Tuesday, November 11, 2014

Overview of CXF Transformation Feature
The CXF Transformation feature allows you to declaratively define a transformation to change namespaces or append/change/drop elements. The transformation is very much tied to the CXFEndpoint.

Intended Use-cases 
The use-cases for which the CXF transformation feature works well are the following use-cases.
  1. Transforming Incoming Requests from clients - A route exposes an interface to clients via a WSDL.  Some clients may not be capable of sending messages (or receive response) that adhere to the interface defined by the route’s WSDL. The CXF transform feature will allow you to define transformations on the CXFEndpoint so that incoming non-conforming messages from clients can be transformed and “fixed” to adhere to the expected interface defined by the WSDL. This allows clients to send non-conforming messages to the route (e.g. incorrect namespace or element names) – messages that would otherwise cause an error when sent to the route - and you can define the appropriate transformation on the CXFEndpoint to “fix” the incoming requests to conform to WSDL, and likewise modify outgoing (well-formed) response to something client expects.
  2.  Transforming Outgoing Requests to backend - A route sends requests to backend service that are well-formed according to a specific WSDL, but the backend service expects some tweak to what the route is sending. This could apply for example, in the case where the route’s CXFEndpoint (client) is ahead of the backend service on changes to element names or namespaces that adhere to an enterprise standard. At some point the backend will be updated to accept these changes. However until the backend service is conformant, the transformation applied to the CXFEndpoint (client) in the route will tweak the outgoing messages so that they can be understood by the backend service, and likewise the responses from the backend service can be transformed as needed into the well-formed format according to the enterprise WSDL.
Transforming Incoming Requests from clients
Let's say we want our route to expose the interface defined by a new "enterprise" WSDL, and we want to go ahead and deploy that. Even if there are clients that are not ready to adhere to the new WSDL, clients can continue to send requests that adhere to the "old" WSDL (e.g. operation names, different namespace, different element names, etc.)  Existing clients can continue to send requests containing non-conforming messages.
In our new route, we just need to include a mapping in the spring file that defines the transformation on incoming messages. So as long as we add the transformation rules into our route, it is able continue to service clients that don't necessarily adhere to the new WSDL, and (if needed) the route will modify the messages as defined in the transformation.
Below shows the Spring configuration a route that exposes the interface defined in ‘deqy.wsdl’.  Imagine this is our “new” interface. If there are clients that must continue to call our route that can’t change to the new WSDL, then the route can be configured (via the Transformation mapping) to accommodate those existing clients.

The above code snippet is a ‘camel-context.xml’ (Spring configuration) for a Camel route that listens on a CXF endpoint. The CXF endpoint points to a WSDL that defines the route’s interface. The route includes a CXF transformation map that declaratively defines mappings that should be done to non-conforming messages as they enter the route. Note the mappings are defined in such a way that they won’t affect messages that do adhere to the WSDL.
The above Spring defines a route starting at (line 43) that begins with a CXF endpoint (listener). The CXF endpoint is defined at lines 32-38 with the transform feature being configured starting at line 35.  The individual transformations are defined at lines (20-29.)
As you can see it is pretty straightforward to configure a route to accommodate for payload transformation to be applied to incoming messages. Likewise it is just as easy to apply a transformation for the corresponding response back to client that will convert the compliant response from the route into a (potentially non-conformant) response that the client expects.
Transforming Outgoing Requests to backend
For a scenario where a route sends a request to a backend server, you may have written your route CXFEndpoint (client) that adheres to a particular WSDL but the backend service for some reason or another needs the message to be tweaked slightly.  To apply a transformation to a request message before it is written to the wire and a response before it gets dispatched to the route you can use an outgoing transform.
The following code shows the implementation of an outgoing transform feature.
Lines 29-42 define the transform mapping for ‘outTranformElements’ – the mapping on requests as they are sent from route to the backend service, followed by ‘inTransformElements’ – the mapping on response before the message re-enters the route.
See lines 44-51 for the CXFEndpoint ‘deqyService’, where the transform feature is applied (lines 48-50).
The route starts at line 54. The route reads a message from a file in ‘inbox’ directory, and sends it to the backend service. The route receives the response, and writes response to a file in ‘outbox’ directory.
Consider a test where the message content of the file dropped into ‘inbox’ conforms to the  WSDL defined on the outgoing CXFEndpoint – i.e. ‘deqy.wsdl’.    Imagine the backend service requires a tweak to a well-structured request. For example the namespace the backend service expects in incoming request and writes to response is different from what is defined in the WSDL. We can apply any such tweak required by backend service using a transform defined on the endpoint.  In this scenario the well-formed message we drop into inbox will get transformed just before it is sent to the backend service based on the transform defined in ‘outTranformElements’. Likewise, the response from backend service gets transformed based on the ‘inTransformElements’ before the response message re-enters the route.
A Workable Solution using CXF transform Feature
The following diagram shows how a message flows through a route that consists of incoming and outgoing CXF Endpoints. Both incoming and outgoing CXF endpoints have transforms defined.  The location in the flow where the transforms are executed is shown in the diagram.
Consider a client that is written based on the interface defined by Service ‘A’ WSDL.  The client sends a request to the route using the format based on Service ‘A’ WSDL. The incoming ‘inTransform’ (see position 1 in diagram) will convert messages from the (non-conforming) format sent by clients, into a format adhering to the Service ‘B’ WSDL. Without this transform, the incorrectly formatted message sent from client would cause an operation/binding mis-match when it hit the route’s CXFEndpoint that begins the route, and this would cause the request to fail on entry to the route.
The message continues through the route, through Processor A and B and then reaches the consumer CXFEndpoint. This CXFEndpoint will call the backend Service ‘B’. The format of the message is correct (adheres to Service ‘B’ WSDL) and Service ‘B’ will respond with correctly formatted message. In this solution, the ‘outTransform’ and ‘inTransform’ at position 2 and 3 are not required.
The response from Service ‘B’ flows back into the route, and the response goes through Processor ‘C’ where a camel-based transformation (e.g. xquery, XSLT, java code) can be used to transform/enrich the response in any way required. This transformed message can now be returned back to client. The ‘outTransform’ (position 4) is not necessary in this solution.
Using CXF in Generic Provider Mode
One aspect of the CXF transformation feature that is problematic is the requirement that you are restricted to performing transformations that convert message into something that conforms with the WSDL the endpoint is tied to. If the message is transformed in a way that makes it non-conformant, an error is thrown when it reaches the endpoint.
In order to make the use of CXF transform easier within Camel, there is a way to disable the requirement that the incoming message match the operation/binding of the associated CXFEndpoint. That way, if this message-level validation is disabled on the Endpoint you can freely apply tranformations to any format at the endpoint and it will continue to pass through the endpoint without failure.
You should be careful when over-riding this as the purpose of a WSDL is to do the operation checking and some special handling based on the determined operation and also to provide a way for the consumer to get the typed interface information (i.e., retrieving the WSDL from the endpoint). By disabling, you will do neither of these, so you could just simply set up your endpoint using the generic provider mode (i.e., omitting the wsdlURL and serviceClass properties of the endpoint).
In this case, the message will be transformed using CXF's transform feature and directly transferred to your camel route where you can do further modification necessary and set the operationName header to match the operation at the outbound endpoint
In this way, you can allow messages to be transformed using CXF transform feature without the problem of adhering to a particular interface.
The transformation capabilities of the CXF transform mapping language are limited. The transform language works well for minimal changes such as element name changes or namespace changes. More involved e.g. structural changes to complex types or multi-faceted elements are not easily implemented using the mapping language.
A CXF transform feature is applied to a specific CXFEndpoint in a route. Because the transform is tightly tied to the specific CXFEndpoint, it cannot be used to apply a transformation at random spots on a message flow through the route. This limits the usage of the CXF transform feature as a general mapping tool for transformations within a camel route.
In order to make the CXF transform feature a useful technique for performing general transformations to messages as they flow through a route, you will likely need to disable the opearation validation of messages at the CXFEndpoint and use the generic provider mode. This will allow for further modifications in the transform feature than would normally be possible if you used the default strictly-typed message checks that typically are part of the message entry into the route.

No comments:

Post a Comment