In this section, we will review a sample application that is included in the Spring Integration distribution. This sample is inspired by one of the samples featured in Gregor Hohpe's Ramblings.

The domain is that of a Cafe, and the basic flow is depicted in the following diagram:

The Order object may contain multiple OrderItems. Once the order is placed, a Splitter will break the composite order message into a single message per drink. Each of these is then processed by a Router that determines whether the drink is hot or cold (checking the OrderItem object's 'isIced' property). The Barista prepares each drink, but hot and cold drink preparation are handled by two distinct methods: 'prepareHotDrink' and 'prepareColdDrink'. The prepared drinks are then sent to the Waiter where they are aggregated into a Delivery object.

Here is the XML configuration:

<?xml version="1.0" encoding="UTF-8"?>
<beans:beans xmlns="http://www.springframework.org/schema/integration"
    xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
    xmlns:beans="http://www.springframework.org/schema/beans"
    xmlns:stream="http://www.springframework.org/schema/integration/stream"
    xsi:schemaLocation="http://www.springframework.org/schema/beans
            http://www.springframework.org/schema/beans/spring-beans-2.5.xsd
            http://www.springframework.org/schema/integration
            http://www.springframework.org/schema/integration/spring-integration-1.0.xsd
            http://www.springframework.org/schema/integration/stream
            http://www.springframework.org/schema/integration/stream/spring-integration-stream-1.0.xsd">

    <gateway id="cafe" service-interface="org.springframework.integration.samples.cafe.Cafe"/>

    <channel id="orders"/>
    <splitter input-channel="orders" ref="orderSplitter" method="split" output-channel="drinks"/>

    <channel id="drinks"/>
    <router input-channel="drinks" ref="drinkRouter" method="resolveOrderItemChannel"/>

    <channel id="coldDrinks">
        <queue capacity="10"/>
    </channel>
    <service-activator input-channel="coldDrinks" ref="barista"
                       method="prepareColdDrink" output-channel="preparedDrinks"/>

    <channel id="hotDrinks">
        <queue capacity="10"/>
    </channel>
    <service-activator input-channel="hotDrinks" ref="barista"
                       method="prepareHotDrink" output-channel="preparedDrinks"/>

    <channel id="preparedDrinks"/>
    <aggregator input-channel="preparedDrinks" ref="waiter"
                method="prepareDelivery" output-channel="deliveries"/>

    <stream:stdout-channel-adapter id="deliveries"/>

    <beans:bean id="orderSplitter"
                class="org.springframework.integration.samples.cafe.xml.OrderSplitter"/>

    <beans:bean id="drinkRouter"
                class="org.springframework.integration.samples.cafe.xml.DrinkRouter"/>

    <beans:bean id="barista" class="org.springframework.integration.samples.cafe.xml.Barista"/>

    <beans:bean id="waiter" class="org.springframework.integration.samples.cafe.xml.Waiter"/>

    <poller id="poller" default="true">
        <interval-trigger interval="1000"/>
    </poller>

</beans:beans>

As you can see, each Message Endpoint is connected to input and/or output channels. Each endpoint will manage its own Lifecycle (by default endpoints start automatically upon initialization - to prevent that add the "auto-startup" attribute with a value of "false"). Most importantly, notice that the objects are simple POJOs with strongly typed method arguments. For example, here is the Splitter:

public class OrderSplitter {

    public List<OrderItem> split(Order order) {
        return order.getItems();
    }
}

In the case of the Router, the return value does not have to be a MessageChannel instance (although it can be). As you see in this example, a String-value representing the channel name is returned instead.

public class DrinkRouter {

    public String resolveOrderItemChannel(OrderItem orderItem) {
        return (orderItem.isIced()) ? "coldDrinks" : "hotDrinks";
    }
}

Now turning back to the XML, you see that there are two <service-activator> elements. Each of these is delegating to the same Barista instance but different methods: 'prepareHotDrink' or 'prepareColdDrink' corresponding to the two channels where order items have been routed.

public class Barista {

    private long hotDrinkDelay = 5000;
    private long coldDrinkDelay = 1000;

    private AtomicInteger hotDrinkCounter = new AtomicInteger();
    private AtomicInteger coldDrinkCounter = new AtomicInteger();

    public void setHotDrinkDelay(long hotDrinkDelay) {
        this.hotDrinkDelay = hotDrinkDelay;
    }

    public void setColdDrinkDelay(long coldDrinkDelay) {
        this.coldDrinkDelay = coldDrinkDelay;
    }

    public Drink prepareHotDrink(OrderItem orderItem) {
        try {
            Thread.sleep(this.hotDrinkDelay);
            System.out.println(Thread.currentThread().getName()
                    + " prepared hot drink #" + hotDrinkCounter.incrementAndGet()
                    + " for order #" + orderItem.getOrder().getNumber() + ": " + orderItem);
            return new Drink(orderItem.getOrder().getNumber(), orderItem.getDrinkType(),
                    orderItem.isIced(), orderItem.getShots());
        }
        catch (InterruptedException e) {
            Thread.currentThread().interrupt();
            return null;
        }
    }

    public Drink prepareColdDrink(OrderItem orderItem) {
        try {
            Thread.sleep(this.coldDrinkDelay);
            System.out.println(Thread.currentThread().getName()
                    + " prepared cold drink #" + coldDrinkCounter.incrementAndGet()
                    + " for order #" + orderItem.getOrder().getNumber() + ": " + orderItem);
            return new Drink(orderItem.getOrder().getNumber(), orderItem.getDrinkType(),
                    orderItem.isIced(), orderItem.getShots());
        }
        catch (InterruptedException e) {
            Thread.currentThread().interrupt();
            return null;
        }
    }

}

As you can see from the code excerpt above, the barista methods have different delays (the hot drinks take 5 times as long to prepare). This simulates work being completed at different rates. When the CafeDemo 'main' method runs, it will loop 100 times sending a single hot drink and a single cold drink each time. It actually sends the messages by invoking the 'placeOrder' method on the Cafe interface. Above, you will see that the <gateway> element is specified in the configuration file. This triggers the creation of a proxy that implements the given 'service-interface' and connects it to a channel. The channel name is provided on the @Gateway annotation of the Cafe interface.

public interface Cafe {

    @Gateway(requestChannel="orders")
    void placeOrder(Order order);

}

Finally, have a look at the main() method of the CafeDemo itself.

public static void main(String[] args) {
    AbstractApplicationContext context = null;
    if (args.length > 0) {
        context = new FileSystemXmlApplicationContext(args);
    }
    else {
        context = new ClassPathXmlApplicationContext("cafeDemo.xml", CafeDemo.class);
    }
    Cafe cafe = (Cafe) context.getBean("cafe");
    for (int i = 1; i <= 100; i++) {
        Order order = new Order(i);
        order.addItem(DrinkType.LATTE, 2, false);
        order.addItem(DrinkType.MOCHA, 3, true);
        cafe.placeOrder(order);
    }
}

	Tip
	To run this sample as well as 8 others, refer to the README.txt within the "samples" directory of the main distribution as described at the beginning of this chapter.

When you run cafeDemo, you will see that the cold drinks are initially prepared more quickly than the hot drinks. Because there is an aggregator, the cold drinks are effectively limited by the rate of the hot drink preparation. This is to be expected based on their respective delays of 1000 and 5000 milliseconds. However, by configuring a poller with a concurrent task executor, you can dramatically change the results. For example, you could use a thread pool executor with 5 workers for the hot drink barista while keeping the cold drink barista as it is:

<service-activator input-channel="hotDrinks"
                   ref="barista"
                   method="prepareHotDrink"
                   output-channel="preparedDrinks"/>

<service-activator input-channel="hotDrinks"
                   ref="barista"
                   method="prepareHotDrink"
                   output-channel="preparedDrinks">
    <poller task-executor="pool">
        <interval-trigger interval="1000"/>
    </poller>
</service-activator>

<thread-pool-task-executor id="pool" core-size="5"/>

Also, notice that the worker thread name is displayed with each invocation. You will see that the hot drinks are prepared by the task-executor threads. If you provide a much shorter poller interval (such as 100 milliseconds), then you will notice that occasionally it throttles the input by forcing the task-scheduler (the caller) to invoke the operation.

	Note
	In addition to experimenting with the poller's concurrency settings, you can also add the 'transactional' sub-element and then refer to any PlatformTransactionManager instance within the context.