Contact Us 1-800-596-4880

Migrating Splitter/Aggregator

Splitters are not longer available in Mule 4. To process a collection of elements sequentially, use the For Each component.

Aggregators remain in Mule 4 but as a separate module with very different behavior.

Splitters

The main difference between Mule 3 splitters and the Mule 4 For Each component is that the foreach element is a scope, meaning that the components used to process the iterated elements must be defined within foreach.

Mule 3 Example
<flow name="flow">
    ...
    <collection-splitter/>
    <component-1/>
    <component-2/>
    ...
    <component-n>
</flow>
Mule 4 Example
<flow name="flow">
    ...
    <foreach>
        <component-1/>
        <component-2/>
        ...
        <component-n>
    </foreach>
</flow>

Also, the foreach element does not return a collection with the results of all the processed values. Instead, it returns the same collection received as input. So, the behavior of previous examples is not the same.

It does not matter if a Mule 3 aggregator is present. To store the results of the processed values, you need to use a Mule 4 aggregator.

Simple Migration

The default format of the migration should be:

Mule 3 Example
<flow name="flow">
    ...
    <collection-splitter/>
    <component-1/>
    <component-2/>
    ...
    <component-n>
    <collection-aggregator/>
</flow>
Mule 4 Example
<flow name="flow">
    ...
    <set-variable variableName="collection-splitter0-group-size" value="#[sizeOf(payload)]"/>
    <foreach>
        <component-1/>
        <component-2/>
        ...
        <component-n>
        <aggregators:group-based-aggregator name="collection-splitter0-aggregator"
                                            groupSize="#[vars.collection-splitter0-group-size]"
                                            evictionTime="0">
            <aggregators:aggregation-complete>
                <set-variable variableName="aggregation0" value="#[payload]"/>
            </aggregators:aggregation-complete>
        </aggregators:group-based-aggregator>
    </foreach>
    <set-payload value="#[vars.aggregation0]"/>
</flow>

In the example:

  • A variable collection-splitter0-group-size should be created to store the size of the collection. This is needed because inside the foreach scope, the payload will correspond to the element from the collection being processed.

  • Every component between the collection-splitter and collection-aggregator is wrapped inside a foreach scope.

  • A group-based-aggregator is added at the end of the foreach scope to store all the results. The size of the aggregation expected will correspond to the value stored in the collection-splitter0-group-size variable. set-variable is added inside the aggregation-complete to store the result of the aggregation once it is complete.

  • Since the payload after the foreach element will be the same as the input received, set-payload should be used to set it as the stored value in the variable defined insider the aggregation-complete route.

Complex Migration

In some cases, the Mule 3 configuration with splitters and aggregators might set more parameters, so the migration will be a bit more complicated to achieve the Mule 3 behavior in Mule 4.

Mule 3 aggregators allowed for 2 attributes to be configured: timeout and failOnTimeout. If the timeout attribute is present (with a value greater than 0), the migration should be as follows in the examples.

Even when there is a way of adding extra logic when there is a timed-out aggregation and failOnTimeout is set to true, the behavior will not be the same as in Mule 3. Aggregators in Mule 4 do not raise any errors or dispatch any notifications where there is a timed-out aggregation.
Mule 3 Example
<flow name="flow">
    ...
    <collection-splitter/>
    <component-1/>
    <component-2/>
    ...
    <component-n>
    <collection-aggregator timeout="1000" failOnTimeout="false"/>
</flow>
Mule 4 example
<vm:config name="splitter-aggregator-vm-config">
    <vm:queues>
        <vm:queue queueName="collection-splitter0-vm-queue"/>
    </vm:queues>
</vm:config>
...
<flow name="flow">
    ...
    <set-variable variableName="collection-splitter0-group-size" value="#[sizeOf(payload)]"/>
    <set-variable variableName="collection-splitter0-aggregator-complete-aggregation" value="#[false]"/>
    <foreach>
        <component-1/>
        <component-2/>
        ...
        <component-n>
        <aggregators:group-based-aggregator name="collection-splitter0-aggregator"
                                            groupSize="#[vars.collection-splitter0-group-size]"
                                            timeout="1000"
                                            timeoutUni="MILLISECONDS"
                                            evictionTime="0">
            <aggregators:aggregation-complete>
                <set-variable variableName="collection-splitter0-aggregator-complete-aggregation" value="#[true]"/>
            </aggregators:aggregation-complete>
        </aggregators:group-based-aggregator>
    </foreach>
    <vm:consume queueName="collection-splitter0-vm-queue" config-ref="splitter-aggregator-vm-config"/>
    <choice>
        <when expression="#[vars.collection-splitter0-aggregator-complete-aggregation == false]">
            <!--place to add logic for when an aggregation timed out and the failOnTimeout flag was true-->
        </when>
    </choice>
</flow>

<flow name="collection-splitter0-aggregator-listener-flow">
    <aggregators:aggregator-listener aggregatorName="collection-splitter0-aggregator" includeTimedOutGroups="true"/>
    <vm:publish config-ref="splitter-aggregator-vm-config" queueName="collection-splitter0-vm-queue"/>
</flow>

In the example:

  • A variable collection-splitter0-group-size should be created to store the size of the collection. This is needed because inside the foreach scope, the payload will correspond to the element from the collection being processed.

  • Another variable collection-splitter0-aggregator-complete-aggregation should be created if there is a need for storing an indication of whether or not the aggregation terminated due to a timeout or a complete processing.

  • Every component between the collection-splitter and collection-aggregator is wrapped inside a foreach scope.

  • A group-based-aggregator is added at the end of the foreach scope to store all the results. The expected size of the aggregation will correspond to the value stored in the collection-splitter0-group-size variable. The timeout for the aggregation should be set to the timeout that the Mule 3 aggregator contained. set-variable is added inside the aggregation-complete to set that the aggregation was complete. (That route is only executed with a complete aggregation).

    Another flow should be created and an aggregator:aggregator-listener should be added as source. All aggregations (complete or not) will trigger that listener. Then, they will be sent to the main flow via a vm:publish operation.

  • A vm:consume operation is added after the foreach element to wait for the aggregation. Once a value is read, the flow will continue with its execution.

  • Because the collection-splitter0-aggregator-complete-aggregation will only be set to true if the aggregation completes, additional logic can be added to check the value of that variable (with choice and when). This can simulate logic that was previously added in a Mule 3 error handler for when a timeout error was received.

Additional Attributes

All splitters allowed for the configuration of the enableCorrelation attribute.

There is no way to migrate an enableCorrelation="NEVER" configuration. A correlation ID will always be set to the events processed from the collection by the foreach element.

Expression splitters (<splitter/>) allowed for the configuration of two attributes: evaluator and custom-evaluator. Expression evaluators can’t be configured anymore; only DataWeave can be used.

Available Splitters Migration

Mule 3 had the following splitters, some of which require some extra logic to be migrated or cannot be migrated at all.

  • collection-splitter: As defined in the main example.

  • custom-splitter: The use of custom classes is not allowed in Mule 4 anymore so this splitter can’t be migrated.

  • splitter (expression splitter) : The expression used should be set as the collection attribute for the foreach element.

    • Mule 3: <splitter expression="#[payload.someKey]"/>

    • Mule 4: <foreach collection="#[payload.someKey]">

  • map-splitter : The DataWeave expression #[dw::core::Objects::entrySet(payload)] should be used as the collection attribute in the foreach element.

    • Mule 3: <map-splitter/>

    • Mule 4: <foreach collection="#[dw::core::Objects::entrySet(payload)]">

  • message-chunk-splitter: There is no similar behavior in Mule 4. It cannot be migrated.

Aggregators

Most parameters allowed by Mule 3 aggregators are not allowed in Mule 4:

  • timeout: Though the timeout parameter can be set in any new aggregator, there are some differences. First, timeoutUnit should also be set to specify the unit of time. Also, the behavior of aggregation timeouts is very different in a Mule 4 aggregator. This is explained in Aggregators Module Reference.

  • failOnTimeout: No longer available. However similar behavior can be achieved, as explained in the Complex Migration.

  • processed-groups-object-store-ref: Processed groups are no longer stored in a separate object store. This paremeter is not allowed anymore.

  • event-groups-object-store-ref: You can set this object store using the objectStore attribute in Mule 4 aggregators.

  • persistentStores: Not allowed anymore. If a persistent object store is needed, the objectStore attribute should be the name of a persistent object store.

  • storePrefix: Not allowed anymore.

Available Aggregators Migration

  • collection-aggregator: Should be migrated as defined in the main example.

  • custom-aggregator: Cannot be migrated. There is no way in Mule 4 to add custom implementations for an interface.

  • message-chunk-aggregator: There is no similar behavior in Mule 4. It cannot be migrated.