends ItemListenerSupport<NumberInfo, Integer>
  implements ItemProcessor<NumberInfo, Integer> {

    private StepExecution stepExecution;

    @BeforeStep
    public void beforeStep(StepExecution stepExecution) {
        this.stepExecution = stepExecution;
        this.stepExecution.setExitStatus(new ExitStatus(QUIET));
    }

    @Override
    public Integer process(NumberInfo numberInfo) throws Exception {
        return Integer.valueOf(numberInfo.getNumber());
    }

    @Override
    public void afterProcess(NumberInfo item, Integer result) {
        super.afterProcess(item, result);
        if (item.isPositive()) {
            stepExecution.setExitStatus(new ExitStatus(NOTIFY));
        }
    }
}

Note: we use the ItemProcessor to set the ExitStatus in this example, but we could just as easily do it in our step's ItemReader or ItemWriter.

Finally, when we create our job, we tell our JobBuilderFactory to send notifications for any step that exits with a status of NOTIFY:

jobBuilderFactory.get("Number generator - second dataset")
    .start(dataProviderStep)
    .on("NOTIFY").to(notificationStep)
    .end()
    .build();

Also note that when we have additional conditional branches and multiple exit codes, we can add them to our job with the from and on methods of the JobBuilderFacotry:

jobBuilderFactory.get("Number generator - second dataset")
    .start(dataProviderStep)
    .on("NOTIFY").to(notificationStep)
    .from(step).on("LOG_ERROR").to(errorLoggingStep)
    .end()
    .build();

Now, any time our ItemProcessor sees a positive number, it will direct our job to run the notificationStep, which simply prints a message to System.out:

Second Dataset Processor 11
Second Dataset Processor -2
Second Dataset Processor -3
[Number generator - second dataset] contains interesting data!!

If we had a data set without a positive number, we would not see our notificationStep message:

Second Dataset Processor -1
Second Dataset Processor -2
Second Dataset Processor -3

3.2. Programmatic Branching With JobExecutionDecider

Alternatively, we can use a class that implements JobExecutionDecider to determine job flow. This is especially useful if we have external factors for determining execution flow.

To use this method, we first need to modify our ItemProcessor to remove the ItemListenerSupport interface and @BeforeStep method:

public class NumberInfoClassifierWithDecider extends ItemListenerSupport<NumberInfo, Integer>
  implements ItemProcessor<NumberInfo, Integer> {

    @Override
    public Integer process(NumberInfo numberInfo) throws Exception {
        return Integer.valueOf(numberInfo.getNumber());
    }
}

Next, we create a decider class that determines the notification status of our step:

public class NumberInfoDecider implements JobExecutionDecider {

    private boolean shouldNotify() {
        return true;
    }

    @Override
    public FlowExecutionStatus decide(JobExecution jobExecution, StepExecution stepExecution) {
        if (shouldNotify()) {
            return new FlowExecutionStatus(NOTIFY);
        } else {
            return new FlowExecutionStatus(QUIET);
        }
    }
}

Then, we set up our Job to use the decider in the flow:

jobBuilderFactory.get("Number generator - third dataset")
    .start(dataProviderStep)
    .next(new NumberInfoDecider()).on("NOTIFY").to(notificationStep)
    .end()
    .build();

4. Conclusion

In this quick tutorial, we explored two options for implementing conditional flows with Spring Batch. First, we looked at how to use the ExitStatus to control the flow of our job.

Then we took a look at how we can control flow programmatically by defining our own JobExecutionDecider.

As always, the full source code of the article is available over on GitHub.