Geeks With Blogs

News View Michael Stephenson's profile on BizTalk Blog Doc View Michael Stephenson's profile on LinkedIn
Michael Stephenson keeping your feet on premise while your heads in the cloud

The next in my series on BizTalk testing will cover testing pipeline components.  I will begin this post by discussing some of the traditional ways I have seen pipeline components tested, then continue to 2 more recent techniques which I believe to offer significant advantages.  To begin with the traditional techniques are: 

Traditional Approach 1 - Testing as part of a larger process

In this technique the Pipeline component is developed and then deployed along with a BizTalk solution.  Tests are then conducted against the overall process and it is assumed that if the end to end test is successful then the pipeline component has been adequately tested.

The key points about this approach are:

  • Often problems with the pipeline component are not detected during development because the end to end test does not cover all cases in the component
  • It is difficult to obtain code coverage information for the pipeline component
  • It is time consuming as it requires a deployment to BizTalk to be able to test
  • It is error prone because often you forget to restart host processes and think you havent fixed something that you really have
  • The component has limited reusability as it is only tested within the context of this process

Traditional Approach 2 - Using abstraction to make the component more testable

In this technique you find the developer has abstracted the logic using the facade pattern which the pipeline component then uses.  This means the code in the pipeline component is as simple as possible.  The more complex code is in other classes which do not depend on the BizTalk classes and interfaces such as IBaseMessage.  This in turn makes these classes easier to test outside BizTalk.

I think this pattern in general isnt a bad thing.  but I often see this technique used in conjunction with technique 1.  So we end up with a situation where the underlying classes are tested with unit tests and the pipeline component itself is assumed as tested as part of the larger process.  The key points of this technique are:

  • It is better than technique 1 as we are performing some unit tests which validates most of the functionality of the component before BizTalk becomes involved
  • We still cant test the pipeline component interface without having to deploy to BizTalk
  • Most of the other points from technique 1 still apply

Traditional Approach 3 - Using Pipeline.exe

I sometimes have seen the technique where a developer will create the pipeline component and then some pipelines.  The developer will then use Pipeline.exe to execute the test cases. 

The key points of this approach are as follows:

  • This does not require the artifacts to be deployed to BizTalk
  • It requires additional BizTalk pipelines to be defined to test the pipeline component
  • This tool needs to be used from the command line (although you could use the process object to call it from a C# test)
  • When using this approach you would probably want to validate the output document from the pipeline.exe call to ensure the message is as expected
  • You cant really interact with the message context before or after the test so this might limit your testing capability or require additional components needing to be added to the pipeline.

The challenge of the traditional approaches

The main challenge which limited the traditional approach to how you would test pipeline components was the ability for the developer to create and setup the IBaseMessage and IPipelineContext objects which you would then use for testing.  This resulted in the above 3 approaches being (in my opinion) the most popular way of testing pipeline components.

As result developers were often making their best effort at being able to test pipeline components but always knowing that they could only effectively test so much and there was always a reasonable chance that the component is going to have problems when used.

Newer Approachs

As with previous posts in this series I'm trying to encourage the following desired practices when testing:

  • We want to make testing the component relatively simple
  • We want to test the component as much as possible before we start using it in BizTalk
  • We want the tests to be automated and part of a continuous integration process

In order to implement the approach to testing pipeline components I would recommend either of the following 2 techniques (outlined below) when testing pipeline components.  Before I discuss the two techniques some background on the sample (available for download at the bottom of the article):

The sample contains a simple pipeline component which will read the input message using the XPathMutatorStream.  When it finds an element matching the desired XPath query the value from this element will be promoted to the File.ReceivedFileName promoted property.

The sample pipeline component is intended to be a fairly straightforward component which can be used to demonstrate how to test a component.  The following picture shows the main part of the pipeline component:

In the tests project there are 2 test classes each one demonstrating each technique.

Approach 1 - Testing with the Pipeline Component Test Library

The Pipeline Component Test Library has been around for a little while, but I dont think its used as much as it should be.  The library basically provides a simpler API to the PipelineObjects.dll which comes with the Pipeline.exe tool in the SDK.

This means you can interact with the Pipeline.exe type facilities in a simpler way directly from your C# test.  You can also access the message and its context much easier than you would be able to by using Pipeline.exe.  The following picture shows the code snippet which forms the test of the pipeline component using the pipeline component test library:


 In the test you can see you use the Pipeline Library to help tackle the key challenges of the IBaseMessage and the IPipelineContext.  In terms of the message you use the libraries helpers to create this message from an input document.  For the IPipelineContext this is handled by the library internally because you are creating a pipeline in code to execute the component in.

The advantages of this technique are:

  • You have full access to the proper IBaseMessage before the test.  This lets you remove the dependancy on things like components before yours in a pipeline or adapters because you can do things like set properties yourself.
  • The technique uses objects that a BizTalk person will be familiar with so the learning curve is not that steep
  • The tests can be developed very quickly
  • You control the pipeline so you can add additional components as required

With this technique it allows you to treat the pipeline component like a black box.  You put a message in and check the message and context that comes out. 

Useful Resources:

Some useful resources on this technique are:

Tomas Restrepo  - Creator of the Pipeline Component Test Library

Nick Heppleson - Has an article on how he tests his Message Archive component using this technique

Approach 2 - Testing with Rhino Mocks

In approach 2 i'm going to demonstrate how you can use a mocking framework to help you test the pipeline component.  In this example I am using Rhino Mocks.  In this technique you are basically defining a dynamic mock for the objects which will be used by the pipeline component.  On the mock objects you set expectations for what should happen each time a method is called on the mock object.  You then execute the pipeline component and then verify that all of the expectations happened as you planned.

The below code sample shows the equivelent test implemented using Rhino Mocks.



The advantage of this technique is:

  • It is a very powerful technique which gives you full control over pretty much all of the objects
  • It is a technique which is common to C# developers
  • Encourages the developer to think more about the component
  • Again this does not require the code to be deployed to BizTalk

This technique is much more white box, requiring the developer to have a much more intermate knowledge of what the component is doing when creating the test or as we are all test driven developers this makes you think a little harder about what the component does internally.   

Useful resources:

For more info on BizTalk and Rhino Mocks check out the following (click here)


 I think the key differences between the pipeline component test library and Rhino Mock techniques are as follows (i will refer to the pipeline component test library as PCTL):

  • The PCTL offers a technique which has a shallower learning curve and will be familiar to most BizTalk developers
  • Rhino mocks offers probably more control over things for very complicated tests
  • The PCTL is a much quicker way of developing tests, i find that using Rhino Mocks is quite time consuming in working out all of the expectations (especially when you are new to the technique)
  • It would be easier to refactor PCTL tests when there are changes to your component
  • In my opinion the PCTL just gives me a little more confidence than Rhino Mocks.  This is mainly because the test technique gives me the gut feeling that it is performing like how it will in BizTalk.  Where as with Rhino Mocks it sometimes feels that there is a bit of a gap between the mocking and what will happen when it is in BizTalk.  I dont really have any hard evidence to back this up but I think the fact that the tests themselves are that bit more complicated to write that they almost need testing in their own right.

So based on this article I would make the following recommendations for your approach to testing pipeline components:

  1. Use the traditional abstraction technique anyway as this is a pattern that can make your component simpler to understand and test
  2. As a default technique use the Pipeline Component Test Library
  3. When you have a special case or unusual component that has advanced testing requirements, compliment the Pipeline Component Test Library tests with ones which use Rhino Mocks to help you do those more advanced things
  4. Use a code coverage tool to ensure you dont miss any tests
  5. Remember to test more than just the core interface such as IComponent as the rest of the code needs testing too!

The sample code is available below:



Posted on Sunday, March 30, 2008 12:14 AM BizTalk , BizTalk Testing | Back to top

Comments on this post: BizTalk Testing Series - Testing Pipeline Components

# re: BizTalk Testing Series - Testing Pipeline Components
Requesting Gravatar...
where is the source for this? the link does nothing but point to the root of your blog.
Left by Ryan CrawCour on Feb 11, 2009 2:06 AM

# re: BizTalk Testing Series - Testing Pipeline Components
Requesting Gravatar...
On using Pipeline Component Test Library, I've noticed that if you read IPipelineContext.StageID from the pipeline component, this is not correctly set. It always has a {0000-...-0000} value. Is there a way to get the actual satge id?

Left by Lucian on Nov 20, 2009 9:24 AM

# re: BizTalk Testing Series - Testing Pipeline Components
Requesting Gravatar...
Actually I do need the guid to be set to correctly point to the correct stage, cause the pipeline component I am testing is supposed to act based on the stageID of the Pipelinecontext
Left by Arijit Basu on Mar 08, 2010 8:32 AM

# re: BizTalk Testing Series - Testing Pipeline Components
Requesting Gravatar...
It's Very Nice Blog
Left by Yash on Apr 23, 2011 5:20 AM

Your comment:
 (will show your gravatar)

Copyright © Michael Stephenson | Powered by: