Routers (Flow Controls in Anypoint Studio) route messages to various destinations in a Mule flow. Some routers incorporate logic to analyze and possibly transform messages before routing takes place. For example, various flow controls can:
Split a message into several segments, then route each segment to a different building block
Combine several messages into a single message before sending it to the next building block in the flow
Reorder a list of messages before sending it to the next building block
Evaluate a message to determine which of several possible building blocks it should be routed to next
Broadcast the same message to multiple building blocks
Click a link in the Reference table below for details on a specific message processor.
Broadcast a message to multiple targets
Run a chain of message processors in a separate thread
Evaluates a message against specified criteria, then sends it to the first message processor that matches those criteria.
Checks the group tag (known as a Correlation ID) attached to each message in a group to create a collection of messages which share the same Correlation ID.
Accepts a collection of messages (or parts of messages), splits them into individual messages, then sends each new message, in sequence, to the next message processor in a flow.
Lets you write you own Java code to determine how messages are constructed and sent.
A custom-written message processor
Sends a message to the next message processor within a "circular" list of processor targets.
Filter out duplicate message by message ID
Filter out duplicate message by message content
Checks the group tag (Correlation ID) of each message in a collection, selects all the messages whose group tag matches the specified value, then combines those messages into a single message which is then sent to the next message processor in an application flow. This is particularly useful for re-assembling the segments of a long message that has been received as multiple messages, each one consisting of a segment of fixed length created and sent by the Message Chunk Splitter.
Sections a message into segments of a specified length, then sends each segment, in sequence, to the next message processor in a flow. This is particularly useful when the message recipient cannot accept messages longer than a specified length.
Filter messages using a filter
Create a message chain from multiple targets
Send a message to multiple connectors
Receive a message for asynchronous processing and accept the asynchronous response on a different channel
Accepts a collection of messages, then uses the Sequence ID of each message to reorder those messages. It then sends the messages (in order of their new sequence), to the next message processor in an application flow.
Iterates through a list of two or more message processors, sending successive messages to the next message processor on the list. When it reaches the end of the list, it jumps to the start of the list and resumes the iteration.
Repeatedly attempt to process a message until successful
Sends a request message to multiple targets concurrently. It collects the responses from all routes, and aggregates them into a single message.
Evaluates an expression which determines how it sections a message into two or more parts. The Splitter then sends each of these message parts, in sequence, to the next message processor in an application flow.
Send a message to an extra message processor as well as to the next message processor in the chain
The All message processor can be used to send the same message to multiple targets. Configuration for this router is as follows:
<all> <jms:endpoint queue="test.queue" transformer-refs="StringToJmsMessage"/> <http:endpoint host="10.192.111.11" transformer-refs="StringToHttpClientRequest"/> <tcp:endpoint host="10.192.111.12" transformer-refs="StringToByteArray"/> </all>
If any of the targets specified is an connector that has a filter configured on it, only messages accepted by that filter are sent to that connector.
All messages (if any) returned by the targets are aggregated together and form the response from this processor.
The Async message processor runs a chain of message processors in another thread, optionally specifying a threading profile for the thread to be used. The message processor is configured as follows:
<async> <append-string-transformer message="-async" /> <vm:outbound-endpoint path="async-async-out" exchange-pattern="one-way" /> <threading-profile doThreading="true" maxThreadsActive="16"/> </async>
This transforms the current message and sends it to the specified connector, using a threadpool that contains up to 16 concurrent threads.
The Choice message processor sends a message to the first message processor that matches. If none match and a message processor has been configured as "otherwise", the message is sent there. If none match and no otherwise message processor has been configured, an exception is thrown.
Choice is configured as follows:
<choice> <when expression="payload=='foo'" evaluator="groovy"> <append-string-transformer message=" Hello foo" /> </when> <when expression="payload=='bar'" evaluator="groovy"> <append-string-transformer message=" Hello bar" /> </when> <otherwise> <append-string-transformer message=" Hello ?" /> </otherwise> </choice>
If the message payload is "foo" or "bar", the corresponding transformer is run. If not, the transformer specified under "otherwise" is run.
The Collection Aggregator groups incoming messages that have matching group IDs before forwarding them. The group ID can come from the correlation ID or another property that links messages together, often this ID is first assigned when splitting a message with a Collection Splitter.
You can specify the
timeout attribute to determine how long the router waits in milliseconds for messages to complete the group. By default, if the expected messages are not received by the
timeout time, an exception is thrown and the messages are not forwarded. You can also set the
failOnTimeout attribute to
false to prevent the exception from being thrown and simply forward whatever messages have been received so far.
Configuration for the Collection Aggregator is as follows:
<collection-aggregator timeout="6000" failOnTimeout="false"/>
The Collection Aggregator pays attention to the following outbound properties in the mule messages that arrive to it:
MULE_CORRELATION_ID defines the ID of the batch to which the message belongs, so it knows what messages to group
MULE_CORRELATION_GROUP_SIZE defines the number of messages in the batch, so it knows when it has the complete set
MULE_CORRELATION_SEQUENCE is optional, and is added if you want to preserve the original order
The Collection Splitter acts on messages whose payload is a Collection type. It sends each member of the collection to the next message processor as separate messages. You can specify the attribute
enableCorrelation to determine whether a correlation ID is set on each individual message.
Configuration for the Collection Splitter is as follows:
A Custom Aggregator is an instance of a user-written class that aggregates messages. This class must implement the interface MessageProcessor. Often, it will be useful for it to subclass AbstractAggregator, which provides the skeleton of a thread-safe aggregator implementation, requiring only specific correlation logic. As with most custom objects in Mule, it can be configured either with a fully specified class name or as a reference to a Spring bean. It can also be configured with the same
failOnTimeout attributes described under Collection Aggregator.
Configuration for a Custom Aggregator is as follows:
<custom-aggregator failOnTimeout="true" class="com.mycompany.utils.PurchaseOrderAggregator"/>
A Custom Processor is an instance of a user-written class that acts as a message processor. This class must implement the interface MessageProcessor. As with most custom objects in Mule, it can be configured either with a fully specified class name or as a reference to a Spring bean.
Configuration for a Custom Processor is as follows:
The First Successful message processor iterates through its list of child message processors, routing a received message to each of them in order until one processes the message successfully. If none succeed, an exception is thrown.
Success is defined as:
If the child message processor thows an exception, this is a failure.
If the child message processor returns a message that contains an exception payload, this is a failure.
If the child message processor returns a message that does not contain an exception payload, this is a success.
If the child message processor does not return a message (e.g. is a one-way connector), this is a success.
This message processor was added in Mule 3.0.1.
<first-successful> <http:request config-ref="Config_port90" path="weather-forecast" method="GET" doc:name="HTTP"/> <http:request config-ref="Config_port91" path="weather-forecast" method="GET" doc:name="HTTP"/> <http:request config-ref="Config_port92" path="weather-forecast" method="GET" doc:name="HTTP"/> <vm:outbound-endpoint path="dead-letter-queue" /> </first-successful>
From 3.1.0 you can further customize the behavior of this router by specifying a 'failureExpression' that allows you to use Mule Expressions to define a failure. The failureExpression attribute is configured as follows:
<first-successful failureExpression="exception-type:java.net.SocketTimeoutException"> <http:request config-ref="Config_port90" path="weather-forecast" method="GET" doc:name="HTTP"/> <http:request config-ref="Config_port91" path="weather-forecast" method="GET" doc:name="HTTP"/> <http:request config-ref="Config_port92" path="weather-forecast" method="GET" doc:name="HTTP"/> <vm:outbound-endpoint path="dead-letter-queue" /> </first-successful>
In the above example a failure expression is being used to more exactly define the exception type that will be considered a failure, alternatively you can use any other Mule expression that can be used with expression filters, just remember that the expression denotes failure rather than success.
An idempotent filter checks the unique message ID of the incoming message to ensure that only unique messages are received by the flow. The ID can be generated from the message using an expression defined in the
idExpression attribute. By default, the expression used is
#[message:id], which means the underlying connector must support unique message IDs for this to work. Otherwise, a
UniqueIdNotSupportedException is thrown.
There is a simple idempotent filter implementation provided at org.mule.routers.IdempotentMessageFilter. The default implementation uses a simple file-based mechanism for storing message IDs, but you can extend this class to store the IDs in a database instead by implementing the ObjectStore interface.
Configuration for this router is as follows:
<idempotent-message-filter idExpression="#[message:id]-#[header:foo]"> <simple-text-file-store directory="./idempotent"/> </idempotent-message-filter>
idExpression attribute determines what should be used as the unique message ID. If this attribute is not used,
#[message:id] is used by default.
The nested element shown above configures the location where the received message IDs are stored. In this example, they are stored to disk so that the router can remember state between restarts. If the
directory attribute is not specified, the default value used is
mule.working.dir is the working directory configured for the Mule instance.
If no store is configured, the InMemoryObjectStore is used by default.
This filter calculates the hash of the message itself using a message digest algorithm to ensure that only unique messages are received by the flow. This approach provides a value with an infinitesimally small chance of a collision and can be used to filter message duplicates. Note that the hash is calculated over the entire byte array representing the message, so any leading or trailing spaces or extraneous bytes (like padding) can produce different hash values for the same semantic message content. Therefore, you should ensure that messages do not contain extraneous bytes. This router is useful when the message does not support unique identifiers.
Configuration for this filter is as follows:
<idempotent-secure-hash-filter messageDigestAlgorithm="SHA26"> <simple-text-file-store directory="./idempotent"/> </idempotent-secure-hash-filter>
Idempotent Secure Hash Message Filter also uses object stores, which are configured the same way as the Idempotent Message Filter. The optional
messageDigestAlgorithm attribute determines the hashing algorithm that will be used. If this attribute is not specified, the default algorithm SHA-256 is used.
After a splitter such as the Message Chunk Splitter splits a message into parts, the message chunk aggregator router reassembles those parts back into a single message. The aggregator uses the message’s correlation ID to identify which parts belong to the same message.
Configuration for the Message Chunk Aggregator is as follows:
<message-chunk-aggregator> <expression-message-info-mapping messageIdExpression="#[header:id]" correlationIdExpression="#[header:correlation]"/> </message-chunk-aggregator>
expression-message-info-mapping element allows you to identify the correlation ID in the message using an expression. If this element is not specified,
MuleMessage.getCorrelationId() is used.
The Message Chunk Aggregator also accepts the
failOnTimeout attributes as described under Collection Aggregator.
The Message Chunk Splitter allows you to split a single message into a number of fixed-length messages that will all be sent to the same message processor. It will split the message up into a number of smaller chunks according to the messageSize attribute that you configure for the router. The message is split by first converting it to a byte array and then splitting this array into chunks. If the message cannot be converted into a byte array, a RoutingException is raised.
A message chunk splitter is useful if you have bandwidth problems (or size limitations) when using a particular transport.
To put the chunked items back together again, you can use the Message Chunk Aggregator.
Configuration for the Message Chunk Splitter is as follows:
The Message Filter is used to control whether a message is processed by using a Idempotent Secure Hash Message Filter. In addition to the filter, you can configure whether to throw an exception if the filter does not accept the message and an optional message processor to send unaccepted messages to.
Configuration for the Message Filter is as follows:
<message-filter throwOnUnaccepted="false" onUnaccepted="rejectedMessageLogger"> <message-property-filter pattern="Content-Type=text/xml" caseSensitive="false"/> </message-filter>
A Processor Chain is a linear chain of message processors which process a message in order. A Processor Chain can be configured wherever a message processor appears in a Mule Schema. For example, to allow a WireTap to transform the current message before sending it off, you can configure the following:
<wire-tap> <processor-chain> <append-string-transformer message="tap" /> <vm:outbound-endpoint path="wiretap-tap" exchange-pattern="one-way" /> </processor-chain> </wire-tap>
The Recipient List message processor allows you to send a message to multiple connectors by specifying an expression that, when evaluated, provides the list of connectors. These messages can optionally be given a correlation ID, as in the Collection Splitter. An example is
<recipient-list enableCorrelation="ALWAYS" evaluator="header" expression="myRecipients"/>
which finds the list of connectors in the message header named
The Request Reply message processor receives a message on one channel, allows the back-end process to be forked to invoke other flows asynchronously, and accepts the asynchronous result on another channel.
Here is an example that uses the Request Reply message processor:
<flow name="main"> <vm:inbound-endpoint path="input"/> <request-reply storePrefix="mainFlow"> <vm:outbound-endpoint path="request"/> <vm:inbound-endpoint path="reply"/> </request-reply> <component class="com.mycompany.OrderProcessor"/> </flow> <flow name="handle-request-reply"> <vm:inbound-endpoint path="request"/> <component class="come.mycompany.AsyncOrderGenerator"/> </flow>
The request is received in the main flow and passed to the request-reply router, which implicitly sets the MULE_REPLYTO message property to the URL of its inbound connector (vm://reply) and asynchronously dispatches the message to the (one-way) vm://request connector, where it is processed by the handle-request-reply flow. The main flow then waits for a reply. The handle-request-reply flow passes the message to the AsynchOrderGenerator component. When this processing is complete, the message is sent to vm://reply (the value of the MULE_REPLYTO property.) The asynchronous reply is received and given to the OrderProcessor component to complete the order processing.
In more advanced cases, you might not want the automatic forwarding of the second flow’s response to the request-reply inbound connector. For instance, the second flow might trigger the running of a third flow, which then generates and sends the reply. In these cases, you can remove the MULE_REPLYTO property with a Message Properties Transformer:
<request-reply storePrefix="mainFlow"> <vm:outbound-endpoint path="request"> <message-properties-transformer scope="outbound"> <delete-message-property key="MULE_REPLYTO"/> </message-properties-transformer> </vm:outbound-endpoint> <vm:inbound-endpoint path="reply"/> </request-reply>
The Resequencer sorts a set of received messages by their correlation sequence property and issues them in the correct order. It uses the
failOnTimeout attributes described in Collection Aggregator to determine when all the messages in the set have been received.
The Resequencer is configured as follows:
<resequencer timeout="6000" failOnTimeout="false"/>
The Round Robin message processor iterates through a list of child message processors in round-robin fashion: the first message received is routed to the first child, the second message to the second child, and so on. After a message has been routed to each child, the next is routed to the first child again, restarting the iteration.
This message processor was added in Mule 3.0.1.
<round-robin> <http:request config-ref="Config_port90" path="weather-forecast" method="GET" doc:name="HTTP"/> <http:request config-ref="Config_port91" path="weather-forecast" method="GET" doc:name="HTTP"/> <http:request config-ref="Config_port92" path="weather-forecast" method="GET" doc:name="HTTP"/> </round-robin>
A Splitter uses an expression to split a message into pieces, all of which are then sent to the next message processor. Like other splitters, it can optionally specify non-0default locations within the message for the message ID and correlation ID.
The Splitter is configured as shown below:
<splitter expression="#[xpath3('//acme:Trade')]" doc:name="Splitter"/>
This uses the specified XPath expression wrapped inside a MEL expression to find a list of nodes in the current message and sends each of them as a separate message.
The WireTap message processor allows you to route certain messages to a different message processor as well as to the next one in the chain. For instance, To copy all messages to a specific connector, you configure it as an outbound connector on the WireTap routing processor:
<wire-tap> <vm:outbound-endpoint path="tapped.channel"/> </wire-tap>
The WireTap routing processor is useful both with and without filtering. If filtered, it can be used to record or take note of particular messages or to copy only messages that require additional processing. If filters aren’t used, you can make a backup copy of all messages received. The behavior here is similar to that of an interceptor, but interceptors can alter the message flow by preventing the message from reaching the component. WireTap routers cannot alter message flow but just copy on demand. In this example, only messages that match the filter expression are copied to the vm connector.
<wire-tap> <vm:outbound-endpoint path="tapped.channel"/> <wildcard-filter pattern="the quick brown*"/> </wire-tap>