Spring Batch: One reader, multiple processors and writers

问题

In Spring batch I need to pass the items read by an ItemReader to two different processors and writer. What I'm trying to achieve is that...

                        +---> ItemProcessor#1 ---> ItemWriter#1
                        |
ItemReader ---> item ---+
                        |
                        +---> ItemProcessor#2 ---> ItemWriter#2

This is needed because items written by ItemWriter#1 should be processed in a completely different way compared to the ones written by ItemWriter#2. Moreover, ItemReader reads item from a database, and the queries it executes are so computational expensive that executing the same query twice should be discarded.

Any hint about how to achieve such set up ? Or, at least, a logically equivalent set up ?

回答1:

This solution is valid if your item should be processed by processor #1 and processor #2

You have to create a processor #0 with this signature:

class Processor0<Item, CompositeResultBean>

where CompositeResultBean is a bean defined as

class CompositeResultBean {
  Processor1ResultBean result1;
  Processor2ResultBean result2;
}

In your Processor #0 just delegate work to processors #1 and #2 and put result in CompositeResultBean

CompositeResultBean Processor0.process(Item item) {
  final CompositeResultBean r = new CompositeResultBean();
  r.setResult1(processor1.process(item));
  r.setResult2(processor2.process(item));
  return r;
}

Your own writer is a CompositeItemWriter that delegate to writer CompositeResultBean.result1 or CompositeResultBean.result2 (look at PropertyExtractingDelegatingItemWriter, maybe can help)

回答2:

I followed Luca's suggestion to use PropertyExtractingDelegatingItemWriter as writer and I was able to work with two different entities in one single step.

First of all what I did was to define a DTO that stores the two entities/results from the processor

public class DatabaseEntry {
    private AccessLogEntry accessLogEntry;
    private BlockedIp blockedIp;

    public AccessLogEntry getAccessLogEntry() {
        return accessLogEntry;
    }

    public void setAccessLogEntry(AccessLogEntry accessLogEntry) {
        this.accessLogEntry = accessLogEntry;
    }

    public BlockedIp getBlockedIp() {
        return blockedIp;
    }

    public void setBlockedIp(BlockedIp blockedIp) {
        this.blockedIp = blockedIp;
    }
}

Then I passed this DTO to the writer, a PropertyExtractingDelegatingItemWriter class where I define two customized methods to write the entities into the database, see my writer code below:

@Configuration
public class LogWriter extends LogAbstract {
    @Autowired
    private DataSource dataSource;

    @Bean()
    public PropertyExtractingDelegatingItemWriter<DatabaseEntry> itemWriterAccessLogEntry() {
        PropertyExtractingDelegatingItemWriter<DatabaseEntry> propertyExtractingDelegatingItemWriter = new PropertyExtractingDelegatingItemWriter<DatabaseEntry>();
        propertyExtractingDelegatingItemWriter.setFieldsUsedAsTargetMethodArguments(new String[]{"accessLogEntry", "blockedIp"});
        propertyExtractingDelegatingItemWriter.setTargetObject(this);
        propertyExtractingDelegatingItemWriter.setTargetMethod("saveTransaction");
        return propertyExtractingDelegatingItemWriter;
    }

    public void saveTransaction(AccessLogEntry accessLogEntry, BlockedIp blockedIp) throws SQLException {
        writeAccessLogTable(accessLogEntry);
        if (blockedIp != null) {
            writeBlockedIp(blockedIp);
        }

    }

    private void writeBlockedIp(BlockedIp entry) throws SQLException {
        PreparedStatement statement = dataSource.getConnection().prepareStatement("INSERT INTO blocked_ips (ip,threshold,startDate,endDate,comment) VALUES (?,?,?,?,?)");
        statement.setString(1, entry.getIp());
        statement.setInt(2, threshold);
        statement.setTimestamp(3, Timestamp.valueOf(startDate));
        statement.setTimestamp(4, Timestamp.valueOf(endDate));
        statement.setString(5, entry.getComment());
        statement.execute();
    }

    private void writeAccessLogTable(AccessLogEntry entry) throws SQLException {
        PreparedStatement statement = dataSource.getConnection().prepareStatement("INSERT INTO log_entries (date,ip,request,status,userAgent) VALUES (?,?,?,?,?)");
        statement.setTimestamp(1, Timestamp.valueOf(entry.getDate()));
        statement.setString(2, entry.getIp());
        statement.setString(3, entry.getRequest());
        statement.setString(4, entry.getStatus());
        statement.setString(5, entry.getUserAgent());
        statement.execute();
    }
}

With this approach you can get the wanted inital behaviour from a single reader for processing multiple entities and save them in a single step.

回答3:

You can use a CompositeItemProcessor and CompositeItemWriter

It won't look exactly like your schema, it will be sequential, but it will do the job.

回答4:

this is the solution I came up with.

So, the idea is to code a new Writer that "contains" both an ItemProcessor and an ItemWriter. Just to give you an idea, we called it PreprocessoWriter, and that's the core code.

private ItemWriter<O> writer;
private ItemProcessor<I, O> processor;

@Override
public void write(List<? extends I> items) throws Exception {
    List<O> toWrite = new ArrayList<O>();
    for (I item : items) {
        toWrite.add(processor.process(item));
    }
    writer.write(toWrite);
}

There's a lot of things being left aside. Management of ItemStream, for instance. But in our particular scenario this was enough.

So you can just combine multiple PreprocessorWriter with CompositeWriter.

回答5:

There is an other solution if you have a reasonable amount of items (like less than 1 Go) : you can cache the result of your select into a collection wrapped in a Spring bean.

Then u can just read the collection twice with no cost.

来源：https://stackoverflow.com/questions/18999724/spring-batch-one-reader-multiple-processors-and-writers