XSLT vs. PURE - different number of failed-asserts

[robertmarkolwitzatnortal]

Hi Philip,

in the project I am currently working on we have been using the Pure-Validation until now, but are currently evaluating XSLT-Validation since the number of rules and asserts are quite large and XSLT performance is way better on complex structures.

I have pre-generated the XSLT via and am loading the XSLT via SchematronResourceXSLT.fromClassPath.

While running our existing test-suite for both, pure and xslt validation, and comparing results, I found that the result schematron-output differs in terms of numbers of <fired-rule/> elements as well as number of <failed-assert/> elements.

Now, I understand that xslt requires a phase but the schematron files do not contain any <phase/> elements nor a defaultPhase attribute.

I was generating the xslt without specifying an explicit phase, since the schematron rules do not contain <phase/> elements nor a defaultPhase attribute.

I am attaching a zip archive with some files for further analysis (since they are quite large):

• cn_24_maximal_error.xml: (the test input)
• *.sch: (the schematron files)
• schematron_eu_1.0.0.xslt: (the xslt pre-generated from schematron files)
• result_SVRL_XSLT.xml: (the schematron-output after XSLT-Validation)
• result_SVRL_PURE.xml: (the schematron-output after PURE-Validation)

It would be fantastic if you could take some time to investigate this matter and provide a hint on where I might have taken a wrong turn.

From my understanding, outputs of XSLT-Validation and PURE-Validation can actually differ. But if no <phase/> is given in the schematron files and can hence not be specified for XSLT pre-generation, how can I enforce the same number of rules fired or at least the identical set of failed-asserts across both validation modes?

Thank you very much in advance for your help! :)

Cheers, Robert schematron_eu-1_0_0.zip