As you analyze the six requirements listed earlier, you can see that they are all integration concerns. For example, in the consumer pre-screening step, we need to gather additional information about the consumer and the consumer’s desires and enrich the loan request with additional meta-information. We then have to filter such information to select the most appropriate list of banks and so on. Enrich, filter, and select are all integration concerns for which EIP defines a solution in the form of patterns. Spring Integration provides an implementation of these patterns.

The following image shows a representation of a messaging gateway:

The messaging gateway pattern provides a simple mechanism to access messaging systems, including our loan broker. In Spring Integration, you can define the gateway as a plain old java interface (you need not provide an implementation), configure it with the XML <gateway> element or with an annotation in Java, and use it as you would any other Spring bean. Spring Integration takes care of delegating and mapping method invocations to the messaging infrastructure by generating a message (the payload is mapped to an input parameter of the method) and sending it to the designated channel. The following example shows how to define such a gateway with XML:

Our current gateway provides two methods that could be invoked. One that returns the best single quote and another one that returns all quotes. Somehow, downstream, we need to know what type of reply the caller needs. The best way to achieve this in messaging architecture is to enrich the content of the message with some metadata that describes your intentions. Content Enricher is one of the patterns that addresses this. Spring Integration does, as a convenience, provide a separate configuration element to enrich message headers with arbitrary data (described later) However, since the gateway element is responsible for constructing the initial message, it includes ability to enrich the newly created message with arbitrary message headers. In our example, we add a RESPONSE_TYPE header with a value of BEST whenever the getBestQuote() method is invoked. For other methods, we do not add any header. Now we can check downstream for the existence of this header. Based on its presence and its value, we can determine what type of reply the caller wants.

Based on the use case, we also know tat some pre-screening steps need to be performed, such as getting and evaluating the consumer’s credit score, because some premiere banks only accept quote requests from consumers that meet a minimum credit score requirement. So it would be nice if the message would be enriched with such information before it is forwarded to the banks. It would also be nice if, when several processes need to be completed to provide such meta-information, those processes could be grouped in a single unit. In our use case, we need to determine the credit score and, based on the credit score and some rule, select a list of message channels (bank channels) to which to send quote request.

The composed message processor pattern describes rules around building endpoints that maintain control over message flow, which consists of multiple message processors. In Spring Integration, the composed message processor pattern is implemented by the <chain> element.

The following image shows the chain pattern:

The preceding image shows that we have a chain with an inner header-enricher element that further enriches the content of the message with the CREDIT_SCORE header and the value (which is determined by the call to a credit service — a simple POJO spring bean identified by 'creditBureau' name). Then it delegates to the message router.

The following image shows the message router pattern:

Spring Integration offers several implementations of the message routing pattern. In this case, we use a router that determines a list of channels based on evaluating an expression (in Spring Expression Language) that looks at the credit score (determined in the previous step) and selects the list of channels from the Map bean with an id of banks whose values are premier or secondary, based on the value of credit score. Once the list of channels is selected, the message is routed to those channels.

Now, one last thing the loan broker needs to receive the loan quotes form the banks, aggregate them by consumer (we do not want to show quotes from one consumer to another), assemble the response based on the consumer’s selection criteria (single best quote or all quotes) and send the reply to the consumer.

The following image shows the message aggregator pattern:

An aggregator pattern describes an endpoint that groups related messages into a single message. Criteria and rules can be provided to determine an aggregation and correlation strategy. Spring Integration provides several implementations of the aggregator pattern as well as a convenient namespace-based configuration.

The following example shows how to define an aggregator:

Our Loan Broker defines a 'quotesAggregator' bean with the <aggregator> element, which provides a default aggregation and correlation strategy. The default correlation strategy correlates messages based on the correlationId header (see the correlation identifier pattern in the EIP book). Note that we never provided the value for this header. It was automatically set earlier by the router, when it generated a separate message for each bank channel.

Once the messages are correlated, they are released to the actual aggregator implementation. Although Spring Integration provides a default aggregator, its strategy (gather the list of payloads from all messages and construct a new message with this list as its payload) does not satisfy our requirement. Having all the results in the message is a problem, because our consumer might require a single best quote or all quotes. To communicate the consumer’s intention, earlier in the process we set the RESPONSE_TYPE header. Now we have to evaluate this header and return either all the quotes (the default aggregation strategy would work) or the best quote (the default aggregation strategy does not work because we have to determine which loan quote is the best).

In a more realistic application, selecting the best quote might be based on complex criteria that might influence the complexity of the aggregator implementation and configuration. For now, though, we are making it simple. If the consumer wants the best quote, we select a quote with the lowest interest rate. To accomplish that, the LoanQuoteAggregator class sorts all the quotes by interest rate and returns the first one. The LoanQuote class implements Comparable to compare quotes based on the rate attribute. Once the response message is created, it is sent to the default reply channel of the messaging gateway (and, thus, to the consumer) that started the process. Our consumer got the loan quote!

In conclusion, a rather complex process was assembled based on POJO (that is existing or legacy) logic and a light-weight, embeddable messaging framework (Spring Integration) with a loosely coupled programming model intended to simplify integration of heterogeneous systems without requiring a heavy-weight ESB-like engine or a proprietary development and deployment environment. As a developer, you should not need to port your Swing or console-based application to an ESB-like server or implement proprietary interfaces just because you have an integration concern.

This sample and the other samples in this section are built on top of Enterprise Integration Patterns. You can consider them to be “building blocks” for your solution. They are not intended to be complete solutions. Integration concerns exist in all types of application (whether server-based or not). Our goal is to make is so that integrating applications does not require changes in design, testing, and deployment strategy.