camel-aggregate-eip.md

Aggregate-eip.md

The Aggregator from the EIP patterns allows you to combine a number of messages into a single message.

How do we combine the results of individual, but related, messages so that they can be processed as a whole?

Use a stateful filter, an Aggregator, to collect and store individual messages until a complete set of related messages has been received. Then, the Aggregator publishes a single message distilled from the individual messages.

The aggregator is one of the most complex EIP and has many features and configurations.

The logic for combing messages together is correlated in buckets based on a correlation key. Messages with the same correlation key are aggregated together, using an AggregationStrategy.

Aggregate options

Exchange properties

Worker pools

The aggregate EIP will always use a worker pool used to process all the outgoing messages from the aggregator. The worker pool is determined accordingly:

• If a custom ExecutorService has been configured, then this is used as worker pool.

• If parallelProcessing=true then a default worker pool (is 10 worker threads by default) is created. However, the thread pool size and other configurations can be configured using thread pool profiles.

• Otherwise, a single threaded worker pool is created.

• To achieve synchronous aggregation, use an instance of SynchronousExecutorService for the executorService option. The aggregated output will execute in the same thread that called the aggregator.

Aggregating

The AggregationStrategy is used for aggregating the old, and the new exchanges together into a single exchange; that becomes the next old, when the next message is aggregated, and so forth.

Possible implementations include performing some kind of combining or delta processing. For instance, adding line items together into an invoice or just using the newest exchange and removing old exchanges such as for state tracking or market data prices, where old values are of little use.

Notice the aggregation strategy is a mandatory option and must be provided to the aggregator.

In the aggregate method, do not create a new exchange instance to return, instead return either the old or new exchange from the input parameters; favor returning the old exchange whenever possible.

Here are a few example AggregationStrategy implementations that should help you create your own custom strategy.

//simply combines Exchange String body values using '+' as a delimiter
class StringAggregationStrategy implements AggregationStrategy {

    public Exchange aggregate(Exchange oldExchange, Exchange newExchange) {
        if (oldExchange == null) {
            return newExchange;
        }

        String oldBody = oldExchange.getIn().getBody(String.class);
        String newBody = newExchange.getIn().getBody(String.class);
        oldExchange.getIn().setBody(oldBody + "+" + newBody);
        return oldExchange;
    }
}

//simply combines Exchange body values into an ArrayList<Object>
class ArrayListAggregationStrategy implements AggregationStrategy {

    public Exchange aggregate(Exchange oldExchange, Exchange newExchange) {
        Object newBody = newExchange.getIn().getBody();
        ArrayList<Object> list = null;
        if (oldExchange == null) {
            list = new ArrayList<Object>();
            list.add(newBody);
            newExchange.getIn().setBody(list);
            return newExchange;
        } else {
            list = oldExchange.getIn().getBody(ArrayList.class);
            list.add(newBody);
            return oldExchange;
        }
    }
}

The org.apache.camel.builder.AggregationStrategies is a builder that can be used for creating commonly used aggregation strategies without having to create a class.

Aggregate by grouping exchanges

In the route below we group all the exchanges together using GroupedExchangeAggregationStrategy:

from("direct:start")
    // aggregates all using the same expression and group the
    // exchanges, so we get one single exchange containing all
    // the others
    .aggregate(new GroupedExchangeAggregationStrategy()).constant(true)
    // wait for 0.5 seconds to aggregate
    .completionTimeout(500L).to("mock:result");

As a result we have one outgoing Exchange being routed to the "mock:result" endpoint. The exchange is a holder containing all the incoming Exchanges.

The output of the aggregator will then contain the exchanges grouped together in a list as shown below:

List<Exchange> grouped = exchange.getMessage().getBody(List.class);

Aggregating into a List

If you want to aggregate some value from the messages <V> into a List<V> then you can use the org.apache.camel.processor.aggregate.AbstractListAggregationStrategy abstract class.

The completed Exchange sent out of the aggregator will contain the List<V> in the message body.

For example, to aggregate a List<Integer> you can extend this class as shown below, and implement the getValue method:

public class MyListOfNumbersStrategy extends AbstractListAggregationStrategy<Integer> {

    @Override
    public Integer getValue(Exchange exchange) {
        // the message body contains a number, so return that as-is
        return exchange.getIn().getBody(Integer.class);
    }
}

The org.apache.camel.builder.AggregationStrategies is a builder that can be used for creating commonly used aggregation strategies without having to create a class.

The previous example can also be built using the builder as shown:

AggregationStrategy agg = AggregationStrategies.flexible(Integer.class)
    .accumulateInCollection(ArrayList.class)
    .pick(body());

Aggregating on timeout

If your aggregation strategy implements TimeoutAwareAggregationStrategy, then Camel will invoke the timeout method when the timeout occurs. Notice that the values for index and total parameters will be -1, and the timeout parameter will be provided only if configured as a fixed value. You must not throw any exceptions from the timeout method.

Aggregate with persistent repository

The aggregator provides a pluggable repository which you can implement your own org.apache.camel.spi.AggregationRepository.

If you need a persistent repository, then Camel provides numerous implementations, such as from the Caffeine, CassandraQL, EHCache, Infinispan, JCache, LevelDB, Redis, or SQL components.

Completion

When aggregation Exchanges at some point, you need to indicate that the aggregated exchanges are complete, so they can be sent out of the aggregator. Camel allows you to indicate completion in various ways as follows:

• completionTimeout: Is an inactivity timeout in that is triggered if no new exchanges have been aggregated for that particular correlation key within the period.

• completionInterval: Once every X period all the current aggregated exchanges are completed.

• completionSize: Is a number indicating that after X aggregated exchanges its complete.

• completionPredicate: Runs a Predicate when a new exchange is aggregated to determine if we are complete or not. The configured aggregationStrategy can implement the Predicate interface and will be used as the completionPredicate if no completionPredicate is configured. The configured aggregationStrategy can override the preComplete method and will be used as the completionPredicate in pre-complete check mode. See further below for more details.

• completionFromBatchConsumer: Special option for Batch Consumer, which allows you to complete when all the messages from the batch have been aggregated.

• forceCompletionOnStop: Indicates to complete all current aggregated exchanges when the context is stopped

• AggregateController: which allows to use an external source (AggregateController implementation) to complete groups or all groups. This can be done using Java or JMX API.

All the different completions are per correlation key. You can combine them in any way you like. It’s basically the first that triggers that wins. So you can use a completion size together with a completion timeout. Only completionTimeout and completionInterval cannot be used at the same time.

Completion is mandatory and must be configured on the aggregation.

Pre-completion mode

There can be use-cases where you want the incoming Exchange to determine if the correlation group should pre-complete, and then the incoming Exchange is starting a new group from scratch. The pre-completion mode must be enabled by the AggregationStrategy by overriding the canPreComplete method to return a true value.

When pre-completion is enabled then the preComplete method is invoked:

/**
 * Determines if the aggregation should complete the current group, and start a new group, or the aggregation
 * should continue using the current group.
 *
 * @param oldExchange the oldest exchange (is <tt>null</tt> on first aggregation as we only have the new exchange)
 * @param newExchange the newest exchange (can be <tt>null</tt> if there was no data possible to acquire)
 * @return <tt>true</tt> to complete current group and start a new group, or <tt>false</tt> to keep using current
 */
boolean preComplete(Exchange oldExchange, Exchange newExchange);

If the preComplete method returns true, then the existing correlation group is completed (without aggregating the incoming exchange (newExchange). Then the newExchange is used to start the correlation group from scratch, so the group would contain only that new incoming exchange. This is known as pre-completion mode.

The newExchange contains the following exchange properties, which can be used to determine whether to pre complete.

Property	Type	Description
CamelAggregatedSize	int	The total number of messages aggregated.
CamelAggregatedCorrelationKey	String	The correlation identifier as a String.

When the aggregation is in pre-completion mode, then only the following completions are in use:

• completionTimeout or completionInterval can also be used as fallback completions

• any other completions are not used (such as by size, from batch consumer, etc.)

• eagerCheckCompletion is implied as true, but the option has no effect

CompletionAwareAggregationStrategy

If your aggregation strategy implements CompletionAwareAggregationStrategy, then Camel will invoke the onComplete method when the aggregated Exchange is completed. This allows you to do any last minute custom logic such as to clean up some resources, or additional work on the exchange as it’s now completed. You must not throw any exceptions from the onCompletion method.

Completing the current group decided from the AggregationStrategy

The AggregationStrategy supports checking for the exchange property (Exchange.AGGREGATION_COMPLETE_CURRENT_GROUP) on the returned Exchange that contains a boolean to indicate if the current group should be completed. This allows to overrule any existing completion predicates / sizes / timeouts etc., and complete the group.

For example, the following logic will complete the group if the message body size is larger than 5. This is done by setting the exchange property Exchange.AGGREGATION_COMPLETE_CURRENT_GROUP to true.

public final class MyCompletionStrategy implements AggregationStrategy {
    @Override
    public Exchange aggregate(Exchange oldExchange, Exchange newExchange) {
        if (oldExchange == null) {
            return newExchange;
        }
        String body = oldExchange.getIn().getBody(String.class) + "+"
            + newExchange.getIn().getBody(String.class);
        oldExchange.getIn().setBody(body);
        if (body.length() >= 5) {
            oldExchange.setProperty(Exchange.AGGREGATION_COMPLETE_CURRENT_GROUP, true);
        }
        return oldExchange;
    }
}

Completing all previous group decided from the AggregationStrategy

The AggregationStrategy checks an exchange property, from the returned exchange, indicating if all previous groups should be completed.

This allows to overrule any existing completion predicates / sizes / timeouts etc., and complete all the existing previous group.

The following logic will complete all the previous groups, and start a new aggregation group.

This is done by setting the property Exchange.AGGREGATION_COMPLETE_ALL_GROUPS to true on the returned exchange.

public final class MyCompletionStrategy implements AggregationStrategy {
    @Override
    public Exchange aggregate(Exchange oldExchange, Exchange newExchange) {
        if (oldExchange == null) {
            // we start a new correlation group, so complete all previous groups
            newExchange.setProperty(Exchange.AGGREGATION_COMPLETE_ALL_GROUPS, true);
            return newExchange;
        }

        String body1 = oldExchange.getIn().getBody(String.class);
        String body2 = newExchange.getIn().getBody(String.class);

        oldExchange.getIn().setBody(body1 + body2);
        return oldExchange;
    }
}

Manually force the completion of all aggregated Exchanges immediately

You can manually trigger completion of all current aggregated exchanges by sending an exchange containing the exchange property Exchange.AGGREGATION_COMPLETE_ALL_GROUPS set to true. The message is considered a signal message only, the message headers/contents will not be processed otherwise.

You can alternatively set the exchange property Exchange.AGGREGATION_COMPLETE_ALL_GROUPS_INCLUSIVE to true to trigger completion of all groups after processing the current message.

Using a controller to force the aggregator to complete

The org.apache.camel.processor.aggregate.AggregateController allows you to control the aggregate at runtime using Java or JMX API. This can be used to force completing groups of exchanges, or query its current runtime statistics.

The aggregator provides a default implementation if no custom one has been configured, which can be accessed using getAggregateController() method. Though it may be easier to configure a controller in the route using aggregateController as shown below:

private AggregateController controller = new DefaultAggregateController();

from("direct:start")
   .aggregate(header("id"), new MyAggregationStrategy())
      .completionSize(10).id("myAggregator")
      .aggregateController(controller)
      .to("mock:aggregated");

Then there is API on AggregateController to force completion. For example, to complete a group with key foo:

int groups = controller.forceCompletionOfGroup("foo");

The returned value is the number of groups completed. A value of 1 is returned if the foo group existed, otherwise 0 is returned.

There is also a method to complete all groups:

int groups = controller.forceCompletionOfAllGroups();

The controller can also be used in XML DSL using the aggregateController to refer to a bean with the controller implementation, which is looked up in the registry.

When using Spring XML, you can create the bean with <bean> as shown:

<bean id="myController" class="org.apache.camel.processor.aggregate.DefaultAggregateController"/>

<camelContext xmlns="http://camel.apache.org/schema/spring">
    <route>
        <from uri="direct:start"/>
        <aggregate aggregationStrategy="myAppender" completionSize="10"
                   aggregateController="myController">
            <correlationExpression>
                <header>id</header>
            </correlationExpression>
            <to uri="mock:result"/>
        </aggregate>
    </route>
</camelContext>

There is also JMX API on the aggregator which is available under the processors node in the Camel JMX tree.

Aggregating with Beans

To use the AggregationStrategy you had to implement the org.apache.camel.AggregationStrategy interface, which means your logic would be tied to the Camel API. You can use a bean for the logic and let Camel adapt to your bean. To use a bean, then a convention must be followed:

• there must be a public method to use

• the method must not be void

• the method can be static or non-static

• the method must have two or more parameters

• the parameters are paired, so the first half is applied to the oldExchange, and the reminder half is for the newExchange. Therefore, there must be an equal number of parameters, e.g., 2, 4, 6, etc.

The paired methods are expected to be ordered as follows:

• the first parameter is the message body

• optional, the second parameter is a Map of the headers

• optional, the third parameter is a Map of the exchange properties

This convention is best explained with some examples.

In the method below, we have only two parameters, so the first parameter is the body of the oldExchange, and the second is paired to the body of the newExchange:

public String append(String existing, String next) {
    return existing + next;
}

In the method below, we have only four parameters, so the first parameter is the body of the oldExchange, and the second is the Map of the oldExchange headers, and the third is paired to the body of the newExchange, and the fourth parameter is the Map of the newExchange headers:

public String append(String existing, Map existingHeaders, String next, Map nextHeaders) {
    return existing + next;
}

And finally, if we have six parameters, that includes the exchange properties:

public String append(String existing, Map existingHeaders, Map existingProperties,
                     String next, Map nextHeaders, Map nextProperties) {
    return existing + next;
}

To use this with the aggregate EIP, we can use a bean with the aggregate logic as follows:

public class MyBodyAppender {

    public String append(String existing, String next) {
        return next + existing;
    }

}

And then in the Camel route we create an instance of our bean, and then refer to the bean in the route using bean method from org.apache.camel.builder.AggregationStrategies as shown:

private MyBodyAppender appender = new MyBodyAppender();

public void configure() throws Exception {
    from("direct:start")
        .aggregate(constant(true), AggregationStrategies.bean(appender, "append"))
            .completionSize(3)
            .to("mock:result");
}

We can also provide the bean class type directly:

public void configure() throws Exception {
    from("direct:start")
        .aggregate(constant(true), AggregationStrategies.bean(MyBodyAppender.class, "append"))
            .completionSize(3)
            .to("mock:result");
}

And if the bean has only one method, we do not need to specify the name of the method:

public void configure() throws Exception {
    from("direct:start")
        .aggregate(constant(true), AggregationStrategies.bean(MyBodyAppender.class))
            .completionSize(3)
            .to("mock:result");
}

And the append method could be static:

public class MyBodyAppender {

    public static String append(String existing, String next) {
        return next + existing;
    }

}

If you are using XML DSL, then we need to declare a <bean> with the bean:

<bean id="myAppender" class="com.foo.MyBodyAppender"/>

And in the Camel route we use aggregationStrategy to refer to the bean by its id, and the strategyMethodName can be used to define the method name to call:

<camelContext xmlns="http://camel.apache.org/schema/spring">
    <route>
        <from uri="direct:start"/>
        <aggregate aggregationStrategy="myAppender" aggregationStrategyMethodName="append" completionSize="3">
            <correlationExpression>
                <constant>true</constant>
            </correlationExpression>
            <to uri="mock:result"/>
        </aggregate>
    </route>
</camelContext>

When using XML DSL, you can also specify the bean class directly in aggregationStrategy using the #class: syntax as shown:

<route>
    <from uri="direct:start"/>
    <aggregate aggregationStrategy="#class:com.foo.MyBodyAppender" aggregationStrategyMethodName="append" completionSize="3">
        <correlationExpression>
            <constant>true</constant>
        </correlationExpression>
        <to uri="mock:result"/>
    </aggregate>
</route>

You can use this in XML DSL when you are not using the classic Spring XML files ( where you use XML only for Camel routes).

Aggregating when no data

When using bean as AggregationStrategy, then the method is only invoked when there is data to be aggregated, meaning that the message body is not null. In cases where you want to have the method invoked, even when there are no data (message body is null), then set the strategyMethodAllowNull to true.

When using beans, this can be configured a bit easier using the beanAllowNull method from AggregationStrategies as shown:

public void configure() throws Exception {
    from("direct:start")
        .pollEnrich("seda:foo", 1000, AggregationStrategies.beanAllowNull(appender, "append"))
            .to("mock:result");
}

Then the append method in the bean would need to deal with the situation that newExchange can be null:

public class MyBodyAppender {

    public String append(String existing, String next) {
        if (next == null) {
            return "NewWasNull" + existing;
        } else {
            return existing + next;
        }
    }

}

In the example above we use the Content Enricher EIP using pollEnrich. The newExchange will be null in the situation we could not get any data from the "seda:foo" endpoint, and a timeout was hit after 1 second.

So if we need to do special merge logic, we would need to set setAllowNullNewExchange=true. If we didn’t do this, then on timeout the append method would normally not be invoked, meaning the Content Enricher did not merge/change the message.

In XML DSL you would configure the strategyMethodAllowNull option and set it to true as shown below:

<camelContext xmlns="http://camel.apache.org/schema/spring">
    <route>
        <from uri="direct:start"/>
        <aggregate aggregationStrategy="myAppender"
                   aggregationStrategyMethodName="append"
                   aggregationStrategyMethodAllowNull="true"
                   completionSize="3">
            <correlationExpression>
                <constant>true</constant>
            </correlationExpression>
            <to uri="mock:result"/>
        </aggregate>
    </route>
</camelContext>

Aggregating with different body types

When, for example, using strategyMethodAllowNull as true, then the parameter type of the message bodies does not have to be the same. For example suppose we want to aggregate from a com.foo.User type to a List<String> that contains the name of the user. We could code a bean as follows:

public final class MyUserAppender {

    public List addUsers(List names, User user) {
        if (names == null) {
            names = new ArrayList();
        }
        names.add(user.getName());
        return names;
    }
}

Notice that the return type is a List which we want to contain the name of the users. The first parameter is the List of names, and the second parameter is the incoming com.foo.User type.