Introduction

Ever since starting work with BizTalk using a custom XSLT file for doing mapping, we've been relying on the handy element at the root of our XSLT:

<xsl:strip-space elements="*"/>

This made it so we didn't get any extra whitespace in elements where it didn't belong. Because if there were extra line-feed/carriage-return/tab characters, BizTalk would choke on the errors, because would read those elements as actually having a value and throw an error for it having invalid data.

The Cause

Well it turns out that some of the underlying libraries that BizTalk uses were changed for 2013, most importantly the following:

The Mapper uses the XSLCompiledTransform class. Previous BizTalk Server versions used the XslTransform class, which is obsolete

As outlined on the What's New in BizTalk 2013 page on MSDN.

Well I decided to try mucking around with the test app, I found it all came down to this line at the top of the XSL file:

<xsl:strip-space elements="*"/>

I ended up getting the error

Whitespace cannot be stripped from documents that have already been loaded. Provide the input as an XmlReader instead

So, I did what it says, in my test app and it worked! However, I couldn't do so (or didn't know how to specify) in BizTalk.

The Fix

The fix as it turns out was to stop being lazy basically. Since we can't dig into BizTalk (we aren't using custom pipelines unfortunately) it came down to using a whole bunch of [normalize-space(.)] selectors around basically anything that selects text in the XSLT file. It's not at all ideal, but relatively easy and got me in the habit of doing it now.

I know I also enabled the legacy Whitespace behavior in my Host --> Settings dialog box. The combination of these 2 things did the trick.

If I get the chance to do this app over again, it will be done using custom pipelines, instead of orchestrations and I believe this will allow us to avoid this situation.

If anyone else has suggestions or other tips, please let me know.