Custom Pipeline Woes

In one of my project, the source system was sending 2 almost similar messages to Biztalk. The only difference was the root node name. I had 2 ways of dealing with it.

Solution 1:

1) Create separate document schemas
2) Create separate Receive Pipelines referring to each document schema
3) Use a map to change the root node name so to have exactly same message that can pass through the common orchestration thereafter.

Pros:

1) Less Development time
2) Easy message retrieving in case of failure

Cons:

1) Additional artifacts required in form of additional schemas and pipelines.
2) Low in performance and high in memory usage as we have to use extra map.

Solution 2:

Use a custom pipeline component that will replace the “root node” name to a common node name.

Pros:

1) Cleaner and Faster

Cons:

1) More development time
2) If error handling is not proper then the entire message is lost. This is because there is no persistence to messagebox in the pipelines.

Whereas the first solution is very easy, the second is relatively difficult. I tried the second one for simple reason that it was challenging and I thought I could do it within reasonable time that I had. I started by following the standard procedure for developing custom pipeline components. I used the “FixMsg” custom pipeline sample from “SDK” as my base code. It is located at “..\Program Files\Microsoft BizTalk Server 2004\SDK\Samples\Pipelines\CustomComponent”. This saved me some time in not “reinventing” the wheel. I worked on the “Execute” method and converted the Stream to Byte and then Byte to string. Then I tried replacing the required node in string and then tried to convert back the new string back to stream. Since pipeline only deals with streaming data. I ensured that the return is a “stream”. My approached looked quiet logical to me but the results were NADA. This string replacement did not actually work as I was not able to pass the “stream” back to the next stage of the pipeline. Then, I thought of using XPath. But since pipeline deal with streaming data whereas Xpath requires the whole message to be loaded into memory. I could not use the normal Xpath but I could “streaming Xpath”. This is called “forward only” Xpath. I tried “cloning” the stream so I can use XMLReader class on it. But could not go anywhere, then I gave up the solution as I was running out of time. I lost this challenge!! But, I will definitely revisit it once I get some “breathing space”.

Any inputs will be welcome.