How to debug?

[tgraham-antenna]

What is the recommended technique or techniques for debugging xslTNG?

[ndw]

It's a good question. If there's an appropriate debug flag, I turn that on. Often I wind up just adding xsl:message statements in various places. Debugging complex XSLT transformations is frustratingly difficult. (Ask me about the XX compiler some day.)

[fsteimke]

Do you think that Exception handling with try - catch would help?

For example, the function f:absolute-length in units.xsl parses a length and returns xs:double. When $length cant be parsed, it does not fail, but emits a message and returns 0`. Using the xslTNG Stylesheetes within XPROC (xmlcalabash implementation) means that the error message will be discarded. Maybe the whole XProc steps fails, or it passes, but you get unexpected and wrong results without indication what went wrong.

It will take some time to find out, that the error was caused by an invalid length. And even if you found out, you don't know the reason, because you don't know the object with an invalid length specification (a CALS colwidth? An imange dimension?)

I think it would be better to throw an exception.

Greetings, Frank

[ndw]

Perhaps. I think years of exposure to the behavior that browsers always render something even if the input is "broken" has left some folks with the conviction that stylesheets should always produce some result. Treating an invalid thing like it was absent is probably motivated partly by trying to satisfy that goal. But I think a case could be made for being more strict. Or, sigh, perhaps having a flag that enabled more strict behavior.

With respect to "the error message will be discarded", can you explain what you mean? XML Calabash should print the messages generated in XSLT. Consider:

<p:declare-step xmlns:p="http://www.w3.org/ns/xproc" 
	        xmlns:xs="http://www.w3.org/2001/XMLSchema"
	        version="3.1">
  <p:output port="result" primary="true"  sequence="true"/>

  <p:xslt>
    <p:with-input port="source">
      <doc>
        <p>Test</p>
      </doc>
    </p:with-input>
    <p:with-input port="stylesheet">
      <xsl:stylesheet xmlns:xsl="http://www.w3.org/1999/XSL/Transform"
                      xmlns:xs="http://www.w3.org/2001/XMLSchema"
                      exclude-result-prefixes="xs"
                      version="3.0">
        <xsl:output method="xml" encoding="utf-8" indent="no"/>
        <xsl:mode on-no-match="shallow-copy"/>
        <xsl:template match="p">
          <xsl:if test="not(@xml:id)">
            <xsl:message>Paragraph lacks an id.</xsl:message>
          </xsl:if>
          <xsl:next-match/>
        </xsl:template>
      </xsl:stylesheet>
    </p:with-input>
  </p:xslt>
	
</p:declare-step>

When I run that, I get:

$ app/build/stage/xmlcalabash.sh /tmp/test.xpl -o:/dev/null
Paragraph lacks an id.

XML Calabash shouldn't be swallowing the messages.

Your observation about the fact that "length was wrong" isn't really very specific about which length is definitely relevant as well.

[fsteimke]

I have to investigate the report of error messages with xmlCalabash.

I have built a framework around the xslTNG Stylesheets. It uses xproc and xmlcalabash to call an XSLT Stylesheet, which acts as a customization layer for docbooc.xsl from the xslTNG Stylesheets. Sometimes I find it difficult to keep track of it all.

And then i test the whole thing with a really large document, and all i get is xproc step faild. Dont't know why. It took about an hour to find out, that the width of a CALS column was given as 20% instead of 2*. Apologies for being a bit annoyed.

Maybe i can make a suggestion for f:absolute-length. And i will try to find the XSLT messages somewhere in the layers of my framework.

Greetings, Frank