The Problem
If you have been a BizTalk developer long enough, you will know that application lifecycle management has always been a pain point when it comes to managing BizTalk projects.
The Out-of-Box support available in BizTalk Server is not too great for things like automated builds, continuous integration, code analysis, code coverage, unit testing… I can go on and on, but I am sure you get the picture.
Thankfully, most of these are set to improve with the upcoming release of BizTalk Server i.e. BizTalk Server 2009, but there are still improvement areas that I hope the product group will address soon.
One of the other hard things (IMHO) that one needs to do early-on in a BizTalk project is to decide on the naming convention and projects structure for modular deployment. Sometime this is dictated by your customer and sometimes you just follow what has been decided by your company, simple isn’t it? Wish life was that easyJ.
In reality you will find that customers want you to use their own guidelines that are often not complete or not fully thought-out, and you end up using that and try to fill the gap or modify it by adding your own stuff.
Is this so bad?… not if you have gone through this exercise and got it frozen and signed off before you start your development / coding, else, this can turn out to be the cause for endless frustrations for you and your team.
This is especially true for BizTalk projects, where you know that small changes to things like namespaces or typenames can cause a rippling effect across the project, and you end up rebuilding / reconfiguring all of the artefacts.
I have been on multiple projects in the past where naming conventions changed almost every week while the project was halfway through the development phase, so this is not an uncommon thing; however you need to be aware of this and take measures to avoid getting into such situations.
Another challenge is in sticking to the guidelines or conventions that you have so painstakingly chosen and frozen. If you have a large team, you will often see that not all developers are fully aware of the guidelines and conventions that are being followed in the team, especially when developers are constantly moving in and out of the team.
Solution
So what can be done in such situations? One of the things you can do on your BizTalk project is to run a static code analyser as part of your daily builds. I have always been a huge fan of build automation and the niceties that come along with it like automated testing, static code analysis report, code coverage report etc. There is nothing that makes you happier than getting up in the morning gazing at your PDA / mobile device to see that there were no errors or build breaks on the nightly build, especially when you are the Dev Lead / Architect on the project.
You must be thinking… “all this is great… but isn’t this is a BizTalk Solution?… how do you automate checking naming conventions?… especially on things like orchestration shapes, map names, physical ports etc…”. well you are in luck, you can now do static code analysis on BizTalk solutions too, thanks to Elton Stoneman for this cool new tool/plug-in called BizTalkCop available on codeplex.
Check out the links:
http://geekswithblogs.net/EltonStoneman/archive/2008/11/14/introducing-biztalkcop.aspx
BizTalkCop is essentially a set of FxCop rules that will allow you to inspect BizTalk assemblies for structure and naming convention of your BizTalk artefacts such as Orchestrations, Pipelines, Schemas, Maps etc. It can even look into the deployed solutions inside BizTalk Server (using BizTalk Object Model and BizTalkMgmtDB) to validate the names of your physical ports and receive locations on your deployed BizTalk Server application.
If you are already familiar with FxCop, BizTalkCop provides you the same experience since it is just an extension of the FxCop ruleset.
BizTalkCop is currently is release 1.0 and contains a set of rules based on Scott Colestock’s naming conventions, but they’re (mostly) configurable so you can modify them to suit your own standards. You can create your new custom rules as well, since the full source code is provided with base classes and frameworks.
Another thing to note is that since BizTalk (2006/R2) projects are not integrated with Code Analysis tool in Visual Studio 2005, you need to download and install FxCop separately before installing BizTalkCop (even if you are using the Team Suite edition of Visual Studio 2005). I have not tried it with BizTalk 2009 Beta yet since I have it on Professional edition of Visual Studio 2008 installed on my VPC (Code Analysis is available only on Visual Studio Team Suite).
The link above provides instructions on how to install and configure BizTalkCop so I am not going to detail it out here.
Demo – The Bad Project
Let me illustrate the usefulness of BizTalkCop by running it on a badly done solution. I created a very simple BizTalk application a la “Hello World” in BizTalk (I call it the ’BadProject’) without giving much attention to the project structure or naming convention (just used the default names, like we do most of the time). The application basically receives an inventory replenishment request; the request is transformed into a purchase order request and is sent out. For simplicity sake, both the receive port and the send port use FILE adapters.
So essentially, my BizTalk project consists of a single BizTalk project within which there are two Schemas, a Map and an Orchestration as shown in the solution explorer screenshot below:
The orchestration, transformation and schemas are all very elementary as shown below (notice the naming of the artefacts highlighted).
After I have built and deployed the solutions, I fire up FxCop, load the target BizTalk application assembly and hit the Analyze button. I selected only BizTalkCop rules and unchecked others
|
Note: You also need to make sure that BizTalkCop is configured to point to the right BizTalk application so that it can validate the application level artifacts like the physical post names etc.
|
You see from the screenshots that many errors (about 28 of them) are generated for such a simple application.
Here are the errors:
The errors are pretty self explanatory so I don’t want to go through each of them.
|
1. Port names should be prefixed with their direction – start ‘ReceivePort1’ with ‘Receive.’
|
|
2. Port names should be prefixed with their direction – start ‘SendPort1’ with ‘Send.’
|
|
3. Receive Location names should be prefixed with their Receive Port name – start ‘Receive Location1’ with ‘ReceivePort1’
|
|
4. Receive Location names should be suffixed with their transport type – end ‘Receive Location1’ with ‘.FILE’
|
|
5. Artifacts should be declared in modules with the correct suffix – consider module name ‘BadProject.Schemas’
|
|
6. Schema names should end with the data format. Format: ‘inventoryRequest’ is unknown
|
|
7. Schema names should begin with the root node. Start: ‘inventoryRequest’ with: ‘Root’
|
|
8. Artifacts should be declared in modules with the correct suffix – consider module name ‘BadProject.Transforms’
|
|
9. Map names should have the format “SourceSchema_DestinationSchema”
|
|
10.Artifacts should be declared in modules with the correct suffix – consider module name ‘BadProject.Orchestrations’
|
|
11.Orchestration members should be Camel cased. Replace Message name: ‘InvReqMsg’ with: ‘invReqMsg’
|
|
12.Orchestration members should be Camel cased. Replace Message name: ‘PurchaseOrderMsg’ with: ‘purchaseOrderMsg’
|
|
13.Orchestration members should be Camel cased. Replace Port name: ‘Port_1’ with: ‘port1’
|
|
14.Orchestration members should be Camel cased. Replace Port name: ‘Port_2’ with: ‘port2’
|
|
15.Orchestration shapes should be correctly named – replace ‘ConstructMessage_PurchaseOrder’ with ‘Construct_PurchaseOrderMsg’
|
|
16.Orchestration shapes should be correctly named – replace ‘Transform_InvToPurchaseOrder’ with ‘Transform_inventoryRequest_purchaseOrderSchema’
|
|
17.Orchestration Shapes should have the correct prefix – start ‘MessageAssignment_1’ with ‘Assign_’
|
|
18.Orchestration Shapes should have the correct prefix – start ‘Receive_InvRequest’ with ‘Rcv_’
|
|
19.Orchestration Shapes should have the correct prefix – start ‘Send_POMsg’ with ‘Snd_’
|
|
20.Orchestration types should be Pascal cased. Replace Port Type name: ‘PortType_1’ with: ‘PortType1’
|
|
21.Orchestration types should be Pascal cased. Replace Port Type name: ‘PortType_2’ with: ‘PortType2’
|
|
22.Orchestration Types should have the correct suffix – end ‘PortType_1’ with ‘PortType’
|
|
23.Orchestration Types should have the correct suffix – end ‘PortType_2’ with ‘PortType’
|
|
24.Orchestration Types should have the correct suffix – end ‘Port_1’ with ‘Port’
|
|
25.Orchestration Types should have the correct suffix – end ‘Port_2’ with ‘Port’
|
|
26.Artifacts should be declared in modules with the correct suffix – consider module name ‘BadProject.Schemas’
|
|
27.Schema names should end with the data format. Format: ‘purchaseOrderSchema’ is unknown
|
|
28.Schema names should begin with the root node. Start: ‘purchaseOrderSchema’ with: ‘Root’
|
You can also save the FxCop project and use it in the command line option to generate a pretty cool report as show below.
Demo – The Good Solution
Now I took the same solution and made changes by following the naming conventions and also restructuring the solution to be suitable for modular deployment. You see from the illustration below, I have split the application into three separate projects so that it is not only easier to maintain and saves time during development, but it is absolutely critical that it is broken in this manner from an administration point of view when the solution is in production. Any seasoned BizTalk developer / administrator will understand what I mean here, it’s a whole different topic and don’t want to get swayed into that.
You also see in the solution explorer above that the name of some of the artefacts have also been changed in accordance with the naming convention. For example, the schema names indicate the data format and the root node name of the schema. I also made sure that the physical ports confirm to the naming standards.
The orchestration, map and the schemas below show that changes have been made in the names of the artefacts.
You can see from the below screenshots that the errors have been reduced to just one error.
Conclusion
To sum it all up, I feel FxCop / BizTalkCop makes a great tool to have in your arsenal for that next BizTalk project you are waiting for or even in your current project. Saves you (or the SQA team) a lot of time from having to check for the naming conventions manually on a routine basis and provides you reports on demand.
It can be used both on the individual machines as well as on the build server. You can integrate it with your build scripts so that your application health report now also includes static code analysis among other things. This way you can be assured and can sleep peacefully knowing that the code drop you just sent to the customer conforms to all the naming conventions without having to manually review it at the eleventh-hour, not to mention the amount of time saved. I hope to see many more new rules and improvements from Elton in future releases of BizTalkCop.
Benny Mathew
Sr.Consultant, MGSI (Microsoft Global Services, India).
While I acknowledge that I am an employee of Microsoft., any & all views expressed in this article are mine and do not necessarily reflect the views of Microsoft.